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 482f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 483f7039b1dSLi Dongyang { 4840b246afaSJeff Mahoney struct inode *inode = file_inode(file); 4850b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 486f7039b1dSLi Dongyang struct btrfs_device *device; 487f7039b1dSLi Dongyang struct request_queue *q; 488f7039b1dSLi Dongyang struct fstrim_range range; 489f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 490f7039b1dSLi Dongyang u64 num_devices = 0; 491f7039b1dSLi Dongyang int ret; 492f7039b1dSLi Dongyang 493f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 494f7039b1dSLi Dongyang return -EPERM; 495f7039b1dSLi Dongyang 496f35f06c3SFilipe Manana /* 497f35f06c3SFilipe Manana * If the fs is mounted with nologreplay, which requires it to be 498f35f06c3SFilipe Manana * mounted in RO mode as well, we can not allow discard on free space 499f35f06c3SFilipe Manana * inside block groups, because log trees refer to extents that are not 500f35f06c3SFilipe Manana * pinned in a block group's free space cache (pinning the extents is 501f35f06c3SFilipe Manana * precisely the first phase of replaying a log tree). 502f35f06c3SFilipe Manana */ 503f35f06c3SFilipe Manana if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 504f35f06c3SFilipe Manana return -EROFS; 505f35f06c3SFilipe Manana 5061f78160cSXiao Guangrong rcu_read_lock(); 5071f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 5081f78160cSXiao Guangrong dev_list) { 509f7039b1dSLi Dongyang if (!device->bdev) 510f7039b1dSLi Dongyang continue; 511f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 512f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 513f7039b1dSLi Dongyang num_devices++; 51450d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 515f7039b1dSLi Dongyang minlen); 516f7039b1dSLi Dongyang } 517f7039b1dSLi Dongyang } 5181f78160cSXiao Guangrong rcu_read_unlock(); 519f4c697e6SLukas Czerner 520f7039b1dSLi Dongyang if (!num_devices) 521f7039b1dSLi Dongyang return -EOPNOTSUPP; 522f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 523f7039b1dSLi Dongyang return -EFAULT; 5246ba9fc8eSQu Wenruo 5256ba9fc8eSQu Wenruo /* 5266ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 5276ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 5286ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 5296ba9fc8eSQu Wenruo */ 5306ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 531f4c697e6SLukas Czerner return -EINVAL; 532f7039b1dSLi Dongyang 533f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5342ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 535f7039b1dSLi Dongyang if (ret < 0) 536f7039b1dSLi Dongyang return ret; 537f7039b1dSLi Dongyang 538f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 539f7039b1dSLi Dongyang return -EFAULT; 540f7039b1dSLi Dongyang 541f7039b1dSLi Dongyang return 0; 542f7039b1dSLi Dongyang } 543f7039b1dSLi Dongyang 544dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 545dd5f9615SStefan Behrens { 54646e0f66aSChris Mason int i; 54746e0f66aSChris Mason 54846e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 54946e0f66aSChris Mason if (uuid[i]) 55046e0f66aSChris Mason return 0; 55146e0f66aSChris Mason } 55246e0f66aSChris Mason return 1; 553dd5f9615SStefan Behrens } 554dd5f9615SStefan Behrens 555d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 556cb8e7090SChristoph Hellwig struct dentry *dentry, 55752f75f4fSDavid Sterba const char *name, int namelen, 5586f72c7e2SArne Jansen u64 *async_transid, 5598696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 560f46b5a66SChristoph Hellwig { 5610b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 562f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 563f46b5a66SChristoph Hellwig struct btrfs_key key; 56449a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 565f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 566f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 567d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 56876dda93cSYan, Zheng struct btrfs_root *new_root; 569d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 57095582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5715662344bSTsutomu Itoh struct inode *inode; 572f46b5a66SChristoph Hellwig int ret; 573f46b5a66SChristoph Hellwig int err; 574f46b5a66SChristoph Hellwig u64 objectid; 575f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 5763de4586cSChris Mason u64 index = 0; 5778ea05e3aSAlexander Block uuid_le new_uuid; 578f46b5a66SChristoph Hellwig 57949a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 58049a3c4d9SDavid Sterba if (!root_item) 58149a3c4d9SDavid Sterba return -ENOMEM; 58249a3c4d9SDavid Sterba 5830b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5842fbe8c8aSAl Viro if (ret) 58549a3c4d9SDavid Sterba goto fail_free; 5866a912213SJosef Bacik 587e09fe2d2SQu Wenruo /* 588e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 58901327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 590e09fe2d2SQu Wenruo */ 59149a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 59249a3c4d9SDavid Sterba ret = -ENOSPC; 59349a3c4d9SDavid Sterba goto fail_free; 59449a3c4d9SDavid Sterba } 595e09fe2d2SQu Wenruo 596d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 5979ed74f2dSJosef Bacik /* 598d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 599d5c12070SMiao Xie * of create_snapshot(). 