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" 48aac0023cSJosef Bacik #include "block-group.h" 49f46b5a66SChristoph Hellwig 50abccd00fSHugo Mills #ifdef CONFIG_64BIT 51abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 52abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 53abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 54abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 55abccd00fSHugo Mills */ 56abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 57abccd00fSHugo Mills __u64 sec; 58abccd00fSHugo Mills __u32 nsec; 59abccd00fSHugo Mills } __attribute__ ((__packed__)); 60abccd00fSHugo Mills 61abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 62abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 63abccd00fSHugo Mills __u64 stransid; /* in */ 64abccd00fSHugo Mills __u64 rtransid; /* out */ 65abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 66abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 67abccd00fSHugo Mills __u64 flags; /* in */ 68abccd00fSHugo Mills __u64 reserved[16]; /* in */ 69abccd00fSHugo Mills } __attribute__ ((__packed__)); 70abccd00fSHugo Mills 71abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 72abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 73abccd00fSHugo Mills #endif 74abccd00fSHugo Mills 752351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 762351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 772351f431SJosef Bacik __s64 send_fd; /* in */ 782351f431SJosef Bacik __u64 clone_sources_count; /* in */ 792351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 802351f431SJosef Bacik __u64 parent_root; /* in */ 812351f431SJosef Bacik __u64 flags; /* in */ 822351f431SJosef Bacik __u64 reserved[4]; /* in */ 832351f431SJosef Bacik } __attribute__ ((__packed__)); 842351f431SJosef Bacik 852351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 862351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 872351f431SJosef Bacik #endif 88abccd00fSHugo Mills 89416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 901c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 911c919a5eSMark Fasheh int no_time_update); 92416161dbSMark Fasheh 936cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 941905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 951905a0f7SDavid Sterba unsigned int flags) 966cbff00fSChristoph Hellwig { 971905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 986cbff00fSChristoph Hellwig return flags; 991905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 1006cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1016cbff00fSChristoph Hellwig else 1026cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1036cbff00fSChristoph Hellwig } 104f46b5a66SChristoph Hellwig 1056cbff00fSChristoph Hellwig /* 106a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 107a157d4fdSDavid Sterba * ioctl. 1086cbff00fSChristoph Hellwig */ 109a157d4fdSDavid Sterba static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags) 1106cbff00fSChristoph Hellwig { 1116cbff00fSChristoph Hellwig unsigned int iflags = 0; 1126cbff00fSChristoph Hellwig 1136cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1146cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1156cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1166cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1176cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1186cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1196cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1206cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1216cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1226cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1236cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1246cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 125d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 126d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 127d0092bddSLi Zefan 12813f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 129d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 13013f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13113f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1326cbff00fSChristoph Hellwig 1336cbff00fSChristoph Hellwig return iflags; 1346cbff00fSChristoph Hellwig } 1356cbff00fSChristoph Hellwig 1366cbff00fSChristoph Hellwig /* 1376cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1386cbff00fSChristoph Hellwig */ 1397b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1406cbff00fSChristoph Hellwig { 1415c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1423cc79392SFilipe Manana unsigned int new_fl = 0; 1436cbff00fSChristoph Hellwig 1445c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1453cc79392SFilipe Manana new_fl |= S_SYNC; 1465c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1473cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1485c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1493cc79392SFilipe Manana new_fl |= S_APPEND; 1505c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1513cc79392SFilipe Manana new_fl |= S_NOATIME; 1525c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1533cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1543cc79392SFilipe Manana 1553cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1563cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1573cc79392SFilipe Manana new_fl); 1586cbff00fSChristoph Hellwig } 1596cbff00fSChristoph Hellwig 1606cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1616cbff00fSChristoph Hellwig { 1625c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 1635c57b8b6SDavid Sterba unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags); 1646cbff00fSChristoph Hellwig 1656cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1666cbff00fSChristoph Hellwig return -EFAULT; 1676cbff00fSChristoph Hellwig return 0; 1686cbff00fSChristoph Hellwig } 1696cbff00fSChristoph Hellwig 1705ba76abfSDavid Sterba /* Check if @flags are a supported and valid set of FS_*_FL flags */ 1715ba76abfSDavid Sterba static int check_fsflags(unsigned int flags) 17275e7cb7fSLiu Bo { 17375e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17475e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17575e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 176e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 177e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17875e7cb7fSLiu Bo return -EOPNOTSUPP; 17975e7cb7fSLiu Bo 18075e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 18175e7cb7fSLiu Bo return -EINVAL; 18275e7cb7fSLiu Bo 18375e7cb7fSLiu Bo return 0; 18475e7cb7fSLiu Bo } 18575e7cb7fSLiu Bo 1866cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1876cbff00fSChristoph Hellwig { 188496ad9aaSAl Viro struct inode *inode = file_inode(file); 1890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1905c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1915c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 1926cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1935aca2842SDarrick J. Wong unsigned int fsflags, old_fsflags; 1946cbff00fSChristoph Hellwig int ret; 195ff9fef55SAnand Jain const char *comp = NULL; 196d2b8fcfeSAnand Jain u32 binode_flags = binode->flags; 1976cbff00fSChristoph Hellwig 198bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 199bd60ea0fSDavid Sterba return -EPERM; 200bd60ea0fSDavid Sterba 201b83cc969SLi Zefan if (btrfs_root_readonly(root)) 202b83cc969SLi Zefan return -EROFS; 203b83cc969SLi Zefan 2045c57b8b6SDavid Sterba if (copy_from_user(&fsflags, arg, sizeof(fsflags))) 2056cbff00fSChristoph Hellwig return -EFAULT; 2066cbff00fSChristoph Hellwig 2075c57b8b6SDavid Sterba ret = check_fsflags(fsflags); 20875e7cb7fSLiu Bo if (ret) 20975e7cb7fSLiu Bo return ret; 2106cbff00fSChristoph Hellwig 211e7848683SJan Kara ret = mnt_want_write_file(file); 212e7848683SJan Kara if (ret) 213e7848683SJan Kara return ret; 214e7848683SJan Kara 2155955102cSAl Viro inode_lock(inode); 2166cbff00fSChristoph Hellwig 2175c57b8b6SDavid Sterba fsflags = btrfs_mask_fsflags_for_type(inode, fsflags); 2185aca2842SDarrick J. Wong old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags); 2195aca2842SDarrick J. Wong ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags); 2205aca2842SDarrick J. Wong if (ret) 2216cbff00fSChristoph Hellwig goto out_unlock; 2226cbff00fSChristoph Hellwig 2235c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 224d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_SYNC; 2256cbff00fSChristoph Hellwig else 226d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_SYNC; 2275c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 228d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_IMMUTABLE; 2296cbff00fSChristoph Hellwig else 230d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_IMMUTABLE; 2315c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 232d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_APPEND; 2336cbff00fSChristoph Hellwig else 234d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_APPEND; 2355c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 236d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODUMP; 2376cbff00fSChristoph Hellwig else 238d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODUMP; 2395c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 240d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOATIME; 2416cbff00fSChristoph Hellwig else 242d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOATIME; 2435c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 244d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_DIRSYNC; 2456cbff00fSChristoph Hellwig else 246d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_DIRSYNC; 2475c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 24844e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2497e97b8daSDavid Sterba /* 2507e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2517e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2527e97b8daSDavid Sterba * status of the file and will not set it. 2537e97b8daSDavid Sterba */ 2547e97b8daSDavid Sterba if (inode->i_size == 0) 255d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW | 256d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM; 2577e97b8daSDavid Sterba } else { 258d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW; 2597e97b8daSDavid Sterba } 2607e97b8daSDavid Sterba } else { 2617e97b8daSDavid Sterba /* 26201327610SNicholas D Steeves * Revert back under same assumptions as above 2637e97b8daSDavid Sterba */ 26444e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2657e97b8daSDavid Sterba if (inode->i_size == 0) 266d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_NODATACOW | 267d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM); 2687e97b8daSDavid Sterba } else { 269d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODATACOW; 2707e97b8daSDavid Sterba } 2717e97b8daSDavid Sterba } 2726cbff00fSChristoph Hellwig 27375e7cb7fSLiu Bo /* 27475e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 27575e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 27675e7cb7fSLiu Bo * things smaller. 27775e7cb7fSLiu Bo */ 2785c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 279d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_COMPRESS; 280d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOCOMPRESS; 2815c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 28263541927SFilipe David Borba Manana 283eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 284eede2bf3SOmar Sandoval ret = -ETXTBSY; 285eede2bf3SOmar Sandoval goto out_unlock; 286eede2bf3SOmar Sandoval } 287eede2bf3SOmar Sandoval 288d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_COMPRESS; 289d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOCOMPRESS; 29063541927SFilipe David Borba Manana 29193370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 29293370509SDavid Sterba if (!comp || comp[0] == 0) 29393370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 294ebcb904dSLi Zefan } else { 295d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 29675e7cb7fSLiu Bo } 2976cbff00fSChristoph Hellwig 298ff9fef55SAnand Jain /* 299ff9fef55SAnand Jain * 1 for inode item 300ff9fef55SAnand Jain * 2 for properties 301ff9fef55SAnand Jain */ 302ff9fef55SAnand Jain trans = btrfs_start_transaction(root, 3); 303f062abf0SLi Zefan if (IS_ERR(trans)) { 304f062abf0SLi Zefan ret = PTR_ERR(trans); 305d2b8fcfeSAnand Jain goto out_unlock; 306f062abf0SLi Zefan } 3076cbff00fSChristoph Hellwig 308ff9fef55SAnand Jain if (comp) { 309ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, 310ff9fef55SAnand Jain strlen(comp), 0); 311ff9fef55SAnand Jain if (ret) { 312ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 313ff9fef55SAnand Jain goto out_end_trans; 314ff9fef55SAnand Jain } 315ff9fef55SAnand Jain } else { 316ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 317ff9fef55SAnand Jain 0, 0); 318ff9fef55SAnand Jain if (ret && ret != -ENODATA) { 319ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 320ff9fef55SAnand Jain goto out_end_trans; 321ff9fef55SAnand Jain } 322ff9fef55SAnand Jain } 323ff9fef55SAnand Jain 324d2b8fcfeSAnand Jain binode->flags = binode_flags; 3257b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3260c4d2d95SJosef Bacik inode_inc_iversion(inode); 327c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3286cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3296cbff00fSChristoph Hellwig 330ff9fef55SAnand Jain out_end_trans: 3313a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3326cbff00fSChristoph Hellwig out_unlock: 3335955102cSAl Viro inode_unlock(inode); 334e7848683SJan Kara mnt_drop_write_file(file); 3352d4e6f6aSliubo return ret; 3366cbff00fSChristoph Hellwig } 3376cbff00fSChristoph Hellwig 33819f93b3cSDavid Sterba /* 33919f93b3cSDavid Sterba * Translate btrfs internal inode flags to xflags as expected by the 34019f93b3cSDavid Sterba * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are 34119f93b3cSDavid Sterba * silently dropped. 34219f93b3cSDavid Sterba */ 34319f93b3cSDavid Sterba static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags) 34419f93b3cSDavid Sterba { 34519f93b3cSDavid Sterba unsigned int xflags = 0; 34619f93b3cSDavid Sterba 34719f93b3cSDavid Sterba if (flags & BTRFS_INODE_APPEND) 34819f93b3cSDavid Sterba xflags |= FS_XFLAG_APPEND; 34919f93b3cSDavid Sterba if (flags & BTRFS_INODE_IMMUTABLE) 35019f93b3cSDavid Sterba xflags |= FS_XFLAG_IMMUTABLE; 35119f93b3cSDavid Sterba if (flags & BTRFS_INODE_NOATIME) 35219f93b3cSDavid Sterba xflags |= FS_XFLAG_NOATIME; 35319f93b3cSDavid Sterba if (flags & BTRFS_INODE_NODUMP) 35419f93b3cSDavid Sterba xflags |= FS_XFLAG_NODUMP; 35519f93b3cSDavid Sterba if (flags & BTRFS_INODE_SYNC) 35619f93b3cSDavid Sterba xflags |= FS_XFLAG_SYNC; 35719f93b3cSDavid Sterba 35819f93b3cSDavid Sterba return xflags; 35919f93b3cSDavid Sterba } 36019f93b3cSDavid Sterba 36119f93b3cSDavid Sterba /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */ 36219f93b3cSDavid Sterba static int check_xflags(unsigned int flags) 36319f93b3cSDavid Sterba { 36419f93b3cSDavid Sterba if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME | 36519f93b3cSDavid Sterba FS_XFLAG_NODUMP | FS_XFLAG_SYNC)) 36619f93b3cSDavid Sterba return -EOPNOTSUPP; 36719f93b3cSDavid Sterba return 0; 36819f93b3cSDavid Sterba } 36919f93b3cSDavid Sterba 370e4202ac9SDavid Sterba /* 371e4202ac9SDavid Sterba * Set the xflags from the internal inode flags. The remaining items of fsxattr 372e4202ac9SDavid Sterba * are zeroed. 373e4202ac9SDavid Sterba */ 374e4202ac9SDavid Sterba static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg) 375e4202ac9SDavid Sterba { 376e4202ac9SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 377e4202ac9SDavid Sterba struct fsxattr fa; 378e4202ac9SDavid Sterba 3797b0e492eSDarrick J. Wong simple_fill_fsxattr(&fa, btrfs_inode_flags_to_xflags(binode->flags)); 380e4202ac9SDavid Sterba if (copy_to_user(arg, &fa, sizeof(fa))) 381e4202ac9SDavid Sterba return -EFAULT; 382e4202ac9SDavid Sterba 383e4202ac9SDavid Sterba return 0; 384e4202ac9SDavid Sterba } 385e4202ac9SDavid Sterba 386025f2121SDavid Sterba static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg) 387025f2121SDavid Sterba { 388025f2121SDavid Sterba struct inode *inode = file_inode(file); 389025f2121SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 390025f2121SDavid Sterba struct btrfs_root *root = binode->root; 391025f2121SDavid Sterba struct btrfs_trans_handle *trans; 3927b0e492eSDarrick J. Wong struct fsxattr fa, old_fa; 393025f2121SDavid Sterba unsigned old_flags; 394025f2121SDavid Sterba unsigned old_i_flags; 395025f2121SDavid Sterba int ret = 0; 396025f2121SDavid Sterba 397025f2121SDavid Sterba if (!inode_owner_or_capable(inode)) 398025f2121SDavid Sterba return -EPERM; 399025f2121SDavid Sterba 400025f2121SDavid Sterba if (btrfs_root_readonly(root)) 401025f2121SDavid Sterba return -EROFS; 402025f2121SDavid Sterba 403025f2121SDavid Sterba if (copy_from_user(&fa, arg, sizeof(fa))) 404025f2121SDavid Sterba return -EFAULT; 405025f2121SDavid Sterba 406025f2121SDavid Sterba ret = check_xflags(fa.fsx_xflags); 407025f2121SDavid Sterba if (ret) 408025f2121SDavid Sterba return ret; 409025f2121SDavid Sterba 410025f2121SDavid Sterba if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0) 411025f2121SDavid Sterba return -EOPNOTSUPP; 412025f2121SDavid Sterba 413025f2121SDavid Sterba ret = mnt_want_write_file(file); 414025f2121SDavid Sterba if (ret) 415025f2121SDavid Sterba return ret; 416025f2121SDavid Sterba 417025f2121SDavid Sterba inode_lock(inode); 418025f2121SDavid Sterba 419025f2121SDavid Sterba old_flags = binode->flags; 420025f2121SDavid Sterba old_i_flags = inode->i_flags; 421025f2121SDavid Sterba 4227b0e492eSDarrick J. Wong simple_fill_fsxattr(&old_fa, 4237b0e492eSDarrick J. Wong btrfs_inode_flags_to_xflags(binode->flags)); 4247b0e492eSDarrick J. Wong ret = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa); 4257b0e492eSDarrick J. Wong if (ret) 426025f2121SDavid Sterba goto out_unlock; 427025f2121SDavid Sterba 428025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_SYNC) 429025f2121SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 430025f2121SDavid Sterba else 431025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 432025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE) 433025f2121SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 434025f2121SDavid Sterba else 435025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 436025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_APPEND) 437025f2121SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 438025f2121SDavid Sterba else 439025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 440025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NODUMP) 441025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 442025f2121SDavid Sterba else 443025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 444025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NOATIME) 445025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 446025f2121SDavid Sterba else 447025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 448025f2121SDavid Sterba 449025f2121SDavid Sterba /* 1 item for the inode */ 450025f2121SDavid Sterba trans = btrfs_start_transaction(root, 1); 451025f2121SDavid Sterba if (IS_ERR(trans)) { 452025f2121SDavid Sterba ret = PTR_ERR(trans); 453025f2121SDavid Sterba goto out_unlock; 454025f2121SDavid Sterba } 455025f2121SDavid Sterba 456025f2121SDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 457025f2121SDavid Sterba inode_inc_iversion(inode); 458025f2121SDavid Sterba inode->i_ctime = current_time(inode); 459025f2121SDavid Sterba ret = btrfs_update_inode(trans, root, inode); 460025f2121SDavid Sterba 461025f2121SDavid Sterba btrfs_end_transaction(trans); 462025f2121SDavid Sterba 463025f2121SDavid Sterba out_unlock: 464025f2121SDavid Sterba if (ret) { 465025f2121SDavid Sterba binode->flags = old_flags; 466025f2121SDavid Sterba inode->i_flags = old_i_flags; 467025f2121SDavid Sterba } 468025f2121SDavid Sterba 469025f2121SDavid Sterba inode_unlock(inode); 470025f2121SDavid Sterba mnt_drop_write_file(file); 471025f2121SDavid Sterba 472025f2121SDavid Sterba return ret; 473025f2121SDavid Sterba } 474025f2121SDavid Sterba 4756cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4766cbff00fSChristoph Hellwig { 477496ad9aaSAl Viro struct inode *inode = file_inode(file); 4786cbff00fSChristoph Hellwig 4796cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4806cbff00fSChristoph Hellwig } 481f46b5a66SChristoph Hellwig 482b929c1d8SMarcos Paulo de Souza static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, 483b929c1d8SMarcos Paulo de Souza void __user *arg) 484f7039b1dSLi Dongyang { 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 543e1f60a65SDavid Sterba int __pure 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); 707c7e54b51SJosef Bacik if (ret) { 708c7e54b51SJosef Bacik btrfs_abort_transaction(trans, ret); 709c7e54b51SJosef Bacik goto fail; 710c7e54b51SJosef Bacik } 71152c26179SYan Zheng 7126025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7134a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 714c7e54b51SJosef Bacik if (ret) { 715c7e54b51SJosef Bacik btrfs_abort_transaction(trans, ret); 716c7e54b51SJosef Bacik goto fail; 717c7e54b51SJosef Bacik } 7180660b5afSChris Mason 719cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7206bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 721dd5f9615SStefan Behrens if (ret) 72266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 723dd5f9615SStefan Behrens 724f46b5a66SChristoph Hellwig fail: 72549a3c4d9SDavid Sterba kfree(root_item); 726d5c12070SMiao Xie trans->block_rsv = NULL; 727d5c12070SMiao Xie trans->bytes_reserved = 0; 7287775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 729de6e8200SLiu Bo 73072fd032eSSage Weil if (async_transid) { 73172fd032eSSage Weil *async_transid = trans->transid; 7323a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 73300d71c9cSMiao Xie if (err) 7343a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73572fd032eSSage Weil } else { 7363a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73772fd032eSSage Weil } 738f46b5a66SChristoph Hellwig if (err && !ret) 739f46b5a66SChristoph Hellwig ret = err; 7401a65e24bSChris Mason 7415662344bSTsutomu Itoh if (!ret) { 7425662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 743de6e8200SLiu Bo if (IS_ERR(inode)) 744de6e8200SLiu Bo return PTR_ERR(inode); 7455662344bSTsutomu Itoh d_instantiate(dentry, inode); 7465662344bSTsutomu Itoh } 747f46b5a66SChristoph Hellwig return ret; 74849a3c4d9SDavid Sterba 74949a3c4d9SDavid Sterba fail_free: 75049a3c4d9SDavid Sterba kfree(root_item); 75149a3c4d9SDavid Sterba return ret; 752f46b5a66SChristoph Hellwig } 753f46b5a66SChristoph Hellwig 754e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 75561d7e4cbSDavid Sterba struct dentry *dentry, 756e9662f70SMiao Xie u64 *async_transid, bool readonly, 757e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 758f46b5a66SChristoph Hellwig { 7590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7602e4bfab9SYan, Zheng struct inode *inode; 761f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 762f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7632e4bfab9SYan, Zheng int ret; 7648ecebf4dSRobbie Ko bool snapshot_force_cow = false; 765f46b5a66SChristoph Hellwig 76627cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 767f46b5a66SChristoph Hellwig return -EINVAL; 768f46b5a66SChristoph Hellwig 769eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 770eede2bf3SOmar Sandoval btrfs_warn(fs_info, 771eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 772eede2bf3SOmar Sandoval return -ETXTBSY; 773eede2bf3SOmar Sandoval } 774eede2bf3SOmar Sandoval 77523269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 776a1ee7362SDavid Sterba if (!pending_snapshot) 777a1ee7362SDavid Sterba return -ENOMEM; 778a1ee7362SDavid Sterba 779b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 78023269bf5SDavid Sterba GFP_KERNEL); 7818546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7828546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 783b0c0ea63SDavid Sterba ret = -ENOMEM; 784b0c0ea63SDavid Sterba goto free_pending; 785b0c0ea63SDavid Sterba } 786b0c0ea63SDavid Sterba 7878ecebf4dSRobbie Ko /* 7888ecebf4dSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 7898ecebf4dSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 7908ecebf4dSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 7918ecebf4dSRobbie Ko */ 792ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7934e857c58SPeter Zijlstra smp_mb__after_atomic(); 79445bac0f3SLiu Bo /* wait for no snapshot writes */ 79545bac0f3SLiu Bo wait_event(root->subv_writers->wait, 79645bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 7978257b2dcSMiao Xie 7983cd24c69SEthan Lien ret = btrfs_start_delalloc_snapshot(root); 7996a03843dSMiao Xie if (ret) 800a1ee7362SDavid Sterba goto dec_and_free; 8016a03843dSMiao Xie 8028ecebf4dSRobbie Ko /* 8038ecebf4dSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 8048ecebf4dSRobbie Ko * we force future writes to fallback to COW mode during snapshot 8058ecebf4dSRobbie Ko * creation. 8068ecebf4dSRobbie Ko */ 8078ecebf4dSRobbie Ko atomic_inc(&root->snapshot_force_cow); 8088ecebf4dSRobbie Ko snapshot_force_cow = true; 8098ecebf4dSRobbie Ko 8106374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 8116a03843dSMiao Xie 81266d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 81366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 814d5c12070SMiao Xie /* 815d5c12070SMiao Xie * 1 - parent dir inode 816d5c12070SMiao Xie * 2 - dir entries 817d5c12070SMiao Xie * 1 - root item 818d5c12070SMiao Xie * 2 - root ref/backref 819d5c12070SMiao Xie * 1 - root of snapshot 820dd5f9615SStefan Behrens * 1 - UUID item 821d5c12070SMiao Xie */ 822d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 823dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 824ee3441b4SJeff Mahoney false); 825d5c12070SMiao Xie if (ret) 826a1ee7362SDavid Sterba goto dec_and_free; 827d5c12070SMiao Xie 828a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 829a22285a6SYan, Zheng pending_snapshot->root = root; 830b83cc969SLi Zefan pending_snapshot->readonly = readonly; 831e9662f70SMiao Xie pending_snapshot->dir = dir; 8328696c533SMiao Xie pending_snapshot->inherit = inherit; 833a22285a6SYan, Zheng 834d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 835a22285a6SYan, Zheng if (IS_ERR(trans)) { 836a22285a6SYan, Zheng ret = PTR_ERR(trans); 837a22285a6SYan, Zheng goto fail; 838a22285a6SYan, Zheng } 839a22285a6SYan, Zheng 8400b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 841a22285a6SYan, Zheng list_add(&pending_snapshot->list, 842a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8430b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 84472fd032eSSage Weil if (async_transid) { 84572fd032eSSage Weil *async_transid = trans->transid; 8463a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 84700d71c9cSMiao Xie if (ret) 8483a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 84972fd032eSSage Weil } else { 8503a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 85172fd032eSSage Weil } 852aec8030aSMiao Xie if (ret) 853c37b2b62SJosef Bacik goto fail; 854a22285a6SYan, Zheng 855a22285a6SYan, Zheng ret = pending_snapshot->error; 856f46b5a66SChristoph Hellwig if (ret) 8572e4bfab9SYan, Zheng goto fail; 858f46b5a66SChristoph Hellwig 859d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 860d3797308SChris Mason if (ret) 861d3797308SChris Mason goto fail; 862d3797308SChris Mason 8632b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8642e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8652e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8662e4bfab9SYan, Zheng goto fail; 8672e4bfab9SYan, Zheng } 8685662344bSTsutomu Itoh 8692e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8702e4bfab9SYan, Zheng ret = 0; 8712e4bfab9SYan, Zheng fail: 8727775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 873a1ee7362SDavid Sterba dec_and_free: 8748ecebf4dSRobbie Ko if (snapshot_force_cow) 8758ecebf4dSRobbie Ko atomic_dec(&root->snapshot_force_cow); 876ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8774625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 878b0c0ea63SDavid Sterba free_pending: 879b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8808546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 881a1ee7362SDavid Sterba kfree(pending_snapshot); 882a1ee7362SDavid Sterba 883f46b5a66SChristoph Hellwig return ret; 884f46b5a66SChristoph Hellwig } 885f46b5a66SChristoph Hellwig 8864260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8874260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8884260f7c7SSage Weil * whether the type of victim is right. 8894260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8904260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8914260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8924260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8934260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8944260f7c7SSage Weil * a. be owner of dir, or 8954260f7c7SSage Weil * b. be owner of victim, or 8964260f7c7SSage Weil * c. have CAP_FOWNER capability 89701327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8984260f7c7SSage Weil * links pointing to it. 8994260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 9004260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 9014260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 9024260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 9034260f7c7SSage Weil * nfs_async_unlink(). 