1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 2f46b5a66SChristoph Hellwig /* 3f46b5a66SChristoph Hellwig * Copyright (C) 2007 Oracle. All rights reserved. 4f46b5a66SChristoph Hellwig */ 5f46b5a66SChristoph Hellwig 6f46b5a66SChristoph Hellwig #include <linux/kernel.h> 7f46b5a66SChristoph Hellwig #include <linux/bio.h> 8f46b5a66SChristoph Hellwig #include <linux/file.h> 9f46b5a66SChristoph Hellwig #include <linux/fs.h> 10cb8e7090SChristoph Hellwig #include <linux/fsnotify.h> 11f46b5a66SChristoph Hellwig #include <linux/pagemap.h> 12f46b5a66SChristoph Hellwig #include <linux/highmem.h> 13f46b5a66SChristoph Hellwig #include <linux/time.h> 14f46b5a66SChristoph Hellwig #include <linux/string.h> 15f46b5a66SChristoph Hellwig #include <linux/backing-dev.h> 16cb8e7090SChristoph Hellwig #include <linux/mount.h> 17cb8e7090SChristoph Hellwig #include <linux/namei.h> 18f46b5a66SChristoph Hellwig #include <linux/writeback.h> 19f46b5a66SChristoph Hellwig #include <linux/compat.h> 20cb8e7090SChristoph Hellwig #include <linux/security.h> 21f46b5a66SChristoph Hellwig #include <linux/xattr.h> 22f54de068SDavid Sterba #include <linux/mm.h> 235a0e3ad6STejun Heo #include <linux/slab.h> 24f7039b1dSLi Dongyang #include <linux/blkdev.h> 258ea05e3aSAlexander Block #include <linux/uuid.h> 2655e301fdSFilipe Brandenburger #include <linux/btrfs.h> 27416161dbSMark Fasheh #include <linux/uaccess.h> 28ae5e165dSJeff Layton #include <linux/iversion.h> 29f46b5a66SChristoph Hellwig #include "ctree.h" 30f46b5a66SChristoph Hellwig #include "disk-io.h" 31f46b5a66SChristoph Hellwig #include "transaction.h" 32f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 33f46b5a66SChristoph Hellwig #include "print-tree.h" 34f46b5a66SChristoph Hellwig #include "volumes.h" 35925baeddSChris Mason #include "locking.h" 36581bb050SLi Zefan #include "inode-map.h" 37d7728c96SJan Schmidt #include "backref.h" 38606686eeSJosef Bacik #include "rcu-string.h" 3931db9f7cSAlexander Block #include "send.h" 403f6bcfbdSStefan Behrens #include "dev-replace.h" 4163541927SFilipe David Borba Manana #include "props.h" 423b02a68aSJeff Mahoney #include "sysfs.h" 43fcebe456SJosef Bacik #include "qgroup.h" 441ec9a1aeSFilipe Manana #include "tree-log.h" 45ebb8765bSAnand Jain #include "compression.h" 46f46b5a66SChristoph Hellwig 47abccd00fSHugo Mills #ifdef CONFIG_64BIT 48abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 49abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 50abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 51abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 52abccd00fSHugo Mills */ 53abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 54abccd00fSHugo Mills __u64 sec; 55abccd00fSHugo Mills __u32 nsec; 56abccd00fSHugo Mills } __attribute__ ((__packed__)); 57abccd00fSHugo Mills 58abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 59abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 60abccd00fSHugo Mills __u64 stransid; /* in */ 61abccd00fSHugo Mills __u64 rtransid; /* out */ 62abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 63abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 64abccd00fSHugo Mills __u64 flags; /* in */ 65abccd00fSHugo Mills __u64 reserved[16]; /* in */ 66abccd00fSHugo Mills } __attribute__ ((__packed__)); 67abccd00fSHugo Mills 68abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 69abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 70abccd00fSHugo Mills #endif 71abccd00fSHugo Mills 722351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 732351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 742351f431SJosef Bacik __s64 send_fd; /* in */ 752351f431SJosef Bacik __u64 clone_sources_count; /* in */ 762351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 772351f431SJosef Bacik __u64 parent_root; /* in */ 782351f431SJosef Bacik __u64 flags; /* in */ 792351f431SJosef Bacik __u64 reserved[4]; /* in */ 802351f431SJosef Bacik } __attribute__ ((__packed__)); 812351f431SJosef Bacik 822351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 832351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 842351f431SJosef Bacik #endif 85abccd00fSHugo Mills 86416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 871c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 881c919a5eSMark Fasheh int no_time_update); 89416161dbSMark Fasheh 906cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 911905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 921905a0f7SDavid Sterba unsigned int flags) 936cbff00fSChristoph Hellwig { 941905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 956cbff00fSChristoph Hellwig return flags; 961905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 976cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 986cbff00fSChristoph Hellwig else 996cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1006cbff00fSChristoph Hellwig } 101f46b5a66SChristoph Hellwig 1026cbff00fSChristoph Hellwig /* 103a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 104a157d4fdSDavid Sterba * ioctl. 1056cbff00fSChristoph Hellwig */ 106a157d4fdSDavid Sterba static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags) 1076cbff00fSChristoph Hellwig { 1086cbff00fSChristoph Hellwig unsigned int iflags = 0; 1096cbff00fSChristoph Hellwig 1106cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1116cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1126cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1136cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1156cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1176cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1196cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1216cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 122d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 123d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 124d0092bddSLi Zefan 12513f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 126d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 12713f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 12813f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1296cbff00fSChristoph Hellwig 1306cbff00fSChristoph Hellwig return iflags; 1316cbff00fSChristoph Hellwig } 1326cbff00fSChristoph Hellwig 1336cbff00fSChristoph Hellwig /* 1346cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1356cbff00fSChristoph Hellwig */ 1367b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1376cbff00fSChristoph Hellwig { 1385c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1393cc79392SFilipe Manana unsigned int new_fl = 0; 1406cbff00fSChristoph Hellwig 1415c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1423cc79392SFilipe Manana new_fl |= S_SYNC; 1435c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1443cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1455c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1463cc79392SFilipe Manana new_fl |= S_APPEND; 1475c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1483cc79392SFilipe Manana new_fl |= S_NOATIME; 1495c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1503cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1513cc79392SFilipe Manana 1523cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1533cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1543cc79392SFilipe Manana new_fl); 1556cbff00fSChristoph Hellwig } 1566cbff00fSChristoph Hellwig 1576cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1586cbff00fSChristoph Hellwig { 1595c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 1605c57b8b6SDavid Sterba unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags); 1616cbff00fSChristoph Hellwig 1626cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1636cbff00fSChristoph Hellwig return -EFAULT; 1646cbff00fSChristoph Hellwig return 0; 1656cbff00fSChristoph Hellwig } 1666cbff00fSChristoph Hellwig 1675ba76abfSDavid Sterba /* Check if @flags are a supported and valid set of FS_*_FL flags */ 1685ba76abfSDavid Sterba static int check_fsflags(unsigned int flags) 16975e7cb7fSLiu Bo { 17075e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17175e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17275e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 173e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 174e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17575e7cb7fSLiu Bo return -EOPNOTSUPP; 17675e7cb7fSLiu Bo 17775e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 17875e7cb7fSLiu Bo return -EINVAL; 17975e7cb7fSLiu Bo 18075e7cb7fSLiu Bo return 0; 18175e7cb7fSLiu Bo } 18275e7cb7fSLiu Bo 1836cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1846cbff00fSChristoph Hellwig { 185496ad9aaSAl Viro struct inode *inode = file_inode(file); 1860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1875c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1885c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 1896cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1905c57b8b6SDavid Sterba unsigned int fsflags, old_fsflags; 1916cbff00fSChristoph Hellwig int ret; 1927e97b8daSDavid Sterba umode_t mode; 193ff9fef55SAnand Jain const char *comp = NULL; 194d2b8fcfeSAnand Jain u32 binode_flags = binode->flags; 1956cbff00fSChristoph Hellwig 196bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 197bd60ea0fSDavid Sterba return -EPERM; 198bd60ea0fSDavid Sterba 199b83cc969SLi Zefan if (btrfs_root_readonly(root)) 200b83cc969SLi Zefan return -EROFS; 201b83cc969SLi Zefan 2025c57b8b6SDavid Sterba if (copy_from_user(&fsflags, arg, sizeof(fsflags))) 2036cbff00fSChristoph Hellwig return -EFAULT; 2046cbff00fSChristoph Hellwig 2055c57b8b6SDavid Sterba ret = check_fsflags(fsflags); 20675e7cb7fSLiu Bo if (ret) 20775e7cb7fSLiu Bo return ret; 2086cbff00fSChristoph Hellwig 209e7848683SJan Kara ret = mnt_want_write_file(file); 210e7848683SJan Kara if (ret) 211e7848683SJan Kara return ret; 212e7848683SJan Kara 2135955102cSAl Viro inode_lock(inode); 2146cbff00fSChristoph Hellwig 2157e97b8daSDavid Sterba mode = inode->i_mode; 216f062abf0SLi Zefan 2175c57b8b6SDavid Sterba fsflags = btrfs_mask_fsflags_for_type(inode, fsflags); 2185c57b8b6SDavid Sterba old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags); 2195c57b8b6SDavid Sterba if ((fsflags ^ old_fsflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2206cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2216cbff00fSChristoph Hellwig ret = -EPERM; 2226cbff00fSChristoph Hellwig goto out_unlock; 2236cbff00fSChristoph Hellwig } 2246cbff00fSChristoph Hellwig } 2256cbff00fSChristoph Hellwig 2265c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 227d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_SYNC; 2286cbff00fSChristoph Hellwig else 229d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_SYNC; 2305c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 231d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_IMMUTABLE; 2326cbff00fSChristoph Hellwig else 233d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_IMMUTABLE; 2345c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 235d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_APPEND; 2366cbff00fSChristoph Hellwig else 237d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_APPEND; 2385c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 239d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODUMP; 2406cbff00fSChristoph Hellwig else 241d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODUMP; 2425c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 243d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOATIME; 2446cbff00fSChristoph Hellwig else 245d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOATIME; 2465c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 247d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_DIRSYNC; 2486cbff00fSChristoph Hellwig else 249d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_DIRSYNC; 2505c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 2517e97b8daSDavid Sterba if (S_ISREG(mode)) { 2527e97b8daSDavid Sterba /* 2537e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2547e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2557e97b8daSDavid Sterba * status of the file and will not set it. 2567e97b8daSDavid Sterba */ 2577e97b8daSDavid Sterba if (inode->i_size == 0) 258d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW | 259d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM; 2607e97b8daSDavid Sterba } else { 261d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW; 2627e97b8daSDavid Sterba } 2637e97b8daSDavid Sterba } else { 2647e97b8daSDavid Sterba /* 26501327610SNicholas D Steeves * Revert back under same assumptions as above 2667e97b8daSDavid Sterba */ 2677e97b8daSDavid Sterba if (S_ISREG(mode)) { 2687e97b8daSDavid Sterba if (inode->i_size == 0) 269d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_NODATACOW | 270d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM); 2717e97b8daSDavid Sterba } else { 272d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODATACOW; 2737e97b8daSDavid Sterba } 2747e97b8daSDavid Sterba } 2756cbff00fSChristoph Hellwig 27675e7cb7fSLiu Bo /* 27775e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 27875e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 27975e7cb7fSLiu Bo * things smaller. 28075e7cb7fSLiu Bo */ 2815c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 282d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_COMPRESS; 283d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOCOMPRESS; 2845c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 28563541927SFilipe David Borba Manana 286eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 287eede2bf3SOmar Sandoval ret = -ETXTBSY; 288eede2bf3SOmar Sandoval goto out_unlock; 289eede2bf3SOmar Sandoval } 290eede2bf3SOmar Sandoval 291d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_COMPRESS; 292d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOCOMPRESS; 29363541927SFilipe David Borba Manana 29493370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 29593370509SDavid Sterba if (!comp || comp[0] == 0) 29693370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 297ebcb904dSLi Zefan } else { 298d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 29975e7cb7fSLiu Bo } 3006cbff00fSChristoph Hellwig 301ff9fef55SAnand Jain /* 302ff9fef55SAnand Jain * 1 for inode item 303ff9fef55SAnand Jain * 2 for properties 304ff9fef55SAnand Jain */ 305ff9fef55SAnand Jain trans = btrfs_start_transaction(root, 3); 306f062abf0SLi Zefan if (IS_ERR(trans)) { 307f062abf0SLi Zefan ret = PTR_ERR(trans); 308d2b8fcfeSAnand Jain goto out_unlock; 309f062abf0SLi Zefan } 3106cbff00fSChristoph Hellwig 311ff9fef55SAnand Jain if (comp) { 312ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, 313ff9fef55SAnand Jain strlen(comp), 0); 314ff9fef55SAnand Jain if (ret) { 315ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 316ff9fef55SAnand Jain goto out_end_trans; 317ff9fef55SAnand Jain } 318ff9fef55SAnand Jain set_bit(BTRFS_INODE_COPY_EVERYTHING, 319ff9fef55SAnand Jain &BTRFS_I(inode)->runtime_flags); 320ff9fef55SAnand Jain } else { 321ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 322ff9fef55SAnand Jain 0, 0); 323ff9fef55SAnand Jain if (ret && ret != -ENODATA) { 324ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 325ff9fef55SAnand Jain goto out_end_trans; 326ff9fef55SAnand Jain } 327ff9fef55SAnand Jain } 328ff9fef55SAnand Jain 329d2b8fcfeSAnand Jain binode->flags = binode_flags; 3307b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3310c4d2d95SJosef Bacik inode_inc_iversion(inode); 332c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3336cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3346cbff00fSChristoph Hellwig 335ff9fef55SAnand Jain out_end_trans: 3363a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3376cbff00fSChristoph Hellwig out_unlock: 3385955102cSAl Viro inode_unlock(inode); 339e7848683SJan Kara mnt_drop_write_file(file); 3402d4e6f6aSliubo return ret; 3416cbff00fSChristoph Hellwig } 3426cbff00fSChristoph Hellwig 34319f93b3cSDavid Sterba /* 34419f93b3cSDavid Sterba * Translate btrfs internal inode flags to xflags as expected by the 34519f93b3cSDavid Sterba * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are 34619f93b3cSDavid Sterba * silently dropped. 34719f93b3cSDavid Sterba */ 34819f93b3cSDavid Sterba static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags) 34919f93b3cSDavid Sterba { 35019f93b3cSDavid Sterba unsigned int xflags = 0; 35119f93b3cSDavid Sterba 35219f93b3cSDavid Sterba if (flags & BTRFS_INODE_APPEND) 35319f93b3cSDavid Sterba xflags |= FS_XFLAG_APPEND; 35419f93b3cSDavid Sterba if (flags & BTRFS_INODE_IMMUTABLE) 35519f93b3cSDavid Sterba xflags |= FS_XFLAG_IMMUTABLE; 35619f93b3cSDavid Sterba if (flags & BTRFS_INODE_NOATIME) 35719f93b3cSDavid Sterba xflags |= FS_XFLAG_NOATIME; 35819f93b3cSDavid Sterba if (flags & BTRFS_INODE_NODUMP) 35919f93b3cSDavid Sterba xflags |= FS_XFLAG_NODUMP; 36019f93b3cSDavid Sterba if (flags & BTRFS_INODE_SYNC) 36119f93b3cSDavid Sterba xflags |= FS_XFLAG_SYNC; 36219f93b3cSDavid Sterba 36319f93b3cSDavid Sterba return xflags; 36419f93b3cSDavid Sterba } 36519f93b3cSDavid Sterba 36619f93b3cSDavid Sterba /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */ 36719f93b3cSDavid Sterba static int check_xflags(unsigned int flags) 36819f93b3cSDavid Sterba { 36919f93b3cSDavid Sterba if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME | 37019f93b3cSDavid Sterba FS_XFLAG_NODUMP | FS_XFLAG_SYNC)) 37119f93b3cSDavid Sterba return -EOPNOTSUPP; 37219f93b3cSDavid Sterba return 0; 37319f93b3cSDavid Sterba } 37419f93b3cSDavid Sterba 375e4202ac9SDavid Sterba /* 376e4202ac9SDavid Sterba * Set the xflags from the internal inode flags. The remaining items of fsxattr 377e4202ac9SDavid Sterba * are zeroed. 378e4202ac9SDavid Sterba */ 379e4202ac9SDavid Sterba static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg) 380e4202ac9SDavid Sterba { 381e4202ac9SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 382e4202ac9SDavid Sterba struct fsxattr fa; 383e4202ac9SDavid Sterba 384e4202ac9SDavid Sterba memset(&fa, 0, sizeof(fa)); 385e4202ac9SDavid Sterba fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags); 386e4202ac9SDavid Sterba 387e4202ac9SDavid Sterba if (copy_to_user(arg, &fa, sizeof(fa))) 388e4202ac9SDavid Sterba return -EFAULT; 389e4202ac9SDavid Sterba 390e4202ac9SDavid Sterba return 0; 391e4202ac9SDavid Sterba } 392e4202ac9SDavid Sterba 393025f2121SDavid Sterba static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg) 394025f2121SDavid Sterba { 395025f2121SDavid Sterba struct inode *inode = file_inode(file); 396025f2121SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 397025f2121SDavid Sterba struct btrfs_root *root = binode->root; 398025f2121SDavid Sterba struct btrfs_trans_handle *trans; 399025f2121SDavid Sterba struct fsxattr fa; 400025f2121SDavid Sterba unsigned old_flags; 401025f2121SDavid Sterba unsigned old_i_flags; 402025f2121SDavid Sterba int ret = 0; 403025f2121SDavid Sterba 404025f2121SDavid Sterba if (!inode_owner_or_capable(inode)) 405025f2121SDavid Sterba return -EPERM; 406025f2121SDavid Sterba 407025f2121SDavid Sterba if (btrfs_root_readonly(root)) 408025f2121SDavid Sterba return -EROFS; 409025f2121SDavid Sterba 410025f2121SDavid Sterba memset(&fa, 0, sizeof(fa)); 411025f2121SDavid Sterba if (copy_from_user(&fa, arg, sizeof(fa))) 412025f2121SDavid Sterba return -EFAULT; 413025f2121SDavid Sterba 414025f2121SDavid Sterba ret = check_xflags(fa.fsx_xflags); 415025f2121SDavid Sterba if (ret) 416025f2121SDavid Sterba return ret; 417025f2121SDavid Sterba 418025f2121SDavid Sterba if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0) 419025f2121SDavid Sterba return -EOPNOTSUPP; 420025f2121SDavid Sterba 421025f2121SDavid Sterba ret = mnt_want_write_file(file); 422025f2121SDavid Sterba if (ret) 423025f2121SDavid Sterba return ret; 424025f2121SDavid Sterba 425025f2121SDavid Sterba inode_lock(inode); 426025f2121SDavid Sterba 427025f2121SDavid Sterba old_flags = binode->flags; 428025f2121SDavid Sterba old_i_flags = inode->i_flags; 429025f2121SDavid Sterba 430025f2121SDavid Sterba /* We need the capabilities to change append-only or immutable inode */ 431025f2121SDavid Sterba if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) || 432025f2121SDavid Sterba (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) && 433025f2121SDavid Sterba !capable(CAP_LINUX_IMMUTABLE)) { 434025f2121SDavid Sterba ret = -EPERM; 435025f2121SDavid Sterba goto out_unlock; 436025f2121SDavid Sterba } 437025f2121SDavid Sterba 438025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_SYNC) 439025f2121SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 440025f2121SDavid Sterba else 441025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 442025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE) 443025f2121SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 444025f2121SDavid Sterba else 445025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 446025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_APPEND) 447025f2121SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 448025f2121SDavid Sterba else 449025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 450025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NODUMP) 451025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 452025f2121SDavid Sterba else 453025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 454025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NOATIME) 455025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 456025f2121SDavid Sterba else 457025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 458025f2121SDavid Sterba 459025f2121SDavid Sterba /* 1 item for the inode */ 460025f2121SDavid Sterba trans = btrfs_start_transaction(root, 1); 461025f2121SDavid Sterba if (IS_ERR(trans)) { 462025f2121SDavid Sterba ret = PTR_ERR(trans); 463025f2121SDavid Sterba goto out_unlock; 464025f2121SDavid Sterba } 465025f2121SDavid Sterba 466025f2121SDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 467025f2121SDavid Sterba inode_inc_iversion(inode); 468025f2121SDavid Sterba inode->i_ctime = current_time(inode); 469025f2121SDavid Sterba ret = btrfs_update_inode(trans, root, inode); 470025f2121SDavid Sterba 471025f2121SDavid Sterba btrfs_end_transaction(trans); 472025f2121SDavid Sterba 473025f2121SDavid Sterba out_unlock: 474025f2121SDavid Sterba if (ret) { 475025f2121SDavid Sterba binode->flags = old_flags; 476025f2121SDavid Sterba inode->i_flags = old_i_flags; 477025f2121SDavid Sterba } 478025f2121SDavid Sterba 479025f2121SDavid Sterba inode_unlock(inode); 480025f2121SDavid Sterba mnt_drop_write_file(file); 481025f2121SDavid Sterba 482025f2121SDavid Sterba return ret; 483025f2121SDavid Sterba } 484025f2121SDavid Sterba 4856cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4866cbff00fSChristoph Hellwig { 487496ad9aaSAl Viro struct inode *inode = file_inode(file); 4886cbff00fSChristoph Hellwig 4896cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4906cbff00fSChristoph Hellwig } 491f46b5a66SChristoph Hellwig 492f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 493f7039b1dSLi Dongyang { 4940b246afaSJeff Mahoney struct inode *inode = file_inode(file); 4950b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 496f7039b1dSLi Dongyang struct btrfs_device *device; 497f7039b1dSLi Dongyang struct request_queue *q; 498f7039b1dSLi Dongyang struct fstrim_range range; 499f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 500f7039b1dSLi Dongyang u64 num_devices = 0; 501f7039b1dSLi Dongyang int ret; 502f7039b1dSLi Dongyang 503f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 504f7039b1dSLi Dongyang return -EPERM; 505f7039b1dSLi Dongyang 506f35f06c3SFilipe Manana /* 507f35f06c3SFilipe Manana * If the fs is mounted with nologreplay, which requires it to be 508f35f06c3SFilipe Manana * mounted in RO mode as well, we can not allow discard on free space 509f35f06c3SFilipe Manana * inside block groups, because log trees refer to extents that are not 510f35f06c3SFilipe Manana * pinned in a block group's free space cache (pinning the extents is 511f35f06c3SFilipe Manana * precisely the first phase of replaying a log tree). 512f35f06c3SFilipe Manana */ 513f35f06c3SFilipe Manana if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 514f35f06c3SFilipe Manana return -EROFS; 515f35f06c3SFilipe Manana 5161f78160cSXiao Guangrong rcu_read_lock(); 5171f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 5181f78160cSXiao Guangrong dev_list) { 519f7039b1dSLi Dongyang if (!device->bdev) 520f7039b1dSLi Dongyang continue; 521f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 522f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 523f7039b1dSLi Dongyang num_devices++; 52450d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 525f7039b1dSLi Dongyang minlen); 526f7039b1dSLi Dongyang } 527f7039b1dSLi Dongyang } 5281f78160cSXiao Guangrong rcu_read_unlock(); 529f4c697e6SLukas Czerner 530f7039b1dSLi Dongyang if (!num_devices) 531f7039b1dSLi Dongyang return -EOPNOTSUPP; 532f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 533f7039b1dSLi Dongyang return -EFAULT; 5346ba9fc8eSQu Wenruo 5356ba9fc8eSQu Wenruo /* 5366ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 5376ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 5386ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 5396ba9fc8eSQu Wenruo */ 5406ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 541f4c697e6SLukas Czerner return -EINVAL; 542f7039b1dSLi Dongyang 543f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5442ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 545f7039b1dSLi Dongyang if (ret < 0) 546f7039b1dSLi Dongyang return ret; 547f7039b1dSLi Dongyang 548f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 549f7039b1dSLi Dongyang return -EFAULT; 550f7039b1dSLi Dongyang 551f7039b1dSLi Dongyang return 0; 552f7039b1dSLi Dongyang } 553f7039b1dSLi Dongyang 554dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 555dd5f9615SStefan Behrens { 55646e0f66aSChris Mason int i; 55746e0f66aSChris Mason 55846e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 55946e0f66aSChris Mason if (uuid[i]) 56046e0f66aSChris Mason return 0; 56146e0f66aSChris Mason } 56246e0f66aSChris Mason return 1; 563dd5f9615SStefan Behrens } 564dd5f9615SStefan Behrens 565d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 566cb8e7090SChristoph Hellwig struct dentry *dentry, 56752f75f4fSDavid Sterba const char *name, int namelen, 5686f72c7e2SArne Jansen u64 *async_transid, 5698696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 570f46b5a66SChristoph Hellwig { 5710b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 572f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 573f46b5a66SChristoph Hellwig struct btrfs_key key; 57449a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 575f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 576f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 577d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 57876dda93cSYan, Zheng struct btrfs_root *new_root; 579d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 58095582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5815662344bSTsutomu Itoh struct inode *inode; 582f46b5a66SChristoph Hellwig int ret; 583f46b5a66SChristoph Hellwig int err; 584f46b5a66SChristoph Hellwig u64 objectid; 585f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 5863de4586cSChris Mason u64 index = 0; 5878ea05e3aSAlexander Block uuid_le new_uuid; 588f46b5a66SChristoph Hellwig 58949a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 59049a3c4d9SDavid Sterba if (!root_item) 59149a3c4d9SDavid Sterba return -ENOMEM; 59249a3c4d9SDavid Sterba 5930b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5942fbe8c8aSAl Viro if (ret) 59549a3c4d9SDavid Sterba goto fail_free; 5966a912213SJosef Bacik 597e09fe2d2SQu Wenruo /* 598e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 59901327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 600e09fe2d2SQu Wenruo */ 60149a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 60249a3c4d9SDavid Sterba ret = -ENOSPC; 60349a3c4d9SDavid Sterba goto fail_free; 60449a3c4d9SDavid Sterba } 605e09fe2d2SQu Wenruo 606d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 6079ed74f2dSJosef Bacik /* 608d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 609d5c12070SMiao Xie * of create_snapshot(). 