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" 468719aaaeSJosef Bacik #include "space-info.h" 4786736342SJosef Bacik #include "delalloc-space.h" 48f46b5a66SChristoph Hellwig 49abccd00fSHugo Mills #ifdef CONFIG_64BIT 50abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 51abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 52abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 53abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 54abccd00fSHugo Mills */ 55abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 56abccd00fSHugo Mills __u64 sec; 57abccd00fSHugo Mills __u32 nsec; 58abccd00fSHugo Mills } __attribute__ ((__packed__)); 59abccd00fSHugo Mills 60abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 61abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 62abccd00fSHugo Mills __u64 stransid; /* in */ 63abccd00fSHugo Mills __u64 rtransid; /* out */ 64abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 65abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 66abccd00fSHugo Mills __u64 flags; /* in */ 67abccd00fSHugo Mills __u64 reserved[16]; /* in */ 68abccd00fSHugo Mills } __attribute__ ((__packed__)); 69abccd00fSHugo Mills 70abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 71abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 72abccd00fSHugo Mills #endif 73abccd00fSHugo Mills 742351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 752351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 762351f431SJosef Bacik __s64 send_fd; /* in */ 772351f431SJosef Bacik __u64 clone_sources_count; /* in */ 782351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 792351f431SJosef Bacik __u64 parent_root; /* in */ 802351f431SJosef Bacik __u64 flags; /* in */ 812351f431SJosef Bacik __u64 reserved[4]; /* in */ 822351f431SJosef Bacik } __attribute__ ((__packed__)); 832351f431SJosef Bacik 842351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 852351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 862351f431SJosef Bacik #endif 87abccd00fSHugo Mills 88416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 891c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 901c919a5eSMark Fasheh int no_time_update); 91416161dbSMark Fasheh 926cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 931905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 941905a0f7SDavid Sterba unsigned int flags) 956cbff00fSChristoph Hellwig { 961905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 976cbff00fSChristoph Hellwig return flags; 981905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 996cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1006cbff00fSChristoph Hellwig else 1016cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1026cbff00fSChristoph Hellwig } 103f46b5a66SChristoph Hellwig 1046cbff00fSChristoph Hellwig /* 105a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 106a157d4fdSDavid Sterba * ioctl. 1076cbff00fSChristoph Hellwig */ 108a157d4fdSDavid Sterba static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags) 1096cbff00fSChristoph Hellwig { 1106cbff00fSChristoph Hellwig unsigned int iflags = 0; 1116cbff00fSChristoph Hellwig 1126cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1136cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1156cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1176cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1196cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1216cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1226cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1236cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 124d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 125d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 126d0092bddSLi Zefan 12713f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 128d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 12913f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13013f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1316cbff00fSChristoph Hellwig 1326cbff00fSChristoph Hellwig return iflags; 1336cbff00fSChristoph Hellwig } 1346cbff00fSChristoph Hellwig 1356cbff00fSChristoph Hellwig /* 1366cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1376cbff00fSChristoph Hellwig */ 1387b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1396cbff00fSChristoph Hellwig { 1405c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1413cc79392SFilipe Manana unsigned int new_fl = 0; 1426cbff00fSChristoph Hellwig 1435c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1443cc79392SFilipe Manana new_fl |= S_SYNC; 1455c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1463cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1475c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1483cc79392SFilipe Manana new_fl |= S_APPEND; 1495c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1503cc79392SFilipe Manana new_fl |= S_NOATIME; 1515c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1523cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1533cc79392SFilipe Manana 1543cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1553cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1563cc79392SFilipe Manana new_fl); 1576cbff00fSChristoph Hellwig } 1586cbff00fSChristoph Hellwig 1596cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1606cbff00fSChristoph Hellwig { 1615c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 1625c57b8b6SDavid Sterba unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags); 1636cbff00fSChristoph Hellwig 1646cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1656cbff00fSChristoph Hellwig return -EFAULT; 1666cbff00fSChristoph Hellwig return 0; 1676cbff00fSChristoph Hellwig } 1686cbff00fSChristoph Hellwig 1695ba76abfSDavid Sterba /* Check if @flags are a supported and valid set of FS_*_FL flags */ 1705ba76abfSDavid Sterba static int check_fsflags(unsigned int flags) 17175e7cb7fSLiu Bo { 17275e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17375e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17475e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 175e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 176e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17775e7cb7fSLiu Bo return -EOPNOTSUPP; 17875e7cb7fSLiu Bo 17975e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 18075e7cb7fSLiu Bo return -EINVAL; 18175e7cb7fSLiu Bo 18275e7cb7fSLiu Bo return 0; 18375e7cb7fSLiu Bo } 18475e7cb7fSLiu Bo 1856cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1866cbff00fSChristoph Hellwig { 187496ad9aaSAl Viro struct inode *inode = file_inode(file); 1880b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1895c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1905c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 1916cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1925aca2842SDarrick J. Wong unsigned int fsflags, old_fsflags; 1936cbff00fSChristoph Hellwig int ret; 194ff9fef55SAnand Jain const char *comp = NULL; 195d2b8fcfeSAnand Jain u32 binode_flags = binode->flags; 1966cbff00fSChristoph Hellwig 197bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 198bd60ea0fSDavid Sterba return -EPERM; 199bd60ea0fSDavid Sterba 200b83cc969SLi Zefan if (btrfs_root_readonly(root)) 201b83cc969SLi Zefan return -EROFS; 202b83cc969SLi Zefan 2035c57b8b6SDavid Sterba if (copy_from_user(&fsflags, arg, sizeof(fsflags))) 2046cbff00fSChristoph Hellwig return -EFAULT; 2056cbff00fSChristoph Hellwig 2065c57b8b6SDavid Sterba ret = check_fsflags(fsflags); 20775e7cb7fSLiu Bo if (ret) 20875e7cb7fSLiu Bo return ret; 2096cbff00fSChristoph Hellwig 210e7848683SJan Kara ret = mnt_want_write_file(file); 211e7848683SJan Kara if (ret) 212e7848683SJan Kara return ret; 213e7848683SJan Kara 2145955102cSAl Viro inode_lock(inode); 2156cbff00fSChristoph Hellwig 2165c57b8b6SDavid Sterba fsflags = btrfs_mask_fsflags_for_type(inode, fsflags); 2175aca2842SDarrick J. Wong old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags); 2185aca2842SDarrick J. Wong ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags); 2195aca2842SDarrick J. Wong if (ret) 2206cbff00fSChristoph Hellwig goto out_unlock; 2216cbff00fSChristoph Hellwig 2225c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 223d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_SYNC; 2246cbff00fSChristoph Hellwig else 225d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_SYNC; 2265c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 227d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_IMMUTABLE; 2286cbff00fSChristoph Hellwig else 229d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_IMMUTABLE; 2305c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 231d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_APPEND; 2326cbff00fSChristoph Hellwig else 233d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_APPEND; 2345c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 235d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODUMP; 2366cbff00fSChristoph Hellwig else 237d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODUMP; 2385c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 239d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOATIME; 2406cbff00fSChristoph Hellwig else 241d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOATIME; 2425c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 243d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_DIRSYNC; 2446cbff00fSChristoph Hellwig else 245d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_DIRSYNC; 2465c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 24744e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2487e97b8daSDavid Sterba /* 2497e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2507e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2517e97b8daSDavid Sterba * status of the file and will not set it. 2527e97b8daSDavid Sterba */ 2537e97b8daSDavid Sterba if (inode->i_size == 0) 254d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW | 255d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM; 2567e97b8daSDavid Sterba } else { 257d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW; 2587e97b8daSDavid Sterba } 2597e97b8daSDavid Sterba } else { 2607e97b8daSDavid Sterba /* 26101327610SNicholas D Steeves * Revert back under same assumptions as above 2627e97b8daSDavid Sterba */ 26344e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2647e97b8daSDavid Sterba if (inode->i_size == 0) 265d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_NODATACOW | 266d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM); 2677e97b8daSDavid Sterba } else { 268d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODATACOW; 2697e97b8daSDavid Sterba } 2707e97b8daSDavid Sterba } 2716cbff00fSChristoph Hellwig 27275e7cb7fSLiu Bo /* 27375e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 27475e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 27575e7cb7fSLiu Bo * things smaller. 27675e7cb7fSLiu Bo */ 2775c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 278d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_COMPRESS; 279d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOCOMPRESS; 2805c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 28163541927SFilipe David Borba Manana 282eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 283eede2bf3SOmar Sandoval ret = -ETXTBSY; 284eede2bf3SOmar Sandoval goto out_unlock; 285eede2bf3SOmar Sandoval } 286eede2bf3SOmar Sandoval 287d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_COMPRESS; 288d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOCOMPRESS; 28963541927SFilipe David Borba Manana 29093370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 29193370509SDavid Sterba if (!comp || comp[0] == 0) 29293370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 293ebcb904dSLi Zefan } else { 294d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 29575e7cb7fSLiu Bo } 2966cbff00fSChristoph Hellwig 297ff9fef55SAnand Jain /* 298ff9fef55SAnand Jain * 1 for inode item 299ff9fef55SAnand Jain * 2 for properties 300ff9fef55SAnand Jain */ 301ff9fef55SAnand Jain trans = btrfs_start_transaction(root, 3); 302f062abf0SLi Zefan if (IS_ERR(trans)) { 303f062abf0SLi Zefan ret = PTR_ERR(trans); 304d2b8fcfeSAnand Jain goto out_unlock; 305f062abf0SLi Zefan } 3066cbff00fSChristoph Hellwig 307ff9fef55SAnand Jain if (comp) { 308ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, 309ff9fef55SAnand Jain strlen(comp), 0); 310ff9fef55SAnand Jain if (ret) { 311ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 312ff9fef55SAnand Jain goto out_end_trans; 313ff9fef55SAnand Jain } 314ff9fef55SAnand Jain } else { 315ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 316ff9fef55SAnand Jain 0, 0); 317ff9fef55SAnand Jain if (ret && ret != -ENODATA) { 318ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 319ff9fef55SAnand Jain goto out_end_trans; 320ff9fef55SAnand Jain } 321ff9fef55SAnand Jain } 322ff9fef55SAnand Jain 323d2b8fcfeSAnand Jain binode->flags = binode_flags; 3247b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3250c4d2d95SJosef Bacik inode_inc_iversion(inode); 326c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3276cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3286cbff00fSChristoph Hellwig 329ff9fef55SAnand Jain out_end_trans: 3303a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3316cbff00fSChristoph Hellwig out_unlock: 3325955102cSAl Viro inode_unlock(inode); 333e7848683SJan Kara mnt_drop_write_file(file); 3342d4e6f6aSliubo return ret; 3356cbff00fSChristoph Hellwig } 3366cbff00fSChristoph Hellwig 33719f93b3cSDavid Sterba /* 33819f93b3cSDavid Sterba * Translate btrfs internal inode flags to xflags as expected by the 33919f93b3cSDavid Sterba * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are 34019f93b3cSDavid Sterba * silently dropped. 34119f93b3cSDavid Sterba */ 34219f93b3cSDavid Sterba static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags) 34319f93b3cSDavid Sterba { 34419f93b3cSDavid Sterba unsigned int xflags = 0; 34519f93b3cSDavid Sterba 34619f93b3cSDavid Sterba if (flags & BTRFS_INODE_APPEND) 34719f93b3cSDavid Sterba xflags |= FS_XFLAG_APPEND; 34819f93b3cSDavid Sterba if (flags & BTRFS_INODE_IMMUTABLE) 34919f93b3cSDavid Sterba xflags |= FS_XFLAG_IMMUTABLE; 35019f93b3cSDavid Sterba if (flags & BTRFS_INODE_NOATIME) 35119f93b3cSDavid Sterba xflags |= FS_XFLAG_NOATIME; 35219f93b3cSDavid Sterba if (flags & BTRFS_INODE_NODUMP) 35319f93b3cSDavid Sterba xflags |= FS_XFLAG_NODUMP; 35419f93b3cSDavid Sterba if (flags & BTRFS_INODE_SYNC) 35519f93b3cSDavid Sterba xflags |= FS_XFLAG_SYNC; 35619f93b3cSDavid Sterba 35719f93b3cSDavid Sterba return xflags; 35819f93b3cSDavid Sterba } 35919f93b3cSDavid Sterba 36019f93b3cSDavid Sterba /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */ 36119f93b3cSDavid Sterba static int check_xflags(unsigned int flags) 36219f93b3cSDavid Sterba { 36319f93b3cSDavid Sterba if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME | 36419f93b3cSDavid Sterba FS_XFLAG_NODUMP | FS_XFLAG_SYNC)) 36519f93b3cSDavid Sterba return -EOPNOTSUPP; 36619f93b3cSDavid Sterba return 0; 36719f93b3cSDavid Sterba } 36819f93b3cSDavid Sterba 369e4202ac9SDavid Sterba /* 370e4202ac9SDavid Sterba * Set the xflags from the internal inode flags. The remaining items of fsxattr 371e4202ac9SDavid Sterba * are zeroed. 372e4202ac9SDavid Sterba */ 373e4202ac9SDavid Sterba static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg) 374e4202ac9SDavid Sterba { 375e4202ac9SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 376e4202ac9SDavid Sterba struct fsxattr fa; 377e4202ac9SDavid Sterba 3787b0e492eSDarrick J. Wong simple_fill_fsxattr(&fa, btrfs_inode_flags_to_xflags(binode->flags)); 379e4202ac9SDavid Sterba if (copy_to_user(arg, &fa, sizeof(fa))) 380e4202ac9SDavid Sterba return -EFAULT; 381e4202ac9SDavid Sterba 382e4202ac9SDavid Sterba return 0; 383e4202ac9SDavid Sterba } 384e4202ac9SDavid Sterba 385025f2121SDavid Sterba static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg) 386025f2121SDavid Sterba { 387025f2121SDavid Sterba struct inode *inode = file_inode(file); 388025f2121SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 389025f2121SDavid Sterba struct btrfs_root *root = binode->root; 390025f2121SDavid Sterba struct btrfs_trans_handle *trans; 3917b0e492eSDarrick J. Wong struct fsxattr fa, old_fa; 392025f2121SDavid Sterba unsigned old_flags; 393025f2121SDavid Sterba unsigned old_i_flags; 394025f2121SDavid Sterba int ret = 0; 395025f2121SDavid Sterba 396025f2121SDavid Sterba if (!inode_owner_or_capable(inode)) 397025f2121SDavid Sterba return -EPERM; 398025f2121SDavid Sterba 399025f2121SDavid Sterba if (btrfs_root_readonly(root)) 400025f2121SDavid Sterba return -EROFS; 401025f2121SDavid Sterba 402025f2121SDavid Sterba if (copy_from_user(&fa, arg, sizeof(fa))) 403025f2121SDavid Sterba return -EFAULT; 404025f2121SDavid Sterba 405025f2121SDavid Sterba ret = check_xflags(fa.fsx_xflags); 406025f2121SDavid Sterba if (ret) 407025f2121SDavid Sterba return ret; 408025f2121SDavid Sterba 409025f2121SDavid Sterba if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0) 410025f2121SDavid Sterba return -EOPNOTSUPP; 411025f2121SDavid Sterba 412025f2121SDavid Sterba ret = mnt_want_write_file(file); 413025f2121SDavid Sterba if (ret) 414025f2121SDavid Sterba return ret; 415025f2121SDavid Sterba 416025f2121SDavid Sterba inode_lock(inode); 417025f2121SDavid Sterba 418025f2121SDavid Sterba old_flags = binode->flags; 419025f2121SDavid Sterba old_i_flags = inode->i_flags; 420025f2121SDavid Sterba 4217b0e492eSDarrick J. Wong simple_fill_fsxattr(&old_fa, 4227b0e492eSDarrick J. Wong btrfs_inode_flags_to_xflags(binode->flags)); 4237b0e492eSDarrick J. Wong ret = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa); 4247b0e492eSDarrick J. Wong if (ret) 425025f2121SDavid Sterba goto out_unlock; 426025f2121SDavid Sterba 427025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_SYNC) 428025f2121SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 429025f2121SDavid Sterba else 430025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 431025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE) 432025f2121SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 433025f2121SDavid Sterba else 434025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 435025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_APPEND) 436025f2121SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 437025f2121SDavid Sterba else 438025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 439025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NODUMP) 440025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 441025f2121SDavid Sterba else 442025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 443025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NOATIME) 444025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 445025f2121SDavid Sterba else 446025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 447025f2121SDavid Sterba 448025f2121SDavid Sterba /* 1 item for the inode */ 449025f2121SDavid Sterba trans = btrfs_start_transaction(root, 1); 450025f2121SDavid Sterba if (IS_ERR(trans)) { 451025f2121SDavid Sterba ret = PTR_ERR(trans); 452025f2121SDavid Sterba goto out_unlock; 453025f2121SDavid Sterba } 454025f2121SDavid Sterba 455025f2121SDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 456025f2121SDavid Sterba inode_inc_iversion(inode); 457025f2121SDavid Sterba inode->i_ctime = current_time(inode); 458025f2121SDavid Sterba ret = btrfs_update_inode(trans, root, inode); 459025f2121SDavid Sterba 460025f2121SDavid Sterba btrfs_end_transaction(trans); 461025f2121SDavid Sterba 462025f2121SDavid Sterba out_unlock: 463025f2121SDavid Sterba if (ret) { 464025f2121SDavid Sterba binode->flags = old_flags; 465025f2121SDavid Sterba inode->i_flags = old_i_flags; 466025f2121SDavid Sterba } 467025f2121SDavid Sterba 468025f2121SDavid Sterba inode_unlock(inode); 469025f2121SDavid Sterba mnt_drop_write_file(file); 470025f2121SDavid Sterba 471025f2121SDavid Sterba return ret; 472025f2121SDavid Sterba } 473025f2121SDavid Sterba 4746cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4756cbff00fSChristoph Hellwig { 476496ad9aaSAl Viro struct inode *inode = file_inode(file); 4776cbff00fSChristoph Hellwig 4786cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4796cbff00fSChristoph Hellwig } 480f46b5a66SChristoph Hellwig 481f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 482f7039b1dSLi Dongyang { 4830b246afaSJeff Mahoney struct inode *inode = file_inode(file); 4840b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 485f7039b1dSLi Dongyang struct btrfs_device *device; 486f7039b1dSLi Dongyang struct request_queue *q; 487f7039b1dSLi Dongyang struct fstrim_range range; 488f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 489f7039b1dSLi Dongyang u64 num_devices = 0; 490f7039b1dSLi Dongyang int ret; 491f7039b1dSLi Dongyang 492f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 493f7039b1dSLi Dongyang return -EPERM; 494f7039b1dSLi Dongyang 495f35f06c3SFilipe Manana /* 496f35f06c3SFilipe Manana * If the fs is mounted with nologreplay, which requires it to be 497f35f06c3SFilipe Manana * mounted in RO mode as well, we can not allow discard on free space 498f35f06c3SFilipe Manana * inside block groups, because log trees refer to extents that are not 499f35f06c3SFilipe Manana * pinned in a block group's free space cache (pinning the extents is 500f35f06c3SFilipe Manana * precisely the first phase of replaying a log tree). 501f35f06c3SFilipe Manana */ 502f35f06c3SFilipe Manana if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 503f35f06c3SFilipe Manana return -EROFS; 504f35f06c3SFilipe Manana 5051f78160cSXiao Guangrong rcu_read_lock(); 5061f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 5071f78160cSXiao Guangrong dev_list) { 508f7039b1dSLi Dongyang if (!device->bdev) 509f7039b1dSLi Dongyang continue; 510f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 511f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 512f7039b1dSLi Dongyang num_devices++; 51350d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 514f7039b1dSLi Dongyang minlen); 515f7039b1dSLi Dongyang } 516f7039b1dSLi Dongyang } 5171f78160cSXiao Guangrong rcu_read_unlock(); 518f4c697e6SLukas Czerner 519f7039b1dSLi Dongyang if (!num_devices) 520f7039b1dSLi Dongyang return -EOPNOTSUPP; 521f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 522f7039b1dSLi Dongyang return -EFAULT; 5236ba9fc8eSQu Wenruo 5246ba9fc8eSQu Wenruo /* 5256ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 5266ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 5276ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 5286ba9fc8eSQu Wenruo */ 5296ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 530f4c697e6SLukas Czerner return -EINVAL; 531f7039b1dSLi Dongyang 532f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5332ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 534f7039b1dSLi Dongyang if (ret < 0) 535f7039b1dSLi Dongyang return ret; 536f7039b1dSLi Dongyang 537f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 538f7039b1dSLi Dongyang return -EFAULT; 539f7039b1dSLi Dongyang 540f7039b1dSLi Dongyang return 0; 541f7039b1dSLi Dongyang } 542f7039b1dSLi Dongyang 543dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 544dd5f9615SStefan Behrens { 54546e0f66aSChris Mason int i; 54646e0f66aSChris Mason 54746e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 54846e0f66aSChris Mason if (uuid[i]) 54946e0f66aSChris Mason return 0; 55046e0f66aSChris Mason } 55146e0f66aSChris Mason return 1; 552dd5f9615SStefan Behrens } 553dd5f9615SStefan Behrens 554d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 555cb8e7090SChristoph Hellwig struct dentry *dentry, 55652f75f4fSDavid Sterba const char *name, int namelen, 5576f72c7e2SArne Jansen u64 *async_transid, 5588696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 559f46b5a66SChristoph Hellwig { 5600b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 561f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 562f46b5a66SChristoph Hellwig struct btrfs_key key; 56349a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 564f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 565f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 566d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 56776dda93cSYan, Zheng struct btrfs_root *new_root; 568d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 56995582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5705662344bSTsutomu Itoh struct inode *inode; 571f46b5a66SChristoph Hellwig int ret; 572f46b5a66SChristoph Hellwig int err; 573f46b5a66SChristoph Hellwig u64 objectid; 574f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 5753de4586cSChris Mason u64 index = 0; 5768ea05e3aSAlexander Block uuid_le new_uuid; 577f46b5a66SChristoph Hellwig 57849a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 57949a3c4d9SDavid Sterba if (!root_item) 58049a3c4d9SDavid Sterba return -ENOMEM; 58149a3c4d9SDavid Sterba 5820b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5832fbe8c8aSAl Viro if (ret) 58449a3c4d9SDavid Sterba goto fail_free; 5856a912213SJosef Bacik 586e09fe2d2SQu Wenruo /* 587e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 58801327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 589e09fe2d2SQu Wenruo */ 59049a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 59149a3c4d9SDavid Sterba ret = -ENOSPC; 59249a3c4d9SDavid Sterba goto fail_free; 59349a3c4d9SDavid Sterba } 594e09fe2d2SQu Wenruo 595d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 5969ed74f2dSJosef Bacik /* 597d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 598d5c12070SMiao Xie * of create_snapshot(). 5999ed74f2dSJosef Bacik */ 600c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 601d5c12070SMiao Xie if (ret) 60249a3c4d9SDavid Sterba goto fail_free; 603f46b5a66SChristoph Hellwig 604d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 605d5c12070SMiao Xie if (IS_ERR(trans)) { 606d5c12070SMiao Xie ret = PTR_ERR(trans); 6077775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 60849a3c4d9SDavid Sterba goto fail_free; 609d5c12070SMiao Xie } 610d5c12070SMiao Xie trans->block_rsv = &block_rsv; 611d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 612f46b5a66SChristoph Hellwig 613a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 6146f72c7e2SArne Jansen if (ret) 6156f72c7e2SArne Jansen goto fail; 6166f72c7e2SArne Jansen 6174d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6188e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6198e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6208e8a1e31SJosef Bacik goto fail; 6218e8a1e31SJosef Bacik } 622f46b5a66SChristoph Hellwig 623f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 624f46b5a66SChristoph Hellwig 62549a3c4d9SDavid Sterba inode_item = &root_item->inode; 6263cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6273cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6283cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 629da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6300b246afaSJeff Mahoney fs_info->nodesize); 6313cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 632f46b5a66SChristoph Hellwig 63349a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 63449a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6353cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 63608fe4db1SLi Zefan 63749a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 63849a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 63949a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 64049a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 64149a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 64249a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 643f46b5a66SChristoph Hellwig 64449a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 64549a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6468ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 64749a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 64849a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 64949a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 65049a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 65149a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 65249a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 653f46b5a66SChristoph Hellwig 654925baeddSChris Mason btrfs_tree_unlock(leaf); 655f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 656f46b5a66SChristoph Hellwig leaf = NULL; 657f46b5a66SChristoph Hellwig 65849a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 659f46b5a66SChristoph Hellwig 660f46b5a66SChristoph Hellwig key.objectid = objectid; 6615d4f98a2SYan Zheng key.offset = 0; 662962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6630b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 66449a3c4d9SDavid Sterba root_item); 665f46b5a66SChristoph Hellwig if (ret) 666f46b5a66SChristoph Hellwig goto fail; 667f46b5a66SChristoph Hellwig 66876dda93cSYan, Zheng key.offset = (u64)-1; 6690b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 67079787eaaSJeff Mahoney if (IS_ERR(new_root)) { 67179787eaaSJeff Mahoney ret = PTR_ERR(new_root); 67266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 67379787eaaSJeff Mahoney goto fail; 67479787eaaSJeff Mahoney } 67576dda93cSYan, Zheng 67676dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 67776dda93cSYan, Zheng 67863541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 679ce598979SMark Fasheh if (ret) { 680ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 68166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 682ce598979SMark Fasheh goto fail; 683ce598979SMark Fasheh } 684ce598979SMark Fasheh 685f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 686f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 687f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 688f32e48e9SChandan Rajendra 689f46b5a66SChristoph Hellwig /* 690f46b5a66SChristoph Hellwig * insert the directory item 691f46b5a66SChristoph Hellwig */ 692877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 69379787eaaSJeff Mahoney if (ret) { 69466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 69579787eaaSJeff Mahoney goto fail; 69679787eaaSJeff Mahoney } 6973de4586cSChris Mason 698684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 6993de4586cSChris Mason BTRFS_FT_DIR, index); 70079787eaaSJeff Mahoney if (ret) { 70166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 702f46b5a66SChristoph Hellwig goto fail; 70379787eaaSJeff Mahoney } 7040660b5afSChris Mason 7056ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 70652c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 70752c26179SYan Zheng BUG_ON(ret); 70852c26179SYan Zheng 7096025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7104a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7110660b5afSChris Mason BUG_ON(ret); 7120660b5afSChris Mason 713cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7146bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 715dd5f9615SStefan Behrens if (ret) 71666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 717dd5f9615SStefan Behrens 718f46b5a66SChristoph Hellwig fail: 71949a3c4d9SDavid Sterba kfree(root_item); 720d5c12070SMiao Xie trans->block_rsv = NULL; 721d5c12070SMiao Xie trans->bytes_reserved = 0; 7227775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 723de6e8200SLiu Bo 72472fd032eSSage Weil if (async_transid) { 72572fd032eSSage Weil *async_transid = trans->transid; 7263a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 72700d71c9cSMiao Xie if (err) 7283a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 72972fd032eSSage Weil } else { 7303a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73172fd032eSSage Weil } 732f46b5a66SChristoph Hellwig if (err && !ret) 733f46b5a66SChristoph Hellwig ret = err; 7341a65e24bSChris Mason 7355662344bSTsutomu Itoh if (!ret) { 7365662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 737de6e8200SLiu Bo if (IS_ERR(inode)) 738de6e8200SLiu Bo return PTR_ERR(inode); 7395662344bSTsutomu Itoh d_instantiate(dentry, inode); 7405662344bSTsutomu Itoh } 741f46b5a66SChristoph Hellwig return ret; 74249a3c4d9SDavid Sterba 74349a3c4d9SDavid Sterba fail_free: 74449a3c4d9SDavid Sterba kfree(root_item); 74549a3c4d9SDavid Sterba return ret; 746f46b5a66SChristoph Hellwig } 747f46b5a66SChristoph Hellwig 748e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 74961d7e4cbSDavid Sterba struct dentry *dentry, 750e9662f70SMiao Xie u64 *async_transid, bool readonly, 751e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 752f46b5a66SChristoph Hellwig { 7530b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7542e4bfab9SYan, Zheng struct inode *inode; 755f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 756f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7572e4bfab9SYan, Zheng int ret; 7588ecebf4dSRobbie Ko bool snapshot_force_cow = false; 759f46b5a66SChristoph Hellwig 76027cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 761f46b5a66SChristoph Hellwig return -EINVAL; 762f46b5a66SChristoph Hellwig 763eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 764eede2bf3SOmar Sandoval btrfs_warn(fs_info, 765eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 766eede2bf3SOmar Sandoval return -ETXTBSY; 767eede2bf3SOmar Sandoval } 768eede2bf3SOmar Sandoval 76923269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 770a1ee7362SDavid Sterba if (!pending_snapshot) 771a1ee7362SDavid Sterba return -ENOMEM; 772a1ee7362SDavid Sterba 773b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 77423269bf5SDavid Sterba GFP_KERNEL); 7758546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7768546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 777b0c0ea63SDavid Sterba ret = -ENOMEM; 778b0c0ea63SDavid Sterba goto free_pending; 779b0c0ea63SDavid Sterba } 780b0c0ea63SDavid Sterba 7818ecebf4dSRobbie Ko /* 7828ecebf4dSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 7838ecebf4dSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 7848ecebf4dSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 7858ecebf4dSRobbie Ko */ 786ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7874e857c58SPeter Zijlstra smp_mb__after_atomic(); 78845bac0f3SLiu Bo /* wait for no snapshot writes */ 78945bac0f3SLiu Bo wait_event(root->subv_writers->wait, 79045bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 7918257b2dcSMiao Xie 7923cd24c69SEthan Lien ret = btrfs_start_delalloc_snapshot(root); 7936a03843dSMiao Xie if (ret) 794a1ee7362SDavid Sterba goto dec_and_free; 7956a03843dSMiao Xie 7968ecebf4dSRobbie Ko /* 7978ecebf4dSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 7988ecebf4dSRobbie Ko * we force future writes to fallback to COW mode during snapshot 7998ecebf4dSRobbie Ko * creation. 