6009ed74f2dSJosef Bacik */ 601c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 602d5c12070SMiao Xie if (ret) 60349a3c4d9SDavid Sterba goto fail_free; 604f46b5a66SChristoph Hellwig 605d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 606d5c12070SMiao Xie if (IS_ERR(trans)) { 607d5c12070SMiao Xie ret = PTR_ERR(trans); 6087775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 60949a3c4d9SDavid Sterba goto fail_free; 610d5c12070SMiao Xie } 611d5c12070SMiao Xie trans->block_rsv = &block_rsv; 612d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 613f46b5a66SChristoph Hellwig 614a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 6156f72c7e2SArne Jansen if (ret) 6166f72c7e2SArne Jansen goto fail; 6176f72c7e2SArne Jansen 6184d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6198e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6208e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6218e8a1e31SJosef Bacik goto fail; 6228e8a1e31SJosef Bacik } 623f46b5a66SChristoph Hellwig 624f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 625f46b5a66SChristoph Hellwig 62649a3c4d9SDavid Sterba inode_item = &root_item->inode; 6273cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6283cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6293cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 630da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6310b246afaSJeff Mahoney fs_info->nodesize); 6323cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 633f46b5a66SChristoph Hellwig 63449a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 63549a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6363cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 63708fe4db1SLi Zefan 63849a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 63949a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 64049a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 64149a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 64249a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 64349a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 644f46b5a66SChristoph Hellwig 64549a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 64649a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6478ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 64849a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 64949a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 65049a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 65149a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 65249a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 65349a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 654f46b5a66SChristoph Hellwig 655925baeddSChris Mason btrfs_tree_unlock(leaf); 656f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 657f46b5a66SChristoph Hellwig leaf = NULL; 658f46b5a66SChristoph Hellwig 65949a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 660f46b5a66SChristoph Hellwig 661f46b5a66SChristoph Hellwig key.objectid = objectid; 6625d4f98a2SYan Zheng key.offset = 0; 663962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6640b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 66549a3c4d9SDavid Sterba root_item); 666f46b5a66SChristoph Hellwig if (ret) 667f46b5a66SChristoph Hellwig goto fail; 668f46b5a66SChristoph Hellwig 66976dda93cSYan, Zheng key.offset = (u64)-1; 6700b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 67179787eaaSJeff Mahoney if (IS_ERR(new_root)) { 67279787eaaSJeff Mahoney ret = PTR_ERR(new_root); 67366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 67479787eaaSJeff Mahoney goto fail; 67579787eaaSJeff Mahoney } 67676dda93cSYan, Zheng 67776dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 67876dda93cSYan, Zheng 67963541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 680ce598979SMark Fasheh if (ret) { 681ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 68266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 683ce598979SMark Fasheh goto fail; 684ce598979SMark Fasheh } 685ce598979SMark Fasheh 686f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 687f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 688f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 689f32e48e9SChandan Rajendra 690f46b5a66SChristoph Hellwig /* 691f46b5a66SChristoph Hellwig * insert the directory item 692f46b5a66SChristoph Hellwig */ 693877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 69479787eaaSJeff Mahoney if (ret) { 69566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 69679787eaaSJeff Mahoney goto fail; 69779787eaaSJeff Mahoney } 6983de4586cSChris Mason 699684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 7003de4586cSChris Mason BTRFS_FT_DIR, index); 70179787eaaSJeff Mahoney if (ret) { 70266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 703f46b5a66SChristoph Hellwig goto fail; 70479787eaaSJeff Mahoney } 7050660b5afSChris Mason 7066ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 70752c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 70852c26179SYan Zheng BUG_ON(ret); 70952c26179SYan Zheng 7106025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7114a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7120660b5afSChris Mason BUG_ON(ret); 7130660b5afSChris Mason 714cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7156bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 716dd5f9615SStefan Behrens if (ret) 71766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 718dd5f9615SStefan Behrens 719f46b5a66SChristoph Hellwig fail: 72049a3c4d9SDavid Sterba kfree(root_item); 721d5c12070SMiao Xie trans->block_rsv = NULL; 722d5c12070SMiao Xie trans->bytes_reserved = 0; 7237775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 724de6e8200SLiu Bo 72572fd032eSSage Weil if (async_transid) { 72672fd032eSSage Weil *async_transid = trans->transid; 7273a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 72800d71c9cSMiao Xie if (err) 7293a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73072fd032eSSage Weil } else { 7313a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73272fd032eSSage Weil } 733f46b5a66SChristoph Hellwig if (err && !ret) 734f46b5a66SChristoph Hellwig ret = err; 7351a65e24bSChris Mason 7365662344bSTsutomu Itoh if (!ret) { 7375662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 738de6e8200SLiu Bo if (IS_ERR(inode)) 739de6e8200SLiu Bo return PTR_ERR(inode); 7405662344bSTsutomu Itoh d_instantiate(dentry, inode); 7415662344bSTsutomu Itoh } 742f46b5a66SChristoph Hellwig return ret; 74349a3c4d9SDavid Sterba 74449a3c4d9SDavid Sterba fail_free: 74549a3c4d9SDavid Sterba kfree(root_item); 74649a3c4d9SDavid Sterba return ret; 747f46b5a66SChristoph Hellwig } 748f46b5a66SChristoph