6109ed74f2dSJosef Bacik */ 611c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 612d5c12070SMiao Xie if (ret) 61349a3c4d9SDavid Sterba goto fail_free; 614f46b5a66SChristoph Hellwig 615d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 616d5c12070SMiao Xie if (IS_ERR(trans)) { 617d5c12070SMiao Xie ret = PTR_ERR(trans); 6187775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 61949a3c4d9SDavid Sterba goto fail_free; 620d5c12070SMiao Xie } 621d5c12070SMiao Xie trans->block_rsv = &block_rsv; 622d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 623f46b5a66SChristoph Hellwig 624a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 6256f72c7e2SArne Jansen if (ret) 6266f72c7e2SArne Jansen goto fail; 6276f72c7e2SArne Jansen 6284d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6298e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6308e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6318e8a1e31SJosef Bacik goto fail; 6328e8a1e31SJosef Bacik } 633f46b5a66SChristoph Hellwig 634f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 635f46b5a66SChristoph Hellwig 63649a3c4d9SDavid Sterba inode_item = &root_item->inode; 6373cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6383cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6393cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 640da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6410b246afaSJeff Mahoney fs_info->nodesize); 6423cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 643f46b5a66SChristoph Hellwig 64449a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 64549a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6463cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 64708fe4db1SLi Zefan 64849a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 64949a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 65049a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 65149a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 65249a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 65349a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 654f46b5a66SChristoph Hellwig 65549a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 65649a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6578ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 65849a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 65949a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 66049a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 66149a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 66249a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 66349a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 664f46b5a66SChristoph Hellwig 665925baeddSChris Mason btrfs_tree_unlock(leaf); 666f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 667f46b5a66SChristoph Hellwig leaf = NULL; 668f46b5a66SChristoph Hellwig 66949a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 670f46b5a66SChristoph Hellwig 671f46b5a66SChristoph Hellwig key.objectid = objectid; 6725d4f98a2SYan Zheng key.offset = 0; 673962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6740b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 67549a3c4d9SDavid Sterba root_item); 676f46b5a66SChristoph Hellwig if (ret) 677f46b5a66SChristoph Hellwig goto fail; 678f46b5a66SChristoph Hellwig 67976dda93cSYan, Zheng key.offset = (u64)-1; 6800b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 68179787eaaSJeff Mahoney if (IS_ERR(new_root)) { 68279787eaaSJeff Mahoney ret = PTR_ERR(new_root); 68366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 68479787eaaSJeff Mahoney goto fail; 68579787eaaSJeff Mahoney } 68676dda93cSYan, Zheng 68776dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 68876dda93cSYan, Zheng 68963541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 690ce598979SMark Fasheh if (ret) { 691ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 69266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 693ce598979SMark Fasheh goto fail; 694ce598979SMark Fasheh } 695ce598979SMark Fasheh 696f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 697f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 698f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 699f32e48e9SChandan Rajendra 700f46b5a66SChristoph Hellwig /* 701f46b5a66SChristoph Hellwig * insert the directory item 702f46b5a66SChristoph Hellwig */ 703877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 70479787eaaSJeff Mahoney if (ret) { 70566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 70679787eaaSJeff Mahoney goto fail; 70779787eaaSJeff Mahoney } 7083de4586cSChris Mason 709684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 7103de4586cSChris Mason BTRFS_FT_DIR, index); 71179787eaaSJeff Mahoney if (ret) { 71266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 713f46b5a66SChristoph Hellwig goto fail; 71479787eaaSJeff Mahoney } 7150660b5afSChris Mason 7166ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 71752c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 71852c26179SYan Zheng BUG_ON(ret); 71952c26179SYan Zheng 7206025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7214a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7220660b5afSChris Mason BUG_ON(ret); 7230660b5afSChris Mason 724cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7256bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 726dd5f9615SStefan Behrens if (ret) 72766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 728dd5f9615SStefan Behrens 729f46b5a66SChristoph Hellwig fail: 73049a3c4d9SDavid Sterba kfree(root_item); 731d5c12070SMiao Xie trans->block_rsv = NULL; 732d5c12070SMiao Xie trans->bytes_reserved = 0; 7337775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 734de6e8200SLiu Bo 73572fd032eSSage Weil if (async_transid) { 73672fd032eSSage Weil *async_transid = trans->transid; 7373a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 73800d71c9cSMiao Xie if (err) 7393a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 74072fd032eSSage Weil } else { 7413a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 74272fd032eSSage Weil } 743f46b5a66SChristoph Hellwig if (err && !ret) 744f46b5a66SChristoph Hellwig ret = err; 7451a65e24bSChris Mason 7465662344bSTsutomu Itoh if (!ret) { 7475662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 748de6e8200SLiu Bo if (IS_ERR(inode)) 749de6e8200SLiu Bo return PTR_ERR(inode); 7505662344bSTsutomu Itoh d_instantiate(dentry, inode); 7515662344bSTsutomu Itoh } 752f46b5a66SChristoph Hellwig return ret; 75349a3c4d9SDavid Sterba 75449a3c4d9SDavid Sterba fail_free: 75549a3c4d9SDavid Sterba kfree(root_item); 75649a3c4d9SDavid Sterba return ret; 757f46b5a66SChristoph Hellwig } 758f46b5a66SChristoph Hellwig 759e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 76061d7e4cbSDavid Sterba struct dentry *dentry, 761e9662f70SMiao Xie u64 *async_transid, bool readonly, 762e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 763f46b5a66SChristoph Hellwig { 7640b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7652e4bfab9SYan, Zheng struct inode *inode; 766f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 767f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7682e4bfab9SYan, Zheng int ret; 7698ecebf4dSRobbie Ko bool snapshot_force_cow = false; 770f46b5a66SChristoph Hellwig 77127cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 772f46b5a66SChristoph Hellwig return -EINVAL; 773f46b5a66SChristoph Hellwig 774eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 775eede2bf3SOmar Sandoval btrfs_warn(fs_info, 776eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 777eede2bf3SOmar Sandoval return -ETXTBSY; 778eede2bf3SOmar Sandoval } 779eede2bf3SOmar Sandoval 78023269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 781a1ee7362SDavid Sterba if (!pending_snapshot) 782a1ee7362SDavid Sterba return -ENOMEM; 783a1ee7362SDavid Sterba 784b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 78523269bf5SDavid Sterba GFP_KERNEL); 7868546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7878546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 788b0c0ea63SDavid Sterba ret = -ENOMEM; 789b0c0ea63SDavid Sterba goto free_pending; 790b0c0ea63SDavid Sterba } 791b0c0ea63SDavid Sterba 7928ecebf4dSRobbie Ko /* 7938ecebf4dSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 7948ecebf4dSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 7958ecebf4dSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 7968ecebf4dSRobbie Ko */ 797ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7984e857c58SPeter Zijlstra smp_mb__after_atomic(); 79945bac0f3SLiu Bo /* wait for no snapshot writes */ 80045bac0f3SLiu Bo wait_event(root->subv_writers->wait, 80145bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 8028257b2dcSMiao Xie 8033cd24c69SEthan Lien ret = btrfs_start_delalloc_snapshot(root); 8046a03843dSMiao Xie if (ret) 805a1ee7362SDavid Sterba goto dec_and_free; 8066a03843dSMiao Xie 8078ecebf4dSRobbie Ko /* 8088ecebf4dSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 8098ecebf4dSRobbie Ko * we force future writes to fallback to COW mode during snapshot 8108ecebf4dSRobbie Ko * creation. 8118ecebf4dSRobbie Ko */ 8128ecebf4dSRobbie Ko atomic_inc(&root->snapshot_force_cow); 8138ecebf4dSRobbie Ko snapshot_force_cow = true; 8148ecebf4dSRobbie Ko 8156374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 8166a03843dSMiao Xie 81766d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 81866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 819d5c12070SMiao Xie /* 820d5c12070SMiao Xie * 1 - parent dir inode 821d5c12070SMiao Xie * 2 - dir entries 822d5c12070SMiao Xie * 1 - root item 823d5c12070SMiao Xie * 2 - root ref/backref 824d5c12070SMiao Xie * 1 - root of snapshot 825dd5f9615SStefan Behrens * 1 - UUID item 826d5c12070SMiao Xie */ 827d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 828dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 829ee3441b4SJeff Mahoney false); 830d5c12070SMiao Xie if (ret) 831a1ee7362SDavid Sterba goto dec_and_free; 832d5c12070SMiao Xie 833a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 834a22285a6SYan, Zheng pending_snapshot->root = root; 835b83cc969SLi Zefan pending_snapshot->readonly = readonly; 836e9662f70SMiao Xie pending_snapshot->dir = dir; 8378696c533SMiao Xie pending_snapshot->inherit = inherit; 838a22285a6SYan, Zheng 839d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 840a22285a6SYan, Zheng if (IS_ERR(trans)) { 841a22285a6SYan, Zheng ret = PTR_ERR(trans); 842a22285a6SYan, Zheng goto fail; 843a22285a6SYan, Zheng } 844a22285a6SYan, Zheng 8450b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 846a22285a6SYan, Zheng list_add(&pending_snapshot->list, 847a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8480b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 84972fd032eSSage Weil if (async_transid) { 85072fd032eSSage Weil *async_transid = trans->transid; 8513a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 85200d71c9cSMiao Xie if (ret) 8533a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 85472fd032eSSage Weil } else { 8553a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 85672fd032eSSage Weil } 857aec8030aSMiao Xie if (ret) 858c37b2b62SJosef Bacik goto fail; 859a22285a6SYan, Zheng 860a22285a6SYan, Zheng ret = pending_snapshot->error; 861f46b5a66SChristoph Hellwig if (ret) 8622e4bfab9SYan, Zheng goto fail; 863f46b5a66SChristoph Hellwig 864d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 865d3797308SChris Mason if (ret) 866d3797308SChris Mason goto fail; 867d3797308SChris Mason 8682b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8692e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8702e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8712e4bfab9SYan, Zheng goto fail; 8722e4bfab9SYan, Zheng } 8735662344bSTsutomu Itoh 8742e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8752e4bfab9SYan, Zheng ret = 0; 8762e4bfab9SYan, Zheng fail: 8777775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 878a1ee7362SDavid Sterba dec_and_free: 8798ecebf4dSRobbie Ko if (snapshot_force_cow) 8808ecebf4dSRobbie Ko atomic_dec(&root->snapshot_force_cow); 881ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8824625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 883b0c0ea63SDavid Sterba free_pending: 884b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8858546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 886a1ee7362SDavid Sterba kfree(pending_snapshot); 887a1ee7362SDavid Sterba 888f46b5a66SChristoph Hellwig return ret; 889f46b5a66SChristoph Hellwig } 890f46b5a66SChristoph Hellwig 8914260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8924260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8934260f7c7SSage Weil * whether the type of victim is right. 8944260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8954260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8964260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8974260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8984260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8994260f7c7SSage Weil * a. be owner of dir, or 9004260f7c7SSage Weil * b. be owner of victim, or 9014260f7c7SSage Weil * c. have CAP_FOWNER capability 90201327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 9034260f7c7SSage Weil * links pointing to it. 9044260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 9054260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 9064260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 9074260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 9084260f7c7SSage Weil * nfs_async_unlink(). 9094260f7c7SSage Weil */ 9104260f7c7SSage Weil 9114260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 9124260f7c7SSage Weil { 9134260f7c7SSage Weil int error; 9144260f7c7SSage Weil 9152b0143b5SDavid Howells if (d_really_is_negative(victim)) 9164260f7c7SSage Weil return -ENOENT; 9174260f7c7SSage Weil 9182b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 9194fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 9204260f7c7SSage Weil 9214260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 9224260f7c7SSage Weil if (error) 9234260f7c7SSage Weil return error; 9244260f7c7SSage Weil if (IS_APPEND(dir)) 9254260f7c7SSage Weil return -EPERM; 9262b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9272b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9284260f7c7SSage Weil return -EPERM; 9294260f7c7SSage Weil if (isdir) { 930e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9314260f7c7SSage Weil return -ENOTDIR; 9324260f7c7SSage Weil if (IS_ROOT(victim)) 9334260f7c7SSage Weil return -EBUSY; 934e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9354260f7c7SSage Weil return -EISDIR; 9364260f7c7SSage Weil if (IS_DEADDIR(dir)) 9374260f7c7SSage Weil return -ENOENT; 9384260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9394260f7c7SSage Weil return -EBUSY; 9404260f7c7SSage Weil return 0; 9414260f7c7SSage Weil } 9424260f7c7SSage Weil 943cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 944cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 945cb8e7090SChristoph Hellwig { 9462b0143b5SDavid Howells if (d_really_is_positive(child)) 947cb8e7090SChristoph Hellwig return -EEXIST; 948cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 949cb8e7090SChristoph Hellwig return -ENOENT; 950cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 951cb8e7090SChristoph Hellwig } 952cb8e7090SChristoph Hellwig 953cb8e7090SChristoph Hellwig /* 954cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 955cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 956cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 957cb8e7090SChristoph Hellwig */ 95892872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 95952f75f4fSDavid Sterba const char *name, int namelen, 96072fd032eSSage Weil struct btrfs_root *snap_src, 9616f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9628696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 963cb8e7090SChristoph Hellwig { 9642b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 966cb8e7090SChristoph Hellwig struct dentry *dentry; 967cb8e7090SChristoph Hellwig int error; 968cb8e7090SChristoph Hellwig 96900235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 97000235411SAl Viro if (error == -EINTR) 97100235411SAl Viro return error; 972cb8e7090SChristoph Hellwig 973cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 974cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 975cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 976cb8e7090SChristoph Hellwig goto out_unlock; 977cb8e7090SChristoph Hellwig 97876dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 979cb8e7090SChristoph Hellwig if (error) 980a874a63eSLiu Bo goto out_dput; 981cb8e7090SChristoph Hellwig 9829c52057cSChris Mason /* 9839c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9849c52057cSChris Mason * check for them now when we can safely fail 9859c52057cSChris Mason */ 9869c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9879c52057cSChris Mason dir->i_ino, name, 9889c52057cSChris Mason namelen); 9899c52057cSChris Mason if (error) 9909c52057cSChris Mason goto out_dput; 9919c52057cSChris Mason 9920b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 99376dda93cSYan, Zheng 99476dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 99576dda93cSYan, Zheng goto out_up_read; 99676dda93cSYan, Zheng 9973de4586cSChris Mason if (snap_src) { 99861d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9996f72c7e2SArne Jansen async_transid, readonly, inherit); 10003de4586cSChris Mason } else { 1001d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 1002d5c12070SMiao Xie async_transid, inherit); 10033de4586cSChris Mason } 100476dda93cSYan, Zheng if (!error) 100576dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 100676dda93cSYan, Zheng out_up_read: 10070b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 1008cb8e7090SChristoph Hellwig out_dput: 1009cb8e7090SChristoph Hellwig dput(dentry); 1010cb8e7090SChristoph Hellwig out_unlock: 10115955102cSAl Viro inode_unlock(dir); 1012cb8e7090SChristoph Hellwig return error; 1013cb8e7090SChristoph Hellwig } 1014cb8e7090SChristoph Hellwig 10154cb5300bSChris Mason /* 10164cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 10174cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 10184cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 10194cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 10204cb5300bSChris Mason * part of the file 10214cb5300bSChris Mason */ 1022aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 10234cb5300bSChris Mason { 10244cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10254cb5300bSChris Mason struct extent_map *em = NULL; 10264cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10274cb5300bSChris Mason u64 end; 10284cb5300bSChris Mason 10294cb5300bSChris Mason read_lock(&em_tree->lock); 103009cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10314cb5300bSChris Mason read_unlock(&em_tree->lock); 10324cb5300bSChris Mason 10334cb5300bSChris Mason if (em) { 10344cb5300bSChris Mason end = extent_map_end(em); 10354cb5300bSChris Mason free_extent_map(em); 10364cb5300bSChris Mason if (end - offset > thresh) 10374cb5300bSChris Mason return 0; 10384cb5300bSChris Mason } 10394cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10404cb5300bSChris Mason thresh /= 2; 10414cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10424cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10434cb5300bSChris Mason if (end >= thresh) 10444cb5300bSChris Mason return 0; 10454cb5300bSChris Mason return 1; 10464cb5300bSChris Mason } 10474cb5300bSChris Mason 10484cb5300bSChris Mason /* 10494cb5300bSChris Mason * helper function to walk through a file and find extents 10504cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10514cb5300bSChris Mason * 10524cb5300bSChris Mason * This is used by the defragging code to find new and small 10534cb5300bSChris Mason * extents 10544cb5300bSChris Mason */ 10554cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10564cb5300bSChris Mason struct inode *inode, u64 newer_than, 1057aab110abSDavid Sterba u64 *off, u32 thresh) 10584cb5300bSChris Mason { 10594cb5300bSChris Mason struct btrfs_path *path; 10604cb5300bSChris Mason struct btrfs_key min_key; 10614cb5300bSChris Mason struct extent_buffer *leaf; 10624cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10634cb5300bSChris Mason int type; 10644cb5300bSChris Mason int ret; 10654a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10664cb5300bSChris Mason 10674cb5300bSChris Mason path = btrfs_alloc_path(); 10684cb5300bSChris Mason if (!path) 10694cb5300bSChris Mason return -ENOMEM; 10704cb5300bSChris Mason 1071a4689d2bSDavid Sterba min_key.objectid = ino; 10724cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10734cb5300bSChris Mason min_key.offset = *off; 10744cb5300bSChris Mason 10754cb5300bSChris Mason while (1) { 10766174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10774cb5300bSChris Mason if (ret != 0) 10784cb5300bSChris Mason goto none; 1079f094c9bdSFilipe Manana process_slot: 1080a4689d2bSDavid Sterba if (min_key.objectid != ino) 10814cb5300bSChris Mason goto none; 10824cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10834cb5300bSChris Mason goto none; 10844cb5300bSChris Mason 10854cb5300bSChris Mason leaf = path->nodes[0]; 10864cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10874cb5300bSChris Mason struct btrfs_file_extent_item); 10884cb5300bSChris Mason 10894cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10904cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10914cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10924cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10934cb5300bSChris Mason *off = min_key.offset; 10944cb5300bSChris Mason btrfs_free_path(path); 10954cb5300bSChris Mason return 0; 10964cb5300bSChris Mason } 10974cb5300bSChris Mason 1098f094c9bdSFilipe Manana path->slots[0]++; 1099f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1100f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1101f094c9bdSFilipe Manana goto process_slot; 1102f094c9bdSFilipe Manana } 1103f094c9bdSFilipe Manana 11044cb5300bSChris Mason if (min_key.offset == (u64)-1) 11054cb5300bSChris Mason goto none; 11064cb5300bSChris Mason 11074cb5300bSChris Mason min_key.offset++; 11084cb5300bSChris Mason btrfs_release_path(path); 11094cb5300bSChris Mason } 11104cb5300bSChris Mason none: 11114cb5300bSChris Mason btrfs_free_path(path); 11124cb5300bSChris Mason return -ENOENT; 11134cb5300bSChris Mason } 11144cb5300bSChris Mason 11156c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 111617ce6ef8SLiu Bo { 111717ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1118940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 11196c282eb4SLi Zefan struct extent_map *em; 112009cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1121940100a4SChris Mason 1122940100a4SChris Mason /* 1123940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1124940100a4SChris Mason * the full extent lock 1125940100a4SChris Mason */ 1126940100a4SChris Mason read_lock(&em_tree->lock); 1127940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1128940100a4SChris Mason read_unlock(&em_tree->lock); 1129940100a4SChris Mason 1130940100a4SChris Mason if (!em) { 1131308d9800SFilipe Manana struct extent_state *cached = NULL; 1132308d9800SFilipe Manana u64 end = start + len - 1; 1133308d9800SFilipe Manana 1134940100a4SChris Mason /* get the big lock and read metadata off disk */ 1135ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1136fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1137e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1138940100a4SChris Mason 11396cf8bfbfSDan Carpenter if (IS_ERR(em)) 11406c282eb4SLi Zefan return NULL; 1141940100a4SChris Mason } 1142940100a4SChris Mason 11436c282eb4SLi Zefan return em; 11446c282eb4SLi Zefan } 11456c282eb4SLi Zefan 11466c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11476c282eb4SLi Zefan { 11486c282eb4SLi Zefan struct extent_map *next; 11496c282eb4SLi Zefan bool ret = true; 11506c282eb4SLi Zefan 11516c282eb4SLi Zefan /* this is the last extent */ 11526c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11536c282eb4SLi Zefan return false; 11546c282eb4SLi Zefan 11556c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1156e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1157e9512d72SChris Mason ret = false; 1158e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1159ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11606c282eb4SLi Zefan ret = false; 11616c282eb4SLi Zefan 11626c282eb4SLi Zefan free_extent_map(next); 11636c282eb4SLi Zefan return ret; 11646c282eb4SLi Zefan } 11656c282eb4SLi Zefan 1166aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1167a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1168a43a2111SAndrew Mahone int compress) 11696c282eb4SLi Zefan { 11706c282eb4SLi Zefan struct extent_map *em; 11716c282eb4SLi Zefan int ret = 1; 11726c282eb4SLi Zefan bool next_mergeable = true; 11734a3560c4SLiu Bo bool prev_mergeable = true; 11746c282eb4SLi Zefan 11756c282eb4SLi Zefan /* 11766c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11776c282eb4SLi Zefan * defragging it 11786c282eb4SLi Zefan */ 11796c282eb4SLi Zefan if (start < *defrag_end) 11806c282eb4SLi Zefan return 1; 11816c282eb4SLi Zefan 11826c282eb4SLi Zefan *skip = 0; 11836c282eb4SLi Zefan 11846c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11856c282eb4SLi Zefan if (!em) 11866c282eb4SLi Zefan return 0; 11876c282eb4SLi Zefan 1188940100a4SChris Mason /* this will cover holes, and inline extents */ 118917ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1190940100a4SChris Mason ret = 0; 119117ce6ef8SLiu Bo goto out; 119217ce6ef8SLiu Bo } 119317ce6ef8SLiu Bo 11944a3560c4SLiu Bo if (!*defrag_end) 11954a3560c4SLiu Bo prev_mergeable = false; 11964a3560c4SLiu Bo 11976c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1198940100a4SChris Mason /* 11996c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 12006c282eb4SLi Zefan * real extent, don't bother defragging it 1201940100a4SChris Mason */ 1202a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 12034a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1204940100a4SChris Mason ret = 0; 120517ce6ef8SLiu Bo out: 1206940100a4SChris Mason /* 1207940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1208940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1209940100a4SChris Mason * so that the next tiny extent will force a defrag. 1210940100a4SChris Mason * 1211940100a4SChris Mason * The end result of this is that tiny extents before a single big 1212940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1213940100a4SChris Mason */ 1214940100a4SChris Mason if (ret) { 1215940100a4SChris Mason *defrag_end = extent_map_end(em); 1216940100a4SChris Mason } else { 1217940100a4SChris Mason *last_len = 0; 1218940100a4SChris Mason *skip = extent_map_end(em); 1219940100a4SChris Mason *defrag_end = 0; 1220940100a4SChris Mason } 1221940100a4SChris Mason 1222940100a4SChris Mason free_extent_map(em); 1223940100a4SChris Mason return ret; 1224940100a4SChris Mason } 1225940100a4SChris Mason 12264cb5300bSChris Mason /* 12274cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12284cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12294cb5300bSChris Mason * to COW and defrag. 12304cb5300bSChris Mason * 12314cb5300bSChris Mason * It also makes sure the delalloc code has enough 12324cb5300bSChris Mason * dirty data to avoid making new small extents as part 12334cb5300bSChris Mason * of the defrag 12344cb5300bSChris Mason * 12354cb5300bSChris Mason * It's a good idea to start RA on this range 12364cb5300bSChris Mason * before calling this. 12374cb5300bSChris Mason */ 12384cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12394cb5300bSChris Mason struct page **pages, 12404cb5300bSChris Mason unsigned long start_index, 1241c41570c9SJustin Maggard unsigned long num_pages) 1242f46b5a66SChristoph Hellwig { 12434cb5300bSChris Mason unsigned long file_end; 12444cb5300bSChris Mason u64 isize = i_size_read(inode); 1245f46b5a66SChristoph Hellwig u64 page_start; 1246f46b5a66SChristoph Hellwig u64 page_end; 12471f12bd06SLiu Bo u64 page_cnt; 12484cb5300bSChris Mason int ret; 12494cb5300bSChris Mason int i; 12504cb5300bSChris Mason int i_done; 12514cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12524cb5300bSChris Mason struct extent_state *cached_state = NULL; 1253600a45e1SMiao Xie struct extent_io_tree *tree; 1254364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12553b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12564cb5300bSChris Mason 125709cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12581f12bd06SLiu Bo if (!isize || start_index > file_end) 12591f12bd06SLiu Bo return 0; 12601f12bd06SLiu Bo 12611f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12624cb5300bSChris Mason 1263364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 126409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 126509cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12664cb5300bSChris Mason if (ret) 12674cb5300bSChris Mason return ret; 12684cb5300bSChris Mason i_done = 0; 1269600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12704cb5300bSChris Mason 12714cb5300bSChris Mason /* step one, lock all the pages */ 12721f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12734cb5300bSChris Mason struct page *page; 1274600a45e1SMiao Xie again: 1275a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12763b16a4e3SJosef Bacik start_index + i, mask); 12774cb5300bSChris Mason if (!page) 12784cb5300bSChris Mason break; 12794cb5300bSChris Mason 1280600a45e1SMiao Xie page_start = page_offset(page); 128109cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1282600a45e1SMiao Xie while (1) { 1283308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1284ff13db41SDavid Sterba &cached_state); 1285600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1286600a45e1SMiao Xie page_start); 1287308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1288e43bbe5eSDavid Sterba &cached_state); 1289600a45e1SMiao Xie if (!ordered) 1290600a45e1SMiao Xie break; 1291600a45e1SMiao Xie 1292600a45e1SMiao Xie unlock_page(page); 1293600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1294600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1295600a45e1SMiao Xie lock_page(page); 12961f12bd06SLiu Bo /* 12971f12bd06SLiu Bo * we unlocked the page above, so we need check if 12981f12bd06SLiu Bo * it was released or not. 12991f12bd06SLiu Bo */ 13001f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 13011f12bd06SLiu Bo unlock_page(page); 130209cbfeafSKirill A. Shutemov put_page(page); 13031f12bd06SLiu Bo goto again; 13041f12bd06SLiu Bo } 1305600a45e1SMiao Xie } 1306600a45e1SMiao Xie 13074cb5300bSChris Mason if (!PageUptodate(page)) { 13084cb5300bSChris Mason btrfs_readpage(NULL, page); 13094cb5300bSChris Mason lock_page(page); 13104cb5300bSChris Mason if (!PageUptodate(page)) { 13114cb5300bSChris Mason unlock_page(page); 131209cbfeafSKirill A. Shutemov put_page(page); 13134cb5300bSChris Mason ret = -EIO; 13144cb5300bSChris Mason break; 13154cb5300bSChris Mason } 13164cb5300bSChris Mason } 1317600a45e1SMiao Xie 1318600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1319600a45e1SMiao Xie unlock_page(page); 132009cbfeafSKirill A. Shutemov put_page(page); 1321600a45e1SMiao Xie goto again; 1322600a45e1SMiao Xie } 1323600a45e1SMiao Xie 13244cb5300bSChris Mason pages[i] = page; 13254cb5300bSChris Mason i_done++; 13264cb5300bSChris Mason } 13274cb5300bSChris Mason if (!i_done || ret) 13284cb5300bSChris Mason goto out; 13294cb5300bSChris Mason 13301751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13314cb5300bSChris Mason goto out; 13324cb5300bSChris Mason 13334cb5300bSChris Mason /* 13344cb5300bSChris Mason * so now we have a nice long stream of locked 13354cb5300bSChris Mason * and up to date pages, lets wait on them 13364cb5300bSChris Mason */ 13374cb5300bSChris Mason for (i = 0; i < i_done; i++) 13384cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13394cb5300bSChris Mason 13404cb5300bSChris Mason page_start = page_offset(pages[0]); 134109cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13424cb5300bSChris Mason 13434cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1344ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13454cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 13464cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 13479e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1348ae0f1625SDavid Sterba &cached_state); 13494cb5300bSChris Mason 13501f12bd06SLiu Bo if (i_done != page_cnt) { 13519e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 135228c4a3e2SSu Yue btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); 13539e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1354bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 135509cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 135643b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13574cb5300bSChris Mason } 13584cb5300bSChris Mason 13594cb5300bSChris Mason 13609e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1361018ed4f7SDavid Sterba &cached_state); 13624cb5300bSChris Mason 13634cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1364e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13654cb5300bSChris Mason 13664cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13674cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13684cb5300bSChris Mason ClearPageChecked(pages[i]); 13694cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13704cb5300bSChris Mason set_page_dirty(pages[i]); 13714cb5300bSChris Mason unlock_page(pages[i]); 137209cbfeafSKirill A. Shutemov put_page(pages[i]); 13734cb5300bSChris Mason } 137443b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 137543b18595SQu Wenruo false); 1376364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13774cb5300bSChris Mason return i_done; 13784cb5300bSChris Mason out: 13794cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13804cb5300bSChris Mason unlock_page(pages[i]); 138109cbfeafSKirill A. Shutemov put_page(pages[i]); 13824cb5300bSChris Mason } 1383bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 138409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 138543b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 138643b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 138743b18595SQu Wenruo true); 1388364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13894cb5300bSChris Mason return ret; 13904cb5300bSChris Mason 13914cb5300bSChris Mason } 13924cb5300bSChris Mason 13934cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13944cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13954cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13964cb5300bSChris Mason { 13970b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13984cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13994cb5300bSChris Mason struct file_ra_state *ra = NULL; 14004cb5300bSChris Mason unsigned long last_index; 1401151a31b2SLi Zefan u64 isize = i_size_read(inode); 1402940100a4SChris Mason u64 last_len = 0; 1403940100a4SChris Mason u64 skip = 0; 1404940100a4SChris Mason u64 defrag_end = 0; 14054cb5300bSChris Mason u64 newer_off = range->start; 1406f46b5a66SChristoph Hellwig unsigned long i; 1407008873eaSLi Zefan unsigned long ra_index = 0; 1408f46b5a66SChristoph Hellwig int ret; 14094cb5300bSChris Mason int defrag_count = 0; 14101a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1411aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 141209cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1413c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1414ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 14154cb5300bSChris Mason struct page **pages = NULL; 14161e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14174cb5300bSChris Mason 14180abd5b17SLiu Bo if (isize == 0) 14190abd5b17SLiu Bo return 0; 14200abd5b17SLiu Bo 14210abd5b17SLiu Bo if (range->start >= isize) 14220abd5b17SLiu Bo return -EINVAL; 14231a419d85SLi Zefan 14241e2ef46dSDavid Sterba if (do_compress) { 14251a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 14261a419d85SLi Zefan return -EINVAL; 14271a419d85SLi Zefan if (range->compress_type) 14281a419d85SLi Zefan compress_type = range->compress_type; 14291a419d85SLi Zefan } 1430f46b5a66SChristoph Hellwig 14310abd5b17SLiu Bo if (extent_thresh == 0) 1432ee22184bSByongho Lee extent_thresh = SZ_256K; 1433f46b5a66SChristoph Hellwig 14344cb5300bSChris Mason /* 14350a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14360a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14370a52d108SDavid Sterba * we don't error out. 14384cb5300bSChris Mason */ 14394cb5300bSChris Mason if (!file) { 144063e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14410a52d108SDavid Sterba if (ra) 14424cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14434cb5300bSChris Mason } else { 14444cb5300bSChris Mason ra = &file->f_ra; 14454cb5300bSChris Mason } 14464cb5300bSChris Mason 144763e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14484cb5300bSChris Mason if (!pages) { 14494cb5300bSChris Mason ret = -ENOMEM; 14504cb5300bSChris Mason goto out_ra; 14514cb5300bSChris Mason } 14524cb5300bSChris Mason 14534cb5300bSChris Mason /* find the last page to defrag */ 14541e701a32SChris Mason if (range->start + range->len > range->start) { 1455151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 145609cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14571e701a32SChris Mason } else { 145809cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14591e701a32SChris Mason } 14601e701a32SChris Mason 14614cb5300bSChris Mason if (newer_than) { 14624cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1463ee22184bSByongho Lee &newer_off, SZ_64K); 14644cb5300bSChris Mason if (!ret) { 14654cb5300bSChris Mason range->start = newer_off; 14664cb5300bSChris Mason /* 14674cb5300bSChris Mason * we always align our defrag to help keep 14684cb5300bSChris Mason * the extents in the file evenly spaced 14694cb5300bSChris Mason */ 147009cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14714cb5300bSChris Mason } else 14724cb5300bSChris Mason goto out_ra; 14734cb5300bSChris Mason } else { 147409cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14754cb5300bSChris Mason } 14764cb5300bSChris Mason if (!max_to_defrag) 1477070034bdSchandan max_to_defrag = last_index - i + 1; 14784cb5300bSChris Mason 14792a0f7f57SLi Zefan /* 14802a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14812a0f7f57SLi Zefan * written sequentially. 