8008ecebf4dSRobbie Ko */ 8018ecebf4dSRobbie Ko atomic_inc(&root->snapshot_force_cow); 8028ecebf4dSRobbie Ko snapshot_force_cow = true; 8038ecebf4dSRobbie Ko 8046374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 8056a03843dSMiao Xie 80666d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 80766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 808d5c12070SMiao Xie /* 809d5c12070SMiao Xie * 1 - parent dir inode 810d5c12070SMiao Xie * 2 - dir entries 811d5c12070SMiao Xie * 1 - root item 812d5c12070SMiao Xie * 2 - root ref/backref 813d5c12070SMiao Xie * 1 - root of snapshot 814dd5f9615SStefan Behrens * 1 - UUID item 815d5c12070SMiao Xie */ 816d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 817dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 818ee3441b4SJeff Mahoney false); 819d5c12070SMiao Xie if (ret) 820a1ee7362SDavid Sterba goto dec_and_free; 821d5c12070SMiao Xie 822a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 823a22285a6SYan, Zheng pending_snapshot->root = root; 824b83cc969SLi Zefan pending_snapshot->readonly = readonly; 825e9662f70SMiao Xie pending_snapshot->dir = dir; 8268696c533SMiao Xie pending_snapshot->inherit = inherit; 827a22285a6SYan, Zheng 828d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 829a22285a6SYan, Zheng if (IS_ERR(trans)) { 830a22285a6SYan, Zheng ret = PTR_ERR(trans); 831a22285a6SYan, Zheng goto fail; 832a22285a6SYan, Zheng } 833a22285a6SYan, Zheng 8340b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 835a22285a6SYan, Zheng list_add(&pending_snapshot->list, 836a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8370b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 83872fd032eSSage Weil if (async_transid) { 83972fd032eSSage Weil *async_transid = trans->transid; 8403a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 84100d71c9cSMiao Xie if (ret) 8423a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 84372fd032eSSage Weil } else { 8443a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 84572fd032eSSage Weil } 846aec8030aSMiao Xie if (ret) 847c37b2b62SJosef Bacik goto fail; 848a22285a6SYan, Zheng 849a22285a6SYan, Zheng ret = pending_snapshot->error; 850f46b5a66SChristoph Hellwig if (ret) 8512e4bfab9SYan, Zheng goto fail; 852f46b5a66SChristoph Hellwig 853d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 854d3797308SChris Mason if (ret) 855d3797308SChris Mason goto fail; 856d3797308SChris Mason 8572b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8582e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8592e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8602e4bfab9SYan, Zheng goto fail; 8612e4bfab9SYan, Zheng } 8625662344bSTsutomu Itoh 8632e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8642e4bfab9SYan, Zheng ret = 0; 8652e4bfab9SYan, Zheng fail: 8667775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 867a1ee7362SDavid Sterba dec_and_free: 8688ecebf4dSRobbie Ko if (snapshot_force_cow) 8698ecebf4dSRobbie Ko atomic_dec(&root->snapshot_force_cow); 870ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8714625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 872b0c0ea63SDavid Sterba free_pending: 873b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8748546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 875a1ee7362SDavid Sterba kfree(pending_snapshot); 876a1ee7362SDavid Sterba 877f46b5a66SChristoph Hellwig return ret; 878f46b5a66SChristoph Hellwig } 879f46b5a66SChristoph Hellwig 8804260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8814260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8824260f7c7SSage Weil * whether the type of victim is right. 8834260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8844260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8854260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8864260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8874260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8884260f7c7SSage Weil * a. be owner of dir, or 8894260f7c7SSage Weil * b. be owner of victim, or 8904260f7c7SSage Weil * c. have CAP_FOWNER capability 89101327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8924260f7c7SSage Weil * links pointing to it. 8934260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8944260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8954260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8964260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8974260f7c7SSage Weil * nfs_async_unlink(). 8984260f7c7SSage Weil */ 8994260f7c7SSage Weil 9004260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 9014260f7c7SSage Weil { 9024260f7c7SSage Weil int error; 9034260f7c7SSage Weil 9042b0143b5SDavid Howells if (d_really_is_negative(victim)) 9054260f7c7SSage Weil return -ENOENT; 9064260f7c7SSage Weil 9072b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 9084fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 9094260f7c7SSage Weil 9104260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 9114260f7c7SSage Weil if (error) 9124260f7c7SSage Weil return error; 9134260f7c7SSage Weil if (IS_APPEND(dir)) 9144260f7c7SSage Weil return -EPERM; 9152b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9162b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9174260f7c7SSage Weil return -EPERM; 9184260f7c7SSage Weil if (isdir) { 919e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9204260f7c7SSage Weil return -ENOTDIR; 9214260f7c7SSage Weil if (IS_ROOT(victim)) 9224260f7c7SSage Weil return -EBUSY; 923e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9244260f7c7SSage Weil return -EISDIR; 9254260f7c7SSage Weil if (IS_DEADDIR(dir)) 9264260f7c7SSage Weil return -ENOENT; 9274260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9284260f7c7SSage Weil return -EBUSY; 9294260f7c7SSage Weil return 0; 9304260f7c7SSage Weil } 9314260f7c7SSage Weil 932cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 933cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 934cb8e7090SChristoph Hellwig { 9352b0143b5SDavid Howells if (d_really_is_positive(child)) 936cb8e7090SChristoph Hellwig return -EEXIST; 937cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 938cb8e7090SChristoph Hellwig return -ENOENT; 939cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 940cb8e7090SChristoph Hellwig } 941cb8e7090SChristoph Hellwig 942cb8e7090SChristoph Hellwig /* 943cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 944cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 945cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 946cb8e7090SChristoph Hellwig */ 94792872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 94852f75f4fSDavid Sterba const char *name, int namelen, 94972fd032eSSage Weil struct btrfs_root *snap_src, 9506f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9518696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 952cb8e7090SChristoph Hellwig { 9532b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 955cb8e7090SChristoph Hellwig struct dentry *dentry; 956cb8e7090SChristoph Hellwig int error; 957cb8e7090SChristoph Hellwig 95800235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 95900235411SAl Viro if (error == -EINTR) 96000235411SAl Viro return error; 961cb8e7090SChristoph Hellwig 962cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 963cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 964cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 965cb8e7090SChristoph Hellwig goto out_unlock; 966cb8e7090SChristoph Hellwig 96776dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 968cb8e7090SChristoph Hellwig if (error) 969a874a63eSLiu Bo goto out_dput; 970cb8e7090SChristoph Hellwig 9719c52057cSChris Mason /* 9729c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9739c52057cSChris Mason * check for them now when we can safely fail 9749c52057cSChris Mason */ 9759c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9769c52057cSChris Mason dir->i_ino, name, 9779c52057cSChris Mason namelen); 9789c52057cSChris Mason if (error) 9799c52057cSChris Mason goto out_dput; 9809c52057cSChris Mason 9810b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 98276dda93cSYan, Zheng 98376dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 98476dda93cSYan, Zheng goto out_up_read; 98576dda93cSYan, Zheng 9863de4586cSChris Mason if (snap_src) { 98761d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9886f72c7e2SArne Jansen async_transid, readonly, inherit); 9893de4586cSChris Mason } else { 990d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 991d5c12070SMiao Xie async_transid, inherit); 9923de4586cSChris Mason } 99376dda93cSYan, Zheng if (!error) 99476dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 99576dda93cSYan, Zheng out_up_read: 9960b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 997cb8e7090SChristoph Hellwig out_dput: 998cb8e7090SChristoph Hellwig dput(dentry); 999cb8e7090SChristoph Hellwig out_unlock: 10005955102cSAl Viro inode_unlock(dir); 1001cb8e7090SChristoph Hellwig return error; 1002cb8e7090SChristoph Hellwig } 1003cb8e7090SChristoph Hellwig 10044cb5300bSChris Mason /* 10054cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 10064cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 10074cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 10084cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 10094cb5300bSChris Mason * part of the file 10104cb5300bSChris Mason */ 1011aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 10124cb5300bSChris Mason { 10134cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10144cb5300bSChris Mason struct extent_map *em = NULL; 10154cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10164cb5300bSChris Mason u64 end; 10174cb5300bSChris Mason 10184cb5300bSChris Mason read_lock(&em_tree->lock); 101909cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10204cb5300bSChris Mason read_unlock(&em_tree->lock); 10214cb5300bSChris Mason 10224cb5300bSChris Mason if (em) { 10234cb5300bSChris Mason end = extent_map_end(em); 10244cb5300bSChris Mason free_extent_map(em); 10254cb5300bSChris Mason if (end - offset > thresh) 10264cb5300bSChris Mason return 0; 10274cb5300bSChris Mason } 10284cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10294cb5300bSChris Mason thresh /= 2; 10304cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10314cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10324cb5300bSChris Mason if (end >= thresh) 10334cb5300bSChris Mason return 0; 10344cb5300bSChris Mason return 1; 10354cb5300bSChris Mason } 10364cb5300bSChris Mason 10374cb5300bSChris Mason /* 10384cb5300bSChris Mason * helper function to walk through a file and find extents 10394cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10404cb5300bSChris Mason * 10414cb5300bSChris Mason * This is used by the defragging code to find new and small 10424cb5300bSChris Mason * extents 10434cb5300bSChris Mason */ 10444cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10454cb5300bSChris Mason struct inode *inode, u64 newer_than, 1046aab110abSDavid Sterba u64 *off, u32 thresh) 10474cb5300bSChris Mason { 10484cb5300bSChris Mason struct btrfs_path *path; 10494cb5300bSChris Mason struct btrfs_key min_key; 10504cb5300bSChris Mason struct extent_buffer *leaf; 10514cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10524cb5300bSChris Mason int type; 10534cb5300bSChris Mason int ret; 10544a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10554cb5300bSChris Mason 10564cb5300bSChris Mason path = btrfs_alloc_path(); 10574cb5300bSChris Mason if (!path) 10584cb5300bSChris Mason return -ENOMEM; 10594cb5300bSChris Mason 1060a4689d2bSDavid Sterba min_key.objectid = ino; 10614cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10624cb5300bSChris Mason min_key.offset = *off; 10634cb5300bSChris Mason 10644cb5300bSChris Mason while (1) { 10656174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10664cb5300bSChris Mason if (ret != 0) 10674cb5300bSChris Mason goto none; 1068f094c9bdSFilipe Manana process_slot: 1069a4689d2bSDavid Sterba if (min_key.objectid != ino) 10704cb5300bSChris Mason goto none; 10714cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10724cb5300bSChris Mason goto none; 10734cb5300bSChris Mason 10744cb5300bSChris Mason leaf = path->nodes[0]; 10754cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10764cb5300bSChris Mason struct btrfs_file_extent_item); 10774cb5300bSChris Mason 10784cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10794cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10804cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10814cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10824cb5300bSChris Mason *off = min_key.offset; 10834cb5300bSChris Mason btrfs_free_path(path); 10844cb5300bSChris Mason return 0; 10854cb5300bSChris Mason } 10864cb5300bSChris Mason 1087f094c9bdSFilipe Manana path->slots[0]++; 1088f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1089f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1090f094c9bdSFilipe Manana goto process_slot; 1091f094c9bdSFilipe Manana } 1092f094c9bdSFilipe Manana 10934cb5300bSChris Mason if (min_key.offset == (u64)-1) 10944cb5300bSChris Mason goto none; 10954cb5300bSChris Mason 10964cb5300bSChris Mason min_key.offset++; 10974cb5300bSChris Mason btrfs_release_path(path); 10984cb5300bSChris Mason } 10994cb5300bSChris Mason none: 11004cb5300bSChris Mason btrfs_free_path(path); 11014cb5300bSChris Mason return -ENOENT; 11024cb5300bSChris Mason } 11034cb5300bSChris Mason 11046c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 110517ce6ef8SLiu Bo { 110617ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1107940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 11086c282eb4SLi Zefan struct extent_map *em; 110909cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1110940100a4SChris Mason 1111940100a4SChris Mason /* 1112940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1113940100a4SChris Mason * the full extent lock 1114940100a4SChris Mason */ 1115940100a4SChris Mason read_lock(&em_tree->lock); 1116940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1117940100a4SChris Mason read_unlock(&em_tree->lock); 1118940100a4SChris Mason 1119940100a4SChris Mason if (!em) { 1120308d9800SFilipe Manana struct extent_state *cached = NULL; 1121308d9800SFilipe Manana u64 end = start + len - 1; 1122308d9800SFilipe Manana 1123940100a4SChris Mason /* get the big lock and read metadata off disk */ 1124ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1125fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1126e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1127940100a4SChris Mason 11286cf8bfbfSDan Carpenter if (IS_ERR(em)) 11296c282eb4SLi Zefan return NULL; 1130940100a4SChris Mason } 1131940100a4SChris Mason 11326c282eb4SLi Zefan return em; 11336c282eb4SLi Zefan } 11346c282eb4SLi Zefan 11356c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11366c282eb4SLi Zefan { 11376c282eb4SLi Zefan struct extent_map *next; 11386c282eb4SLi Zefan bool ret = true; 11396c282eb4SLi Zefan 11406c282eb4SLi Zefan /* this is the last extent */ 11416c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11426c282eb4SLi Zefan return false; 11436c282eb4SLi Zefan 11446c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1145e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1146e9512d72SChris Mason ret = false; 1147e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1148ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11496c282eb4SLi Zefan ret = false; 11506c282eb4SLi Zefan 11516c282eb4SLi Zefan free_extent_map(next); 11526c282eb4SLi Zefan return ret; 11536c282eb4SLi Zefan } 11546c282eb4SLi Zefan 1155aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1156a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1157a43a2111SAndrew Mahone int compress) 11586c282eb4SLi Zefan { 11596c282eb4SLi Zefan struct extent_map *em; 11606c282eb4SLi Zefan int ret = 1; 11616c282eb4SLi Zefan bool next_mergeable = true; 11624a3560c4SLiu Bo bool prev_mergeable = true; 11636c282eb4SLi Zefan 11646c282eb4SLi Zefan /* 11656c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11666c282eb4SLi Zefan * defragging it 11676c282eb4SLi Zefan */ 11686c282eb4SLi Zefan if (start < *defrag_end) 11696c282eb4SLi Zefan return 1; 11706c282eb4SLi Zefan 11716c282eb4SLi Zefan *skip = 0; 11726c282eb4SLi Zefan 11736c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11746c282eb4SLi Zefan if (!em) 11756c282eb4SLi Zefan return 0; 11766c282eb4SLi Zefan 1177940100a4SChris Mason /* this will cover holes, and inline extents */ 117817ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1179940100a4SChris Mason ret = 0; 118017ce6ef8SLiu Bo goto out; 118117ce6ef8SLiu Bo } 118217ce6ef8SLiu Bo 11834a3560c4SLiu Bo if (!*defrag_end) 11844a3560c4SLiu Bo prev_mergeable = false; 11854a3560c4SLiu Bo 11866c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1187940100a4SChris Mason /* 11886c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11896c282eb4SLi Zefan * real extent, don't bother defragging it 1190940100a4SChris Mason */ 1191a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11924a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1193940100a4SChris Mason ret = 0; 119417ce6ef8SLiu Bo out: 1195940100a4SChris Mason /* 1196940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1197940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1198940100a4SChris Mason * so that the next tiny extent will force a defrag. 1199940100a4SChris Mason * 1200940100a4SChris Mason * The end result of this is that tiny extents before a single big 1201940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1202940100a4SChris Mason */ 1203940100a4SChris Mason if (ret) { 1204940100a4SChris Mason *defrag_end = extent_map_end(em); 1205940100a4SChris Mason } else { 1206940100a4SChris Mason *last_len = 0; 1207940100a4SChris Mason *skip = extent_map_end(em); 1208940100a4SChris Mason *defrag_end = 0; 1209940100a4SChris Mason } 1210940100a4SChris Mason 1211940100a4SChris Mason free_extent_map(em); 1212940100a4SChris Mason return ret; 1213940100a4SChris Mason } 1214940100a4SChris Mason 12154cb5300bSChris Mason /* 12164cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12174cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12184cb5300bSChris Mason * to COW and defrag. 12194cb5300bSChris Mason * 12204cb5300bSChris Mason * It also makes sure the delalloc code has enough 12214cb5300bSChris Mason * dirty data to avoid making new small extents as part 12224cb5300bSChris Mason * of the defrag 12234cb5300bSChris Mason * 12244cb5300bSChris Mason * It's a good idea to start RA on this range 12254cb5300bSChris Mason * before calling this. 12264cb5300bSChris Mason */ 12274cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12284cb5300bSChris Mason struct page **pages, 12294cb5300bSChris Mason unsigned long start_index, 1230c41570c9SJustin Maggard unsigned long num_pages) 1231f46b5a66SChristoph Hellwig { 12324cb5300bSChris Mason unsigned long file_end; 12334cb5300bSChris Mason u64 isize = i_size_read(inode); 1234f46b5a66SChristoph Hellwig u64 page_start; 1235f46b5a66SChristoph Hellwig u64 page_end; 12361f12bd06SLiu Bo u64 page_cnt; 12374cb5300bSChris Mason int ret; 12384cb5300bSChris Mason int i; 12394cb5300bSChris Mason int i_done; 12404cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12414cb5300bSChris Mason struct extent_state *cached_state = NULL; 1242600a45e1SMiao Xie struct extent_io_tree *tree; 1243364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12443b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12454cb5300bSChris Mason 124609cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12471f12bd06SLiu Bo if (!isize || start_index > file_end) 12481f12bd06SLiu Bo return 0; 12491f12bd06SLiu Bo 12501f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12514cb5300bSChris Mason 1252364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 125309cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 125409cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12554cb5300bSChris Mason if (ret) 12564cb5300bSChris Mason return ret; 12574cb5300bSChris Mason i_done = 0; 1258600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12594cb5300bSChris Mason 12604cb5300bSChris Mason /* step one, lock all the pages */ 12611f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12624cb5300bSChris Mason struct page *page; 1263600a45e1SMiao Xie again: 1264a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12653b16a4e3SJosef Bacik start_index + i, mask); 12664cb5300bSChris Mason if (!page) 12674cb5300bSChris Mason break; 12684cb5300bSChris Mason 1269600a45e1SMiao Xie page_start = page_offset(page); 127009cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1271600a45e1SMiao Xie while (1) { 1272308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1273ff13db41SDavid Sterba &cached_state); 1274600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1275600a45e1SMiao Xie page_start); 1276308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1277e43bbe5eSDavid Sterba &cached_state); 1278600a45e1SMiao Xie if (!ordered) 1279600a45e1SMiao Xie break; 1280600a45e1SMiao Xie 1281600a45e1SMiao Xie unlock_page(page); 1282600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1283600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1284600a45e1SMiao Xie lock_page(page); 12851f12bd06SLiu Bo /* 12861f12bd06SLiu Bo * we unlocked the page above, so we need check if 12871f12bd06SLiu Bo * it was released or not. 12881f12bd06SLiu Bo */ 12891f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12901f12bd06SLiu Bo unlock_page(page); 129109cbfeafSKirill A. Shutemov put_page(page); 12921f12bd06SLiu Bo goto again; 12931f12bd06SLiu Bo } 1294600a45e1SMiao Xie } 1295600a45e1SMiao Xie 12964cb5300bSChris Mason if (!PageUptodate(page)) { 12974cb5300bSChris Mason btrfs_readpage(NULL, page); 12984cb5300bSChris Mason lock_page(page); 12994cb5300bSChris Mason if (!PageUptodate(page)) { 13004cb5300bSChris Mason unlock_page(page); 130109cbfeafSKirill A. Shutemov put_page(page); 13024cb5300bSChris Mason ret = -EIO; 13034cb5300bSChris Mason break; 13044cb5300bSChris Mason } 13054cb5300bSChris Mason } 1306600a45e1SMiao Xie 1307600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1308600a45e1SMiao Xie unlock_page(page); 130909cbfeafSKirill A. Shutemov put_page(page); 1310600a45e1SMiao Xie goto again; 1311600a45e1SMiao Xie } 1312600a45e1SMiao Xie 13134cb5300bSChris Mason pages[i] = page; 13144cb5300bSChris Mason i_done++; 13154cb5300bSChris Mason } 13164cb5300bSChris Mason if (!i_done || ret) 13174cb5300bSChris Mason goto out; 13184cb5300bSChris Mason 13191751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13204cb5300bSChris Mason goto out; 13214cb5300bSChris Mason 13224cb5300bSChris Mason /* 13234cb5300bSChris Mason * so now we have a nice long stream of locked 13244cb5300bSChris Mason * and up to date pages, lets wait on them 13254cb5300bSChris Mason */ 13264cb5300bSChris Mason for (i = 0; i < i_done; i++) 13274cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13284cb5300bSChris Mason 13294cb5300bSChris Mason page_start = page_offset(pages[0]); 133009cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13314cb5300bSChris Mason 13324cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1333ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13344cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 13354cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 13369e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1337ae0f1625SDavid Sterba &cached_state); 13384cb5300bSChris Mason 13391f12bd06SLiu Bo if (i_done != page_cnt) { 13409e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 134128c4a3e2SSu Yue btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); 13429e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1343bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 134409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 134543b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13464cb5300bSChris Mason } 13474cb5300bSChris Mason 13484cb5300bSChris Mason 13499e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1350018ed4f7SDavid Sterba &cached_state); 13514cb5300bSChris Mason 13524cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1353e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13544cb5300bSChris Mason 13554cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13564cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13574cb5300bSChris Mason ClearPageChecked(pages[i]); 13584cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13594cb5300bSChris Mason set_page_dirty(pages[i]); 13604cb5300bSChris Mason unlock_page(pages[i]); 136109cbfeafSKirill A. Shutemov put_page(pages[i]); 13624cb5300bSChris Mason } 136343b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 136443b18595SQu Wenruo false); 1365364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13664cb5300bSChris Mason return i_done; 13674cb5300bSChris Mason out: 13684cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13694cb5300bSChris Mason unlock_page(pages[i]); 137009cbfeafSKirill A. Shutemov put_page(pages[i]); 13714cb5300bSChris Mason } 1372bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 137309cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 137443b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 137543b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 137643b18595SQu Wenruo true); 1377364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13784cb5300bSChris Mason return ret; 13794cb5300bSChris Mason 13804cb5300bSChris Mason } 13814cb5300bSChris Mason 13824cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13834cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13844cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13854cb5300bSChris Mason { 13860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13874cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13884cb5300bSChris Mason struct file_ra_state *ra = NULL; 13894cb5300bSChris Mason unsigned long last_index; 1390151a31b2SLi Zefan u64 isize = i_size_read(inode); 1391940100a4SChris Mason u64 last_len = 0; 1392940100a4SChris Mason u64 skip = 0; 1393940100a4SChris Mason u64 defrag_end = 0; 13944cb5300bSChris Mason u64 newer_off = range->start; 1395f46b5a66SChristoph Hellwig unsigned long i; 1396008873eaSLi Zefan unsigned long ra_index = 0; 1397f46b5a66SChristoph Hellwig int ret; 13984cb5300bSChris Mason int defrag_count = 0; 13991a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1400aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 140109cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1402c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1403ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 14044cb5300bSChris Mason struct page **pages = NULL; 14051e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14064cb5300bSChris Mason 14070abd5b17SLiu Bo if (isize == 0) 14080abd5b17SLiu Bo return 0; 14090abd5b17SLiu Bo 14100abd5b17SLiu Bo if (range->start >= isize) 14110abd5b17SLiu Bo return -EINVAL; 14121a419d85SLi Zefan 14131e2ef46dSDavid Sterba if (do_compress) { 14141a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 14151a419d85SLi Zefan return -EINVAL; 14161a419d85SLi Zefan if (range->compress_type) 14171a419d85SLi Zefan compress_type = range->compress_type; 14181a419d85SLi Zefan } 1419f46b5a66SChristoph Hellwig 14200abd5b17SLiu Bo if (extent_thresh == 0) 1421ee22184bSByongho Lee extent_thresh = SZ_256K; 1422f46b5a66SChristoph Hellwig 14234cb5300bSChris Mason /* 14240a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14250a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14260a52d108SDavid Sterba * we don't error out. 14274cb5300bSChris Mason */ 14284cb5300bSChris Mason if (!file) { 142963e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14300a52d108SDavid Sterba if (ra) 14314cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14324cb5300bSChris Mason } else { 14334cb5300bSChris Mason ra = &file->f_ra; 14344cb5300bSChris Mason } 14354cb5300bSChris Mason 143663e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14374cb5300bSChris Mason if (!pages) { 14384cb5300bSChris Mason ret = -ENOMEM; 14394cb5300bSChris Mason goto out_ra; 14404cb5300bSChris Mason } 14414cb5300bSChris Mason 14424cb5300bSChris Mason /* find the last page to defrag */ 14431e701a32SChris Mason if (range->start + range->len > range->start) { 1444151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 144509cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14461e701a32SChris Mason } else { 144709cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14481e701a32SChris Mason } 14491e701a32SChris Mason 14504cb5300bSChris Mason if (newer_than) { 14514cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1452ee22184bSByongho Lee &newer_off, SZ_64K); 14534cb5300bSChris Mason if (!ret) { 14544cb5300bSChris Mason range->start = newer_off; 14554cb5300bSChris Mason /* 14564cb5300bSChris Mason * we always align our defrag to help keep 14574cb5300bSChris Mason * the extents in the file evenly spaced 14584cb5300bSChris Mason */ 145909cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14604cb5300bSChris Mason } else 14614cb5300bSChris Mason goto out_ra; 14624cb5300bSChris Mason } else { 146309cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14644cb5300bSChris Mason } 14654cb5300bSChris Mason if (!max_to_defrag) 1466070034bdSchandan max_to_defrag = last_index - i + 1; 14674cb5300bSChris Mason 14682a0f7f57SLi Zefan /* 14692a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14702a0f7f57SLi Zefan * written sequentially. 