9044260f7c7SSage Weil */ 9054260f7c7SSage Weil 9064260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 9074260f7c7SSage Weil { 9084260f7c7SSage Weil int error; 9094260f7c7SSage Weil 9102b0143b5SDavid Howells if (d_really_is_negative(victim)) 9114260f7c7SSage Weil return -ENOENT; 9124260f7c7SSage Weil 9132b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 9144fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 9154260f7c7SSage Weil 9164260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 9174260f7c7SSage Weil if (error) 9184260f7c7SSage Weil return error; 9194260f7c7SSage Weil if (IS_APPEND(dir)) 9204260f7c7SSage Weil return -EPERM; 9212b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9222b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9234260f7c7SSage Weil return -EPERM; 9244260f7c7SSage Weil if (isdir) { 925e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9264260f7c7SSage Weil return -ENOTDIR; 9274260f7c7SSage Weil if (IS_ROOT(victim)) 9284260f7c7SSage Weil return -EBUSY; 929e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9304260f7c7SSage Weil return -EISDIR; 9314260f7c7SSage Weil if (IS_DEADDIR(dir)) 9324260f7c7SSage Weil return -ENOENT; 9334260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9344260f7c7SSage Weil return -EBUSY; 9354260f7c7SSage Weil return 0; 9364260f7c7SSage Weil } 9374260f7c7SSage Weil 938cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 939cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 940cb8e7090SChristoph Hellwig { 9412b0143b5SDavid Howells if (d_really_is_positive(child)) 942cb8e7090SChristoph Hellwig return -EEXIST; 943cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 944cb8e7090SChristoph Hellwig return -ENOENT; 945cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 946cb8e7090SChristoph Hellwig } 947cb8e7090SChristoph Hellwig 948cb8e7090SChristoph Hellwig /* 949cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 950cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 951cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 952cb8e7090SChristoph Hellwig */ 95392872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 95452f75f4fSDavid Sterba const char *name, int namelen, 95572fd032eSSage Weil struct btrfs_root *snap_src, 9566f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9578696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 958cb8e7090SChristoph Hellwig { 9592b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9600b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 961cb8e7090SChristoph Hellwig struct dentry *dentry; 962cb8e7090SChristoph Hellwig int error; 963cb8e7090SChristoph Hellwig 96400235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 96500235411SAl Viro if (error == -EINTR) 96600235411SAl Viro return error; 967cb8e7090SChristoph Hellwig 968cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 969cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 970cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 971cb8e7090SChristoph Hellwig goto out_unlock; 972cb8e7090SChristoph Hellwig 97376dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 974cb8e7090SChristoph Hellwig if (error) 975a874a63eSLiu Bo goto out_dput; 976cb8e7090SChristoph Hellwig 9779c52057cSChris Mason /* 9789c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9799c52057cSChris Mason * check for them now when we can safely fail 9809c52057cSChris Mason */ 9819c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9829c52057cSChris Mason dir->i_ino, name, 9839c52057cSChris Mason namelen); 9849c52057cSChris Mason if (error) 9859c52057cSChris Mason goto out_dput; 9869c52057cSChris Mason 9870b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 98876dda93cSYan, Zheng 98976dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 99076dda93cSYan, Zheng goto out_up_read; 99176dda93cSYan, Zheng 9923de4586cSChris Mason if (snap_src) { 99361d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9946f72c7e2SArne Jansen async_transid, readonly, inherit); 9953de4586cSChris Mason } else { 996d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 997d5c12070SMiao Xie async_transid, inherit); 9983de4586cSChris Mason } 99976dda93cSYan, Zheng if (!error) 100076dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 100176dda93cSYan, Zheng out_up_read: 10020b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 1003cb8e7090SChristoph Hellwig out_dput: 1004cb8e7090SChristoph Hellwig dput(dentry); 1005cb8e7090SChristoph Hellwig out_unlock: 10065955102cSAl Viro inode_unlock(dir); 1007cb8e7090SChristoph Hellwig return error; 1008cb8e7090SChristoph Hellwig } 1009cb8e7090SChristoph Hellwig 10104cb5300bSChris Mason /* 10114cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 10124cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 10134cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 10144cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 10154cb5300bSChris Mason * part of the file 10164cb5300bSChris Mason */ 1017aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 10184cb5300bSChris Mason { 10194cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10204cb5300bSChris Mason struct extent_map *em = NULL; 10214cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10224cb5300bSChris Mason u64 end; 10234cb5300bSChris Mason 10244cb5300bSChris Mason read_lock(&em_tree->lock); 102509cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10264cb5300bSChris Mason read_unlock(&em_tree->lock); 10274cb5300bSChris Mason 10284cb5300bSChris Mason if (em) { 10294cb5300bSChris Mason end = extent_map_end(em); 10304cb5300bSChris Mason free_extent_map(em); 10314cb5300bSChris Mason if (end - offset > thresh) 10324cb5300bSChris Mason return 0; 10334cb5300bSChris Mason } 10344cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10354cb5300bSChris Mason thresh /= 2; 10364cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10374cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10384cb5300bSChris Mason if (end >= thresh) 10394cb5300bSChris Mason return 0; 10404cb5300bSChris Mason return 1; 10414cb5300bSChris Mason } 10424cb5300bSChris Mason 10434cb5300bSChris Mason /* 10444cb5300bSChris Mason * helper function to walk through a file and find extents 10454cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10464cb5300bSChris Mason * 10474cb5300bSChris Mason * This is used by the defragging code to find new and small 10484cb5300bSChris Mason * extents 10494cb5300bSChris Mason */ 10504cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10514cb5300bSChris Mason struct inode *inode, u64 newer_than, 1052aab110abSDavid Sterba u64 *off, u32 thresh) 10534cb5300bSChris Mason { 10544cb5300bSChris Mason struct btrfs_path *path; 10554cb5300bSChris Mason struct btrfs_key min_key; 10564cb5300bSChris Mason struct extent_buffer *leaf; 10574cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10584cb5300bSChris Mason int type; 10594cb5300bSChris Mason int ret; 10604a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10614cb5300bSChris Mason 10624cb5300bSChris Mason path = btrfs_alloc_path(); 10634cb5300bSChris Mason if (!path) 10644cb5300bSChris Mason return -ENOMEM; 10654cb5300bSChris Mason 1066a4689d2bSDavid Sterba min_key.objectid = ino; 10674cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10684cb5300bSChris Mason min_key.offset = *off; 10694cb5300bSChris Mason 10704cb5300bSChris Mason while (1) { 10716174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10724cb5300bSChris Mason if (ret != 0) 10734cb5300bSChris Mason goto none; 1074f094c9bdSFilipe Manana process_slot: 1075a4689d2bSDavid Sterba if (min_key.objectid != ino) 10764cb5300bSChris Mason goto none; 10774cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10784cb5300bSChris Mason goto none; 10794cb5300bSChris Mason 10804cb5300bSChris Mason leaf = path->nodes[0]; 10814cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10824cb5300bSChris Mason struct btrfs_file_extent_item); 10834cb5300bSChris Mason 10844cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10854cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10864cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10874cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10884cb5300bSChris Mason *off = min_key.offset; 10894cb5300bSChris Mason btrfs_free_path(path); 10904cb5300bSChris Mason return 0; 10914cb5300bSChris Mason } 10924cb5300bSChris Mason 1093f094c9bdSFilipe Manana path->slots[0]++; 1094f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1095f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1096f094c9bdSFilipe Manana goto process_slot; 1097f094c9bdSFilipe Manana } 1098f094c9bdSFilipe Manana 10994cb5300bSChris Mason if (min_key.offset == (u64)-1) 11004cb5300bSChris Mason goto none; 11014cb5300bSChris Mason 11024cb5300bSChris Mason min_key.offset++; 11034cb5300bSChris Mason btrfs_release_path(path); 11044cb5300bSChris Mason } 11054cb5300bSChris Mason none: 11064cb5300bSChris Mason btrfs_free_path(path); 11074cb5300bSChris Mason return -ENOENT; 11084cb5300bSChris Mason } 11094cb5300bSChris Mason 11106c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 111117ce6ef8SLiu Bo { 111217ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1113940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 11146c282eb4SLi Zefan struct extent_map *em; 111509cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1116940100a4SChris Mason 1117940100a4SChris Mason /* 1118940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1119940100a4SChris Mason * the full extent lock 1120940100a4SChris Mason */ 1121940100a4SChris Mason read_lock(&em_tree->lock); 1122940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1123940100a4SChris Mason read_unlock(&em_tree->lock); 1124940100a4SChris Mason 1125940100a4SChris Mason if (!em) { 1126308d9800SFilipe Manana struct extent_state *cached = NULL; 1127308d9800SFilipe Manana u64 end = start + len - 1; 1128308d9800SFilipe Manana 1129940100a4SChris Mason /* get the big lock and read metadata off disk */ 1130ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1131fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1132e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1133940100a4SChris Mason 11346cf8bfbfSDan Carpenter if (IS_ERR(em)) 11356c282eb4SLi Zefan return NULL; 1136940100a4SChris Mason } 1137940100a4SChris Mason 11386c282eb4SLi Zefan return em; 11396c282eb4SLi Zefan } 11406c282eb4SLi Zefan 11416c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11426c282eb4SLi Zefan { 11436c282eb4SLi Zefan struct extent_map *next; 11446c282eb4SLi Zefan bool ret = true; 11456c282eb4SLi Zefan 11466c282eb4SLi Zefan /* this is the last extent */ 11476c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11486c282eb4SLi Zefan return false; 11496c282eb4SLi Zefan 11506c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1151e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1152e9512d72SChris Mason ret = false; 1153e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1154ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11556c282eb4SLi Zefan ret = false; 11566c282eb4SLi Zefan 11576c282eb4SLi Zefan free_extent_map(next); 11586c282eb4SLi Zefan return ret; 11596c282eb4SLi Zefan } 11606c282eb4SLi Zefan 1161aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1162a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1163a43a2111SAndrew Mahone int compress) 11646c282eb4SLi Zefan { 11656c282eb4SLi Zefan struct extent_map *em; 11666c282eb4SLi Zefan int ret = 1; 11676c282eb4SLi Zefan bool next_mergeable = true; 11684a3560c4SLiu Bo bool prev_mergeable = true; 11696c282eb4SLi Zefan 11706c282eb4SLi Zefan /* 11716c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11726c282eb4SLi Zefan * defragging it 11736c282eb4SLi Zefan */ 11746c282eb4SLi Zefan if (start < *defrag_end) 11756c282eb4SLi Zefan return 1; 11766c282eb4SLi Zefan 11776c282eb4SLi Zefan *skip = 0; 11786c282eb4SLi Zefan 11796c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11806c282eb4SLi Zefan if (!em) 11816c282eb4SLi Zefan return 0; 11826c282eb4SLi Zefan 1183940100a4SChris Mason /* this will cover holes, and inline extents */ 118417ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1185940100a4SChris Mason ret = 0; 118617ce6ef8SLiu Bo goto out; 118717ce6ef8SLiu Bo } 118817ce6ef8SLiu Bo 11894a3560c4SLiu Bo if (!*defrag_end) 11904a3560c4SLiu Bo prev_mergeable = false; 11914a3560c4SLiu Bo 11926c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1193940100a4SChris Mason /* 11946c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11956c282eb4SLi Zefan * real extent, don't bother defragging it 1196940100a4SChris Mason */ 1197a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11984a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1199940100a4SChris Mason ret = 0; 120017ce6ef8SLiu Bo out: 1201940100a4SChris Mason /* 1202940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1203940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1204940100a4SChris Mason * so that the next tiny extent will force a defrag. 1205940100a4SChris Mason * 1206940100a4SChris Mason * The end result of this is that tiny extents before a single big 1207940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1208940100a4SChris Mason */ 1209940100a4SChris Mason if (ret) { 1210940100a4SChris Mason *defrag_end = extent_map_end(em); 1211940100a4SChris Mason } else { 1212940100a4SChris Mason *last_len = 0; 1213940100a4SChris Mason *skip = extent_map_end(em); 1214940100a4SChris Mason *defrag_end = 0; 1215940100a4SChris Mason } 1216940100a4SChris Mason 1217940100a4SChris Mason free_extent_map(em); 1218940100a4SChris Mason return ret; 1219940100a4SChris Mason } 1220940100a4SChris Mason 12214cb5300bSChris Mason /* 12224cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12234cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12244cb5300bSChris Mason * to COW and defrag. 12254cb5300bSChris Mason * 12264cb5300bSChris Mason * It also makes sure the delalloc code has enough 12274cb5300bSChris Mason * dirty data to avoid making new small extents as part 12284cb5300bSChris Mason * of the defrag 12294cb5300bSChris Mason * 12304cb5300bSChris Mason * It's a good idea to start RA on this range 12314cb5300bSChris Mason * before calling this. 12324cb5300bSChris Mason */ 12334cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12344cb5300bSChris Mason struct page **pages, 12354cb5300bSChris Mason unsigned long start_index, 1236c41570c9SJustin Maggard unsigned long num_pages) 1237f46b5a66SChristoph Hellwig { 12384cb5300bSChris Mason unsigned long file_end; 12394cb5300bSChris Mason u64 isize = i_size_read(inode); 1240f46b5a66SChristoph Hellwig u64 page_start; 1241f46b5a66SChristoph Hellwig u64 page_end; 12421f12bd06SLiu Bo u64 page_cnt; 12434cb5300bSChris Mason int ret; 12444cb5300bSChris Mason int i; 12454cb5300bSChris Mason int i_done; 12464cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12474cb5300bSChris Mason struct extent_state *cached_state = NULL; 1248600a45e1SMiao Xie struct extent_io_tree *tree; 1249364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12503b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12514cb5300bSChris Mason 125209cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12531f12bd06SLiu Bo if (!isize || start_index > file_end) 12541f12bd06SLiu Bo return 0; 12551f12bd06SLiu Bo 12561f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12574cb5300bSChris Mason 1258364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 125909cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 126009cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12614cb5300bSChris Mason if (ret) 12624cb5300bSChris Mason return ret; 12634cb5300bSChris Mason i_done = 0; 1264600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12654cb5300bSChris Mason 12664cb5300bSChris Mason /* step one, lock all the pages */ 12671f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12684cb5300bSChris Mason struct page *page; 1269600a45e1SMiao Xie again: 1270a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12713b16a4e3SJosef Bacik start_index + i, mask); 12724cb5300bSChris Mason if (!page) 12734cb5300bSChris Mason break; 12744cb5300bSChris Mason 1275600a45e1SMiao Xie page_start = page_offset(page); 127609cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1277600a45e1SMiao Xie while (1) { 1278308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1279ff13db41SDavid Sterba &cached_state); 1280600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1281600a45e1SMiao Xie page_start); 1282308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1283e43bbe5eSDavid Sterba &cached_state); 1284600a45e1SMiao Xie if (!ordered) 1285600a45e1SMiao Xie break; 1286600a45e1SMiao Xie 1287600a45e1SMiao Xie unlock_page(page); 1288600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1289600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1290600a45e1SMiao Xie lock_page(page); 12911f12bd06SLiu Bo /* 12921f12bd06SLiu Bo * we unlocked the page above, so we need check if 12931f12bd06SLiu Bo * it was released or not. 12941f12bd06SLiu Bo */ 12951f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12961f12bd06SLiu Bo unlock_page(page); 129709cbfeafSKirill A. Shutemov put_page(page); 12981f12bd06SLiu Bo goto again; 12991f12bd06SLiu Bo } 1300600a45e1SMiao Xie } 1301600a45e1SMiao Xie 13024cb5300bSChris Mason if (!PageUptodate(page)) { 13034cb5300bSChris Mason btrfs_readpage(NULL, page); 13044cb5300bSChris Mason lock_page(page); 13054cb5300bSChris Mason if (!PageUptodate(page)) { 13064cb5300bSChris Mason unlock_page(page); 130709cbfeafSKirill A. Shutemov put_page(page); 13084cb5300bSChris Mason ret = -EIO; 13094cb5300bSChris Mason break; 13104cb5300bSChris Mason } 13114cb5300bSChris Mason } 1312600a45e1SMiao Xie 1313600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1314600a45e1SMiao Xie unlock_page(page); 131509cbfeafSKirill A. Shutemov put_page(page); 1316600a45e1SMiao Xie goto again; 1317600a45e1SMiao Xie } 1318600a45e1SMiao Xie 13194cb5300bSChris Mason pages[i] = page; 13204cb5300bSChris Mason i_done++; 13214cb5300bSChris Mason } 13224cb5300bSChris Mason if (!i_done || ret) 13234cb5300bSChris Mason goto out; 13244cb5300bSChris Mason 13251751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13264cb5300bSChris Mason goto out; 13274cb5300bSChris Mason 13284cb5300bSChris Mason /* 13294cb5300bSChris Mason * so now we have a nice long stream of locked 13304cb5300bSChris Mason * and up to date pages, lets wait on them 13314cb5300bSChris Mason */ 13324cb5300bSChris Mason for (i = 0; i < i_done; i++) 13334cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13344cb5300bSChris Mason 13354cb5300bSChris Mason page_start = page_offset(pages[0]); 133609cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13374cb5300bSChris Mason 13384cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1339ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13404cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 1341e182163dSOmar Sandoval page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | 1342e182163dSOmar Sandoval EXTENT_DEFRAG, 0, 0, &cached_state); 13434cb5300bSChris Mason 13441f12bd06SLiu Bo if (i_done != page_cnt) { 13459e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 134628c4a3e2SSu Yue btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); 13479e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1348bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 134909cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 135043b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13514cb5300bSChris Mason } 13524cb5300bSChris Mason 13534cb5300bSChris Mason 13549e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1355018ed4f7SDavid Sterba &cached_state); 13564cb5300bSChris Mason 13574cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1358e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13594cb5300bSChris Mason 13604cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13614cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13624cb5300bSChris Mason ClearPageChecked(pages[i]); 13634cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13644cb5300bSChris Mason set_page_dirty(pages[i]); 13654cb5300bSChris Mason unlock_page(pages[i]); 136609cbfeafSKirill A. Shutemov put_page(pages[i]); 13674cb5300bSChris Mason } 13688702ba93SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT); 1369364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13704cb5300bSChris Mason return i_done; 13714cb5300bSChris Mason out: 13724cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13734cb5300bSChris Mason unlock_page(pages[i]); 137409cbfeafSKirill A. Shutemov put_page(pages[i]); 13754cb5300bSChris Mason } 1376bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 137709cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 137843b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 13798702ba93SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT); 1380364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13814cb5300bSChris Mason return ret; 13824cb5300bSChris Mason 13834cb5300bSChris Mason } 13844cb5300bSChris Mason 13854cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13864cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13874cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13884cb5300bSChris Mason { 13890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13904cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13914cb5300bSChris Mason struct file_ra_state *ra = NULL; 13924cb5300bSChris Mason unsigned long last_index; 1393151a31b2SLi Zefan u64 isize = i_size_read(inode); 1394940100a4SChris Mason u64 last_len = 0; 1395940100a4SChris Mason u64 skip = 0; 1396940100a4SChris Mason u64 defrag_end = 0; 13974cb5300bSChris Mason u64 newer_off = range->start; 1398f46b5a66SChristoph Hellwig unsigned long i; 1399008873eaSLi Zefan unsigned long ra_index = 0; 1400f46b5a66SChristoph Hellwig int ret; 14014cb5300bSChris Mason int defrag_count = 0; 14021a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1403aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 140409cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1405c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1406ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 14074cb5300bSChris Mason struct page **pages = NULL; 14081e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14094cb5300bSChris Mason 14100abd5b17SLiu Bo if (isize == 0) 14110abd5b17SLiu Bo return 0; 14120abd5b17SLiu Bo 14130abd5b17SLiu Bo if (range->start >= isize) 14140abd5b17SLiu Bo return -EINVAL; 14151a419d85SLi Zefan 14161e2ef46dSDavid Sterba if (do_compress) { 1417ce96b7ffSChengguang Xu if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES) 14181a419d85SLi Zefan return -EINVAL; 14191a419d85SLi Zefan if (range->compress_type) 14201a419d85SLi Zefan compress_type = range->compress_type; 14211a419d85SLi Zefan } 1422f46b5a66SChristoph Hellwig 14230abd5b17SLiu Bo if (extent_thresh == 0) 1424ee22184bSByongho Lee extent_thresh = SZ_256K; 1425f46b5a66SChristoph Hellwig 14264cb5300bSChris Mason /* 14270a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14280a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14290a52d108SDavid Sterba * we don't error out. 14304cb5300bSChris Mason */ 14314cb5300bSChris Mason if (!file) { 143263e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14330a52d108SDavid Sterba if (ra) 14344cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14354cb5300bSChris Mason } else { 14364cb5300bSChris Mason ra = &file->f_ra; 14374cb5300bSChris Mason } 14384cb5300bSChris Mason 143963e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14404cb5300bSChris Mason if (!pages) { 14414cb5300bSChris Mason ret = -ENOMEM; 14424cb5300bSChris Mason goto out_ra; 14434cb5300bSChris Mason } 14444cb5300bSChris Mason 14454cb5300bSChris Mason /* find the last page to defrag */ 14461e701a32SChris Mason if (range->start + range->len > range->start) { 1447151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 144809cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14491e701a32SChris Mason } else { 145009cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14511e701a32SChris Mason } 14521e701a32SChris Mason 14534cb5300bSChris Mason if (newer_than) { 14544cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1455ee22184bSByongho Lee &newer_off, SZ_64K); 14564cb5300bSChris Mason if (!ret) { 14574cb5300bSChris Mason range->start = newer_off; 14584cb5300bSChris Mason /* 14594cb5300bSChris Mason * we always align our defrag to help keep 14604cb5300bSChris Mason * the extents in the file evenly spaced 14614cb5300bSChris Mason */ 146209cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14634cb5300bSChris Mason } else 14644cb5300bSChris Mason goto out_ra; 14654cb5300bSChris Mason } else { 146609cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14674cb5300bSChris Mason } 14684cb5300bSChris Mason if (!max_to_defrag) 1469070034bdSchandan max_to_defrag = last_index - i + 1; 14704cb5300bSChris Mason 14712a0f7f57SLi Zefan /* 14722a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14732a0f7f57SLi Zefan * written sequentially. 14742a0f7f57SLi Zefan */ 14752a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14762a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14772a0f7f57SLi Zefan 1478f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 147909cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14804cb5300bSChris Mason /* 14814cb5300bSChris Mason * make sure we stop running if someone unmounts 14824cb5300bSChris Mason * the FS 14834cb5300bSChris Mason */ 14841751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14854cb5300bSChris Mason break; 14864cb5300bSChris Mason 14870b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14880b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1489210549ebSDavid Sterba ret = -EAGAIN; 1490210549ebSDavid Sterba break; 1491210549ebSDavid Sterba } 1492210549ebSDavid Sterba 149309cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14946c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14951e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1496940100a4SChris Mason unsigned long next; 1497940100a4SChris Mason /* 1498940100a4SChris Mason * the should_defrag function tells us how much to skip 1499940100a4SChris Mason * bump our counter by the suggested amount 1500940100a4SChris Mason */ 150109cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1502940100a4SChris Mason i = max(i + 1, next); 1503940100a4SChris Mason continue; 1504940100a4SChris Mason } 1505008873eaSLi Zefan 1506008873eaSLi Zefan if (!newer_than) { 150709cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 150809cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1509008873eaSLi Zefan cluster = min(cluster, max_cluster); 1510008873eaSLi Zefan } else { 1511008873eaSLi Zefan cluster = max_cluster; 1512008873eaSLi Zefan } 1513008873eaSLi Zefan 1514008873eaSLi Zefan if (i + cluster > ra_index) { 1515008873eaSLi Zefan ra_index = max(i, ra_index); 15160a52d108SDavid Sterba if (ra) 1517d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1518d3c0bab5SDavid Sterba file, ra_index, cluster); 1519e4826a5bSchandan ra_index += cluster; 1520008873eaSLi Zefan } 15214cb5300bSChris Mason 15225955102cSAl Viro inode_lock(inode); 1523eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1524eede2bf3SOmar Sandoval ret = -ETXTBSY; 1525eede2bf3SOmar Sandoval } else { 15261e2ef46dSDavid Sterba if (do_compress) 1527eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1528008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1529eede2bf3SOmar Sandoval } 1530ecb8bea8SLiu Bo if (ret < 0) { 15315955102cSAl Viro inode_unlock(inode); 15324cb5300bSChris Mason goto out_ra; 1533ecb8bea8SLiu Bo } 15344cb5300bSChris Mason 15354cb5300bSChris Mason defrag_count += ret; 1536d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15375955102cSAl Viro inode_unlock(inode); 15384cb5300bSChris Mason 15394cb5300bSChris Mason if (newer_than) { 15404cb5300bSChris Mason if (newer_off == (u64)-1) 15414cb5300bSChris Mason break; 15424cb5300bSChris Mason 1543e1f041e1SLiu Bo if (ret > 0) 1544e1f041e1SLiu Bo i += ret; 1545e1f041e1SLiu Bo 15464cb5300bSChris Mason newer_off = max(newer_off + 1, 154709cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15484cb5300bSChris Mason 1549ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1550ee22184bSByongho Lee &newer_off, SZ_64K); 15514cb5300bSChris Mason if (!ret) { 15524cb5300bSChris Mason range->start = newer_off; 155309cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15544cb5300bSChris Mason } else { 15554cb5300bSChris Mason break; 1556940100a4SChris Mason } 15574cb5300bSChris Mason } else { 1558008873eaSLi Zefan if (ret > 0) { 1559cbcc8326SLi Zefan i += ret; 156009cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1561008873eaSLi Zefan } else { 1562940100a4SChris Mason i++; 1563008873eaSLi Zefan last_len = 0; 1564008873eaSLi Zefan } 1565f46b5a66SChristoph Hellwig } 15664cb5300bSChris Mason } 1567f46b5a66SChristoph Hellwig 1568dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15691e701a32SChris Mason filemap_flush(inode->i_mapping); 1570dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1571dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1572dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1573dec8ef90SFilipe Manana } 15741e701a32SChris Mason 15751a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15760b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15775c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15785c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15791a419d85SLi Zefan } 15801a419d85SLi Zefan 158160ccf82fSDiego Calleja ret = defrag_count; 1582940100a4SChris Mason 15834cb5300bSChris Mason out_ra: 15841e2ef46dSDavid Sterba if (do_compress) { 15855955102cSAl Viro inode_lock(inode); 1586eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15875955102cSAl Viro inode_unlock(inode); 1588633085c7SFilipe David Borba Manana } 15894cb5300bSChris Mason if (!file) 15904cb5300bSChris Mason kfree(ra); 15914cb5300bSChris Mason kfree(pages); 1592940100a4SChris Mason return ret; 1593f46b5a66SChristoph Hellwig } 1594f46b5a66SChristoph Hellwig 1595198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 159676dda93cSYan, Zheng void __user *arg) 1597f46b5a66SChristoph Hellwig { 15980b246afaSJeff Mahoney struct inode *inode = file_inode(file); 15990b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1600f46b5a66SChristoph Hellwig u64 new_size; 1601f46b5a66SChristoph Hellwig u64 old_size; 1602f46b5a66SChristoph Hellwig u64 devid = 1; 16030b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1604f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1605f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1606f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1607f46b5a66SChristoph