14822a0f7f57SLi Zefan */ 14832a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14842a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14852a0f7f57SLi Zefan 1486f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 148709cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14884cb5300bSChris Mason /* 14894cb5300bSChris Mason * make sure we stop running if someone unmounts 14904cb5300bSChris Mason * the FS 14914cb5300bSChris Mason */ 14921751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14934cb5300bSChris Mason break; 14944cb5300bSChris Mason 14950b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14960b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1497210549ebSDavid Sterba ret = -EAGAIN; 1498210549ebSDavid Sterba break; 1499210549ebSDavid Sterba } 1500210549ebSDavid Sterba 150109cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 15026c282eb4SLi Zefan extent_thresh, &last_len, &skip, 15031e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1504940100a4SChris Mason unsigned long next; 1505940100a4SChris Mason /* 1506940100a4SChris Mason * the should_defrag function tells us how much to skip 1507940100a4SChris Mason * bump our counter by the suggested amount 1508940100a4SChris Mason */ 150909cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1510940100a4SChris Mason i = max(i + 1, next); 1511940100a4SChris Mason continue; 1512940100a4SChris Mason } 1513008873eaSLi Zefan 1514008873eaSLi Zefan if (!newer_than) { 151509cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 151609cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1517008873eaSLi Zefan cluster = min(cluster, max_cluster); 1518008873eaSLi Zefan } else { 1519008873eaSLi Zefan cluster = max_cluster; 1520008873eaSLi Zefan } 1521008873eaSLi Zefan 1522008873eaSLi Zefan if (i + cluster > ra_index) { 1523008873eaSLi Zefan ra_index = max(i, ra_index); 15240a52d108SDavid Sterba if (ra) 1525d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1526d3c0bab5SDavid Sterba file, ra_index, cluster); 1527e4826a5bSchandan ra_index += cluster; 1528008873eaSLi Zefan } 15294cb5300bSChris Mason 15305955102cSAl Viro inode_lock(inode); 1531eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1532eede2bf3SOmar Sandoval ret = -ETXTBSY; 1533eede2bf3SOmar Sandoval } else { 15341e2ef46dSDavid Sterba if (do_compress) 1535eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1536008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1537eede2bf3SOmar Sandoval } 1538ecb8bea8SLiu Bo if (ret < 0) { 15395955102cSAl Viro inode_unlock(inode); 15404cb5300bSChris Mason goto out_ra; 1541ecb8bea8SLiu Bo } 15424cb5300bSChris Mason 15434cb5300bSChris Mason defrag_count += ret; 1544d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15455955102cSAl Viro inode_unlock(inode); 15464cb5300bSChris Mason 15474cb5300bSChris Mason if (newer_than) { 15484cb5300bSChris Mason if (newer_off == (u64)-1) 15494cb5300bSChris Mason break; 15504cb5300bSChris Mason 1551e1f041e1SLiu Bo if (ret > 0) 1552e1f041e1SLiu Bo i += ret; 1553e1f041e1SLiu Bo 15544cb5300bSChris Mason newer_off = max(newer_off + 1, 155509cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15564cb5300bSChris Mason 1557ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1558ee22184bSByongho Lee &newer_off, SZ_64K); 15594cb5300bSChris Mason if (!ret) { 15604cb5300bSChris Mason range->start = newer_off; 156109cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15624cb5300bSChris Mason } else { 15634cb5300bSChris Mason break; 1564940100a4SChris Mason } 15654cb5300bSChris Mason } else { 1566008873eaSLi Zefan if (ret > 0) { 1567cbcc8326SLi Zefan i += ret; 156809cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1569008873eaSLi Zefan } else { 1570940100a4SChris Mason i++; 1571008873eaSLi Zefan last_len = 0; 1572008873eaSLi Zefan } 1573f46b5a66SChristoph Hellwig } 15744cb5300bSChris Mason } 1575f46b5a66SChristoph Hellwig 1576dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15771e701a32SChris Mason filemap_flush(inode->i_mapping); 1578dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1579dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1580dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1581dec8ef90SFilipe Manana } 15821e701a32SChris Mason 15831a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15840b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15855c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15865c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15871a419d85SLi Zefan } 15881a419d85SLi Zefan 158960ccf82fSDiego Calleja ret = defrag_count; 1590940100a4SChris Mason 15914cb5300bSChris Mason out_ra: 15921e2ef46dSDavid Sterba if (do_compress) { 15935955102cSAl Viro inode_lock(inode); 1594eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15955955102cSAl Viro inode_unlock(inode); 1596633085c7SFilipe David Borba Manana } 15974cb5300bSChris Mason if (!file) 15984cb5300bSChris Mason kfree(ra); 15994cb5300bSChris Mason kfree(pages); 1600940100a4SChris Mason return ret; 1601f46b5a66SChristoph Hellwig } 1602f46b5a66SChristoph Hellwig 1603198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 160476dda93cSYan, Zheng void __user *arg) 1605f46b5a66SChristoph Hellwig { 16060b246afaSJeff Mahoney struct inode *inode = file_inode(file); 16070b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1608f46b5a66SChristoph Hellwig u64 new_size; 1609f46b5a66SChristoph Hellwig u64 old_size; 1610f46b5a66SChristoph Hellwig u64 devid = 1; 16110b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1612f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1613f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1614f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1615f46b5a66SChristoph Hellwig char *sizestr; 16169a40f122SGui Hecheng char *retptr; 1617f46b5a66SChristoph Hellwig char *devstr = NULL; 1618f46b5a66SChristoph Hellwig int ret = 0; 1619f46b5a66SChristoph Hellwig int mod = 0; 1620f46b5a66SChristoph Hellwig 1621e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1622e441d54dSChris Mason return -EPERM; 1623e441d54dSChris Mason 1624198605a8SMiao Xie ret = mnt_want_write_file(file); 1625198605a8SMiao Xie if (ret) 1626198605a8SMiao Xie return ret; 1627198605a8SMiao Xie 1628171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 162997547676SMiao Xie mnt_drop_write_file(file); 1630e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1631c9e9f97bSIlya Dryomov } 1632c9e9f97bSIlya Dryomov 1633dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1634c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1635c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1636c9e9f97bSIlya Dryomov goto out; 1637c9e9f97bSIlya Dryomov } 16385516e595SMark Fasheh 16395516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1640f46b5a66SChristoph Hellwig 1641f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1642f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1643f46b5a66SChristoph Hellwig if (devstr) { 1644f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1645f46b5a66SChristoph Hellwig *devstr = '\0'; 1646f46b5a66SChristoph Hellwig devstr = vol_args->name; 164758dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 164858dfae63SZhangZhen if (ret) 164958dfae63SZhangZhen goto out_free; 1650dfd79829SMiao Xie if (!devid) { 1651dfd79829SMiao Xie ret = -EINVAL; 1652dfd79829SMiao Xie goto out_free; 1653dfd79829SMiao Xie } 16540b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1655f46b5a66SChristoph Hellwig } 1656dba60f3fSMiao Xie 165709ba3bc9SAnand Jain device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true); 1658f46b5a66SChristoph Hellwig if (!device) { 16590b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1660c1c9ff7cSGeert Uytterhoeven devid); 1661dfd79829SMiao Xie ret = -ENODEV; 1662c9e9f97bSIlya Dryomov goto out_free; 1663f46b5a66SChristoph Hellwig } 1664dba60f3fSMiao Xie 1665ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16660b246afaSJeff Mahoney btrfs_info(fs_info, 1667efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1668c1c9ff7cSGeert Uytterhoeven devid); 1669dfd79829SMiao Xie ret = -EPERM; 16704e42ae1bSLiu Bo goto out_free; 16714e42ae1bSLiu Bo } 16724e42ae1bSLiu Bo 1673f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1674f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1675f46b5a66SChristoph Hellwig else { 1676f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1677f46b5a66SChristoph Hellwig mod = -1; 1678f46b5a66SChristoph Hellwig sizestr++; 1679f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1680f46b5a66SChristoph Hellwig mod = 1; 1681f46b5a66SChristoph Hellwig sizestr++; 1682f46b5a66SChristoph Hellwig } 16839a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16849a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1685f46b5a66SChristoph Hellwig ret = -EINVAL; 1686c9e9f97bSIlya Dryomov goto out_free; 1687f46b5a66SChristoph Hellwig } 1688f46b5a66SChristoph Hellwig } 1689f46b5a66SChristoph Hellwig 1690401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1691dfd79829SMiao Xie ret = -EPERM; 169263a212abSStefan Behrens goto out_free; 169363a212abSStefan Behrens } 169463a212abSStefan Behrens 16957cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1696f46b5a66SChristoph Hellwig 1697f46b5a66SChristoph Hellwig if (mod < 0) { 1698f46b5a66SChristoph Hellwig if (new_size > old_size) { 1699f46b5a66SChristoph Hellwig ret = -EINVAL; 1700c9e9f97bSIlya Dryomov goto out_free; 1701f46b5a66SChristoph Hellwig } 1702f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1703f46b5a66SChristoph Hellwig } else if (mod > 0) { 1704eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1705902c68a4SGui Hecheng ret = -ERANGE; 1706eb8052e0SWenliang Fan goto out_free; 1707eb8052e0SWenliang Fan } 1708f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1709f46b5a66SChristoph Hellwig } 1710f46b5a66SChristoph Hellwig 1711ee22184bSByongho Lee if (new_size < SZ_256M) { 1712f46b5a66SChristoph Hellwig ret = -EINVAL; 1713c9e9f97bSIlya Dryomov goto out_free; 1714f46b5a66SChristoph Hellwig } 1715f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1716f46b5a66SChristoph Hellwig ret = -EFBIG; 1717c9e9f97bSIlya Dryomov goto out_free; 1718f46b5a66SChristoph Hellwig } 1719f46b5a66SChristoph Hellwig 172047f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1721f46b5a66SChristoph Hellwig 17220b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1723c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1724f46b5a66SChristoph Hellwig 1725f46b5a66SChristoph Hellwig if (new_size > old_size) { 1726a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 172798d5dc13STsutomu Itoh if (IS_ERR(trans)) { 172898d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1729c9e9f97bSIlya Dryomov goto out_free; 173098d5dc13STsutomu Itoh } 1731f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17323a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1733ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1734f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17350253f40eSjeff.liu } /* equal, nothing need to do */ 1736f46b5a66SChristoph Hellwig 1737c9e9f97bSIlya Dryomov out_free: 1738f46b5a66SChristoph Hellwig kfree(vol_args); 1739c9e9f97bSIlya Dryomov out: 1740171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 174118f39c41SIlya Dryomov mnt_drop_write_file(file); 1742f46b5a66SChristoph Hellwig return ret; 1743f46b5a66SChristoph Hellwig } 1744f46b5a66SChristoph Hellwig 174572fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 174652f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17476f72c7e2SArne Jansen u64 *transid, bool readonly, 17488696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1749f46b5a66SChristoph Hellwig { 1750f46b5a66SChristoph Hellwig int namelen; 17513de4586cSChris Mason int ret = 0; 1752f46b5a66SChristoph Hellwig 1753325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1754325c50e3SJeff Mahoney return -ENOTDIR; 1755325c50e3SJeff Mahoney 1756a874a63eSLiu Bo ret = mnt_want_write_file(file); 1757a874a63eSLiu Bo if (ret) 1758a874a63eSLiu Bo goto out; 1759a874a63eSLiu Bo 176072fd032eSSage Weil namelen = strlen(name); 176172fd032eSSage Weil if (strchr(name, '/')) { 1762f46b5a66SChristoph Hellwig ret = -EINVAL; 1763a874a63eSLiu Bo goto out_drop_write; 1764f46b5a66SChristoph Hellwig } 1765f46b5a66SChristoph Hellwig 176616780cabSChris Mason if (name[0] == '.' && 176716780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 176816780cabSChris Mason ret = -EEXIST; 1769a874a63eSLiu Bo goto out_drop_write; 177016780cabSChris Mason } 177116780cabSChris Mason 17723de4586cSChris Mason if (subvol) { 177372fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17746f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1775cb8e7090SChristoph Hellwig } else { 17762903ff01SAl Viro struct fd src = fdget(fd); 17773de4586cSChris Mason struct inode *src_inode; 17782903ff01SAl Viro if (!src.file) { 17793de4586cSChris Mason ret = -EINVAL; 1780a874a63eSLiu Bo goto out_drop_write; 17813de4586cSChris Mason } 17823de4586cSChris Mason 1783496ad9aaSAl Viro src_inode = file_inode(src.file); 1784496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1785c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1786efe120a0SFrank Holton "Snapshot src from another FS"); 178723ad5b17SKusanagi Kouichi ret = -EXDEV; 1788d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1789d0242061SDavid Sterba /* 1790d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1791d0242061SDavid Sterba * are limited to own subvolumes only 1792d0242061SDavid Sterba */ 1793d0242061SDavid Sterba ret = -EPERM; 1794ecd18815SAl Viro } else { 179572fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 179672fd032eSSage Weil BTRFS_I(src_inode)->root, 17976f72c7e2SArne Jansen transid, readonly, inherit); 1798ecd18815SAl Viro } 17992903ff01SAl Viro fdput(src); 1800cb8e7090SChristoph Hellwig } 1801a874a63eSLiu Bo out_drop_write: 1802a874a63eSLiu Bo mnt_drop_write_file(file); 1803f46b5a66SChristoph Hellwig out: 180472fd032eSSage Weil return ret; 180572fd032eSSage Weil } 180672fd032eSSage Weil 180772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1808fa0d2b9bSLi Zefan void __user *arg, int subvol) 180972fd032eSSage Weil { 1810fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 181172fd032eSSage Weil int ret; 181272fd032eSSage Weil 1813325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1814325c50e3SJeff Mahoney return -ENOTDIR; 1815325c50e3SJeff Mahoney 1816fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1817fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1818fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1819fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1820fa0d2b9bSLi Zefan 1821fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1822b83cc969SLi Zefan vol_args->fd, subvol, 18236f72c7e2SArne Jansen NULL, false, NULL); 1824fa0d2b9bSLi Zefan 1825fa0d2b9bSLi Zefan kfree(vol_args); 1826fa0d2b9bSLi Zefan return ret; 1827fa0d2b9bSLi Zefan } 1828fa0d2b9bSLi Zefan 1829fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1830fa0d2b9bSLi Zefan void __user *arg, int subvol) 1831fa0d2b9bSLi Zefan { 1832fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1833fa0d2b9bSLi Zefan int ret; 1834fdfb1e4fSLi Zefan u64 transid = 0; 1835fdfb1e4fSLi Zefan u64 *ptr = NULL; 1836b83cc969SLi Zefan bool readonly = false; 18376f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 183872fd032eSSage Weil 1839325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1840325c50e3SJeff Mahoney return -ENOTDIR; 1841325c50e3SJeff Mahoney 1842fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1843fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1844fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1845fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1846fdfb1e4fSLi Zefan 1847b83cc969SLi Zefan if (vol_args->flags & 18486f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18496f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1850b83cc969SLi Zefan ret = -EOPNOTSUPP; 1851c47ca32dSDan Carpenter goto free_args; 1852fdfb1e4fSLi Zefan } 1853fdfb1e4fSLi Zefan 1854fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1855fdfb1e4fSLi Zefan ptr = &transid; 1856b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1857b83cc969SLi Zefan readonly = true; 18586f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 185909cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18606f72c7e2SArne Jansen ret = -EINVAL; 1861c47ca32dSDan Carpenter goto free_args; 18626f72c7e2SArne Jansen } 18636f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18646f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18656f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1866c47ca32dSDan Carpenter goto free_args; 18676f72c7e2SArne Jansen } 18686f72c7e2SArne Jansen } 186975eaa0e2SSage Weil 1870fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18716f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18728696c533SMiao Xie readonly, inherit); 1873c47ca32dSDan Carpenter if (ret) 1874c47ca32dSDan Carpenter goto free_inherit; 187575eaa0e2SSage Weil 1876c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1877fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1878c47ca32dSDan Carpenter transid), 1879c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 188075eaa0e2SSage Weil ret = -EFAULT; 1881c47ca32dSDan Carpenter 1882c47ca32dSDan Carpenter free_inherit: 18836f72c7e2SArne Jansen kfree(inherit); 1884c47ca32dSDan Carpenter free_args: 1885c47ca32dSDan Carpenter kfree(vol_args); 1886f46b5a66SChristoph Hellwig return ret; 1887f46b5a66SChristoph Hellwig } 1888f46b5a66SChristoph Hellwig 18890caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18900caa102dSLi Zefan void __user *arg) 18910caa102dSLi Zefan { 1892496ad9aaSAl Viro struct inode *inode = file_inode(file); 18930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18940caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18950caa102dSLi Zefan int ret = 0; 18960caa102dSLi Zefan u64 flags = 0; 18970caa102dSLi Zefan 18984a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18990caa102dSLi Zefan return -EINVAL; 19000caa102dSLi Zefan 19010b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 19020caa102dSLi Zefan if (btrfs_root_readonly(root)) 19030caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 19040b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 19050caa102dSLi Zefan 19060caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 19070caa102dSLi Zefan ret = -EFAULT; 19080caa102dSLi Zefan 19090caa102dSLi Zefan return ret; 19100caa102dSLi Zefan } 19110caa102dSLi Zefan 19120caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19130caa102dSLi Zefan void __user *arg) 19140caa102dSLi Zefan { 1915496ad9aaSAl Viro struct inode *inode = file_inode(file); 19160b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19170caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19180caa102dSLi Zefan struct btrfs_trans_handle *trans; 19190caa102dSLi Zefan u64 root_flags; 19200caa102dSLi Zefan u64 flags; 19210caa102dSLi Zefan int ret = 0; 19220caa102dSLi Zefan 1923bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1924bd60ea0fSDavid Sterba return -EPERM; 1925bd60ea0fSDavid Sterba 1926b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1927b9ca0664SLiu Bo if (ret) 1928b9ca0664SLiu Bo goto out; 19290caa102dSLi Zefan 19304a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1931b9ca0664SLiu Bo ret = -EINVAL; 1932b9ca0664SLiu Bo goto out_drop_write; 1933b9ca0664SLiu Bo } 19340caa102dSLi Zefan 1935b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1936b9ca0664SLiu Bo ret = -EFAULT; 1937b9ca0664SLiu Bo goto out_drop_write; 1938b9ca0664SLiu Bo } 19390caa102dSLi Zefan 1940b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1941b9ca0664SLiu Bo ret = -EINVAL; 1942b9ca0664SLiu Bo goto out_drop_write; 1943b9ca0664SLiu Bo } 19440caa102dSLi Zefan 1945b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1946b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1947b9ca0664SLiu Bo goto out_drop_write; 1948b9ca0664SLiu Bo } 19490caa102dSLi Zefan 19500b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19510caa102dSLi Zefan 19520caa102dSLi Zefan /* nothing to do */ 19530caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1954b9ca0664SLiu Bo goto out_drop_sem; 19550caa102dSLi Zefan 19560caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19572c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19580caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19590caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19602c686537SDavid Sterba } else { 19612c686537SDavid Sterba /* 19622c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19632c686537SDavid Sterba * send 19642c686537SDavid Sterba */ 19652c686537SDavid Sterba spin_lock(&root->root_item_lock); 19662c686537SDavid Sterba if (root->send_in_progress == 0) { 19670caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19680caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19692c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19702c686537SDavid Sterba } else { 19712c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19720b246afaSJeff Mahoney btrfs_warn(fs_info, 19732c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19742c686537SDavid Sterba root->root_key.objectid); 19752c686537SDavid Sterba ret = -EPERM; 19762c686537SDavid Sterba goto out_drop_sem; 19772c686537SDavid Sterba } 19782c686537SDavid Sterba } 19790caa102dSLi Zefan 19800caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19810caa102dSLi Zefan if (IS_ERR(trans)) { 19820caa102dSLi Zefan ret = PTR_ERR(trans); 19830caa102dSLi Zefan goto out_reset; 19840caa102dSLi Zefan } 19850caa102dSLi Zefan 19860b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19870caa102dSLi Zefan &root->root_key, &root->root_item); 19889417ebc8SNikolay Borisov if (ret < 0) { 19899417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19909417ebc8SNikolay Borisov goto out_reset; 19919417ebc8SNikolay Borisov } 19920caa102dSLi Zefan 19939417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19949417ebc8SNikolay Borisov 19950caa102dSLi Zefan out_reset: 19960caa102dSLi Zefan if (ret) 19970caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1998b9ca0664SLiu Bo out_drop_sem: 19990b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 2000b9ca0664SLiu Bo out_drop_write: 2001b9ca0664SLiu Bo mnt_drop_write_file(file); 2002b9ca0664SLiu Bo out: 20030caa102dSLi Zefan return ret; 20040caa102dSLi Zefan } 20050caa102dSLi Zefan 2006ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2007ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2008ac8e9819SChris Mason { 2009abc6e134SChris Mason struct btrfs_key test; 2010abc6e134SChris Mason int ret; 2011abc6e134SChris Mason 2012abc6e134SChris Mason test.objectid = sk->min_objectid; 2013abc6e134SChris Mason test.type = sk->min_type; 2014abc6e134SChris Mason test.offset = sk->min_offset; 2015abc6e134SChris Mason 2016abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2017abc6e134SChris Mason if (ret < 0) 2018ac8e9819SChris Mason return 0; 2019abc6e134SChris Mason 2020abc6e134SChris Mason test.objectid = sk->max_objectid; 2021abc6e134SChris Mason test.type = sk->max_type; 2022abc6e134SChris Mason test.offset = sk->max_offset; 2023abc6e134SChris Mason 2024abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2025abc6e134SChris Mason if (ret > 0) 2026ac8e9819SChris Mason return 0; 2027ac8e9819SChris Mason return 1; 2028ac8e9819SChris Mason } 2029ac8e9819SChris Mason 2030df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2031ac8e9819SChris Mason struct btrfs_key *key, 2032ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20339b6e817dSGerhard Heift size_t *buf_size, 2034ba346b35SGerhard Heift char __user *ubuf, 2035ac8e9819SChris Mason unsigned long *sk_offset, 2036ac8e9819SChris Mason int *num_found) 2037ac8e9819SChris Mason { 2038ac8e9819SChris Mason u64 found_transid; 2039ac8e9819SChris Mason struct extent_buffer *leaf; 2040ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2041dd81d459SNaohiro Aota struct btrfs_key test; 2042ac8e9819SChris Mason unsigned long item_off; 2043ac8e9819SChris Mason unsigned long item_len; 2044ac8e9819SChris Mason int nritems; 2045ac8e9819SChris Mason int i; 2046ac8e9819SChris Mason int slot; 2047ac8e9819SChris Mason int ret = 0; 2048ac8e9819SChris Mason 2049ac8e9819SChris Mason leaf = path->nodes[0]; 2050ac8e9819SChris Mason slot = path->slots[0]; 2051ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2052ac8e9819SChris Mason 2053ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2054ac8e9819SChris Mason i = nritems; 2055ac8e9819SChris Mason goto advance_key; 2056ac8e9819SChris Mason } 2057ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2058ac8e9819SChris Mason 2059ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2060ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2061ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2062ac8e9819SChris Mason 206303b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 206403b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 206503b71c6cSGabriel de Perthuis continue; 206603b71c6cSGabriel de Perthuis 20679b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20688f5f6178SGerhard Heift if (*num_found) { 2069ac8e9819SChris Mason ret = 1; 20708f5f6178SGerhard Heift goto out; 20718f5f6178SGerhard Heift } 20728f5f6178SGerhard Heift 20738f5f6178SGerhard Heift /* 20748f5f6178SGerhard Heift * return one empty item back for v1, which does not 20758f5f6178SGerhard Heift * handle -EOVERFLOW 20768f5f6178SGerhard Heift */ 20778f5f6178SGerhard Heift 20789b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2079ac8e9819SChris Mason item_len = 0; 20808f5f6178SGerhard Heift ret = -EOVERFLOW; 20818f5f6178SGerhard Heift } 2082ac8e9819SChris Mason 20839b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2084ac8e9819SChris Mason ret = 1; 208525c9bc2eSGerhard Heift goto out; 2086ac8e9819SChris Mason } 2087ac8e9819SChris Mason 2088ac8e9819SChris Mason sh.objectid = key->objectid; 2089ac8e9819SChris Mason sh.offset = key->offset; 2090ac8e9819SChris Mason sh.type = key->type; 2091ac8e9819SChris Mason sh.len = item_len; 2092ac8e9819SChris Mason sh.transid = found_transid; 2093ac8e9819SChris Mason 2094ac8e9819SChris Mason /* copy search result header */ 2095ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2096ba346b35SGerhard Heift ret = -EFAULT; 2097ba346b35SGerhard Heift goto out; 2098ba346b35SGerhard Heift } 2099ba346b35SGerhard Heift 2100ac8e9819SChris Mason *sk_offset += sizeof(sh); 2101ac8e9819SChris Mason 2102ac8e9819SChris Mason if (item_len) { 2103ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2104ac8e9819SChris Mason /* copy the item */ 2105ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2106ba346b35SGerhard Heift item_off, item_len)) { 2107ba346b35SGerhard Heift ret = -EFAULT; 2108ba346b35SGerhard Heift goto out; 2109ba346b35SGerhard Heift } 2110ba346b35SGerhard Heift 2111ac8e9819SChris Mason *sk_offset += item_len; 2112ac8e9819SChris Mason } 2113e2156867SHugo Mills (*num_found)++; 2114ac8e9819SChris Mason 21158f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21168f5f6178SGerhard Heift goto out; 21178f5f6178SGerhard Heift 211825c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 211925c9bc2eSGerhard Heift ret = 1; 212025c9bc2eSGerhard Heift goto out; 212125c9bc2eSGerhard Heift } 2122ac8e9819SChris Mason } 2123ac8e9819SChris Mason advance_key: 2124ac8e9819SChris Mason ret = 0; 2125dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2126dd81d459SNaohiro Aota test.type = sk->max_type; 2127dd81d459SNaohiro Aota test.offset = sk->max_offset; 2128dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2129dd81d459SNaohiro Aota ret = 1; 2130dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2131abc6e134SChris Mason key->offset++; 2132dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2133abc6e134SChris Mason key->offset = 0; 2134abc6e134SChris Mason key->type++; 2135dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2136abc6e134SChris Mason key->offset = 0; 2137abc6e134SChris Mason key->type = 0; 2138abc6e134SChris Mason key->objectid++; 2139abc6e134SChris Mason } else 2140abc6e134SChris Mason ret = 1; 214125c9bc2eSGerhard Heift out: 2142ba346b35SGerhard Heift /* 2143ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2144ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2145ba346b35SGerhard Heift * * all items were found 2146ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2147ba346b35SGerhard Heift * leaf 2148ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2149ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2150ba346b35SGerhard Heift */ 2151ac8e9819SChris Mason return ret; 2152ac8e9819SChris Mason } 2153ac8e9819SChris Mason 2154ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 215512544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21569b6e817dSGerhard Heift size_t *buf_size, 2157ba346b35SGerhard Heift char __user *ubuf) 2158ac8e9819SChris Mason { 21590b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2160ac8e9819SChris Mason struct btrfs_root *root; 2161ac8e9819SChris Mason struct btrfs_key key; 2162ac8e9819SChris Mason struct btrfs_path *path; 2163ac8e9819SChris Mason int ret; 2164ac8e9819SChris Mason int num_found = 0; 2165ac8e9819SChris Mason unsigned long sk_offset = 0; 2166ac8e9819SChris Mason 21679b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21689b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 