14712a0f7f57SLi Zefan */ 14722a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14732a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14742a0f7f57SLi Zefan 1475f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 147609cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14774cb5300bSChris Mason /* 14784cb5300bSChris Mason * make sure we stop running if someone unmounts 14794cb5300bSChris Mason * the FS 14804cb5300bSChris Mason */ 14811751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14824cb5300bSChris Mason break; 14834cb5300bSChris Mason 14840b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14850b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1486210549ebSDavid Sterba ret = -EAGAIN; 1487210549ebSDavid Sterba break; 1488210549ebSDavid Sterba } 1489210549ebSDavid Sterba 149009cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14916c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14921e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1493940100a4SChris Mason unsigned long next; 1494940100a4SChris Mason /* 1495940100a4SChris Mason * the should_defrag function tells us how much to skip 1496940100a4SChris Mason * bump our counter by the suggested amount 1497940100a4SChris Mason */ 149809cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1499940100a4SChris Mason i = max(i + 1, next); 1500940100a4SChris Mason continue; 1501940100a4SChris Mason } 1502008873eaSLi Zefan 1503008873eaSLi Zefan if (!newer_than) { 150409cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 150509cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1506008873eaSLi Zefan cluster = min(cluster, max_cluster); 1507008873eaSLi Zefan } else { 1508008873eaSLi Zefan cluster = max_cluster; 1509008873eaSLi Zefan } 1510008873eaSLi Zefan 1511008873eaSLi Zefan if (i + cluster > ra_index) { 1512008873eaSLi Zefan ra_index = max(i, ra_index); 15130a52d108SDavid Sterba if (ra) 1514d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1515d3c0bab5SDavid Sterba file, ra_index, cluster); 1516e4826a5bSchandan ra_index += cluster; 1517008873eaSLi Zefan } 15184cb5300bSChris Mason 15195955102cSAl Viro inode_lock(inode); 1520eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1521eede2bf3SOmar Sandoval ret = -ETXTBSY; 1522eede2bf3SOmar Sandoval } else { 15231e2ef46dSDavid Sterba if (do_compress) 1524eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1525008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1526eede2bf3SOmar Sandoval } 1527ecb8bea8SLiu Bo if (ret < 0) { 15285955102cSAl Viro inode_unlock(inode); 15294cb5300bSChris Mason goto out_ra; 1530ecb8bea8SLiu Bo } 15314cb5300bSChris Mason 15324cb5300bSChris Mason defrag_count += ret; 1533d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15345955102cSAl Viro inode_unlock(inode); 15354cb5300bSChris Mason 15364cb5300bSChris Mason if (newer_than) { 15374cb5300bSChris Mason if (newer_off == (u64)-1) 15384cb5300bSChris Mason break; 15394cb5300bSChris Mason 1540e1f041e1SLiu Bo if (ret > 0) 1541e1f041e1SLiu Bo i += ret; 1542e1f041e1SLiu Bo 15434cb5300bSChris Mason newer_off = max(newer_off + 1, 154409cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15454cb5300bSChris Mason 1546ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1547ee22184bSByongho Lee &newer_off, SZ_64K); 15484cb5300bSChris Mason if (!ret) { 15494cb5300bSChris Mason range->start = newer_off; 155009cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15514cb5300bSChris Mason } else { 15524cb5300bSChris Mason break; 1553940100a4SChris Mason } 15544cb5300bSChris Mason } else { 1555008873eaSLi Zefan if (ret > 0) { 1556cbcc8326SLi Zefan i += ret; 155709cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1558008873eaSLi Zefan } else { 1559940100a4SChris Mason i++; 1560008873eaSLi Zefan last_len = 0; 1561008873eaSLi Zefan } 1562f46b5a66SChristoph Hellwig } 15634cb5300bSChris Mason } 1564f46b5a66SChristoph Hellwig 1565dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15661e701a32SChris Mason filemap_flush(inode->i_mapping); 1567dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1568dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1569dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1570dec8ef90SFilipe Manana } 15711e701a32SChris Mason 15721a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15730b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15745c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15755c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15761a419d85SLi Zefan } 15771a419d85SLi Zefan 157860ccf82fSDiego Calleja ret = defrag_count; 1579940100a4SChris Mason 15804cb5300bSChris Mason out_ra: 15811e2ef46dSDavid Sterba if (do_compress) { 15825955102cSAl Viro inode_lock(inode); 1583eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15845955102cSAl Viro inode_unlock(inode); 1585633085c7SFilipe David Borba Manana } 15864cb5300bSChris Mason if (!file) 15874cb5300bSChris Mason kfree(ra); 15884cb5300bSChris Mason kfree(pages); 1589940100a4SChris Mason return ret; 1590f46b5a66SChristoph Hellwig } 1591f46b5a66SChristoph Hellwig 1592198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 159376dda93cSYan, Zheng void __user *arg) 1594f46b5a66SChristoph Hellwig { 15950b246afaSJeff Mahoney struct inode *inode = file_inode(file); 15960b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1597f46b5a66SChristoph Hellwig u64 new_size; 1598f46b5a66SChristoph Hellwig u64 old_size; 1599f46b5a66SChristoph Hellwig u64 devid = 1; 16000b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1601f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1602f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1603f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1604f46b5a66SChristoph Hellwig char *sizestr; 16059a40f122SGui Hecheng char *retptr; 1606f46b5a66SChristoph Hellwig char *devstr = NULL; 1607f46b5a66SChristoph Hellwig int ret = 0; 1608f46b5a66SChristoph Hellwig int mod = 0; 1609f46b5a66SChristoph Hellwig 1610e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1611e441d54dSChris Mason return -EPERM; 1612e441d54dSChris Mason 1613198605a8SMiao Xie ret = mnt_want_write_file(file); 1614198605a8SMiao Xie if (ret) 1615198605a8SMiao Xie return ret; 1616198605a8SMiao Xie 1617171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 161897547676SMiao Xie mnt_drop_write_file(file); 1619e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1620c9e9f97bSIlya Dryomov } 1621c9e9f97bSIlya Dryomov 1622dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1623c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1624c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1625c9e9f97bSIlya Dryomov goto out; 1626c9e9f97bSIlya Dryomov } 16275516e595SMark Fasheh 16285516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1629f46b5a66SChristoph Hellwig 1630f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1631f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1632f46b5a66SChristoph Hellwig if (devstr) { 1633f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1634f46b5a66SChristoph Hellwig *devstr = '\0'; 1635f46b5a66SChristoph Hellwig devstr = vol_args->name; 163658dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 163758dfae63SZhangZhen if (ret) 163858dfae63SZhangZhen goto out_free; 1639dfd79829SMiao Xie if (!devid) { 1640dfd79829SMiao Xie ret = -EINVAL; 1641dfd79829SMiao Xie goto out_free; 1642dfd79829SMiao Xie } 16430b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1644f46b5a66SChristoph Hellwig } 1645dba60f3fSMiao Xie 164609ba3bc9SAnand Jain device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true); 1647f46b5a66SChristoph Hellwig if (!device) { 16480b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1649c1c9ff7cSGeert Uytterhoeven devid); 1650dfd79829SMiao Xie ret = -ENODEV; 1651c9e9f97bSIlya Dryomov goto out_free; 1652f46b5a66SChristoph Hellwig } 1653dba60f3fSMiao Xie 1654ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16550b246afaSJeff Mahoney btrfs_info(fs_info, 1656efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1657c1c9ff7cSGeert Uytterhoeven devid); 1658dfd79829SMiao Xie ret = -EPERM; 16594e42ae1bSLiu Bo goto out_free; 16604e42ae1bSLiu Bo } 16614e42ae1bSLiu Bo 1662f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1663f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1664f46b5a66SChristoph Hellwig else { 1665f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1666f46b5a66SChristoph Hellwig mod = -1; 1667f46b5a66SChristoph Hellwig sizestr++; 1668f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1669f46b5a66SChristoph Hellwig mod = 1; 1670f46b5a66SChristoph Hellwig sizestr++; 1671f46b5a66SChristoph Hellwig } 16729a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16739a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1674f46b5a66SChristoph Hellwig ret = -EINVAL; 1675c9e9f97bSIlya Dryomov goto out_free; 1676f46b5a66SChristoph Hellwig } 1677f46b5a66SChristoph Hellwig } 1678f46b5a66SChristoph Hellwig 1679401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1680dfd79829SMiao Xie ret = -EPERM; 168163a212abSStefan Behrens goto out_free; 168263a212abSStefan Behrens } 168363a212abSStefan Behrens 16847cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1685f46b5a66SChristoph Hellwig 1686f46b5a66SChristoph Hellwig if (mod < 0) { 1687f46b5a66SChristoph Hellwig if (new_size > old_size) { 1688f46b5a66SChristoph Hellwig ret = -EINVAL; 1689c9e9f97bSIlya Dryomov goto out_free; 1690f46b5a66SChristoph Hellwig } 1691f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1692f46b5a66SChristoph Hellwig } else if (mod > 0) { 1693eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1694902c68a4SGui Hecheng ret = -ERANGE; 1695eb8052e0SWenliang Fan goto out_free; 1696eb8052e0SWenliang Fan } 1697f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1698f46b5a66SChristoph Hellwig } 1699f46b5a66SChristoph Hellwig 1700ee22184bSByongho Lee if (new_size < SZ_256M) { 1701f46b5a66SChristoph Hellwig ret = -EINVAL; 1702c9e9f97bSIlya Dryomov goto out_free; 1703f46b5a66SChristoph Hellwig } 1704f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1705f46b5a66SChristoph Hellwig ret = -EFBIG; 1706c9e9f97bSIlya Dryomov goto out_free; 1707f46b5a66SChristoph Hellwig } 1708f46b5a66SChristoph Hellwig 170947f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1710f46b5a66SChristoph Hellwig 17110b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1712c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1713f46b5a66SChristoph Hellwig 1714f46b5a66SChristoph Hellwig if (new_size > old_size) { 1715a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 171698d5dc13STsutomu Itoh if (IS_ERR(trans)) { 171798d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1718c9e9f97bSIlya Dryomov goto out_free; 171998d5dc13STsutomu Itoh } 1720f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17213a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1722ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1723f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17240253f40eSjeff.liu } /* equal, nothing need to do */ 1725f46b5a66SChristoph Hellwig 1726c9e9f97bSIlya Dryomov out_free: 1727f46b5a66SChristoph Hellwig kfree(vol_args); 1728c9e9f97bSIlya Dryomov out: 1729171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 173018f39c41SIlya Dryomov mnt_drop_write_file(file); 1731f46b5a66SChristoph Hellwig return ret; 1732f46b5a66SChristoph Hellwig } 1733f46b5a66SChristoph Hellwig 173472fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 173552f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17366f72c7e2SArne Jansen u64 *transid, bool readonly, 17378696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1738f46b5a66SChristoph Hellwig { 1739f46b5a66SChristoph Hellwig int namelen; 17403de4586cSChris Mason int ret = 0; 1741f46b5a66SChristoph Hellwig 1742325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1743325c50e3SJeff Mahoney return -ENOTDIR; 1744325c50e3SJeff Mahoney 1745a874a63eSLiu Bo ret = mnt_want_write_file(file); 1746a874a63eSLiu Bo if (ret) 1747a874a63eSLiu Bo goto out; 1748a874a63eSLiu Bo 174972fd032eSSage Weil namelen = strlen(name); 175072fd032eSSage Weil if (strchr(name, '/')) { 1751f46b5a66SChristoph Hellwig ret = -EINVAL; 1752a874a63eSLiu Bo goto out_drop_write; 1753f46b5a66SChristoph Hellwig } 1754f46b5a66SChristoph Hellwig 175516780cabSChris Mason if (name[0] == '.' && 175616780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 175716780cabSChris Mason ret = -EEXIST; 1758a874a63eSLiu Bo goto out_drop_write; 175916780cabSChris Mason } 176016780cabSChris Mason 17613de4586cSChris Mason if (subvol) { 176272fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17636f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1764cb8e7090SChristoph Hellwig } else { 17652903ff01SAl Viro struct fd src = fdget(fd); 17663de4586cSChris Mason struct inode *src_inode; 17672903ff01SAl Viro if (!src.file) { 17683de4586cSChris Mason ret = -EINVAL; 1769a874a63eSLiu Bo goto out_drop_write; 17703de4586cSChris Mason } 17713de4586cSChris Mason 1772496ad9aaSAl Viro src_inode = file_inode(src.file); 1773496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1774c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1775efe120a0SFrank Holton "Snapshot src from another FS"); 177623ad5b17SKusanagi Kouichi ret = -EXDEV; 1777d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1778d0242061SDavid Sterba /* 1779d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1780d0242061SDavid Sterba * are limited to own subvolumes only 1781d0242061SDavid Sterba */ 1782d0242061SDavid Sterba ret = -EPERM; 1783ecd18815SAl Viro } else { 178472fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 178572fd032eSSage Weil BTRFS_I(src_inode)->root, 17866f72c7e2SArne Jansen transid, readonly, inherit); 1787ecd18815SAl Viro } 17882903ff01SAl Viro fdput(src); 1789cb8e7090SChristoph Hellwig } 1790a874a63eSLiu Bo out_drop_write: 1791a874a63eSLiu Bo mnt_drop_write_file(file); 1792f46b5a66SChristoph Hellwig out: 179372fd032eSSage Weil return ret; 179472fd032eSSage Weil } 179572fd032eSSage Weil 179672fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1797fa0d2b9bSLi Zefan void __user *arg, int subvol) 179872fd032eSSage Weil { 1799fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 180072fd032eSSage Weil int ret; 180172fd032eSSage Weil 1802325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1803325c50e3SJeff Mahoney return -ENOTDIR; 1804325c50e3SJeff Mahoney 1805fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1806fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1807fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1808fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1809fa0d2b9bSLi Zefan 1810fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1811b83cc969SLi Zefan vol_args->fd, subvol, 18126f72c7e2SArne Jansen NULL, false, NULL); 1813fa0d2b9bSLi Zefan 1814fa0d2b9bSLi Zefan kfree(vol_args); 1815fa0d2b9bSLi Zefan return ret; 1816fa0d2b9bSLi Zefan } 1817fa0d2b9bSLi Zefan 1818fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1819fa0d2b9bSLi Zefan void __user *arg, int subvol) 1820fa0d2b9bSLi Zefan { 1821fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1822fa0d2b9bSLi Zefan int ret; 1823fdfb1e4fSLi Zefan u64 transid = 0; 1824fdfb1e4fSLi Zefan u64 *ptr = NULL; 1825b83cc969SLi Zefan bool readonly = false; 18266f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 182772fd032eSSage Weil 1828325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1829325c50e3SJeff Mahoney return -ENOTDIR; 1830325c50e3SJeff Mahoney 1831fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1832fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1833fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1834fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1835fdfb1e4fSLi Zefan 1836b83cc969SLi Zefan if (vol_args->flags & 18376f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18386f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1839b83cc969SLi Zefan ret = -EOPNOTSUPP; 1840c47ca32dSDan Carpenter goto free_args; 1841fdfb1e4fSLi Zefan } 1842fdfb1e4fSLi Zefan 1843fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1844fdfb1e4fSLi Zefan ptr = &transid; 1845b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1846b83cc969SLi Zefan readonly = true; 18476f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 184809cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18496f72c7e2SArne Jansen ret = -EINVAL; 1850c47ca32dSDan Carpenter goto free_args; 18516f72c7e2SArne Jansen } 18526f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18536f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18546f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1855c47ca32dSDan Carpenter goto free_args; 18566f72c7e2SArne Jansen } 18576f72c7e2SArne Jansen } 185875eaa0e2SSage Weil 1859fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18606f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18618696c533SMiao Xie readonly, inherit); 1862c47ca32dSDan Carpenter if (ret) 1863c47ca32dSDan Carpenter goto free_inherit; 186475eaa0e2SSage Weil 1865c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1866fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1867c47ca32dSDan Carpenter transid), 1868c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 186975eaa0e2SSage Weil ret = -EFAULT; 1870c47ca32dSDan Carpenter 1871c47ca32dSDan Carpenter free_inherit: 18726f72c7e2SArne Jansen kfree(inherit); 1873c47ca32dSDan Carpenter free_args: 1874c47ca32dSDan Carpenter kfree(vol_args); 1875f46b5a66SChristoph Hellwig return ret; 1876f46b5a66SChristoph Hellwig } 1877f46b5a66SChristoph Hellwig 18780caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18790caa102dSLi Zefan void __user *arg) 18800caa102dSLi Zefan { 1881496ad9aaSAl Viro struct inode *inode = file_inode(file); 18820b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18830caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18840caa102dSLi Zefan int ret = 0; 18850caa102dSLi Zefan u64 flags = 0; 18860caa102dSLi Zefan 18874a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18880caa102dSLi Zefan return -EINVAL; 18890caa102dSLi Zefan 18900b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 18910caa102dSLi Zefan if (btrfs_root_readonly(root)) 18920caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 18930b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 18940caa102dSLi Zefan 18950caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 18960caa102dSLi Zefan ret = -EFAULT; 18970caa102dSLi Zefan 18980caa102dSLi Zefan return ret; 18990caa102dSLi Zefan } 19000caa102dSLi Zefan 19010caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19020caa102dSLi Zefan void __user *arg) 19030caa102dSLi Zefan { 1904496ad9aaSAl Viro struct inode *inode = file_inode(file); 19050b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19060caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19070caa102dSLi Zefan struct btrfs_trans_handle *trans; 19080caa102dSLi Zefan u64 root_flags; 19090caa102dSLi Zefan u64 flags; 19100caa102dSLi Zefan int ret = 0; 19110caa102dSLi Zefan 1912bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1913bd60ea0fSDavid Sterba return -EPERM; 1914bd60ea0fSDavid Sterba 1915b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1916b9ca0664SLiu Bo if (ret) 1917b9ca0664SLiu Bo goto out; 19180caa102dSLi Zefan 19194a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1920b9ca0664SLiu Bo ret = -EINVAL; 1921b9ca0664SLiu Bo goto out_drop_write; 1922b9ca0664SLiu Bo } 19230caa102dSLi Zefan 1924b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1925b9ca0664SLiu Bo ret = -EFAULT; 1926b9ca0664SLiu Bo goto out_drop_write; 1927b9ca0664SLiu Bo } 19280caa102dSLi Zefan 1929b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1930b9ca0664SLiu Bo ret = -EINVAL; 1931b9ca0664SLiu Bo goto out_drop_write; 1932b9ca0664SLiu Bo } 19330caa102dSLi Zefan 1934b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1935b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1936b9ca0664SLiu Bo goto out_drop_write; 1937b9ca0664SLiu Bo } 19380caa102dSLi Zefan 19390b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19400caa102dSLi Zefan 19410caa102dSLi Zefan /* nothing to do */ 19420caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1943b9ca0664SLiu Bo goto out_drop_sem; 19440caa102dSLi Zefan 19450caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19462c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19470caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19480caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19492c686537SDavid Sterba } else { 19502c686537SDavid Sterba /* 19512c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19522c686537SDavid Sterba * send 19532c686537SDavid Sterba */ 19542c686537SDavid Sterba spin_lock(&root->root_item_lock); 19552c686537SDavid Sterba if (root->send_in_progress == 0) { 19560caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19570caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19582c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19592c686537SDavid Sterba } else { 19602c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19610b246afaSJeff Mahoney btrfs_warn(fs_info, 19622c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19632c686537SDavid Sterba root->root_key.objectid); 19642c686537SDavid Sterba ret = -EPERM; 19652c686537SDavid Sterba goto out_drop_sem; 19662c686537SDavid Sterba } 19672c686537SDavid Sterba } 19680caa102dSLi Zefan 19690caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19700caa102dSLi Zefan if (IS_ERR(trans)) { 19710caa102dSLi Zefan ret = PTR_ERR(trans); 19720caa102dSLi Zefan goto out_reset; 19730caa102dSLi Zefan } 19740caa102dSLi Zefan 19750b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19760caa102dSLi Zefan &root->root_key, &root->root_item); 19779417ebc8SNikolay Borisov if (ret < 0) { 19789417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19799417ebc8SNikolay Borisov goto out_reset; 19809417ebc8SNikolay Borisov } 19810caa102dSLi Zefan 19829417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19839417ebc8SNikolay Borisov 19840caa102dSLi Zefan out_reset: 19850caa102dSLi Zefan if (ret) 19860caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1987b9ca0664SLiu Bo out_drop_sem: 19880b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1989b9ca0664SLiu Bo out_drop_write: 1990b9ca0664SLiu Bo mnt_drop_write_file(file); 1991b9ca0664SLiu Bo out: 19920caa102dSLi Zefan return ret; 19930caa102dSLi Zefan } 19940caa102dSLi Zefan 1995ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1996ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1997ac8e9819SChris Mason { 1998abc6e134SChris Mason struct btrfs_key test; 1999abc6e134SChris Mason int ret; 2000abc6e134SChris Mason 2001abc6e134SChris Mason test.objectid = sk->min_objectid; 2002abc6e134SChris Mason test.type = sk->min_type; 2003abc6e134SChris Mason test.offset = sk->min_offset; 2004abc6e134SChris Mason 2005abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2006abc6e134SChris Mason if (ret < 0) 2007ac8e9819SChris Mason return 0; 2008abc6e134SChris Mason 2009abc6e134SChris Mason test.objectid = sk->max_objectid; 2010abc6e134SChris Mason test.type = sk->max_type; 2011abc6e134SChris Mason test.offset = sk->max_offset; 2012abc6e134SChris Mason 2013abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2014abc6e134SChris Mason if (ret > 0) 2015ac8e9819SChris Mason return 0; 2016ac8e9819SChris Mason return 1; 2017ac8e9819SChris Mason } 2018ac8e9819SChris Mason 2019df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2020ac8e9819SChris Mason struct btrfs_key *key, 2021ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20229b6e817dSGerhard Heift size_t *buf_size, 2023ba346b35SGerhard Heift char __user *ubuf, 2024ac8e9819SChris Mason unsigned long *sk_offset, 2025ac8e9819SChris Mason int *num_found) 2026ac8e9819SChris Mason { 2027ac8e9819SChris Mason u64 found_transid; 2028ac8e9819SChris Mason struct extent_buffer *leaf; 2029ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2030dd81d459SNaohiro Aota struct btrfs_key test; 2031ac8e9819SChris Mason unsigned long item_off; 2032ac8e9819SChris Mason unsigned long item_len; 2033ac8e9819SChris Mason int nritems; 2034ac8e9819SChris Mason int i; 2035ac8e9819SChris Mason int slot; 2036ac8e9819SChris Mason int ret = 0; 2037ac8e9819SChris Mason 2038ac8e9819SChris Mason leaf = path->nodes[0]; 2039ac8e9819SChris Mason slot = path->slots[0]; 2040ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2041ac8e9819SChris Mason 2042ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2043ac8e9819SChris Mason i = nritems; 2044ac8e9819SChris Mason goto advance_key; 2045ac8e9819SChris Mason } 2046ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2047ac8e9819SChris Mason 2048ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2049ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2050ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2051ac8e9819SChris Mason 205203b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 205303b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 205403b71c6cSGabriel de Perthuis continue; 205503b71c6cSGabriel de Perthuis 20569b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20578f5f6178SGerhard Heift if (*num_found) { 2058ac8e9819SChris Mason ret = 1; 20598f5f6178SGerhard Heift goto out; 20608f5f6178SGerhard Heift } 20618f5f6178SGerhard Heift 20628f5f6178SGerhard Heift /* 20638f5f6178SGerhard Heift * return one empty item back for v1, which does not 20648f5f6178SGerhard Heift * handle -EOVERFLOW 20658f5f6178SGerhard Heift */ 20668f5f6178SGerhard Heift 20679b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2068ac8e9819SChris Mason item_len = 0; 20698f5f6178SGerhard Heift ret = -EOVERFLOW; 20708f5f6178SGerhard Heift } 2071ac8e9819SChris Mason 20729b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2073ac8e9819SChris Mason ret = 1; 207425c9bc2eSGerhard Heift goto out; 2075ac8e9819SChris Mason } 2076ac8e9819SChris Mason 2077ac8e9819SChris Mason sh.objectid = key->objectid; 2078ac8e9819SChris Mason sh.offset = key->offset; 2079ac8e9819SChris Mason sh.type = key->type; 2080ac8e9819SChris Mason sh.len = item_len; 2081ac8e9819SChris Mason sh.transid = found_transid; 2082ac8e9819SChris Mason 2083ac8e9819SChris Mason /* copy search result header */ 2084ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2085ba346b35SGerhard Heift ret = -EFAULT; 2086ba346b35SGerhard Heift goto out; 2087ba346b35SGerhard Heift } 2088ba346b35SGerhard Heift 2089ac8e9819SChris Mason *sk_offset += sizeof(sh); 2090ac8e9819SChris Mason 2091ac8e9819SChris Mason if (item_len) { 2092ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2093ac8e9819SChris Mason /* copy the item */ 2094ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2095ba346b35SGerhard Heift item_off, item_len)) { 2096ba346b35SGerhard Heift ret = -EFAULT; 2097ba346b35SGerhard Heift goto out; 2098ba346b35SGerhard Heift } 2099ba346b35SGerhard Heift 2100ac8e9819SChris Mason *sk_offset += item_len; 2101ac8e9819SChris Mason } 2102e2156867SHugo Mills (*num_found)++; 2103ac8e9819SChris Mason 21048f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21058f5f6178SGerhard Heift goto out; 21068f5f6178SGerhard Heift 210725c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 210825c9bc2eSGerhard Heift ret = 1; 210925c9bc2eSGerhard Heift goto out; 211025c9bc2eSGerhard Heift } 2111ac8e9819SChris Mason } 2112ac8e9819SChris Mason advance_key: 2113ac8e9819SChris Mason ret = 0; 2114dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2115dd81d459SNaohiro Aota test.type = sk->max_type; 2116dd81d459SNaohiro Aota test.offset = sk->max_offset; 2117dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2118dd81d459SNaohiro Aota ret = 1; 2119dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2120abc6e134SChris Mason key->offset++; 2121dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2122abc6e134SChris Mason key->offset = 0; 2123abc6e134SChris Mason key->type++; 2124dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2125abc6e134SChris Mason key->offset = 0; 2126abc6e134SChris Mason key->type = 0; 2127abc6e134SChris Mason key->objectid++; 2128abc6e134SChris Mason } else 2129abc6e134SChris Mason ret = 1; 213025c9bc2eSGerhard Heift out: 2131ba346b35SGerhard Heift /* 2132ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2133ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2134ba346b35SGerhard Heift * * all items were found 2135ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2136ba346b35SGerhard Heift * leaf 2137ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2138ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2139ba346b35SGerhard Heift */ 2140ac8e9819SChris Mason return ret; 2141ac8e9819SChris Mason } 2142ac8e9819SChris Mason 2143ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 214412544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21459b6e817dSGerhard Heift size_t *buf_size, 2146ba346b35SGerhard Heift char __user *ubuf) 2147ac8e9819SChris Mason { 21480b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2149ac8e9819SChris Mason struct btrfs_root *root; 2150ac8e9819SChris Mason struct btrfs_key key; 2151ac8e9819SChris Mason struct btrfs_path *path; 2152ac8e9819SChris Mason int ret; 2153ac8e9819SChris Mason int num_found = 0; 2154ac8e9819SChris Mason unsigned long sk_offset = 0; 2155ac8e9819SChris Mason 21569b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21579b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 215812544442SGerhard Heift return -EOVERFLOW; 21599b6e817dSGerhard Heift } 216012544442SGerhard Heift 2161ac8e9819SChris Mason path = btrfs_alloc_path(); 2162ac8e9819SChris Mason if (!path) 2163ac8e9819SChris Mason return -ENOMEM; 2164ac8e9819SChris Mason 2165ac8e9819SChris Mason if (sk->tree_id == 0) { 2166ac8e9819SChris Mason /* search the root of the inode that was passed */ 