Hellwig char *sizestr; 16089a40f122SGui Hecheng char *retptr; 1609f46b5a66SChristoph Hellwig char *devstr = NULL; 1610f46b5a66SChristoph Hellwig int ret = 0; 1611f46b5a66SChristoph Hellwig int mod = 0; 1612f46b5a66SChristoph Hellwig 1613e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1614e441d54dSChris Mason return -EPERM; 1615e441d54dSChris Mason 1616198605a8SMiao Xie ret = mnt_want_write_file(file); 1617198605a8SMiao Xie if (ret) 1618198605a8SMiao Xie return ret; 1619198605a8SMiao Xie 1620171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 162197547676SMiao Xie mnt_drop_write_file(file); 1622e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1623c9e9f97bSIlya Dryomov } 1624c9e9f97bSIlya Dryomov 1625dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1626c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1627c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1628c9e9f97bSIlya Dryomov goto out; 1629c9e9f97bSIlya Dryomov } 16305516e595SMark Fasheh 16315516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1632f46b5a66SChristoph Hellwig 1633f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1634f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1635f46b5a66SChristoph Hellwig if (devstr) { 1636f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1637f46b5a66SChristoph Hellwig *devstr = '\0'; 1638f46b5a66SChristoph Hellwig devstr = vol_args->name; 163958dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 164058dfae63SZhangZhen if (ret) 164158dfae63SZhangZhen goto out_free; 1642dfd79829SMiao Xie if (!devid) { 1643dfd79829SMiao Xie ret = -EINVAL; 1644dfd79829SMiao Xie goto out_free; 1645dfd79829SMiao Xie } 16460b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1647f46b5a66SChristoph Hellwig } 1648dba60f3fSMiao Xie 164909ba3bc9SAnand Jain device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true); 1650f46b5a66SChristoph Hellwig if (!device) { 16510b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1652c1c9ff7cSGeert Uytterhoeven devid); 1653dfd79829SMiao Xie ret = -ENODEV; 1654c9e9f97bSIlya Dryomov goto out_free; 1655f46b5a66SChristoph Hellwig } 1656dba60f3fSMiao Xie 1657ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16580b246afaSJeff Mahoney btrfs_info(fs_info, 1659efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1660c1c9ff7cSGeert Uytterhoeven devid); 1661dfd79829SMiao Xie ret = -EPERM; 16624e42ae1bSLiu Bo goto out_free; 16634e42ae1bSLiu Bo } 16644e42ae1bSLiu Bo 1665f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1666f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1667f46b5a66SChristoph Hellwig else { 1668f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1669f46b5a66SChristoph Hellwig mod = -1; 1670f46b5a66SChristoph Hellwig sizestr++; 1671f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1672f46b5a66SChristoph Hellwig mod = 1; 1673f46b5a66SChristoph Hellwig sizestr++; 1674f46b5a66SChristoph Hellwig } 16759a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16769a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1677f46b5a66SChristoph Hellwig ret = -EINVAL; 1678c9e9f97bSIlya Dryomov goto out_free; 1679f46b5a66SChristoph Hellwig } 1680f46b5a66SChristoph Hellwig } 1681f46b5a66SChristoph Hellwig 1682401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1683dfd79829SMiao Xie ret = -EPERM; 168463a212abSStefan Behrens goto out_free; 168563a212abSStefan Behrens } 168663a212abSStefan Behrens 16877cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1688f46b5a66SChristoph Hellwig 1689f46b5a66SChristoph Hellwig if (mod < 0) { 1690f46b5a66SChristoph Hellwig if (new_size > old_size) { 1691f46b5a66SChristoph Hellwig ret = -EINVAL; 1692c9e9f97bSIlya Dryomov goto out_free; 1693f46b5a66SChristoph Hellwig } 1694f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1695f46b5a66SChristoph Hellwig } else if (mod > 0) { 1696eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1697902c68a4SGui Hecheng ret = -ERANGE; 1698eb8052e0SWenliang Fan goto out_free; 1699eb8052e0SWenliang Fan } 1700f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1701f46b5a66SChristoph Hellwig } 1702f46b5a66SChristoph Hellwig 1703ee22184bSByongho Lee if (new_size < SZ_256M) { 1704f46b5a66SChristoph Hellwig ret = -EINVAL; 1705c9e9f97bSIlya Dryomov goto out_free; 1706f46b5a66SChristoph Hellwig } 1707f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1708f46b5a66SChristoph Hellwig ret = -EFBIG; 1709c9e9f97bSIlya Dryomov goto out_free; 1710f46b5a66SChristoph Hellwig } 1711f46b5a66SChristoph Hellwig 171247f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1713f46b5a66SChristoph Hellwig 17140b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1715c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1716f46b5a66SChristoph Hellwig 1717f46b5a66SChristoph Hellwig if (new_size > old_size) { 1718a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 171998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 172098d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1721c9e9f97bSIlya Dryomov goto out_free; 172298d5dc13STsutomu Itoh } 1723f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17243a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1725ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1726f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17270253f40eSjeff.liu } /* equal, nothing need to do */ 1728f46b5a66SChristoph Hellwig 1729c9e9f97bSIlya Dryomov out_free: 1730f46b5a66SChristoph Hellwig kfree(vol_args); 1731c9e9f97bSIlya Dryomov out: 1732171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 173318f39c41SIlya Dryomov mnt_drop_write_file(file); 1734f46b5a66SChristoph Hellwig return ret; 1735f46b5a66SChristoph Hellwig } 1736f46b5a66SChristoph Hellwig 173772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 173852f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17396f72c7e2SArne Jansen u64 *transid, bool readonly, 17408696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1741f46b5a66SChristoph Hellwig { 1742f46b5a66SChristoph Hellwig int namelen; 17433de4586cSChris Mason int ret = 0; 1744f46b5a66SChristoph Hellwig 1745325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1746325c50e3SJeff Mahoney return -ENOTDIR; 1747325c50e3SJeff Mahoney 1748a874a63eSLiu Bo ret = mnt_want_write_file(file); 1749a874a63eSLiu Bo if (ret) 1750a874a63eSLiu Bo goto out; 1751a874a63eSLiu Bo 175272fd032eSSage Weil namelen = strlen(name); 175372fd032eSSage Weil if (strchr(name, '/')) { 1754f46b5a66SChristoph Hellwig ret = -EINVAL; 1755a874a63eSLiu Bo goto out_drop_write; 1756f46b5a66SChristoph Hellwig } 1757f46b5a66SChristoph Hellwig 175816780cabSChris Mason if (name[0] == '.' && 175916780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 176016780cabSChris Mason ret = -EEXIST; 1761a874a63eSLiu Bo goto out_drop_write; 176216780cabSChris Mason } 176316780cabSChris Mason 17643de4586cSChris Mason if (subvol) { 176572fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17666f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1767cb8e7090SChristoph Hellwig } else { 17682903ff01SAl Viro struct fd src = fdget(fd); 17693de4586cSChris Mason struct inode *src_inode; 17702903ff01SAl Viro if (!src.file) { 17713de4586cSChris Mason ret = -EINVAL; 1772a874a63eSLiu Bo goto out_drop_write; 17733de4586cSChris Mason } 17743de4586cSChris Mason 1775496ad9aaSAl Viro src_inode = file_inode(src.file); 1776496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1777c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1778efe120a0SFrank Holton "Snapshot src from another FS"); 177923ad5b17SKusanagi Kouichi ret = -EXDEV; 1780d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1781d0242061SDavid Sterba /* 1782d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1783d0242061SDavid Sterba * are limited to own subvolumes only 1784d0242061SDavid Sterba */ 1785d0242061SDavid Sterba ret = -EPERM; 1786ecd18815SAl Viro } else { 178772fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 178872fd032eSSage Weil BTRFS_I(src_inode)->root, 17896f72c7e2SArne Jansen transid, readonly, inherit); 1790ecd18815SAl Viro } 17912903ff01SAl Viro fdput(src); 1792cb8e7090SChristoph Hellwig } 1793a874a63eSLiu Bo out_drop_write: 1794a874a63eSLiu Bo mnt_drop_write_file(file); 1795f46b5a66SChristoph Hellwig out: 179672fd032eSSage Weil return ret; 179772fd032eSSage Weil } 179872fd032eSSage Weil 179972fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1800fa0d2b9bSLi Zefan void __user *arg, int subvol) 180172fd032eSSage Weil { 1802fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 180372fd032eSSage Weil int ret; 180472fd032eSSage Weil 1805325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1806325c50e3SJeff Mahoney return -ENOTDIR; 1807325c50e3SJeff Mahoney 1808fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1809fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1810fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1811fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1812fa0d2b9bSLi Zefan 1813fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1814b83cc969SLi Zefan vol_args->fd, subvol, 18156f72c7e2SArne Jansen NULL, false, NULL); 1816fa0d2b9bSLi Zefan 1817fa0d2b9bSLi Zefan kfree(vol_args); 1818fa0d2b9bSLi Zefan return ret; 1819fa0d2b9bSLi Zefan } 1820fa0d2b9bSLi Zefan 1821fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1822fa0d2b9bSLi Zefan void __user *arg, int subvol) 1823fa0d2b9bSLi Zefan { 1824fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1825fa0d2b9bSLi Zefan int ret; 1826fdfb1e4fSLi Zefan u64 transid = 0; 1827fdfb1e4fSLi Zefan u64 *ptr = NULL; 1828b83cc969SLi Zefan bool readonly = false; 18296f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 183072fd032eSSage Weil 1831325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1832325c50e3SJeff Mahoney return -ENOTDIR; 1833325c50e3SJeff Mahoney 1834fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1835fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1836fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1837fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1838fdfb1e4fSLi Zefan 1839b83cc969SLi Zefan if (vol_args->flags & 18406f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18416f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1842b83cc969SLi Zefan ret = -EOPNOTSUPP; 1843c47ca32dSDan Carpenter goto free_args; 1844fdfb1e4fSLi Zefan } 1845fdfb1e4fSLi Zefan 1846ebc87351SNikolay Borisov if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1847ebc87351SNikolay Borisov struct inode *inode = file_inode(file); 1848ebc87351SNikolay Borisov struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1849ebc87351SNikolay Borisov 1850ebc87351SNikolay Borisov btrfs_warn(fs_info, 1851ebc87351SNikolay Borisov "SNAP_CREATE_V2 ioctl with CREATE_ASYNC is deprecated and will be removed in kernel 5.7"); 1852ebc87351SNikolay Borisov 1853fdfb1e4fSLi Zefan ptr = &transid; 1854ebc87351SNikolay Borisov } 1855b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1856b83cc969SLi Zefan readonly = true; 18576f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 185809cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18596f72c7e2SArne Jansen ret = -EINVAL; 1860c47ca32dSDan Carpenter goto free_args; 18616f72c7e2SArne Jansen } 18626f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18636f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18646f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1865c47ca32dSDan Carpenter goto free_args; 18666f72c7e2SArne Jansen } 18676f72c7e2SArne Jansen } 186875eaa0e2SSage Weil 1869fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18706f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18718696c533SMiao Xie readonly, inherit); 1872c47ca32dSDan Carpenter if (ret) 1873c47ca32dSDan Carpenter goto free_inherit; 187475eaa0e2SSage Weil 1875c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1876fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1877c47ca32dSDan Carpenter transid), 1878c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 187975eaa0e2SSage Weil ret = -EFAULT; 1880c47ca32dSDan Carpenter 1881c47ca32dSDan Carpenter free_inherit: 18826f72c7e2SArne Jansen kfree(inherit); 1883c47ca32dSDan Carpenter free_args: 1884c47ca32dSDan Carpenter kfree(vol_args); 1885f46b5a66SChristoph Hellwig return ret; 1886f46b5a66SChristoph Hellwig } 1887f46b5a66SChristoph Hellwig 18880caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18890caa102dSLi Zefan void __user *arg) 18900caa102dSLi Zefan { 1891496ad9aaSAl Viro struct inode *inode = file_inode(file); 18920b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18930caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18940caa102dSLi Zefan int ret = 0; 18950caa102dSLi Zefan u64 flags = 0; 18960caa102dSLi Zefan 18974a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18980caa102dSLi Zefan return -EINVAL; 18990caa102dSLi Zefan 19000b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 19010caa102dSLi Zefan if (btrfs_root_readonly(root)) 19020caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 19030b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 19040caa102dSLi Zefan 19050caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 19060caa102dSLi Zefan ret = -EFAULT; 19070caa102dSLi Zefan 19080caa102dSLi Zefan return ret; 19090caa102dSLi Zefan } 19100caa102dSLi Zefan 19110caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19120caa102dSLi Zefan void __user *arg) 19130caa102dSLi Zefan { 1914496ad9aaSAl Viro struct inode *inode = file_inode(file); 19150b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19160caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19170caa102dSLi Zefan struct btrfs_trans_handle *trans; 19180caa102dSLi Zefan u64 root_flags; 19190caa102dSLi Zefan u64 flags; 19200caa102dSLi Zefan int ret = 0; 19210caa102dSLi Zefan 1922bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1923bd60ea0fSDavid Sterba return -EPERM; 1924bd60ea0fSDavid Sterba 1925b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1926b9ca0664SLiu Bo if (ret) 1927b9ca0664SLiu Bo goto out; 19280caa102dSLi Zefan 19294a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1930b9ca0664SLiu Bo ret = -EINVAL; 1931b9ca0664SLiu Bo goto out_drop_write; 1932b9ca0664SLiu Bo } 19330caa102dSLi Zefan 1934b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1935b9ca0664SLiu Bo ret = -EFAULT; 1936b9ca0664SLiu Bo goto out_drop_write; 1937b9ca0664SLiu Bo } 19380caa102dSLi Zefan 1939b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1940b9ca0664SLiu Bo ret = -EINVAL; 1941b9ca0664SLiu Bo goto out_drop_write; 1942b9ca0664SLiu Bo } 19430caa102dSLi Zefan 1944b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1945b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1946b9ca0664SLiu Bo goto out_drop_write; 1947b9ca0664SLiu Bo } 19480caa102dSLi Zefan 19490b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19500caa102dSLi Zefan 19510caa102dSLi Zefan /* nothing to do */ 19520caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1953b9ca0664SLiu Bo goto out_drop_sem; 19540caa102dSLi Zefan 19550caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19562c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19570caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19580caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19592c686537SDavid Sterba } else { 19602c686537SDavid Sterba /* 19612c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19622c686537SDavid Sterba * send 19632c686537SDavid Sterba */ 19642c686537SDavid Sterba spin_lock(&root->root_item_lock); 19652c686537SDavid Sterba if (root->send_in_progress == 0) { 19660caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19670caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19682c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19692c686537SDavid Sterba } else { 19702c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19710b246afaSJeff Mahoney btrfs_warn(fs_info, 19722c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19732c686537SDavid Sterba root->root_key.objectid); 19742c686537SDavid Sterba ret = -EPERM; 19752c686537SDavid Sterba goto out_drop_sem; 19762c686537SDavid Sterba } 19772c686537SDavid Sterba } 19780caa102dSLi Zefan 19790caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19800caa102dSLi Zefan if (IS_ERR(trans)) { 19810caa102dSLi Zefan ret = PTR_ERR(trans); 19820caa102dSLi Zefan goto out_reset; 19830caa102dSLi Zefan } 19840caa102dSLi Zefan 19850b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19860caa102dSLi Zefan &root->root_key, &root->root_item); 19879417ebc8SNikolay Borisov if (ret < 0) { 19889417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19899417ebc8SNikolay Borisov goto out_reset; 19909417ebc8SNikolay Borisov } 19910caa102dSLi Zefan 19929417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19939417ebc8SNikolay Borisov 19940caa102dSLi Zefan out_reset: 19950caa102dSLi Zefan if (ret) 19960caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1997b9ca0664SLiu Bo out_drop_sem: 19980b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1999b9ca0664SLiu Bo out_drop_write: 2000b9ca0664SLiu Bo mnt_drop_write_file(file); 2001b9ca0664SLiu Bo out: 20020caa102dSLi Zefan return ret; 20030caa102dSLi Zefan } 20040caa102dSLi Zefan 2005ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2006ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2007ac8e9819SChris Mason { 2008abc6e134SChris Mason struct btrfs_key test; 2009abc6e134SChris Mason int ret; 2010abc6e134SChris Mason 2011abc6e134SChris Mason test.objectid = sk->min_objectid; 2012abc6e134SChris Mason test.type = sk->min_type; 2013abc6e134SChris Mason test.offset = sk->min_offset; 2014abc6e134SChris Mason 2015abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2016abc6e134SChris Mason if (ret < 0) 2017ac8e9819SChris Mason return 0; 2018abc6e134SChris Mason 2019abc6e134SChris Mason test.objectid = sk->max_objectid; 2020abc6e134SChris Mason test.type = sk->max_type; 2021abc6e134SChris Mason test.offset = sk->max_offset; 2022abc6e134SChris Mason 2023abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2024abc6e134SChris Mason if (ret > 0) 2025ac8e9819SChris Mason return 0; 2026ac8e9819SChris Mason return 1; 2027ac8e9819SChris Mason } 2028ac8e9819SChris Mason 2029df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2030ac8e9819SChris Mason struct btrfs_key *key, 2031ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20329b6e817dSGerhard Heift size_t *buf_size, 2033ba346b35SGerhard Heift char __user *ubuf, 2034ac8e9819SChris Mason unsigned long *sk_offset, 2035ac8e9819SChris Mason int *num_found) 2036ac8e9819SChris Mason { 2037ac8e9819SChris Mason u64 found_transid; 2038ac8e9819SChris Mason struct extent_buffer *leaf; 2039ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2040dd81d459SNaohiro Aota struct btrfs_key test; 2041ac8e9819SChris Mason unsigned long item_off; 2042ac8e9819SChris Mason unsigned long item_len; 2043ac8e9819SChris Mason int nritems; 2044ac8e9819SChris Mason int i; 2045ac8e9819SChris Mason int slot; 2046ac8e9819SChris Mason int ret = 0; 2047ac8e9819SChris Mason 2048ac8e9819SChris Mason leaf = path->nodes[0]; 2049ac8e9819SChris Mason slot = path->slots[0]; 2050ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2051ac8e9819SChris Mason 2052ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2053ac8e9819SChris Mason i = nritems; 2054ac8e9819SChris Mason goto advance_key; 2055ac8e9819SChris Mason } 2056ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2057ac8e9819SChris Mason 2058ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2059ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2060ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2061ac8e9819SChris Mason 206203b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 206303b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 206403b71c6cSGabriel de Perthuis continue; 206503b71c6cSGabriel de Perthuis 20669b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20678f5f6178SGerhard Heift if (*num_found) { 2068ac8e9819SChris Mason ret = 1; 20698f5f6178SGerhard Heift goto out; 20708f5f6178SGerhard Heift } 20718f5f6178SGerhard Heift 20728f5f6178SGerhard Heift /* 20738f5f6178SGerhard Heift * return one empty item back for v1, which does not 20748f5f6178SGerhard Heift * handle -EOVERFLOW 20758f5f6178SGerhard Heift */ 20768f5f6178SGerhard Heift 20779b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2078ac8e9819SChris Mason item_len = 0; 20798f5f6178SGerhard Heift ret = -EOVERFLOW; 20808f5f6178SGerhard Heift } 2081ac8e9819SChris Mason 20829b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2083ac8e9819SChris Mason ret = 1; 208425c9bc2eSGerhard Heift goto out; 2085ac8e9819SChris Mason } 2086ac8e9819SChris Mason 2087ac8e9819SChris Mason sh.objectid = key->objectid; 2088ac8e9819SChris Mason sh.offset = key->offset; 2089ac8e9819SChris Mason sh.type = key->type; 2090ac8e9819SChris Mason sh.len = item_len; 2091ac8e9819SChris Mason sh.transid = found_transid; 2092ac8e9819SChris Mason 2093ac8e9819SChris Mason /* copy search result header */ 2094ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2095ba346b35SGerhard Heift ret = -EFAULT; 2096ba346b35SGerhard Heift goto out; 2097ba346b35SGerhard Heift } 2098ba346b35SGerhard Heift 2099ac8e9819SChris Mason *sk_offset += sizeof(sh); 2100ac8e9819SChris Mason 2101ac8e9819SChris Mason if (item_len) { 2102ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2103ac8e9819SChris Mason /* copy the item */ 2104ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2105ba346b35SGerhard Heift item_off, item_len)) { 2106ba346b35SGerhard Heift ret = -EFAULT; 2107ba346b35SGerhard Heift goto out; 2108ba346b35SGerhard Heift } 2109ba346b35SGerhard Heift 2110ac8e9819SChris Mason *sk_offset += item_len; 2111ac8e9819SChris Mason } 2112e2156867SHugo Mills (*num_found)++; 2113ac8e9819SChris Mason 21148f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21158f5f6178SGerhard Heift goto out; 21168f5f6178SGerhard Heift 211725c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 211825c9bc2eSGerhard Heift ret = 1; 211925c9bc2eSGerhard Heift goto out; 212025c9bc2eSGerhard Heift } 2121ac8e9819SChris Mason } 2122ac8e9819SChris Mason advance_key: 2123ac8e9819SChris Mason ret = 0; 2124dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2125dd81d459SNaohiro Aota test.type = sk->max_type; 2126dd81d459SNaohiro Aota test.offset = sk->max_offset; 2127dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2128dd81d459SNaohiro Aota ret = 1; 2129dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2130abc6e134SChris Mason key->offset++; 2131dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2132abc6e134SChris Mason key->offset = 0; 2133abc6e134SChris Mason key->type++; 2134dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2135abc6e134SChris Mason key->offset = 0; 2136abc6e134SChris Mason key->type = 0; 2137abc6e134SChris Mason key->objectid++; 2138abc6e134SChris Mason } else 2139abc6e134SChris Mason ret = 1; 214025c9bc2eSGerhard Heift out: 2141ba346b35SGerhard Heift /* 2142ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2143ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2144ba346b35SGerhard Heift * * all items were found 2145ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2146ba346b35SGerhard Heift * leaf 2147ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2148ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2149ba346b35SGerhard Heift */ 2150ac8e9819SChris Mason return ret; 2151ac8e9819SChris Mason } 2152ac8e9819SChris Mason 2153ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 215412544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21559b6e817dSGerhard Heift size_t *buf_size, 2156ba346b35SGerhard Heift char __user *ubuf) 2157ac8e9819SChris Mason { 21580b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2159ac8e9819SChris Mason struct btrfs_root *root; 2160ac8e9819SChris Mason struct btrfs_key key; 2161ac8e9819SChris Mason struct btrfs_path *path; 2162ac8e9819SChris Mason int ret; 2163ac8e9819SChris Mason int num_found = 0; 2164ac8e9819SChris Mason unsigned long sk_offset = 0; 2165ac8e9819SChris Mason 21669b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21679b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 216812544442SGerhard Heift return -EOVERFLOW; 21699b6e817dSGerhard Heift } 217012544442SGerhard Heift 2171ac8e9819SChris Mason path = btrfs_alloc_path(); 2172ac8e9819SChris Mason if (!path) 2173ac8e9819SChris Mason return -ENOMEM; 2174ac8e9819SChris Mason 2175ac8e9819SChris Mason if (sk->tree_id == 0) { 2176ac8e9819SChris Mason /* search the root of the inode that was passed */ 2177ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2178ac8e9819SChris Mason } else { 2179ac8e9819SChris Mason key.objectid = sk->tree_id; 2180ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2181ac8e9819SChris Mason key.offset = (u64)-1; 2182ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2183ac8e9819SChris Mason if (IS_ERR(root)) { 2184ac8e9819SChris Mason btrfs_free_path(path); 2185ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2186ac8e9819SChris Mason } 2187ac8e9819SChris Mason } 2188ac8e9819SChris Mason 2189ac8e9819SChris Mason key.objectid = sk->min_objectid; 2190ac8e9819SChris Mason key.type = sk->min_type; 2191ac8e9819SChris Mason key.offset = sk->min_offset; 2192ac8e9819SChris Mason 2193ac8e9819SChris Mason while (1) { 21946174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2195ac8e9819SChris Mason if (ret != 0) { 2196ac8e9819SChris Mason if (ret > 0) 2197ac8e9819SChris Mason ret = 0; 2198ac8e9819SChris Mason goto err; 2199ac8e9819SChris Mason } 2200df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2201ac8e9819SChris Mason &sk_offset, &num_found); 2202b3b4aa74SDavid Sterba btrfs_release_path(path); 220325c9bc2eSGerhard Heift if (ret) 2204ac8e9819SChris Mason break; 2205ac8e9819SChris Mason 2206ac8e9819SChris Mason } 22078f5f6178SGerhard Heift if (ret > 0) 2208ac8e9819SChris Mason ret = 0; 2209ac8e9819SChris Mason err: 2210ac8e9819SChris Mason sk->nr_items = num_found; 2211ac8e9819SChris Mason btrfs_free_path(path); 2212ac8e9819SChris Mason return ret; 2213ac8e9819SChris Mason } 2214ac8e9819SChris Mason 2215ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2216ac8e9819SChris Mason void __user *argp) 2217ac8e9819SChris Mason { 2218ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2219ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2220ac8e9819SChris Mason struct inode *inode; 2221ac8e9819SChris Mason int ret; 22229b6e817dSGerhard Heift size_t buf_size; 2223ac8e9819SChris Mason 2224ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2225ac8e9819SChris Mason return -EPERM; 2226ac8e9819SChris Mason 2227ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2228ac8e9819SChris Mason 2229ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2230ba346b35SGerhard Heift return -EFAULT; 2231ba346b35SGerhard Heift 2232ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2233ac8e9819SChris Mason 2234496ad9aaSAl Viro inode = file_inode(file); 2235ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22368f5f6178SGerhard Heift 22378f5f6178SGerhard Heift /* 22388f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22398f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22408f5f6178SGerhard Heift */ 22418f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22428f5f6178SGerhard Heift ret = 0; 22438f5f6178SGerhard Heift 2244ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2245ac8e9819SChris Mason ret = -EFAULT; 2246ac8e9819SChris Mason return ret; 2247ac8e9819SChris Mason } 2248ac8e9819SChris Mason 2249cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2250cc68a8a5SGerhard Heift void __user *argp) 2251cc68a8a5SGerhard Heift { 2252cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2253cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2254cc68a8a5SGerhard Heift struct inode *inode; 2255cc68a8a5SGerhard Heift int ret; 2256cc68a8a5SGerhard Heift size_t buf_size; 2257ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2258cc68a8a5SGerhard Heift 2259cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2260cc68a8a5SGerhard Heift return -EPERM; 2261cc68a8a5SGerhard Heift 2262cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2263cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2264cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2265cc68a8a5SGerhard Heift return -EFAULT; 2266cc68a8a5SGerhard Heift 2267cc68a8a5SGerhard Heift buf_size = args.buf_size; 2268cc68a8a5SGerhard Heift 2269cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2270cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2271cc68a8a5SGerhard Heift buf_size = buf_limit; 2272cc68a8a5SGerhard Heift 2273cc68a8a5SGerhard Heift inode = file_inode(file); 2274cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2275718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2276cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2277cc68a8a5SGerhard Heift ret = -EFAULT; 2278cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2279cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2280cc68a8a5SGerhard Heift ret = -EFAULT; 2281cc68a8a5SGerhard Heift 228276dda93cSYan, Zheng return ret; 228376dda93cSYan, Zheng } 228476dda93cSYan, Zheng 228598d377a0STARUISI Hiroaki /* 2286ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2287ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 228898d377a0STARUISI Hiroaki */ 228998d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 229098d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 229198d377a0STARUISI Hiroaki { 229298d377a0STARUISI Hiroaki struct btrfs_root *root; 229398d377a0STARUISI Hiroaki struct btrfs_key key; 2294ac8e9819SChris Mason char *ptr; 229598d377a0STARUISI Hiroaki int ret = -1; 229698d377a0STARUISI Hiroaki int slot; 229798d377a0STARUISI Hiroaki int len; 229898d377a0STARUISI Hiroaki int total_len = 0; 229998d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 230098d377a0STARUISI Hiroaki struct extent_buffer *l; 230198d377a0STARUISI Hiroaki struct btrfs_path *path; 230298d377a0STARUISI Hiroaki 230398d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 230498d377a0STARUISI Hiroaki name[0]='\0'; 230598d377a0STARUISI Hiroaki