216912544442SGerhard Heift return -EOVERFLOW; 21709b6e817dSGerhard Heift } 217112544442SGerhard Heift 2172ac8e9819SChris Mason path = btrfs_alloc_path(); 2173ac8e9819SChris Mason if (!path) 2174ac8e9819SChris Mason return -ENOMEM; 2175ac8e9819SChris Mason 2176ac8e9819SChris Mason if (sk->tree_id == 0) { 2177ac8e9819SChris Mason /* search the root of the inode that was passed */ 2178ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2179ac8e9819SChris Mason } else { 2180ac8e9819SChris Mason key.objectid = sk->tree_id; 2181ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2182ac8e9819SChris Mason key.offset = (u64)-1; 2183ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2184ac8e9819SChris Mason if (IS_ERR(root)) { 2185ac8e9819SChris Mason btrfs_free_path(path); 2186ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2187ac8e9819SChris Mason } 2188ac8e9819SChris Mason } 2189ac8e9819SChris Mason 2190ac8e9819SChris Mason key.objectid = sk->min_objectid; 2191ac8e9819SChris Mason key.type = sk->min_type; 2192ac8e9819SChris Mason key.offset = sk->min_offset; 2193ac8e9819SChris Mason 2194ac8e9819SChris Mason while (1) { 21956174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2196ac8e9819SChris Mason if (ret != 0) { 2197ac8e9819SChris Mason if (ret > 0) 2198ac8e9819SChris Mason ret = 0; 2199ac8e9819SChris Mason goto err; 2200ac8e9819SChris Mason } 2201df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2202ac8e9819SChris Mason &sk_offset, &num_found); 2203b3b4aa74SDavid Sterba btrfs_release_path(path); 220425c9bc2eSGerhard Heift if (ret) 2205ac8e9819SChris Mason break; 2206ac8e9819SChris Mason 2207ac8e9819SChris Mason } 22088f5f6178SGerhard Heift if (ret > 0) 2209ac8e9819SChris Mason ret = 0; 2210ac8e9819SChris Mason err: 2211ac8e9819SChris Mason sk->nr_items = num_found; 2212ac8e9819SChris Mason btrfs_free_path(path); 2213ac8e9819SChris Mason return ret; 2214ac8e9819SChris Mason } 2215ac8e9819SChris Mason 2216ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2217ac8e9819SChris Mason void __user *argp) 2218ac8e9819SChris Mason { 2219ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2220ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2221ac8e9819SChris Mason struct inode *inode; 2222ac8e9819SChris Mason int ret; 22239b6e817dSGerhard Heift size_t buf_size; 2224ac8e9819SChris Mason 2225ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2226ac8e9819SChris Mason return -EPERM; 2227ac8e9819SChris Mason 2228ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2229ac8e9819SChris Mason 2230ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2231ba346b35SGerhard Heift return -EFAULT; 2232ba346b35SGerhard Heift 2233ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2234ac8e9819SChris Mason 2235496ad9aaSAl Viro inode = file_inode(file); 2236ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22378f5f6178SGerhard Heift 22388f5f6178SGerhard Heift /* 22398f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22408f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22418f5f6178SGerhard Heift */ 22428f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22438f5f6178SGerhard Heift ret = 0; 22448f5f6178SGerhard Heift 2245ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2246ac8e9819SChris Mason ret = -EFAULT; 2247ac8e9819SChris Mason return ret; 2248ac8e9819SChris Mason } 2249ac8e9819SChris Mason 2250cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2251cc68a8a5SGerhard Heift void __user *argp) 2252cc68a8a5SGerhard Heift { 2253cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2254cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2255cc68a8a5SGerhard Heift struct inode *inode; 2256cc68a8a5SGerhard Heift int ret; 2257cc68a8a5SGerhard Heift size_t buf_size; 2258ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2259cc68a8a5SGerhard Heift 2260cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2261cc68a8a5SGerhard Heift return -EPERM; 2262cc68a8a5SGerhard Heift 2263cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2264cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2265cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2266cc68a8a5SGerhard Heift return -EFAULT; 2267cc68a8a5SGerhard Heift 2268cc68a8a5SGerhard Heift buf_size = args.buf_size; 2269cc68a8a5SGerhard Heift 2270cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2271cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2272cc68a8a5SGerhard Heift buf_size = buf_limit; 2273cc68a8a5SGerhard Heift 2274cc68a8a5SGerhard Heift inode = file_inode(file); 2275cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2276718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2277cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2278cc68a8a5SGerhard Heift ret = -EFAULT; 2279cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2280cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2281cc68a8a5SGerhard Heift ret = -EFAULT; 2282cc68a8a5SGerhard Heift 228376dda93cSYan, Zheng return ret; 228476dda93cSYan, Zheng } 228576dda93cSYan, Zheng 228698d377a0STARUISI Hiroaki /* 2287ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2288ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 228998d377a0STARUISI Hiroaki */ 229098d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 229198d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 229298d377a0STARUISI Hiroaki { 229398d377a0STARUISI Hiroaki struct btrfs_root *root; 229498d377a0STARUISI Hiroaki struct btrfs_key key; 2295ac8e9819SChris Mason char *ptr; 229698d377a0STARUISI Hiroaki int ret = -1; 229798d377a0STARUISI Hiroaki int slot; 229898d377a0STARUISI Hiroaki int len; 229998d377a0STARUISI Hiroaki int total_len = 0; 230098d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 230198d377a0STARUISI Hiroaki struct extent_buffer *l; 230298d377a0STARUISI Hiroaki struct btrfs_path *path; 230398d377a0STARUISI Hiroaki 230498d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 230598d377a0STARUISI Hiroaki name[0]='\0'; 230698d377a0STARUISI Hiroaki return 0; 230798d377a0STARUISI Hiroaki } 230898d377a0STARUISI Hiroaki 230998d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 231098d377a0STARUISI Hiroaki if (!path) 231198d377a0STARUISI Hiroaki return -ENOMEM; 231298d377a0STARUISI Hiroaki 2313c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 231498d377a0STARUISI Hiroaki 231598d377a0STARUISI Hiroaki key.objectid = tree_id; 231698d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 231798d377a0STARUISI Hiroaki key.offset = (u64)-1; 231898d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 231998d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2320ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 23218ad6fcabSChris Mason goto out; 232298d377a0STARUISI Hiroaki } 232398d377a0STARUISI Hiroaki 232498d377a0STARUISI Hiroaki key.objectid = dirid; 232598d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23268ad6fcabSChris Mason key.offset = (u64)-1; 232798d377a0STARUISI Hiroaki 232898d377a0STARUISI Hiroaki while (1) { 232998d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 233098d377a0STARUISI Hiroaki if (ret < 0) 233198d377a0STARUISI Hiroaki goto out; 233218674c6cSFilipe David Borba Manana else if (ret > 0) { 233318674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 233418674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 233518674c6cSFilipe David Borba Manana if (ret < 0) 233618674c6cSFilipe David Borba Manana goto out; 233718674c6cSFilipe David Borba Manana else if (ret > 0) { 2338ac8e9819SChris Mason ret = -ENOENT; 233998d377a0STARUISI Hiroaki goto out; 2340ac8e9819SChris Mason } 234118674c6cSFilipe David Borba Manana } 234218674c6cSFilipe David Borba Manana 234318674c6cSFilipe David Borba Manana l = path->nodes[0]; 234418674c6cSFilipe David Borba Manana slot = path->slots[0]; 234518674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 234698d377a0STARUISI Hiroaki 234798d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 234898d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 234998d377a0STARUISI Hiroaki ptr -= len + 1; 235098d377a0STARUISI Hiroaki total_len += len + 1; 2351a696cf35SFilipe David Borba Manana if (ptr < name) { 2352a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 235398d377a0STARUISI Hiroaki goto out; 2354a696cf35SFilipe David Borba Manana } 235598d377a0STARUISI Hiroaki 235698d377a0STARUISI Hiroaki *(ptr + len) = '/'; 235798d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 235898d377a0STARUISI Hiroaki 235998d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 236098d377a0STARUISI Hiroaki break; 236198d377a0STARUISI Hiroaki 2362b3b4aa74SDavid Sterba btrfs_release_path(path); 236398d377a0STARUISI Hiroaki key.objectid = key.offset; 23648ad6fcabSChris Mason key.offset = (u64)-1; 236598d377a0STARUISI Hiroaki dirid = key.objectid; 236698d377a0STARUISI Hiroaki } 236777906a50SLi Zefan memmove(name, ptr, total_len); 236898d377a0STARUISI Hiroaki name[total_len] = '\0'; 236998d377a0STARUISI Hiroaki ret = 0; 237098d377a0STARUISI Hiroaki out: 237198d377a0STARUISI Hiroaki btrfs_free_path(path); 2372ac8e9819SChris Mason return ret; 2373ac8e9819SChris Mason } 2374ac8e9819SChris Mason 237523d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 237623d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 237723d0b79dSTomohiro Misono { 237823d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 237923d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 238023d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 238123d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 238223d0b79dSTomohiro Misono u64 dirid = args->dirid; 238323d0b79dSTomohiro Misono unsigned long item_off; 238423d0b79dSTomohiro Misono unsigned long item_len; 238523d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 238623d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 238723d0b79dSTomohiro Misono struct btrfs_root *root; 238823d0b79dSTomohiro Misono struct btrfs_path *path; 238923d0b79dSTomohiro Misono struct btrfs_key key, key2; 239023d0b79dSTomohiro Misono struct extent_buffer *leaf; 239123d0b79dSTomohiro Misono struct inode *temp_inode; 239223d0b79dSTomohiro Misono char *ptr; 239323d0b79dSTomohiro Misono int slot; 239423d0b79dSTomohiro Misono int len; 239523d0b79dSTomohiro Misono int total_len = 0; 239623d0b79dSTomohiro Misono int ret; 239723d0b79dSTomohiro Misono 239823d0b79dSTomohiro Misono path = btrfs_alloc_path(); 239923d0b79dSTomohiro Misono if (!path) 240023d0b79dSTomohiro Misono return -ENOMEM; 240123d0b79dSTomohiro Misono 240223d0b79dSTomohiro Misono /* 240323d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 240423d0b79dSTomohiro Misono * construct the path in from the bottom up. 240523d0b79dSTomohiro Misono */ 240623d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 240723d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 240823d0b79dSTomohiro Misono 240923d0b79dSTomohiro Misono key.objectid = treeid; 241023d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 241123d0b79dSTomohiro Misono key.offset = (u64)-1; 241223d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 241323d0b79dSTomohiro Misono if (IS_ERR(root)) { 241423d0b79dSTomohiro Misono ret = PTR_ERR(root); 241523d0b79dSTomohiro Misono goto out; 241623d0b79dSTomohiro Misono } 241723d0b79dSTomohiro Misono 241823d0b79dSTomohiro Misono key.objectid = dirid; 241923d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 242023d0b79dSTomohiro Misono key.offset = (u64)-1; 242123d0b79dSTomohiro Misono while (1) { 242223d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 242323d0b79dSTomohiro Misono if (ret < 0) { 242423d0b79dSTomohiro Misono goto out; 242523d0b79dSTomohiro Misono } else if (ret > 0) { 242623d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 242723d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 242823d0b79dSTomohiro Misono if (ret < 0) { 242923d0b79dSTomohiro Misono goto out; 243023d0b79dSTomohiro Misono } else if (ret > 0) { 243123d0b79dSTomohiro Misono ret = -ENOENT; 243223d0b79dSTomohiro Misono goto out; 243323d0b79dSTomohiro Misono } 243423d0b79dSTomohiro Misono } 243523d0b79dSTomohiro Misono 243623d0b79dSTomohiro Misono leaf = path->nodes[0]; 243723d0b79dSTomohiro Misono slot = path->slots[0]; 243823d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 243923d0b79dSTomohiro Misono 244023d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 244123d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 244223d0b79dSTomohiro Misono ptr -= len + 1; 244323d0b79dSTomohiro Misono total_len += len + 1; 244423d0b79dSTomohiro Misono if (ptr < args->path) { 244523d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 244623d0b79dSTomohiro Misono goto out; 244723d0b79dSTomohiro Misono } 244823d0b79dSTomohiro Misono 244923d0b79dSTomohiro Misono *(ptr + len) = '/'; 245023d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 245123d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 245223d0b79dSTomohiro Misono 245323d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 245423d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 245523d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 245623d0b79dSTomohiro Misono if (ret < 0) { 245723d0b79dSTomohiro Misono goto out; 245823d0b79dSTomohiro Misono } else if (ret > 0) { 245923d0b79dSTomohiro Misono ret = -ENOENT; 246023d0b79dSTomohiro Misono goto out; 246123d0b79dSTomohiro Misono } 246223d0b79dSTomohiro Misono 246323d0b79dSTomohiro Misono leaf = path->nodes[0]; 246423d0b79dSTomohiro Misono slot = path->slots[0]; 246523d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 246623d0b79dSTomohiro Misono if (key2.objectid != dirid) { 246723d0b79dSTomohiro Misono ret = -ENOENT; 246823d0b79dSTomohiro Misono goto out; 246923d0b79dSTomohiro Misono } 247023d0b79dSTomohiro Misono 247123d0b79dSTomohiro Misono temp_inode = btrfs_iget(sb, &key2, root, NULL); 24723ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24733ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24743ca57bd6SMisono Tomohiro goto out; 24753ca57bd6SMisono Tomohiro } 247623d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 247723d0b79dSTomohiro Misono iput(temp_inode); 247823d0b79dSTomohiro Misono if (ret) { 247923d0b79dSTomohiro Misono ret = -EACCES; 248023d0b79dSTomohiro Misono goto out; 248123d0b79dSTomohiro Misono } 248223d0b79dSTomohiro Misono 248323d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 248423d0b79dSTomohiro Misono break; 248523d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 248623d0b79dSTomohiro Misono ret = -EACCES; 248723d0b79dSTomohiro Misono goto out; 248823d0b79dSTomohiro Misono } 248923d0b79dSTomohiro Misono 249023d0b79dSTomohiro Misono btrfs_release_path(path); 249123d0b79dSTomohiro Misono key.objectid = key.offset; 249223d0b79dSTomohiro Misono key.offset = (u64)-1; 249323d0b79dSTomohiro Misono dirid = key.objectid; 249423d0b79dSTomohiro Misono } 249523d0b79dSTomohiro Misono 249623d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 249723d0b79dSTomohiro Misono args->path[total_len] = '\0'; 249823d0b79dSTomohiro Misono btrfs_release_path(path); 249923d0b79dSTomohiro Misono } 250023d0b79dSTomohiro Misono 250123d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 250223d0b79dSTomohiro Misono root = fs_info->tree_root; 250323d0b79dSTomohiro Misono key.objectid = treeid; 250423d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 250523d0b79dSTomohiro Misono key.offset = args->treeid; 250623d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 250723d0b79dSTomohiro Misono if (ret < 0) { 250823d0b79dSTomohiro Misono goto out; 250923d0b79dSTomohiro Misono } else if (ret > 0) { 251023d0b79dSTomohiro Misono ret = -ENOENT; 251123d0b79dSTomohiro Misono goto out; 251223d0b79dSTomohiro Misono } 251323d0b79dSTomohiro Misono 251423d0b79dSTomohiro Misono leaf = path->nodes[0]; 251523d0b79dSTomohiro Misono slot = path->slots[0]; 251623d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 251723d0b79dSTomohiro Misono 251823d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 251923d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 252023d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 252123d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 252223d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 252323d0b79dSTomohiro Misono ret = -EINVAL; 252423d0b79dSTomohiro Misono goto out; 252523d0b79dSTomohiro Misono } 252623d0b79dSTomohiro Misono 252723d0b79dSTomohiro Misono /* Copy subvolume's name */ 252823d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 252923d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 253023d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 253123d0b79dSTomohiro Misono args->name[item_len] = 0; 253223d0b79dSTomohiro Misono 253323d0b79dSTomohiro Misono out: 253423d0b79dSTomohiro Misono btrfs_free_path(path); 253523d0b79dSTomohiro Misono return ret; 253623d0b79dSTomohiro Misono } 253723d0b79dSTomohiro Misono 2538ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2539ac8e9819SChris Mason void __user *argp) 2540ac8e9819SChris Mason { 2541ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2542ac8e9819SChris Mason struct inode *inode; 254301b810b8SDavid Sterba int ret = 0; 2544ac8e9819SChris Mason 25452354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25462354d08fSJulia Lawall if (IS_ERR(args)) 25472354d08fSJulia Lawall return PTR_ERR(args); 2548c2b96929SDan Carpenter 2549496ad9aaSAl Viro inode = file_inode(file); 2550ac8e9819SChris Mason 255101b810b8SDavid Sterba /* 255201b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 255301b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 255401b810b8SDavid Sterba */ 25551b53ac4dSChris Mason if (args->treeid == 0) 25561b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25571b53ac4dSChris Mason 255801b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 255901b810b8SDavid Sterba args->name[0] = 0; 256001b810b8SDavid Sterba goto out; 256101b810b8SDavid Sterba } 256201b810b8SDavid Sterba 256301b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 256401b810b8SDavid Sterba ret = -EPERM; 256501b810b8SDavid Sterba goto out; 256601b810b8SDavid Sterba } 256701b810b8SDavid Sterba 2568ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2569ac8e9819SChris Mason args->treeid, args->objectid, 2570ac8e9819SChris Mason args->name); 2571ac8e9819SChris Mason 257201b810b8SDavid Sterba out: 2573ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2574ac8e9819SChris Mason ret = -EFAULT; 2575ac8e9819SChris Mason 2576ac8e9819SChris Mason kfree(args); 257798d377a0STARUISI Hiroaki return ret; 257898d377a0STARUISI Hiroaki } 257998d377a0STARUISI Hiroaki 258023d0b79dSTomohiro Misono /* 258123d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 258223d0b79dSTomohiro Misono * 258323d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 258423d0b79dSTomohiro Misono * 258523d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 258623d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 258723d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 258823d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 258923d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 259023d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 259123d0b79dSTomohiro Misono */ 259223d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 259323d0b79dSTomohiro Misono { 259423d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 259523d0b79dSTomohiro Misono struct inode *inode; 259623d0b79dSTomohiro Misono int ret; 259723d0b79dSTomohiro Misono 259823d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 259923d0b79dSTomohiro Misono if (IS_ERR(args)) 260023d0b79dSTomohiro Misono return PTR_ERR(args); 260123d0b79dSTomohiro Misono 260223d0b79dSTomohiro Misono inode = file_inode(file); 260323d0b79dSTomohiro Misono 260423d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 260523d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 260623d0b79dSTomohiro Misono /* 260723d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 260823d0b79dSTomohiro Misono * called 260923d0b79dSTomohiro Misono */ 261023d0b79dSTomohiro Misono kfree(args); 261123d0b79dSTomohiro Misono return -EACCES; 261223d0b79dSTomohiro Misono } 261323d0b79dSTomohiro Misono 261423d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 261523d0b79dSTomohiro Misono 261623d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 261723d0b79dSTomohiro Misono ret = -EFAULT; 261823d0b79dSTomohiro Misono 261923d0b79dSTomohiro Misono kfree(args); 262023d0b79dSTomohiro Misono return ret; 262123d0b79dSTomohiro Misono } 262223d0b79dSTomohiro Misono 2623b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2624b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2625b64ec075STomohiro Misono { 2626b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2627b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2628b64ec075STomohiro Misono struct btrfs_root *root; 2629b64ec075STomohiro Misono struct btrfs_path *path; 2630b64ec075STomohiro Misono struct btrfs_key key; 2631b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2632b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2633b64ec075STomohiro Misono struct extent_buffer *leaf; 2634b64ec075STomohiro Misono unsigned long item_off; 2635b64ec075STomohiro Misono unsigned long item_len; 2636b64ec075STomohiro Misono struct inode *inode; 2637b64ec075STomohiro Misono int slot; 2638b64ec075STomohiro Misono int ret = 0; 2639b64ec075STomohiro Misono 2640b64ec075STomohiro Misono path = btrfs_alloc_path(); 2641b64ec075STomohiro Misono if (!path) 2642b64ec075STomohiro Misono return -ENOMEM; 2643b64ec075STomohiro Misono 2644b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2645b64ec075STomohiro Misono if (!subvol_info) { 2646b64ec075STomohiro Misono btrfs_free_path(path); 2647b64ec075STomohiro Misono return -ENOMEM; 2648b64ec075STomohiro Misono } 2649b64ec075STomohiro Misono 2650b64ec075STomohiro Misono inode = file_inode(file); 2651b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2652b64ec075STomohiro Misono 2653b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2654b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2655b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2656b64ec075STomohiro Misono key.offset = (u64)-1; 2657b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2658b64ec075STomohiro Misono if (IS_ERR(root)) { 2659b64ec075STomohiro Misono ret = PTR_ERR(root); 2660b64ec075STomohiro Misono goto out; 2661b64ec075STomohiro Misono } 2662b64ec075STomohiro Misono root_item = &root->root_item; 2663b64ec075STomohiro Misono 2664b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2665b64ec075STomohiro Misono 2666b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2667b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2668b64ec075STomohiro Misono 2669b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2670b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2671b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2672b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2673b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2674b64ec075STomohiro Misono 2675b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2676b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2677b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2678b64ec075STomohiro Misono 2679b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2680b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2681b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2682b64ec075STomohiro Misono 2683b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2684b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2685b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2686b64ec075STomohiro Misono 2687b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2688b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2689b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2690b64ec075STomohiro Misono 2691b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2692b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2693b64ec075STomohiro Misono root = fs_info->tree_root; 2694b64ec075STomohiro Misono 2695b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2696b64ec075STomohiro Misono key.offset = 0; 2697b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2698b64ec075STomohiro Misono if (ret < 0) { 2699b64ec075STomohiro Misono goto out; 2700b64ec075STomohiro Misono } else if (path->slots[0] >= 2701b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2702b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2703b64ec075STomohiro Misono if (ret < 0) { 2704b64ec075STomohiro Misono goto out; 2705b64ec075STomohiro Misono } else if (ret > 0) { 2706b64ec075STomohiro Misono ret = -EUCLEAN; 2707b64ec075STomohiro Misono goto out; 2708b64ec075STomohiro Misono } 2709b64ec075STomohiro Misono } 2710b64ec075STomohiro Misono 2711b64ec075STomohiro Misono leaf = path->nodes[0]; 2712b64ec075STomohiro Misono slot = path->slots[0]; 2713b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2714b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2715b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2716b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2717b64ec075STomohiro Misono 2718b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2719b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2720b64ec075STomohiro Misono 2721b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2722b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2723b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2724b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2725b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2726b64ec075STomohiro Misono item_off, item_len); 2727b64ec075STomohiro Misono } else { 2728b64ec075STomohiro Misono ret = -ENOENT; 2729b64ec075STomohiro Misono goto out; 2730b64ec075STomohiro Misono } 2731b64ec075STomohiro Misono } 2732b64ec075STomohiro Misono 2733b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2734b64ec075STomohiro Misono ret = -EFAULT; 2735b64ec075STomohiro Misono 2736b64ec075STomohiro Misono out: 2737b64ec075STomohiro Misono btrfs_free_path(path); 2738b64ec075STomohiro Misono kzfree(subvol_info); 2739b64ec075STomohiro Misono return ret; 2740b64ec075STomohiro Misono } 2741b64ec075STomohiro Misono 274242e4b520STomohiro Misono /* 274342e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 274442e4b520STomohiro Misono * except the subvolume name. 274542e4b520STomohiro Misono */ 274642e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 274742e4b520STomohiro Misono { 274842e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 274942e4b520STomohiro Misono struct btrfs_root_ref *rref; 275042e4b520STomohiro Misono struct btrfs_root *root; 275142e4b520STomohiro Misono struct btrfs_path *path; 275242e4b520STomohiro Misono struct btrfs_key key; 275342e4b520STomohiro Misono struct extent_buffer *leaf; 275442e4b520STomohiro Misono struct inode *inode; 275542e4b520STomohiro Misono u64 objectid; 275642e4b520STomohiro Misono int slot; 275742e4b520STomohiro Misono int ret; 275842e4b520STomohiro Misono u8 found; 275942e4b520STomohiro Misono 276042e4b520STomohiro Misono path = btrfs_alloc_path(); 276142e4b520STomohiro Misono if (!path) 276242e4b520STomohiro Misono return -ENOMEM; 276342e4b520STomohiro Misono 276442e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 276542e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 276642e4b520STomohiro Misono btrfs_free_path(path); 276742e4b520STomohiro Misono return PTR_ERR(rootrefs); 276842e4b520STomohiro Misono } 276942e4b520STomohiro Misono 277042e4b520STomohiro Misono inode = file_inode(file); 277142e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 277242e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 277342e4b520STomohiro Misono 277442e4b520STomohiro Misono key.objectid = objectid; 277542e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 277642e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 277742e4b520STomohiro Misono found = 0; 277842e4b520STomohiro Misono 277942e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 278042e4b520STomohiro Misono if (ret < 0) { 278142e4b520STomohiro Misono goto out; 278242e4b520STomohiro Misono } else if (path->slots[0] >= 278342e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 278442e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 278542e4b520STomohiro Misono if (ret < 0) { 278642e4b520STomohiro Misono goto out; 278742e4b520STomohiro Misono } else if (ret > 0) { 278842e4b520STomohiro Misono ret = -EUCLEAN; 278942e4b520STomohiro Misono goto out; 279042e4b520STomohiro Misono } 279142e4b520STomohiro Misono } 279242e4b520STomohiro Misono while (1) { 279342e4b520STomohiro Misono leaf = path->nodes[0]; 279442e4b520STomohiro Misono slot = path->slots[0]; 279542e4b520STomohiro Misono 279642e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 279742e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 279842e4b520STomohiro Misono ret = 0; 279942e4b520STomohiro Misono goto out; 280042e4b520STomohiro Misono } 280142e4b520STomohiro Misono 280242e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 280342e4b520STomohiro Misono ret = -EOVERFLOW; 280442e4b520STomohiro Misono goto out; 280542e4b520STomohiro Misono } 280642e4b520STomohiro Misono 280742e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 280842e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 280942e4b520STomohiro Misono rootrefs->rootref[found].dirid = 281042e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 281142e4b520STomohiro Misono found++; 281242e4b520STomohiro Misono 281342e4b520STomohiro Misono ret = btrfs_next_item(root, path); 281442e4b520STomohiro Misono if (ret < 0) { 281542e4b520STomohiro Misono goto out; 281642e4b520STomohiro Misono } else if (ret > 0) { 281742e4b520STomohiro Misono ret = -EUCLEAN; 281842e4b520STomohiro Misono goto out; 281942e4b520STomohiro Misono } 282042e4b520STomohiro Misono } 282142e4b520STomohiro Misono 282242e4b520STomohiro Misono out: 282342e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 282442e4b520STomohiro Misono rootrefs->num_items = found; 282542e4b520STomohiro Misono /* update min_treeid for next search */ 282642e4b520STomohiro Misono if (found) 282742e4b520STomohiro Misono rootrefs->min_treeid = 282842e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 282942e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 283042e4b520STomohiro Misono ret = -EFAULT; 283142e4b520STomohiro Misono } 283242e4b520STomohiro Misono 283342e4b520STomohiro Misono kfree(rootrefs); 283442e4b520STomohiro Misono btrfs_free_path(path); 283542e4b520STomohiro Misono 283642e4b520STomohiro Misono return ret; 283742e4b520STomohiro Misono } 283842e4b520STomohiro Misono 283976dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 284076dda93cSYan, Zheng void __user *arg) 284176dda93cSYan, Zheng { 284254563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28430b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 284476dda93cSYan, Zheng struct dentry *dentry; 28452b0143b5SDavid Howells struct inode *dir = d_inode(parent); 284676dda93cSYan, Zheng struct inode *inode; 284776dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 284876dda93cSYan, Zheng struct btrfs_root *dest = NULL; 284976dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 285076dda93cSYan, Zheng int namelen; 285176dda93cSYan, Zheng int err = 0; 285276dda93cSYan, Zheng 2853325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2854325c50e3SJeff Mahoney return -ENOTDIR; 2855325c50e3SJeff Mahoney 285676dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 285776dda93cSYan, Zheng if (IS_ERR(vol_args)) 285876dda93cSYan, Zheng return PTR_ERR(vol_args); 285976dda93cSYan, Zheng 286076dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 286176dda93cSYan, Zheng namelen = strlen(vol_args->name); 286276dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 286376dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 286476dda93cSYan, Zheng err = -EINVAL; 286576dda93cSYan, Zheng goto out; 286676dda93cSYan, Zheng } 286776dda93cSYan, Zheng 2868a561be71SAl Viro err = mnt_want_write_file(file); 286976dda93cSYan, Zheng if (err) 287076dda93cSYan, Zheng goto out; 287176dda93cSYan, Zheng 2872521e0546SDavid Sterba 287300235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 287400235411SAl Viro if (err == -EINTR) 287500235411SAl Viro goto out_drop_write; 287676dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 287776dda93cSYan, Zheng if (IS_ERR(dentry)) { 287876dda93cSYan, Zheng err = PTR_ERR(dentry); 287976dda93cSYan, Zheng goto out_unlock_dir; 288076dda93cSYan, Zheng } 288176dda93cSYan, Zheng 28822b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 288376dda93cSYan, Zheng err = -ENOENT; 288476dda93cSYan, Zheng goto out_dput; 288576dda93cSYan, Zheng } 288676dda93cSYan, Zheng 28872b0143b5SDavid Howells inode = d_inode(dentry); 28884260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28894260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28904260f7c7SSage Weil /* 28914260f7c7SSage Weil * Regular user. Only allow this with a special mount 28924260f7c7SSage Weil * option, when the user has write+exec access to the 28934260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28944260f7c7SSage Weil * allowed. 