2167ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2168ac8e9819SChris Mason } else { 2169ac8e9819SChris Mason key.objectid = sk->tree_id; 2170ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2171ac8e9819SChris Mason key.offset = (u64)-1; 2172ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2173ac8e9819SChris Mason if (IS_ERR(root)) { 2174ac8e9819SChris Mason btrfs_free_path(path); 2175ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2176ac8e9819SChris Mason } 2177ac8e9819SChris Mason } 2178ac8e9819SChris Mason 2179ac8e9819SChris Mason key.objectid = sk->min_objectid; 2180ac8e9819SChris Mason key.type = sk->min_type; 2181ac8e9819SChris Mason key.offset = sk->min_offset; 2182ac8e9819SChris Mason 2183ac8e9819SChris Mason while (1) { 21846174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2185ac8e9819SChris Mason if (ret != 0) { 2186ac8e9819SChris Mason if (ret > 0) 2187ac8e9819SChris Mason ret = 0; 2188ac8e9819SChris Mason goto err; 2189ac8e9819SChris Mason } 2190df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2191ac8e9819SChris Mason &sk_offset, &num_found); 2192b3b4aa74SDavid Sterba btrfs_release_path(path); 219325c9bc2eSGerhard Heift if (ret) 2194ac8e9819SChris Mason break; 2195ac8e9819SChris Mason 2196ac8e9819SChris Mason } 21978f5f6178SGerhard Heift if (ret > 0) 2198ac8e9819SChris Mason ret = 0; 2199ac8e9819SChris Mason err: 2200ac8e9819SChris Mason sk->nr_items = num_found; 2201ac8e9819SChris Mason btrfs_free_path(path); 2202ac8e9819SChris Mason return ret; 2203ac8e9819SChris Mason } 2204ac8e9819SChris Mason 2205ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2206ac8e9819SChris Mason void __user *argp) 2207ac8e9819SChris Mason { 2208ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2209ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2210ac8e9819SChris Mason struct inode *inode; 2211ac8e9819SChris Mason int ret; 22129b6e817dSGerhard Heift size_t buf_size; 2213ac8e9819SChris Mason 2214ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2215ac8e9819SChris Mason return -EPERM; 2216ac8e9819SChris Mason 2217ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2218ac8e9819SChris Mason 2219ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2220ba346b35SGerhard Heift return -EFAULT; 2221ba346b35SGerhard Heift 2222ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2223ac8e9819SChris Mason 2224496ad9aaSAl Viro inode = file_inode(file); 2225ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22268f5f6178SGerhard Heift 22278f5f6178SGerhard Heift /* 22288f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22298f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22308f5f6178SGerhard Heift */ 22318f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22328f5f6178SGerhard Heift ret = 0; 22338f5f6178SGerhard Heift 2234ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2235ac8e9819SChris Mason ret = -EFAULT; 2236ac8e9819SChris Mason return ret; 2237ac8e9819SChris Mason } 2238ac8e9819SChris Mason 2239cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2240cc68a8a5SGerhard Heift void __user *argp) 2241cc68a8a5SGerhard Heift { 2242cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2243cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2244cc68a8a5SGerhard Heift struct inode *inode; 2245cc68a8a5SGerhard Heift int ret; 2246cc68a8a5SGerhard Heift size_t buf_size; 2247ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2248cc68a8a5SGerhard Heift 2249cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2250cc68a8a5SGerhard Heift return -EPERM; 2251cc68a8a5SGerhard Heift 2252cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2253cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2254cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2255cc68a8a5SGerhard Heift return -EFAULT; 2256cc68a8a5SGerhard Heift 2257cc68a8a5SGerhard Heift buf_size = args.buf_size; 2258cc68a8a5SGerhard Heift 2259cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2260cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2261cc68a8a5SGerhard Heift buf_size = buf_limit; 2262cc68a8a5SGerhard Heift 2263cc68a8a5SGerhard Heift inode = file_inode(file); 2264cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2265718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2266cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2267cc68a8a5SGerhard Heift ret = -EFAULT; 2268cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2269cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2270cc68a8a5SGerhard Heift ret = -EFAULT; 2271cc68a8a5SGerhard Heift 227276dda93cSYan, Zheng return ret; 227376dda93cSYan, Zheng } 227476dda93cSYan, Zheng 227598d377a0STARUISI Hiroaki /* 2276ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2277ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 227898d377a0STARUISI Hiroaki */ 227998d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 228098d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 228198d377a0STARUISI Hiroaki { 228298d377a0STARUISI Hiroaki struct btrfs_root *root; 228398d377a0STARUISI Hiroaki struct btrfs_key key; 2284ac8e9819SChris Mason char *ptr; 228598d377a0STARUISI Hiroaki int ret = -1; 228698d377a0STARUISI Hiroaki int slot; 228798d377a0STARUISI Hiroaki int len; 228898d377a0STARUISI Hiroaki int total_len = 0; 228998d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 229098d377a0STARUISI Hiroaki struct extent_buffer *l; 229198d377a0STARUISI Hiroaki struct btrfs_path *path; 229298d377a0STARUISI Hiroaki 229398d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 229498d377a0STARUISI Hiroaki name[0]='\0'; 229598d377a0STARUISI Hiroaki return 0; 229698d377a0STARUISI Hiroaki } 229798d377a0STARUISI Hiroaki 229898d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 229998d377a0STARUISI Hiroaki if (!path) 230098d377a0STARUISI Hiroaki return -ENOMEM; 230198d377a0STARUISI Hiroaki 2302c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 230398d377a0STARUISI Hiroaki 230498d377a0STARUISI Hiroaki key.objectid = tree_id; 230598d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 230698d377a0STARUISI Hiroaki key.offset = (u64)-1; 230798d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 230898d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2309ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 23108ad6fcabSChris Mason goto out; 231198d377a0STARUISI Hiroaki } 231298d377a0STARUISI Hiroaki 231398d377a0STARUISI Hiroaki key.objectid = dirid; 231498d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23158ad6fcabSChris Mason key.offset = (u64)-1; 231698d377a0STARUISI Hiroaki 231798d377a0STARUISI Hiroaki while (1) { 231898d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 231998d377a0STARUISI Hiroaki if (ret < 0) 232098d377a0STARUISI Hiroaki goto out; 232118674c6cSFilipe David Borba Manana else if (ret > 0) { 232218674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 232318674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 232418674c6cSFilipe David Borba Manana if (ret < 0) 232518674c6cSFilipe David Borba Manana goto out; 232618674c6cSFilipe David Borba Manana else if (ret > 0) { 2327ac8e9819SChris Mason ret = -ENOENT; 232898d377a0STARUISI Hiroaki goto out; 2329ac8e9819SChris Mason } 233018674c6cSFilipe David Borba Manana } 233118674c6cSFilipe David Borba Manana 233218674c6cSFilipe David Borba Manana l = path->nodes[0]; 233318674c6cSFilipe David Borba Manana slot = path->slots[0]; 233418674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 233598d377a0STARUISI Hiroaki 233698d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 233798d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 233898d377a0STARUISI Hiroaki ptr -= len + 1; 233998d377a0STARUISI Hiroaki total_len += len + 1; 2340a696cf35SFilipe David Borba Manana if (ptr < name) { 2341a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 234298d377a0STARUISI Hiroaki goto out; 2343a696cf35SFilipe David Borba Manana } 234498d377a0STARUISI Hiroaki 234598d377a0STARUISI Hiroaki *(ptr + len) = '/'; 234698d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 234798d377a0STARUISI Hiroaki 234898d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 234998d377a0STARUISI Hiroaki break; 235098d377a0STARUISI Hiroaki 2351b3b4aa74SDavid Sterba btrfs_release_path(path); 235298d377a0STARUISI Hiroaki key.objectid = key.offset; 23538ad6fcabSChris Mason key.offset = (u64)-1; 235498d377a0STARUISI Hiroaki dirid = key.objectid; 235598d377a0STARUISI Hiroaki } 235677906a50SLi Zefan memmove(name, ptr, total_len); 235798d377a0STARUISI Hiroaki name[total_len] = '\0'; 235898d377a0STARUISI Hiroaki ret = 0; 235998d377a0STARUISI Hiroaki out: 236098d377a0STARUISI Hiroaki btrfs_free_path(path); 2361ac8e9819SChris Mason return ret; 2362ac8e9819SChris Mason } 2363ac8e9819SChris Mason 236423d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 236523d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 236623d0b79dSTomohiro Misono { 236723d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 236823d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 236923d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 237023d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 237123d0b79dSTomohiro Misono u64 dirid = args->dirid; 237223d0b79dSTomohiro Misono unsigned long item_off; 237323d0b79dSTomohiro Misono unsigned long item_len; 237423d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 237523d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 237623d0b79dSTomohiro Misono struct btrfs_root *root; 237723d0b79dSTomohiro Misono struct btrfs_path *path; 237823d0b79dSTomohiro Misono struct btrfs_key key, key2; 237923d0b79dSTomohiro Misono struct extent_buffer *leaf; 238023d0b79dSTomohiro Misono struct inode *temp_inode; 238123d0b79dSTomohiro Misono char *ptr; 238223d0b79dSTomohiro Misono int slot; 238323d0b79dSTomohiro Misono int len; 238423d0b79dSTomohiro Misono int total_len = 0; 238523d0b79dSTomohiro Misono int ret; 238623d0b79dSTomohiro Misono 238723d0b79dSTomohiro Misono path = btrfs_alloc_path(); 238823d0b79dSTomohiro Misono if (!path) 238923d0b79dSTomohiro Misono return -ENOMEM; 239023d0b79dSTomohiro Misono 239123d0b79dSTomohiro Misono /* 239223d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 239323d0b79dSTomohiro Misono * construct the path in from the bottom up. 239423d0b79dSTomohiro Misono */ 239523d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 239623d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 239723d0b79dSTomohiro Misono 239823d0b79dSTomohiro Misono key.objectid = treeid; 239923d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 240023d0b79dSTomohiro Misono key.offset = (u64)-1; 240123d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 240223d0b79dSTomohiro Misono if (IS_ERR(root)) { 240323d0b79dSTomohiro Misono ret = PTR_ERR(root); 240423d0b79dSTomohiro Misono goto out; 240523d0b79dSTomohiro Misono } 240623d0b79dSTomohiro Misono 240723d0b79dSTomohiro Misono key.objectid = dirid; 240823d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 240923d0b79dSTomohiro Misono key.offset = (u64)-1; 241023d0b79dSTomohiro Misono while (1) { 241123d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 241223d0b79dSTomohiro Misono if (ret < 0) { 241323d0b79dSTomohiro Misono goto out; 241423d0b79dSTomohiro Misono } else if (ret > 0) { 241523d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 241623d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 241723d0b79dSTomohiro Misono if (ret < 0) { 241823d0b79dSTomohiro Misono goto out; 241923d0b79dSTomohiro Misono } else if (ret > 0) { 242023d0b79dSTomohiro Misono ret = -ENOENT; 242123d0b79dSTomohiro Misono goto out; 242223d0b79dSTomohiro Misono } 242323d0b79dSTomohiro Misono } 242423d0b79dSTomohiro Misono 242523d0b79dSTomohiro Misono leaf = path->nodes[0]; 242623d0b79dSTomohiro Misono slot = path->slots[0]; 242723d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 242823d0b79dSTomohiro Misono 242923d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 243023d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 243123d0b79dSTomohiro Misono ptr -= len + 1; 243223d0b79dSTomohiro Misono total_len += len + 1; 243323d0b79dSTomohiro Misono if (ptr < args->path) { 243423d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 243523d0b79dSTomohiro Misono goto out; 243623d0b79dSTomohiro Misono } 243723d0b79dSTomohiro Misono 243823d0b79dSTomohiro Misono *(ptr + len) = '/'; 243923d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 244023d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 244123d0b79dSTomohiro Misono 244223d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 244323d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 244423d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 244523d0b79dSTomohiro Misono if (ret < 0) { 244623d0b79dSTomohiro Misono goto out; 244723d0b79dSTomohiro Misono } else if (ret > 0) { 244823d0b79dSTomohiro Misono ret = -ENOENT; 244923d0b79dSTomohiro Misono goto out; 245023d0b79dSTomohiro Misono } 245123d0b79dSTomohiro Misono 245223d0b79dSTomohiro Misono leaf = path->nodes[0]; 245323d0b79dSTomohiro Misono slot = path->slots[0]; 245423d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 245523d0b79dSTomohiro Misono if (key2.objectid != dirid) { 245623d0b79dSTomohiro Misono ret = -ENOENT; 245723d0b79dSTomohiro Misono goto out; 245823d0b79dSTomohiro Misono } 245923d0b79dSTomohiro Misono 246023d0b79dSTomohiro Misono temp_inode = btrfs_iget(sb, &key2, root, NULL); 24613ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24623ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24633ca57bd6SMisono Tomohiro goto out; 24643ca57bd6SMisono Tomohiro } 246523d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 246623d0b79dSTomohiro Misono iput(temp_inode); 246723d0b79dSTomohiro Misono if (ret) { 246823d0b79dSTomohiro Misono ret = -EACCES; 246923d0b79dSTomohiro Misono goto out; 247023d0b79dSTomohiro Misono } 247123d0b79dSTomohiro Misono 247223d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 247323d0b79dSTomohiro Misono break; 247423d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 247523d0b79dSTomohiro Misono ret = -EACCES; 247623d0b79dSTomohiro Misono goto out; 247723d0b79dSTomohiro Misono } 247823d0b79dSTomohiro Misono 247923d0b79dSTomohiro Misono btrfs_release_path(path); 248023d0b79dSTomohiro Misono key.objectid = key.offset; 248123d0b79dSTomohiro Misono key.offset = (u64)-1; 248223d0b79dSTomohiro Misono dirid = key.objectid; 248323d0b79dSTomohiro Misono } 248423d0b79dSTomohiro Misono 248523d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 248623d0b79dSTomohiro Misono args->path[total_len] = '\0'; 248723d0b79dSTomohiro Misono btrfs_release_path(path); 248823d0b79dSTomohiro Misono } 248923d0b79dSTomohiro Misono 249023d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 249123d0b79dSTomohiro Misono root = fs_info->tree_root; 249223d0b79dSTomohiro Misono key.objectid = treeid; 249323d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 249423d0b79dSTomohiro Misono key.offset = args->treeid; 249523d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 249623d0b79dSTomohiro Misono if (ret < 0) { 249723d0b79dSTomohiro Misono goto out; 249823d0b79dSTomohiro Misono } else if (ret > 0) { 249923d0b79dSTomohiro Misono ret = -ENOENT; 250023d0b79dSTomohiro Misono goto out; 250123d0b79dSTomohiro Misono } 250223d0b79dSTomohiro Misono 250323d0b79dSTomohiro Misono leaf = path->nodes[0]; 250423d0b79dSTomohiro Misono slot = path->slots[0]; 250523d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 250623d0b79dSTomohiro Misono 250723d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 250823d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 250923d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 251023d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 251123d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 251223d0b79dSTomohiro Misono ret = -EINVAL; 251323d0b79dSTomohiro Misono goto out; 251423d0b79dSTomohiro Misono } 251523d0b79dSTomohiro Misono 251623d0b79dSTomohiro Misono /* Copy subvolume's name */ 251723d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 251823d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 251923d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 252023d0b79dSTomohiro Misono args->name[item_len] = 0; 252123d0b79dSTomohiro Misono 252223d0b79dSTomohiro Misono out: 252323d0b79dSTomohiro Misono btrfs_free_path(path); 252423d0b79dSTomohiro Misono return ret; 252523d0b79dSTomohiro Misono } 252623d0b79dSTomohiro Misono 2527ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2528ac8e9819SChris Mason void __user *argp) 2529ac8e9819SChris Mason { 2530ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2531ac8e9819SChris Mason struct inode *inode; 253201b810b8SDavid Sterba int ret = 0; 2533ac8e9819SChris Mason 25342354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25352354d08fSJulia Lawall if (IS_ERR(args)) 25362354d08fSJulia Lawall return PTR_ERR(args); 2537c2b96929SDan Carpenter 2538496ad9aaSAl Viro inode = file_inode(file); 2539ac8e9819SChris Mason 254001b810b8SDavid Sterba /* 254101b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 254201b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 254301b810b8SDavid Sterba */ 25441b53ac4dSChris Mason if (args->treeid == 0) 25451b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25461b53ac4dSChris Mason 254701b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 254801b810b8SDavid Sterba args->name[0] = 0; 254901b810b8SDavid Sterba goto out; 255001b810b8SDavid Sterba } 255101b810b8SDavid Sterba 255201b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 255301b810b8SDavid Sterba ret = -EPERM; 255401b810b8SDavid Sterba goto out; 255501b810b8SDavid Sterba } 255601b810b8SDavid Sterba 2557ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2558ac8e9819SChris Mason args->treeid, args->objectid, 2559ac8e9819SChris Mason args->name); 2560ac8e9819SChris Mason 256101b810b8SDavid Sterba out: 2562ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2563ac8e9819SChris Mason ret = -EFAULT; 2564ac8e9819SChris Mason 2565ac8e9819SChris Mason kfree(args); 256698d377a0STARUISI Hiroaki return ret; 256798d377a0STARUISI Hiroaki } 256898d377a0STARUISI Hiroaki 256923d0b79dSTomohiro Misono /* 257023d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 257123d0b79dSTomohiro Misono * 257223d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 257323d0b79dSTomohiro Misono * 257423d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 257523d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 257623d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 257723d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 257823d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 257923d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 258023d0b79dSTomohiro Misono */ 258123d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 258223d0b79dSTomohiro Misono { 258323d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 258423d0b79dSTomohiro Misono struct inode *inode; 258523d0b79dSTomohiro Misono int ret; 258623d0b79dSTomohiro Misono 258723d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 258823d0b79dSTomohiro Misono if (IS_ERR(args)) 258923d0b79dSTomohiro Misono return PTR_ERR(args); 259023d0b79dSTomohiro Misono 259123d0b79dSTomohiro Misono inode = file_inode(file); 259223d0b79dSTomohiro Misono 259323d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 259423d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 259523d0b79dSTomohiro Misono /* 259623d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 259723d0b79dSTomohiro Misono * called 259823d0b79dSTomohiro Misono */ 259923d0b79dSTomohiro Misono kfree(args); 260023d0b79dSTomohiro Misono return -EACCES; 260123d0b79dSTomohiro Misono } 260223d0b79dSTomohiro Misono 260323d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 260423d0b79dSTomohiro Misono 260523d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 260623d0b79dSTomohiro Misono ret = -EFAULT; 260723d0b79dSTomohiro Misono 260823d0b79dSTomohiro Misono kfree(args); 260923d0b79dSTomohiro Misono return ret; 261023d0b79dSTomohiro Misono } 261123d0b79dSTomohiro Misono 2612b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2613b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2614b64ec075STomohiro Misono { 2615b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2616b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2617b64ec075STomohiro Misono struct btrfs_root *root; 2618b64ec075STomohiro Misono struct btrfs_path *path; 2619b64ec075STomohiro Misono struct btrfs_key key; 2620b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2621b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2622b64ec075STomohiro Misono struct extent_buffer *leaf; 2623b64ec075STomohiro Misono unsigned long item_off; 2624b64ec075STomohiro Misono unsigned long item_len; 2625b64ec075STomohiro Misono struct inode *inode; 2626b64ec075STomohiro Misono int slot; 2627b64ec075STomohiro Misono int ret = 0; 2628b64ec075STomohiro Misono 2629b64ec075STomohiro Misono path = btrfs_alloc_path(); 2630b64ec075STomohiro Misono if (!path) 2631b64ec075STomohiro Misono return -ENOMEM; 2632b64ec075STomohiro Misono 2633b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2634b64ec075STomohiro Misono if (!subvol_info) { 2635b64ec075STomohiro Misono btrfs_free_path(path); 2636b64ec075STomohiro Misono return -ENOMEM; 2637b64ec075STomohiro Misono } 2638b64ec075STomohiro Misono 2639b64ec075STomohiro Misono inode = file_inode(file); 2640b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2641b64ec075STomohiro Misono 2642b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2643b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2644b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2645b64ec075STomohiro Misono key.offset = (u64)-1; 2646b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2647b64ec075STomohiro Misono if (IS_ERR(root)) { 2648b64ec075STomohiro Misono ret = PTR_ERR(root); 2649b64ec075STomohiro Misono goto out; 2650b64ec075STomohiro Misono } 2651b64ec075STomohiro Misono root_item = &root->root_item; 2652b64ec075STomohiro Misono 2653b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2654b64ec075STomohiro Misono 2655b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2656b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2657b64ec075STomohiro Misono 2658b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2659b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2660b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2661b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2662b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2663b64ec075STomohiro Misono 2664b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2665b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2666b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2667b64ec075STomohiro Misono 2668b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2669b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2670b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2671b64ec075STomohiro Misono 2672b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2673b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2674b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2675b64ec075STomohiro Misono 2676b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2677b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2678b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2679b64ec075STomohiro Misono 2680b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2681b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2682b64ec075STomohiro Misono root = fs_info->tree_root; 2683b64ec075STomohiro Misono 2684b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2685b64ec075STomohiro Misono key.offset = 0; 2686b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2687b64ec075STomohiro Misono if (ret < 0) { 2688b64ec075STomohiro Misono goto out; 2689b64ec075STomohiro Misono } else if (path->slots[0] >= 2690b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2691b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2692b64ec075STomohiro Misono if (ret < 0) { 2693b64ec075STomohiro Misono goto out; 2694b64ec075STomohiro Misono } else if (ret > 0) { 2695b64ec075STomohiro Misono ret = -EUCLEAN; 2696b64ec075STomohiro Misono goto out; 2697b64ec075STomohiro Misono } 2698b64ec075STomohiro Misono } 2699b64ec075STomohiro Misono 2700b64ec075STomohiro Misono leaf = path->nodes[0]; 2701b64ec075STomohiro Misono slot = path->slots[0]; 2702b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2703b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2704b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2705b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2706b64ec075STomohiro Misono 2707b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2708b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2709b64ec075STomohiro Misono 2710b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2711b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2712b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2713b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2714b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2715b64ec075STomohiro Misono item_off, item_len); 2716b64ec075STomohiro Misono } else { 2717b64ec075STomohiro Misono ret = -ENOENT; 2718b64ec075STomohiro Misono goto out; 2719b64ec075STomohiro Misono } 2720b64ec075STomohiro Misono } 2721b64ec075STomohiro Misono 2722b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2723b64ec075STomohiro Misono ret = -EFAULT; 2724b64ec075STomohiro Misono 2725b64ec075STomohiro Misono out: 2726b64ec075STomohiro Misono btrfs_free_path(path); 2727b64ec075STomohiro Misono kzfree(subvol_info); 2728b64ec075STomohiro Misono return ret; 2729b64ec075STomohiro Misono } 2730b64ec075STomohiro Misono 273142e4b520STomohiro Misono /* 273242e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 273342e4b520STomohiro Misono * except the subvolume name. 273442e4b520STomohiro Misono */ 273542e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 273642e4b520STomohiro Misono { 273742e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 273842e4b520STomohiro Misono struct btrfs_root_ref *rref; 273942e4b520STomohiro Misono struct btrfs_root *root; 274042e4b520STomohiro Misono struct btrfs_path *path; 274142e4b520STomohiro Misono struct btrfs_key key; 274242e4b520STomohiro Misono struct extent_buffer *leaf; 274342e4b520STomohiro Misono struct inode *inode; 274442e4b520STomohiro Misono u64 objectid; 274542e4b520STomohiro Misono int slot; 274642e4b520STomohiro Misono int ret; 274742e4b520STomohiro Misono u8 found; 274842e4b520STomohiro Misono 274942e4b520STomohiro Misono path = btrfs_alloc_path(); 275042e4b520STomohiro Misono if (!path) 275142e4b520STomohiro Misono return -ENOMEM; 275242e4b520STomohiro Misono 275342e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 275442e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 275542e4b520STomohiro Misono btrfs_free_path(path); 275642e4b520STomohiro Misono return PTR_ERR(rootrefs); 275742e4b520STomohiro Misono } 275842e4b520STomohiro Misono 275942e4b520STomohiro Misono inode = file_inode(file); 276042e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 276142e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 276242e4b520STomohiro Misono 276342e4b520STomohiro Misono key.objectid = objectid; 276442e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 276542e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 276642e4b520STomohiro Misono found = 0; 276742e4b520STomohiro Misono 276842e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 276942e4b520STomohiro Misono if (ret < 0) { 277042e4b520STomohiro Misono goto out; 277142e4b520STomohiro Misono } else if (path->slots[0] >= 277242e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 277342e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 277442e4b520STomohiro Misono if (ret < 0) { 277542e4b520STomohiro Misono goto out; 277642e4b520STomohiro Misono } else if (ret > 0) { 277742e4b520STomohiro Misono ret = -EUCLEAN; 277842e4b520STomohiro Misono goto out; 277942e4b520STomohiro Misono } 278042e4b520STomohiro Misono } 278142e4b520STomohiro Misono while (1) { 278242e4b520STomohiro Misono leaf = path->nodes[0]; 278342e4b520STomohiro Misono slot = path->slots[0]; 278442e4b520STomohiro Misono 278542e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 278642e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 278742e4b520STomohiro Misono ret = 0; 278842e4b520STomohiro Misono goto out; 278942e4b520STomohiro Misono } 279042e4b520STomohiro Misono 279142e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 279242e4b520STomohiro Misono ret = -EOVERFLOW; 279342e4b520STomohiro Misono goto out; 279442e4b520STomohiro Misono } 279542e4b520STomohiro Misono 279642e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 279742e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 279842e4b520STomohiro Misono rootrefs->rootref[found].dirid = 279942e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 280042e4b520STomohiro Misono found++; 280142e4b520STomohiro Misono 280242e4b520STomohiro Misono ret = btrfs_next_item(root, path); 280342e4b520STomohiro Misono if (ret < 0) { 280442e4b520STomohiro Misono goto out; 280542e4b520STomohiro Misono } else if (ret > 0) { 280642e4b520STomohiro Misono ret = -EUCLEAN; 280742e4b520STomohiro Misono goto out; 280842e4b520STomohiro Misono } 280942e4b520STomohiro Misono } 281042e4b520STomohiro Misono 281142e4b520STomohiro Misono out: 281242e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 281342e4b520STomohiro Misono rootrefs->num_items = found; 281442e4b520STomohiro Misono /* update min_treeid for next search */ 281542e4b520STomohiro Misono if (found) 281642e4b520STomohiro Misono rootrefs->min_treeid = 281742e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 281842e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 281942e4b520STomohiro Misono ret = -EFAULT; 282042e4b520STomohiro Misono } 282142e4b520STomohiro Misono 282242e4b520STomohiro Misono kfree(rootrefs); 282342e4b520STomohiro Misono btrfs_free_path(path); 282442e4b520STomohiro Misono 282542e4b520STomohiro Misono return ret; 282642e4b520STomohiro Misono } 282742e4b520STomohiro Misono 282876dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 282976dda93cSYan, Zheng void __user *arg) 283076dda93cSYan, Zheng { 283154563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 283376dda93cSYan, Zheng struct dentry *dentry; 28342b0143b5SDavid Howells struct inode *dir = d_inode(parent); 283576dda93cSYan, Zheng struct inode *inode; 283676dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 283776dda93cSYan, Zheng struct btrfs_root *dest = NULL; 283876dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 283976dda93cSYan, Zheng int namelen; 284076dda93cSYan, Zheng int err = 0; 284176dda93cSYan, Zheng 2842325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2843325c50e3SJeff Mahoney return -ENOTDIR; 2844325c50e3SJeff Mahoney 284576dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 284676dda93cSYan, Zheng if (IS_ERR(vol_args)) 284776dda93cSYan, Zheng return PTR_ERR(vol_args); 284876dda93cSYan, Zheng 284976dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 285076dda93cSYan, Zheng namelen = strlen(vol_args->name); 285176dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 285276dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 285376dda93cSYan, Zheng err = -EINVAL; 285476dda93cSYan, Zheng goto out; 285576dda93cSYan, Zheng } 285676dda93cSYan, Zheng 2857a561be71SAl Viro err = mnt_want_write_file(file); 285876dda93cSYan, Zheng if (err) 285976dda93cSYan, Zheng goto out; 286076dda93cSYan, Zheng 2861521e0546SDavid Sterba 286200235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 286300235411SAl Viro if (err == -EINTR) 286400235411SAl Viro goto out_drop_write; 286576dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 286676dda93cSYan, Zheng if (IS_ERR(dentry)) { 286776dda93cSYan, Zheng err = PTR_ERR(dentry); 286876dda93cSYan, Zheng goto out_unlock_dir; 286976dda93cSYan, Zheng } 287076dda93cSYan, Zheng 28712b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 287276dda93cSYan, Zheng err = -ENOENT; 287376dda93cSYan, Zheng goto out_dput; 287476dda93cSYan, Zheng } 287576dda93cSYan, Zheng 28762b0143b5SDavid Howells inode = d_inode(dentry); 28774260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28784260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28794260f7c7SSage Weil /* 28804260f7c7SSage Weil * Regular user. Only allow this with a special mount 28814260f7c7SSage Weil * option, when the user has write+exec access to the 28824260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28834260f7c7SSage Weil * allowed. 