return 0; 230698d377a0STARUISI Hiroaki } 230798d377a0STARUISI Hiroaki 230898d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 230998d377a0STARUISI Hiroaki if (!path) 231098d377a0STARUISI Hiroaki return -ENOMEM; 231198d377a0STARUISI Hiroaki 2312c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 231398d377a0STARUISI Hiroaki 231498d377a0STARUISI Hiroaki key.objectid = tree_id; 231598d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 231698d377a0STARUISI Hiroaki key.offset = (u64)-1; 231798d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 231898d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2319ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 23208ad6fcabSChris Mason goto out; 232198d377a0STARUISI Hiroaki } 232298d377a0STARUISI Hiroaki 232398d377a0STARUISI Hiroaki key.objectid = dirid; 232498d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23258ad6fcabSChris Mason key.offset = (u64)-1; 232698d377a0STARUISI Hiroaki 232798d377a0STARUISI Hiroaki while (1) { 232898d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 232998d377a0STARUISI Hiroaki if (ret < 0) 233098d377a0STARUISI Hiroaki goto out; 233118674c6cSFilipe David Borba Manana else if (ret > 0) { 233218674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 233318674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 233418674c6cSFilipe David Borba Manana if (ret < 0) 233518674c6cSFilipe David Borba Manana goto out; 233618674c6cSFilipe David Borba Manana else if (ret > 0) { 2337ac8e9819SChris Mason ret = -ENOENT; 233898d377a0STARUISI Hiroaki goto out; 2339ac8e9819SChris Mason } 234018674c6cSFilipe David Borba Manana } 234118674c6cSFilipe David Borba Manana 234218674c6cSFilipe David Borba Manana l = path->nodes[0]; 234318674c6cSFilipe David Borba Manana slot = path->slots[0]; 234418674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 234598d377a0STARUISI Hiroaki 234698d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 234798d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 234898d377a0STARUISI Hiroaki ptr -= len + 1; 234998d377a0STARUISI Hiroaki total_len += len + 1; 2350a696cf35SFilipe David Borba Manana if (ptr < name) { 2351a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 235298d377a0STARUISI Hiroaki goto out; 2353a696cf35SFilipe David Borba Manana } 235498d377a0STARUISI Hiroaki 235598d377a0STARUISI Hiroaki *(ptr + len) = '/'; 235698d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 235798d377a0STARUISI Hiroaki 235898d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 235998d377a0STARUISI Hiroaki break; 236098d377a0STARUISI Hiroaki 2361b3b4aa74SDavid Sterba btrfs_release_path(path); 236298d377a0STARUISI Hiroaki key.objectid = key.offset; 23638ad6fcabSChris Mason key.offset = (u64)-1; 236498d377a0STARUISI Hiroaki dirid = key.objectid; 236598d377a0STARUISI Hiroaki } 236677906a50SLi Zefan memmove(name, ptr, total_len); 236798d377a0STARUISI Hiroaki name[total_len] = '\0'; 236898d377a0STARUISI Hiroaki ret = 0; 236998d377a0STARUISI Hiroaki out: 237098d377a0STARUISI Hiroaki btrfs_free_path(path); 2371ac8e9819SChris Mason return ret; 2372ac8e9819SChris Mason } 2373ac8e9819SChris Mason 237423d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 237523d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 237623d0b79dSTomohiro Misono { 237723d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 237823d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 237923d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 238023d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 238123d0b79dSTomohiro Misono u64 dirid = args->dirid; 238223d0b79dSTomohiro Misono unsigned long item_off; 238323d0b79dSTomohiro Misono unsigned long item_len; 238423d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 238523d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 238623d0b79dSTomohiro Misono struct btrfs_root *root; 238723d0b79dSTomohiro Misono struct btrfs_path *path; 238823d0b79dSTomohiro Misono struct btrfs_key key, key2; 238923d0b79dSTomohiro Misono struct extent_buffer *leaf; 239023d0b79dSTomohiro Misono struct inode *temp_inode; 239123d0b79dSTomohiro Misono char *ptr; 239223d0b79dSTomohiro Misono int slot; 239323d0b79dSTomohiro Misono int len; 239423d0b79dSTomohiro Misono int total_len = 0; 239523d0b79dSTomohiro Misono int ret; 239623d0b79dSTomohiro Misono 239723d0b79dSTomohiro Misono path = btrfs_alloc_path(); 239823d0b79dSTomohiro Misono if (!path) 239923d0b79dSTomohiro Misono return -ENOMEM; 240023d0b79dSTomohiro Misono 240123d0b79dSTomohiro Misono /* 240223d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 240323d0b79dSTomohiro Misono * construct the path in from the bottom up. 240423d0b79dSTomohiro Misono */ 240523d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 240623d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 240723d0b79dSTomohiro Misono 240823d0b79dSTomohiro Misono key.objectid = treeid; 240923d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 241023d0b79dSTomohiro Misono key.offset = (u64)-1; 241123d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 241223d0b79dSTomohiro Misono if (IS_ERR(root)) { 241323d0b79dSTomohiro Misono ret = PTR_ERR(root); 241423d0b79dSTomohiro Misono goto out; 241523d0b79dSTomohiro Misono } 241623d0b79dSTomohiro Misono 241723d0b79dSTomohiro Misono key.objectid = dirid; 241823d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 241923d0b79dSTomohiro Misono key.offset = (u64)-1; 242023d0b79dSTomohiro Misono while (1) { 242123d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 242223d0b79dSTomohiro Misono if (ret < 0) { 242323d0b79dSTomohiro Misono goto out; 242423d0b79dSTomohiro Misono } else if (ret > 0) { 242523d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 242623d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 242723d0b79dSTomohiro Misono if (ret < 0) { 242823d0b79dSTomohiro Misono goto out; 242923d0b79dSTomohiro Misono } else if (ret > 0) { 243023d0b79dSTomohiro Misono ret = -ENOENT; 243123d0b79dSTomohiro Misono goto out; 243223d0b79dSTomohiro Misono } 243323d0b79dSTomohiro Misono } 243423d0b79dSTomohiro Misono 243523d0b79dSTomohiro Misono leaf = path->nodes[0]; 243623d0b79dSTomohiro Misono slot = path->slots[0]; 243723d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 243823d0b79dSTomohiro Misono 243923d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 244023d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 244123d0b79dSTomohiro Misono ptr -= len + 1; 244223d0b79dSTomohiro Misono total_len += len + 1; 244323d0b79dSTomohiro Misono if (ptr < args->path) { 244423d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 244523d0b79dSTomohiro Misono goto out; 244623d0b79dSTomohiro Misono } 244723d0b79dSTomohiro Misono 244823d0b79dSTomohiro Misono *(ptr + len) = '/'; 244923d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 245023d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 245123d0b79dSTomohiro Misono 245223d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 245323d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 245423d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 245523d0b79dSTomohiro Misono if (ret < 0) { 245623d0b79dSTomohiro Misono goto out; 245723d0b79dSTomohiro Misono } else if (ret > 0) { 245823d0b79dSTomohiro Misono ret = -ENOENT; 245923d0b79dSTomohiro Misono goto out; 246023d0b79dSTomohiro Misono } 246123d0b79dSTomohiro Misono 246223d0b79dSTomohiro Misono leaf = path->nodes[0]; 246323d0b79dSTomohiro Misono slot = path->slots[0]; 246423d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 246523d0b79dSTomohiro Misono if (key2.objectid != dirid) { 246623d0b79dSTomohiro Misono ret = -ENOENT; 246723d0b79dSTomohiro Misono goto out; 246823d0b79dSTomohiro Misono } 246923d0b79dSTomohiro Misono 24704c66e0d4SDavid Sterba temp_inode = btrfs_iget(sb, &key2, root); 24713ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24723ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24733ca57bd6SMisono Tomohiro goto out; 24743ca57bd6SMisono Tomohiro } 247523d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 247623d0b79dSTomohiro Misono iput(temp_inode); 247723d0b79dSTomohiro Misono if (ret) { 247823d0b79dSTomohiro Misono ret = -EACCES; 247923d0b79dSTomohiro Misono goto out; 248023d0b79dSTomohiro Misono } 248123d0b79dSTomohiro Misono 248223d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 248323d0b79dSTomohiro Misono break; 248423d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 248523d0b79dSTomohiro Misono ret = -EACCES; 248623d0b79dSTomohiro Misono goto out; 248723d0b79dSTomohiro Misono } 248823d0b79dSTomohiro Misono 248923d0b79dSTomohiro Misono btrfs_release_path(path); 249023d0b79dSTomohiro Misono key.objectid = key.offset; 249123d0b79dSTomohiro Misono key.offset = (u64)-1; 249223d0b79dSTomohiro Misono dirid = key.objectid; 249323d0b79dSTomohiro Misono } 249423d0b79dSTomohiro Misono 249523d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 249623d0b79dSTomohiro Misono args->path[total_len] = '\0'; 249723d0b79dSTomohiro Misono btrfs_release_path(path); 249823d0b79dSTomohiro Misono } 249923d0b79dSTomohiro Misono 250023d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 250123d0b79dSTomohiro Misono root = fs_info->tree_root; 250223d0b79dSTomohiro Misono key.objectid = treeid; 250323d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 250423d0b79dSTomohiro Misono key.offset = args->treeid; 250523d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 250623d0b79dSTomohiro Misono if (ret < 0) { 250723d0b79dSTomohiro Misono goto out; 250823d0b79dSTomohiro Misono } else if (ret > 0) { 250923d0b79dSTomohiro Misono ret = -ENOENT; 251023d0b79dSTomohiro Misono goto out; 251123d0b79dSTomohiro Misono } 251223d0b79dSTomohiro Misono 251323d0b79dSTomohiro Misono leaf = path->nodes[0]; 251423d0b79dSTomohiro Misono slot = path->slots[0]; 251523d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 251623d0b79dSTomohiro Misono 251723d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 251823d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 251923d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 252023d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 252123d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 252223d0b79dSTomohiro Misono ret = -EINVAL; 252323d0b79dSTomohiro Misono goto out; 252423d0b79dSTomohiro Misono } 252523d0b79dSTomohiro Misono 252623d0b79dSTomohiro Misono /* Copy subvolume's name */ 252723d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 252823d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 252923d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 253023d0b79dSTomohiro Misono args->name[item_len] = 0; 253123d0b79dSTomohiro Misono 253223d0b79dSTomohiro Misono out: 253323d0b79dSTomohiro Misono btrfs_free_path(path); 253423d0b79dSTomohiro Misono return ret; 253523d0b79dSTomohiro Misono } 253623d0b79dSTomohiro Misono 2537ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2538ac8e9819SChris Mason void __user *argp) 2539ac8e9819SChris Mason { 2540ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2541ac8e9819SChris Mason struct inode *inode; 254201b810b8SDavid Sterba int ret = 0; 2543ac8e9819SChris Mason 25442354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25452354d08fSJulia Lawall if (IS_ERR(args)) 25462354d08fSJulia Lawall return PTR_ERR(args); 2547c2b96929SDan Carpenter 2548496ad9aaSAl Viro inode = file_inode(file); 2549ac8e9819SChris Mason 255001b810b8SDavid Sterba /* 255101b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 255201b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 255301b810b8SDavid Sterba */ 25541b53ac4dSChris Mason if (args->treeid == 0) 25551b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25561b53ac4dSChris Mason 255701b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 255801b810b8SDavid Sterba args->name[0] = 0; 255901b810b8SDavid Sterba goto out; 256001b810b8SDavid Sterba } 256101b810b8SDavid Sterba 256201b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 256301b810b8SDavid Sterba ret = -EPERM; 256401b810b8SDavid Sterba goto out; 256501b810b8SDavid Sterba } 256601b810b8SDavid Sterba 2567ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2568ac8e9819SChris Mason args->treeid, args->objectid, 2569ac8e9819SChris Mason args->name); 2570ac8e9819SChris Mason 257101b810b8SDavid Sterba out: 2572ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2573ac8e9819SChris Mason ret = -EFAULT; 2574ac8e9819SChris Mason 2575ac8e9819SChris Mason kfree(args); 257698d377a0STARUISI Hiroaki return ret; 257798d377a0STARUISI Hiroaki } 257898d377a0STARUISI Hiroaki 257923d0b79dSTomohiro Misono /* 258023d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 258123d0b79dSTomohiro Misono * 258223d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 258323d0b79dSTomohiro Misono * 258423d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 258523d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 258623d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 258723d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 258823d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 258923d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 259023d0b79dSTomohiro Misono */ 259123d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 259223d0b79dSTomohiro Misono { 259323d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 259423d0b79dSTomohiro Misono struct inode *inode; 259523d0b79dSTomohiro Misono int ret; 259623d0b79dSTomohiro Misono 259723d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 259823d0b79dSTomohiro Misono if (IS_ERR(args)) 259923d0b79dSTomohiro Misono return PTR_ERR(args); 260023d0b79dSTomohiro Misono 260123d0b79dSTomohiro Misono inode = file_inode(file); 260223d0b79dSTomohiro Misono 260323d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 260423d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 260523d0b79dSTomohiro Misono /* 260623d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 260723d0b79dSTomohiro Misono * called 260823d0b79dSTomohiro Misono */ 260923d0b79dSTomohiro Misono kfree(args); 261023d0b79dSTomohiro Misono return -EACCES; 261123d0b79dSTomohiro Misono } 261223d0b79dSTomohiro Misono 261323d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 261423d0b79dSTomohiro Misono 261523d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 261623d0b79dSTomohiro Misono ret = -EFAULT; 261723d0b79dSTomohiro Misono 261823d0b79dSTomohiro Misono kfree(args); 261923d0b79dSTomohiro Misono return ret; 262023d0b79dSTomohiro Misono } 262123d0b79dSTomohiro Misono 2622b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2623b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2624b64ec075STomohiro Misono { 2625b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2626b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2627b64ec075STomohiro Misono struct btrfs_root *root; 2628b64ec075STomohiro Misono struct btrfs_path *path; 2629b64ec075STomohiro Misono struct btrfs_key key; 2630b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2631b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2632b64ec075STomohiro Misono struct extent_buffer *leaf; 2633b64ec075STomohiro Misono unsigned long item_off; 2634b64ec075STomohiro Misono unsigned long item_len; 2635b64ec075STomohiro Misono struct inode *inode; 2636b64ec075STomohiro Misono int slot; 2637b64ec075STomohiro Misono int ret = 0; 2638b64ec075STomohiro Misono 2639b64ec075STomohiro Misono path = btrfs_alloc_path(); 2640b64ec075STomohiro Misono if (!path) 2641b64ec075STomohiro Misono return -ENOMEM; 2642b64ec075STomohiro Misono 2643b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2644b64ec075STomohiro Misono if (!subvol_info) { 2645b64ec075STomohiro Misono btrfs_free_path(path); 2646b64ec075STomohiro Misono return -ENOMEM; 2647b64ec075STomohiro Misono } 2648b64ec075STomohiro Misono 2649b64ec075STomohiro Misono inode = file_inode(file); 2650b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2651b64ec075STomohiro Misono 2652b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2653b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2654b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2655b64ec075STomohiro Misono key.offset = (u64)-1; 2656b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2657b64ec075STomohiro Misono if (IS_ERR(root)) { 2658b64ec075STomohiro Misono ret = PTR_ERR(root); 2659b64ec075STomohiro Misono goto out; 2660b64ec075STomohiro Misono } 2661b64ec075STomohiro Misono root_item = &root->root_item; 2662b64ec075STomohiro Misono 2663b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2664b64ec075STomohiro Misono 2665b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2666b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2667b64ec075STomohiro Misono 2668b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2669b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2670b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2671b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2672b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2673b64ec075STomohiro Misono 2674b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2675b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2676b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2677b64ec075STomohiro Misono 2678b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2679b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2680b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2681b64ec075STomohiro Misono 2682b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2683b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2684b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2685b64ec075STomohiro Misono 2686b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2687b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2688b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2689b64ec075STomohiro Misono 2690b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2691b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2692b64ec075STomohiro Misono root = fs_info->tree_root; 2693b64ec075STomohiro Misono 2694b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2695b64ec075STomohiro Misono key.offset = 0; 2696b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2697b64ec075STomohiro Misono if (ret < 0) { 2698b64ec075STomohiro Misono goto out; 2699b64ec075STomohiro Misono } else if (path->slots[0] >= 2700b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2701b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2702b64ec075STomohiro Misono if (ret < 0) { 2703b64ec075STomohiro Misono goto out; 2704b64ec075STomohiro Misono } else if (ret > 0) { 2705b64ec075STomohiro Misono ret = -EUCLEAN; 2706b64ec075STomohiro Misono goto out; 2707b64ec075STomohiro Misono } 2708b64ec075STomohiro Misono } 2709b64ec075STomohiro Misono 2710b64ec075STomohiro Misono leaf = path->nodes[0]; 2711b64ec075STomohiro Misono slot = path->slots[0]; 2712b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2713b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2714b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2715b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2716b64ec075STomohiro Misono 2717b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2718b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2719b64ec075STomohiro Misono 2720b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2721b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2722b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2723b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2724b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2725b64ec075STomohiro Misono item_off, item_len); 2726b64ec075STomohiro Misono } else { 2727b64ec075STomohiro Misono ret = -ENOENT; 2728b64ec075STomohiro Misono goto out; 2729b64ec075STomohiro Misono } 2730b64ec075STomohiro Misono } 2731b64ec075STomohiro Misono 2732b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2733b64ec075STomohiro Misono ret = -EFAULT; 2734b64ec075STomohiro Misono 2735b64ec075STomohiro Misono out: 2736b64ec075STomohiro Misono btrfs_free_path(path); 2737b64ec075STomohiro Misono kzfree(subvol_info); 2738b64ec075STomohiro Misono return ret; 2739b64ec075STomohiro Misono } 2740b64ec075STomohiro Misono 274142e4b520STomohiro Misono /* 274242e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 274342e4b520STomohiro Misono * except the subvolume name. 274442e4b520STomohiro Misono */ 274542e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 274642e4b520STomohiro Misono { 274742e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 274842e4b520STomohiro Misono struct btrfs_root_ref *rref; 274942e4b520STomohiro Misono struct btrfs_root *root; 275042e4b520STomohiro Misono struct btrfs_path *path; 275142e4b520STomohiro Misono struct btrfs_key key; 275242e4b520STomohiro Misono struct extent_buffer *leaf; 275342e4b520STomohiro Misono struct inode *inode; 275442e4b520STomohiro Misono u64 objectid; 275542e4b520STomohiro Misono int slot; 275642e4b520STomohiro Misono int ret; 275742e4b520STomohiro Misono u8 found; 275842e4b520STomohiro Misono 275942e4b520STomohiro Misono path = btrfs_alloc_path(); 276042e4b520STomohiro Misono if (!path) 276142e4b520STomohiro Misono return -ENOMEM; 276242e4b520STomohiro Misono 276342e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 276442e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 276542e4b520STomohiro Misono btrfs_free_path(path); 276642e4b520STomohiro Misono return PTR_ERR(rootrefs); 276742e4b520STomohiro Misono } 276842e4b520STomohiro Misono 276942e4b520STomohiro Misono inode = file_inode(file); 277042e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 277142e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 277242e4b520STomohiro Misono 277342e4b520STomohiro Misono key.objectid = objectid; 277442e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 277542e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 277642e4b520STomohiro Misono found = 0; 277742e4b520STomohiro Misono 277842e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 277942e4b520STomohiro Misono if (ret < 0) { 278042e4b520STomohiro Misono goto out; 278142e4b520STomohiro Misono } else if (path->slots[0] >= 278242e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 278342e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 278442e4b520STomohiro Misono if (ret < 0) { 278542e4b520STomohiro Misono goto out; 278642e4b520STomohiro Misono } else if (ret > 0) { 278742e4b520STomohiro Misono ret = -EUCLEAN; 278842e4b520STomohiro Misono goto out; 278942e4b520STomohiro Misono } 279042e4b520STomohiro Misono } 279142e4b520STomohiro Misono while (1) { 279242e4b520STomohiro Misono leaf = path->nodes[0]; 279342e4b520STomohiro Misono slot = path->slots[0]; 279442e4b520STomohiro Misono 279542e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 279642e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 279742e4b520STomohiro Misono ret = 0; 279842e4b520STomohiro Misono goto out; 279942e4b520STomohiro Misono } 280042e4b520STomohiro Misono 280142e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 280242e4b520STomohiro Misono ret = -EOVERFLOW; 280342e4b520STomohiro Misono goto out; 280442e4b520STomohiro Misono } 280542e4b520STomohiro Misono 280642e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 280742e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 280842e4b520STomohiro Misono rootrefs->rootref[found].dirid = 280942e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 281042e4b520STomohiro Misono found++; 281142e4b520STomohiro Misono 281242e4b520STomohiro Misono ret = btrfs_next_item(root, path); 281342e4b520STomohiro Misono if (ret < 0) { 281442e4b520STomohiro Misono goto out; 281542e4b520STomohiro Misono } else if (ret > 0) { 281642e4b520STomohiro Misono ret = -EUCLEAN; 281742e4b520STomohiro Misono goto out; 281842e4b520STomohiro Misono } 281942e4b520STomohiro Misono } 282042e4b520STomohiro Misono 282142e4b520STomohiro Misono out: 282242e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 282342e4b520STomohiro Misono rootrefs->num_items = found; 282442e4b520STomohiro Misono /* update min_treeid for next search */ 282542e4b520STomohiro Misono if (found) 282642e4b520STomohiro Misono rootrefs->min_treeid = 282742e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 282842e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 282942e4b520STomohiro Misono ret = -EFAULT; 283042e4b520STomohiro Misono } 283142e4b520STomohiro Misono 283242e4b520STomohiro Misono kfree(rootrefs); 283342e4b520STomohiro Misono btrfs_free_path(path); 283442e4b520STomohiro Misono 283542e4b520STomohiro Misono return ret; 283642e4b520STomohiro Misono } 283742e4b520STomohiro Misono 283876dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 283976dda93cSYan, Zheng void __user *arg) 284076dda93cSYan, Zheng { 284154563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28420b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 284376dda93cSYan, Zheng struct dentry *dentry; 28442b0143b5SDavid Howells struct inode *dir = d_inode(parent); 284576dda93cSYan, Zheng struct inode *inode; 284676dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 284776dda93cSYan, Zheng struct btrfs_root *dest = NULL; 284876dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 284976dda93cSYan, Zheng int namelen; 285076dda93cSYan, Zheng int err = 0; 285176dda93cSYan, Zheng 2852325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2853325c50e3SJeff Mahoney return -ENOTDIR; 2854325c50e3SJeff Mahoney 285576dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 285676dda93cSYan, Zheng if (IS_ERR(vol_args)) 285776dda93cSYan, Zheng return PTR_ERR(vol_args); 285876dda93cSYan, Zheng 285976dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 286076dda93cSYan, Zheng namelen = strlen(vol_args->name); 286176dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 286276dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 286376dda93cSYan, Zheng err = -EINVAL; 286476dda93cSYan, Zheng goto out; 286576dda93cSYan, Zheng } 286676dda93cSYan, Zheng 2867a561be71SAl Viro err = mnt_want_write_file(file); 286876dda93cSYan, Zheng if (err) 286976dda93cSYan, Zheng goto out; 287076dda93cSYan, Zheng 2871521e0546SDavid Sterba 287200235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 287300235411SAl Viro if (err == -EINTR) 287400235411SAl Viro goto out_drop_write; 287576dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 287676dda93cSYan, Zheng if (IS_ERR(dentry)) { 287776dda93cSYan, Zheng err = PTR_ERR(dentry); 287876dda93cSYan, Zheng goto out_unlock_dir; 287976dda93cSYan, Zheng } 288076dda93cSYan, Zheng 28812b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 288276dda93cSYan, Zheng err = -ENOENT; 288376dda93cSYan, Zheng goto out_dput; 288476dda93cSYan, Zheng } 288576dda93cSYan, Zheng 28862b0143b5SDavid Howells inode = d_inode(dentry); 28874260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28884260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28894260f7c7SSage Weil /* 28904260f7c7SSage Weil * Regular user. Only allow this with a special mount 28914260f7c7SSage Weil * option, when the user has write+exec access to the 28924260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28934260f7c7SSage Weil * allowed. 28944260f7c7SSage Weil * 28954260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28964260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28974260f7c7SSage Weil * otherwise be able to delete. 28984260f7c7SSage Weil * 28994260f7c7SSage Weil * Users who want to delete empty subvols should try 29004260f7c7SSage Weil * rmdir(2). 