28954260f7c7SSage Weil * 28964260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28974260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28984260f7c7SSage Weil * otherwise be able to delete. 28994260f7c7SSage Weil * 29004260f7c7SSage Weil * Users who want to delete empty subvols should try 29014260f7c7SSage Weil * rmdir(2). 29024260f7c7SSage Weil */ 29034260f7c7SSage Weil err = -EPERM; 29040b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 29054260f7c7SSage Weil goto out_dput; 29064260f7c7SSage Weil 29074260f7c7SSage Weil /* 29084260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 29094260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 29104260f7c7SSage Weil * must be called on the dentry referencing the root 29114260f7c7SSage Weil * of the subvol, not a random directory contained 29124260f7c7SSage Weil * within it. 29134260f7c7SSage Weil */ 29144260f7c7SSage Weil err = -EINVAL; 29154260f7c7SSage Weil if (root == dest) 29164260f7c7SSage Weil goto out_dput; 29174260f7c7SSage Weil 29184260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 29194260f7c7SSage Weil if (err) 29204260f7c7SSage Weil goto out_dput; 29215c39da5bSMiao Xie } 29224260f7c7SSage Weil 29235c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 29244260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 29254260f7c7SSage Weil if (err) 29264260f7c7SSage Weil goto out_dput; 29274260f7c7SSage Weil 29284a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 292976dda93cSYan, Zheng err = -EINVAL; 293076dda93cSYan, Zheng goto out_dput; 293176dda93cSYan, Zheng } 293276dda93cSYan, Zheng 29335955102cSAl Viro inode_lock(inode); 2934f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 29355955102cSAl Viro inode_unlock(inode); 2936f60a2364SMisono Tomohiro if (!err) 293776dda93cSYan, Zheng d_delete(dentry); 2938fa6ac876SLiu Bo 293976dda93cSYan, Zheng out_dput: 294076dda93cSYan, Zheng dput(dentry); 294176dda93cSYan, Zheng out_unlock_dir: 29425955102cSAl Viro inode_unlock(dir); 294300235411SAl Viro out_drop_write: 29442a79f17eSAl Viro mnt_drop_write_file(file); 294576dda93cSYan, Zheng out: 294676dda93cSYan, Zheng kfree(vol_args); 294776dda93cSYan, Zheng return err; 294876dda93cSYan, Zheng } 294976dda93cSYan, Zheng 29501e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2951f46b5a66SChristoph Hellwig { 2952496ad9aaSAl Viro struct inode *inode = file_inode(file); 2953f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29541e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2955c146afadSYan Zheng int ret; 2956c146afadSYan Zheng 295725122d15SIlya Dryomov ret = mnt_want_write_file(file); 295825122d15SIlya Dryomov if (ret) 295925122d15SIlya Dryomov return ret; 2960b83cc969SLi Zefan 296125122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 296225122d15SIlya Dryomov ret = -EROFS; 296325122d15SIlya Dryomov goto out; 29645ac00addSStefan Behrens } 2965f46b5a66SChristoph Hellwig 2966f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2967f46b5a66SChristoph Hellwig case S_IFDIR: 2968e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2969e441d54dSChris Mason ret = -EPERM; 2970e441d54dSChris Mason goto out; 2971e441d54dSChris Mason } 2972de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2973f46b5a66SChristoph Hellwig break; 2974f46b5a66SChristoph Hellwig case S_IFREG: 2975616d374eSAdam Borowski /* 2976616d374eSAdam Borowski * Note that this does not check the file descriptor for write 2977616d374eSAdam Borowski * access. This prevents defragmenting executables that are 2978616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 2979616d374eSAdam Borowski */ 2980616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 2981616d374eSAdam Borowski inode_permission(inode, MAY_WRITE)) { 2982616d374eSAdam Borowski ret = -EPERM; 2983e441d54dSChris Mason goto out; 2984e441d54dSChris Mason } 29851e701a32SChris Mason 29861e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29871e701a32SChris Mason if (!range) { 29881e701a32SChris Mason ret = -ENOMEM; 29891e701a32SChris Mason goto out; 29901e701a32SChris Mason } 29911e701a32SChris Mason 29921e701a32SChris Mason if (argp) { 29931e701a32SChris Mason if (copy_from_user(range, argp, 29941e701a32SChris Mason sizeof(*range))) { 29951e701a32SChris Mason ret = -EFAULT; 29961e701a32SChris Mason kfree(range); 2997683be16eSDan Carpenter goto out; 29981e701a32SChris Mason } 29991e701a32SChris Mason /* compression requires us to start the IO */ 30001e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 30011e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 30021e701a32SChris Mason range->extent_thresh = (u32)-1; 30031e701a32SChris Mason } 30041e701a32SChris Mason } else { 30051e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 30061e701a32SChris Mason range->len = (u64)-1; 30071e701a32SChris Mason } 3008496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 30097c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 30104cb5300bSChris Mason if (ret > 0) 30114cb5300bSChris Mason ret = 0; 30121e701a32SChris Mason kfree(range); 3013f46b5a66SChristoph Hellwig break; 30148929ecfaSYan, Zheng default: 30158929ecfaSYan, Zheng ret = -EINVAL; 3016f46b5a66SChristoph Hellwig } 3017e441d54dSChris Mason out: 301825122d15SIlya Dryomov mnt_drop_write_file(file); 3019e441d54dSChris Mason return ret; 3020f46b5a66SChristoph Hellwig } 3021f46b5a66SChristoph Hellwig 30222ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3023f46b5a66SChristoph Hellwig { 3024f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3025f46b5a66SChristoph Hellwig int ret; 3026f46b5a66SChristoph Hellwig 3027e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3028e441d54dSChris Mason return -EPERM; 3029e441d54dSChris Mason 3030171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 3031e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3032c9e9f97bSIlya Dryomov 3033dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3034c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3035c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3036c9e9f97bSIlya Dryomov goto out; 3037c9e9f97bSIlya Dryomov } 3038f46b5a66SChristoph Hellwig 30395516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30402ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3041f46b5a66SChristoph Hellwig 304243d20761SAnand Jain if (!ret) 30430b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 304443d20761SAnand Jain 3045f46b5a66SChristoph Hellwig kfree(vol_args); 3046c9e9f97bSIlya Dryomov out: 3047171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3048f46b5a66SChristoph Hellwig return ret; 3049f46b5a66SChristoph Hellwig } 3050f46b5a66SChristoph Hellwig 30516b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3052f46b5a66SChristoph Hellwig { 30530b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30556b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3056f46b5a66SChristoph Hellwig int ret; 3057f46b5a66SChristoph Hellwig 3058e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3059e441d54dSChris Mason return -EPERM; 3060e441d54dSChris Mason 3061da24927bSMiao Xie ret = mnt_want_write_file(file); 3062da24927bSMiao Xie if (ret) 3063da24927bSMiao Xie return ret; 3064c146afadSYan Zheng 3065dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3066c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3067c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3068c47ca32dSDan Carpenter goto err_drop; 3069c9e9f97bSIlya Dryomov } 3070f46b5a66SChristoph Hellwig 30716b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3072fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3073fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3074fd4e994bSOmar Sandoval goto out; 3075fd4e994bSOmar Sandoval } 3076f46b5a66SChristoph Hellwig 3077171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3078183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3079183860f6SAnand Jain goto out; 3080183860f6SAnand Jain } 3081183860f6SAnand Jain 3082735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30832ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30846b526ed7SAnand Jain } else { 30856b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30862ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30876b526ed7SAnand Jain } 3088171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3089183860f6SAnand Jain 30906b526ed7SAnand Jain if (!ret) { 3091735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30920b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30936b526ed7SAnand Jain vol_args->devid); 30946b526ed7SAnand Jain else 30950b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30966b526ed7SAnand Jain vol_args->name); 30976b526ed7SAnand Jain } 3098183860f6SAnand Jain out: 3099183860f6SAnand Jain kfree(vol_args); 3100c47ca32dSDan Carpenter err_drop: 31014ac20c70SIlya Dryomov mnt_drop_write_file(file); 3102f46b5a66SChristoph Hellwig return ret; 3103f46b5a66SChristoph Hellwig } 3104f46b5a66SChristoph Hellwig 3105f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3106f46b5a66SChristoph Hellwig { 31070b246afaSJeff Mahoney struct inode *inode = file_inode(file); 31080b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3109f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3110f46b5a66SChristoph Hellwig int ret; 3111f46b5a66SChristoph Hellwig 3112f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3113f46b5a66SChristoph Hellwig return -EPERM; 3114f46b5a66SChristoph Hellwig 3115f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3116f46b5a66SChristoph Hellwig if (ret) 3117f46b5a66SChristoph Hellwig return ret; 3118f46b5a66SChristoph Hellwig 3119171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3120f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 312158d7bbf8SDavid Sterba goto out_drop_write; 312258d7bbf8SDavid Sterba } 312358d7bbf8SDavid Sterba 312458d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 312558d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 312658d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3127f46b5a66SChristoph Hellwig goto out; 3128f46b5a66SChristoph Hellwig } 3129f46b5a66SChristoph Hellwig 313058d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31312ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3132f46b5a66SChristoph Hellwig 3133f46b5a66SChristoph Hellwig if (!ret) 31340b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3135f46b5a66SChristoph Hellwig kfree(vol_args); 313658d7bbf8SDavid Sterba out: 3137171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 313858d7bbf8SDavid Sterba out_drop_write: 3139f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 314058d7bbf8SDavid Sterba 3141f46b5a66SChristoph Hellwig return ret; 3142f46b5a66SChristoph Hellwig } 3143f46b5a66SChristoph Hellwig 31442ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 31452ff7e61eSJeff Mahoney void __user *arg) 3146475f6387SJan Schmidt { 3147027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3148475f6387SJan Schmidt struct btrfs_device *device; 31490b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3150027ed2f0SLi Zefan int ret = 0; 3151475f6387SJan Schmidt 3152027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3153027ed2f0SLi Zefan if (!fi_args) 3154027ed2f0SLi Zefan return -ENOMEM; 3155027ed2f0SLi Zefan 3156d03262c7SDavid Sterba rcu_read_lock(); 3157027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3158475f6387SJan Schmidt 3159d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3160027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3161027ed2f0SLi Zefan fi_args->max_id = device->devid; 3162475f6387SJan Schmidt } 3163d03262c7SDavid Sterba rcu_read_unlock(); 3164475f6387SJan Schmidt 3165de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3166bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3167bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3168bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 316980a773fbSDavid Sterba 3170027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3171027ed2f0SLi Zefan ret = -EFAULT; 3172475f6387SJan Schmidt 3173027ed2f0SLi Zefan kfree(fi_args); 3174027ed2f0SLi Zefan return ret; 3175475f6387SJan Schmidt } 3176475f6387SJan Schmidt 31772ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31782ff7e61eSJeff Mahoney void __user *arg) 3179475f6387SJan Schmidt { 3180475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3181475f6387SJan Schmidt struct btrfs_device *dev; 3182475f6387SJan Schmidt int ret = 0; 3183475f6387SJan Schmidt char *s_uuid = NULL; 3184475f6387SJan Schmidt 3185475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3186475f6387SJan Schmidt if (IS_ERR(di_args)) 3187475f6387SJan Schmidt return PTR_ERR(di_args); 3188475f6387SJan Schmidt 3189dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3190475f6387SJan Schmidt s_uuid = di_args->uuid; 3191475f6387SJan Schmidt 3192c5593ca3SDavid Sterba rcu_read_lock(); 3193e4319cd9SAnand Jain dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid, 319409ba3bc9SAnand Jain NULL, true); 3195475f6387SJan Schmidt 3196475f6387SJan Schmidt if (!dev) { 3197475f6387SJan Schmidt ret = -ENODEV; 3198475f6387SJan Schmidt goto out; 3199475f6387SJan Schmidt } 3200475f6387SJan Schmidt 3201475f6387SJan Schmidt di_args->devid = dev->devid; 32027cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 32037cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3204475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3205a27202fbSJim Meyering if (dev->name) { 3206672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3207672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3208a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3209a27202fbSJim Meyering } else { 321099ba55adSStefan Behrens di_args->path[0] = '\0'; 3211a27202fbSJim Meyering } 3212475f6387SJan Schmidt 3213475f6387SJan Schmidt out: 3214c5593ca3SDavid Sterba rcu_read_unlock(); 3215475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3216475f6387SJan Schmidt ret = -EFAULT; 3217475f6387SJan Schmidt 3218475f6387SJan Schmidt kfree(di_args); 3219475f6387SJan Schmidt return ret; 3220475f6387SJan Schmidt } 3221475f6387SJan Schmidt 3222d8b55242SFilipe Manana static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3223d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3224d8b55242SFilipe Manana { 3225d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3226d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3227d8b55242SFilipe Manana } 3228d8b55242SFilipe Manana 3229d8b55242SFilipe Manana static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3230d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3231d8b55242SFilipe Manana { 3232d8b55242SFilipe Manana if (inode1 < inode2) { 3233d8b55242SFilipe Manana swap(inode1, inode2); 3234d8b55242SFilipe Manana swap(loff1, loff2); 3235d8b55242SFilipe Manana } else if (inode1 == inode2 && loff2 < loff1) { 3236d8b55242SFilipe Manana swap(loff1, loff2); 3237d8b55242SFilipe Manana } 3238d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3239d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3240d8b55242SFilipe Manana } 3241d8b55242SFilipe Manana 324257a50e25SFilipe Manana static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len, 324334a28e3dSFilipe Manana struct inode *dst, u64 dst_loff) 3244416161dbSMark Fasheh { 3245416161dbSMark Fasheh int ret; 32460efa9f48SMark Fasheh 3247de02b9f6SFilipe Manana /* 3248d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3249d8b55242SFilipe Manana * source range to serialize with relocation. 3250de02b9f6SFilipe Manana */ 3251d8b55242SFilipe Manana btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 325257a50e25SFilipe Manana ret = btrfs_clone(src, dst, loff, len, len, dst_loff, 1); 3253d8b55242SFilipe Manana btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 32543973909dSTimofey Titovets 32553973909dSTimofey Titovets return ret; 32563973909dSTimofey Titovets } 32573973909dSTimofey Titovets 3258b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3259b6728768STimofey Titovets 32603973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 32613973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 32623973909dSTimofey Titovets { 32633973909dSTimofey Titovets int ret; 3264b6728768STimofey Titovets u64 i, tail_len, chunk_count; 326562d54f3aSFilipe Manana struct btrfs_root *root_dst = BTRFS_I(dst)->root; 326662d54f3aSFilipe Manana 326762d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 326862d54f3aSFilipe Manana if (root_dst->send_in_progress) { 326962d54f3aSFilipe Manana btrfs_warn_rl(root_dst->fs_info, 327062d54f3aSFilipe Manana "cannot deduplicate to root %llu while send operations are using it (%d in progress)", 327162d54f3aSFilipe Manana root_dst->root_key.objectid, 327262d54f3aSFilipe Manana root_dst->send_in_progress); 327362d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 327462d54f3aSFilipe Manana return -EAGAIN; 327562d54f3aSFilipe Manana } 327662d54f3aSFilipe Manana root_dst->dedupe_in_progress++; 327762d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 32783973909dSTimofey Titovets 3279b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3280b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 328167b07bd4STimofey Titovets 3282b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3283b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 328434a28e3dSFilipe Manana dst, dst_loff); 3285b6728768STimofey Titovets if (ret) 328662d54f3aSFilipe Manana goto out; 3287b6728768STimofey Titovets 3288b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3289b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3290b6728768STimofey Titovets } 3291b6728768STimofey Titovets 3292b6728768STimofey Titovets if (tail_len > 0) 329367b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 329434a28e3dSFilipe Manana dst_loff); 329562d54f3aSFilipe Manana out: 329662d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 329762d54f3aSFilipe Manana root_dst->dedupe_in_progress--; 329862d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 3299416161dbSMark Fasheh 3300416161dbSMark Fasheh return ret; 3301416161dbSMark Fasheh } 3302416161dbSMark Fasheh 3303f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3304f82a9901SFilipe Manana struct inode *inode, 3305f82a9901SFilipe Manana u64 endoff, 3306f82a9901SFilipe Manana const u64 destoff, 33071c919a5eSMark Fasheh const u64 olen, 33081c919a5eSMark Fasheh int no_time_update) 3309f82a9901SFilipe Manana { 3310f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3311f82a9901SFilipe Manana int ret; 3312f82a9901SFilipe Manana 3313f82a9901SFilipe Manana inode_inc_iversion(inode); 33141c919a5eSMark Fasheh if (!no_time_update) 3315c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3316f82a9901SFilipe Manana /* 3317f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3318f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3319f82a9901SFilipe Manana */ 3320f82a9901SFilipe Manana if (endoff > destoff + olen) 3321f82a9901SFilipe Manana endoff = destoff + olen; 3322f82a9901SFilipe Manana if (endoff > inode->i_size) 33236ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3324f82a9901SFilipe Manana 3325f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3326f82a9901SFilipe Manana if (ret) { 332766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 33283a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3329f82a9901SFilipe Manana goto out; 3330f82a9901SFilipe Manana } 33313a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3332f82a9901SFilipe Manana out: 3333f82a9901SFilipe Manana return ret; 3334f82a9901SFilipe Manana } 3335f82a9901SFilipe Manana 3336a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 33377ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 33387ffbb598SFilipe Manana const struct btrfs_path *path, 33397ffbb598SFilipe Manana const u64 hole_offset, 33407ffbb598SFilipe Manana const u64 hole_len) 33417ffbb598SFilipe Manana { 3342a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 33437ffbb598SFilipe Manana struct extent_map *em; 33447ffbb598SFilipe Manana int ret; 33457ffbb598SFilipe Manana 33467ffbb598SFilipe Manana em = alloc_extent_map(); 33477ffbb598SFilipe Manana if (!em) { 3348a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33497ffbb598SFilipe Manana return; 33507ffbb598SFilipe Manana } 33517ffbb598SFilipe Manana 335214f59796SFilipe Manana if (path) { 335314f59796SFilipe Manana struct btrfs_file_extent_item *fi; 335414f59796SFilipe Manana 335514f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 335614f59796SFilipe Manana struct btrfs_file_extent_item); 3357a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 33587ffbb598SFilipe Manana em->generation = -1; 33597ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 33607ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33617ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3362a2f392e4SNikolay Borisov &inode->runtime_flags); 33637ffbb598SFilipe Manana } else { 33647ffbb598SFilipe Manana em->start = hole_offset; 33657ffbb598SFilipe Manana em->len = hole_len; 33667ffbb598SFilipe Manana em->ram_bytes = em->len; 33677ffbb598SFilipe Manana em->orig_start = hole_offset; 33687ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33697ffbb598SFilipe Manana em->block_len = 0; 33707ffbb598SFilipe Manana em->orig_block_len = 0; 33717ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33727ffbb598SFilipe Manana em->generation = trans->transid; 33737ffbb598SFilipe Manana } 33747ffbb598SFilipe Manana 33757ffbb598SFilipe Manana while (1) { 33767ffbb598SFilipe Manana write_lock(&em_tree->lock); 33777ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33787ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33797ffbb598SFilipe Manana if (ret != -EEXIST) { 33807ffbb598SFilipe Manana free_extent_map(em); 33817ffbb598SFilipe Manana break; 33827ffbb598SFilipe Manana } 3383a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 33847ffbb598SFilipe Manana em->start + em->len - 1, 0); 33857ffbb598SFilipe Manana } 33867ffbb598SFilipe Manana 3387ee39b432SDavid Sterba if (ret) 3388a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33897ffbb598SFilipe Manana } 33907ffbb598SFilipe Manana 33918039d87dSFilipe Manana /* 33928039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33938039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33948039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33958039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33968039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33978039d87dSFilipe Manana * 33988039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33998039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 34008039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 34018039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 34028039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 34038039d87dSFilipe Manana * i_mutex, which we are holding here). 34048039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 34058039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 34068039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 34078039d87dSFilipe Manana * 34088039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 34098039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 34108039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 34118039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 34128039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 34138039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 34148039d87dSFilipe Manana * the former. 34158039d87dSFilipe Manana */ 34164a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 34178039d87dSFilipe Manana struct btrfs_trans_handle *trans, 34188039d87dSFilipe Manana struct btrfs_path *path, 34198039d87dSFilipe Manana struct btrfs_key *new_key, 34208039d87dSFilipe Manana const u64 drop_start, 34218039d87dSFilipe Manana const u64 datal, 34228039d87dSFilipe Manana const u64 skip, 34238039d87dSFilipe Manana const u64 size, 34248039d87dSFilipe Manana char *inline_data) 34258039d87dSFilipe Manana { 34260b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 34278039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 34288039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 34290b246afaSJeff Mahoney fs_info->sectorsize); 34308039d87dSFilipe Manana int ret; 34318039d87dSFilipe Manana struct btrfs_key key; 34328039d87dSFilipe Manana 34338039d87dSFilipe Manana if (new_key->offset > 0) 34348039d87dSFilipe Manana return -EOPNOTSUPP; 34358039d87dSFilipe Manana 34364a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 34378039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 34388039d87dSFilipe Manana key.offset = 0; 34398039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 34408039d87dSFilipe Manana if (ret < 0) { 34418039d87dSFilipe Manana return ret; 34428039d87dSFilipe Manana } else if (ret > 0) { 34438039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 34448039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 34458039d87dSFilipe Manana if (ret < 0) 34468039d87dSFilipe Manana return ret; 34478039d87dSFilipe Manana else if (ret > 0) 34488039d87dSFilipe Manana goto copy_inline_extent; 34498039d87dSFilipe Manana } 34508039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 34514a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34528039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 34538039d87dSFilipe Manana ASSERT(key.offset > 0); 34548039d87dSFilipe Manana return -EOPNOTSUPP; 34558039d87dSFilipe Manana } 34568039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34578039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34588039d87dSFilipe Manana u64 ext_len; 34598039d87dSFilipe Manana 34608039d87dSFilipe Manana /* 34618039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34628039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34638039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34648039d87dSFilipe Manana */ 34658039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34668039d87dSFilipe Manana struct btrfs_file_extent_item); 34678039d87dSFilipe Manana /* 34688039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34698039d87dSFilipe Manana * following it. 34708039d87dSFilipe Manana */ 34718039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34728039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34738039d87dSFilipe Manana goto copy_inline_extent; 34748039d87dSFilipe Manana 34758039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34768039d87dSFilipe Manana if (ext_len > aligned_end) 34778039d87dSFilipe Manana return -EOPNOTSUPP; 34788039d87dSFilipe Manana 34798039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34808039d87dSFilipe Manana if (ret < 0) { 34818039d87dSFilipe Manana return ret; 34828039d87dSFilipe Manana } else if (ret == 0) { 34838039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34848039d87dSFilipe Manana path->slots[0]); 34854a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34868039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34878039d87dSFilipe Manana return -EOPNOTSUPP; 34888039d87dSFilipe Manana } 34898039d87dSFilipe Manana } 34908039d87dSFilipe Manana 34918039d87dSFilipe Manana copy_inline_extent: 34928039d87dSFilipe Manana /* 34938039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34948039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34958039d87dSFilipe Manana */ 34968039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34978039d87dSFilipe Manana /* 34988039d87dSFilipe Manana * If the destination inode has an inline extent... 34998039d87dSFilipe Manana * This would require copying the data from the source inline 35008039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 35018039d87dSFilipe Manana * But this is really complex, both extents can be compressed 35028039d87dSFilipe Manana * or just one of them, which would require decompressing and 35038039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 35048039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 35058039d87dSFilipe Manana * inline extent). 