Hellwig 749e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 75061d7e4cbSDavid Sterba struct dentry *dentry, 751e9662f70SMiao Xie u64 *async_transid, bool readonly, 752e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 753f46b5a66SChristoph Hellwig { 7540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7552e4bfab9SYan, Zheng struct inode *inode; 756f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 757f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7582e4bfab9SYan, Zheng int ret; 7598ecebf4dSRobbie Ko bool snapshot_force_cow = false; 760f46b5a66SChristoph Hellwig 76127cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 762f46b5a66SChristoph Hellwig return -EINVAL; 763f46b5a66SChristoph Hellwig 764eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 765eede2bf3SOmar Sandoval btrfs_warn(fs_info, 766eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 767eede2bf3SOmar Sandoval return -ETXTBSY; 768eede2bf3SOmar Sandoval } 769eede2bf3SOmar Sandoval 77023269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 771a1ee7362SDavid Sterba if (!pending_snapshot) 772a1ee7362SDavid Sterba return -ENOMEM; 773a1ee7362SDavid Sterba 774b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 77523269bf5SDavid Sterba GFP_KERNEL); 7768546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7778546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 778b0c0ea63SDavid Sterba ret = -ENOMEM; 779b0c0ea63SDavid Sterba goto free_pending; 780b0c0ea63SDavid Sterba } 781b0c0ea63SDavid Sterba 7828ecebf4dSRobbie Ko /* 7838ecebf4dSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 7848ecebf4dSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 7858ecebf4dSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 7868ecebf4dSRobbie Ko */ 787ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7884e857c58SPeter Zijlstra smp_mb__after_atomic(); 78945bac0f3SLiu Bo /* wait for no snapshot writes */ 79045bac0f3SLiu Bo wait_event(root->subv_writers->wait, 79145bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 7928257b2dcSMiao Xie 7933cd24c69SEthan Lien ret = btrfs_start_delalloc_snapshot(root); 7946a03843dSMiao Xie if (ret) 795a1ee7362SDavid Sterba goto dec_and_free; 7966a03843dSMiao Xie 7978ecebf4dSRobbie Ko /* 7988ecebf4dSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 7998ecebf4dSRobbie Ko * we force future writes to fallback to COW mode during snapshot 8008ecebf4dSRobbie Ko * creation. 8018ecebf4dSRobbie Ko */ 8028ecebf4dSRobbie Ko atomic_inc(&root->snapshot_force_cow); 8038ecebf4dSRobbie Ko snapshot_force_cow = true; 8048ecebf4dSRobbie Ko 8056374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 8066a03843dSMiao Xie 80766d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 80866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 809d5c12070SMiao Xie /* 810d5c12070SMiao Xie * 1 - parent dir inode 811d5c12070SMiao Xie * 2 - dir entries 812d5c12070SMiao Xie * 1 - root item 813d5c12070SMiao Xie * 2 - root ref/backref 814d5c12070SMiao Xie * 1 - root of snapshot 815dd5f9615SStefan Behrens * 1 - UUID item 816d5c12070SMiao Xie */ 817d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 818dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 819ee3441b4SJeff Mahoney false); 820d5c12070SMiao Xie if (ret) 821a1ee7362SDavid Sterba goto dec_and_free; 822d5c12070SMiao Xie 823a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 824a22285a6SYan, Zheng pending_snapshot->root = root; 825b83cc969SLi Zefan pending_snapshot->readonly = readonly; 826e9662f70SMiao Xie pending_snapshot->dir = dir; 8278696c533SMiao Xie pending_snapshot->inherit = inherit; 828a22285a6SYan, Zheng 829d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 830a22285a6SYan, Zheng if (IS_ERR(trans)) { 831a22285a6SYan, Zheng ret = PTR_ERR(trans); 832a22285a6SYan, Zheng goto fail; 833a22285a6SYan, Zheng } 834a22285a6SYan, Zheng 8350b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 836a22285a6SYan, Zheng list_add(&pending_snapshot->list, 837a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8380b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 83972fd032eSSage Weil if (async_transid) { 84072fd032eSSage Weil *async_transid = trans->transid; 8413a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 84200d71c9cSMiao Xie if (ret) 8433a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 84472fd032eSSage Weil } else { 8453a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 84672fd032eSSage Weil } 847aec8030aSMiao Xie if (ret) 848c37b2b62SJosef Bacik goto fail; 849a22285a6SYan, Zheng 850a22285a6SYan, Zheng ret = pending_snapshot->error; 851f46b5a66SChristoph Hellwig if (ret) 8522e4bfab9SYan, Zheng goto fail; 853f46b5a66SChristoph Hellwig 854d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 855d3797308SChris Mason if (ret) 856d3797308SChris Mason goto fail; 857d3797308SChris Mason 8582b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8592e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8602e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8612e4bfab9SYan, Zheng goto fail; 8622e4bfab9SYan, Zheng } 8635662344bSTsutomu Itoh 8642e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8652e4bfab9SYan, Zheng ret = 0; 8662e4bfab9SYan, Zheng fail: 8677775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 868a1ee7362SDavid Sterba dec_and_free: 8698ecebf4dSRobbie Ko if (snapshot_force_cow) 8708ecebf4dSRobbie Ko atomic_dec(&root->snapshot_force_cow); 871ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8724625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 873b0c0ea63SDavid Sterba free_pending: 874b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8758546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 876a1ee7362SDavid Sterba kfree(pending_snapshot); 877a1ee7362SDavid Sterba 878f46b5a66SChristoph Hellwig return ret; 879f46b5a66SChristoph Hellwig } 880f46b5a66SChristoph Hellwig 8814260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8824260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8834260f7c7SSage Weil * whether the type of victim is right. 