28844260f7c7SSage Weil * 28854260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28864260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28874260f7c7SSage Weil * otherwise be able to delete. 28884260f7c7SSage Weil * 28894260f7c7SSage Weil * Users who want to delete empty subvols should try 28904260f7c7SSage Weil * rmdir(2). 28914260f7c7SSage Weil */ 28924260f7c7SSage Weil err = -EPERM; 28930b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 28944260f7c7SSage Weil goto out_dput; 28954260f7c7SSage Weil 28964260f7c7SSage Weil /* 28974260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 28984260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 28994260f7c7SSage Weil * must be called on the dentry referencing the root 29004260f7c7SSage Weil * of the subvol, not a random directory contained 29014260f7c7SSage Weil * within it. 29024260f7c7SSage Weil */ 29034260f7c7SSage Weil err = -EINVAL; 29044260f7c7SSage Weil if (root == dest) 29054260f7c7SSage Weil goto out_dput; 29064260f7c7SSage Weil 29074260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 29084260f7c7SSage Weil if (err) 29094260f7c7SSage Weil goto out_dput; 29105c39da5bSMiao Xie } 29114260f7c7SSage Weil 29125c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 29134260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 29144260f7c7SSage Weil if (err) 29154260f7c7SSage Weil goto out_dput; 29164260f7c7SSage Weil 29174a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 291876dda93cSYan, Zheng err = -EINVAL; 291976dda93cSYan, Zheng goto out_dput; 292076dda93cSYan, Zheng } 292176dda93cSYan, Zheng 29225955102cSAl Viro inode_lock(inode); 2923f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 29245955102cSAl Viro inode_unlock(inode); 292546008d9dSAmir Goldstein if (!err) { 292646008d9dSAmir Goldstein fsnotify_rmdir(dir, dentry); 292776dda93cSYan, Zheng d_delete(dentry); 292846008d9dSAmir Goldstein } 2929fa6ac876SLiu Bo 293076dda93cSYan, Zheng out_dput: 293176dda93cSYan, Zheng dput(dentry); 293276dda93cSYan, Zheng out_unlock_dir: 29335955102cSAl Viro inode_unlock(dir); 293400235411SAl Viro out_drop_write: 29352a79f17eSAl Viro mnt_drop_write_file(file); 293676dda93cSYan, Zheng out: 293776dda93cSYan, Zheng kfree(vol_args); 293876dda93cSYan, Zheng return err; 293976dda93cSYan, Zheng } 294076dda93cSYan, Zheng 29411e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2942f46b5a66SChristoph Hellwig { 2943496ad9aaSAl Viro struct inode *inode = file_inode(file); 2944f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29451e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2946c146afadSYan Zheng int ret; 2947c146afadSYan Zheng 294825122d15SIlya Dryomov ret = mnt_want_write_file(file); 294925122d15SIlya Dryomov if (ret) 295025122d15SIlya Dryomov return ret; 2951b83cc969SLi Zefan 295225122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 295325122d15SIlya Dryomov ret = -EROFS; 295425122d15SIlya Dryomov goto out; 29555ac00addSStefan Behrens } 2956f46b5a66SChristoph Hellwig 2957f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2958f46b5a66SChristoph Hellwig case S_IFDIR: 2959e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2960e441d54dSChris Mason ret = -EPERM; 2961e441d54dSChris Mason goto out; 2962e441d54dSChris Mason } 2963de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2964f46b5a66SChristoph Hellwig break; 2965f46b5a66SChristoph Hellwig case S_IFREG: 2966616d374eSAdam Borowski /* 2967616d374eSAdam Borowski * Note that this does not check the file descriptor for write 2968616d374eSAdam Borowski * access. This prevents defragmenting executables that are 2969616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 2970616d374eSAdam Borowski */ 2971616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 2972616d374eSAdam Borowski inode_permission(inode, MAY_WRITE)) { 2973616d374eSAdam Borowski ret = -EPERM; 2974e441d54dSChris Mason goto out; 2975e441d54dSChris Mason } 29761e701a32SChris Mason 29771e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29781e701a32SChris Mason if (!range) { 29791e701a32SChris Mason ret = -ENOMEM; 29801e701a32SChris Mason goto out; 29811e701a32SChris Mason } 29821e701a32SChris Mason 29831e701a32SChris Mason if (argp) { 29841e701a32SChris Mason if (copy_from_user(range, argp, 29851e701a32SChris Mason sizeof(*range))) { 29861e701a32SChris Mason ret = -EFAULT; 29871e701a32SChris Mason kfree(range); 2988683be16eSDan Carpenter goto out; 29891e701a32SChris Mason } 29901e701a32SChris Mason /* compression requires us to start the IO */ 29911e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 29921e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 29931e701a32SChris Mason range->extent_thresh = (u32)-1; 29941e701a32SChris Mason } 29951e701a32SChris Mason } else { 29961e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 29971e701a32SChris Mason range->len = (u64)-1; 29981e701a32SChris Mason } 2999496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 30007c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 30014cb5300bSChris Mason if (ret > 0) 30024cb5300bSChris Mason ret = 0; 30031e701a32SChris Mason kfree(range); 3004f46b5a66SChristoph Hellwig break; 30058929ecfaSYan, Zheng default: 30068929ecfaSYan, Zheng ret = -EINVAL; 3007f46b5a66SChristoph Hellwig } 3008e441d54dSChris Mason out: 300925122d15SIlya Dryomov mnt_drop_write_file(file); 3010e441d54dSChris Mason return ret; 3011f46b5a66SChristoph Hellwig } 3012f46b5a66SChristoph Hellwig 30132ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3014f46b5a66SChristoph Hellwig { 3015f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3016f46b5a66SChristoph Hellwig int ret; 3017f46b5a66SChristoph Hellwig 3018e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3019e441d54dSChris Mason return -EPERM; 3020e441d54dSChris Mason 3021171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 3022e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3023c9e9f97bSIlya Dryomov 3024dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3025c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3026c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3027c9e9f97bSIlya Dryomov goto out; 3028c9e9f97bSIlya Dryomov } 3029f46b5a66SChristoph Hellwig 30305516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30312ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3032f46b5a66SChristoph Hellwig 303343d20761SAnand Jain if (!ret) 30340b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 303543d20761SAnand Jain 3036f46b5a66SChristoph Hellwig kfree(vol_args); 3037c9e9f97bSIlya Dryomov out: 3038171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3039f46b5a66SChristoph Hellwig return ret; 3040f46b5a66SChristoph Hellwig } 3041f46b5a66SChristoph Hellwig 30426b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3043f46b5a66SChristoph Hellwig { 30440b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30450b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30466b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3047f46b5a66SChristoph Hellwig int ret; 3048f46b5a66SChristoph Hellwig 3049e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3050e441d54dSChris Mason return -EPERM; 3051e441d54dSChris Mason 3052da24927bSMiao Xie ret = mnt_want_write_file(file); 3053da24927bSMiao Xie if (ret) 3054da24927bSMiao Xie return ret; 3055c146afadSYan Zheng 3056dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3057c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3058c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3059c47ca32dSDan Carpenter goto err_drop; 3060c9e9f97bSIlya Dryomov } 3061f46b5a66SChristoph Hellwig 30626b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3063fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3064fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3065fd4e994bSOmar Sandoval goto out; 3066fd4e994bSOmar Sandoval } 3067f46b5a66SChristoph Hellwig 3068171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3069183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3070183860f6SAnand Jain goto out; 3071183860f6SAnand Jain } 3072183860f6SAnand Jain 3073735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30742ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30756b526ed7SAnand Jain } else { 30766b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30772ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30786b526ed7SAnand Jain } 3079171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3080183860f6SAnand Jain 30816b526ed7SAnand Jain if (!ret) { 3082735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30830b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30846b526ed7SAnand Jain vol_args->devid); 30856b526ed7SAnand Jain else 30860b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30876b526ed7SAnand Jain vol_args->name); 30886b526ed7SAnand Jain } 3089183860f6SAnand Jain out: 3090183860f6SAnand Jain kfree(vol_args); 3091c47ca32dSDan Carpenter err_drop: 30924ac20c70SIlya Dryomov mnt_drop_write_file(file); 3093f46b5a66SChristoph Hellwig return ret; 3094f46b5a66SChristoph Hellwig } 3095f46b5a66SChristoph Hellwig 3096f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3097f46b5a66SChristoph Hellwig { 30980b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30990b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3100f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3101f46b5a66SChristoph Hellwig int ret; 3102f46b5a66SChristoph Hellwig 3103f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3104f46b5a66SChristoph Hellwig return -EPERM; 3105f46b5a66SChristoph Hellwig 3106f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3107f46b5a66SChristoph Hellwig if (ret) 3108f46b5a66SChristoph Hellwig return ret; 3109f46b5a66SChristoph Hellwig 3110171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3111f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 311258d7bbf8SDavid Sterba goto out_drop_write; 311358d7bbf8SDavid Sterba } 311458d7bbf8SDavid Sterba 311558d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 311658d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 311758d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3118f46b5a66SChristoph Hellwig goto out; 3119f46b5a66SChristoph Hellwig } 3120f46b5a66SChristoph Hellwig 312158d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31222ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3123f46b5a66SChristoph Hellwig 3124f46b5a66SChristoph Hellwig if (!ret) 31250b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3126f46b5a66SChristoph Hellwig kfree(vol_args); 312758d7bbf8SDavid Sterba out: 3128171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 312958d7bbf8SDavid Sterba out_drop_write: 3130f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 313158d7bbf8SDavid Sterba 3132f46b5a66SChristoph Hellwig return ret; 3133f46b5a66SChristoph Hellwig } 3134f46b5a66SChristoph Hellwig 31352ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 31362ff7e61eSJeff Mahoney void __user *arg) 3137475f6387SJan Schmidt { 3138027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3139475f6387SJan Schmidt struct btrfs_device *device; 31400b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3141027ed2f0SLi Zefan int ret = 0; 3142475f6387SJan Schmidt 3143027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3144027ed2f0SLi Zefan if (!fi_args) 3145027ed2f0SLi Zefan return -ENOMEM; 3146027ed2f0SLi Zefan 3147d03262c7SDavid Sterba rcu_read_lock(); 3148027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3149475f6387SJan Schmidt 3150d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3151027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3152027ed2f0SLi Zefan fi_args->max_id = device->devid; 3153475f6387SJan Schmidt } 3154d03262c7SDavid Sterba rcu_read_unlock(); 3155475f6387SJan Schmidt 3156de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3157bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3158bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3159bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 316080a773fbSDavid Sterba 3161027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3162027ed2f0SLi Zefan ret = -EFAULT; 3163475f6387SJan Schmidt 3164027ed2f0SLi Zefan kfree(fi_args); 3165027ed2f0SLi Zefan return ret; 3166475f6387SJan Schmidt } 3167475f6387SJan Schmidt 31682ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31692ff7e61eSJeff Mahoney void __user *arg) 3170475f6387SJan Schmidt { 3171475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3172475f6387SJan Schmidt struct btrfs_device *dev; 3173475f6387SJan Schmidt int ret = 0; 3174475f6387SJan Schmidt char *s_uuid = NULL; 3175475f6387SJan Schmidt 3176475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3177475f6387SJan Schmidt if (IS_ERR(di_args)) 3178475f6387SJan Schmidt return PTR_ERR(di_args); 3179475f6387SJan Schmidt 3180dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3181475f6387SJan Schmidt s_uuid = di_args->uuid; 3182475f6387SJan Schmidt 3183c5593ca3SDavid Sterba rcu_read_lock(); 3184e4319cd9SAnand Jain dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid, 318509ba3bc9SAnand Jain NULL, true); 3186475f6387SJan Schmidt 3187475f6387SJan Schmidt if (!dev) { 3188475f6387SJan Schmidt ret = -ENODEV; 3189475f6387SJan Schmidt goto out; 3190475f6387SJan Schmidt } 3191475f6387SJan Schmidt 3192475f6387SJan Schmidt di_args->devid = dev->devid; 31937cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 31947cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3195475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3196a27202fbSJim Meyering if (dev->name) { 3197672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3198672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3199a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3200a27202fbSJim Meyering } else { 320199ba55adSStefan Behrens di_args->path[0] = '\0'; 3202a27202fbSJim Meyering } 3203475f6387SJan Schmidt 3204475f6387SJan Schmidt out: 3205c5593ca3SDavid Sterba rcu_read_unlock(); 3206475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3207475f6387SJan Schmidt ret = -EFAULT; 3208475f6387SJan Schmidt 3209475f6387SJan Schmidt kfree(di_args); 3210475f6387SJan Schmidt return ret; 3211475f6387SJan Schmidt } 3212475f6387SJan Schmidt 3213d8b55242SFilipe Manana static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3214d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3215d8b55242SFilipe Manana { 3216d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3217d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3218d8b55242SFilipe Manana } 3219d8b55242SFilipe Manana 3220d8b55242SFilipe Manana static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3221d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3222d8b55242SFilipe Manana { 3223d8b55242SFilipe Manana if (inode1 < inode2) { 3224d8b55242SFilipe Manana swap(inode1, inode2); 3225d8b55242SFilipe Manana swap(loff1, loff2); 3226d8b55242SFilipe Manana } else if (inode1 == inode2 && loff2 < loff1) { 3227d8b55242SFilipe Manana swap(loff1, loff2); 3228d8b55242SFilipe Manana } 3229d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3230d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3231d8b55242SFilipe Manana } 3232d8b55242SFilipe Manana 323357a50e25SFilipe Manana static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len, 323434a28e3dSFilipe Manana struct inode *dst, u64 dst_loff) 3235416161dbSMark Fasheh { 3236416161dbSMark Fasheh int ret; 32370efa9f48SMark Fasheh 3238de02b9f6SFilipe Manana /* 3239d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3240d8b55242SFilipe Manana * source range to serialize with relocation. 3241de02b9f6SFilipe Manana */ 3242d8b55242SFilipe Manana btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 324357a50e25SFilipe Manana ret = btrfs_clone(src, dst, loff, len, len, dst_loff, 1); 3244d8b55242SFilipe Manana btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 32453973909dSTimofey Titovets 32463973909dSTimofey Titovets return ret; 32473973909dSTimofey Titovets } 32483973909dSTimofey Titovets 3249b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3250b6728768STimofey Titovets 32513973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 32523973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 32533973909dSTimofey Titovets { 32543973909dSTimofey Titovets int ret; 3255b6728768STimofey Titovets u64 i, tail_len, chunk_count; 325662d54f3aSFilipe Manana struct btrfs_root *root_dst = BTRFS_I(dst)->root; 325762d54f3aSFilipe Manana 325862d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 325962d54f3aSFilipe Manana if (root_dst->send_in_progress) { 326062d54f3aSFilipe Manana btrfs_warn_rl(root_dst->fs_info, 326162d54f3aSFilipe Manana "cannot deduplicate to root %llu while send operations are using it (%d in progress)", 326262d54f3aSFilipe Manana root_dst->root_key.objectid, 326362d54f3aSFilipe Manana root_dst->send_in_progress); 326462d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 326562d54f3aSFilipe Manana return -EAGAIN; 326662d54f3aSFilipe Manana } 326762d54f3aSFilipe Manana root_dst->dedupe_in_progress++; 326862d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 32693973909dSTimofey Titovets 3270b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3271b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 327267b07bd4STimofey Titovets 3273b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3274b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 327534a28e3dSFilipe Manana dst, dst_loff); 3276b6728768STimofey Titovets if (ret) 327762d54f3aSFilipe Manana goto out; 3278b6728768STimofey Titovets 3279b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3280b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3281b6728768STimofey Titovets } 3282b6728768STimofey Titovets 3283b6728768STimofey Titovets if (tail_len > 0) 328467b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 328534a28e3dSFilipe Manana dst_loff); 328662d54f3aSFilipe Manana out: 328762d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 328862d54f3aSFilipe Manana root_dst->dedupe_in_progress--; 328962d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 3290416161dbSMark Fasheh 3291416161dbSMark Fasheh return ret; 3292416161dbSMark Fasheh } 3293416161dbSMark Fasheh 3294f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3295f82a9901SFilipe Manana struct inode *inode, 3296f82a9901SFilipe Manana u64 endoff, 3297f82a9901SFilipe Manana const u64 destoff, 32981c919a5eSMark Fasheh const u64 olen, 32991c919a5eSMark Fasheh int no_time_update) 3300f82a9901SFilipe Manana { 3301f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3302f82a9901SFilipe Manana int ret; 3303f82a9901SFilipe Manana 3304f82a9901SFilipe Manana inode_inc_iversion(inode); 33051c919a5eSMark Fasheh if (!no_time_update) 3306c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3307f82a9901SFilipe Manana /* 3308f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3309f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3310f82a9901SFilipe Manana */ 3311f82a9901SFilipe Manana if (endoff > destoff + olen) 3312f82a9901SFilipe Manana endoff = destoff + olen; 3313f82a9901SFilipe Manana if (endoff > inode->i_size) 33146ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3315f82a9901SFilipe Manana 3316f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3317f82a9901SFilipe Manana if (ret) { 331866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 33193a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3320f82a9901SFilipe Manana goto out; 3321f82a9901SFilipe Manana } 33223a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3323f82a9901SFilipe Manana out: 3324f82a9901SFilipe Manana return ret; 3325f82a9901SFilipe Manana } 3326f82a9901SFilipe Manana 33278039d87dSFilipe Manana /* 33288039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33298039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33308039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33318039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33328039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33338039d87dSFilipe Manana * 33348039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33358039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33368039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33378039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33388039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33398039d87dSFilipe Manana * i_mutex, which we are holding here). 33408039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33418039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33428039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33438039d87dSFilipe Manana * 33448039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33458039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33468039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33478039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33488039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33498039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33508039d87dSFilipe Manana * the former. 33518039d87dSFilipe Manana */ 33524a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 33538039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33548039d87dSFilipe Manana struct btrfs_path *path, 33558039d87dSFilipe Manana struct btrfs_key *new_key, 33568039d87dSFilipe Manana const u64 drop_start, 33578039d87dSFilipe Manana const u64 datal, 33588039d87dSFilipe Manana const u64 skip, 33598039d87dSFilipe Manana const u64 size, 33608039d87dSFilipe Manana char *inline_data) 33618039d87dSFilipe Manana { 33620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 33638039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 33648039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 33650b246afaSJeff Mahoney fs_info->sectorsize); 33668039d87dSFilipe Manana int ret; 33678039d87dSFilipe Manana struct btrfs_key key; 33688039d87dSFilipe Manana 33698039d87dSFilipe Manana if (new_key->offset > 0) 33708039d87dSFilipe Manana return -EOPNOTSUPP; 33718039d87dSFilipe Manana 33724a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 33738039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 33748039d87dSFilipe Manana key.offset = 0; 33758039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 33768039d87dSFilipe Manana if (ret < 0) { 33778039d87dSFilipe Manana return ret; 33788039d87dSFilipe Manana } else if (ret > 0) { 33798039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 33808039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 33818039d87dSFilipe Manana if (ret < 0) 33828039d87dSFilipe Manana return ret; 33838039d87dSFilipe Manana else if (ret > 0) 33848039d87dSFilipe Manana goto copy_inline_extent; 33858039d87dSFilipe Manana } 33868039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 33874a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 33888039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 33898039d87dSFilipe Manana ASSERT(key.offset > 0); 33908039d87dSFilipe Manana return -EOPNOTSUPP; 33918039d87dSFilipe Manana } 33928039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 33938039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 33948039d87dSFilipe Manana u64 ext_len; 33958039d87dSFilipe Manana 33968039d87dSFilipe Manana /* 33978039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 33988039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 33998039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34008039d87dSFilipe Manana */ 34018039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34028039d87dSFilipe Manana struct btrfs_file_extent_item); 34038039d87dSFilipe Manana /* 34048039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34058039d87dSFilipe Manana * following it. 34068039d87dSFilipe Manana */ 34078039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34088039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34098039d87dSFilipe Manana goto copy_inline_extent; 34108039d87dSFilipe Manana 34118039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34128039d87dSFilipe Manana if (ext_len > aligned_end) 34138039d87dSFilipe Manana return -EOPNOTSUPP; 34148039d87dSFilipe Manana 34158039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34168039d87dSFilipe Manana if (ret < 0) { 34178039d87dSFilipe Manana return ret; 34188039d87dSFilipe Manana } else if (ret == 0) { 34198039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34208039d87dSFilipe Manana path->slots[0]); 34214a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34228039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34238039d87dSFilipe Manana return -EOPNOTSUPP; 34248039d87dSFilipe Manana } 34258039d87dSFilipe Manana } 34268039d87dSFilipe Manana 34278039d87dSFilipe Manana copy_inline_extent: 34288039d87dSFilipe Manana /* 34298039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34308039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34318039d87dSFilipe Manana */ 34328039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34338039d87dSFilipe Manana /* 34348039d87dSFilipe Manana * If the destination inode has an inline extent... 34358039d87dSFilipe Manana * This would require copying the data from the source inline 34368039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34378039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34388039d87dSFilipe Manana * or just one of them, which would require decompressing and 34398039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34408039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34418039d87dSFilipe Manana * inline extent). 34428039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34438039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34448039d87dSFilipe Manana */ 34458039d87dSFilipe Manana return -EOPNOTSUPP; 34468039d87dSFilipe Manana } 34478039d87dSFilipe Manana 34488039d87dSFilipe Manana btrfs_release_path(path); 34498039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34508039d87dSFilipe Manana if (ret) 34518039d87dSFilipe Manana return ret; 34528039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34538039d87dSFilipe Manana if (ret) 34548039d87dSFilipe Manana return ret; 34558039d87dSFilipe Manana 34568039d87dSFilipe Manana if (skip) { 34578039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 34588039d87dSFilipe Manana 34598039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 34608039d87dSFilipe Manana } 34618039d87dSFilipe Manana 34628039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 34638039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 34648039d87dSFilipe Manana path->slots[0]), 34658039d87dSFilipe Manana size); 34668039d87dSFilipe Manana inode_add_bytes(dst, datal); 3467690a5dbfSFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(dst)->runtime_flags); 34688039d87dSFilipe Manana 34698039d87dSFilipe Manana return 0; 34708039d87dSFilipe Manana } 34718039d87dSFilipe Manana 347232b7c687SMark Fasheh /** 347332b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 347432b7c687SMark Fasheh * 347532b7c687SMark Fasheh * @src: Inode to clone from 347632b7c687SMark Fasheh * @inode: Inode to clone to 347732b7c687SMark Fasheh * @off: Offset within source to start clone from 347832b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 34791c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 348032b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 34811c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 348232b7c687SMark Fasheh */ 348332b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3484f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 34851c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3486f46b5a66SChristoph Hellwig { 34870b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3488f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 348932b7c687SMark Fasheh struct btrfs_path *path = NULL; 3490f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 349132b7c687SMark Fasheh struct btrfs_trans_handle *trans; 349232b7c687SMark Fasheh char *buf = NULL; 3493ae01a0abSYan Zheng struct btrfs_key key; 3494f46b5a66SChristoph Hellwig u32 nritems; 3495f46b5a66SChristoph Hellwig int slot; 3496ae01a0abSYan Zheng int ret; 3497f82a9901SFilipe Manana const u64 len = olen_aligned; 3498f82a9901SFilipe Manana u64 last_dest_end = destoff; 3499ae01a0abSYan Zheng 3500ae01a0abSYan Zheng ret = -ENOMEM; 3501752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3502ae01a0abSYan Zheng if (!buf) 350332b7c687SMark Fasheh return ret; 3504ae01a0abSYan Zheng 3505ae01a0abSYan Zheng path = btrfs_alloc_path(); 3506ae01a0abSYan Zheng if (!path) { 350715351955SDavid Sterba kvfree(buf); 350832b7c687SMark Fasheh return ret; 3509ae01a0abSYan Zheng } 351032b7c687SMark Fasheh 3511e4058b54SDavid Sterba path->reada = READA_FORWARD; 3512c5c9cd4dSSage Weil /* clone data */ 35134a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3514ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35152c463823SFilipe Manana key.offset = off; 3516f46b5a66SChristoph Hellwig 3517f46b5a66SChristoph Hellwig while (1) { 3518de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3519df858e76SFilipe Manana 3520f46b5a66SChristoph Hellwig /* 3521f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3522f46b5a66SChristoph Hellwig * tree. 3523f46b5a66SChristoph Hellwig */ 3524e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3525362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3526362a20c5SDavid Sterba 0, 0); 3527f46b5a66SChristoph Hellwig if (ret < 0) 3528f46b5a66SChristoph Hellwig goto out; 35292c463823SFilipe Manana /* 35302c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35312c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35322c463823SFilipe Manana * it might overlap our target range, therefore process it. 35332c463823SFilipe Manana */ 35342c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35352c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35362c463823SFilipe Manana path->slots[0] - 1); 35372c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35382c463823SFilipe Manana path->slots[0]--; 35392c463823SFilipe Manana } 3540f46b5a66SChristoph Hellwig 3541ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3542e4355f34SFilipe David Borba Manana process_slot: 3543ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3544362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3545f46b5a66SChristoph Hellwig if (ret < 0) 3546f46b5a66SChristoph Hellwig goto out; 3547f46b5a66SChristoph Hellwig if (ret > 0) 3548f46b5a66SChristoph Hellwig break; 3549ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3550f46b5a66SChristoph Hellwig } 3551f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3552f46b5a66SChristoph Hellwig slot = path->slots[0]; 3553f46b5a66SChristoph Hellwig 3554ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3555962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 35564a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3557f46b5a66SChristoph Hellwig break; 3558f46b5a66SChristoph Hellwig 3559962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3560c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3561c5c9cd4dSSage Weil int type; 356231840ae1SZheng Yan u32 size; 356331840ae1SZheng Yan struct btrfs_key new_key; 3564c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3565c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3566c5c9cd4dSSage Weil u8 comp; 3567f82a9901SFilipe Manana u64 drop_start; 356831840ae1SZheng Yan 3569c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3570c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3571c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3572c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3573c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3574c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3575d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3576d397712bSChris Mason extent); 3577d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3578d397712bSChris Mason extent); 3579c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3580d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3581d397712bSChris Mason extent); 3582c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3583c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3584c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3585c5c9cd4dSSage Weil extent); 3586c5c9cd4dSSage Weil } 358731840ae1SZheng Yan 35882c463823SFilipe Manana /* 35892c463823SFilipe Manana * The first search might have left us at an extent 35902c463823SFilipe Manana * item that ends before our target range's start, can 35912c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 35922c463823SFilipe Manana */ 35932c463823SFilipe Manana if (key.offset + datal <= off) { 3594e4355f34SFilipe David Borba Manana path->slots[0]++; 3595e4355f34SFilipe David Borba Manana goto process_slot; 35962c463823SFilipe Manana } else if (key.offset >= off + len) { 35972c463823SFilipe Manana break; 3598e4355f34SFilipe David Borba Manana } 3599df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3600e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3601e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3602e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3603e4355f34SFilipe David Borba Manana size); 3604e4355f34SFilipe David Borba Manana 3605e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3606e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3607c5c9cd4dSSage Weil 360831840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36094a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36104d728ec7SLi Zefan if (off <= key.offset) 3611c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36124d728ec7SLi Zefan else 36134d728ec7SLi Zefan new_key.offset = destoff; 3614c5c9cd4dSSage Weil 3615b6f3409bSSage Weil /* 3616f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3617f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3618f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3619f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3620f82a9901SFilipe Manana * partially overlaps it). 