35068039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 35078039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 35088039d87dSFilipe Manana */ 35098039d87dSFilipe Manana return -EOPNOTSUPP; 35108039d87dSFilipe Manana } 35118039d87dSFilipe Manana 35128039d87dSFilipe Manana btrfs_release_path(path); 35138039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 35148039d87dSFilipe Manana if (ret) 35158039d87dSFilipe Manana return ret; 35168039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 35178039d87dSFilipe Manana if (ret) 35188039d87dSFilipe Manana return ret; 35198039d87dSFilipe Manana 35208039d87dSFilipe Manana if (skip) { 35218039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 35228039d87dSFilipe Manana 35238039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 35248039d87dSFilipe Manana } 35258039d87dSFilipe Manana 35268039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 35278039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 35288039d87dSFilipe Manana path->slots[0]), 35298039d87dSFilipe Manana size); 35308039d87dSFilipe Manana inode_add_bytes(dst, datal); 35318039d87dSFilipe Manana 35328039d87dSFilipe Manana return 0; 35338039d87dSFilipe Manana } 35348039d87dSFilipe Manana 353532b7c687SMark Fasheh /** 353632b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 353732b7c687SMark Fasheh * 353832b7c687SMark Fasheh * @src: Inode to clone from 353932b7c687SMark Fasheh * @inode: Inode to clone to 354032b7c687SMark Fasheh * @off: Offset within source to start clone from 354132b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 35421c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 354332b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 35441c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 354532b7c687SMark Fasheh */ 354632b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3547f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 35481c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3549f46b5a66SChristoph Hellwig { 35500b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3551f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 355232b7c687SMark Fasheh struct btrfs_path *path = NULL; 3553f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 355432b7c687SMark Fasheh struct btrfs_trans_handle *trans; 355532b7c687SMark Fasheh char *buf = NULL; 3556ae01a0abSYan Zheng struct btrfs_key key; 3557f46b5a66SChristoph Hellwig u32 nritems; 3558f46b5a66SChristoph Hellwig int slot; 3559ae01a0abSYan Zheng int ret; 3560f82a9901SFilipe Manana const u64 len = olen_aligned; 3561f82a9901SFilipe Manana u64 last_dest_end = destoff; 3562ae01a0abSYan Zheng 3563ae01a0abSYan Zheng ret = -ENOMEM; 3564752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3565ae01a0abSYan Zheng if (!buf) 356632b7c687SMark Fasheh return ret; 3567ae01a0abSYan Zheng 3568ae01a0abSYan Zheng path = btrfs_alloc_path(); 3569ae01a0abSYan Zheng if (!path) { 357015351955SDavid Sterba kvfree(buf); 357132b7c687SMark Fasheh return ret; 3572ae01a0abSYan Zheng } 357332b7c687SMark Fasheh 3574e4058b54SDavid Sterba path->reada = READA_FORWARD; 3575c5c9cd4dSSage Weil /* clone data */ 35764a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3577ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35782c463823SFilipe Manana key.offset = off; 3579f46b5a66SChristoph Hellwig 3580f46b5a66SChristoph Hellwig while (1) { 3581de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3582df858e76SFilipe Manana 3583f46b5a66SChristoph Hellwig /* 3584f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3585f46b5a66SChristoph Hellwig * tree. 3586f46b5a66SChristoph Hellwig */ 3587e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3588362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3589362a20c5SDavid Sterba 0, 0); 3590f46b5a66SChristoph Hellwig if (ret < 0) 3591f46b5a66SChristoph Hellwig goto out; 35922c463823SFilipe Manana /* 35932c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35942c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35952c463823SFilipe Manana * it might overlap our target range, therefore process it. 35962c463823SFilipe Manana */ 35972c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35982c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35992c463823SFilipe Manana path->slots[0] - 1); 36002c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 36012c463823SFilipe Manana path->slots[0]--; 36022c463823SFilipe Manana } 3603f46b5a66SChristoph Hellwig 3604ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3605e4355f34SFilipe David Borba Manana process_slot: 3606ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3607362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3608f46b5a66SChristoph Hellwig if (ret < 0) 3609f46b5a66SChristoph Hellwig goto out; 3610f46b5a66SChristoph Hellwig if (ret > 0) 3611f46b5a66SChristoph Hellwig break; 3612ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3613f46b5a66SChristoph Hellwig } 3614f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3615f46b5a66SChristoph Hellwig slot = path->slots[0]; 3616f46b5a66SChristoph Hellwig 3617ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3618962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 36194a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3620f46b5a66SChristoph Hellwig break; 3621f46b5a66SChristoph Hellwig 3622962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3623c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3624c5c9cd4dSSage Weil int type; 362531840ae1SZheng Yan u32 size; 362631840ae1SZheng Yan struct btrfs_key new_key; 3627c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3628c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3629c5c9cd4dSSage Weil u8 comp; 3630f82a9901SFilipe Manana u64 drop_start; 363131840ae1SZheng Yan 3632c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3633c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3634c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3635c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3636c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3637c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3638d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3639d397712bSChris Mason extent); 3640d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3641d397712bSChris Mason extent); 3642c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3643d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3644d397712bSChris Mason extent); 3645c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3646c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3647c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3648c5c9cd4dSSage Weil extent); 3649c5c9cd4dSSage Weil } 365031840ae1SZheng Yan 36512c463823SFilipe Manana /* 36522c463823SFilipe Manana * The first search might have left us at an extent 36532c463823SFilipe Manana * item that ends before our target range's start, can 36542c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 36552c463823SFilipe Manana */ 36562c463823SFilipe Manana if (key.offset + datal <= off) { 3657e4355f34SFilipe David Borba Manana path->slots[0]++; 3658e4355f34SFilipe David Borba Manana goto process_slot; 36592c463823SFilipe Manana } else if (key.offset >= off + len) { 36602c463823SFilipe Manana break; 3661e4355f34SFilipe David Borba Manana } 3662df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3663e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3664e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3665e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3666e4355f34SFilipe David Borba Manana size); 3667e4355f34SFilipe David Borba Manana 3668e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3669e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3670c5c9cd4dSSage Weil 367131840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36724a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36734d728ec7SLi Zefan if (off <= key.offset) 3674c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36754d728ec7SLi Zefan else 36764d728ec7SLi Zefan new_key.offset = destoff; 3677c5c9cd4dSSage Weil 3678b6f3409bSSage Weil /* 3679f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3680f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3681f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3682f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3683f82a9901SFilipe Manana * partially overlaps it). 3684f82a9901SFilipe Manana */ 3685f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3686f82a9901SFilipe Manana drop_start = last_dest_end; 3687f82a9901SFilipe Manana else 3688f82a9901SFilipe Manana drop_start = new_key.offset; 3689f82a9901SFilipe Manana 3690f82a9901SFilipe Manana /* 3691b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3692b6f3409bSSage Weil * 1 - add new extent 3693b6f3409bSSage Weil * 1 - inode update 3694b6f3409bSSage Weil */ 3695b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3696a22285a6SYan, Zheng if (IS_ERR(trans)) { 3697a22285a6SYan, Zheng ret = PTR_ERR(trans); 3698a22285a6SYan, Zheng goto out; 3699a22285a6SYan, Zheng } 3700a22285a6SYan, Zheng 3701c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3702c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3703d72c0842SLi Zefan /* 3704d72c0842SLi Zefan * a | --- range to clone ---| b 3705d72c0842SLi Zefan * | ------------- extent ------------- | 3706d72c0842SLi Zefan */ 3707d72c0842SLi Zefan 370893915584SAntonio Ospite /* subtract range b */ 3709d72c0842SLi Zefan if (key.offset + datal > off + len) 3710d72c0842SLi Zefan datal = off + len - key.offset; 3711d72c0842SLi Zefan 371293915584SAntonio Ospite /* subtract range a */ 3713a22285a6SYan, Zheng if (off > key.offset) { 3714a22285a6SYan, Zheng datao += off - key.offset; 3715a22285a6SYan, Zheng datal -= off - key.offset; 3716a22285a6SYan, Zheng } 3717a22285a6SYan, Zheng 37185dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3719f82a9901SFilipe Manana drop_start, 3720a22285a6SYan, Zheng new_key.offset + datal, 37212671485dSJosef Bacik 1); 372279787eaaSJeff Mahoney if (ret) { 37233f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 372400fdf13aSLiu Bo btrfs_abort_transaction(trans, 372566642832SJeff Mahoney ret); 37263a45bb20SJeff Mahoney btrfs_end_transaction(trans); 372779787eaaSJeff Mahoney goto out; 372879787eaaSJeff Mahoney } 3729a22285a6SYan, Zheng 373031840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 373131840ae1SZheng Yan &new_key, size); 373279787eaaSJeff Mahoney if (ret) { 373366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 37343a45bb20SJeff Mahoney btrfs_end_transaction(trans); 373579787eaaSJeff Mahoney goto out; 373679787eaaSJeff Mahoney } 373731840ae1SZheng Yan 373831840ae1SZheng Yan leaf = path->nodes[0]; 373931840ae1SZheng Yan slot = path->slots[0]; 374031840ae1SZheng Yan write_extent_buffer(leaf, buf, 374131840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 374231840ae1SZheng Yan size); 3743ae01a0abSYan Zheng 3744f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3745f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3746c5c9cd4dSSage Weil 3747c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3748c5c9cd4dSSage Weil if (!disko) 3749c5c9cd4dSSage Weil datao = 0; 3750c5c9cd4dSSage Weil 3751c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3752c5c9cd4dSSage Weil datao); 3753c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3754c5c9cd4dSSage Weil datal); 3755fcebe456SJosef Bacik 3756c5c9cd4dSSage Weil if (disko) { 375782fa113fSQu Wenruo struct btrfs_ref ref = { 0 }; 3758c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 375982fa113fSQu Wenruo btrfs_init_generic_ref(&ref, 376082fa113fSQu Wenruo BTRFS_ADD_DELAYED_REF, disko, 376182fa113fSQu Wenruo diskl, 0); 376282fa113fSQu Wenruo btrfs_init_data_ref(&ref, 3763f46b5a66SChristoph Hellwig root->root_key.objectid, 37644a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 3765b06c4bf5SFilipe Manana new_key.offset - datao); 376682fa113fSQu Wenruo ret = btrfs_inc_extent_ref(trans, &ref); 376779787eaaSJeff Mahoney if (ret) { 376879787eaaSJeff Mahoney btrfs_abort_transaction(trans, 376979787eaaSJeff Mahoney ret); 37703a45bb20SJeff Mahoney btrfs_end_transaction(trans); 377179787eaaSJeff Mahoney goto out; 377279787eaaSJeff Mahoney 377379787eaaSJeff Mahoney } 3774f46b5a66SChristoph Hellwig } 3775c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3776c5c9cd4dSSage Weil u64 skip = 0; 3777c5c9cd4dSSage Weil u64 trim = 0; 3778ed958762SFilipe Manana 3779c5c9cd4dSSage Weil if (off > key.offset) { 3780c5c9cd4dSSage Weil skip = off - key.offset; 3781c5c9cd4dSSage Weil new_key.offset += skip; 378231840ae1SZheng Yan } 3783d397712bSChris Mason 3784c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3785c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3786d397712bSChris Mason 3787c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3788c5c9cd4dSSage Weil ret = -EINVAL; 37893a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3790c5c9cd4dSSage Weil goto out; 379131840ae1SZheng Yan } 3792c5c9cd4dSSage Weil size -= skip + trim; 3793c5c9cd4dSSage Weil datal -= skip + trim; 3794a22285a6SYan, Zheng 37954a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 37968039d87dSFilipe Manana trans, path, 37978039d87dSFilipe Manana &new_key, 3798f82a9901SFilipe Manana drop_start, 37998039d87dSFilipe Manana datal, 38008039d87dSFilipe Manana skip, size, buf); 380179787eaaSJeff Mahoney if (ret) { 38023f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 38033a29bc09SChris Mason btrfs_abort_transaction(trans, 380479787eaaSJeff Mahoney ret); 38053a45bb20SJeff Mahoney btrfs_end_transaction(trans); 380679787eaaSJeff Mahoney goto out; 380779787eaaSJeff Mahoney } 3808c5c9cd4dSSage Weil leaf = path->nodes[0]; 3809c5c9cd4dSSage Weil slot = path->slots[0]; 3810c5c9cd4dSSage Weil } 3811c5c9cd4dSSage Weil 38127ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 38137ffbb598SFilipe Manana if (drop_start < new_key.offset) 3814a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 381514f59796SFilipe Manana NULL, drop_start, 38167ffbb598SFilipe Manana new_key.offset - drop_start); 38177ffbb598SFilipe Manana 3818a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 3819a2f392e4SNikolay Borisov path, 0, 0); 38207ffbb598SFilipe Manana 3821c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3822b3b4aa74SDavid Sterba btrfs_release_path(path); 3823c5c9cd4dSSage Weil 382462e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 38250b246afaSJeff Mahoney fs_info->sectorsize); 3826f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3827f82a9901SFilipe Manana last_dest_end, 38281c919a5eSMark Fasheh destoff, olen, 38291c919a5eSMark Fasheh no_time_update); 3830f82a9901SFilipe Manana if (ret) 383179787eaaSJeff Mahoney goto out; 38322c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 38332c463823SFilipe Manana break; 3834a22285a6SYan, Zheng } 3835b3b4aa74SDavid Sterba btrfs_release_path(path); 3836df858e76SFilipe Manana key.offset = next_key_min_offset; 383769ae5e44SWang Xiaoguang 383869ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 383969ae5e44SWang Xiaoguang ret = -EINTR; 384069ae5e44SWang Xiaoguang goto out; 384169ae5e44SWang Xiaoguang } 3842ae01a0abSYan Zheng } 3843f46b5a66SChristoph Hellwig ret = 0; 384432b7c687SMark Fasheh 3845f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3846f82a9901SFilipe Manana /* 3847f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3848f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3849f82a9901SFilipe Manana */ 3850b3b4aa74SDavid Sterba btrfs_release_path(path); 3851f82a9901SFilipe Manana 3852f82a9901SFilipe Manana /* 3853f82a9901SFilipe Manana * 1 - remove extent(s) 3854f82a9901SFilipe Manana * 1 - inode update 3855f82a9901SFilipe Manana */ 3856f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3857f82a9901SFilipe Manana if (IS_ERR(trans)) { 3858f82a9901SFilipe Manana ret = PTR_ERR(trans); 3859f82a9901SFilipe Manana goto out; 3860f82a9901SFilipe Manana } 3861f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3862f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3863f82a9901SFilipe Manana if (ret) { 3864f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 386566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 38663a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3867f82a9901SFilipe Manana goto out; 3868f82a9901SFilipe Manana } 3869a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 3870a2f392e4SNikolay Borisov last_dest_end, 387114f59796SFilipe Manana destoff + len - last_dest_end); 3872f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38731c919a5eSMark Fasheh destoff, olen, no_time_update); 3874f82a9901SFilipe Manana } 3875f82a9901SFilipe Manana 3876f46b5a66SChristoph Hellwig out: 387732b7c687SMark Fasheh btrfs_free_path(path); 387815351955SDavid Sterba kvfree(buf); 387932b7c687SMark Fasheh return ret; 388032b7c687SMark Fasheh } 388132b7c687SMark Fasheh 38823db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 388332b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 388432b7c687SMark Fasheh { 388554563d41SAl Viro struct inode *inode = file_inode(file); 38863db11b2eSZach Brown struct inode *src = file_inode(file_src); 38870b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 388832b7c687SMark Fasheh int ret; 388932b7c687SMark Fasheh u64 len = olen; 38900b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 389132b7c687SMark Fasheh 389232b7c687SMark Fasheh /* 389332b7c687SMark Fasheh * TODO: 389432b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 389532b7c687SMark Fasheh * decompress into destination's address_space (the file offset 389632b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 389732b7c687SMark Fasheh * otherwise reinsert) a subrange. 389800fdf13aSLiu Bo * 389900fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 390000fdf13aSLiu Bo * be either compressed or non-compressed. 390132b7c687SMark Fasheh */ 390232b7c687SMark Fasheh 3903ac765f83SFilipe Manana /* 390434a28e3dSFilipe Manana * VFS's generic_remap_file_range_prep() protects us from cloning the 390534a28e3dSFilipe Manana * eof block into the middle of a file, which would result in corruption 390634a28e3dSFilipe Manana * if the file size is not blocksize aligned. So we don't need to check 390734a28e3dSFilipe Manana * for that case here. 3908ac765f83SFilipe Manana */ 390934a28e3dSFilipe Manana if (off + len == src->i_size) 391032b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 391132b7c687SMark Fasheh 391232b7c687SMark Fasheh if (destoff > inode->i_size) { 3913f7fa1107SFilipe Manana const u64 wb_start = ALIGN_DOWN(inode->i_size, bs); 3914f7fa1107SFilipe Manana 391532b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 391632b7c687SMark Fasheh if (ret) 391734a28e3dSFilipe Manana return ret; 3918f7fa1107SFilipe Manana /* 3919f7fa1107SFilipe Manana * We may have truncated the last block if the inode's size is 3920f7fa1107SFilipe Manana * not sector size aligned, so we need to wait for writeback to 3921f7fa1107SFilipe Manana * complete before proceeding further, otherwise we can race 3922f7fa1107SFilipe Manana * with cloning and attempt to increment a reference to an 3923f7fa1107SFilipe Manana * extent that no longer exists (writeback completed right after 3924f7fa1107SFilipe Manana * we found the previous extent covering eof and before we 3925f7fa1107SFilipe Manana * attempted to increment its reference count). 3926f7fa1107SFilipe Manana */ 3927f7fa1107SFilipe Manana ret = btrfs_wait_ordered_range(inode, wb_start, 3928f7fa1107SFilipe Manana destoff - wb_start); 3929f7fa1107SFilipe Manana if (ret) 3930f7fa1107SFilipe Manana return ret; 393132b7c687SMark Fasheh } 393232b7c687SMark Fasheh 3933c125b8bfSFilipe Manana /* 3934d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3935d8b55242SFilipe Manana * source range to serialize with relocation. 3936c125b8bfSFilipe Manana */ 3937d8b55242SFilipe Manana btrfs_double_extent_lock(src, off, inode, destoff, len); 39381c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 3939d8b55242SFilipe Manana btrfs_double_extent_unlock(src, off, inode, destoff, len); 3940c125b8bfSFilipe Manana /* 3941c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3942c125b8bfSFilipe Manana * data immediately and not the previous data. 3943c125b8bfSFilipe Manana */ 394465bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 394509cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 394609cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 394734a28e3dSFilipe Manana 394834a28e3dSFilipe Manana return ret; 394934a28e3dSFilipe Manana } 395034a28e3dSFilipe Manana 395134a28e3dSFilipe Manana static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in, 395234a28e3dSFilipe Manana struct file *file_out, loff_t pos_out, 395334a28e3dSFilipe Manana loff_t *len, unsigned int remap_flags) 395434a28e3dSFilipe Manana { 395534a28e3dSFilipe Manana struct inode *inode_in = file_inode(file_in); 395634a28e3dSFilipe Manana struct inode *inode_out = file_inode(file_out); 395734a28e3dSFilipe Manana u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize; 395834a28e3dSFilipe Manana bool same_inode = inode_out == inode_in; 395934a28e3dSFilipe Manana u64 wb_len; 396034a28e3dSFilipe Manana int ret; 396134a28e3dSFilipe Manana 396234a28e3dSFilipe Manana if (!(remap_flags & REMAP_FILE_DEDUP)) { 396334a28e3dSFilipe Manana struct btrfs_root *root_out = BTRFS_I(inode_out)->root; 396434a28e3dSFilipe Manana 396534a28e3dSFilipe Manana if (btrfs_root_readonly(root_out)) 396634a28e3dSFilipe Manana return -EROFS; 396734a28e3dSFilipe Manana 396834a28e3dSFilipe Manana if (file_in->f_path.mnt != file_out->f_path.mnt || 396934a28e3dSFilipe Manana inode_in->i_sb != inode_out->i_sb) 397034a28e3dSFilipe Manana return -EXDEV; 397134a28e3dSFilipe Manana } 397234a28e3dSFilipe Manana 3973500710d3SFilipe Manana /* don't make the dst file partly checksummed */ 3974500710d3SFilipe Manana if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) != 3975500710d3SFilipe Manana (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) { 39767984ae52SGoldwyn Rodrigues return -EINVAL; 3977500710d3SFilipe Manana } 3978500710d3SFilipe Manana 397934a28e3dSFilipe Manana /* 398034a28e3dSFilipe Manana * Now that the inodes are locked, we need to start writeback ourselves 398134a28e3dSFilipe Manana * and can not rely on the writeback from the VFS's generic helper 398234a28e3dSFilipe Manana * generic_remap_file_range_prep() because: 398334a28e3dSFilipe Manana * 398434a28e3dSFilipe Manana * 1) For compression we must call filemap_fdatawrite_range() range 398534a28e3dSFilipe Manana * twice (btrfs_fdatawrite_range() does it for us), and the generic 398634a28e3dSFilipe Manana * helper only calls it once; 398734a28e3dSFilipe Manana * 398834a28e3dSFilipe Manana * 2) filemap_fdatawrite_range(), called by the generic helper only 398934a28e3dSFilipe Manana * waits for the writeback to complete, i.e. for IO to be done, and 399034a28e3dSFilipe Manana * not for the ordered extents to complete. We need to wait for them 399134a28e3dSFilipe Manana * to complete so that new file extent items are in the fs tree. 399234a28e3dSFilipe Manana */ 399334a28e3dSFilipe Manana if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP)) 399434a28e3dSFilipe Manana wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs); 399534a28e3dSFilipe Manana else 399634a28e3dSFilipe Manana wb_len = ALIGN(*len, bs); 399734a28e3dSFilipe Manana 399834a28e3dSFilipe Manana /* 399934a28e3dSFilipe Manana * Since we don't lock ranges, wait for ongoing lockless dio writes (as 400034a28e3dSFilipe Manana * any in progress could create its ordered extents after we wait for 400134a28e3dSFilipe Manana * existing ordered extents below). 400234a28e3dSFilipe Manana */ 400334a28e3dSFilipe Manana inode_dio_wait(inode_in); 400434a28e3dSFilipe Manana if (!same_inode) 400534a28e3dSFilipe Manana inode_dio_wait(inode_out); 400634a28e3dSFilipe Manana 400734a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs), 400834a28e3dSFilipe Manana wb_len); 400934a28e3dSFilipe Manana if (ret < 0) 40107984ae52SGoldwyn Rodrigues return ret; 401134a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs), 401234a28e3dSFilipe Manana wb_len); 401334a28e3dSFilipe Manana if (ret < 0) 40143db11b2eSZach Brown return ret; 40157984ae52SGoldwyn Rodrigues 40167984ae52SGoldwyn Rodrigues return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out, 40177984ae52SGoldwyn Rodrigues len, remap_flags); 40183db11b2eSZach Brown } 40193db11b2eSZach Brown 402042ec3d4cSDarrick J. Wong loff_t btrfs_remap_file_range(struct file *src_file, loff_t off, 402142ec3d4cSDarrick J. Wong struct file *dst_file, loff_t destoff, loff_t len, 40222e5dfc99SDarrick J. Wong unsigned int remap_flags) 40233db11b2eSZach Brown { 402434a28e3dSFilipe Manana struct inode *src_inode = file_inode(src_file); 402534a28e3dSFilipe Manana struct inode *dst_inode = file_inode(dst_file); 402634a28e3dSFilipe Manana bool same_inode = dst_inode == src_inode; 402742ec3d4cSDarrick J. Wong int ret; 402842ec3d4cSDarrick J. Wong 40292e5dfc99SDarrick J. Wong if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY)) 40302e5dfc99SDarrick J. Wong return -EINVAL; 40312e5dfc99SDarrick J. Wong 40327984ae52SGoldwyn Rodrigues if (same_inode) 40337984ae52SGoldwyn Rodrigues inode_lock(src_inode); 40347984ae52SGoldwyn Rodrigues else 40357984ae52SGoldwyn Rodrigues lock_two_nondirectories(src_inode, dst_inode); 40367984ae52SGoldwyn Rodrigues 403734a28e3dSFilipe Manana ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff, 403834a28e3dSFilipe Manana &len, remap_flags); 403934a28e3dSFilipe Manana if (ret < 0 || len == 0) 40407984ae52SGoldwyn Rodrigues goto out_unlock; 40412e5dfc99SDarrick J. Wong 404234a28e3dSFilipe Manana if (remap_flags & REMAP_FILE_DEDUP) 404334a28e3dSFilipe Manana ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff); 404434a28e3dSFilipe Manana else 404542ec3d4cSDarrick J. Wong ret = btrfs_clone_files(dst_file, src_file, off, len, destoff); 404634a28e3dSFilipe Manana 40477984ae52SGoldwyn Rodrigues out_unlock: 404834a28e3dSFilipe Manana if (same_inode) 404934a28e3dSFilipe Manana inode_unlock(src_inode); 405034a28e3dSFilipe Manana else 40514ea748e1SFilipe Manana unlock_two_nondirectories(src_inode, dst_inode); 405234a28e3dSFilipe Manana 405342ec3d4cSDarrick J. Wong return ret < 0 ? ret : len; 4054c5c9cd4dSSage Weil } 4055c5c9cd4dSSage Weil 40566ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40576ef5ed0dSJosef Bacik { 4058496ad9aaSAl Viro struct inode *inode = file_inode(file); 40590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 40606ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40616ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40626ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40636ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40646ef5ed0dSJosef Bacik struct btrfs_path *path; 40656ef5ed0dSJosef Bacik struct btrfs_key location; 40666ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40676ef5ed0dSJosef Bacik u64 objectid = 0; 40686ef5ed0dSJosef Bacik u64 dir_id; 40693c04ce01SMiao Xie int ret; 40706ef5ed0dSJosef Bacik 40716ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40726ef5ed0dSJosef Bacik return -EPERM; 40736ef5ed0dSJosef Bacik 40743c04ce01SMiao Xie ret = mnt_want_write_file(file); 40753c04ce01SMiao Xie if (ret) 40763c04ce01SMiao Xie return ret; 40773c04ce01SMiao Xie 40783c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40793c04ce01SMiao Xie ret = -EFAULT; 40803c04ce01SMiao Xie goto out; 40813c04ce01SMiao Xie } 40826ef5ed0dSJosef Bacik 40836ef5ed0dSJosef Bacik if (!objectid) 40841cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40856ef5ed0dSJosef Bacik 40866ef5ed0dSJosef Bacik location.objectid = objectid; 40876ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40886ef5ed0dSJosef Bacik location.offset = (u64)-1; 40896ef5ed0dSJosef Bacik 40900b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 40913c04ce01SMiao Xie if (IS_ERR(new_root)) { 40923c04ce01SMiao Xie ret = PTR_ERR(new_root); 40933c04ce01SMiao Xie goto out; 40943c04ce01SMiao Xie } 40954fd786e6SMisono Tomohiro if (!is_fstree(new_root->root_key.objectid)) { 40966d6d2829Ssatoru takeuchi ret = -ENOENT; 40976d6d2829Ssatoru takeuchi goto out; 40986d6d2829Ssatoru takeuchi } 40996ef5ed0dSJosef Bacik 41006ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 41013c04ce01SMiao Xie if (!path) { 41023c04ce01SMiao Xie ret = -ENOMEM; 41033c04ce01SMiao Xie goto out; 41043c04ce01SMiao Xie } 41056ef5ed0dSJosef Bacik path->leave_spinning = 1; 41066ef5ed0dSJosef Bacik 41076ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 410898d5dc13STsutomu Itoh if (IS_ERR(trans)) { 41096ef5ed0dSJosef Bacik btrfs_free_path(path); 41103c04ce01SMiao Xie ret = PTR_ERR(trans); 41113c04ce01SMiao Xie goto out; 41126ef5ed0dSJosef Bacik } 41136ef5ed0dSJosef Bacik 41140b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 41150b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 41166ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4117cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 41186ef5ed0dSJosef Bacik btrfs_free_path(path); 41193a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41200b246afaSJeff Mahoney btrfs_err(fs_info, 41215d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 41223c04ce01SMiao Xie ret = -ENOENT; 41233c04ce01SMiao Xie goto out; 41246ef5ed0dSJosef Bacik } 41256ef5ed0dSJosef Bacik 41266ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 41276ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 41286ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 41296ef5ed0dSJosef Bacik btrfs_free_path(path); 41306ef5ed0dSJosef Bacik 41310b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 41323a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41333c04ce01SMiao Xie out: 41343c04ce01SMiao Xie mnt_drop_write_file(file); 41353c04ce01SMiao Xie return ret; 41366ef5ed0dSJosef Bacik } 41376ef5ed0dSJosef Bacik 4138c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4139bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4140bf5fc093SJosef Bacik { 4141bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4142bf5fc093SJosef Bacik 4143bf5fc093SJosef Bacik space->total_bytes = 0; 4144bf5fc093SJosef Bacik space->used_bytes = 0; 4145bf5fc093SJosef Bacik space->flags = 0; 4146bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4147bf5fc093SJosef Bacik space->flags = block_group->flags; 4148bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4149bf5fc093SJosef Bacik space->used_bytes += 4150bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4151bf5fc093SJosef Bacik } 4152bf5fc093SJosef Bacik } 4153bf5fc093SJosef Bacik 41542ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 41552ff7e61eSJeff Mahoney void __user *arg) 41561406e432SJosef Bacik { 41571406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41581406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41591406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41607fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 416113f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41621406e432SJosef Bacik struct btrfs_space_info *info; 4163315d8e98SColin Ian King static const u64 types[] = { 4164315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4165bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4166bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4167315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4168315d8e98SColin Ian King }; 4169bf5fc093SJosef Bacik int num_types = 4; 41707fde62bfSChris Mason int alloc_size; 41711406e432SJosef Bacik int ret = 0; 417251788b1bSDan Rosenberg u64 slot_count = 0; 4173bf5fc093SJosef Bacik int i, c; 41741406e432SJosef Bacik 41751406e432SJosef Bacik if (copy_from_user(&space_args, 41761406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41771406e432SJosef Bacik sizeof(space_args))) 41781406e432SJosef Bacik return -EFAULT; 41791406e432SJosef Bacik 4180bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4181bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4182bf5fc093SJosef Bacik 4183bf5fc093SJosef Bacik info = NULL; 41847fde62bfSChris Mason rcu_read_lock(); 41850b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4186bf5fc093SJosef Bacik list) { 4187bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4188bf5fc093SJosef Bacik info = tmp; 4189bf5fc093SJosef Bacik break; 4190bf5fc093SJosef Bacik } 4191bf5fc093SJosef Bacik } 41927fde62bfSChris Mason rcu_read_unlock(); 41931406e432SJosef Bacik 4194bf5fc093SJosef Bacik if (!info) 4195bf5fc093SJosef Bacik continue; 4196bf5fc093SJosef Bacik 4197bf5fc093SJosef Bacik down_read(&info->groups_sem); 4198bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4199bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4200bf5fc093SJosef Bacik slot_count++; 4201bf5fc093SJosef Bacik } 4202bf5fc093SJosef Bacik up_read(&info->groups_sem); 4203bf5fc093SJosef Bacik } 4204bf5fc093SJosef Bacik 420536523e95SDavid Sterba /* 420636523e95SDavid Sterba * Global block reserve, exported as a space_info 420736523e95SDavid Sterba */ 420836523e95SDavid Sterba slot_count++; 420936523e95SDavid Sterba 42107fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 42117fde62bfSChris Mason if (space_args.space_slots == 0) { 42127fde62bfSChris Mason space_args.total_spaces = slot_count; 42137fde62bfSChris Mason goto out; 42147fde62bfSChris Mason } 4215bf5fc093SJosef Bacik 421651788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4217bf5fc093SJosef Bacik 42187fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4219bf5fc093SJosef Bacik 42207fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 42217fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 42227fde62bfSChris Mason */ 422309cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 42247fde62bfSChris Mason return -ENOMEM; 42257fde62bfSChris Mason 42267fde62bfSChris Mason space_args.total_spaces = 0; 42278d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 42287fde62bfSChris Mason if (!dest) 42297fde62bfSChris Mason return -ENOMEM; 42307fde62bfSChris Mason dest_orig = dest; 42317fde62bfSChris Mason 42327fde62bfSChris Mason /* now we have a buffer to copy into */ 4233bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4234bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4235bf5fc093SJosef Bacik 423651788b1bSDan Rosenberg if (!slot_count) 423751788b1bSDan Rosenberg break; 423851788b1bSDan Rosenberg 4239bf5fc093SJosef Bacik info = NULL; 42401406e432SJosef Bacik rcu_read_lock(); 42410b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4242bf5fc093SJosef Bacik list) { 4243bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4244bf5fc093SJosef Bacik info = tmp; 42457fde62bfSChris Mason break; 4246bf5fc093SJosef Bacik } 4247bf5fc093SJosef Bacik } 4248bf5fc093SJosef Bacik rcu_read_unlock(); 42497fde62bfSChris Mason 4250bf5fc093SJosef Bacik if (!info) 4251bf5fc093SJosef Bacik continue; 4252bf5fc093SJosef Bacik down_read(&info->groups_sem); 4253bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4254bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4255c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4256c065f5b1SSu Yue &space); 42577fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42581406e432SJosef Bacik dest++; 42591406e432SJosef Bacik space_args.total_spaces++; 426051788b1bSDan Rosenberg slot_count--; 42611406e432SJosef