8844260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8854260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8864260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8874260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8884260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8894260f7c7SSage Weil * a. be owner of dir, or 8904260f7c7SSage Weil * b. be owner of victim, or 8914260f7c7SSage Weil * c. have CAP_FOWNER capability 89201327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8934260f7c7SSage Weil * links pointing to it. 8944260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8954260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8964260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8974260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8984260f7c7SSage Weil * nfs_async_unlink(). 8994260f7c7SSage Weil */ 9004260f7c7SSage Weil 9014260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 9024260f7c7SSage Weil { 9034260f7c7SSage Weil int error; 9044260f7c7SSage Weil 9052b0143b5SDavid Howells if (d_really_is_negative(victim)) 9064260f7c7SSage Weil return -ENOENT; 9074260f7c7SSage Weil 9082b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 9094fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 9104260f7c7SSage Weil 9114260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 9124260f7c7SSage Weil if (error) 9134260f7c7SSage Weil return error; 9144260f7c7SSage Weil if (IS_APPEND(dir)) 9154260f7c7SSage Weil return -EPERM; 9162b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9172b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9184260f7c7SSage Weil return -EPERM; 9194260f7c7SSage Weil if (isdir) { 920e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9214260f7c7SSage Weil return -ENOTDIR; 9224260f7c7SSage Weil if (IS_ROOT(victim)) 9234260f7c7SSage Weil return -EBUSY; 924e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9254260f7c7SSage Weil return -EISDIR; 9264260f7c7SSage Weil if (IS_DEADDIR(dir)) 9274260f7c7SSage Weil return -ENOENT; 9284260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9294260f7c7SSage Weil return -EBUSY; 9304260f7c7SSage Weil return 0; 9314260f7c7SSage Weil } 9324260f7c7SSage Weil 933cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 934cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 935cb8e7090SChristoph Hellwig { 9362b0143b5SDavid Howells if (d_really_is_positive(child)) 937cb8e7090SChristoph Hellwig return -EEXIST; 938cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 939cb8e7090SChristoph Hellwig return -ENOENT; 940cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 941cb8e7090SChristoph Hellwig } 942cb8e7090SChristoph Hellwig 943cb8e7090SChristoph Hellwig /* 944cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 945cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 946cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 947cb8e7090SChristoph Hellwig */ 94892872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 94952f75f4fSDavid Sterba const char *name, int namelen, 95072fd032eSSage Weil struct btrfs_root *snap_src, 9516f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9528696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 953cb8e7090SChristoph Hellwig { 9542b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9550b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 956cb8e7090SChristoph Hellwig struct dentry *dentry; 957cb8e7090SChristoph Hellwig int error; 958cb8e7090SChristoph Hellwig 95900235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 96000235411SAl Viro if (error == -EINTR) 96100235411SAl Viro return error; 962cb8e7090SChristoph Hellwig 963cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 964cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 965cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 966cb8e7090SChristoph Hellwig goto out_unlock; 967cb8e7090SChristoph Hellwig 96876dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 969cb8e7090SChristoph Hellwig if (error) 970a874a63eSLiu Bo goto out_dput; 971cb8e7090SChristoph Hellwig 9729c52057cSChris Mason /* 9739c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9749c52057cSChris Mason * check for them now when we can safely fail 9759c52057cSChris Mason */ 9769c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9779c52057cSChris Mason dir->i_ino, name, 9789c52057cSChris Mason namelen); 9799c52057cSChris Mason if (error) 9809c52057cSChris Mason goto out_dput; 9819c52057cSChris Mason 9820b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 98376dda93cSYan, Zheng 98476dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 98576dda93cSYan, Zheng goto out_up_read; 98676dda93cSYan, Zheng 9873de4586cSChris Mason if (snap_src) { 98861d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9896f72c7e2SArne Jansen async_transid, readonly, inherit); 9903de4586cSChris Mason } else { 991d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 992d5c12070SMiao Xie async_transid, inherit); 9933de4586cSChris Mason } 99476dda93cSYan, Zheng if (!error) 99576dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 99676dda93cSYan, Zheng out_up_read: 9970b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 998cb8e7090SChristoph Hellwig out_dput: 999cb8e7090SChristoph Hellwig dput(dentry); 1000cb8e7090SChristoph Hellwig out_unlock: 10015955102cSAl Viro inode_unlock(dir); 1002cb8e7090SChristoph Hellwig return error; 1003cb8e7090SChristoph Hellwig } 1004cb8e7090SChristoph Hellwig 10054cb5300bSChris Mason /* 10064cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 10074cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 10084cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 10094cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 10104cb5300bSChris Mason * part of the file 10114cb5300bSChris Mason */ 1012aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 10134cb5300bSChris Mason { 10144cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10154cb5300bSChris Mason struct extent_map *em = NULL; 