29014260f7c7SSage Weil */ 29024260f7c7SSage Weil err = -EPERM; 29030b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 29044260f7c7SSage Weil goto out_dput; 29054260f7c7SSage Weil 29064260f7c7SSage Weil /* 29074260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 29084260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 29094260f7c7SSage Weil * must be called on the dentry referencing the root 29104260f7c7SSage Weil * of the subvol, not a random directory contained 29114260f7c7SSage Weil * within it. 29124260f7c7SSage Weil */ 29134260f7c7SSage Weil err = -EINVAL; 29144260f7c7SSage Weil if (root == dest) 29154260f7c7SSage Weil goto out_dput; 29164260f7c7SSage Weil 29174260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 29184260f7c7SSage Weil if (err) 29194260f7c7SSage Weil goto out_dput; 29205c39da5bSMiao Xie } 29214260f7c7SSage Weil 29225c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 29234260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 29244260f7c7SSage Weil if (err) 29254260f7c7SSage Weil goto out_dput; 29264260f7c7SSage Weil 29274a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 292876dda93cSYan, Zheng err = -EINVAL; 292976dda93cSYan, Zheng goto out_dput; 293076dda93cSYan, Zheng } 293176dda93cSYan, Zheng 29325955102cSAl Viro inode_lock(inode); 2933f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 29345955102cSAl Viro inode_unlock(inode); 293546008d9dSAmir Goldstein if (!err) { 293646008d9dSAmir Goldstein fsnotify_rmdir(dir, dentry); 293776dda93cSYan, Zheng d_delete(dentry); 293846008d9dSAmir Goldstein } 2939fa6ac876SLiu Bo 294076dda93cSYan, Zheng out_dput: 294176dda93cSYan, Zheng dput(dentry); 294276dda93cSYan, Zheng out_unlock_dir: 29435955102cSAl Viro inode_unlock(dir); 294400235411SAl Viro out_drop_write: 29452a79f17eSAl Viro mnt_drop_write_file(file); 294676dda93cSYan, Zheng out: 294776dda93cSYan, Zheng kfree(vol_args); 294876dda93cSYan, Zheng return err; 294976dda93cSYan, Zheng } 295076dda93cSYan, Zheng 29511e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2952f46b5a66SChristoph Hellwig { 2953496ad9aaSAl Viro struct inode *inode = file_inode(file); 2954f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29551e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2956c146afadSYan Zheng int ret; 2957c146afadSYan Zheng 295825122d15SIlya Dryomov ret = mnt_want_write_file(file); 295925122d15SIlya Dryomov if (ret) 296025122d15SIlya Dryomov return ret; 2961b83cc969SLi Zefan 296225122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 296325122d15SIlya Dryomov ret = -EROFS; 296425122d15SIlya Dryomov goto out; 29655ac00addSStefan Behrens } 2966f46b5a66SChristoph Hellwig 2967f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2968f46b5a66SChristoph Hellwig case S_IFDIR: 2969e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2970e441d54dSChris Mason ret = -EPERM; 2971e441d54dSChris Mason goto out; 2972e441d54dSChris Mason } 2973de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2974f46b5a66SChristoph Hellwig break; 2975f46b5a66SChristoph Hellwig case S_IFREG: 2976616d374eSAdam Borowski /* 2977616d374eSAdam Borowski * Note that this does not check the file descriptor for write 2978616d374eSAdam Borowski * access. This prevents defragmenting executables that are 2979616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 2980616d374eSAdam Borowski */ 2981616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 2982616d374eSAdam Borowski inode_permission(inode, MAY_WRITE)) { 2983616d374eSAdam Borowski ret = -EPERM; 2984e441d54dSChris Mason goto out; 2985e441d54dSChris Mason } 29861e701a32SChris Mason 29871e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29881e701a32SChris Mason if (!range) { 29891e701a32SChris Mason ret = -ENOMEM; 29901e701a32SChris Mason goto out; 29911e701a32SChris Mason } 29921e701a32SChris Mason 29931e701a32SChris Mason if (argp) { 29941e701a32SChris Mason if (copy_from_user(range, argp, 29951e701a32SChris Mason sizeof(*range))) { 29961e701a32SChris Mason ret = -EFAULT; 29971e701a32SChris Mason kfree(range); 2998683be16eSDan Carpenter goto out; 29991e701a32SChris Mason } 30001e701a32SChris Mason /* compression requires us to start the IO */ 30011e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 30021e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 30031e701a32SChris Mason range->extent_thresh = (u32)-1; 30041e701a32SChris Mason } 30051e701a32SChris Mason } else { 30061e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 30071e701a32SChris Mason range->len = (u64)-1; 30081e701a32SChris Mason } 3009496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 30107c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 30114cb5300bSChris Mason if (ret > 0) 30124cb5300bSChris Mason ret = 0; 30131e701a32SChris Mason kfree(range); 3014f46b5a66SChristoph Hellwig break; 30158929ecfaSYan, Zheng default: 30168929ecfaSYan, Zheng ret = -EINVAL; 3017f46b5a66SChristoph Hellwig } 3018e441d54dSChris Mason out: 301925122d15SIlya Dryomov mnt_drop_write_file(file); 3020e441d54dSChris Mason return ret; 3021f46b5a66SChristoph Hellwig } 3022f46b5a66SChristoph Hellwig 30232ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3024f46b5a66SChristoph Hellwig { 3025f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3026f46b5a66SChristoph Hellwig int ret; 3027f46b5a66SChristoph Hellwig 3028e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3029e441d54dSChris Mason return -EPERM; 3030e441d54dSChris Mason 3031171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 3032e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3033c9e9f97bSIlya Dryomov 3034dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3035c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3036c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3037c9e9f97bSIlya Dryomov goto out; 3038c9e9f97bSIlya Dryomov } 3039f46b5a66SChristoph Hellwig 30405516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30412ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3042f46b5a66SChristoph Hellwig 304343d20761SAnand Jain if (!ret) 30440b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 304543d20761SAnand Jain 3046f46b5a66SChristoph Hellwig kfree(vol_args); 3047c9e9f97bSIlya Dryomov out: 3048171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3049f46b5a66SChristoph Hellwig return ret; 3050f46b5a66SChristoph Hellwig } 3051f46b5a66SChristoph Hellwig 30526b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3053f46b5a66SChristoph Hellwig { 30540b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30550b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30566b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3057f46b5a66SChristoph Hellwig int ret; 3058f46b5a66SChristoph Hellwig 3059e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3060e441d54dSChris Mason return -EPERM; 3061e441d54dSChris Mason 3062da24927bSMiao Xie ret = mnt_want_write_file(file); 3063da24927bSMiao Xie if (ret) 3064da24927bSMiao Xie return ret; 3065c146afadSYan Zheng 3066dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3067c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3068c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3069c47ca32dSDan Carpenter goto err_drop; 3070c9e9f97bSIlya Dryomov } 3071f46b5a66SChristoph Hellwig 30726b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3073fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3074fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3075fd4e994bSOmar Sandoval goto out; 3076fd4e994bSOmar Sandoval } 3077f46b5a66SChristoph Hellwig 3078171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3079183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3080183860f6SAnand Jain goto out; 3081183860f6SAnand Jain } 3082183860f6SAnand Jain 3083735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30842ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30856b526ed7SAnand Jain } else { 30866b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30872ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30886b526ed7SAnand Jain } 3089171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3090183860f6SAnand Jain 30916b526ed7SAnand Jain if (!ret) { 3092735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30930b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30946b526ed7SAnand Jain vol_args->devid); 30956b526ed7SAnand Jain else 30960b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30976b526ed7SAnand Jain vol_args->name); 30986b526ed7SAnand Jain } 3099183860f6SAnand Jain out: 3100183860f6SAnand Jain kfree(vol_args); 3101c47ca32dSDan Carpenter err_drop: 31024ac20c70SIlya Dryomov mnt_drop_write_file(file); 3103f46b5a66SChristoph Hellwig return ret; 3104f46b5a66SChristoph Hellwig } 3105f46b5a66SChristoph Hellwig 3106f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3107f46b5a66SChristoph Hellwig { 31080b246afaSJeff Mahoney struct inode *inode = file_inode(file); 31090b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3110f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3111f46b5a66SChristoph Hellwig int ret; 3112f46b5a66SChristoph Hellwig 3113f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3114f46b5a66SChristoph Hellwig return -EPERM; 3115f46b5a66SChristoph Hellwig 3116f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3117f46b5a66SChristoph Hellwig if (ret) 3118f46b5a66SChristoph Hellwig return ret; 3119f46b5a66SChristoph Hellwig 3120171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3121f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 312258d7bbf8SDavid Sterba goto out_drop_write; 312358d7bbf8SDavid Sterba } 312458d7bbf8SDavid Sterba 312558d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 312658d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 312758d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3128f46b5a66SChristoph Hellwig goto out; 3129f46b5a66SChristoph Hellwig } 3130f46b5a66SChristoph Hellwig 313158d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31322ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3133f46b5a66SChristoph Hellwig 3134f46b5a66SChristoph Hellwig if (!ret) 31350b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3136f46b5a66SChristoph Hellwig kfree(vol_args); 313758d7bbf8SDavid Sterba out: 3138171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 313958d7bbf8SDavid Sterba out_drop_write: 3140f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 314158d7bbf8SDavid Sterba 3142f46b5a66SChristoph Hellwig return ret; 3143f46b5a66SChristoph Hellwig } 3144f46b5a66SChristoph Hellwig 31452ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 31462ff7e61eSJeff Mahoney void __user *arg) 3147475f6387SJan Schmidt { 3148027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3149475f6387SJan Schmidt struct btrfs_device *device; 31500b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3151027ed2f0SLi Zefan int ret = 0; 3152475f6387SJan Schmidt 3153027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3154027ed2f0SLi Zefan if (!fi_args) 3155027ed2f0SLi Zefan return -ENOMEM; 3156027ed2f0SLi Zefan 3157d03262c7SDavid Sterba rcu_read_lock(); 3158027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3159475f6387SJan Schmidt 3160d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3161027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3162027ed2f0SLi Zefan fi_args->max_id = device->devid; 3163475f6387SJan Schmidt } 3164d03262c7SDavid Sterba rcu_read_unlock(); 3165475f6387SJan Schmidt 3166de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3167bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3168bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3169bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 317080a773fbSDavid Sterba 3171027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3172027ed2f0SLi Zefan ret = -EFAULT; 3173475f6387SJan Schmidt 3174027ed2f0SLi Zefan kfree(fi_args); 3175027ed2f0SLi Zefan return ret; 3176475f6387SJan Schmidt } 3177475f6387SJan Schmidt 31782ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31792ff7e61eSJeff Mahoney void __user *arg) 3180475f6387SJan Schmidt { 3181475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3182475f6387SJan Schmidt struct btrfs_device *dev; 3183475f6387SJan Schmidt int ret = 0; 3184475f6387SJan Schmidt char *s_uuid = NULL; 3185475f6387SJan Schmidt 3186475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3187475f6387SJan Schmidt if (IS_ERR(di_args)) 3188475f6387SJan Schmidt return PTR_ERR(di_args); 3189475f6387SJan Schmidt 3190dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3191475f6387SJan Schmidt s_uuid = di_args->uuid; 3192475f6387SJan Schmidt 3193c5593ca3SDavid Sterba rcu_read_lock(); 3194e4319cd9SAnand Jain dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid, 319509ba3bc9SAnand Jain NULL, true); 3196475f6387SJan Schmidt 3197475f6387SJan Schmidt if (!dev) { 3198475f6387SJan Schmidt ret = -ENODEV; 3199475f6387SJan Schmidt goto out; 3200475f6387SJan Schmidt } 3201475f6387SJan Schmidt 3202475f6387SJan Schmidt di_args->devid = dev->devid; 32037cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 32047cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3205475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3206a27202fbSJim Meyering if (dev->name) { 3207672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3208672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3209a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3210a27202fbSJim Meyering } else { 321199ba55adSStefan Behrens di_args->path[0] = '\0'; 3212a27202fbSJim Meyering } 3213475f6387SJan Schmidt 3214475f6387SJan Schmidt out: 3215c5593ca3SDavid Sterba rcu_read_unlock(); 3216475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3217475f6387SJan Schmidt ret = -EFAULT; 3218475f6387SJan Schmidt 3219475f6387SJan Schmidt kfree(di_args); 3220475f6387SJan Schmidt return ret; 3221475f6387SJan Schmidt } 3222475f6387SJan Schmidt 3223d8b55242SFilipe Manana static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3224d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3225d8b55242SFilipe Manana { 3226d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3227d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3228d8b55242SFilipe Manana } 3229d8b55242SFilipe Manana 3230d8b55242SFilipe Manana static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3231d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3232d8b55242SFilipe Manana { 3233d8b55242SFilipe Manana if (inode1 < inode2) { 3234d8b55242SFilipe Manana swap(inode1, inode2); 3235d8b55242SFilipe Manana swap(loff1, loff2); 3236d8b55242SFilipe Manana } else if (inode1 == inode2 && loff2 < loff1) { 3237d8b55242SFilipe Manana swap(loff1, loff2); 3238d8b55242SFilipe Manana } 3239d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3240d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3241d8b55242SFilipe Manana } 3242d8b55242SFilipe Manana 324357a50e25SFilipe Manana static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len, 324434a28e3dSFilipe Manana struct inode *dst, u64 dst_loff) 3245416161dbSMark Fasheh { 3246416161dbSMark Fasheh int ret; 32470efa9f48SMark Fasheh 3248de02b9f6SFilipe Manana /* 3249d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3250d8b55242SFilipe Manana * source range to serialize with relocation. 3251de02b9f6SFilipe Manana */ 3252d8b55242SFilipe Manana btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 325357a50e25SFilipe Manana ret = btrfs_clone(src, dst, loff, len, len, dst_loff, 1); 3254d8b55242SFilipe Manana btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 32553973909dSTimofey Titovets 32563973909dSTimofey Titovets return ret; 32573973909dSTimofey Titovets } 32583973909dSTimofey Titovets 3259b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3260b6728768STimofey Titovets 32613973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 32623973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 32633973909dSTimofey Titovets { 32643973909dSTimofey Titovets int ret; 3265b6728768STimofey Titovets u64 i, tail_len, chunk_count; 326662d54f3aSFilipe Manana struct btrfs_root *root_dst = BTRFS_I(dst)->root; 326762d54f3aSFilipe Manana 326862d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 326962d54f3aSFilipe Manana if (root_dst->send_in_progress) { 327062d54f3aSFilipe Manana btrfs_warn_rl(root_dst->fs_info, 327162d54f3aSFilipe Manana "cannot deduplicate to root %llu while send operations are using it (%d in progress)", 327262d54f3aSFilipe Manana root_dst->root_key.objectid, 327362d54f3aSFilipe Manana root_dst->send_in_progress); 327462d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 327562d54f3aSFilipe Manana return -EAGAIN; 327662d54f3aSFilipe Manana } 327762d54f3aSFilipe Manana root_dst->dedupe_in_progress++; 327862d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 32793973909dSTimofey Titovets 3280b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3281b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 328267b07bd4STimofey Titovets 3283b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3284b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 328534a28e3dSFilipe Manana dst, dst_loff); 3286b6728768STimofey Titovets if (ret) 328762d54f3aSFilipe Manana goto out; 3288b6728768STimofey Titovets 3289b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3290b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3291b6728768STimofey Titovets } 3292b6728768STimofey Titovets 3293b6728768STimofey Titovets if (tail_len > 0) 329467b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 329534a28e3dSFilipe Manana dst_loff); 329662d54f3aSFilipe Manana out: 329762d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 329862d54f3aSFilipe Manana root_dst->dedupe_in_progress--; 329962d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 3300416161dbSMark Fasheh 3301416161dbSMark Fasheh return ret; 3302416161dbSMark Fasheh } 3303416161dbSMark Fasheh 3304f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3305f82a9901SFilipe Manana struct inode *inode, 3306f82a9901SFilipe Manana u64 endoff, 3307f82a9901SFilipe Manana const u64 destoff, 33081c919a5eSMark Fasheh const u64 olen, 33091c919a5eSMark Fasheh int no_time_update) 3310f82a9901SFilipe Manana { 3311f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3312f82a9901SFilipe Manana int ret; 3313f82a9901SFilipe Manana 3314f82a9901SFilipe Manana inode_inc_iversion(inode); 33151c919a5eSMark Fasheh if (!no_time_update) 3316c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3317f82a9901SFilipe Manana /* 3318f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3319f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3320f82a9901SFilipe Manana */ 3321f82a9901SFilipe Manana if (endoff > destoff + olen) 3322f82a9901SFilipe Manana endoff = destoff + olen; 3323f82a9901SFilipe Manana if (endoff > inode->i_size) 33246ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3325f82a9901SFilipe Manana 3326f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3327f82a9901SFilipe Manana if (ret) { 332866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 33293a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3330f82a9901SFilipe Manana goto out; 3331f82a9901SFilipe Manana } 33323a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3333f82a9901SFilipe Manana out: 3334f82a9901SFilipe Manana return ret; 3335f82a9901SFilipe Manana } 3336f82a9901SFilipe Manana 33378039d87dSFilipe Manana /* 33388039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33398039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33408039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33418039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33428039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33438039d87dSFilipe Manana * 33448039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33458039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33468039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33478039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33488039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33498039d87dSFilipe Manana * i_mutex, which we are holding here). 33508039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33518039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33528039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33538039d87dSFilipe Manana * 33548039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33558039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33568039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33578039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33588039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33598039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33608039d87dSFilipe Manana * the former. 33618039d87dSFilipe Manana */ 33624a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 33638039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33648039d87dSFilipe Manana struct btrfs_path *path, 33658039d87dSFilipe Manana struct btrfs_key *new_key, 33668039d87dSFilipe Manana const u64 drop_start, 33678039d87dSFilipe Manana const u64 datal, 33688039d87dSFilipe Manana const u64 skip, 33698039d87dSFilipe Manana const u64 size, 33708039d87dSFilipe Manana char *inline_data) 33718039d87dSFilipe Manana { 33720b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 33738039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 33748039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 33750b246afaSJeff Mahoney fs_info->sectorsize); 33768039d87dSFilipe Manana int ret; 33778039d87dSFilipe Manana struct btrfs_key key; 33788039d87dSFilipe Manana 33798039d87dSFilipe Manana if (new_key->offset > 0) 33808039d87dSFilipe Manana return -EOPNOTSUPP; 33818039d87dSFilipe Manana 33824a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 33838039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 33848039d87dSFilipe Manana key.offset = 0; 33858039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 33868039d87dSFilipe Manana if (ret < 0) { 33878039d87dSFilipe Manana return ret; 33888039d87dSFilipe Manana } else if (ret > 0) { 33898039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 33908039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 33918039d87dSFilipe Manana if (ret < 0) 33928039d87dSFilipe Manana return ret; 33938039d87dSFilipe Manana else if (ret > 0) 33948039d87dSFilipe Manana goto copy_inline_extent; 33958039d87dSFilipe Manana } 33968039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 33974a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 33988039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 33998039d87dSFilipe Manana ASSERT(key.offset > 0); 34008039d87dSFilipe Manana return -EOPNOTSUPP; 34018039d87dSFilipe Manana } 34028039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34038039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34048039d87dSFilipe Manana u64 ext_len; 34058039d87dSFilipe Manana 34068039d87dSFilipe Manana /* 34078039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34088039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34098039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34108039d87dSFilipe Manana */ 34118039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34128039d87dSFilipe Manana struct btrfs_file_extent_item); 34138039d87dSFilipe Manana /* 34148039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34158039d87dSFilipe Manana * following it. 34168039d87dSFilipe Manana */ 34178039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34188039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34198039d87dSFilipe Manana goto copy_inline_extent; 34208039d87dSFilipe Manana 34218039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34228039d87dSFilipe Manana if (ext_len > aligned_end) 34238039d87dSFilipe Manana return -EOPNOTSUPP; 34248039d87dSFilipe Manana 34258039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34268039d87dSFilipe Manana if (ret < 0) { 34278039d87dSFilipe Manana return ret; 34288039d87dSFilipe Manana } else if (ret == 0) { 34298039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34308039d87dSFilipe Manana path->slots[0]); 34314a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34328039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34338039d87dSFilipe Manana return -EOPNOTSUPP; 34348039d87dSFilipe Manana } 34358039d87dSFilipe Manana } 34368039d87dSFilipe Manana 34378039d87dSFilipe Manana copy_inline_extent: 34388039d87dSFilipe Manana /* 34398039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34408039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34418039d87dSFilipe Manana */ 34428039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34438039d87dSFilipe Manana /* 34448039d87dSFilipe Manana * If the destination inode has an inline extent... 34458039d87dSFilipe Manana * This would require copying the data from the source inline 34468039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34478039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34488039d87dSFilipe Manana * or just one of them, which would require decompressing and 34498039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34508039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34518039d87dSFilipe Manana * inline extent). 34528039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34538039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34548039d87dSFilipe Manana */ 34558039d87dSFilipe Manana return -EOPNOTSUPP; 34568039d87dSFilipe Manana } 34578039d87dSFilipe Manana 34588039d87dSFilipe Manana btrfs_release_path(path); 34598039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34608039d87dSFilipe Manana if (ret) 34618039d87dSFilipe Manana return ret; 34628039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34638039d87dSFilipe Manana if (ret) 34648039d87dSFilipe Manana return ret; 34658039d87dSFilipe Manana 34668039d87dSFilipe Manana if (skip) { 34678039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 34688039d87dSFilipe Manana 34698039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 34708039d87dSFilipe Manana } 34718039d87dSFilipe Manana 34728039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 34738039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 34748039d87dSFilipe Manana path->slots[0]), 34758039d87dSFilipe Manana size); 34768039d87dSFilipe Manana inode_add_bytes(dst, datal); 3477690a5dbfSFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(dst)->runtime_flags); 34788039d87dSFilipe Manana 34798039d87dSFilipe Manana return 0; 34808039d87dSFilipe Manana } 34818039d87dSFilipe Manana 348232b7c687SMark Fasheh /** 348332b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 348432b7c687SMark Fasheh * 348532b7c687SMark Fasheh * @src: Inode to clone from 348632b7c687SMark Fasheh * @inode: Inode to clone to 348732b7c687SMark Fasheh * @off: Offset within source to start clone from 348832b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 34891c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 349032b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 34911c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 349232b7c687SMark Fasheh */ 349332b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3494f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 34951c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3496f46b5a66SChristoph Hellwig { 34970b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3498f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 349932b7c687SMark Fasheh struct btrfs_path *path = NULL; 3500f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 350132b7c687SMark Fasheh struct btrfs_trans_handle *trans; 350232b7c687SMark Fasheh char *buf = NULL; 3503ae01a0abSYan Zheng struct btrfs_key key; 3504f46b5a66SChristoph Hellwig u32 nritems; 3505f46b5a66SChristoph Hellwig int slot; 3506ae01a0abSYan Zheng int ret; 3507f82a9901SFilipe Manana const u64 len = olen_aligned; 3508f82a9901SFilipe Manana u64 last_dest_end = destoff; 3509ae01a0abSYan Zheng 3510ae01a0abSYan Zheng ret = -ENOMEM; 3511752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3512ae01a0abSYan Zheng if (!buf) 351332b7c687SMark Fasheh return ret; 3514ae01a0abSYan Zheng 3515ae01a0abSYan Zheng path = btrfs_alloc_path(); 3516ae01a0abSYan Zheng if (!path) { 351715351955SDavid Sterba kvfree(buf); 351832b7c687SMark Fasheh return ret; 3519ae01a0abSYan Zheng } 352032b7c687SMark Fasheh 3521e4058b54SDavid Sterba path->reada = READA_FORWARD; 3522c5c9cd4dSSage Weil /* clone data */ 35234a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3524ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35252c463823SFilipe Manana key.offset = off; 3526f46b5a66SChristoph Hellwig 3527f46b5a66SChristoph Hellwig while (1) { 3528de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3529b64119b5SFilipe Manana struct btrfs_file_extent_item *extent; 3530b64119b5SFilipe Manana int type; 3531b64119b5SFilipe Manana u32 size; 3532b64119b5SFilipe Manana struct btrfs_key new_key; 3533b64119b5SFilipe Manana u64 disko = 0, diskl = 0; 3534b64119b5SFilipe Manana u64 datao = 0, datal = 0; 3535b64119b5SFilipe Manana u8 comp; 3536b64119b5SFilipe Manana u64 drop_start; 3537df858e76SFilipe Manana 3538f46b5a66SChristoph Hellwig /* 3539f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3540f46b5a66SChristoph Hellwig * tree. 3541f46b5a66SChristoph Hellwig */ 3542e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3543362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3544362a20c5SDavid Sterba 0, 0); 3545f46b5a66SChristoph Hellwig if (ret < 0) 3546f46b5a66SChristoph Hellwig goto out; 35472c463823SFilipe Manana /* 35482c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35492c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35502c463823SFilipe Manana * it might overlap our target range, therefore process it. 35512c463823SFilipe Manana */ 35522c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35532c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35542c463823SFilipe Manana path->slots[0] - 1); 35552c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35562c463823SFilipe Manana path->slots[0]--; 35572c463823SFilipe Manana } 3558f46b5a66SChristoph Hellwig 3559ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3560e4355f34SFilipe David Borba Manana process_slot: 3561ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3562362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3563f46b5a66SChristoph Hellwig if (ret < 0) 3564f46b5a66SChristoph Hellwig goto out; 3565f46b5a66SChristoph Hellwig if (ret > 0) 3566f46b5a66SChristoph Hellwig break; 3567ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3568f46b5a66SChristoph Hellwig } 3569f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3570f46b5a66SChristoph Hellwig slot = path->slots[0]; 3571f46b5a66SChristoph Hellwig 3572ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3573962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 35744a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3575f46b5a66SChristoph Hellwig break; 3576f46b5a66SChristoph Hellwig 3577b64119b5SFilipe Manana ASSERT(key.type == BTRFS_EXTENT_DATA_KEY); 357831840ae1SZheng Yan 3579c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3580c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3581c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3582c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3583c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3584c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3585b64119b5SFilipe Manana disko = btrfs_file_extent_disk_bytenr(leaf, extent); 3586b64119b5SFilipe Manana diskl = btrfs_file_extent_disk_num_bytes(leaf, extent); 3587c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3588b64119b5SFilipe Manana datal = btrfs_file_extent_num_bytes(leaf, extent); 3589c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3590b64119b5SFilipe Manana /* Take upper bound, may be compressed */ 3591b64119b5SFilipe Manana datal = btrfs_file_extent_ram_bytes(leaf, extent); 3592c5c9cd4dSSage Weil } 359331840ae1SZheng Yan 35942c463823SFilipe Manana /* 3595b64119b5SFilipe Manana * The first search might have left us at an extent item that 3596b64119b5SFilipe Manana * ends before our target range's start, can happen if we have 3597b64119b5SFilipe Manana * holes and NO_HOLES feature enabled. 35982c463823SFilipe Manana */ 35992c463823SFilipe Manana if (key.offset + datal <= off) { 3600e4355f34SFilipe David Borba Manana path->slots[0]++; 3601e4355f34SFilipe David Borba Manana goto process_slot; 36022c463823SFilipe Manana } else if (key.offset >= off + len) { 36032c463823SFilipe Manana break; 3604e4355f34SFilipe David Borba Manana } 3605df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3606e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3607b64119b5SFilipe Manana read_extent_buffer(leaf, buf, btrfs_item_ptr_offset(leaf, slot), 3608e4355f34SFilipe David Borba Manana size); 3609e4355f34SFilipe David Borba Manana 3610e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3611e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3612c5c9cd4dSSage Weil 361331840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36144a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36154d728ec7SLi Zefan if (off <= key.offset) 3616c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36174d728ec7SLi Zefan else 36184d728ec7SLi Zefan new_key.offset = destoff; 3619c5c9cd4dSSage Weil 3620b6f3409bSSage Weil /* 3621b64119b5SFilipe Manana * Deal with a hole that doesn't have an extent item that 3622b64119b5SFilipe Manana * represents it (NO_HOLES feature enabled). 