Bacik } 426251788b1bSDan Rosenberg if (!slot_count) 426351788b1bSDan Rosenberg break; 4264bf5fc093SJosef Bacik } 4265bf5fc093SJosef Bacik up_read(&info->groups_sem); 4266bf5fc093SJosef Bacik } 42671406e432SJosef Bacik 426836523e95SDavid Sterba /* 426936523e95SDavid Sterba * Add global block reserve 427036523e95SDavid Sterba */ 427136523e95SDavid Sterba if (slot_count) { 42720b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 427336523e95SDavid Sterba 427436523e95SDavid Sterba spin_lock(&block_rsv->lock); 427536523e95SDavid Sterba space.total_bytes = block_rsv->size; 427636523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 427736523e95SDavid Sterba spin_unlock(&block_rsv->lock); 427836523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 427936523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 428036523e95SDavid Sterba space_args.total_spaces++; 428136523e95SDavid Sterba } 428236523e95SDavid Sterba 42832eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42847fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42857fde62bfSChris Mason 42867fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42877fde62bfSChris Mason ret = -EFAULT; 42887fde62bfSChris Mason 42897fde62bfSChris Mason kfree(dest_orig); 42907fde62bfSChris Mason out: 42917fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42921406e432SJosef Bacik ret = -EFAULT; 42931406e432SJosef Bacik 42941406e432SJosef Bacik return ret; 42951406e432SJosef Bacik } 42961406e432SJosef Bacik 42979a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42989a8c28beSMiao Xie void __user *argp) 429946204592SSage Weil { 430046204592SSage Weil struct btrfs_trans_handle *trans; 430146204592SSage Weil u64 transid; 4302db5b493aSTsutomu Itoh int ret; 430346204592SSage Weil 4304d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4305ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4306ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 430798d5dc13STsutomu Itoh return PTR_ERR(trans); 4308ff7c1d33SMiao Xie 4309ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4310ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4311ff7c1d33SMiao Xie goto out; 4312ff7c1d33SMiao Xie } 431346204592SSage Weil transid = trans->transid; 43143a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 43158b2b2d3cSTsutomu Itoh if (ret) { 43163a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4317db5b493aSTsutomu Itoh return ret; 43188b2b2d3cSTsutomu Itoh } 4319ff7c1d33SMiao Xie out: 432046204592SSage Weil if (argp) 432146204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 432246204592SSage Weil return -EFAULT; 432346204592SSage Weil return 0; 432446204592SSage Weil } 432546204592SSage Weil 43262ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 43279a8c28beSMiao Xie void __user *argp) 432846204592SSage Weil { 432946204592SSage Weil u64 transid; 433046204592SSage Weil 433146204592SSage Weil if (argp) { 433246204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 433346204592SSage Weil return -EFAULT; 433446204592SSage Weil } else { 433546204592SSage Weil transid = 0; /* current trans */ 433646204592SSage Weil } 43372ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 433846204592SSage Weil } 433946204592SSage Weil 4340b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4341475f6387SJan Schmidt { 43420b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4343475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4344b8e95489SMiao Xie int ret; 4345475f6387SJan Schmidt 4346475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4347475f6387SJan Schmidt return -EPERM; 4348475f6387SJan Schmidt 4349475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4350475f6387SJan Schmidt if (IS_ERR(sa)) 4351475f6387SJan Schmidt return PTR_ERR(sa); 4352475f6387SJan Schmidt 4353b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4354b8e95489SMiao Xie ret = mnt_want_write_file(file); 4355b8e95489SMiao Xie if (ret) 4356b8e95489SMiao Xie goto out; 4357b8e95489SMiao Xie } 4358b8e95489SMiao Xie 43590b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 436063a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 436163a212abSStefan Behrens 0); 4362475f6387SJan Schmidt 436306fe39abSFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4364475f6387SJan Schmidt ret = -EFAULT; 4365475f6387SJan Schmidt 4366b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4367b8e95489SMiao Xie mnt_drop_write_file(file); 4368b8e95489SMiao Xie out: 4369475f6387SJan Schmidt kfree(sa); 4370475f6387SJan Schmidt return ret; 4371475f6387SJan Schmidt } 4372475f6387SJan Schmidt 43732ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4374475f6387SJan Schmidt { 4375475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4376475f6387SJan Schmidt return -EPERM; 4377475f6387SJan Schmidt 43782ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4379475f6387SJan Schmidt } 4380475f6387SJan Schmidt 43812ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4382475f6387SJan Schmidt void __user *arg) 4383475f6387SJan Schmidt { 4384475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4385475f6387SJan Schmidt int ret; 4386475f6387SJan Schmidt 4387475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4388475f6387SJan Schmidt return -EPERM; 4389475f6387SJan Schmidt 4390475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4391475f6387SJan Schmidt if (IS_ERR(sa)) 4392475f6387SJan Schmidt return PTR_ERR(sa); 4393475f6387SJan Schmidt 43942ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4395475f6387SJan Schmidt 43964fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4397475f6387SJan Schmidt ret = -EFAULT; 4398475f6387SJan Schmidt 4399475f6387SJan Schmidt kfree(sa); 4400475f6387SJan Schmidt return ret; 4401475f6387SJan Schmidt } 4402475f6387SJan Schmidt 44032ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4404b27f7c0cSDavid Sterba void __user *arg) 4405c11d2c23SStefan Behrens { 4406c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4407c11d2c23SStefan Behrens int ret; 4408c11d2c23SStefan Behrens 4409c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4410c11d2c23SStefan Behrens if (IS_ERR(sa)) 4411c11d2c23SStefan Behrens return PTR_ERR(sa); 4412c11d2c23SStefan Behrens 4413b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4414b27f7c0cSDavid Sterba kfree(sa); 4415b27f7c0cSDavid Sterba return -EPERM; 4416b27f7c0cSDavid Sterba } 4417b27f7c0cSDavid Sterba 44182ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4419c11d2c23SStefan Behrens 4420eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4421c11d2c23SStefan Behrens ret = -EFAULT; 4422c11d2c23SStefan Behrens 4423c11d2c23SStefan Behrens kfree(sa); 4424c11d2c23SStefan Behrens return ret; 4425c11d2c23SStefan Behrens } 4426c11d2c23SStefan Behrens 44272ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 44282ff7e61eSJeff Mahoney void __user *arg) 44293f6bcfbdSStefan Behrens { 44303f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 44313f6bcfbdSStefan Behrens int ret; 44323f6bcfbdSStefan Behrens 44333f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44343f6bcfbdSStefan Behrens return -EPERM; 44353f6bcfbdSStefan Behrens 44363f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44373f6bcfbdSStefan Behrens if (IS_ERR(p)) 44383f6bcfbdSStefan Behrens return PTR_ERR(p); 44393f6bcfbdSStefan Behrens 44403f6bcfbdSStefan Behrens switch (p->cmd) { 44413f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4442bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4443adfa97cbSIlya Dryomov ret = -EROFS; 4444adfa97cbSIlya Dryomov goto out; 4445adfa97cbSIlya Dryomov } 4446171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4447e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44483f6bcfbdSStefan Behrens } else { 44492ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4450171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 44513f6bcfbdSStefan Behrens } 44523f6bcfbdSStefan Behrens break; 44533f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44540b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 44553f6bcfbdSStefan Behrens ret = 0; 44563f6bcfbdSStefan Behrens break; 44573f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 445817d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 445997282031SAnand Jain ret = 0; 44603f6bcfbdSStefan Behrens break; 44613f6bcfbdSStefan Behrens default: 44623f6bcfbdSStefan Behrens ret = -EINVAL; 44633f6bcfbdSStefan Behrens break; 44643f6bcfbdSStefan Behrens } 44653f6bcfbdSStefan Behrens 4466d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 44673f6bcfbdSStefan Behrens ret = -EFAULT; 4468adfa97cbSIlya Dryomov out: 44693f6bcfbdSStefan Behrens kfree(p); 44703f6bcfbdSStefan Behrens return ret; 44713f6bcfbdSStefan Behrens } 44723f6bcfbdSStefan Behrens 4473d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4474d7728c96SJan Schmidt { 4475d7728c96SJan Schmidt int ret = 0; 4476d7728c96SJan Schmidt int i; 4477740c3d22SChris Mason u64 rel_ptr; 4478d7728c96SJan Schmidt int size; 4479806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4480d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4481d7728c96SJan Schmidt struct btrfs_path *path; 4482d7728c96SJan Schmidt 448382b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4484d7728c96SJan Schmidt return -EPERM; 4485d7728c96SJan Schmidt 4486d7728c96SJan Schmidt path = btrfs_alloc_path(); 4487d7728c96SJan Schmidt if (!path) { 4488d7728c96SJan Schmidt ret = -ENOMEM; 4489d7728c96SJan Schmidt goto out; 4490d7728c96SJan Schmidt } 4491d7728c96SJan Schmidt 4492d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4493d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4494d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4495d7728c96SJan Schmidt ipa = NULL; 4496d7728c96SJan Schmidt goto out; 4497d7728c96SJan Schmidt } 4498d7728c96SJan Schmidt 4499d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4500d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4501d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4502d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4503d7728c96SJan Schmidt ipath = NULL; 4504d7728c96SJan Schmidt goto out; 4505d7728c96SJan Schmidt } 4506d7728c96SJan Schmidt 4507d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4508d7728c96SJan Schmidt if (ret < 0) 4509d7728c96SJan Schmidt goto out; 4510d7728c96SJan Schmidt 4511d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4512745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4513745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4514740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4515d7728c96SJan Schmidt } 4516d7728c96SJan Schmidt 4517718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4518718dc5faSOmar Sandoval ipath->fspath, size); 4519d7728c96SJan Schmidt if (ret) { 4520d7728c96SJan Schmidt ret = -EFAULT; 4521d7728c96SJan Schmidt goto out; 4522d7728c96SJan Schmidt } 4523d7728c96SJan Schmidt 4524d7728c96SJan Schmidt out: 4525d7728c96SJan Schmidt btrfs_free_path(path); 4526d7728c96SJan Schmidt free_ipath(ipath); 4527d7728c96SJan Schmidt kfree(ipa); 4528d7728c96SJan Schmidt 4529d7728c96SJan Schmidt return ret; 4530d7728c96SJan Schmidt } 4531d7728c96SJan Schmidt 4532d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4533d7728c96SJan Schmidt { 4534d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4535d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4536d7728c96SJan Schmidt 4537d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4538d7728c96SJan Schmidt inodes->bytes_left -= c; 4539d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4540d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4541d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4542d7728c96SJan Schmidt inodes->elem_cnt += 3; 4543d7728c96SJan Schmidt } else { 4544d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4545d7728c96SJan Schmidt inodes->bytes_left = 0; 4546d7728c96SJan Schmidt inodes->elem_missed += 3; 4547d7728c96SJan Schmidt } 4548d7728c96SJan Schmidt 4549d7728c96SJan Schmidt return 0; 4550d7728c96SJan Schmidt } 4551d7728c96SJan Schmidt 45522ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4553d24a67b2SZygo Blaxell void __user *arg, int version) 4554d7728c96SJan Schmidt { 4555d7728c96SJan Schmidt int ret = 0; 4556d7728c96SJan Schmidt int size; 4557d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4558d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4559d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4560d24a67b2SZygo Blaxell bool ignore_offset; 4561d7728c96SJan Schmidt 4562d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4563d7728c96SJan Schmidt return -EPERM; 4564d7728c96SJan Schmidt 4565d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 45667b9ea627SShailendra Verma if (IS_ERR(loi)) 45677b9ea627SShailendra Verma return PTR_ERR(loi); 4568d7728c96SJan Schmidt 4569d24a67b2SZygo Blaxell if (version == 1) { 4570d24a67b2SZygo Blaxell ignore_offset = false; 4571b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4572d24a67b2SZygo Blaxell } else { 4573d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4574d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4575d24a67b2SZygo Blaxell ret = -EINVAL; 4576d24a67b2SZygo Blaxell goto out_loi; 4577d24a67b2SZygo Blaxell } 4578d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4579d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4580d24a67b2SZygo Blaxell ret = -EINVAL; 4581d24a67b2SZygo Blaxell goto out_loi; 4582d24a67b2SZygo Blaxell } 4583d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4584b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4585d24a67b2SZygo Blaxell } 4586d24a67b2SZygo Blaxell 4587d7728c96SJan Schmidt path = btrfs_alloc_path(); 4588d7728c96SJan Schmidt if (!path) { 4589d7728c96SJan Schmidt ret = -ENOMEM; 4590d7728c96SJan Schmidt goto out; 4591d7728c96SJan Schmidt } 4592d7728c96SJan Schmidt 4593d7728c96SJan Schmidt inodes = init_data_container(size); 4594d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4595d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4596d7728c96SJan Schmidt inodes = NULL; 4597d7728c96SJan Schmidt goto out; 4598d7728c96SJan Schmidt } 4599d7728c96SJan Schmidt 46002ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4601d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4602df031f07SLiu Bo if (ret == -EINVAL) 4603d7728c96SJan Schmidt ret = -ENOENT; 4604d7728c96SJan Schmidt if (ret < 0) 4605d7728c96SJan Schmidt goto out; 4606d7728c96SJan Schmidt 4607718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4608718dc5faSOmar Sandoval size); 4609d7728c96SJan Schmidt if (ret) 4610d7728c96SJan Schmidt ret = -EFAULT; 4611d7728c96SJan Schmidt 4612d7728c96SJan Schmidt out: 4613d7728c96SJan Schmidt btrfs_free_path(path); 4614f54de068SDavid Sterba kvfree(inodes); 4615d24a67b2SZygo Blaxell out_loi: 4616d7728c96SJan Schmidt kfree(loi); 4617d7728c96SJan Schmidt 4618d7728c96SJan Schmidt return ret; 4619d7728c96SJan Schmidt } 4620d7728c96SJan Schmidt 4621008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4622c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4623c9e9f97bSIlya Dryomov { 4624c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4625c9e9f97bSIlya Dryomov 4626c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4627c9e9f97bSIlya Dryomov 46283009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4629837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4630837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4631837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4632a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4633a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4634837d5b6eSIlya Dryomov 4635c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4636c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4637c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 463819a39dceSIlya Dryomov 463919a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 464019a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 464119a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4642c9e9f97bSIlya Dryomov } 4643c9e9f97bSIlya Dryomov 46449ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4645c9e9f97bSIlya Dryomov { 4646496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4647c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4648c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4649c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4650ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4651c9e9f97bSIlya Dryomov int ret; 4652c9e9f97bSIlya Dryomov 4653c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4654c9e9f97bSIlya Dryomov return -EPERM; 4655c9e9f97bSIlya Dryomov 4656e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46579ba1f6e4SLiu Bo if (ret) 46589ba1f6e4SLiu Bo return ret; 46599ba1f6e4SLiu Bo 4660ed0fb78fSIlya Dryomov again: 4661171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4662c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4663ed0fb78fSIlya Dryomov need_unlock = true; 4664ed0fb78fSIlya Dryomov goto locked; 4665ed0fb78fSIlya Dryomov } 4666ed0fb78fSIlya Dryomov 4667ed0fb78fSIlya Dryomov /* 466801327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4669ed0fb78fSIlya Dryomov * (1) some other op is running 4670ed0fb78fSIlya Dryomov * (2) balance is running 4671ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4672ed0fb78fSIlya Dryomov */ 4673ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4674ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4675ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 46763009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4677ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4678dccdb07bSDavid Sterba /* 4679dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4680dccdb07bSDavid Sterba * we'll reexamine the status again. 4681dccdb07bSDavid Sterba */ 4682ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4683ed0fb78fSIlya Dryomov 4684ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 46853009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4686ed0fb78fSIlya Dryomov /* this is (3) */ 4687ed0fb78fSIlya Dryomov need_unlock = false; 4688ed0fb78fSIlya Dryomov goto locked; 4689ed0fb78fSIlya Dryomov } 4690ed0fb78fSIlya Dryomov 4691ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4692ed0fb78fSIlya Dryomov goto again; 4693ed0fb78fSIlya Dryomov } else { 4694ed0fb78fSIlya Dryomov /* this is (2) */ 4695ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4696ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4697ed0fb78fSIlya Dryomov goto out; 4698ed0fb78fSIlya Dryomov } 4699ed0fb78fSIlya Dryomov } else { 4700ed0fb78fSIlya Dryomov /* this is (1) */ 4701ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4702e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4703ed0fb78fSIlya Dryomov goto out; 4704ed0fb78fSIlya Dryomov } 4705ed0fb78fSIlya Dryomov 4706ed0fb78fSIlya Dryomov locked: 4707171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4708c9e9f97bSIlya Dryomov 4709c9e9f97bSIlya Dryomov if (arg) { 4710c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4711c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4712c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4713ed0fb78fSIlya Dryomov goto out_unlock; 4714c9e9f97bSIlya Dryomov } 4715de322263SIlya Dryomov 4716de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4717de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4718de322263SIlya Dryomov ret = -ENOTCONN; 4719de322263SIlya Dryomov goto out_bargs; 4720de322263SIlya Dryomov } 4721de322263SIlya Dryomov 4722de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4723de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4724de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4725de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4726de322263SIlya Dryomov 4727de322263SIlya Dryomov goto do_balance; 4728de322263SIlya Dryomov } 4729c9e9f97bSIlya Dryomov } else { 4730c9e9f97bSIlya Dryomov bargs = NULL; 4731c9e9f97bSIlya Dryomov } 4732c9e9f97bSIlya Dryomov 4733ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4734837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4735837d5b6eSIlya Dryomov goto out_bargs; 4736837d5b6eSIlya Dryomov } 4737837d5b6eSIlya Dryomov 47388d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4739c9e9f97bSIlya Dryomov if (!bctl) { 4740c9e9f97bSIlya Dryomov ret = -ENOMEM; 4741c9e9f97bSIlya Dryomov goto out_bargs; 4742c9e9f97bSIlya Dryomov } 4743c9e9f97bSIlya Dryomov 4744c9e9f97bSIlya Dryomov if (arg) { 4745c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4746c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4747c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4748c9e9f97bSIlya Dryomov 4749c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4750f43ffb60SIlya Dryomov } else { 4751f43ffb60SIlya Dryomov /* balance everything - no filters */ 4752f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4753c9e9f97bSIlya Dryomov } 4754c9e9f97bSIlya Dryomov 47558eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47568eb93459SDavid Sterba ret = -EINVAL; 47570f89abf5SChristian Engelmayer goto out_bctl; 47588eb93459SDavid Sterba } 47598eb93459SDavid Sterba 4760de322263SIlya Dryomov do_balance: 4761c9e9f97bSIlya Dryomov /* 4762149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4763149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4764149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4765149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4766c9e9f97bSIlya Dryomov */ 4767ed0fb78fSIlya Dryomov need_unlock = false; 4768ed0fb78fSIlya Dryomov 47696fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 47700f89abf5SChristian Engelmayer bctl = NULL; 4771ed0fb78fSIlya Dryomov 4772d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4773c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4774c9e9f97bSIlya Dryomov ret = -EFAULT; 4775c9e9f97bSIlya Dryomov } 4776c9e9f97bSIlya Dryomov 47770f89abf5SChristian Engelmayer out_bctl: 47780f89abf5SChristian Engelmayer kfree(bctl); 4779c9e9f97bSIlya Dryomov out_bargs: 4780c9e9f97bSIlya Dryomov kfree(bargs); 4781ed0fb78fSIlya Dryomov out_unlock: 4782c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4783ed0fb78fSIlya Dryomov if (need_unlock) 4784171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4785ed0fb78fSIlya Dryomov out: 4786e54bfa31SLiu Bo mnt_drop_write_file(file); 4787c9e9f97bSIlya Dryomov return ret; 4788c9e9f97bSIlya Dryomov } 4789c9e9f97bSIlya Dryomov 47902ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4791837d5b6eSIlya Dryomov { 4792837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4793837d5b6eSIlya Dryomov return -EPERM; 4794837d5b6eSIlya Dryomov 4795837d5b6eSIlya Dryomov switch (cmd) { 4796837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 47970b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4798a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47990b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4800837d5b6eSIlya Dryomov } 4801837d5b6eSIlya Dryomov 4802837d5b6eSIlya Dryomov return -EINVAL; 4803837d5b6eSIlya Dryomov } 4804837d5b6eSIlya Dryomov 48052ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 480619a39dceSIlya Dryomov void __user *arg) 480719a39dceSIlya Dryomov { 480819a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 480919a39dceSIlya Dryomov int ret = 0; 481019a39dceSIlya Dryomov 481119a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 481219a39dceSIlya Dryomov return -EPERM; 481319a39dceSIlya Dryomov 481419a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 481519a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 481619a39dceSIlya Dryomov ret = -ENOTCONN; 481719a39dceSIlya Dryomov goto out; 481819a39dceSIlya Dryomov } 481919a39dceSIlya Dryomov 48208d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 482119a39dceSIlya Dryomov if (!bargs) { 482219a39dceSIlya Dryomov ret = -ENOMEM; 482319a39dceSIlya Dryomov goto out; 482419a39dceSIlya Dryomov } 482519a39dceSIlya Dryomov 4826008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 482719a39dceSIlya Dryomov 482819a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 482919a39dceSIlya Dryomov ret = -EFAULT; 483019a39dceSIlya Dryomov 483119a39dceSIlya Dryomov kfree(bargs); 483219a39dceSIlya Dryomov out: 483319a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 483419a39dceSIlya Dryomov return ret; 483519a39dceSIlya Dryomov } 483619a39dceSIlya Dryomov 4837905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 48385d13a37bSArne Jansen { 48390b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48400b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48415d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 48425d13a37bSArne Jansen int ret; 48435d13a37bSArne Jansen 48445d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48455d13a37bSArne Jansen return -EPERM; 48465d13a37bSArne Jansen 4847905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4848905b0ddaSMiao Xie if (ret) 4849905b0ddaSMiao Xie return ret; 48505d13a37bSArne Jansen 48515d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4852905b0ddaSMiao Xie if (IS_ERR(sa)) { 4853905b0ddaSMiao Xie ret = PTR_ERR(sa); 4854905b0ddaSMiao Xie goto drop_write; 4855905b0ddaSMiao Xie } 48565d13a37bSArne Jansen 48570b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48585d13a37bSArne Jansen 48595d13a37bSArne Jansen switch (sa->cmd) { 48605d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4861340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 48625d13a37bSArne Jansen break; 48635d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4864340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 48655d13a37bSArne Jansen break; 48665d13a37bSArne Jansen default: 48675d13a37bSArne Jansen ret = -EINVAL; 48685d13a37bSArne Jansen break; 48695d13a37bSArne Jansen } 48705d13a37bSArne Jansen 48715d13a37bSArne Jansen kfree(sa); 48720b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4873905b0ddaSMiao Xie drop_write: 4874905b0ddaSMiao Xie mnt_drop_write_file(file); 48755d13a37bSArne Jansen return ret; 48765d13a37bSArne Jansen } 48775d13a37bSArne Jansen 4878905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48795d13a37bSArne Jansen { 48800b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48810b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48820b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48835d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48845d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48855d13a37bSArne Jansen int ret; 48865d13a37bSArne Jansen int err; 48875d13a37bSArne Jansen 48885d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48895d13a37bSArne Jansen return -EPERM; 48905d13a37bSArne Jansen 4891905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4892905b0ddaSMiao Xie if (ret) 4893905b0ddaSMiao Xie return ret; 48945d13a37bSArne Jansen 48955d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4896905b0ddaSMiao Xie if (IS_ERR(sa)) { 4897905b0ddaSMiao Xie ret = PTR_ERR(sa); 4898905b0ddaSMiao Xie goto drop_write; 4899905b0ddaSMiao Xie } 49005d13a37bSArne Jansen 49015d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49025d13a37bSArne Jansen if (IS_ERR(trans)) { 49035d13a37bSArne Jansen ret = PTR_ERR(trans); 49045d13a37bSArne Jansen goto out; 49055d13a37bSArne Jansen } 49065d13a37bSArne Jansen 49075d13a37bSArne Jansen if (sa->assign) { 49089f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 49095d13a37bSArne Jansen } else { 491039616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 49115d13a37bSArne Jansen } 49125d13a37bSArne Jansen 4913e082f563SQu Wenruo /* update qgroup status and info */ 4914280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4915e082f563SQu Wenruo if (err < 0) 49160b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4917ad8403dfSAnand Jain "failed to update qgroup status and info"); 49183a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49195d13a37bSArne Jansen if (err && !ret) 49205d13a37bSArne Jansen ret = err; 49215d13a37bSArne Jansen 49225d13a37bSArne Jansen out: 49235d13a37bSArne Jansen kfree(sa); 4924905b0ddaSMiao Xie drop_write: 4925905b0ddaSMiao Xie mnt_drop_write_file(file); 49265d13a37bSArne Jansen return ret; 49275d13a37bSArne Jansen } 49285d13a37bSArne Jansen 4929905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 49305d13a37bSArne Jansen { 49310b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49320b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49335d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49345d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49355d13a37bSArne Jansen int ret; 49365d13a37bSArne Jansen int err; 49375d13a37bSArne Jansen 49385d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49395d13a37bSArne Jansen return -EPERM; 49405d13a37bSArne Jansen 4941905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4942905b0ddaSMiao Xie if (ret) 4943905b0ddaSMiao Xie return ret; 49445d13a37bSArne Jansen 49455d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4946905b0ddaSMiao Xie if (IS_ERR(sa)) { 4947905b0ddaSMiao Xie ret = PTR_ERR(sa); 4948905b0ddaSMiao Xie goto drop_write; 4949905b0ddaSMiao Xie } 49505d13a37bSArne Jansen 4951d86e56cfSMiao Xie if (!sa->qgroupid) { 4952d86e56cfSMiao Xie ret = -EINVAL; 4953d86e56cfSMiao Xie goto out; 4954d86e56cfSMiao Xie } 4955d86e56cfSMiao Xie 49565d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49575d13a37bSArne Jansen if (IS_ERR(trans)) { 49585d13a37bSArne Jansen ret = PTR_ERR(trans); 49595d13a37bSArne Jansen goto out; 49605d13a37bSArne Jansen } 49615d13a37bSArne Jansen 49625d13a37bSArne Jansen if (sa->create) { 496349a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 49645d13a37bSArne Jansen } else { 49653efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 49665d13a37bSArne Jansen } 49675d13a37bSArne Jansen 49683a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49695d13a37bSArne Jansen if (err && !ret) 49705d13a37bSArne Jansen ret = err; 49715d13a37bSArne Jansen 49725d13a37bSArne Jansen out: 49735d13a37bSArne Jansen kfree(sa); 4974905b0ddaSMiao Xie drop_write: 4975905b0ddaSMiao Xie mnt_drop_write_file(file); 49765d13a37bSArne Jansen return ret; 49775d13a37bSArne Jansen } 49785d13a37bSArne Jansen 4979905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49805d13a37bSArne Jansen { 49810b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49820b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49835d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49845d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49855d13a37bSArne Jansen int ret; 49865d13a37bSArne Jansen int err; 49875d13a37bSArne Jansen u64 qgroupid; 49885d13a37bSArne Jansen 49895d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49905d13a37bSArne Jansen return -EPERM; 49915d13a37bSArne Jansen 4992905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4993905b0ddaSMiao Xie if (ret) 4994905b0ddaSMiao Xie return ret; 49955d13a37bSArne Jansen 49965d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4997905b0ddaSMiao Xie if (IS_ERR(sa)) { 4998905b0ddaSMiao Xie ret = PTR_ERR(sa); 4999905b0ddaSMiao Xie goto drop_write; 5000905b0ddaSMiao Xie } 50015d13a37bSArne Jansen 50025d13a37bSArne Jansen trans = btrfs_join_transaction(root); 50035d13a37bSArne Jansen if (IS_ERR(trans)) { 50045d13a37bSArne Jansen ret = PTR_ERR(trans); 50055d13a37bSArne Jansen goto out; 50065d13a37bSArne Jansen } 50075d13a37bSArne Jansen 50085d13a37bSArne Jansen qgroupid = sa->qgroupid; 50095d13a37bSArne Jansen if (!qgroupid) { 50105d13a37bSArne Jansen /* take the current subvol as qgroup */ 50115d13a37bSArne Jansen qgroupid = root->root_key.objectid; 50125d13a37bSArne Jansen } 50135d13a37bSArne Jansen 5014f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 50155d13a37bSArne Jansen 50163a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 50175d13a37bSArne Jansen if (err && !ret) 50185d13a37bSArne Jansen ret = err; 50195d13a37bSArne Jansen 50205d13a37bSArne Jansen out: 50215d13a37bSArne Jansen kfree(sa); 5022905b0ddaSMiao Xie drop_write: 5023905b0ddaSMiao Xie mnt_drop_write_file(file); 50245d13a37bSArne Jansen return ret; 50255d13a37bSArne Jansen } 50265d13a37bSArne Jansen 50272f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 50282f232036SJan Schmidt { 