10164cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10174cb5300bSChris Mason u64 end; 10184cb5300bSChris Mason 10194cb5300bSChris Mason read_lock(&em_tree->lock); 102009cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10214cb5300bSChris Mason read_unlock(&em_tree->lock); 10224cb5300bSChris Mason 10234cb5300bSChris Mason if (em) { 10244cb5300bSChris Mason end = extent_map_end(em); 10254cb5300bSChris Mason free_extent_map(em); 10264cb5300bSChris Mason if (end - offset > thresh) 10274cb5300bSChris Mason return 0; 10284cb5300bSChris Mason } 10294cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10304cb5300bSChris Mason thresh /= 2; 10314cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10324cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10334cb5300bSChris Mason if (end >= thresh) 10344cb5300bSChris Mason return 0; 10354cb5300bSChris Mason return 1; 10364cb5300bSChris Mason } 10374cb5300bSChris Mason 10384cb5300bSChris Mason /* 10394cb5300bSChris Mason * helper function to walk through a file and find extents 10404cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10414cb5300bSChris Mason * 10424cb5300bSChris Mason * This is used by the defragging code to find new and small 10434cb5300bSChris Mason * extents 10444cb5300bSChris Mason */ 10454cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10464cb5300bSChris Mason struct inode *inode, u64 newer_than, 1047aab110abSDavid Sterba u64 *off, u32 thresh) 10484cb5300bSChris Mason { 10494cb5300bSChris Mason struct btrfs_path *path; 10504cb5300bSChris Mason struct btrfs_key min_key; 10514cb5300bSChris Mason struct extent_buffer *leaf; 10524cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10534cb5300bSChris Mason int type; 10544cb5300bSChris Mason int ret; 10554a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10564cb5300bSChris Mason 10574cb5300bSChris Mason path = btrfs_alloc_path(); 10584cb5300bSChris Mason if (!path) 10594cb5300bSChris Mason return -ENOMEM; 10604cb5300bSChris Mason 1061a4689d2bSDavid Sterba min_key.objectid = ino; 10624cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10634cb5300bSChris Mason min_key.offset = *off; 10644cb5300bSChris Mason 10654cb5300bSChris Mason while (1) { 10666174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10674cb5300bSChris Mason if (ret != 0) 10684cb5300bSChris Mason goto none; 1069f094c9bdSFilipe Manana process_slot: 1070a4689d2bSDavid Sterba if (min_key.objectid != ino) 10714cb5300bSChris Mason goto none; 10724cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10734cb5300bSChris Mason goto none; 10744cb5300bSChris Mason 10754cb5300bSChris Mason leaf = path->nodes[0]; 10764cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10774cb5300bSChris Mason struct btrfs_file_extent_item); 10784cb5300bSChris Mason 10794cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10804cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10814cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10824cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10834cb5300bSChris Mason *off = min_key.offset; 10844cb5300bSChris Mason btrfs_free_path(path); 10854cb5300bSChris Mason return 0; 10864cb5300bSChris Mason } 10874cb5300bSChris Mason 1088f094c9bdSFilipe Manana path->slots[0]++; 1089f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1090f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1091f094c9bdSFilipe Manana goto process_slot; 1092f094c9bdSFilipe Manana } 1093f094c9bdSFilipe Manana 10944cb5300bSChris Mason if (min_key.offset == (u64)-1) 10954cb5300bSChris Mason goto none; 10964cb5300bSChris Mason 10974cb5300bSChris Mason min_key.offset++; 10984cb5300bSChris Mason btrfs_release_path(path); 10994cb5300bSChris Mason } 11004cb5300bSChris Mason none: 11014cb5300bSChris Mason btrfs_free_path(path); 11024cb5300bSChris Mason return -ENOENT; 11034cb5300bSChris Mason } 11044cb5300bSChris Mason 11056c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 110617ce6ef8SLiu Bo { 110717ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1108940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 11096c282eb4SLi Zefan struct extent_map *em; 111009cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1111940100a4SChris Mason 1112940100a4SChris Mason /* 1113940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1114940100a4SChris Mason * the full extent lock 1115940100a4SChris Mason */ 1116940100a4SChris Mason read_lock(&em_tree->lock); 1117940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1118940100a4SChris Mason read_unlock(&em_tree->lock); 1119940100a4SChris Mason 1120940100a4SChris Mason if (!em) { 1121308d9800SFilipe Manana struct extent_state *cached = NULL; 1122308d9800SFilipe Manana u64 end = start + len - 1; 1123308d9800SFilipe Manana 1124940100a4SChris Mason /* get the big lock and read metadata off disk */ 1125ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1126fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1127e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1128940100a4SChris Mason 11296cf8bfbfSDan Carpenter if (IS_ERR(em)) 11306c282eb4SLi Zefan return NULL; 1131940100a4SChris Mason } 1132940100a4SChris Mason 11336c282eb4SLi Zefan return em; 11346c282eb4SLi Zefan } 11356c282eb4SLi Zefan 11366c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11376c282eb4SLi Zefan { 11386c282eb4SLi Zefan struct extent_map *next; 11396c282eb4SLi Zefan bool ret = true; 11406c282eb4SLi Zefan 11416c282eb4SLi Zefan /* this is the last extent */ 11426c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11436c282eb4SLi Zefan return false; 11446c282eb4SLi Zefan 11456c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1146e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1147e9512d72SChris Mason ret = false; 1148e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1149ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11506c282eb4SLi Zefan ret = false; 11516c282eb4SLi Zefan 11526c282eb4SLi Zefan free_extent_map(next); 11536c282eb4SLi Zefan return ret; 11546c282eb4SLi Zefan } 11556c282eb4SLi Zefan 