3623b64119b5SFilipe Manana * This hole is either in the middle of the cloning range or at 3624b64119b5SFilipe Manana * the beginning (fully overlaps it or partially overlaps it). 3625f82a9901SFilipe Manana */ 3626f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3627f82a9901SFilipe Manana drop_start = last_dest_end; 3628f82a9901SFilipe Manana else 3629f82a9901SFilipe Manana drop_start = new_key.offset; 3630f82a9901SFilipe Manana 3631c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3632c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3633690a5dbfSFilipe Manana struct btrfs_clone_extent_info clone_info; 3634690a5dbfSFilipe Manana 3635d72c0842SLi Zefan /* 3636d72c0842SLi Zefan * a | --- range to clone ---| b 3637d72c0842SLi Zefan * | ------------- extent ------------- | 3638d72c0842SLi Zefan */ 3639d72c0842SLi Zefan 3640b64119b5SFilipe Manana /* Subtract range b */ 3641d72c0842SLi Zefan if (key.offset + datal > off + len) 3642d72c0842SLi Zefan datal = off + len - key.offset; 3643d72c0842SLi Zefan 3644b64119b5SFilipe Manana /* Subtract range a */ 3645a22285a6SYan, Zheng if (off > key.offset) { 3646a22285a6SYan, Zheng datao += off - key.offset; 3647a22285a6SYan, Zheng datal -= off - key.offset; 3648a22285a6SYan, Zheng } 3649a22285a6SYan, Zheng 3650690a5dbfSFilipe Manana clone_info.disk_offset = disko; 3651690a5dbfSFilipe Manana clone_info.disk_len = diskl; 3652690a5dbfSFilipe Manana clone_info.data_offset = datao; 3653690a5dbfSFilipe Manana clone_info.data_len = datal; 3654690a5dbfSFilipe Manana clone_info.file_offset = new_key.offset; 3655690a5dbfSFilipe Manana clone_info.extent_buf = buf; 3656690a5dbfSFilipe Manana clone_info.item_size = size; 3657690a5dbfSFilipe Manana ret = btrfs_punch_hole_range(inode, path, 3658f82a9901SFilipe Manana drop_start, 3659690a5dbfSFilipe Manana new_key.offset + datal - 1, 3660690a5dbfSFilipe Manana &clone_info, &trans); 3661690a5dbfSFilipe Manana if (ret) 366279787eaaSJeff Mahoney goto out; 3663c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3664c5c9cd4dSSage Weil u64 skip = 0; 3665c5c9cd4dSSage Weil u64 trim = 0; 3666ed958762SFilipe Manana 3667c5c9cd4dSSage Weil if (off > key.offset) { 3668c5c9cd4dSSage Weil skip = off - key.offset; 3669c5c9cd4dSSage Weil new_key.offset += skip; 367031840ae1SZheng Yan } 3671d397712bSChris Mason 3672c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3673c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3674d397712bSChris Mason 3675c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3676c5c9cd4dSSage Weil ret = -EINVAL; 3677c5c9cd4dSSage Weil goto out; 367831840ae1SZheng Yan } 3679c5c9cd4dSSage Weil size -= skip + trim; 3680c5c9cd4dSSage Weil datal -= skip + trim; 3681a22285a6SYan, Zheng 3682690a5dbfSFilipe Manana /* 3683b64119b5SFilipe Manana * If our extent is inline, we know we will drop or 3684b64119b5SFilipe Manana * adjust at most 1 extent item in the destination root. 3685690a5dbfSFilipe Manana * 3686b64119b5SFilipe Manana * 1 - adjusting old extent (we may have to split it) 3687690a5dbfSFilipe Manana * 1 - add new extent 3688690a5dbfSFilipe Manana * 1 - inode update 3689690a5dbfSFilipe Manana */ 3690690a5dbfSFilipe Manana trans = btrfs_start_transaction(root, 3); 3691690a5dbfSFilipe Manana if (IS_ERR(trans)) { 3692690a5dbfSFilipe Manana ret = PTR_ERR(trans); 3693690a5dbfSFilipe Manana goto out; 3694690a5dbfSFilipe Manana } 3695690a5dbfSFilipe Manana 3696b64119b5SFilipe Manana ret = clone_copy_inline_extent(inode, trans, path, 3697b64119b5SFilipe Manana &new_key, drop_start, 3698b64119b5SFilipe Manana datal, skip, size, buf); 369979787eaaSJeff Mahoney if (ret) { 37003f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 3701b64119b5SFilipe Manana btrfs_abort_transaction(trans, ret); 37023a45bb20SJeff Mahoney btrfs_end_transaction(trans); 370379787eaaSJeff Mahoney goto out; 370479787eaaSJeff Mahoney } 3705c5c9cd4dSSage Weil } 3706c5c9cd4dSSage Weil 3707b3b4aa74SDavid Sterba btrfs_release_path(path); 3708c5c9cd4dSSage Weil 370962e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 37100b246afaSJeff Mahoney fs_info->sectorsize); 3711b64119b5SFilipe Manana ret = clone_finish_inode_update(trans, inode, last_dest_end, 3712b64119b5SFilipe Manana destoff, olen, no_time_update); 3713f82a9901SFilipe Manana if (ret) 371479787eaaSJeff Mahoney goto out; 37152c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 37162c463823SFilipe Manana break; 3717b64119b5SFilipe Manana 3718b3b4aa74SDavid Sterba btrfs_release_path(path); 3719df858e76SFilipe Manana key.offset = next_key_min_offset; 372069ae5e44SWang Xiaoguang 372169ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 372269ae5e44SWang Xiaoguang ret = -EINTR; 372369ae5e44SWang Xiaoguang goto out; 372469ae5e44SWang Xiaoguang } 3725ae01a0abSYan Zheng } 3726f46b5a66SChristoph Hellwig ret = 0; 372732b7c687SMark Fasheh 3728f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3729f82a9901SFilipe Manana /* 3730147271e3SFilipe Manana * We have an implicit hole that fully or partially overlaps our 3731147271e3SFilipe Manana * cloning range at its end. This means that we either have the 3732147271e3SFilipe Manana * NO_HOLES feature enabled or the implicit hole happened due to 3733147271e3SFilipe Manana * mixing buffered and direct IO writes against this file. 3734f82a9901SFilipe Manana */ 3735b3b4aa74SDavid Sterba btrfs_release_path(path); 3736690a5dbfSFilipe Manana path->leave_spinning = 0; 3737f82a9901SFilipe Manana 3738690a5dbfSFilipe Manana ret = btrfs_punch_hole_range(inode, path, 3739690a5dbfSFilipe Manana last_dest_end, destoff + len - 1, 3740147271e3SFilipe Manana NULL, &trans); 3741690a5dbfSFilipe Manana if (ret) 3742f82a9901SFilipe Manana goto out; 3743690a5dbfSFilipe Manana 3744f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 37451c919a5eSMark Fasheh destoff, olen, no_time_update); 3746f82a9901SFilipe Manana } 3747f82a9901SFilipe Manana 3748f46b5a66SChristoph Hellwig out: 374932b7c687SMark Fasheh btrfs_free_path(path); 375015351955SDavid Sterba kvfree(buf); 375132b7c687SMark Fasheh return ret; 375232b7c687SMark Fasheh } 375332b7c687SMark Fasheh 37543db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 375532b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 375632b7c687SMark Fasheh { 375754563d41SAl Viro struct inode *inode = file_inode(file); 37583db11b2eSZach Brown struct inode *src = file_inode(file_src); 37590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 376032b7c687SMark Fasheh int ret; 376132b7c687SMark Fasheh u64 len = olen; 37620b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 376332b7c687SMark Fasheh 376432b7c687SMark Fasheh /* 376532b7c687SMark Fasheh * TODO: 376632b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 376732b7c687SMark Fasheh * decompress into destination's address_space (the file offset 376832b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 376932b7c687SMark Fasheh * otherwise reinsert) a subrange. 377000fdf13aSLiu Bo * 377100fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 377200fdf13aSLiu Bo * be either compressed or non-compressed. 377332b7c687SMark Fasheh */ 377432b7c687SMark Fasheh 3775ac765f83SFilipe Manana /* 377634a28e3dSFilipe Manana * VFS's generic_remap_file_range_prep() protects us from cloning the 377734a28e3dSFilipe Manana * eof block into the middle of a file, which would result in corruption 377834a28e3dSFilipe Manana * if the file size is not blocksize aligned. So we don't need to check 377934a28e3dSFilipe Manana * for that case here. 3780ac765f83SFilipe Manana */ 378134a28e3dSFilipe Manana if (off + len == src->i_size) 378232b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 378332b7c687SMark Fasheh 378432b7c687SMark Fasheh if (destoff > inode->i_size) { 3785f7fa1107SFilipe Manana const u64 wb_start = ALIGN_DOWN(inode->i_size, bs); 3786f7fa1107SFilipe Manana 378732b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 378832b7c687SMark Fasheh if (ret) 378934a28e3dSFilipe Manana return ret; 3790f7fa1107SFilipe Manana /* 3791f7fa1107SFilipe Manana * We may have truncated the last block if the inode's size is 3792f7fa1107SFilipe Manana * not sector size aligned, so we need to wait for writeback to 3793f7fa1107SFilipe Manana * complete before proceeding further, otherwise we can race 3794f7fa1107SFilipe Manana * with cloning and attempt to increment a reference to an 3795f7fa1107SFilipe Manana * extent that no longer exists (writeback completed right after 3796f7fa1107SFilipe Manana * we found the previous extent covering eof and before we 3797f7fa1107SFilipe Manana * attempted to increment its reference count). 3798f7fa1107SFilipe Manana */ 3799f7fa1107SFilipe Manana ret = btrfs_wait_ordered_range(inode, wb_start, 3800f7fa1107SFilipe Manana destoff - wb_start); 3801f7fa1107SFilipe Manana if (ret) 3802f7fa1107SFilipe Manana return ret; 380332b7c687SMark Fasheh } 380432b7c687SMark Fasheh 3805c125b8bfSFilipe Manana /* 3806d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3807d8b55242SFilipe Manana * source range to serialize with relocation. 3808c125b8bfSFilipe Manana */ 3809d8b55242SFilipe Manana btrfs_double_extent_lock(src, off, inode, destoff, len); 38101c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 3811d8b55242SFilipe Manana btrfs_double_extent_unlock(src, off, inode, destoff, len); 3812c125b8bfSFilipe Manana /* 3813c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3814c125b8bfSFilipe Manana * data immediately and not the previous data. 3815c125b8bfSFilipe Manana */ 381665bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 381709cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 381809cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 381934a28e3dSFilipe Manana 382034a28e3dSFilipe Manana return ret; 382134a28e3dSFilipe Manana } 382234a28e3dSFilipe Manana 382334a28e3dSFilipe Manana static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in, 382434a28e3dSFilipe Manana struct file *file_out, loff_t pos_out, 382534a28e3dSFilipe Manana loff_t *len, unsigned int remap_flags) 382634a28e3dSFilipe Manana { 382734a28e3dSFilipe Manana struct inode *inode_in = file_inode(file_in); 382834a28e3dSFilipe Manana struct inode *inode_out = file_inode(file_out); 382934a28e3dSFilipe Manana u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize; 383034a28e3dSFilipe Manana bool same_inode = inode_out == inode_in; 383134a28e3dSFilipe Manana u64 wb_len; 383234a28e3dSFilipe Manana int ret; 383334a28e3dSFilipe Manana 383434a28e3dSFilipe Manana if (!(remap_flags & REMAP_FILE_DEDUP)) { 383534a28e3dSFilipe Manana struct btrfs_root *root_out = BTRFS_I(inode_out)->root; 383634a28e3dSFilipe Manana 383734a28e3dSFilipe Manana if (btrfs_root_readonly(root_out)) 383834a28e3dSFilipe Manana return -EROFS; 383934a28e3dSFilipe Manana 384034a28e3dSFilipe Manana if (file_in->f_path.mnt != file_out->f_path.mnt || 384134a28e3dSFilipe Manana inode_in->i_sb != inode_out->i_sb) 384234a28e3dSFilipe Manana return -EXDEV; 384334a28e3dSFilipe Manana } 384434a28e3dSFilipe Manana 3845500710d3SFilipe Manana /* don't make the dst file partly checksummed */ 3846500710d3SFilipe Manana if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) != 3847500710d3SFilipe Manana (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) { 38487984ae52SGoldwyn Rodrigues return -EINVAL; 3849500710d3SFilipe Manana } 3850500710d3SFilipe Manana 385134a28e3dSFilipe Manana /* 385234a28e3dSFilipe Manana * Now that the inodes are locked, we need to start writeback ourselves 385334a28e3dSFilipe Manana * and can not rely on the writeback from the VFS's generic helper 385434a28e3dSFilipe Manana * generic_remap_file_range_prep() because: 385534a28e3dSFilipe Manana * 385634a28e3dSFilipe Manana * 1) For compression we must call filemap_fdatawrite_range() range 385734a28e3dSFilipe Manana * twice (btrfs_fdatawrite_range() does it for us), and the generic 385834a28e3dSFilipe Manana * helper only calls it once; 385934a28e3dSFilipe Manana * 386034a28e3dSFilipe Manana * 2) filemap_fdatawrite_range(), called by the generic helper only 386134a28e3dSFilipe Manana * waits for the writeback to complete, i.e. for IO to be done, and 386234a28e3dSFilipe Manana * not for the ordered extents to complete. We need to wait for them 386334a28e3dSFilipe Manana * to complete so that new file extent items are in the fs tree. 386434a28e3dSFilipe Manana */ 386534a28e3dSFilipe Manana if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP)) 386634a28e3dSFilipe Manana wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs); 386734a28e3dSFilipe Manana else 386834a28e3dSFilipe Manana wb_len = ALIGN(*len, bs); 386934a28e3dSFilipe Manana 387034a28e3dSFilipe Manana /* 387134a28e3dSFilipe Manana * Since we don't lock ranges, wait for ongoing lockless dio writes (as 387234a28e3dSFilipe Manana * any in progress could create its ordered extents after we wait for 387334a28e3dSFilipe Manana * existing ordered extents below). 387434a28e3dSFilipe Manana */ 387534a28e3dSFilipe Manana inode_dio_wait(inode_in); 387634a28e3dSFilipe Manana if (!same_inode) 387734a28e3dSFilipe Manana inode_dio_wait(inode_out); 387834a28e3dSFilipe Manana 3879a94d1d0cSQu Wenruo /* 3880a94d1d0cSQu Wenruo * Workaround to make sure NOCOW buffered write reach disk as NOCOW. 3881a94d1d0cSQu Wenruo * 3882a94d1d0cSQu Wenruo * Btrfs' back references do not have a block level granularity, they 3883a94d1d0cSQu Wenruo * work at the whole extent level. 3884a94d1d0cSQu Wenruo * NOCOW buffered write without data space reserved may not be able 3885a94d1d0cSQu Wenruo * to fall back to CoW due to lack of data space, thus could cause 3886a94d1d0cSQu Wenruo * data loss. 3887a94d1d0cSQu Wenruo * 3888a94d1d0cSQu Wenruo * Here we take a shortcut by flushing the whole inode, so that all 3889a94d1d0cSQu Wenruo * nocow write should reach disk as nocow before we increase the 3890a94d1d0cSQu Wenruo * reference of the extent. We could do better by only flushing NOCOW 3891a94d1d0cSQu Wenruo * data, but that needs extra accounting. 3892a94d1d0cSQu Wenruo * 3893a94d1d0cSQu Wenruo * Also we don't need to check ASYNC_EXTENT, as async extent will be 3894a94d1d0cSQu Wenruo * CoWed anyway, not affecting nocow part. 3895a94d1d0cSQu Wenruo */ 3896a94d1d0cSQu Wenruo ret = filemap_flush(inode_in->i_mapping); 3897a94d1d0cSQu Wenruo if (ret < 0) 3898a94d1d0cSQu Wenruo return ret; 3899a94d1d0cSQu Wenruo 390034a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs), 390134a28e3dSFilipe Manana wb_len); 390234a28e3dSFilipe Manana if (ret < 0) 39037984ae52SGoldwyn Rodrigues return ret; 390434a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs), 390534a28e3dSFilipe Manana wb_len); 390634a28e3dSFilipe Manana if (ret < 0) 39073db11b2eSZach Brown return ret; 39087984ae52SGoldwyn Rodrigues 39097984ae52SGoldwyn Rodrigues return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out, 39107984ae52SGoldwyn Rodrigues len, remap_flags); 39113db11b2eSZach Brown } 39123db11b2eSZach Brown 391342ec3d4cSDarrick J. Wong loff_t btrfs_remap_file_range(struct file *src_file, loff_t off, 391442ec3d4cSDarrick J. Wong struct file *dst_file, loff_t destoff, loff_t len, 39152e5dfc99SDarrick J. Wong unsigned int remap_flags) 39163db11b2eSZach Brown { 391734a28e3dSFilipe Manana struct inode *src_inode = file_inode(src_file); 391834a28e3dSFilipe Manana struct inode *dst_inode = file_inode(dst_file); 391934a28e3dSFilipe Manana bool same_inode = dst_inode == src_inode; 392042ec3d4cSDarrick J. Wong int ret; 392142ec3d4cSDarrick J. Wong 39222e5dfc99SDarrick J. Wong if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY)) 39232e5dfc99SDarrick J. Wong return -EINVAL; 39242e5dfc99SDarrick J. Wong 39257984ae52SGoldwyn Rodrigues if (same_inode) 39267984ae52SGoldwyn Rodrigues inode_lock(src_inode); 39277984ae52SGoldwyn Rodrigues else 39287984ae52SGoldwyn Rodrigues lock_two_nondirectories(src_inode, dst_inode); 39297984ae52SGoldwyn Rodrigues 393034a28e3dSFilipe Manana ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff, 393134a28e3dSFilipe Manana &len, remap_flags); 393234a28e3dSFilipe Manana if (ret < 0 || len == 0) 39337984ae52SGoldwyn Rodrigues goto out_unlock; 39342e5dfc99SDarrick J. Wong 393534a28e3dSFilipe Manana if (remap_flags & REMAP_FILE_DEDUP) 393634a28e3dSFilipe Manana ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff); 393734a28e3dSFilipe Manana else 393842ec3d4cSDarrick J. Wong ret = btrfs_clone_files(dst_file, src_file, off, len, destoff); 393934a28e3dSFilipe Manana 39407984ae52SGoldwyn Rodrigues out_unlock: 394134a28e3dSFilipe Manana if (same_inode) 394234a28e3dSFilipe Manana inode_unlock(src_inode); 394334a28e3dSFilipe Manana else 39444ea748e1SFilipe Manana unlock_two_nondirectories(src_inode, dst_inode); 394534a28e3dSFilipe Manana 394642ec3d4cSDarrick J. Wong return ret < 0 ? ret : len; 3947c5c9cd4dSSage Weil } 3948c5c9cd4dSSage Weil 39496ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 39506ef5ed0dSJosef Bacik { 3951496ad9aaSAl Viro struct inode *inode = file_inode(file); 39520b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 39536ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 39546ef5ed0dSJosef Bacik struct btrfs_root *new_root; 39556ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 39566ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 39576ef5ed0dSJosef Bacik struct btrfs_path *path; 39586ef5ed0dSJosef Bacik struct btrfs_key location; 39596ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 39606ef5ed0dSJosef Bacik u64 objectid = 0; 39616ef5ed0dSJosef Bacik u64 dir_id; 39623c04ce01SMiao Xie int ret; 39636ef5ed0dSJosef Bacik 39646ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 39656ef5ed0dSJosef Bacik return -EPERM; 39666ef5ed0dSJosef Bacik 39673c04ce01SMiao Xie ret = mnt_want_write_file(file); 39683c04ce01SMiao Xie if (ret) 39693c04ce01SMiao Xie return ret; 39703c04ce01SMiao Xie 39713c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 39723c04ce01SMiao Xie ret = -EFAULT; 39733c04ce01SMiao Xie goto out; 39743c04ce01SMiao Xie } 39756ef5ed0dSJosef Bacik 39766ef5ed0dSJosef Bacik if (!objectid) 39771cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 39786ef5ed0dSJosef Bacik 39796ef5ed0dSJosef Bacik location.objectid = objectid; 39806ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 39816ef5ed0dSJosef Bacik location.offset = (u64)-1; 39826ef5ed0dSJosef Bacik 39830b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 39843c04ce01SMiao Xie if (IS_ERR(new_root)) { 39853c04ce01SMiao Xie ret = PTR_ERR(new_root); 39863c04ce01SMiao Xie goto out; 39873c04ce01SMiao Xie } 39884fd786e6SMisono Tomohiro if (!is_fstree(new_root->root_key.objectid)) { 39896d6d2829Ssatoru takeuchi ret = -ENOENT; 39906d6d2829Ssatoru takeuchi goto out; 39916d6d2829Ssatoru takeuchi } 39926ef5ed0dSJosef Bacik 39936ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 39943c04ce01SMiao Xie if (!path) { 39953c04ce01SMiao Xie ret = -ENOMEM; 39963c04ce01SMiao Xie goto out; 39973c04ce01SMiao Xie } 39986ef5ed0dSJosef Bacik path->leave_spinning = 1; 39996ef5ed0dSJosef Bacik 40006ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 400198d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40026ef5ed0dSJosef Bacik btrfs_free_path(path); 40033c04ce01SMiao Xie ret = PTR_ERR(trans); 40043c04ce01SMiao Xie goto out; 40056ef5ed0dSJosef Bacik } 40066ef5ed0dSJosef Bacik 40070b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 40080b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 40096ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4010cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40116ef5ed0dSJosef Bacik btrfs_free_path(path); 40123a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40130b246afaSJeff Mahoney btrfs_err(fs_info, 40145d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 40153c04ce01SMiao Xie ret = -ENOENT; 40163c04ce01SMiao Xie goto out; 40176ef5ed0dSJosef Bacik } 40186ef5ed0dSJosef Bacik 40196ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 40206ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 40216ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 40226ef5ed0dSJosef Bacik btrfs_free_path(path); 40236ef5ed0dSJosef Bacik 40240b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 40253a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40263c04ce01SMiao Xie out: 40273c04ce01SMiao Xie mnt_drop_write_file(file); 40283c04ce01SMiao Xie return ret; 40296ef5ed0dSJosef Bacik } 40306ef5ed0dSJosef Bacik 4031c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4032bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4033bf5fc093SJosef Bacik { 403432da5386SDavid Sterba struct btrfs_block_group *block_group; 4035bf5fc093SJosef Bacik 4036bf5fc093SJosef Bacik space->total_bytes = 0; 4037bf5fc093SJosef Bacik space->used_bytes = 0; 4038bf5fc093SJosef Bacik space->flags = 0; 4039bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4040bf5fc093SJosef Bacik space->flags = block_group->flags; 4041b3470b5dSDavid Sterba space->total_bytes += block_group->length; 4042bf38be65SDavid Sterba space->used_bytes += block_group->used; 4043bf5fc093SJosef Bacik } 4044bf5fc093SJosef Bacik } 4045bf5fc093SJosef Bacik 40462ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 40472ff7e61eSJeff Mahoney void __user *arg) 40481406e432SJosef Bacik { 40491406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 40501406e432SJosef Bacik struct btrfs_ioctl_space_info space; 40511406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 40527fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 405313f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 40541406e432SJosef Bacik struct btrfs_space_info *info; 4055315d8e98SColin Ian King static const u64 types[] = { 4056315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4057bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4058bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4059315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4060315d8e98SColin Ian King }; 4061bf5fc093SJosef Bacik int num_types = 4; 40627fde62bfSChris Mason int alloc_size; 40631406e432SJosef Bacik int ret = 0; 406451788b1bSDan Rosenberg u64 slot_count = 0; 4065bf5fc093SJosef Bacik int i, c; 40661406e432SJosef Bacik 40671406e432SJosef Bacik if (copy_from_user(&space_args, 40681406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 40691406e432SJosef Bacik sizeof(space_args))) 40701406e432SJosef Bacik return -EFAULT; 40711406e432SJosef Bacik 4072bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4073bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4074bf5fc093SJosef Bacik 4075bf5fc093SJosef Bacik info = NULL; 40767fde62bfSChris Mason rcu_read_lock(); 40770b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4078bf5fc093SJosef Bacik list) { 4079bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4080bf5fc093SJosef Bacik info = tmp; 4081bf5fc093SJosef Bacik break; 4082bf5fc093SJosef Bacik } 4083bf5fc093SJosef Bacik } 40847fde62bfSChris Mason rcu_read_unlock(); 40851406e432SJosef Bacik 4086bf5fc093SJosef Bacik if (!info) 4087bf5fc093SJosef Bacik continue; 4088bf5fc093SJosef Bacik 4089bf5fc093SJosef Bacik down_read(&info->groups_sem); 4090bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4091bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4092bf5fc093SJosef Bacik slot_count++; 4093bf5fc093SJosef Bacik } 4094bf5fc093SJosef Bacik up_read(&info->groups_sem); 4095bf5fc093SJosef Bacik } 4096bf5fc093SJosef Bacik 409736523e95SDavid Sterba /* 409836523e95SDavid Sterba * Global block reserve, exported as a space_info 409936523e95SDavid Sterba */ 410036523e95SDavid Sterba slot_count++; 410136523e95SDavid Sterba 41027fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 41037fde62bfSChris Mason if (space_args.space_slots == 0) { 41047fde62bfSChris Mason space_args.total_spaces = slot_count; 41057fde62bfSChris Mason goto out; 41067fde62bfSChris Mason } 4107bf5fc093SJosef Bacik 410851788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4109bf5fc093SJosef Bacik 41107fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4111bf5fc093SJosef Bacik 41127fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41137fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41147fde62bfSChris Mason */ 411509cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 41167fde62bfSChris Mason return -ENOMEM; 41177fde62bfSChris Mason 41187fde62bfSChris Mason space_args.total_spaces = 0; 41198d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 41207fde62bfSChris Mason if (!dest) 41217fde62bfSChris Mason return -ENOMEM; 41227fde62bfSChris Mason dest_orig = dest; 41237fde62bfSChris Mason 41247fde62bfSChris Mason /* now we have a buffer to copy into */ 4125bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4126bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4127bf5fc093SJosef Bacik 412851788b1bSDan Rosenberg if (!slot_count) 412951788b1bSDan Rosenberg break; 413051788b1bSDan Rosenberg 4131bf5fc093SJosef Bacik info = NULL; 41321406e432SJosef Bacik rcu_read_lock(); 41330b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4134bf5fc093SJosef Bacik list) { 4135bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4136bf5fc093SJosef Bacik info = tmp; 41377fde62bfSChris Mason break; 4138bf5fc093SJosef Bacik } 4139bf5fc093SJosef Bacik } 4140bf5fc093SJosef Bacik rcu_read_unlock(); 41417fde62bfSChris Mason 4142bf5fc093SJosef Bacik if (!info) 4143bf5fc093SJosef Bacik continue; 4144bf5fc093SJosef Bacik down_read(&info->groups_sem); 4145bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4146bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4147c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4148c065f5b1SSu Yue &space); 41497fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 41501406e432SJosef Bacik dest++; 41511406e432SJosef Bacik space_args.total_spaces++; 415251788b1bSDan Rosenberg slot_count--; 41531406e432SJosef Bacik } 415451788b1bSDan Rosenberg if (!slot_count) 415551788b1bSDan Rosenberg break; 4156bf5fc093SJosef Bacik } 4157bf5fc093SJosef Bacik up_read(&info->groups_sem); 4158bf5fc093SJosef Bacik } 41591406e432SJosef Bacik 416036523e95SDavid Sterba /* 416136523e95SDavid Sterba * Add global block reserve 416236523e95SDavid Sterba */ 416336523e95SDavid Sterba if (slot_count) { 41640b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 416536523e95SDavid Sterba 416636523e95SDavid Sterba spin_lock(&block_rsv->lock); 416736523e95SDavid Sterba space.total_bytes = block_rsv->size; 416836523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 416936523e95SDavid Sterba spin_unlock(&block_rsv->lock); 417036523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 417136523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 417236523e95SDavid Sterba space_args.total_spaces++; 417336523e95SDavid Sterba } 417436523e95SDavid Sterba 41752eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 41767fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 41777fde62bfSChris Mason 41787fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 41797fde62bfSChris Mason ret = -EFAULT; 41807fde62bfSChris Mason 41817fde62bfSChris Mason kfree(dest_orig); 41827fde62bfSChris Mason out: 41837fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 41841406e432SJosef Bacik ret = -EFAULT; 41851406e432SJosef Bacik 41861406e432SJosef Bacik return ret; 41871406e432SJosef Bacik } 41881406e432SJosef Bacik 41899a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 41909a8c28beSMiao Xie void __user *argp) 419146204592SSage Weil { 419246204592SSage Weil struct btrfs_trans_handle *trans; 419346204592SSage Weil u64 transid; 4194db5b493aSTsutomu Itoh int ret; 419546204592SSage Weil 4196d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4197ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4198ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 419998d5dc13STsutomu Itoh return PTR_ERR(trans); 4200ff7c1d33SMiao Xie 4201ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4202ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4203ff7c1d33SMiao Xie goto out; 4204ff7c1d33SMiao Xie } 420546204592SSage Weil transid = trans->transid; 42063a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 42078b2b2d3cSTsutomu Itoh if (ret) { 42083a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4209db5b493aSTsutomu Itoh return ret; 42108b2b2d3cSTsutomu Itoh } 4211ff7c1d33SMiao Xie out: 421246204592SSage Weil if (argp) 421346204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 421446204592SSage Weil return -EFAULT; 421546204592SSage Weil return 0; 421646204592SSage Weil } 421746204592SSage Weil 42182ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 42199a8c28beSMiao Xie void __user *argp) 422046204592SSage Weil { 422146204592SSage Weil u64 transid; 422246204592SSage Weil 422346204592SSage Weil if (argp) { 422446204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 422546204592SSage Weil return -EFAULT; 422646204592SSage Weil } else { 422746204592SSage Weil transid = 0; /* current trans */ 422846204592SSage Weil } 42292ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 423046204592SSage Weil } 423146204592SSage Weil 4232b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4233475f6387SJan Schmidt { 42340b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4235475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4236b8e95489SMiao Xie int ret; 4237475f6387SJan Schmidt 4238475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4239475f6387SJan Schmidt return -EPERM; 4240475f6387SJan Schmidt 4241475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4242475f6387SJan Schmidt if (IS_ERR(sa)) 4243475f6387SJan Schmidt return PTR_ERR(sa); 4244475f6387SJan Schmidt 4245b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4246b8e95489SMiao Xie ret = mnt_want_write_file(file); 4247b8e95489SMiao Xie if (ret) 4248b8e95489SMiao Xie goto out; 4249b8e95489SMiao Xie } 4250b8e95489SMiao Xie 42510b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 425263a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 425363a212abSStefan Behrens 0); 4254475f6387SJan Schmidt 42555afe6ce7SFilipe Manana /* 42565afe6ce7SFilipe Manana * Copy scrub args to user space even if btrfs_scrub_dev() returned an 42575afe6ce7SFilipe Manana * error. This is important as it allows user space to know how much 42585afe6ce7SFilipe Manana * progress scrub has done. For example, if scrub is canceled we get 42595afe6ce7SFilipe Manana * -ECANCELED from btrfs_scrub_dev() and return that error back to user 42605afe6ce7SFilipe Manana * space. Later user space can inspect the progress from the structure 42615afe6ce7SFilipe Manana * btrfs_ioctl_scrub_args and resume scrub from where it left off 42625afe6ce7SFilipe Manana * previously (btrfs-progs does this). 