3621f82a9901SFilipe Manana */ 3622f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3623f82a9901SFilipe Manana drop_start = last_dest_end; 3624f82a9901SFilipe Manana else 3625f82a9901SFilipe Manana drop_start = new_key.offset; 3626f82a9901SFilipe Manana 3627c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3628c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3629690a5dbfSFilipe Manana struct btrfs_clone_extent_info clone_info; 3630690a5dbfSFilipe Manana 3631d72c0842SLi Zefan /* 3632d72c0842SLi Zefan * a | --- range to clone ---| b 3633d72c0842SLi Zefan * | ------------- extent ------------- | 3634d72c0842SLi Zefan */ 3635d72c0842SLi Zefan 363693915584SAntonio Ospite /* subtract range b */ 3637d72c0842SLi Zefan if (key.offset + datal > off + len) 3638d72c0842SLi Zefan datal = off + len - key.offset; 3639d72c0842SLi Zefan 364093915584SAntonio Ospite /* subtract range a */ 3641a22285a6SYan, Zheng if (off > key.offset) { 3642a22285a6SYan, Zheng datao += off - key.offset; 3643a22285a6SYan, Zheng datal -= off - key.offset; 3644a22285a6SYan, Zheng } 3645a22285a6SYan, Zheng 3646690a5dbfSFilipe Manana clone_info.disk_offset = disko; 3647690a5dbfSFilipe Manana clone_info.disk_len = diskl; 3648690a5dbfSFilipe Manana clone_info.data_offset = datao; 3649690a5dbfSFilipe Manana clone_info.data_len = datal; 3650690a5dbfSFilipe Manana clone_info.file_offset = new_key.offset; 3651690a5dbfSFilipe Manana clone_info.extent_buf = buf; 3652690a5dbfSFilipe Manana clone_info.item_size = size; 3653690a5dbfSFilipe Manana ret = btrfs_punch_hole_range(inode, path, 3654f82a9901SFilipe Manana drop_start, 3655690a5dbfSFilipe Manana new_key.offset + datal - 1, 3656690a5dbfSFilipe Manana &clone_info, &trans); 3657690a5dbfSFilipe Manana if (ret) 365879787eaaSJeff Mahoney goto out; 3659c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3660c5c9cd4dSSage Weil u64 skip = 0; 3661c5c9cd4dSSage Weil u64 trim = 0; 3662ed958762SFilipe Manana 3663c5c9cd4dSSage Weil if (off > key.offset) { 3664c5c9cd4dSSage Weil skip = off - key.offset; 3665c5c9cd4dSSage Weil new_key.offset += skip; 366631840ae1SZheng Yan } 3667d397712bSChris Mason 3668c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3669c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3670d397712bSChris Mason 3671c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3672c5c9cd4dSSage Weil ret = -EINVAL; 3673c5c9cd4dSSage Weil goto out; 367431840ae1SZheng Yan } 3675c5c9cd4dSSage Weil size -= skip + trim; 3676c5c9cd4dSSage Weil datal -= skip + trim; 3677a22285a6SYan, Zheng 3678690a5dbfSFilipe Manana /* 3679690a5dbfSFilipe Manana * If our extent is inline, we know we will drop 3680690a5dbfSFilipe Manana * or adjust at most 1 extent item in the 3681690a5dbfSFilipe Manana * destination root. 3682690a5dbfSFilipe Manana * 3683690a5dbfSFilipe Manana * 1 - adjusting old extent (we may have to 3684690a5dbfSFilipe Manana * split it) 3685690a5dbfSFilipe Manana * 1 - add new extent 3686690a5dbfSFilipe Manana * 1 - inode update 3687690a5dbfSFilipe Manana */ 3688690a5dbfSFilipe Manana trans = btrfs_start_transaction(root, 3); 3689690a5dbfSFilipe Manana if (IS_ERR(trans)) { 3690690a5dbfSFilipe Manana ret = PTR_ERR(trans); 3691690a5dbfSFilipe Manana goto out; 3692690a5dbfSFilipe Manana } 3693690a5dbfSFilipe Manana 36944a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 36958039d87dSFilipe Manana trans, path, 36968039d87dSFilipe Manana &new_key, 3697f82a9901SFilipe Manana drop_start, 36988039d87dSFilipe Manana datal, 36998039d87dSFilipe Manana skip, size, buf); 370079787eaaSJeff Mahoney if (ret) { 37013f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37023a29bc09SChris Mason btrfs_abort_transaction(trans, 370379787eaaSJeff Mahoney ret); 37043a45bb20SJeff Mahoney btrfs_end_transaction(trans); 370579787eaaSJeff Mahoney goto out; 370679787eaaSJeff Mahoney } 3707c5c9cd4dSSage Weil } 3708c5c9cd4dSSage Weil 3709b3b4aa74SDavid Sterba btrfs_release_path(path); 3710c5c9cd4dSSage Weil 371162e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 37120b246afaSJeff Mahoney fs_info->sectorsize); 3713f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3714f82a9901SFilipe Manana last_dest_end, 37151c919a5eSMark Fasheh destoff, olen, 37161c919a5eSMark Fasheh no_time_update); 3717f82a9901SFilipe Manana if (ret) 371879787eaaSJeff Mahoney goto out; 37192c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 37202c463823SFilipe Manana break; 3721a22285a6SYan, Zheng } 3722b3b4aa74SDavid Sterba btrfs_release_path(path); 3723df858e76SFilipe Manana key.offset = next_key_min_offset; 372469ae5e44SWang Xiaoguang 372569ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 372669ae5e44SWang Xiaoguang ret = -EINTR; 372769ae5e44SWang Xiaoguang goto out; 372869ae5e44SWang Xiaoguang } 3729ae01a0abSYan Zheng } 3730f46b5a66SChristoph Hellwig ret = 0; 373132b7c687SMark Fasheh 3732f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3733690a5dbfSFilipe Manana struct btrfs_clone_extent_info clone_info = { 0 }; 3734f82a9901SFilipe Manana /* 3735f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3736f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3737f82a9901SFilipe Manana */ 3738b3b4aa74SDavid Sterba btrfs_release_path(path); 3739690a5dbfSFilipe Manana path->leave_spinning = 0; 3740f82a9901SFilipe Manana 3741f82a9901SFilipe Manana /* 3742690a5dbfSFilipe Manana * We are dealing with a hole and our clone_info already has a 3743690a5dbfSFilipe Manana * disk_offset of 0, we only need to fill the data length and 3744690a5dbfSFilipe Manana * file offset. 3745f82a9901SFilipe Manana */ 3746690a5dbfSFilipe Manana clone_info.data_len = destoff + len - last_dest_end; 3747690a5dbfSFilipe Manana clone_info.file_offset = last_dest_end; 3748690a5dbfSFilipe Manana ret = btrfs_punch_hole_range(inode, path, 3749690a5dbfSFilipe Manana last_dest_end, destoff + len - 1, 3750690a5dbfSFilipe Manana &clone_info, &trans); 3751690a5dbfSFilipe Manana if (ret) 3752f82a9901SFilipe Manana goto out; 3753690a5dbfSFilipe Manana 3754f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 37551c919a5eSMark Fasheh destoff, olen, no_time_update); 3756f82a9901SFilipe Manana } 3757f82a9901SFilipe Manana 3758f46b5a66SChristoph Hellwig out: 375932b7c687SMark Fasheh btrfs_free_path(path); 376015351955SDavid Sterba kvfree(buf); 376132b7c687SMark Fasheh return ret; 376232b7c687SMark Fasheh } 376332b7c687SMark Fasheh 37643db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 376532b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 376632b7c687SMark Fasheh { 376754563d41SAl Viro struct inode *inode = file_inode(file); 37683db11b2eSZach Brown struct inode *src = file_inode(file_src); 37690b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 377032b7c687SMark Fasheh int ret; 377132b7c687SMark Fasheh u64 len = olen; 37720b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 377332b7c687SMark Fasheh 377432b7c687SMark Fasheh /* 377532b7c687SMark Fasheh * TODO: 377632b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 377732b7c687SMark Fasheh * decompress into destination's address_space (the file offset 377832b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 377932b7c687SMark Fasheh * otherwise reinsert) a subrange. 378000fdf13aSLiu Bo * 378100fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 378200fdf13aSLiu Bo * be either compressed or non-compressed. 378332b7c687SMark Fasheh */ 378432b7c687SMark Fasheh 3785ac765f83SFilipe Manana /* 378634a28e3dSFilipe Manana * VFS's generic_remap_file_range_prep() protects us from cloning the 378734a28e3dSFilipe Manana * eof block into the middle of a file, which would result in corruption 378834a28e3dSFilipe Manana * if the file size is not blocksize aligned. So we don't need to check 378934a28e3dSFilipe Manana * for that case here. 3790ac765f83SFilipe Manana */ 379134a28e3dSFilipe Manana if (off + len == src->i_size) 379232b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 379332b7c687SMark Fasheh 379432b7c687SMark Fasheh if (destoff > inode->i_size) { 3795f7fa1107SFilipe Manana const u64 wb_start = ALIGN_DOWN(inode->i_size, bs); 3796f7fa1107SFilipe Manana 379732b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 379832b7c687SMark Fasheh if (ret) 379934a28e3dSFilipe Manana return ret; 3800f7fa1107SFilipe Manana /* 3801f7fa1107SFilipe Manana * We may have truncated the last block if the inode's size is 3802f7fa1107SFilipe Manana * not sector size aligned, so we need to wait for writeback to 3803f7fa1107SFilipe Manana * complete before proceeding further, otherwise we can race 3804f7fa1107SFilipe Manana * with cloning and attempt to increment a reference to an 3805f7fa1107SFilipe Manana * extent that no longer exists (writeback completed right after 3806f7fa1107SFilipe Manana * we found the previous extent covering eof and before we 3807f7fa1107SFilipe Manana * attempted to increment its reference count). 3808f7fa1107SFilipe Manana */ 3809f7fa1107SFilipe Manana ret = btrfs_wait_ordered_range(inode, wb_start, 3810f7fa1107SFilipe Manana destoff - wb_start); 3811f7fa1107SFilipe Manana if (ret) 3812f7fa1107SFilipe Manana return ret; 381332b7c687SMark Fasheh } 381432b7c687SMark Fasheh 3815c125b8bfSFilipe Manana /* 3816d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3817d8b55242SFilipe Manana * source range to serialize with relocation. 3818c125b8bfSFilipe Manana */ 3819d8b55242SFilipe Manana btrfs_double_extent_lock(src, off, inode, destoff, len); 38201c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 3821d8b55242SFilipe Manana btrfs_double_extent_unlock(src, off, inode, destoff, len); 3822c125b8bfSFilipe Manana /* 3823c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3824c125b8bfSFilipe Manana * data immediately and not the previous data. 3825c125b8bfSFilipe Manana */ 382665bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 382709cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 382809cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 382934a28e3dSFilipe Manana 383034a28e3dSFilipe Manana return ret; 383134a28e3dSFilipe Manana } 383234a28e3dSFilipe Manana 383334a28e3dSFilipe Manana static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in, 383434a28e3dSFilipe Manana struct file *file_out, loff_t pos_out, 383534a28e3dSFilipe Manana loff_t *len, unsigned int remap_flags) 383634a28e3dSFilipe Manana { 383734a28e3dSFilipe Manana struct inode *inode_in = file_inode(file_in); 383834a28e3dSFilipe Manana struct inode *inode_out = file_inode(file_out); 383934a28e3dSFilipe Manana u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize; 384034a28e3dSFilipe Manana bool same_inode = inode_out == inode_in; 384134a28e3dSFilipe Manana u64 wb_len; 384234a28e3dSFilipe Manana int ret; 384334a28e3dSFilipe Manana 384434a28e3dSFilipe Manana if (!(remap_flags & REMAP_FILE_DEDUP)) { 384534a28e3dSFilipe Manana struct btrfs_root *root_out = BTRFS_I(inode_out)->root; 384634a28e3dSFilipe Manana 384734a28e3dSFilipe Manana if (btrfs_root_readonly(root_out)) 384834a28e3dSFilipe Manana return -EROFS; 384934a28e3dSFilipe Manana 385034a28e3dSFilipe Manana if (file_in->f_path.mnt != file_out->f_path.mnt || 385134a28e3dSFilipe Manana inode_in->i_sb != inode_out->i_sb) 385234a28e3dSFilipe Manana return -EXDEV; 385334a28e3dSFilipe Manana } 385434a28e3dSFilipe Manana 3855500710d3SFilipe Manana /* don't make the dst file partly checksummed */ 3856500710d3SFilipe Manana if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) != 3857500710d3SFilipe Manana (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) { 38587984ae52SGoldwyn Rodrigues return -EINVAL; 3859500710d3SFilipe Manana } 3860500710d3SFilipe Manana 386134a28e3dSFilipe Manana /* 386234a28e3dSFilipe Manana * Now that the inodes are locked, we need to start writeback ourselves 386334a28e3dSFilipe Manana * and can not rely on the writeback from the VFS's generic helper 386434a28e3dSFilipe Manana * generic_remap_file_range_prep() because: 386534a28e3dSFilipe Manana * 386634a28e3dSFilipe Manana * 1) For compression we must call filemap_fdatawrite_range() range 386734a28e3dSFilipe Manana * twice (btrfs_fdatawrite_range() does it for us), and the generic 386834a28e3dSFilipe Manana * helper only calls it once; 386934a28e3dSFilipe Manana * 387034a28e3dSFilipe Manana * 2) filemap_fdatawrite_range(), called by the generic helper only 387134a28e3dSFilipe Manana * waits for the writeback to complete, i.e. for IO to be done, and 387234a28e3dSFilipe Manana * not for the ordered extents to complete. We need to wait for them 387334a28e3dSFilipe Manana * to complete so that new file extent items are in the fs tree. 387434a28e3dSFilipe Manana */ 387534a28e3dSFilipe Manana if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP)) 387634a28e3dSFilipe Manana wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs); 387734a28e3dSFilipe Manana else 387834a28e3dSFilipe Manana wb_len = ALIGN(*len, bs); 387934a28e3dSFilipe Manana 388034a28e3dSFilipe Manana /* 388134a28e3dSFilipe Manana * Since we don't lock ranges, wait for ongoing lockless dio writes (as 388234a28e3dSFilipe Manana * any in progress could create its ordered extents after we wait for 388334a28e3dSFilipe Manana * existing ordered extents below). 388434a28e3dSFilipe Manana */ 388534a28e3dSFilipe Manana inode_dio_wait(inode_in); 388634a28e3dSFilipe Manana if (!same_inode) 388734a28e3dSFilipe Manana inode_dio_wait(inode_out); 388834a28e3dSFilipe Manana 3889a94d1d0cSQu Wenruo /* 3890a94d1d0cSQu Wenruo * Workaround to make sure NOCOW buffered write reach disk as NOCOW. 3891a94d1d0cSQu Wenruo * 3892a94d1d0cSQu Wenruo * Btrfs' back references do not have a block level granularity, they 3893a94d1d0cSQu Wenruo * work at the whole extent level. 3894a94d1d0cSQu Wenruo * NOCOW buffered write without data space reserved may not be able 3895a94d1d0cSQu Wenruo * to fall back to CoW due to lack of data space, thus could cause 3896a94d1d0cSQu Wenruo * data loss. 3897a94d1d0cSQu Wenruo * 3898a94d1d0cSQu Wenruo * Here we take a shortcut by flushing the whole inode, so that all 3899a94d1d0cSQu Wenruo * nocow write should reach disk as nocow before we increase the 3900a94d1d0cSQu Wenruo * reference of the extent. We could do better by only flushing NOCOW 3901a94d1d0cSQu Wenruo * data, but that needs extra accounting. 3902a94d1d0cSQu Wenruo * 3903a94d1d0cSQu Wenruo * Also we don't need to check ASYNC_EXTENT, as async extent will be 3904a94d1d0cSQu Wenruo * CoWed anyway, not affecting nocow part. 3905a94d1d0cSQu Wenruo */ 3906a94d1d0cSQu Wenruo ret = filemap_flush(inode_in->i_mapping); 3907a94d1d0cSQu Wenruo if (ret < 0) 3908a94d1d0cSQu Wenruo return ret; 3909a94d1d0cSQu Wenruo 391034a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs), 391134a28e3dSFilipe Manana wb_len); 391234a28e3dSFilipe Manana if (ret < 0) 39137984ae52SGoldwyn Rodrigues return ret; 391434a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs), 391534a28e3dSFilipe Manana wb_len); 391634a28e3dSFilipe Manana if (ret < 0) 39173db11b2eSZach Brown return ret; 39187984ae52SGoldwyn Rodrigues 39197984ae52SGoldwyn Rodrigues return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out, 39207984ae52SGoldwyn Rodrigues len, remap_flags); 39213db11b2eSZach Brown } 39223db11b2eSZach Brown 392342ec3d4cSDarrick J. Wong loff_t btrfs_remap_file_range(struct file *src_file, loff_t off, 392442ec3d4cSDarrick J. Wong struct file *dst_file, loff_t destoff, loff_t len, 39252e5dfc99SDarrick J. Wong unsigned int remap_flags) 39263db11b2eSZach Brown { 392734a28e3dSFilipe Manana struct inode *src_inode = file_inode(src_file); 392834a28e3dSFilipe Manana struct inode *dst_inode = file_inode(dst_file); 392934a28e3dSFilipe Manana bool same_inode = dst_inode == src_inode; 393042ec3d4cSDarrick J. Wong int ret; 393142ec3d4cSDarrick J. Wong 39322e5dfc99SDarrick J. Wong if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY)) 39332e5dfc99SDarrick J. Wong return -EINVAL; 39342e5dfc99SDarrick J. Wong 39357984ae52SGoldwyn Rodrigues if (same_inode) 39367984ae52SGoldwyn Rodrigues inode_lock(src_inode); 39377984ae52SGoldwyn Rodrigues else 39387984ae52SGoldwyn Rodrigues lock_two_nondirectories(src_inode, dst_inode); 39397984ae52SGoldwyn Rodrigues 394034a28e3dSFilipe Manana ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff, 394134a28e3dSFilipe Manana &len, remap_flags); 394234a28e3dSFilipe Manana if (ret < 0 || len == 0) 39437984ae52SGoldwyn Rodrigues goto out_unlock; 39442e5dfc99SDarrick J. Wong 394534a28e3dSFilipe Manana if (remap_flags & REMAP_FILE_DEDUP) 394634a28e3dSFilipe Manana ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff); 394734a28e3dSFilipe Manana else 394842ec3d4cSDarrick J. Wong ret = btrfs_clone_files(dst_file, src_file, off, len, destoff); 394934a28e3dSFilipe Manana 39507984ae52SGoldwyn Rodrigues out_unlock: 395134a28e3dSFilipe Manana if (same_inode) 395234a28e3dSFilipe Manana inode_unlock(src_inode); 395334a28e3dSFilipe Manana else 39544ea748e1SFilipe Manana unlock_two_nondirectories(src_inode, dst_inode); 395534a28e3dSFilipe Manana 395642ec3d4cSDarrick J. Wong return ret < 0 ? ret : len; 3957c5c9cd4dSSage Weil } 3958c5c9cd4dSSage Weil 39596ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 39606ef5ed0dSJosef Bacik { 3961496ad9aaSAl Viro struct inode *inode = file_inode(file); 39620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 39636ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 39646ef5ed0dSJosef Bacik struct btrfs_root *new_root; 39656ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 39666ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 39676ef5ed0dSJosef Bacik struct btrfs_path *path; 39686ef5ed0dSJosef Bacik struct btrfs_key location; 39696ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 39706ef5ed0dSJosef Bacik u64 objectid = 0; 39716ef5ed0dSJosef Bacik u64 dir_id; 39723c04ce01SMiao Xie int ret; 39736ef5ed0dSJosef Bacik 39746ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 39756ef5ed0dSJosef Bacik return -EPERM; 39766ef5ed0dSJosef Bacik 39773c04ce01SMiao Xie ret = mnt_want_write_file(file); 39783c04ce01SMiao Xie if (ret) 39793c04ce01SMiao Xie return ret; 39803c04ce01SMiao Xie 39813c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 39823c04ce01SMiao Xie ret = -EFAULT; 39833c04ce01SMiao Xie goto out; 39843c04ce01SMiao Xie } 39856ef5ed0dSJosef Bacik 39866ef5ed0dSJosef Bacik if (!objectid) 39871cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 39886ef5ed0dSJosef Bacik 39896ef5ed0dSJosef Bacik location.objectid = objectid; 39906ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 39916ef5ed0dSJosef Bacik location.offset = (u64)-1; 39926ef5ed0dSJosef Bacik 39930b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 39943c04ce01SMiao Xie if (IS_ERR(new_root)) { 39953c04ce01SMiao Xie ret = PTR_ERR(new_root); 39963c04ce01SMiao Xie goto out; 39973c04ce01SMiao Xie } 39984fd786e6SMisono Tomohiro if (!is_fstree(new_root->root_key.objectid)) { 39996d6d2829Ssatoru takeuchi ret = -ENOENT; 40006d6d2829Ssatoru takeuchi goto out; 40016d6d2829Ssatoru takeuchi } 40026ef5ed0dSJosef Bacik 40036ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40043c04ce01SMiao Xie if (!path) { 40053c04ce01SMiao Xie ret = -ENOMEM; 40063c04ce01SMiao Xie goto out; 40073c04ce01SMiao Xie } 40086ef5ed0dSJosef Bacik path->leave_spinning = 1; 40096ef5ed0dSJosef Bacik 40106ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 401198d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40126ef5ed0dSJosef Bacik btrfs_free_path(path); 40133c04ce01SMiao Xie ret = PTR_ERR(trans); 40143c04ce01SMiao Xie goto out; 40156ef5ed0dSJosef Bacik } 40166ef5ed0dSJosef Bacik 40170b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 40180b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 40196ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4020cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40216ef5ed0dSJosef Bacik btrfs_free_path(path); 40223a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40230b246afaSJeff Mahoney btrfs_err(fs_info, 40245d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 40253c04ce01SMiao Xie ret = -ENOENT; 40263c04ce01SMiao Xie goto out; 40276ef5ed0dSJosef Bacik } 40286ef5ed0dSJosef Bacik 40296ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 40306ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 40316ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 40326ef5ed0dSJosef Bacik btrfs_free_path(path); 40336ef5ed0dSJosef Bacik 40340b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 40353a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40363c04ce01SMiao Xie out: 40373c04ce01SMiao Xie mnt_drop_write_file(file); 40383c04ce01SMiao Xie return ret; 40396ef5ed0dSJosef Bacik } 40406ef5ed0dSJosef Bacik 4041c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4042bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4043bf5fc093SJosef Bacik { 4044bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4045bf5fc093SJosef Bacik 4046bf5fc093SJosef Bacik space->total_bytes = 0; 4047bf5fc093SJosef Bacik space->used_bytes = 0; 4048bf5fc093SJosef Bacik space->flags = 0; 4049bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4050bf5fc093SJosef Bacik space->flags = block_group->flags; 4051bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4052bf5fc093SJosef Bacik space->used_bytes += 4053bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4054bf5fc093SJosef Bacik } 4055bf5fc093SJosef Bacik } 4056bf5fc093SJosef Bacik 40572ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 40582ff7e61eSJeff Mahoney void __user *arg) 40591406e432SJosef Bacik { 40601406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 40611406e432SJosef Bacik struct btrfs_ioctl_space_info space; 40621406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 40637fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 406413f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 40651406e432SJosef Bacik struct btrfs_space_info *info; 4066315d8e98SColin Ian King static const u64 types[] = { 4067315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4068bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4069bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4070315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4071315d8e98SColin Ian King }; 4072bf5fc093SJosef Bacik int num_types = 4; 40737fde62bfSChris Mason int alloc_size; 40741406e432SJosef Bacik int ret = 0; 407551788b1bSDan Rosenberg u64 slot_count = 0; 4076bf5fc093SJosef Bacik int i, c; 40771406e432SJosef Bacik 40781406e432SJosef Bacik if (copy_from_user(&space_args, 40791406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 40801406e432SJosef Bacik sizeof(space_args))) 40811406e432SJosef Bacik return -EFAULT; 40821406e432SJosef Bacik 4083bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4084bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4085bf5fc093SJosef Bacik 4086bf5fc093SJosef Bacik info = NULL; 40877fde62bfSChris Mason rcu_read_lock(); 40880b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4089bf5fc093SJosef Bacik list) { 4090bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4091bf5fc093SJosef Bacik info = tmp; 4092bf5fc093SJosef Bacik break; 4093bf5fc093SJosef Bacik } 4094bf5fc093SJosef Bacik } 40957fde62bfSChris Mason rcu_read_unlock(); 40961406e432SJosef Bacik 4097bf5fc093SJosef Bacik if (!info) 4098bf5fc093SJosef Bacik continue; 4099bf5fc093SJosef Bacik 4100bf5fc093SJosef Bacik down_read(&info->groups_sem); 4101bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4102bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4103bf5fc093SJosef Bacik slot_count++; 4104bf5fc093SJosef Bacik } 4105bf5fc093SJosef Bacik up_read(&info->groups_sem); 4106bf5fc093SJosef Bacik } 4107bf5fc093SJosef Bacik 410836523e95SDavid Sterba /* 410936523e95SDavid Sterba * Global block reserve, exported as a space_info 411036523e95SDavid Sterba */ 411136523e95SDavid Sterba slot_count++; 411236523e95SDavid Sterba 41137fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 41147fde62bfSChris Mason if (space_args.space_slots == 0) { 41157fde62bfSChris Mason space_args.total_spaces = slot_count; 41167fde62bfSChris Mason goto out; 41177fde62bfSChris Mason } 4118bf5fc093SJosef Bacik 411951788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4120bf5fc093SJosef Bacik 41217fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4122bf5fc093SJosef Bacik 41237fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41247fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41257fde62bfSChris Mason */ 412609cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 41277fde62bfSChris Mason return -ENOMEM; 41287fde62bfSChris Mason 41297fde62bfSChris Mason space_args.total_spaces = 0; 41308d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 41317fde62bfSChris Mason if (!dest) 41327fde62bfSChris Mason return -ENOMEM; 41337fde62bfSChris Mason dest_orig = dest; 41347fde62bfSChris Mason 41357fde62bfSChris Mason /* now we have a buffer to copy into */ 4136bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4137bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4138bf5fc093SJosef Bacik 413951788b1bSDan Rosenberg if (!slot_count) 414051788b1bSDan Rosenberg break; 414151788b1bSDan Rosenberg 4142bf5fc093SJosef Bacik info = NULL; 41431406e432SJosef Bacik rcu_read_lock(); 41440b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4145bf5fc093SJosef Bacik list) { 4146bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4147bf5fc093SJosef Bacik info = tmp; 41487fde62bfSChris Mason break; 4149bf5fc093SJosef Bacik } 4150bf5fc093SJosef Bacik } 4151bf5fc093SJosef Bacik rcu_read_unlock(); 41527fde62bfSChris Mason 4153bf5fc093SJosef Bacik if (!info) 4154bf5fc093SJosef Bacik continue; 4155bf5fc093SJosef Bacik down_read(&info->groups_sem); 4156bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4157bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4158c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4159c065f5b1SSu Yue &space); 41607fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 41611406e432SJosef Bacik dest++; 41621406e432SJosef Bacik space_args.total_spaces++; 416351788b1bSDan Rosenberg slot_count--; 41641406e432SJosef Bacik } 416551788b1bSDan Rosenberg if (!slot_count) 416651788b1bSDan Rosenberg break; 4167bf5fc093SJosef Bacik } 4168bf5fc093SJosef Bacik up_read(&info->groups_sem); 4169bf5fc093SJosef Bacik } 41701406e432SJosef Bacik 417136523e95SDavid Sterba /* 417236523e95SDavid Sterba * Add global block reserve 417336523e95SDavid Sterba */ 417436523e95SDavid Sterba if (slot_count) { 41750b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 417636523e95SDavid Sterba 417736523e95SDavid Sterba spin_lock(&block_rsv->lock); 417836523e95SDavid Sterba space.total_bytes = block_rsv->size; 417936523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 418036523e95SDavid Sterba spin_unlock(&block_rsv->lock); 418136523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 418236523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 418336523e95SDavid Sterba space_args.total_spaces++; 418436523e95SDavid Sterba } 418536523e95SDavid Sterba 41862eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 41877fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 41887fde62bfSChris Mason 41897fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 41907fde62bfSChris Mason ret = -EFAULT; 41917fde62bfSChris Mason 41927fde62bfSChris Mason kfree(dest_orig); 41937fde62bfSChris Mason out: 41947fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 41951406e432SJosef Bacik ret = -EFAULT; 41961406e432SJosef Bacik 41971406e432SJosef Bacik return ret; 41981406e432SJosef Bacik } 41991406e432SJosef Bacik 