1156aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1157a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1158a43a2111SAndrew Mahone int compress) 11596c282eb4SLi Zefan { 11606c282eb4SLi Zefan struct extent_map *em; 11616c282eb4SLi Zefan int ret = 1; 11626c282eb4SLi Zefan bool next_mergeable = true; 11634a3560c4SLiu Bo bool prev_mergeable = true; 11646c282eb4SLi Zefan 11656c282eb4SLi Zefan /* 11666c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11676c282eb4SLi Zefan * defragging it 11686c282eb4SLi Zefan */ 11696c282eb4SLi Zefan if (start < *defrag_end) 11706c282eb4SLi Zefan return 1; 11716c282eb4SLi Zefan 11726c282eb4SLi Zefan *skip = 0; 11736c282eb4SLi Zefan 11746c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11756c282eb4SLi Zefan if (!em) 11766c282eb4SLi Zefan return 0; 11776c282eb4SLi Zefan 1178940100a4SChris Mason /* this will cover holes, and inline extents */ 117917ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1180940100a4SChris Mason ret = 0; 118117ce6ef8SLiu Bo goto out; 118217ce6ef8SLiu Bo } 118317ce6ef8SLiu Bo 11844a3560c4SLiu Bo if (!*defrag_end) 11854a3560c4SLiu Bo prev_mergeable = false; 11864a3560c4SLiu Bo 11876c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1188940100a4SChris Mason /* 11896c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11906c282eb4SLi Zefan * real extent, don't bother defragging it 1191940100a4SChris Mason */ 1192a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11934a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1194940100a4SChris Mason ret = 0; 119517ce6ef8SLiu Bo out: 1196940100a4SChris Mason /* 1197940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1198940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1199940100a4SChris Mason * so that the next tiny extent will force a defrag. 1200940100a4SChris Mason * 1201940100a4SChris Mason * The end result of this is that tiny extents before a single big 1202940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1203940100a4SChris Mason */ 1204940100a4SChris Mason if (ret) { 1205940100a4SChris Mason *defrag_end = extent_map_end(em); 1206940100a4SChris Mason } else { 1207940100a4SChris Mason *last_len = 0; 1208940100a4SChris Mason *skip = extent_map_end(em); 1209940100a4SChris Mason *defrag_end = 0; 1210940100a4SChris Mason } 1211940100a4SChris Mason 1212940100a4SChris Mason free_extent_map(em); 1213940100a4SChris Mason return ret; 1214940100a4SChris Mason } 1215940100a4SChris Mason 12164cb5300bSChris Mason /* 12174cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12184cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12194cb5300bSChris Mason * to COW and defrag. 12204cb5300bSChris Mason * 12214cb5300bSChris Mason * It also makes sure the delalloc code has enough 12224cb5300bSChris Mason * dirty data to avoid making new small extents as part 12234cb5300bSChris Mason * of the defrag 12244cb5300bSChris Mason * 12254cb5300bSChris Mason * It's a good idea to start RA on this range 12264cb5300bSChris Mason * before calling this. 12274cb5300bSChris Mason */ 12284cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12294cb5300bSChris Mason struct page **pages, 12304cb5300bSChris Mason unsigned long start_index, 1231c41570c9SJustin Maggard unsigned long num_pages) 1232f46b5a66SChristoph Hellwig { 12334cb5300bSChris Mason unsigned long file_end; 12344cb5300bSChris Mason u64 isize = i_size_read(inode); 1235f46b5a66SChristoph Hellwig u64 page_start; 1236f46b5a66SChristoph Hellwig u64 page_end; 12371f12bd06SLiu Bo u64 page_cnt; 12384cb5300bSChris Mason int ret; 12394cb5300bSChris Mason int i; 12404cb5300bSChris Mason int i_done; 12414cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12424cb5300bSChris Mason struct extent_state *cached_state = NULL; 1243600a45e1SMiao Xie struct extent_io_tree *tree; 1244364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12453b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12464cb5300bSChris Mason 124709cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12481f12bd06SLiu Bo if (!isize || start_index > file_end) 12491f12bd06SLiu Bo return 0; 12501f12bd06SLiu Bo 12511f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12524cb5300bSChris Mason 1253364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 125409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 125509cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12564cb5300bSChris Mason if (ret) 12574cb5300bSChris Mason return ret; 12584cb5300bSChris Mason i_done = 0; 1259600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12604cb5300bSChris Mason 12614cb5300bSChris Mason /* step one, lock all the pages */ 12621f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12634cb5300bSChris Mason struct page *page; 1264600a45e1SMiao Xie again: 1265a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12663b16a4e3SJosef Bacik start_index + i, mask); 12674cb5300bSChris Mason if (!page) 12684cb5300bSChris Mason break; 12694cb5300bSChris Mason 1270600a45e1SMiao Xie page_start = page_offset(page); 127109cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1272600a45e1SMiao Xie while (1) { 1273308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1274ff13db41SDavid Sterba &cached_state); 1275600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1276600a45e1SMiao Xie page_start); 1277308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1278e43bbe5eSDavid Sterba &cached_state); 1279600a45e1SMiao Xie if (!ordered) 1280600a45e1SMiao Xie break; 1281600a45e1SMiao Xie 1282600a45e1SMiao Xie unlock_page(page); 1283600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1284600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1285600a45e1SMiao Xie lock_page(page); 12861f12bd06SLiu Bo /* 12871f12bd06SLiu Bo * we unlocked the page above, so we need check if 12881f12bd06SLiu Bo * it was released or not. 12891f12bd06SLiu Bo */ 12901f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12911f12bd06SLiu Bo unlock_page(page); 129209cbfeafSKirill A. Shutemov put_page(page); 12931f12bd06SLiu Bo goto again; 12941f12bd06SLiu Bo } 