42635afe6ce7SFilipe Manana * If we fail to copy the btrfs_ioctl_scrub_args structure to user space 42645afe6ce7SFilipe Manana * then return -EFAULT to signal the structure was not copied or it may 42655afe6ce7SFilipe Manana * be corrupt and unreliable due to a partial copy. 42665afe6ce7SFilipe Manana */ 42675afe6ce7SFilipe Manana if (copy_to_user(arg, sa, sizeof(*sa))) 4268475f6387SJan Schmidt ret = -EFAULT; 4269475f6387SJan Schmidt 4270b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4271b8e95489SMiao Xie mnt_drop_write_file(file); 4272b8e95489SMiao Xie out: 4273475f6387SJan Schmidt kfree(sa); 4274475f6387SJan Schmidt return ret; 4275475f6387SJan Schmidt } 4276475f6387SJan Schmidt 42772ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4278475f6387SJan Schmidt { 4279475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4280475f6387SJan Schmidt return -EPERM; 4281475f6387SJan Schmidt 42822ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4283475f6387SJan Schmidt } 4284475f6387SJan Schmidt 42852ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4286475f6387SJan Schmidt void __user *arg) 4287475f6387SJan Schmidt { 4288475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4289475f6387SJan Schmidt int ret; 4290475f6387SJan Schmidt 4291475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4292475f6387SJan Schmidt return -EPERM; 4293475f6387SJan Schmidt 4294475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4295475f6387SJan Schmidt if (IS_ERR(sa)) 4296475f6387SJan Schmidt return PTR_ERR(sa); 4297475f6387SJan Schmidt 42982ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4299475f6387SJan Schmidt 43004fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4301475f6387SJan Schmidt ret = -EFAULT; 4302475f6387SJan Schmidt 4303475f6387SJan Schmidt kfree(sa); 4304475f6387SJan Schmidt return ret; 4305475f6387SJan Schmidt } 4306475f6387SJan Schmidt 43072ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4308b27f7c0cSDavid Sterba void __user *arg) 4309c11d2c23SStefan Behrens { 4310c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4311c11d2c23SStefan Behrens int ret; 4312c11d2c23SStefan Behrens 4313c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4314c11d2c23SStefan Behrens if (IS_ERR(sa)) 4315c11d2c23SStefan Behrens return PTR_ERR(sa); 4316c11d2c23SStefan Behrens 4317b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4318b27f7c0cSDavid Sterba kfree(sa); 4319b27f7c0cSDavid Sterba return -EPERM; 4320b27f7c0cSDavid Sterba } 4321b27f7c0cSDavid Sterba 43222ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4323c11d2c23SStefan Behrens 4324eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4325c11d2c23SStefan Behrens ret = -EFAULT; 4326c11d2c23SStefan Behrens 4327c11d2c23SStefan Behrens kfree(sa); 4328c11d2c23SStefan Behrens return ret; 4329c11d2c23SStefan Behrens } 4330c11d2c23SStefan Behrens 43312ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 43322ff7e61eSJeff Mahoney void __user *arg) 43333f6bcfbdSStefan Behrens { 43343f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 43353f6bcfbdSStefan Behrens int ret; 43363f6bcfbdSStefan Behrens 43373f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 43383f6bcfbdSStefan Behrens return -EPERM; 43393f6bcfbdSStefan Behrens 43403f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 43413f6bcfbdSStefan Behrens if (IS_ERR(p)) 43423f6bcfbdSStefan Behrens return PTR_ERR(p); 43433f6bcfbdSStefan Behrens 43443f6bcfbdSStefan Behrens switch (p->cmd) { 43453f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4346bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4347adfa97cbSIlya Dryomov ret = -EROFS; 4348adfa97cbSIlya Dryomov goto out; 4349adfa97cbSIlya Dryomov } 4350171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4351e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 43523f6bcfbdSStefan Behrens } else { 43532ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4354171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 43553f6bcfbdSStefan Behrens } 43563f6bcfbdSStefan Behrens break; 43573f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 43580b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 43593f6bcfbdSStefan Behrens ret = 0; 43603f6bcfbdSStefan Behrens break; 43613f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 436217d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 436397282031SAnand Jain ret = 0; 43643f6bcfbdSStefan Behrens break; 43653f6bcfbdSStefan Behrens default: 43663f6bcfbdSStefan Behrens ret = -EINVAL; 43673f6bcfbdSStefan Behrens break; 43683f6bcfbdSStefan Behrens } 43693f6bcfbdSStefan Behrens 4370d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 43713f6bcfbdSStefan Behrens ret = -EFAULT; 4372adfa97cbSIlya Dryomov out: 43733f6bcfbdSStefan Behrens kfree(p); 43743f6bcfbdSStefan Behrens return ret; 43753f6bcfbdSStefan Behrens } 43763f6bcfbdSStefan Behrens 4377d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4378d7728c96SJan Schmidt { 4379d7728c96SJan Schmidt int ret = 0; 4380d7728c96SJan Schmidt int i; 4381740c3d22SChris Mason u64 rel_ptr; 4382d7728c96SJan Schmidt int size; 4383806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4384d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4385d7728c96SJan Schmidt struct btrfs_path *path; 4386d7728c96SJan Schmidt 438782b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4388d7728c96SJan Schmidt return -EPERM; 4389d7728c96SJan Schmidt 4390d7728c96SJan Schmidt path = btrfs_alloc_path(); 4391d7728c96SJan Schmidt if (!path) { 4392d7728c96SJan Schmidt ret = -ENOMEM; 4393d7728c96SJan Schmidt goto out; 4394d7728c96SJan Schmidt } 4395d7728c96SJan Schmidt 4396d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4397d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4398d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4399d7728c96SJan Schmidt ipa = NULL; 4400d7728c96SJan Schmidt goto out; 4401d7728c96SJan Schmidt } 4402d7728c96SJan Schmidt 4403d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4404d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4405d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4406d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4407d7728c96SJan Schmidt ipath = NULL; 4408d7728c96SJan Schmidt goto out; 4409d7728c96SJan Schmidt } 4410d7728c96SJan Schmidt 4411d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4412d7728c96SJan Schmidt if (ret < 0) 4413d7728c96SJan Schmidt goto out; 4414d7728c96SJan Schmidt 4415d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4416745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4417745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4418740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4419d7728c96SJan Schmidt } 4420d7728c96SJan Schmidt 4421718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4422718dc5faSOmar Sandoval ipath->fspath, size); 4423d7728c96SJan Schmidt if (ret) { 4424d7728c96SJan Schmidt ret = -EFAULT; 4425d7728c96SJan Schmidt goto out; 4426d7728c96SJan Schmidt } 4427d7728c96SJan Schmidt 4428d7728c96SJan Schmidt out: 4429d7728c96SJan Schmidt btrfs_free_path(path); 4430d7728c96SJan Schmidt free_ipath(ipath); 4431d7728c96SJan Schmidt kfree(ipa); 4432d7728c96SJan Schmidt 4433d7728c96SJan Schmidt return ret; 4434d7728c96SJan Schmidt } 4435d7728c96SJan Schmidt 4436d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4437d7728c96SJan Schmidt { 4438d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4439d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4440d7728c96SJan Schmidt 4441d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4442d7728c96SJan Schmidt inodes->bytes_left -= c; 4443d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4444d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4445d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4446d7728c96SJan Schmidt inodes->elem_cnt += 3; 4447d7728c96SJan Schmidt } else { 4448d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4449d7728c96SJan Schmidt inodes->bytes_left = 0; 4450d7728c96SJan Schmidt inodes->elem_missed += 3; 4451d7728c96SJan Schmidt } 4452d7728c96SJan Schmidt 4453d7728c96SJan Schmidt return 0; 4454d7728c96SJan Schmidt } 4455d7728c96SJan Schmidt 44562ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4457d24a67b2SZygo Blaxell void __user *arg, int version) 4458d7728c96SJan Schmidt { 4459d7728c96SJan Schmidt int ret = 0; 4460d7728c96SJan Schmidt int size; 4461d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4462d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4463d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4464d24a67b2SZygo Blaxell bool ignore_offset; 4465d7728c96SJan Schmidt 4466d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4467d7728c96SJan Schmidt return -EPERM; 4468d7728c96SJan Schmidt 4469d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 44707b9ea627SShailendra Verma if (IS_ERR(loi)) 44717b9ea627SShailendra Verma return PTR_ERR(loi); 4472d7728c96SJan Schmidt 4473d24a67b2SZygo Blaxell if (version == 1) { 4474d24a67b2SZygo Blaxell ignore_offset = false; 4475b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4476d24a67b2SZygo Blaxell } else { 4477d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4478d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4479d24a67b2SZygo Blaxell ret = -EINVAL; 4480d24a67b2SZygo Blaxell goto out_loi; 4481d24a67b2SZygo Blaxell } 4482d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4483d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4484d24a67b2SZygo Blaxell ret = -EINVAL; 4485d24a67b2SZygo Blaxell goto out_loi; 4486d24a67b2SZygo Blaxell } 4487d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4488b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4489d24a67b2SZygo Blaxell } 4490d24a67b2SZygo Blaxell 4491d7728c96SJan Schmidt path = btrfs_alloc_path(); 4492d7728c96SJan Schmidt if (!path) { 4493d7728c96SJan Schmidt ret = -ENOMEM; 4494d7728c96SJan Schmidt goto out; 4495d7728c96SJan Schmidt } 4496d7728c96SJan Schmidt 4497d7728c96SJan Schmidt inodes = init_data_container(size); 4498d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4499d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4500d7728c96SJan Schmidt inodes = NULL; 4501d7728c96SJan Schmidt goto out; 4502d7728c96SJan Schmidt } 4503d7728c96SJan Schmidt 45042ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4505d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4506df031f07SLiu Bo if (ret == -EINVAL) 4507d7728c96SJan Schmidt ret = -ENOENT; 4508d7728c96SJan Schmidt if (ret < 0) 4509d7728c96SJan Schmidt goto out; 4510d7728c96SJan Schmidt 4511718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4512718dc5faSOmar Sandoval size); 4513d7728c96SJan Schmidt if (ret) 4514d7728c96SJan Schmidt ret = -EFAULT; 4515d7728c96SJan Schmidt 4516d7728c96SJan Schmidt out: 4517d7728c96SJan Schmidt btrfs_free_path(path); 4518f54de068SDavid Sterba kvfree(inodes); 4519d24a67b2SZygo Blaxell out_loi: 4520d7728c96SJan Schmidt kfree(loi); 4521d7728c96SJan Schmidt 4522d7728c96SJan Schmidt return ret; 4523d7728c96SJan Schmidt } 4524d7728c96SJan Schmidt 4525008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4526c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4527c9e9f97bSIlya Dryomov { 4528c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4529c9e9f97bSIlya Dryomov 4530c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4531c9e9f97bSIlya Dryomov 45323009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4533837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4534837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4535837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4536a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4537a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4538837d5b6eSIlya Dryomov 4539c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4540c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4541c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 454219a39dceSIlya Dryomov 454319a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 454419a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 454519a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4546c9e9f97bSIlya Dryomov } 4547c9e9f97bSIlya Dryomov 45489ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4549c9e9f97bSIlya Dryomov { 4550496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4551c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4552c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4553c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4554ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4555c9e9f97bSIlya Dryomov int ret; 4556c9e9f97bSIlya Dryomov 4557c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4558c9e9f97bSIlya Dryomov return -EPERM; 4559c9e9f97bSIlya Dryomov 4560e54bfa31SLiu Bo ret = mnt_want_write_file(file); 45619ba1f6e4SLiu Bo if (ret) 45629ba1f6e4SLiu Bo return ret; 45639ba1f6e4SLiu Bo 4564ed0fb78fSIlya Dryomov again: 4565171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4566c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4567ed0fb78fSIlya Dryomov need_unlock = true; 4568ed0fb78fSIlya Dryomov goto locked; 4569ed0fb78fSIlya Dryomov } 4570ed0fb78fSIlya Dryomov 4571ed0fb78fSIlya Dryomov /* 457201327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4573ed0fb78fSIlya Dryomov * (1) some other op is running 4574ed0fb78fSIlya Dryomov * (2) balance is running 4575ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4576ed0fb78fSIlya Dryomov */ 4577ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4578ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4579ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 45803009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4581ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4582dccdb07bSDavid Sterba /* 4583dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4584dccdb07bSDavid Sterba * we'll reexamine the status again. 4585dccdb07bSDavid Sterba */ 4586ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4587ed0fb78fSIlya Dryomov 4588ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 45893009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4590ed0fb78fSIlya Dryomov /* this is (3) */ 4591ed0fb78fSIlya Dryomov need_unlock = false; 4592ed0fb78fSIlya Dryomov goto locked; 4593ed0fb78fSIlya Dryomov } 4594ed0fb78fSIlya Dryomov 4595ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4596ed0fb78fSIlya Dryomov goto again; 4597ed0fb78fSIlya Dryomov } else { 4598ed0fb78fSIlya Dryomov /* this is (2) */ 4599ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4600ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4601ed0fb78fSIlya Dryomov goto out; 4602ed0fb78fSIlya Dryomov } 4603ed0fb78fSIlya Dryomov } else { 4604ed0fb78fSIlya Dryomov /* this is (1) */ 4605ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4606e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4607ed0fb78fSIlya Dryomov goto out; 4608ed0fb78fSIlya Dryomov } 4609ed0fb78fSIlya Dryomov 4610ed0fb78fSIlya Dryomov locked: 4611171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4612c9e9f97bSIlya Dryomov 4613c9e9f97bSIlya Dryomov if (arg) { 4614c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4615c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4616c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4617ed0fb78fSIlya Dryomov goto out_unlock; 4618c9e9f97bSIlya Dryomov } 4619de322263SIlya Dryomov 4620de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4621de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4622de322263SIlya Dryomov ret = -ENOTCONN; 4623de322263SIlya Dryomov goto out_bargs; 4624de322263SIlya Dryomov } 4625de322263SIlya Dryomov 4626de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4627de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4628de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4629de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4630de322263SIlya Dryomov 4631de322263SIlya Dryomov goto do_balance; 4632de322263SIlya Dryomov } 4633c9e9f97bSIlya Dryomov } else { 4634c9e9f97bSIlya Dryomov bargs = NULL; 4635c9e9f97bSIlya Dryomov } 4636c9e9f97bSIlya Dryomov 4637ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4638837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4639837d5b6eSIlya Dryomov goto out_bargs; 4640837d5b6eSIlya Dryomov } 4641837d5b6eSIlya Dryomov 46428d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4643c9e9f97bSIlya Dryomov if (!bctl) { 4644c9e9f97bSIlya Dryomov ret = -ENOMEM; 4645c9e9f97bSIlya Dryomov goto out_bargs; 4646c9e9f97bSIlya Dryomov } 4647c9e9f97bSIlya Dryomov 4648c9e9f97bSIlya Dryomov if (arg) { 4649c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4650c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4651c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4652c9e9f97bSIlya Dryomov 4653c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4654f43ffb60SIlya Dryomov } else { 4655f43ffb60SIlya Dryomov /* balance everything - no filters */ 4656f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4657c9e9f97bSIlya Dryomov } 4658c9e9f97bSIlya Dryomov 46598eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 46608eb93459SDavid Sterba ret = -EINVAL; 46610f89abf5SChristian Engelmayer goto out_bctl; 46628eb93459SDavid Sterba } 46638eb93459SDavid Sterba 4664de322263SIlya Dryomov do_balance: 4665c9e9f97bSIlya Dryomov /* 4666149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4667149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4668149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4669149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4670c9e9f97bSIlya Dryomov */ 4671ed0fb78fSIlya Dryomov need_unlock = false; 4672ed0fb78fSIlya Dryomov 46736fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 46740f89abf5SChristian Engelmayer bctl = NULL; 4675ed0fb78fSIlya Dryomov 4676d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4677c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4678c9e9f97bSIlya Dryomov ret = -EFAULT; 4679c9e9f97bSIlya Dryomov } 4680c9e9f97bSIlya Dryomov 46810f89abf5SChristian Engelmayer out_bctl: 46820f89abf5SChristian Engelmayer kfree(bctl); 4683c9e9f97bSIlya Dryomov out_bargs: 4684c9e9f97bSIlya Dryomov kfree(bargs); 4685ed0fb78fSIlya Dryomov out_unlock: 4686c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4687ed0fb78fSIlya Dryomov if (need_unlock) 4688171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4689ed0fb78fSIlya Dryomov out: 4690e54bfa31SLiu Bo mnt_drop_write_file(file); 4691c9e9f97bSIlya Dryomov return ret; 4692c9e9f97bSIlya Dryomov } 4693c9e9f97bSIlya Dryomov 46942ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4695837d5b6eSIlya Dryomov { 4696837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4697837d5b6eSIlya Dryomov return -EPERM; 4698837d5b6eSIlya Dryomov 4699837d5b6eSIlya Dryomov switch (cmd) { 4700837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 47010b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4702a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47030b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4704837d5b6eSIlya Dryomov } 4705837d5b6eSIlya Dryomov 4706837d5b6eSIlya Dryomov return -EINVAL; 4707837d5b6eSIlya Dryomov } 4708837d5b6eSIlya Dryomov 47092ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 471019a39dceSIlya Dryomov void __user *arg) 471119a39dceSIlya Dryomov { 471219a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 471319a39dceSIlya Dryomov int ret = 0; 471419a39dceSIlya Dryomov 471519a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 471619a39dceSIlya Dryomov return -EPERM; 471719a39dceSIlya Dryomov 471819a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 471919a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 472019a39dceSIlya Dryomov ret = -ENOTCONN; 472119a39dceSIlya Dryomov goto out; 472219a39dceSIlya Dryomov } 472319a39dceSIlya Dryomov 47248d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 472519a39dceSIlya Dryomov if (!bargs) { 472619a39dceSIlya Dryomov ret = -ENOMEM; 472719a39dceSIlya Dryomov goto out; 472819a39dceSIlya Dryomov } 472919a39dceSIlya Dryomov 4730008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 473119a39dceSIlya Dryomov 473219a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 473319a39dceSIlya Dryomov ret = -EFAULT; 473419a39dceSIlya Dryomov 473519a39dceSIlya Dryomov kfree(bargs); 473619a39dceSIlya Dryomov out: 473719a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 473819a39dceSIlya Dryomov return ret; 473919a39dceSIlya Dryomov } 474019a39dceSIlya Dryomov 4741905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 47425d13a37bSArne Jansen { 47430b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47440b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47455d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 47465d13a37bSArne Jansen int ret; 47475d13a37bSArne Jansen 47485d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47495d13a37bSArne Jansen return -EPERM; 47505d13a37bSArne Jansen 4751905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4752905b0ddaSMiao Xie if (ret) 4753905b0ddaSMiao Xie return ret; 47545d13a37bSArne Jansen 47555d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4756905b0ddaSMiao Xie if (IS_ERR(sa)) { 4757905b0ddaSMiao Xie ret = PTR_ERR(sa); 4758905b0ddaSMiao Xie goto drop_write; 4759905b0ddaSMiao Xie } 47605d13a37bSArne Jansen 47610b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 47625d13a37bSArne Jansen 47635d13a37bSArne Jansen switch (sa->cmd) { 47645d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4765340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 47665d13a37bSArne Jansen break; 47675d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4768340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 47695d13a37bSArne Jansen break; 47705d13a37bSArne Jansen default: 47715d13a37bSArne Jansen ret = -EINVAL; 47725d13a37bSArne Jansen break; 47735d13a37bSArne Jansen } 47745d13a37bSArne Jansen 47755d13a37bSArne Jansen kfree(sa); 47760b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4777905b0ddaSMiao Xie drop_write: 4778905b0ddaSMiao Xie mnt_drop_write_file(file); 47795d13a37bSArne Jansen return ret; 47805d13a37bSArne Jansen } 47815d13a37bSArne Jansen 4782905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 47835d13a37bSArne Jansen { 47840b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47850b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47860b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 47875d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 47885d13a37bSArne Jansen struct btrfs_trans_handle *trans; 47895d13a37bSArne Jansen int ret; 47905d13a37bSArne Jansen int err; 47915d13a37bSArne Jansen 47925d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47935d13a37bSArne Jansen return -EPERM; 47945d13a37bSArne Jansen 4795905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4796905b0ddaSMiao Xie if (ret) 4797905b0ddaSMiao Xie return ret; 47985d13a37bSArne Jansen 47995d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4800905b0ddaSMiao Xie if (IS_ERR(sa)) { 4801905b0ddaSMiao Xie ret = PTR_ERR(sa); 4802905b0ddaSMiao Xie goto drop_write; 4803905b0ddaSMiao Xie } 48045d13a37bSArne Jansen 48055d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48065d13a37bSArne Jansen if (IS_ERR(trans)) { 48075d13a37bSArne Jansen ret = PTR_ERR(trans); 48085d13a37bSArne Jansen goto out; 48095d13a37bSArne Jansen } 48105d13a37bSArne Jansen 48115d13a37bSArne Jansen if (sa->assign) { 48129f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 48135d13a37bSArne Jansen } else { 481439616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 48155d13a37bSArne Jansen } 48165d13a37bSArne Jansen 4817e082f563SQu Wenruo /* update qgroup status and info */ 4818280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4819e082f563SQu Wenruo if (err < 0) 48200b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4821ad8403dfSAnand Jain "failed to update qgroup status and info"); 48223a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48235d13a37bSArne Jansen if (err && !ret) 48245d13a37bSArne Jansen ret = err; 48255d13a37bSArne Jansen 48265d13a37bSArne Jansen out: 48275d13a37bSArne Jansen kfree(sa); 4828905b0ddaSMiao Xie drop_write: 4829905b0ddaSMiao Xie mnt_drop_write_file(file); 48305d13a37bSArne Jansen return ret; 48315d13a37bSArne Jansen } 48325d13a37bSArne Jansen 4833905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 48345d13a37bSArne Jansen { 48350b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48360b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48375d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 48385d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48395d13a37bSArne Jansen int ret; 48405d13a37bSArne Jansen int err; 48415d13a37bSArne Jansen 48425d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48435d13a37bSArne Jansen return -EPERM; 48445d13a37bSArne Jansen 4845905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4846905b0ddaSMiao Xie if (ret) 4847905b0ddaSMiao Xie return ret; 48485d13a37bSArne Jansen 48495d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4850905b0ddaSMiao Xie if (IS_ERR(sa)) { 4851905b0ddaSMiao Xie ret = PTR_ERR(sa); 4852905b0ddaSMiao Xie goto drop_write; 4853905b0ddaSMiao Xie } 48545d13a37bSArne Jansen 4855d86e56cfSMiao Xie if (!sa->qgroupid) { 4856d86e56cfSMiao Xie ret = -EINVAL; 4857d86e56cfSMiao Xie goto out; 4858d86e56cfSMiao Xie } 4859d86e56cfSMiao Xie 48605d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48615d13a37bSArne Jansen if (IS_ERR(trans)) { 48625d13a37bSArne Jansen ret = PTR_ERR(trans); 48635d13a37bSArne Jansen goto out; 48645d13a37bSArne Jansen } 48655d13a37bSArne Jansen 48665d13a37bSArne Jansen if (sa->create) { 486749a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 48685d13a37bSArne Jansen } else { 48693efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 48705d13a37bSArne Jansen } 48715d13a37bSArne Jansen 48723a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48735d13a37bSArne Jansen if (err && !ret) 48745d13a37bSArne Jansen ret = err; 48755d13a37bSArne Jansen 48765d13a37bSArne Jansen out: 48775d13a37bSArne Jansen kfree(sa); 4878905b0ddaSMiao Xie drop_write: 4879905b0ddaSMiao Xie mnt_drop_write_file(file); 48805d13a37bSArne Jansen return ret; 48815d13a37bSArne Jansen } 48825d13a37bSArne Jansen 4883905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 48845d13a37bSArne Jansen { 48850b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48860b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48875d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 48885d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48895d13a37bSArne Jansen int ret; 48905d13a37bSArne Jansen int err; 48915d13a37bSArne Jansen u64 qgroupid; 48925d13a37bSArne Jansen 48935d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48945d13a37bSArne Jansen return -EPERM; 48955d13a37bSArne Jansen 4896905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4897905b0ddaSMiao Xie if (ret) 4898905b0ddaSMiao Xie return ret; 48995d13a37bSArne Jansen 49005d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4901905b0ddaSMiao Xie if (IS_ERR(sa)) { 4902905b0ddaSMiao Xie ret = PTR_ERR(sa); 4903905b0ddaSMiao Xie goto drop_write; 4904905b0ddaSMiao Xie } 49055d13a37bSArne Jansen 49065d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49075d13a37bSArne Jansen if (IS_ERR(trans)) { 49085d13a37bSArne Jansen ret = PTR_ERR(trans); 49095d13a37bSArne Jansen goto out; 49105d13a37bSArne Jansen } 49115d13a37bSArne Jansen 49125d13a37bSArne Jansen qgroupid = sa->qgroupid; 49135d13a37bSArne Jansen if (!qgroupid) { 49145d13a37bSArne Jansen /* take the current subvol as qgroup */ 49155d13a37bSArne Jansen qgroupid = root->root_key.objectid; 49165d13a37bSArne Jansen } 49175d13a37bSArne Jansen 4918f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 49195d13a37bSArne Jansen 49203a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49215d13a37bSArne Jansen if (err && !ret) 49225d13a37bSArne Jansen ret = err; 49235d13a37bSArne Jansen 49245d13a37bSArne Jansen out: 49255d13a37bSArne Jansen kfree(sa); 4926905b0ddaSMiao Xie drop_write: 4927905b0ddaSMiao Xie mnt_drop_write_file(file); 49285d13a37bSArne Jansen return ret; 49295d13a37bSArne Jansen } 49305d13a37bSArne Jansen 49312f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 49322f232036SJan Schmidt { 49330b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49340b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49352f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49362f232036SJan Schmidt int ret; 49372f232036SJan Schmidt 49382f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49392f232036SJan Schmidt return -EPERM; 49402f232036SJan Schmidt 49412f232036SJan Schmidt ret = mnt_want_write_file(file); 49422f232036SJan Schmidt if (ret) 49432f232036SJan Schmidt return ret; 49442f232036SJan Schmidt 49452f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 49462f232036SJan Schmidt if (IS_ERR(qsa)) { 49472f232036SJan Schmidt ret = PTR_ERR(qsa); 49482f232036SJan Schmidt goto drop_write; 49492f232036SJan Schmidt } 49502f232036SJan Schmidt 49512f232036SJan Schmidt if (qsa->flags) { 49522f232036SJan Schmidt ret = -EINVAL; 49532f232036SJan Schmidt goto out; 49542f232036SJan Schmidt } 49552f232036SJan Schmidt 49560b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 49572f232036SJan Schmidt 49582f232036SJan Schmidt