50290b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50300b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50312f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50322f232036SJan Schmidt int ret; 50332f232036SJan Schmidt 50342f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50352f232036SJan Schmidt return -EPERM; 50362f232036SJan Schmidt 50372f232036SJan Schmidt ret = mnt_want_write_file(file); 50382f232036SJan Schmidt if (ret) 50392f232036SJan Schmidt return ret; 50402f232036SJan Schmidt 50412f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50422f232036SJan Schmidt if (IS_ERR(qsa)) { 50432f232036SJan Schmidt ret = PTR_ERR(qsa); 50442f232036SJan Schmidt goto drop_write; 50452f232036SJan Schmidt } 50462f232036SJan Schmidt 50472f232036SJan Schmidt if (qsa->flags) { 50482f232036SJan Schmidt ret = -EINVAL; 50492f232036SJan Schmidt goto out; 50502f232036SJan Schmidt } 50512f232036SJan Schmidt 50520b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 50532f232036SJan Schmidt 50542f232036SJan Schmidt out: 50552f232036SJan Schmidt kfree(qsa); 50562f232036SJan Schmidt drop_write: 50572f232036SJan Schmidt mnt_drop_write_file(file); 50582f232036SJan Schmidt return ret; 50592f232036SJan Schmidt } 50602f232036SJan Schmidt 50612f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50622f232036SJan Schmidt { 50630b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50640b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50652f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50662f232036SJan Schmidt int ret = 0; 50672f232036SJan Schmidt 50682f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50692f232036SJan Schmidt return -EPERM; 50702f232036SJan Schmidt 50718d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 50722f232036SJan Schmidt if (!qsa) 50732f232036SJan Schmidt return -ENOMEM; 50742f232036SJan Schmidt 50750b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50762f232036SJan Schmidt qsa->flags = 1; 50770b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 50782f232036SJan Schmidt } 50792f232036SJan Schmidt 50802f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50812f232036SJan Schmidt ret = -EFAULT; 50822f232036SJan Schmidt 50832f232036SJan Schmidt kfree(qsa); 50842f232036SJan Schmidt return ret; 50852f232036SJan Schmidt } 50862f232036SJan Schmidt 508757254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 508857254b6eSJan Schmidt { 50890b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50900b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 509157254b6eSJan Schmidt 509257254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 509357254b6eSJan Schmidt return -EPERM; 509457254b6eSJan Schmidt 50950b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 509657254b6eSJan Schmidt } 509757254b6eSJan Schmidt 5098abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5099abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 51008ea05e3aSAlexander Block { 5101496ad9aaSAl Viro struct inode *inode = file_inode(file); 51020b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51038ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 51048ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 51058ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 510695582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 51078ea05e3aSAlexander Block int ret = 0; 5108dd5f9615SStefan Behrens int received_uuid_changed; 51098ea05e3aSAlexander Block 5110bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5111bd60ea0fSDavid Sterba return -EPERM; 5112bd60ea0fSDavid Sterba 51138ea05e3aSAlexander Block ret = mnt_want_write_file(file); 51148ea05e3aSAlexander Block if (ret < 0) 51158ea05e3aSAlexander Block return ret; 51168ea05e3aSAlexander Block 51170b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51188ea05e3aSAlexander Block 51194a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 51208ea05e3aSAlexander Block ret = -EINVAL; 51218ea05e3aSAlexander Block goto out; 51228ea05e3aSAlexander Block } 51238ea05e3aSAlexander Block 51248ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 51258ea05e3aSAlexander Block ret = -EROFS; 51268ea05e3aSAlexander Block goto out; 51278ea05e3aSAlexander Block } 51288ea05e3aSAlexander Block 5129dd5f9615SStefan Behrens /* 5130dd5f9615SStefan Behrens * 1 - root item 5131dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5132dd5f9615SStefan Behrens */ 5133dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51348ea05e3aSAlexander Block if (IS_ERR(trans)) { 51358ea05e3aSAlexander Block ret = PTR_ERR(trans); 51368ea05e3aSAlexander Block trans = NULL; 51378ea05e3aSAlexander Block goto out; 51388ea05e3aSAlexander Block } 51398ea05e3aSAlexander Block 51408ea05e3aSAlexander Block sa->rtransid = trans->transid; 51418ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51428ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51438ea05e3aSAlexander Block 5144dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5145dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5146dd5f9615SStefan Behrens if (received_uuid_changed && 5147d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5148d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5149dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5150dd5f9615SStefan Behrens root->root_key.objectid); 5151d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5152d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5153d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5154d87ff758SNikolay Borisov goto out; 5155d87ff758SNikolay Borisov } 5156d87ff758SNikolay Borisov } 51578ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51588ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51598ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51603cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51613cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51623cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51633cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51648ea05e3aSAlexander Block 51650b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 51668ea05e3aSAlexander Block &root->root_key, &root->root_item); 51678ea05e3aSAlexander Block if (ret < 0) { 51683a45bb20SJeff Mahoney btrfs_end_transaction(trans); 51698ea05e3aSAlexander Block goto out; 5170dd5f9615SStefan Behrens } 5171dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5172cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5173dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5174dd5f9615SStefan Behrens root->root_key.objectid); 5175dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 517666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5177efd38150SNikolay Borisov btrfs_end_transaction(trans); 5178dd5f9615SStefan Behrens goto out; 5179dd5f9615SStefan Behrens } 5180dd5f9615SStefan Behrens } 51813a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5182abccd00fSHugo Mills out: 51830b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5184abccd00fSHugo Mills mnt_drop_write_file(file); 5185abccd00fSHugo Mills return ret; 5186abccd00fSHugo Mills } 5187abccd00fSHugo Mills 5188abccd00fSHugo Mills #ifdef CONFIG_64BIT 5189abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5190abccd00fSHugo Mills void __user *arg) 5191abccd00fSHugo Mills { 5192abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5193abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5194abccd00fSHugo Mills int ret = 0; 5195abccd00fSHugo Mills 5196abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 51977b9ea627SShailendra Verma if (IS_ERR(args32)) 51987b9ea627SShailendra Verma return PTR_ERR(args32); 5199abccd00fSHugo Mills 52008d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 520184dbeb87SDan Carpenter if (!args64) { 520284dbeb87SDan Carpenter ret = -ENOMEM; 5203abccd00fSHugo Mills goto out; 5204abccd00fSHugo Mills } 5205abccd00fSHugo Mills 5206abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5207abccd00fSHugo Mills args64->stransid = args32->stransid; 5208abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5209abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5210abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5211abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5212abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5213abccd00fSHugo Mills args64->flags = args32->flags; 5214abccd00fSHugo Mills 5215abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5216abccd00fSHugo Mills if (ret) 5217abccd00fSHugo Mills goto out; 5218abccd00fSHugo Mills 5219abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5220abccd00fSHugo Mills args32->stransid = args64->stransid; 5221abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5222abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5223abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5224abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5225abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5226abccd00fSHugo Mills args32->flags = args64->flags; 5227abccd00fSHugo Mills 5228abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5229abccd00fSHugo Mills if (ret) 5230abccd00fSHugo Mills ret = -EFAULT; 5231abccd00fSHugo Mills 5232abccd00fSHugo Mills out: 5233abccd00fSHugo Mills kfree(args32); 5234abccd00fSHugo Mills kfree(args64); 5235abccd00fSHugo Mills return ret; 5236abccd00fSHugo Mills } 5237abccd00fSHugo Mills #endif 5238abccd00fSHugo Mills 5239abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5240abccd00fSHugo Mills void __user *arg) 5241abccd00fSHugo Mills { 5242abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5243abccd00fSHugo Mills int ret = 0; 5244abccd00fSHugo Mills 5245abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 52467b9ea627SShailendra Verma if (IS_ERR(sa)) 52477b9ea627SShailendra Verma return PTR_ERR(sa); 5248abccd00fSHugo Mills 5249abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5250abccd00fSHugo Mills 5251abccd00fSHugo Mills if (ret) 5252abccd00fSHugo Mills goto out; 5253abccd00fSHugo Mills 52548ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52558ea05e3aSAlexander Block if (ret) 52568ea05e3aSAlexander Block ret = -EFAULT; 52578ea05e3aSAlexander Block 52588ea05e3aSAlexander Block out: 52598ea05e3aSAlexander Block kfree(sa); 52608ea05e3aSAlexander Block return ret; 52618ea05e3aSAlexander Block } 52628ea05e3aSAlexander Block 5263867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5264867ab667Sjeff.liu { 52650b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52660b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5267a1b83ac5SAnand Jain size_t len; 5268867ab667Sjeff.liu int ret; 5269a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5270a1b83ac5SAnand Jain 52710b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 52720b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 52730b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5274a1b83ac5SAnand Jain 5275a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5276867ab667Sjeff.liu 5277867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 52780b246afaSJeff Mahoney btrfs_warn(fs_info, 52790b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 52800b246afaSJeff Mahoney --len); 5281867ab667Sjeff.liu } 5282867ab667Sjeff.liu 5283867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5284867ab667Sjeff.liu 5285867ab667Sjeff.liu return ret ? -EFAULT : 0; 5286867ab667Sjeff.liu } 5287867ab667Sjeff.liu 5288a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5289a8bfd4abSjeff.liu { 52900b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52910b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52920b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52930b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5294a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5295a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5296a8bfd4abSjeff.liu int ret; 5297a8bfd4abSjeff.liu 5298a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5299a8bfd4abSjeff.liu return -EPERM; 5300a8bfd4abSjeff.liu 5301a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5302a8bfd4abSjeff.liu return -EFAULT; 5303a8bfd4abSjeff.liu 5304a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 53050b246afaSJeff Mahoney btrfs_err(fs_info, 53065d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5307a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5308a8bfd4abSjeff.liu return -EINVAL; 5309a8bfd4abSjeff.liu } 5310a8bfd4abSjeff.liu 5311a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5312a8bfd4abSjeff.liu if (ret) 5313a8bfd4abSjeff.liu return ret; 5314a8bfd4abSjeff.liu 5315a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5316a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5317a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5318a8bfd4abSjeff.liu goto out_unlock; 5319a8bfd4abSjeff.liu } 5320a8bfd4abSjeff.liu 53210b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5322a8bfd4abSjeff.liu strcpy(super_block->label, label); 53230b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53243a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5325a8bfd4abSjeff.liu 5326a8bfd4abSjeff.liu out_unlock: 5327a8bfd4abSjeff.liu mnt_drop_write_file(file); 5328a8bfd4abSjeff.liu return ret; 5329a8bfd4abSjeff.liu } 5330a8bfd4abSjeff.liu 53312eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 53322eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 53332eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53342eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53352eaa055fSJeff Mahoney 5336d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53372eaa055fSJeff Mahoney { 53384d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53392eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53402eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53412eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53422eaa055fSJeff Mahoney }; 53432eaa055fSJeff Mahoney 53442eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53452eaa055fSJeff Mahoney return -EFAULT; 53462eaa055fSJeff Mahoney 53472eaa055fSJeff Mahoney return 0; 53482eaa055fSJeff Mahoney } 53492eaa055fSJeff Mahoney 53502eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53512eaa055fSJeff Mahoney { 53520b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53530b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53540b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53552eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53562eaa055fSJeff Mahoney 53572eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53582eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53592eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53602eaa055fSJeff Mahoney 53612eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53622eaa055fSJeff Mahoney return -EFAULT; 53632eaa055fSJeff Mahoney 53642eaa055fSJeff Mahoney return 0; 53652eaa055fSJeff Mahoney } 53662eaa055fSJeff Mahoney 53672ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 53683b02a68aSJeff Mahoney enum btrfs_feature_set set, 53692eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53702eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53712eaa055fSJeff Mahoney { 53723b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 53733b02a68aSJeff Mahoney char *names; 53742eaa055fSJeff Mahoney u64 disallowed, unsupported; 53752eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53762eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53772eaa055fSJeff Mahoney 53782eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53792eaa055fSJeff Mahoney if (unsupported) { 53803b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53813b02a68aSJeff Mahoney if (names) { 53820b246afaSJeff Mahoney btrfs_warn(fs_info, 53833b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53843b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53853b02a68aSJeff Mahoney kfree(names); 53863b02a68aSJeff Mahoney } else 53870b246afaSJeff Mahoney btrfs_warn(fs_info, 53882eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53892eaa055fSJeff Mahoney type, unsupported); 53902eaa055fSJeff Mahoney return -EOPNOTSUPP; 53912eaa055fSJeff Mahoney } 53922eaa055fSJeff Mahoney 53932eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53942eaa055fSJeff Mahoney if (disallowed) { 53953b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53963b02a68aSJeff Mahoney if (names) { 53970b246afaSJeff Mahoney btrfs_warn(fs_info, 53983b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53993b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54003b02a68aSJeff Mahoney kfree(names); 54013b02a68aSJeff Mahoney } else 54020b246afaSJeff Mahoney btrfs_warn(fs_info, 54032eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 54042eaa055fSJeff Mahoney type, disallowed); 54052eaa055fSJeff Mahoney return -EPERM; 54062eaa055fSJeff Mahoney } 54072eaa055fSJeff Mahoney 54082eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 54092eaa055fSJeff Mahoney if (disallowed) { 54103b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54113b02a68aSJeff Mahoney if (names) { 54120b246afaSJeff Mahoney btrfs_warn(fs_info, 54133b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 54143b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54153b02a68aSJeff Mahoney kfree(names); 54163b02a68aSJeff Mahoney } else 54170b246afaSJeff Mahoney btrfs_warn(fs_info, 54182eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 54192eaa055fSJeff Mahoney type, disallowed); 54202eaa055fSJeff Mahoney return -EPERM; 54212eaa055fSJeff Mahoney } 54222eaa055fSJeff Mahoney 54232eaa055fSJeff Mahoney return 0; 54242eaa055fSJeff Mahoney } 54252eaa055fSJeff Mahoney 54262ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 54272ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 54282eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 54292eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 54302eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 54312eaa055fSJeff Mahoney 54322eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 54332eaa055fSJeff Mahoney { 54340b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54350b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54360b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54370b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 54382eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54392eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54402eaa055fSJeff Mahoney u64 newflags; 54412eaa055fSJeff Mahoney int ret; 54422eaa055fSJeff Mahoney 54432eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54442eaa055fSJeff Mahoney return -EPERM; 54452eaa055fSJeff Mahoney 54462eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54472eaa055fSJeff Mahoney return -EFAULT; 54482eaa055fSJeff Mahoney 54492eaa055fSJeff Mahoney /* Nothing to do */ 54502eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54512eaa055fSJeff Mahoney !flags[0].incompat_flags) 54522eaa055fSJeff Mahoney return 0; 54532eaa055fSJeff Mahoney 54542ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 54552eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54562eaa055fSJeff Mahoney if (ret) 54572eaa055fSJeff Mahoney return ret; 54582eaa055fSJeff Mahoney 54592ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 54602eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54612eaa055fSJeff Mahoney if (ret) 54622eaa055fSJeff Mahoney return ret; 54632eaa055fSJeff Mahoney 54642ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 54652eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54662eaa055fSJeff Mahoney if (ret) 54672eaa055fSJeff Mahoney return ret; 54682eaa055fSJeff Mahoney 54697ab19625SDavid Sterba ret = mnt_want_write_file(file); 54707ab19625SDavid Sterba if (ret) 54717ab19625SDavid Sterba return ret; 54727ab19625SDavid Sterba 54738051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54747ab19625SDavid Sterba if (IS_ERR(trans)) { 54757ab19625SDavid Sterba ret = PTR_ERR(trans); 54767ab19625SDavid Sterba goto out_drop_write; 54777ab19625SDavid Sterba } 54782eaa055fSJeff Mahoney 54790b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 54802eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54812eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54822eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54832eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54842eaa055fSJeff Mahoney 54852eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54862eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54872eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54882eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54892eaa055fSJeff Mahoney 54902eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54912eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54922eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54932eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54940b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 54952eaa055fSJeff Mahoney 54963a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 54977ab19625SDavid Sterba out_drop_write: 54987ab19625SDavid Sterba mnt_drop_write_file(file); 54997ab19625SDavid Sterba 55007ab19625SDavid Sterba return ret; 55012eaa055fSJeff Mahoney } 55022eaa055fSJeff Mahoney 55032351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 55042351f431SJosef Bacik { 55052351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 55062351f431SJosef Bacik int ret; 55072351f431SJosef Bacik 55082351f431SJosef Bacik if (compat) { 55092351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55102351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 55112351f431SJosef Bacik 55122351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 55132351f431SJosef Bacik if (ret) 55142351f431SJosef Bacik return -EFAULT; 55152351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 55162351f431SJosef Bacik if (!arg) 55172351f431SJosef Bacik return -ENOMEM; 55182351f431SJosef Bacik arg->send_fd = args32.send_fd; 55192351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 55202351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 55212351f431SJosef Bacik arg->parent_root = args32.parent_root; 55222351f431SJosef Bacik arg->flags = args32.flags; 55232351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 55242351f431SJosef Bacik sizeof(args32.reserved)); 55252351f431SJosef Bacik #else 55262351f431SJosef Bacik return -ENOTTY; 55272351f431SJosef Bacik #endif 55282351f431SJosef Bacik } else { 55292351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 55302351f431SJosef Bacik if (IS_ERR(arg)) 55312351f431SJosef Bacik return PTR_ERR(arg); 55322351f431SJosef Bacik } 55332351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 55342351f431SJosef Bacik kfree(arg); 55352351f431SJosef Bacik return ret; 55362351f431SJosef Bacik } 55372351f431SJosef Bacik 5538f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5539f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5540f46b5a66SChristoph Hellwig { 55410b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55420b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55430b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55444bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5545f46b5a66SChristoph Hellwig 5546f46b5a66SChristoph Hellwig switch (cmd) { 55476cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 55486cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 55496cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 55506cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 55516cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 55526cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5553f7039b1dSLi Dongyang case FITRIM: 5554f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5555f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5556fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5557fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5558fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 55593de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5560fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 55616f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 55626f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 556376dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 556476dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 55650caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 55660caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 55670caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 55680caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55696ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55706ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5571f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55721e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55731e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55741e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5575f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5576198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5577f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55782ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5579f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5580da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55816b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55826b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5583475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 55842ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5585475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 55862ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5587f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55889ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5589ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5590ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5591cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5592cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5593ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5594ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5595d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5596d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5597d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5598d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5599d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5600d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 56011406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 56022ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 56039b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 56049b199859SFilipe David Borba Manana int ret; 56059b199859SFilipe David Borba Manana 560682b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 56079b199859SFilipe David Borba Manana if (ret) 56089b199859SFilipe David Borba Manana return ret; 56090b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 56102fad4e83SDavid Sterba /* 56112fad4e83SDavid Sterba * The transaction thread may want to do more work, 561201327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 56132fad4e83SDavid Sterba * processing uncleaned subvols. 56142fad4e83SDavid Sterba */ 56150b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 56169b199859SFilipe David Borba Manana return ret; 56179b199859SFilipe David Borba Manana } 561846204592SSage Weil case BTRFS_IOC_START_SYNC: 56199a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 562046204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 56212ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5622475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5623b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5624475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 56252ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5626475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 56272ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5628c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 56299ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5630837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 56312ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 563219a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 56332ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 56348ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 56358ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5636abccd00fSHugo Mills #ifdef CONFIG_64BIT 5637abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5638abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5639abccd00fSHugo Mills #endif 564031db9f7cSAlexander Block case BTRFS_IOC_SEND: 56412351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 56422351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 56432351f431SJosef Bacik case BTRFS_IOC_SEND_32: 56442351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 56452351f431SJosef Bacik #endif 5646c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 56472ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 56485d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5649905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 56505d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5651905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 56525d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5653905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 56545d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5655905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 56562f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 56572f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 56582f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 56592f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 566057254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 566157254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 56623f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 56632ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5664867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5665867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5666a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5667a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 56682eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5669d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 56702eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 56712eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 56722eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56732eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5674e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5675e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5676025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5677025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5678b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5679b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 568042e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 568142e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 568223d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 568323d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5684f46b5a66SChristoph Hellwig } 5685f46b5a66SChristoph Hellwig 5686f46b5a66SChristoph Hellwig return -ENOTTY; 5687f46b5a66SChristoph Hellwig } 56884c63c245SLuke Dashjr 56894c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56904c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56914c63c245SLuke Dashjr { 56922a362249SJeff Mahoney /* 56932a362249SJeff Mahoney * These all access 32-bit values anyway so no further 56942a362249SJeff Mahoney * handling is necessary. 56952a362249SJeff Mahoney */ 56964c63c245SLuke Dashjr switch (cmd) { 56974c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56984c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56994c63c245SLuke Dashjr break; 57004c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 57014c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 57024c63c245SLuke Dashjr break; 57034c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 57044c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 57054c63c245SLuke Dashjr break; 57064c63c245SLuke Dashjr } 57074c63c245SLuke Dashjr 57084c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 57094c63c245SLuke Dashjr } 57104c63c245SLuke Dashjr #endif 5711