42009a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42019a8c28beSMiao Xie void __user *argp) 420246204592SSage Weil { 420346204592SSage Weil struct btrfs_trans_handle *trans; 420446204592SSage Weil u64 transid; 4205db5b493aSTsutomu Itoh int ret; 420646204592SSage Weil 4207d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4208ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4209ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 421098d5dc13STsutomu Itoh return PTR_ERR(trans); 4211ff7c1d33SMiao Xie 4212ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4213ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4214ff7c1d33SMiao Xie goto out; 4215ff7c1d33SMiao Xie } 421646204592SSage Weil transid = trans->transid; 42173a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 42188b2b2d3cSTsutomu Itoh if (ret) { 42193a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4220db5b493aSTsutomu Itoh return ret; 42218b2b2d3cSTsutomu Itoh } 4222ff7c1d33SMiao Xie out: 422346204592SSage Weil if (argp) 422446204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 422546204592SSage Weil return -EFAULT; 422646204592SSage Weil return 0; 422746204592SSage Weil } 422846204592SSage Weil 42292ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 42309a8c28beSMiao Xie void __user *argp) 423146204592SSage Weil { 423246204592SSage Weil u64 transid; 423346204592SSage Weil 423446204592SSage Weil if (argp) { 423546204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 423646204592SSage Weil return -EFAULT; 423746204592SSage Weil } else { 423846204592SSage Weil transid = 0; /* current trans */ 423946204592SSage Weil } 42402ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 424146204592SSage Weil } 424246204592SSage Weil 4243b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4244475f6387SJan Schmidt { 42450b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4246475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4247b8e95489SMiao Xie int ret; 4248475f6387SJan Schmidt 4249475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4250475f6387SJan Schmidt return -EPERM; 4251475f6387SJan Schmidt 4252475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4253475f6387SJan Schmidt if (IS_ERR(sa)) 4254475f6387SJan Schmidt return PTR_ERR(sa); 4255475f6387SJan Schmidt 4256b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4257b8e95489SMiao Xie ret = mnt_want_write_file(file); 4258b8e95489SMiao Xie if (ret) 4259b8e95489SMiao Xie goto out; 4260b8e95489SMiao Xie } 4261b8e95489SMiao Xie 42620b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 426363a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 426463a212abSStefan Behrens 0); 4265475f6387SJan Schmidt 426606fe39abSFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4267475f6387SJan Schmidt ret = -EFAULT; 4268475f6387SJan Schmidt 4269b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4270b8e95489SMiao Xie mnt_drop_write_file(file); 4271b8e95489SMiao Xie out: 4272475f6387SJan Schmidt kfree(sa); 4273475f6387SJan Schmidt return ret; 4274475f6387SJan Schmidt } 4275475f6387SJan Schmidt 42762ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4277475f6387SJan Schmidt { 4278475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4279475f6387SJan Schmidt return -EPERM; 4280475f6387SJan Schmidt 42812ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4282475f6387SJan Schmidt } 4283475f6387SJan Schmidt 42842ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4285475f6387SJan Schmidt void __user *arg) 4286475f6387SJan Schmidt { 4287475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4288475f6387SJan Schmidt int ret; 4289475f6387SJan Schmidt 4290475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4291475f6387SJan Schmidt return -EPERM; 4292475f6387SJan Schmidt 4293475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4294475f6387SJan Schmidt if (IS_ERR(sa)) 4295475f6387SJan Schmidt return PTR_ERR(sa); 4296475f6387SJan Schmidt 42972ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4298475f6387SJan Schmidt 42994fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4300475f6387SJan Schmidt ret = -EFAULT; 4301475f6387SJan Schmidt 4302475f6387SJan Schmidt kfree(sa); 4303475f6387SJan Schmidt return ret; 4304475f6387SJan Schmidt } 4305475f6387SJan Schmidt 43062ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4307b27f7c0cSDavid Sterba void __user *arg) 4308c11d2c23SStefan Behrens { 4309c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4310c11d2c23SStefan Behrens int ret; 4311c11d2c23SStefan Behrens 4312c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4313c11d2c23SStefan Behrens if (IS_ERR(sa)) 4314c11d2c23SStefan Behrens return PTR_ERR(sa); 4315c11d2c23SStefan Behrens 4316b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4317b27f7c0cSDavid Sterba kfree(sa); 4318b27f7c0cSDavid Sterba return -EPERM; 4319b27f7c0cSDavid Sterba } 4320b27f7c0cSDavid Sterba 43212ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4322c11d2c23SStefan Behrens 4323eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4324c11d2c23SStefan Behrens ret = -EFAULT; 4325c11d2c23SStefan Behrens 4326c11d2c23SStefan Behrens kfree(sa); 4327c11d2c23SStefan Behrens return ret; 4328c11d2c23SStefan Behrens } 4329c11d2c23SStefan Behrens 43302ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 43312ff7e61eSJeff Mahoney void __user *arg) 43323f6bcfbdSStefan Behrens { 43333f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 43343f6bcfbdSStefan Behrens int ret; 43353f6bcfbdSStefan Behrens 43363f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 43373f6bcfbdSStefan Behrens return -EPERM; 43383f6bcfbdSStefan Behrens 43393f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 43403f6bcfbdSStefan Behrens if (IS_ERR(p)) 43413f6bcfbdSStefan Behrens return PTR_ERR(p); 43423f6bcfbdSStefan Behrens 43433f6bcfbdSStefan Behrens switch (p->cmd) { 43443f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4345bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4346adfa97cbSIlya Dryomov ret = -EROFS; 4347adfa97cbSIlya Dryomov goto out; 4348adfa97cbSIlya Dryomov } 4349171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4350e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 43513f6bcfbdSStefan Behrens } else { 43522ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4353171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 43543f6bcfbdSStefan Behrens } 43553f6bcfbdSStefan Behrens break; 43563f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 43570b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 43583f6bcfbdSStefan Behrens ret = 0; 43593f6bcfbdSStefan Behrens break; 43603f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 436117d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 436297282031SAnand Jain ret = 0; 43633f6bcfbdSStefan Behrens break; 43643f6bcfbdSStefan Behrens default: 43653f6bcfbdSStefan Behrens ret = -EINVAL; 43663f6bcfbdSStefan Behrens break; 43673f6bcfbdSStefan Behrens } 43683f6bcfbdSStefan Behrens 4369d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 43703f6bcfbdSStefan Behrens ret = -EFAULT; 4371adfa97cbSIlya Dryomov out: 43723f6bcfbdSStefan Behrens kfree(p); 43733f6bcfbdSStefan Behrens return ret; 43743f6bcfbdSStefan Behrens } 43753f6bcfbdSStefan Behrens 4376d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4377d7728c96SJan Schmidt { 4378d7728c96SJan Schmidt int ret = 0; 4379d7728c96SJan Schmidt int i; 4380740c3d22SChris Mason u64 rel_ptr; 4381d7728c96SJan Schmidt int size; 4382806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4383d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4384d7728c96SJan Schmidt struct btrfs_path *path; 4385d7728c96SJan Schmidt 438682b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4387d7728c96SJan Schmidt return -EPERM; 4388d7728c96SJan Schmidt 4389d7728c96SJan Schmidt path = btrfs_alloc_path(); 4390d7728c96SJan Schmidt if (!path) { 4391d7728c96SJan Schmidt ret = -ENOMEM; 4392d7728c96SJan Schmidt goto out; 4393d7728c96SJan Schmidt } 4394d7728c96SJan Schmidt 4395d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4396d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4397d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4398d7728c96SJan Schmidt ipa = NULL; 4399d7728c96SJan Schmidt goto out; 4400d7728c96SJan Schmidt } 4401d7728c96SJan Schmidt 4402d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4403d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4404d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4405d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4406d7728c96SJan Schmidt ipath = NULL; 4407d7728c96SJan Schmidt goto out; 4408d7728c96SJan Schmidt } 4409d7728c96SJan Schmidt 4410d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4411d7728c96SJan Schmidt if (ret < 0) 4412d7728c96SJan Schmidt goto out; 4413d7728c96SJan Schmidt 4414d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4415745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4416745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4417740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4418d7728c96SJan Schmidt } 4419d7728c96SJan Schmidt 4420718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4421718dc5faSOmar Sandoval ipath->fspath, size); 4422d7728c96SJan Schmidt if (ret) { 4423d7728c96SJan Schmidt ret = -EFAULT; 4424d7728c96SJan Schmidt goto out; 4425d7728c96SJan Schmidt } 4426d7728c96SJan Schmidt 4427d7728c96SJan Schmidt out: 4428d7728c96SJan Schmidt btrfs_free_path(path); 4429d7728c96SJan Schmidt free_ipath(ipath); 4430d7728c96SJan Schmidt kfree(ipa); 4431d7728c96SJan Schmidt 4432d7728c96SJan Schmidt return ret; 4433d7728c96SJan Schmidt } 4434d7728c96SJan Schmidt 4435d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4436d7728c96SJan Schmidt { 4437d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4438d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4439d7728c96SJan Schmidt 4440d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4441d7728c96SJan Schmidt inodes->bytes_left -= c; 4442d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4443d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4444d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4445d7728c96SJan Schmidt inodes->elem_cnt += 3; 4446d7728c96SJan Schmidt } else { 4447d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4448d7728c96SJan Schmidt inodes->bytes_left = 0; 4449d7728c96SJan Schmidt inodes->elem_missed += 3; 4450d7728c96SJan Schmidt } 4451d7728c96SJan Schmidt 4452d7728c96SJan Schmidt return 0; 4453d7728c96SJan Schmidt } 4454d7728c96SJan Schmidt 44552ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4456d24a67b2SZygo Blaxell void __user *arg, int version) 4457d7728c96SJan Schmidt { 4458d7728c96SJan Schmidt int ret = 0; 4459d7728c96SJan Schmidt int size; 4460d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4461d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4462d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4463d24a67b2SZygo Blaxell bool ignore_offset; 4464d7728c96SJan Schmidt 4465d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4466d7728c96SJan Schmidt return -EPERM; 4467d7728c96SJan Schmidt 4468d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 44697b9ea627SShailendra Verma if (IS_ERR(loi)) 44707b9ea627SShailendra Verma return PTR_ERR(loi); 4471d7728c96SJan Schmidt 4472d24a67b2SZygo Blaxell if (version == 1) { 4473d24a67b2SZygo Blaxell ignore_offset = false; 4474b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4475d24a67b2SZygo Blaxell } else { 4476d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4477d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4478d24a67b2SZygo Blaxell ret = -EINVAL; 4479d24a67b2SZygo Blaxell goto out_loi; 4480d24a67b2SZygo Blaxell } 4481d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4482d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4483d24a67b2SZygo Blaxell ret = -EINVAL; 4484d24a67b2SZygo Blaxell goto out_loi; 4485d24a67b2SZygo Blaxell } 4486d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4487b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4488d24a67b2SZygo Blaxell } 4489d24a67b2SZygo Blaxell 4490d7728c96SJan Schmidt path = btrfs_alloc_path(); 4491d7728c96SJan Schmidt if (!path) { 4492d7728c96SJan Schmidt ret = -ENOMEM; 4493d7728c96SJan Schmidt goto out; 4494d7728c96SJan Schmidt } 4495d7728c96SJan Schmidt 4496d7728c96SJan Schmidt inodes = init_data_container(size); 4497d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4498d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4499d7728c96SJan Schmidt inodes = NULL; 4500d7728c96SJan Schmidt goto out; 4501d7728c96SJan Schmidt } 4502d7728c96SJan Schmidt 45032ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4504d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4505df031f07SLiu Bo if (ret == -EINVAL) 4506d7728c96SJan Schmidt ret = -ENOENT; 4507d7728c96SJan Schmidt if (ret < 0) 4508d7728c96SJan Schmidt goto out; 4509d7728c96SJan Schmidt 4510718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4511718dc5faSOmar Sandoval size); 4512d7728c96SJan Schmidt if (ret) 4513d7728c96SJan Schmidt ret = -EFAULT; 4514d7728c96SJan Schmidt 4515d7728c96SJan Schmidt out: 4516d7728c96SJan Schmidt btrfs_free_path(path); 4517f54de068SDavid Sterba kvfree(inodes); 4518d24a67b2SZygo Blaxell out_loi: 4519d7728c96SJan Schmidt kfree(loi); 4520d7728c96SJan Schmidt 4521d7728c96SJan Schmidt return ret; 4522d7728c96SJan Schmidt } 4523d7728c96SJan Schmidt 4524008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4525c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4526c9e9f97bSIlya Dryomov { 4527c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4528c9e9f97bSIlya Dryomov 4529c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4530c9e9f97bSIlya Dryomov 45313009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4532837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4533837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4534837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4535a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4536a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4537837d5b6eSIlya Dryomov 4538c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4539c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4540c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 454119a39dceSIlya Dryomov 454219a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 454319a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 454419a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4545c9e9f97bSIlya Dryomov } 4546c9e9f97bSIlya Dryomov 45479ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4548c9e9f97bSIlya Dryomov { 4549496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4550c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4551c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4552c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4553ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4554c9e9f97bSIlya Dryomov int ret; 4555c9e9f97bSIlya Dryomov 4556c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4557c9e9f97bSIlya Dryomov return -EPERM; 4558c9e9f97bSIlya Dryomov 4559e54bfa31SLiu Bo ret = mnt_want_write_file(file); 45609ba1f6e4SLiu Bo if (ret) 45619ba1f6e4SLiu Bo return ret; 45629ba1f6e4SLiu Bo 4563ed0fb78fSIlya Dryomov again: 4564171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4565c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4566ed0fb78fSIlya Dryomov need_unlock = true; 4567ed0fb78fSIlya Dryomov goto locked; 4568ed0fb78fSIlya Dryomov } 4569ed0fb78fSIlya Dryomov 4570ed0fb78fSIlya Dryomov /* 457101327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4572ed0fb78fSIlya Dryomov * (1) some other op is running 4573ed0fb78fSIlya Dryomov * (2) balance is running 4574ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4575ed0fb78fSIlya Dryomov */ 4576ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4577ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4578ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 45793009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4580ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4581dccdb07bSDavid Sterba /* 4582dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4583dccdb07bSDavid Sterba * we'll reexamine the status again. 4584dccdb07bSDavid Sterba */ 4585ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4586ed0fb78fSIlya Dryomov 4587ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 45883009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4589ed0fb78fSIlya Dryomov /* this is (3) */ 4590ed0fb78fSIlya Dryomov need_unlock = false; 4591ed0fb78fSIlya Dryomov goto locked; 4592ed0fb78fSIlya Dryomov } 4593ed0fb78fSIlya Dryomov 4594ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4595ed0fb78fSIlya Dryomov goto again; 4596ed0fb78fSIlya Dryomov } else { 4597ed0fb78fSIlya Dryomov /* this is (2) */ 4598ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4599ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4600ed0fb78fSIlya Dryomov goto out; 4601ed0fb78fSIlya Dryomov } 4602ed0fb78fSIlya Dryomov } else { 4603ed0fb78fSIlya Dryomov /* this is (1) */ 4604ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4605e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4606ed0fb78fSIlya Dryomov goto out; 4607ed0fb78fSIlya Dryomov } 4608ed0fb78fSIlya Dryomov 4609ed0fb78fSIlya Dryomov locked: 4610171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4611c9e9f97bSIlya Dryomov 4612c9e9f97bSIlya Dryomov if (arg) { 4613c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4614c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4615c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4616ed0fb78fSIlya Dryomov goto out_unlock; 4617c9e9f97bSIlya Dryomov } 4618de322263SIlya Dryomov 4619de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4620de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4621de322263SIlya Dryomov ret = -ENOTCONN; 4622de322263SIlya Dryomov goto out_bargs; 4623de322263SIlya Dryomov } 4624de322263SIlya Dryomov 4625de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4626de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4627de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4628de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4629de322263SIlya Dryomov 4630de322263SIlya Dryomov goto do_balance; 4631de322263SIlya Dryomov } 4632c9e9f97bSIlya Dryomov } else { 4633c9e9f97bSIlya Dryomov bargs = NULL; 4634c9e9f97bSIlya Dryomov } 4635c9e9f97bSIlya Dryomov 4636ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4637837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4638837d5b6eSIlya Dryomov goto out_bargs; 4639837d5b6eSIlya Dryomov } 4640837d5b6eSIlya Dryomov 46418d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4642c9e9f97bSIlya Dryomov if (!bctl) { 4643c9e9f97bSIlya Dryomov ret = -ENOMEM; 4644c9e9f97bSIlya Dryomov goto out_bargs; 4645c9e9f97bSIlya Dryomov } 4646c9e9f97bSIlya Dryomov 4647c9e9f97bSIlya Dryomov if (arg) { 4648c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4649c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4650c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4651c9e9f97bSIlya Dryomov 4652c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4653f43ffb60SIlya Dryomov } else { 4654f43ffb60SIlya Dryomov /* balance everything - no filters */ 4655f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4656c9e9f97bSIlya Dryomov } 4657c9e9f97bSIlya Dryomov 46588eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 46598eb93459SDavid Sterba ret = -EINVAL; 46600f89abf5SChristian Engelmayer goto out_bctl; 46618eb93459SDavid Sterba } 46628eb93459SDavid Sterba 4663de322263SIlya Dryomov do_balance: 4664c9e9f97bSIlya Dryomov /* 4665149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4666149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4667149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4668149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4669c9e9f97bSIlya Dryomov */ 4670ed0fb78fSIlya Dryomov need_unlock = false; 4671ed0fb78fSIlya Dryomov 46726fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 46730f89abf5SChristian Engelmayer bctl = NULL; 4674ed0fb78fSIlya Dryomov 4675d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4676c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4677c9e9f97bSIlya Dryomov ret = -EFAULT; 4678c9e9f97bSIlya Dryomov } 4679c9e9f97bSIlya Dryomov 46800f89abf5SChristian Engelmayer out_bctl: 46810f89abf5SChristian Engelmayer kfree(bctl); 4682c9e9f97bSIlya Dryomov out_bargs: 4683c9e9f97bSIlya Dryomov kfree(bargs); 4684ed0fb78fSIlya Dryomov out_unlock: 4685c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4686ed0fb78fSIlya Dryomov if (need_unlock) 4687171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4688ed0fb78fSIlya Dryomov out: 4689e54bfa31SLiu Bo mnt_drop_write_file(file); 4690c9e9f97bSIlya Dryomov return ret; 4691c9e9f97bSIlya Dryomov } 4692c9e9f97bSIlya Dryomov 46932ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4694837d5b6eSIlya Dryomov { 4695837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4696837d5b6eSIlya Dryomov return -EPERM; 4697837d5b6eSIlya Dryomov 4698837d5b6eSIlya Dryomov switch (cmd) { 4699837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 47000b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4701a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47020b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4703837d5b6eSIlya Dryomov } 4704837d5b6eSIlya Dryomov 4705837d5b6eSIlya Dryomov return -EINVAL; 4706837d5b6eSIlya Dryomov } 4707837d5b6eSIlya Dryomov 47082ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 470919a39dceSIlya Dryomov void __user *arg) 471019a39dceSIlya Dryomov { 471119a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 471219a39dceSIlya Dryomov int ret = 0; 471319a39dceSIlya Dryomov 471419a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 471519a39dceSIlya Dryomov return -EPERM; 471619a39dceSIlya Dryomov 471719a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 471819a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 471919a39dceSIlya Dryomov ret = -ENOTCONN; 472019a39dceSIlya Dryomov goto out; 472119a39dceSIlya Dryomov } 472219a39dceSIlya Dryomov 47238d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 472419a39dceSIlya Dryomov if (!bargs) { 472519a39dceSIlya Dryomov ret = -ENOMEM; 472619a39dceSIlya Dryomov goto out; 472719a39dceSIlya Dryomov } 472819a39dceSIlya Dryomov 4729008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 473019a39dceSIlya Dryomov 473119a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 473219a39dceSIlya Dryomov ret = -EFAULT; 473319a39dceSIlya Dryomov 473419a39dceSIlya Dryomov kfree(bargs); 473519a39dceSIlya Dryomov out: 473619a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 473719a39dceSIlya Dryomov return ret; 473819a39dceSIlya Dryomov } 473919a39dceSIlya Dryomov 4740905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 47415d13a37bSArne Jansen { 47420b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47430b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47445d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 47455d13a37bSArne Jansen int ret; 47465d13a37bSArne Jansen 47475d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47485d13a37bSArne Jansen return -EPERM; 47495d13a37bSArne Jansen 4750905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4751905b0ddaSMiao Xie if (ret) 4752905b0ddaSMiao Xie return ret; 47535d13a37bSArne Jansen 47545d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4755905b0ddaSMiao Xie if (IS_ERR(sa)) { 4756905b0ddaSMiao Xie ret = PTR_ERR(sa); 4757905b0ddaSMiao Xie goto drop_write; 4758905b0ddaSMiao Xie } 47595d13a37bSArne Jansen 47600b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 47615d13a37bSArne Jansen 47625d13a37bSArne Jansen switch (sa->cmd) { 47635d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4764340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 47655d13a37bSArne Jansen break; 47665d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4767340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 47685d13a37bSArne Jansen break; 47695d13a37bSArne Jansen default: 47705d13a37bSArne Jansen ret = -EINVAL; 47715d13a37bSArne Jansen break; 47725d13a37bSArne Jansen } 47735d13a37bSArne Jansen 47745d13a37bSArne Jansen kfree(sa); 47750b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4776905b0ddaSMiao Xie drop_write: 4777905b0ddaSMiao Xie mnt_drop_write_file(file); 47785d13a37bSArne Jansen return ret; 47795d13a37bSArne Jansen } 47805d13a37bSArne Jansen 4781905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 47825d13a37bSArne Jansen { 47830b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47840b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47850b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 47865d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 47875d13a37bSArne Jansen struct btrfs_trans_handle *trans; 47885d13a37bSArne Jansen int ret; 47895d13a37bSArne Jansen int err; 47905d13a37bSArne Jansen 47915d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47925d13a37bSArne Jansen return -EPERM; 47935d13a37bSArne Jansen 4794905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4795905b0ddaSMiao Xie if (ret) 4796905b0ddaSMiao Xie return ret; 47975d13a37bSArne Jansen 47985d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4799905b0ddaSMiao Xie if (IS_ERR(sa)) { 4800905b0ddaSMiao Xie ret = PTR_ERR(sa); 4801905b0ddaSMiao Xie goto drop_write; 4802905b0ddaSMiao Xie } 48035d13a37bSArne Jansen 48045d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48055d13a37bSArne Jansen if (IS_ERR(trans)) { 48065d13a37bSArne Jansen ret = PTR_ERR(trans); 48075d13a37bSArne Jansen goto out; 48085d13a37bSArne Jansen } 48095d13a37bSArne Jansen 48105d13a37bSArne Jansen if (sa->assign) { 48119f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 48125d13a37bSArne Jansen } else { 481339616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 48145d13a37bSArne Jansen } 48155d13a37bSArne Jansen 4816e082f563SQu Wenruo /* update qgroup status and info */ 4817280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4818e082f563SQu Wenruo if (err < 0) 48190b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4820ad8403dfSAnand Jain "failed to update qgroup status and info"); 48213a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48225d13a37bSArne Jansen if (err && !ret) 48235d13a37bSArne Jansen ret = err; 48245d13a37bSArne Jansen 48255d13a37bSArne Jansen out: 48265d13a37bSArne Jansen kfree(sa); 4827905b0ddaSMiao Xie drop_write: 4828905b0ddaSMiao Xie mnt_drop_write_file(file); 48295d13a37bSArne Jansen return ret; 48305d13a37bSArne Jansen } 48315d13a37bSArne Jansen 4832905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 48335d13a37bSArne Jansen { 48340b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48350b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48365d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 48375d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48385d13a37bSArne Jansen int ret; 48395d13a37bSArne Jansen int err; 48405d13a37bSArne Jansen 48415d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48425d13a37bSArne Jansen return -EPERM; 48435d13a37bSArne Jansen 4844905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4845905b0ddaSMiao Xie if (ret) 4846905b0ddaSMiao Xie return ret; 48475d13a37bSArne Jansen 48485d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4849905b0ddaSMiao Xie if (IS_ERR(sa)) { 4850905b0ddaSMiao Xie ret = PTR_ERR(sa); 4851905b0ddaSMiao Xie goto drop_write; 4852905b0ddaSMiao Xie } 48535d13a37bSArne Jansen 4854d86e56cfSMiao Xie if (!sa->qgroupid) { 4855d86e56cfSMiao Xie ret = -EINVAL; 4856d86e56cfSMiao Xie goto out; 4857d86e56cfSMiao Xie } 4858d86e56cfSMiao Xie 48595d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48605d13a37bSArne Jansen if (IS_ERR(trans)) { 48615d13a37bSArne Jansen ret = PTR_ERR(trans); 48625d13a37bSArne Jansen goto out; 48635d13a37bSArne Jansen } 48645d13a37bSArne Jansen 48655d13a37bSArne Jansen if (sa->create) { 486649a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 48675d13a37bSArne Jansen } else { 48683efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 48695d13a37bSArne Jansen } 48705d13a37bSArne Jansen 48713a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48725d13a37bSArne Jansen if (err && !ret) 48735d13a37bSArne Jansen ret = err; 48745d13a37bSArne Jansen 48755d13a37bSArne Jansen out: 48765d13a37bSArne Jansen kfree(sa); 4877905b0ddaSMiao Xie drop_write: 4878905b0ddaSMiao Xie mnt_drop_write_file(file); 48795d13a37bSArne Jansen return ret; 48805d13a37bSArne Jansen } 48815d13a37bSArne Jansen 4882905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 48835d13a37bSArne Jansen { 48840b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48850b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48865d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 48875d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48885d13a37bSArne Jansen int ret; 48895d13a37bSArne Jansen int err; 48905d13a37bSArne Jansen u64 qgroupid; 48915d13a37bSArne Jansen 48925d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48935d13a37bSArne Jansen return -EPERM; 48945d13a37bSArne Jansen 4895905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4896905b0ddaSMiao Xie if (ret) 4897905b0ddaSMiao Xie return ret; 48985d13a37bSArne Jansen 48995d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4900905b0ddaSMiao Xie if (IS_ERR(sa)) { 4901905b0ddaSMiao Xie ret = PTR_ERR(sa); 4902905b0ddaSMiao Xie goto drop_write; 4903905b0ddaSMiao Xie } 49045d13a37bSArne Jansen 49055d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49065d13a37bSArne Jansen if (IS_ERR(trans)) { 49075d13a37bSArne Jansen ret = PTR_ERR(trans); 49085d13a37bSArne Jansen goto out; 49095d13a37bSArne Jansen } 49105d13a37bSArne Jansen 49115d13a37bSArne Jansen qgroupid = sa->qgroupid; 49125d13a37bSArne Jansen if (!qgroupid) { 49135d13a37bSArne Jansen /* take the current subvol as qgroup */ 49145d13a37bSArne Jansen qgroupid = root->root_key.objectid; 49155d13a37bSArne Jansen } 49165d13a37bSArne Jansen 4917f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 49185d13a37bSArne Jansen 49193a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49205d13a37bSArne Jansen if (err && !ret) 49215d13a37bSArne Jansen ret = err; 49225d13a37bSArne Jansen 49235d13a37bSArne Jansen out: 49245d13a37bSArne Jansen kfree(sa); 4925905b0ddaSMiao Xie drop_write: 4926905b0ddaSMiao Xie mnt_drop_write_file(file); 49275d13a37bSArne Jansen return ret; 49285d13a37bSArne Jansen } 49295d13a37bSArne Jansen 49302f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 49312f232036SJan Schmidt { 49320b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49330b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49342f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49352f232036SJan Schmidt int ret; 49362f232036SJan Schmidt 49372f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49382f232036SJan Schmidt return -EPERM; 49392f232036SJan Schmidt 49402f232036SJan Schmidt ret = mnt_want_write_file(file); 49412f232036SJan Schmidt if (ret) 49422f232036SJan Schmidt return ret; 49432f232036SJan Schmidt 49442f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 49452f232036SJan Schmidt if (IS_ERR(qsa)) { 49462f232036SJan Schmidt ret = PTR_ERR(qsa); 49472f232036SJan Schmidt goto drop_write; 49482f232036SJan Schmidt } 49492f232036SJan Schmidt 49502f232036SJan