out: 49592f232036SJan Schmidt kfree(qsa); 49602f232036SJan Schmidt drop_write: 49612f232036SJan Schmidt mnt_drop_write_file(file); 49622f232036SJan Schmidt return ret; 49632f232036SJan Schmidt } 49642f232036SJan Schmidt 4965b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, 4966b929c1d8SMarcos Paulo de Souza void __user *arg) 49672f232036SJan Schmidt { 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 4990b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info, 4991b929c1d8SMarcos Paulo de Souza void __user *arg) 499257254b6eSJan Schmidt { 499357254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 499457254b6eSJan Schmidt return -EPERM; 499557254b6eSJan Schmidt 49960b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 499757254b6eSJan Schmidt } 499857254b6eSJan Schmidt 4999abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5000abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50018ea05e3aSAlexander Block { 5002496ad9aaSAl Viro struct inode *inode = file_inode(file); 50030b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50048ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50058ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50068ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 500795582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 50088ea05e3aSAlexander Block int ret = 0; 5009dd5f9615SStefan Behrens int received_uuid_changed; 50108ea05e3aSAlexander Block 5011bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5012bd60ea0fSDavid Sterba return -EPERM; 5013bd60ea0fSDavid Sterba 50148ea05e3aSAlexander Block ret = mnt_want_write_file(file); 50158ea05e3aSAlexander Block if (ret < 0) 50168ea05e3aSAlexander Block return ret; 50178ea05e3aSAlexander Block 50180b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 50198ea05e3aSAlexander Block 50204a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 50218ea05e3aSAlexander Block ret = -EINVAL; 50228ea05e3aSAlexander Block goto out; 50238ea05e3aSAlexander Block } 50248ea05e3aSAlexander Block 50258ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 50268ea05e3aSAlexander Block ret = -EROFS; 50278ea05e3aSAlexander Block goto out; 50288ea05e3aSAlexander Block } 50298ea05e3aSAlexander Block 5030dd5f9615SStefan Behrens /* 5031dd5f9615SStefan Behrens * 1 - root item 5032dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5033dd5f9615SStefan Behrens */ 5034dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 50358ea05e3aSAlexander Block if (IS_ERR(trans)) { 50368ea05e3aSAlexander Block ret = PTR_ERR(trans); 50378ea05e3aSAlexander Block trans = NULL; 50388ea05e3aSAlexander Block goto out; 50398ea05e3aSAlexander Block } 50408ea05e3aSAlexander Block 50418ea05e3aSAlexander Block sa->rtransid = trans->transid; 50428ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 50438ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 50448ea05e3aSAlexander Block 5045dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5046dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5047dd5f9615SStefan Behrens if (received_uuid_changed && 5048d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5049d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5050dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5051dd5f9615SStefan Behrens root->root_key.objectid); 5052d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5053d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5054d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5055d87ff758SNikolay Borisov goto out; 5056d87ff758SNikolay Borisov } 5057d87ff758SNikolay Borisov } 50588ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 50598ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 50608ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 50613cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 50623cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 50633cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 50643cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 50658ea05e3aSAlexander Block 50660b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 50678ea05e3aSAlexander Block &root->root_key, &root->root_item); 50688ea05e3aSAlexander Block if (ret < 0) { 50693a45bb20SJeff Mahoney btrfs_end_transaction(trans); 50708ea05e3aSAlexander Block goto out; 5071dd5f9615SStefan Behrens } 5072dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5073cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5074dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5075dd5f9615SStefan Behrens root->root_key.objectid); 5076dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 507766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5078efd38150SNikolay Borisov btrfs_end_transaction(trans); 5079dd5f9615SStefan Behrens goto out; 5080dd5f9615SStefan Behrens } 5081dd5f9615SStefan Behrens } 50823a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5083abccd00fSHugo Mills out: 50840b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5085abccd00fSHugo Mills mnt_drop_write_file(file); 5086abccd00fSHugo Mills return ret; 5087abccd00fSHugo Mills } 5088abccd00fSHugo Mills 5089abccd00fSHugo Mills #ifdef CONFIG_64BIT 5090abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5091abccd00fSHugo Mills void __user *arg) 5092abccd00fSHugo Mills { 5093abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5094abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5095abccd00fSHugo Mills int ret = 0; 5096abccd00fSHugo Mills 5097abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 50987b9ea627SShailendra Verma if (IS_ERR(args32)) 50997b9ea627SShailendra Verma return PTR_ERR(args32); 5100abccd00fSHugo Mills 51018d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 510284dbeb87SDan Carpenter if (!args64) { 510384dbeb87SDan Carpenter ret = -ENOMEM; 5104abccd00fSHugo Mills goto out; 5105abccd00fSHugo Mills } 5106abccd00fSHugo Mills 5107abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5108abccd00fSHugo Mills args64->stransid = args32->stransid; 5109abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5110abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5111abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5112abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5113abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5114abccd00fSHugo Mills args64->flags = args32->flags; 5115abccd00fSHugo Mills 5116abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5117abccd00fSHugo Mills if (ret) 5118abccd00fSHugo Mills goto out; 5119abccd00fSHugo Mills 5120abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5121abccd00fSHugo Mills args32->stransid = args64->stransid; 5122abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5123abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5124abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5125abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5126abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5127abccd00fSHugo Mills args32->flags = args64->flags; 5128abccd00fSHugo Mills 5129abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5130abccd00fSHugo Mills if (ret) 5131abccd00fSHugo Mills ret = -EFAULT; 5132abccd00fSHugo Mills 5133abccd00fSHugo Mills out: 5134abccd00fSHugo Mills kfree(args32); 5135abccd00fSHugo Mills kfree(args64); 5136abccd00fSHugo Mills return ret; 5137abccd00fSHugo Mills } 5138abccd00fSHugo Mills #endif 5139abccd00fSHugo Mills 5140abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5141abccd00fSHugo Mills void __user *arg) 5142abccd00fSHugo Mills { 5143abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5144abccd00fSHugo Mills int ret = 0; 5145abccd00fSHugo Mills 5146abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 51477b9ea627SShailendra Verma if (IS_ERR(sa)) 51487b9ea627SShailendra Verma return PTR_ERR(sa); 5149abccd00fSHugo Mills 5150abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5151abccd00fSHugo Mills 5152abccd00fSHugo Mills if (ret) 5153abccd00fSHugo Mills goto out; 5154abccd00fSHugo Mills 51558ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 51568ea05e3aSAlexander Block if (ret) 51578ea05e3aSAlexander Block ret = -EFAULT; 51588ea05e3aSAlexander Block 51598ea05e3aSAlexander Block out: 51608ea05e3aSAlexander Block kfree(sa); 51618ea05e3aSAlexander Block return ret; 51628ea05e3aSAlexander Block } 51638ea05e3aSAlexander Block 5164b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, 5165b929c1d8SMarcos Paulo de Souza void __user *arg) 5166867ab667Sjeff.liu { 5167a1b83ac5SAnand Jain size_t len; 5168867ab667Sjeff.liu int ret; 5169a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5170a1b83ac5SAnand Jain 51710b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 51720b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 51730b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5174a1b83ac5SAnand Jain 5175a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5176867ab667Sjeff.liu 5177867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 51780b246afaSJeff Mahoney btrfs_warn(fs_info, 51790b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 51800b246afaSJeff Mahoney --len); 5181867ab667Sjeff.liu } 5182867ab667Sjeff.liu 5183867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5184867ab667Sjeff.liu 5185867ab667Sjeff.liu return ret ? -EFAULT : 0; 5186867ab667Sjeff.liu } 5187867ab667Sjeff.liu 5188a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5189a8bfd4abSjeff.liu { 51900b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51910b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51920b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51930b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5194a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5195a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5196a8bfd4abSjeff.liu int ret; 5197a8bfd4abSjeff.liu 5198a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5199a8bfd4abSjeff.liu return -EPERM; 5200a8bfd4abSjeff.liu 5201a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5202a8bfd4abSjeff.liu return -EFAULT; 5203a8bfd4abSjeff.liu 5204a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 52050b246afaSJeff Mahoney btrfs_err(fs_info, 52065d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5207a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5208a8bfd4abSjeff.liu return -EINVAL; 5209a8bfd4abSjeff.liu } 5210a8bfd4abSjeff.liu 5211a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5212a8bfd4abSjeff.liu if (ret) 5213a8bfd4abSjeff.liu return ret; 5214a8bfd4abSjeff.liu 5215a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5216a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5217a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5218a8bfd4abSjeff.liu goto out_unlock; 5219a8bfd4abSjeff.liu } 5220a8bfd4abSjeff.liu 52210b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5222a8bfd4abSjeff.liu strcpy(super_block->label, label); 52230b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 52243a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5225a8bfd4abSjeff.liu 5226a8bfd4abSjeff.liu out_unlock: 5227a8bfd4abSjeff.liu mnt_drop_write_file(file); 5228a8bfd4abSjeff.liu return ret; 5229a8bfd4abSjeff.liu } 5230a8bfd4abSjeff.liu 52312eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 52322eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 52332eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 52342eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 52352eaa055fSJeff Mahoney 5236d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 52372eaa055fSJeff Mahoney { 52384d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 52392eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 52402eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 52412eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 52422eaa055fSJeff Mahoney }; 52432eaa055fSJeff Mahoney 52442eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52452eaa055fSJeff Mahoney return -EFAULT; 52462eaa055fSJeff Mahoney 52472eaa055fSJeff Mahoney return 0; 52482eaa055fSJeff Mahoney } 52492eaa055fSJeff Mahoney 5250b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, 5251b929c1d8SMarcos Paulo de Souza void __user *arg) 52522eaa055fSJeff Mahoney { 52530b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 52542eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 52552eaa055fSJeff Mahoney 52562eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 52572eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 52582eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 52592eaa055fSJeff Mahoney 52602eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52612eaa055fSJeff Mahoney return -EFAULT; 52622eaa055fSJeff Mahoney 52632eaa055fSJeff Mahoney return 0; 52642eaa055fSJeff Mahoney } 52652eaa055fSJeff Mahoney 52662ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 52673b02a68aSJeff Mahoney enum btrfs_feature_set set, 52682eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 52692eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 52702eaa055fSJeff Mahoney { 5271f10152bcSDavid Sterba const char *type = btrfs_feature_set_name(set); 52723b02a68aSJeff Mahoney char *names; 52732eaa055fSJeff Mahoney u64 disallowed, unsupported; 52742eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 52752eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 52762eaa055fSJeff Mahoney 52772eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 52782eaa055fSJeff Mahoney if (unsupported) { 52793b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 52803b02a68aSJeff Mahoney if (names) { 52810b246afaSJeff Mahoney btrfs_warn(fs_info, 52823b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 52833b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 52843b02a68aSJeff Mahoney kfree(names); 52853b02a68aSJeff Mahoney } else 52860b246afaSJeff Mahoney btrfs_warn(fs_info, 52872eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 52882eaa055fSJeff Mahoney type, unsupported); 52892eaa055fSJeff Mahoney return -EOPNOTSUPP; 52902eaa055fSJeff Mahoney } 52912eaa055fSJeff Mahoney 52922eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 52932eaa055fSJeff Mahoney if (disallowed) { 52943b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 52953b02a68aSJeff Mahoney if (names) { 52960b246afaSJeff Mahoney btrfs_warn(fs_info, 52973b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 52983b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 52993b02a68aSJeff Mahoney kfree(names); 53003b02a68aSJeff Mahoney } else 53010b246afaSJeff Mahoney btrfs_warn(fs_info, 53022eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53032eaa055fSJeff Mahoney type, disallowed); 53042eaa055fSJeff Mahoney return -EPERM; 53052eaa055fSJeff Mahoney } 53062eaa055fSJeff Mahoney 53072eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 53082eaa055fSJeff Mahoney if (disallowed) { 53093b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53103b02a68aSJeff Mahoney if (names) { 53110b246afaSJeff Mahoney btrfs_warn(fs_info, 53123b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 53133b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53143b02a68aSJeff Mahoney kfree(names); 53153b02a68aSJeff Mahoney } else 53160b246afaSJeff Mahoney btrfs_warn(fs_info, 53172eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 53182eaa055fSJeff Mahoney type, disallowed); 53192eaa055fSJeff Mahoney return -EPERM; 53202eaa055fSJeff Mahoney } 53212eaa055fSJeff Mahoney 53222eaa055fSJeff Mahoney return 0; 53232eaa055fSJeff Mahoney } 53242eaa055fSJeff Mahoney 53252ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 53262ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 53272eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 53282eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 53292eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 53302eaa055fSJeff Mahoney 53312eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 53322eaa055fSJeff Mahoney { 53330b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53340b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53350b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 53360b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53372eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 53382eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 53392eaa055fSJeff Mahoney u64 newflags; 53402eaa055fSJeff Mahoney int ret; 53412eaa055fSJeff Mahoney 53422eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 53432eaa055fSJeff Mahoney return -EPERM; 53442eaa055fSJeff Mahoney 53452eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 53462eaa055fSJeff Mahoney return -EFAULT; 53472eaa055fSJeff Mahoney 53482eaa055fSJeff Mahoney /* Nothing to do */ 53492eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 53502eaa055fSJeff Mahoney !flags[0].incompat_flags) 53512eaa055fSJeff Mahoney return 0; 53522eaa055fSJeff Mahoney 53532ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 53542eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 53552eaa055fSJeff Mahoney if (ret) 53562eaa055fSJeff Mahoney return ret; 53572eaa055fSJeff Mahoney 53582ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 53592eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 53602eaa055fSJeff Mahoney if (ret) 53612eaa055fSJeff Mahoney return ret; 53622eaa055fSJeff Mahoney 53632ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 53642eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 53652eaa055fSJeff Mahoney if (ret) 53662eaa055fSJeff Mahoney return ret; 53672eaa055fSJeff Mahoney 53687ab19625SDavid Sterba ret = mnt_want_write_file(file); 53697ab19625SDavid Sterba if (ret) 53707ab19625SDavid Sterba return ret; 53717ab19625SDavid Sterba 53728051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 53737ab19625SDavid Sterba if (IS_ERR(trans)) { 53747ab19625SDavid Sterba ret = PTR_ERR(trans); 53757ab19625SDavid Sterba goto out_drop_write; 53767ab19625SDavid Sterba } 53772eaa055fSJeff Mahoney 53780b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 53792eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 53802eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 53812eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 53822eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 53832eaa055fSJeff Mahoney 53842eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 53852eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 53862eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 53872eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 53882eaa055fSJeff Mahoney 53892eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 53902eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 53912eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 53922eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 53930b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53942eaa055fSJeff Mahoney 53953a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 53967ab19625SDavid Sterba out_drop_write: 53977ab19625SDavid Sterba mnt_drop_write_file(file); 53987ab19625SDavid Sterba 53997ab19625SDavid Sterba return ret; 54002eaa055fSJeff Mahoney } 54012eaa055fSJeff Mahoney 54022351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 54032351f431SJosef Bacik { 54042351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 54052351f431SJosef Bacik int ret; 54062351f431SJosef Bacik 54072351f431SJosef Bacik if (compat) { 54082351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 54092351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 54102351f431SJosef Bacik 54112351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 54122351f431SJosef Bacik if (ret) 54132351f431SJosef Bacik return -EFAULT; 54142351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 54152351f431SJosef Bacik if (!arg) 54162351f431SJosef Bacik return -ENOMEM; 54172351f431SJosef Bacik arg->send_fd = args32.send_fd; 54182351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 54192351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 54202351f431SJosef Bacik arg->parent_root = args32.parent_root; 54212351f431SJosef Bacik arg->flags = args32.flags; 54222351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 54232351f431SJosef Bacik sizeof(args32.reserved)); 54242351f431SJosef Bacik #else 54252351f431SJosef Bacik return -ENOTTY; 54262351f431SJosef Bacik #endif 54272351f431SJosef Bacik } else { 54282351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 54292351f431SJosef Bacik if (IS_ERR(arg)) 54302351f431SJosef Bacik return PTR_ERR(arg); 54312351f431SJosef Bacik } 54322351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 54332351f431SJosef Bacik kfree(arg); 54342351f431SJosef Bacik return ret; 54352351f431SJosef Bacik } 54362351f431SJosef Bacik 5437f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5438f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5439f46b5a66SChristoph Hellwig { 54400b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54410b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54420b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54434bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5444f46b5a66SChristoph Hellwig 5445f46b5a66SChristoph Hellwig switch (cmd) { 54466cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 54476cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 54486cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 54496cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 54506cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 54516cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 545240cf931fSEric Sandeen case FS_IOC_GETFSLABEL: 5453b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_fslabel(fs_info, argp); 545440cf931fSEric Sandeen case FS_IOC_SETFSLABEL: 545540cf931fSEric Sandeen return btrfs_ioctl_set_fslabel(file, argp); 5456f7039b1dSLi Dongyang case FITRIM: 5457b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_fitrim(fs_info, argp); 5458f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5459fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5460fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5461fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 54623de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5463fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 54646f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 54656f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 546676dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 546776dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 54680caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 54690caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 54700caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 54710caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 54726ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 54736ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5474f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 54751e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 54761e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 54771e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5478f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5479198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5480f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 54812ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5482f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5483da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 54846b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 54856b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5486475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 54872ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5488475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 54892ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5490f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 54919ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5492ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5493ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5494cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5495cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5496ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5497ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5498d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5499d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5500d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5501d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5502d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5503d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 55041406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55052ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 55069b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55079b199859SFilipe David Borba Manana int ret; 55089b199859SFilipe David Borba Manana 550982b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 55109b199859SFilipe David Borba Manana if (ret) 55119b199859SFilipe David Borba Manana return ret; 55120b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 55132fad4e83SDavid Sterba /* 55142fad4e83SDavid Sterba * The transaction thread may want to do more work, 551501327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 55162fad4e83SDavid Sterba * processing uncleaned subvols. 55172fad4e83SDavid Sterba */ 55180b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 55199b199859SFilipe David Borba Manana return ret; 55209b199859SFilipe David Borba Manana } 552146204592SSage Weil case BTRFS_IOC_START_SYNC: 55229a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 552346204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55242ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5525475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5526b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5527475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 55282ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5529475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 55302ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5531c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55329ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5533837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 55342ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 553519a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 55362ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 55378ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55388ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5539abccd00fSHugo Mills #ifdef CONFIG_64BIT 5540abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5541abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5542abccd00fSHugo Mills #endif 554331db9f7cSAlexander Block case BTRFS_IOC_SEND: 55442351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 55452351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55462351f431SJosef Bacik case BTRFS_IOC_SEND_32: 55472351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 55482351f431SJosef Bacik #endif 5549c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 55502ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 55515d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5552905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55535d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5554905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55555d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5556905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55575d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5558905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 55592f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 55602f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 55612f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 5562b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_status(fs_info, argp); 556357254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 5564b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_wait(fs_info, argp); 55653f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 55662ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 55672eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5568d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 55692eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 5570b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_features(fs_info, argp); 55712eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 55722eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5573e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5574e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5575025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5576025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5577b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5578b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 557942e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 558042e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 558123d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 558223d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5583f46b5a66SChristoph Hellwig } 5584f46b5a66SChristoph Hellwig 5585f46b5a66SChristoph Hellwig return -ENOTTY; 5586f46b5a66SChristoph Hellwig } 55874c63c245SLuke Dashjr 55884c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 55894c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 55904c63c245SLuke Dashjr { 55912a362249SJeff Mahoney /* 55922a362249SJeff Mahoney * These all access 32-bit values anyway so no further 55932a362249SJeff Mahoney * handling is necessary. 55942a362249SJeff Mahoney */ 55954c63c245SLuke Dashjr switch (cmd) { 55964c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 55974c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 55984c63c245SLuke Dashjr break; 55994c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56004c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56014c63c245SLuke Dashjr break; 56024c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56034c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56044c63c245SLuke Dashjr break; 56054c63c245SLuke Dashjr } 56064c63c245SLuke Dashjr 56074c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56084c63c245SLuke Dashjr } 56094c63c245SLuke Dashjr #endif 5610