Schmidt if (qsa->flags) { 49512f232036SJan Schmidt ret = -EINVAL; 49522f232036SJan Schmidt goto out; 49532f232036SJan Schmidt } 49542f232036SJan Schmidt 49550b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 49562f232036SJan Schmidt 49572f232036SJan Schmidt out: 49582f232036SJan Schmidt kfree(qsa); 49592f232036SJan Schmidt drop_write: 49602f232036SJan Schmidt mnt_drop_write_file(file); 49612f232036SJan Schmidt return ret; 49622f232036SJan Schmidt } 49632f232036SJan Schmidt 49642f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 49652f232036SJan Schmidt { 49660b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49670b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49682f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49692f232036SJan Schmidt int ret = 0; 49702f232036SJan Schmidt 49712f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49722f232036SJan Schmidt return -EPERM; 49732f232036SJan Schmidt 49748d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 49752f232036SJan Schmidt if (!qsa) 49762f232036SJan Schmidt return -ENOMEM; 49772f232036SJan Schmidt 49780b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 49792f232036SJan Schmidt qsa->flags = 1; 49800b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 49812f232036SJan Schmidt } 49822f232036SJan Schmidt 49832f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 49842f232036SJan Schmidt ret = -EFAULT; 49852f232036SJan Schmidt 49862f232036SJan Schmidt kfree(qsa); 49872f232036SJan Schmidt return ret; 49882f232036SJan Schmidt } 49892f232036SJan Schmidt 499057254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 499157254b6eSJan Schmidt { 49920b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 499457254b6eSJan Schmidt 499557254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 499657254b6eSJan Schmidt return -EPERM; 499757254b6eSJan Schmidt 49980b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 499957254b6eSJan Schmidt } 500057254b6eSJan Schmidt 5001abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5002abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50038ea05e3aSAlexander Block { 5004496ad9aaSAl Viro struct inode *inode = file_inode(file); 50050b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50068ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50078ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50088ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 500995582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 50108ea05e3aSAlexander Block int ret = 0; 5011dd5f9615SStefan Behrens int received_uuid_changed; 50128ea05e3aSAlexander Block 5013bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5014bd60ea0fSDavid Sterba return -EPERM; 5015bd60ea0fSDavid Sterba 50168ea05e3aSAlexander Block ret = mnt_want_write_file(file); 50178ea05e3aSAlexander Block if (ret < 0) 50188ea05e3aSAlexander Block return ret; 50198ea05e3aSAlexander Block 50200b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 50218ea05e3aSAlexander Block 50224a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 50238ea05e3aSAlexander Block ret = -EINVAL; 50248ea05e3aSAlexander Block goto out; 50258ea05e3aSAlexander Block } 50268ea05e3aSAlexander Block 50278ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 50288ea05e3aSAlexander Block ret = -EROFS; 50298ea05e3aSAlexander Block goto out; 50308ea05e3aSAlexander Block } 50318ea05e3aSAlexander Block 5032dd5f9615SStefan Behrens /* 5033dd5f9615SStefan Behrens * 1 - root item 5034dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5035dd5f9615SStefan Behrens */ 5036dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 50378ea05e3aSAlexander Block if (IS_ERR(trans)) { 50388ea05e3aSAlexander Block ret = PTR_ERR(trans); 50398ea05e3aSAlexander Block trans = NULL; 50408ea05e3aSAlexander Block goto out; 50418ea05e3aSAlexander Block } 50428ea05e3aSAlexander Block 50438ea05e3aSAlexander Block sa->rtransid = trans->transid; 50448ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 50458ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 50468ea05e3aSAlexander Block 5047dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5048dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5049dd5f9615SStefan Behrens if (received_uuid_changed && 5050d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5051d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5052dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5053dd5f9615SStefan Behrens root->root_key.objectid); 5054d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5055d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5056d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5057d87ff758SNikolay Borisov goto out; 5058d87ff758SNikolay Borisov } 5059d87ff758SNikolay Borisov } 50608ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 50618ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 50628ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 50633cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 50643cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 50653cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 50663cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 50678ea05e3aSAlexander Block 50680b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 50698ea05e3aSAlexander Block &root->root_key, &root->root_item); 50708ea05e3aSAlexander Block if (ret < 0) { 50713a45bb20SJeff Mahoney btrfs_end_transaction(trans); 50728ea05e3aSAlexander Block goto out; 5073dd5f9615SStefan Behrens } 5074dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5075cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5076dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5077dd5f9615SStefan Behrens root->root_key.objectid); 5078dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 507966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5080efd38150SNikolay Borisov btrfs_end_transaction(trans); 5081dd5f9615SStefan Behrens goto out; 5082dd5f9615SStefan Behrens } 5083dd5f9615SStefan Behrens } 50843a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5085abccd00fSHugo Mills out: 50860b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5087abccd00fSHugo Mills mnt_drop_write_file(file); 5088abccd00fSHugo Mills return ret; 5089abccd00fSHugo Mills } 5090abccd00fSHugo Mills 5091abccd00fSHugo Mills #ifdef CONFIG_64BIT 5092abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5093abccd00fSHugo Mills void __user *arg) 5094abccd00fSHugo Mills { 5095abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5096abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5097abccd00fSHugo Mills int ret = 0; 5098abccd00fSHugo Mills 5099abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 51007b9ea627SShailendra Verma if (IS_ERR(args32)) 51017b9ea627SShailendra Verma return PTR_ERR(args32); 5102abccd00fSHugo Mills 51038d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 510484dbeb87SDan Carpenter if (!args64) { 510584dbeb87SDan Carpenter ret = -ENOMEM; 5106abccd00fSHugo Mills goto out; 5107abccd00fSHugo Mills } 5108abccd00fSHugo Mills 5109abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5110abccd00fSHugo Mills args64->stransid = args32->stransid; 5111abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5112abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5113abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5114abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5115abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5116abccd00fSHugo Mills args64->flags = args32->flags; 5117abccd00fSHugo Mills 5118abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5119abccd00fSHugo Mills if (ret) 5120abccd00fSHugo Mills goto out; 5121abccd00fSHugo Mills 5122abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5123abccd00fSHugo Mills args32->stransid = args64->stransid; 5124abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5125abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5126abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5127abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5128abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5129abccd00fSHugo Mills args32->flags = args64->flags; 5130abccd00fSHugo Mills 5131abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5132abccd00fSHugo Mills if (ret) 5133abccd00fSHugo Mills ret = -EFAULT; 5134abccd00fSHugo Mills 5135abccd00fSHugo Mills out: 5136abccd00fSHugo Mills kfree(args32); 5137abccd00fSHugo Mills kfree(args64); 5138abccd00fSHugo Mills return ret; 5139abccd00fSHugo Mills } 5140abccd00fSHugo Mills #endif 5141abccd00fSHugo Mills 5142abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5143abccd00fSHugo Mills void __user *arg) 5144abccd00fSHugo Mills { 5145abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5146abccd00fSHugo Mills int ret = 0; 5147abccd00fSHugo Mills 5148abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 51497b9ea627SShailendra Verma if (IS_ERR(sa)) 51507b9ea627SShailendra Verma return PTR_ERR(sa); 5151abccd00fSHugo Mills 5152abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5153abccd00fSHugo Mills 5154abccd00fSHugo Mills if (ret) 5155abccd00fSHugo Mills goto out; 5156abccd00fSHugo Mills 51578ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 51588ea05e3aSAlexander Block if (ret) 51598ea05e3aSAlexander Block ret = -EFAULT; 51608ea05e3aSAlexander Block 51618ea05e3aSAlexander Block out: 51628ea05e3aSAlexander Block kfree(sa); 51638ea05e3aSAlexander Block return ret; 51648ea05e3aSAlexander Block } 51658ea05e3aSAlexander Block 5166867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5167867ab667Sjeff.liu { 51680b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51690b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5170a1b83ac5SAnand Jain size_t len; 5171867ab667Sjeff.liu int ret; 5172a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5173a1b83ac5SAnand Jain 51740b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 51750b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 51760b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5177a1b83ac5SAnand Jain 5178a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5179867ab667Sjeff.liu 5180867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 51810b246afaSJeff Mahoney btrfs_warn(fs_info, 51820b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 51830b246afaSJeff Mahoney --len); 5184867ab667Sjeff.liu } 5185867ab667Sjeff.liu 5186867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5187867ab667Sjeff.liu 5188867ab667Sjeff.liu return ret ? -EFAULT : 0; 5189867ab667Sjeff.liu } 5190867ab667Sjeff.liu 5191a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5192a8bfd4abSjeff.liu { 51930b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51940b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51950b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51960b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5197a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5198a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5199a8bfd4abSjeff.liu int ret; 5200a8bfd4abSjeff.liu 5201a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5202a8bfd4abSjeff.liu return -EPERM; 5203a8bfd4abSjeff.liu 5204a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5205a8bfd4abSjeff.liu return -EFAULT; 5206a8bfd4abSjeff.liu 5207a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 52080b246afaSJeff Mahoney btrfs_err(fs_info, 52095d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5210a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5211a8bfd4abSjeff.liu return -EINVAL; 5212a8bfd4abSjeff.liu } 5213a8bfd4abSjeff.liu 5214a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5215a8bfd4abSjeff.liu if (ret) 5216a8bfd4abSjeff.liu return ret; 5217a8bfd4abSjeff.liu 5218a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5219a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5220a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5221a8bfd4abSjeff.liu goto out_unlock; 5222a8bfd4abSjeff.liu } 5223a8bfd4abSjeff.liu 52240b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5225a8bfd4abSjeff.liu strcpy(super_block->label, label); 52260b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 52273a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5228a8bfd4abSjeff.liu 5229a8bfd4abSjeff.liu out_unlock: 5230a8bfd4abSjeff.liu mnt_drop_write_file(file); 5231a8bfd4abSjeff.liu return ret; 5232a8bfd4abSjeff.liu } 5233a8bfd4abSjeff.liu 52342eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 52352eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 52362eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 52372eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 52382eaa055fSJeff Mahoney 5239d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 52402eaa055fSJeff Mahoney { 52414d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 52422eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 52432eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 52442eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 52452eaa055fSJeff Mahoney }; 52462eaa055fSJeff Mahoney 52472eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52482eaa055fSJeff Mahoney return -EFAULT; 52492eaa055fSJeff Mahoney 52502eaa055fSJeff Mahoney return 0; 52512eaa055fSJeff Mahoney } 52522eaa055fSJeff Mahoney 52532eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 52542eaa055fSJeff Mahoney { 52550b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52560b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52570b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 52582eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 52592eaa055fSJeff Mahoney 52602eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 52612eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 52622eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 52632eaa055fSJeff Mahoney 52642eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52652eaa055fSJeff Mahoney return -EFAULT; 52662eaa055fSJeff Mahoney 52672eaa055fSJeff Mahoney return 0; 52682eaa055fSJeff Mahoney } 52692eaa055fSJeff Mahoney 52702ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 52713b02a68aSJeff Mahoney enum btrfs_feature_set set, 52722eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 52732eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 52742eaa055fSJeff Mahoney { 52753b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 52763b02a68aSJeff Mahoney char *names; 52772eaa055fSJeff Mahoney u64 disallowed, unsupported; 52782eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 52792eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 52802eaa055fSJeff Mahoney 52812eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 52822eaa055fSJeff Mahoney if (unsupported) { 52833b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 52843b02a68aSJeff Mahoney if (names) { 52850b246afaSJeff Mahoney btrfs_warn(fs_info, 52863b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 52873b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 52883b02a68aSJeff Mahoney kfree(names); 52893b02a68aSJeff Mahoney } else 52900b246afaSJeff Mahoney btrfs_warn(fs_info, 52912eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 52922eaa055fSJeff Mahoney type, unsupported); 52932eaa055fSJeff Mahoney return -EOPNOTSUPP; 52942eaa055fSJeff Mahoney } 52952eaa055fSJeff Mahoney 52962eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 52972eaa055fSJeff Mahoney if (disallowed) { 52983b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 52993b02a68aSJeff Mahoney if (names) { 53000b246afaSJeff Mahoney btrfs_warn(fs_info, 53013b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53023b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53033b02a68aSJeff Mahoney kfree(names); 53043b02a68aSJeff Mahoney } else 53050b246afaSJeff Mahoney btrfs_warn(fs_info, 53062eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53072eaa055fSJeff Mahoney type, disallowed); 53082eaa055fSJeff Mahoney return -EPERM; 53092eaa055fSJeff Mahoney } 53102eaa055fSJeff Mahoney 53112eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 53122eaa055fSJeff Mahoney if (disallowed) { 53133b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53143b02a68aSJeff Mahoney if (names) { 53150b246afaSJeff Mahoney btrfs_warn(fs_info, 53163b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 53173b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53183b02a68aSJeff Mahoney kfree(names); 53193b02a68aSJeff Mahoney } else 53200b246afaSJeff Mahoney btrfs_warn(fs_info, 53212eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 53222eaa055fSJeff Mahoney type, disallowed); 53232eaa055fSJeff Mahoney return -EPERM; 53242eaa055fSJeff Mahoney } 53252eaa055fSJeff Mahoney 53262eaa055fSJeff Mahoney return 0; 53272eaa055fSJeff Mahoney } 53282eaa055fSJeff Mahoney 53292ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 53302ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 53312eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 53322eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 53332eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 53342eaa055fSJeff Mahoney 53352eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 53362eaa055fSJeff Mahoney { 53370b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53380b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53390b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 53400b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53412eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 53422eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 53432eaa055fSJeff Mahoney u64 newflags; 53442eaa055fSJeff Mahoney int ret; 53452eaa055fSJeff Mahoney 53462eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 53472eaa055fSJeff Mahoney return -EPERM; 53482eaa055fSJeff Mahoney 53492eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 53502eaa055fSJeff Mahoney return -EFAULT; 53512eaa055fSJeff Mahoney 53522eaa055fSJeff Mahoney /* Nothing to do */ 53532eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 53542eaa055fSJeff Mahoney !flags[0].incompat_flags) 53552eaa055fSJeff Mahoney return 0; 53562eaa055fSJeff Mahoney 53572ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 53582eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 53592eaa055fSJeff Mahoney if (ret) 53602eaa055fSJeff Mahoney return ret; 53612eaa055fSJeff Mahoney 53622ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 53632eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 53642eaa055fSJeff Mahoney if (ret) 53652eaa055fSJeff Mahoney return ret; 53662eaa055fSJeff Mahoney 53672ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 53682eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 53692eaa055fSJeff Mahoney if (ret) 53702eaa055fSJeff Mahoney return ret; 53712eaa055fSJeff Mahoney 53727ab19625SDavid Sterba ret = mnt_want_write_file(file); 53737ab19625SDavid Sterba if (ret) 53747ab19625SDavid Sterba return ret; 53757ab19625SDavid Sterba 53768051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 53777ab19625SDavid Sterba if (IS_ERR(trans)) { 53787ab19625SDavid Sterba ret = PTR_ERR(trans); 53797ab19625SDavid Sterba goto out_drop_write; 53807ab19625SDavid Sterba } 53812eaa055fSJeff Mahoney 53820b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 53832eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 53842eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 53852eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 53862eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 53872eaa055fSJeff Mahoney 53882eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 53892eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 53902eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 53912eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 53922eaa055fSJeff Mahoney 53932eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 53942eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 53952eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 53962eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 53970b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53982eaa055fSJeff Mahoney 53993a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 54007ab19625SDavid Sterba out_drop_write: 54017ab19625SDavid Sterba mnt_drop_write_file(file); 54027ab19625SDavid Sterba 54037ab19625SDavid Sterba return ret; 54042eaa055fSJeff Mahoney } 54052eaa055fSJeff Mahoney 54062351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 54072351f431SJosef Bacik { 54082351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 54092351f431SJosef Bacik int ret; 54102351f431SJosef Bacik 54112351f431SJosef Bacik if (compat) { 54122351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 54132351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 54142351f431SJosef Bacik 54152351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 54162351f431SJosef Bacik if (ret) 54172351f431SJosef Bacik return -EFAULT; 54182351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 54192351f431SJosef Bacik if (!arg) 54202351f431SJosef Bacik return -ENOMEM; 54212351f431SJosef Bacik arg->send_fd = args32.send_fd; 54222351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 54232351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 54242351f431SJosef Bacik arg->parent_root = args32.parent_root; 54252351f431SJosef Bacik arg->flags = args32.flags; 54262351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 54272351f431SJosef Bacik sizeof(args32.reserved)); 54282351f431SJosef Bacik #else 54292351f431SJosef Bacik return -ENOTTY; 54302351f431SJosef Bacik #endif 54312351f431SJosef Bacik } else { 54322351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 54332351f431SJosef Bacik if (IS_ERR(arg)) 54342351f431SJosef Bacik return PTR_ERR(arg); 54352351f431SJosef Bacik } 54362351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 54372351f431SJosef Bacik kfree(arg); 54382351f431SJosef Bacik return ret; 54392351f431SJosef Bacik } 54402351f431SJosef Bacik 5441f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5442f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5443f46b5a66SChristoph Hellwig { 54440b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54450b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54460b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54474bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5448f46b5a66SChristoph Hellwig 5449f46b5a66SChristoph Hellwig switch (cmd) { 54506cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 54516cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 54526cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 54536cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 54546cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 54556cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 545640cf931fSEric Sandeen case FS_IOC_GETFSLABEL: 545740cf931fSEric Sandeen return btrfs_ioctl_get_fslabel(file, argp); 545840cf931fSEric Sandeen case FS_IOC_SETFSLABEL: 545940cf931fSEric Sandeen return btrfs_ioctl_set_fslabel(file, argp); 5460f7039b1dSLi Dongyang case FITRIM: 5461f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5462f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5463fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5464fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5465fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 54663de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5467fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 54686f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 54696f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 547076dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 547176dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 54720caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 54730caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 54740caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 54750caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 54766ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 54776ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5478f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 54791e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 54801e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 54811e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5482f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5483198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5484f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 54852ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5486f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5487da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 54886b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 54896b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5490475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 54912ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5492475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 54932ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5494f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 54959ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5496ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5497ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5498cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5499cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5500ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5501ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5502d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5503d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5504d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5505d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5506d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5507d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 55081406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55092ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 55109b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55119b199859SFilipe David Borba Manana int ret; 55129b199859SFilipe David Borba Manana 551382b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 55149b199859SFilipe David Borba Manana if (ret) 55159b199859SFilipe David Borba Manana return ret; 55160b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 55172fad4e83SDavid Sterba /* 55182fad4e83SDavid Sterba * The transaction thread may want to do more work, 551901327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 55202fad4e83SDavid Sterba * processing uncleaned subvols. 55212fad4e83SDavid Sterba */ 55220b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 55239b199859SFilipe David Borba Manana return ret; 55249b199859SFilipe David Borba Manana } 552546204592SSage Weil case BTRFS_IOC_START_SYNC: 55269a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 552746204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55282ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5529475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5530b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5531475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 55322ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5533475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 55342ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5535c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55369ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5537837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 55382ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 553919a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 55402ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 55418ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55428ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5543abccd00fSHugo Mills #ifdef CONFIG_64BIT 5544abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5545abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5546abccd00fSHugo Mills #endif 554731db9f7cSAlexander Block case BTRFS_IOC_SEND: 55482351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 55492351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55502351f431SJosef Bacik case BTRFS_IOC_SEND_32: 55512351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 55522351f431SJosef Bacik #endif 5553c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 55542ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 55555d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5556905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55575d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5558905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55595d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5560905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55615d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5562905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 55632f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 55642f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 55652f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 55662f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 556757254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 556857254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 55693f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 55702ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 55712eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5572d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 55732eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 55742eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 55752eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 55762eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5577e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5578e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5579025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5580025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5581b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5582b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 558342e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 558442e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 558523d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 558623d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5587f46b5a66SChristoph Hellwig } 5588f46b5a66SChristoph Hellwig 5589f46b5a66SChristoph Hellwig return -ENOTTY; 5590f46b5a66SChristoph Hellwig } 55914c63c245SLuke Dashjr 55924c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 55934c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 55944c63c245SLuke Dashjr { 55952a362249SJeff Mahoney /* 55962a362249SJeff Mahoney * These all access 32-bit values anyway so no further 55972a362249SJeff Mahoney * handling is necessary. 55982a362249SJeff Mahoney */ 55994c63c245SLuke Dashjr switch (cmd) { 56004c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56014c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56024c63c245SLuke Dashjr break; 56034c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56044c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56054c63c245SLuke Dashjr break; 56064c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56074c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56084c63c245SLuke Dashjr break; 56094c63c245SLuke Dashjr } 56104c63c245SLuke Dashjr 56114c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56124c63c245SLuke Dashjr } 56134c63c245SLuke Dashjr #endif 5614