1295600a45e1SMiao Xie } 1296600a45e1SMiao Xie 12974cb5300bSChris Mason if (!PageUptodate(page)) { 12984cb5300bSChris Mason btrfs_readpage(NULL, page); 12994cb5300bSChris Mason lock_page(page); 13004cb5300bSChris Mason if (!PageUptodate(page)) { 13014cb5300bSChris Mason unlock_page(page); 130209cbfeafSKirill A. Shutemov put_page(page); 13034cb5300bSChris Mason ret = -EIO; 13044cb5300bSChris Mason break; 13054cb5300bSChris Mason } 13064cb5300bSChris Mason } 1307600a45e1SMiao Xie 1308600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1309600a45e1SMiao Xie unlock_page(page); 131009cbfeafSKirill A. Shutemov put_page(page); 1311600a45e1SMiao Xie goto again; 1312600a45e1SMiao Xie } 1313600a45e1SMiao Xie 13144cb5300bSChris Mason pages[i] = page; 13154cb5300bSChris Mason i_done++; 13164cb5300bSChris Mason } 13174cb5300bSChris Mason if (!i_done || ret) 13184cb5300bSChris Mason goto out; 13194cb5300bSChris Mason 13201751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13214cb5300bSChris Mason goto out; 13224cb5300bSChris Mason 13234cb5300bSChris Mason /* 13244cb5300bSChris Mason * so now we have a nice long stream of locked 13254cb5300bSChris Mason * and up to date pages, lets wait on them 13264cb5300bSChris Mason */ 13274cb5300bSChris Mason for (i = 0; i < i_done; i++) 13284cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13294cb5300bSChris Mason 13304cb5300bSChris Mason page_start = page_offset(pages[0]); 133109cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13324cb5300bSChris Mason 13334cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1334ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13354cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 1336e182163dSOmar Sandoval page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | 1337e182163dSOmar Sandoval EXTENT_DEFRAG, 0, 0, &cached_state); 13384cb5300bSChris Mason 13391f12bd06SLiu Bo if (i_done != page_cnt) { 13409e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 134128c4a3e2SSu Yue btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); 13429e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1343bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 134409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 134543b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13464cb5300bSChris Mason } 13474cb5300bSChris Mason 13484cb5300bSChris Mason 13499e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1350018ed4f7SDavid Sterba &cached_state); 13514cb5300bSChris Mason 13524cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1353e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13544cb5300bSChris Mason 13554cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13564cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13574cb5300bSChris Mason ClearPageChecked(pages[i]); 13584cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13594cb5300bSChris Mason set_page_dirty(pages[i]); 13604cb5300bSChris Mason unlock_page(pages[i]); 136109cbfeafSKirill A. Shutemov put_page(pages[i]); 13624cb5300bSChris Mason } 136343b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 136443b18595SQu Wenruo false); 1365364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13664cb5300bSChris Mason return i_done; 13674cb5300bSChris Mason out: 13684cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13694cb5300bSChris Mason unlock_page(pages[i]); 137009cbfeafSKirill A. Shutemov put_page(pages[i]); 13714cb5300bSChris Mason } 1372bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 137309cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 137443b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 137543b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 137643b18595SQu Wenruo true); 1377364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13784cb5300bSChris Mason return ret; 13794cb5300bSChris Mason 13804cb5300bSChris Mason } 13814cb5300bSChris Mason 13824cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13834cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13844cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13854cb5300bSChris Mason { 13860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13874cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13884cb5300bSChris Mason struct file_ra_state *ra = NULL; 13894cb5300bSChris Mason unsigned long last_index; 1390151a31b2SLi Zefan u64 isize = i_size_read(inode); 1391940100a4SChris Mason u64 last_len = 0; 1392940100a4SChris Mason u64 skip = 0; 1393940100a4SChris Mason u64 defrag_end = 0; 13944cb5300bSChris Mason u64 newer_off = range->start; 1395f46b5a66SChristoph Hellwig unsigned long i; 1396008873eaSLi Zefan unsigned long ra_index = 0; 1397f46b5a66SChristoph Hellwig int ret; 13984cb5300bSChris Mason int defrag_count = 0; 13991a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1400aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 140109cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1402c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1403ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 14044cb5300bSChris Mason struct page **pages = NULL; 14051e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14064cb5300bSChris Mason 14070abd5b17SLiu Bo if (isize == 0) 14080abd5b17SLiu Bo return 0; 14090abd5b17SLiu Bo 14100abd5b17SLiu Bo if (range->start >= isize) 14110abd5b17SLiu Bo return -EINVAL; 14121a419d85SLi Zefan 14131e2ef46dSDavid Sterba if (do_compress) { 14141a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 14151a419d85SLi Zefan return -EINVAL; 14161a419d85SLi Zefan if (range->compress_type) 14171a419d85SLi Zefan compress_type = range->compress_type; 14181a419d85SLi Zefan } 1419f46b5a66SChristoph Hellwig 14200abd5b17SLiu Bo if (extent_thresh == 0) 1421ee22184bSByongho Lee extent_thresh = SZ_256K; 1422f46b5a66SChristoph Hellwig 14234cb5300bSChris Mason /* 14240a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14250a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14260a52d108SDavid Sterba * we don't error out. 14274cb5300bSChris Mason */ 14284cb5300bSChris Mason if (!file) { 142963e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14300a52d108SDavid Sterba if (ra) 14314cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14324cb5300bSChris Mason } else { 14334cb5300bSChris Mason ra = &file->f_ra; 14344cb5300bSChris Mason } 14354cb5300bSChris Mason 143663e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14374cb5300bSChris Mason if (!pages) { 14384cb5300bSChris Mason ret = -ENOMEM; 14394cb5300bSChris Mason goto out_ra; 14404cb5300bSChris Mason } 14414cb5300bSChris Mason 14424cb5300bSChris Mason /* find the last page to defrag */ 14431e701a32SChris Mason if (range->start + range->len > range->start) { 1444151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 144509cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14461e701a32SChris Mason } else { 144709cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14481e701a32SChris Mason } 14491e701a32SChris Mason 14504cb5300bSChris Mason if (newer_than) { 14514cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1452ee22184bSByongho Lee &newer_off, SZ_64K); 14534cb5300bSChris Mason if (!ret) { 14544cb5300bSChris Mason range->start = newer_off; 14554cb5300bSChris Mason /* 14564cb5300bSChris Mason * we always align our defrag to help keep 14574cb5300bSChris Mason * the extents in the file evenly spaced 14584cb5300bSChris Mason */ 145909cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14604cb5300bSChris Mason } else 14614cb5300bSChris Mason goto out_ra; 14624cb5300bSChris Mason } else { 146309cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14644cb5300bSChris Mason } 14654cb5300bSChris Mason if (!max_to_defrag) 1466070034bdSchandan max_to_defrag = last_index - i + 1; 14674cb5300bSChris Mason 14682a0f7f57SLi Zefan /* 14692a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14702a0f7f57SLi Zefan * written sequentially. 14712a0f7f57SLi Zefan */ 14722a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14732a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14742a0f7f57SLi Zefan 1475f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 147609cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14774cb5300bSChris Mason /* 14784cb5300bSChris Mason * make sure we stop running if someone unmounts 14794cb5300bSChris Mason * the FS 14804cb5300bSChris Mason */ 14811751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14824cb5300bSChris Mason break; 14834cb5300bSChris Mason 14840b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14850b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1486210549ebSDavid Sterba ret = -EAGAIN; 1487210549ebSDavid Sterba break; 1488210549ebSDavid Sterba } 1489210549ebSDavid Sterba 149009cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14916c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14921e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1493940100a4SChris Mason unsigned long next; 1494940100a4SChris Mason /* 1495940100a4SChris Mason * the should_defrag function tells us how much to skip 1496940100a4SChris Mason * bump our counter by the suggested amount 1497940100a4SChris Mason */ 149809cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1499940100a4SChris Mason i = max(i + 1, next); 1500940100a4SChris Mason continue; 1501940100a4SChris Mason } 1502008873eaSLi Zefan 1503008873eaSLi Zefan if (!newer_than) { 150409cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 150509cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1506008873eaSLi Zefan cluster = min(cluster, max_cluster); 1507008873eaSLi Zefan } else { 1508008873eaSLi Zefan cluster = max_cluster; 1509008873eaSLi Zefan } 1510008873eaSLi Zefan 1511008873eaSLi Zefan if (i + cluster > ra_index) { 1512008873eaSLi Zefan ra_index = max(i, ra_index); 15130a52d108SDavid Sterba if (ra) 1514d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1515d3c0bab5SDavid Sterba file, ra_index, cluster); 1516e4826a5bSchandan ra_index += cluster; 1517008873eaSLi Zefan } 15184cb5300bSChris Mason 15195955102cSAl Viro inode_lock(inode); 1520eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1521eede2bf3SOmar Sandoval ret = -ETXTBSY; 1522eede2bf3SOmar Sandoval } else { 15231e2ef46dSDavid Sterba if (do_compress) 1524eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1525008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1526eede2bf3SOmar Sandoval } 1527ecb8bea8SLiu Bo if (ret < 0) { 15285955102cSAl Viro inode_unlock(inode); 15294cb5300bSChris Mason goto out_ra; 1530ecb8bea8SLiu Bo } 15314cb5300bSChris Mason 15324cb5300bSChris Mason defrag_count += ret; 1533d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15345955102cSAl Viro inode_unlock(inode); 15354cb5300bSChris Mason 15364cb5300bSChris Mason if (newer_than) { 15374cb5300bSChris Mason if (newer_off == (u64)-1) 15384cb5300bSChris Mason break; 15394cb5300bSChris Mason 1540e1f041e1SLiu Bo if (ret > 0) 1541e1f041e1SLiu Bo i += ret; 1542e1f041e1SLiu Bo 15434cb5300bSChris Mason newer_off = max(newer_off + 1, 154409cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15454cb5300bSChris Mason 1546ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1547ee22184bSByongho Lee &newer_off, SZ_64K); 15484cb5300bSChris Mason if (!ret) { 15494cb5300bSChris Mason range->start = newer_off; 155009cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15514cb5300bSChris Mason } else { 15524cb5300bSChris Mason break; 1553940100a4SChris Mason } 15544cb5300bSChris Mason } else { 1555008873eaSLi Zefan if (ret > 0) { 1556cbcc8326SLi Zefan i += ret; 155709cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1558008873eaSLi Zefan } else { 1559940100a4SChris Mason i++; 1560008873eaSLi Zefan last_len = 0; 1561008873eaSLi Zefan } 1562f46b5a66SChristoph Hellwig } 15634cb5300bSChris Mason } 1564f46b5a66SChristoph Hellwig 1565dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15661e701a32SChris Mason filemap_flush(inode->i_mapping); 1567dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1568dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1569dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1570dec8ef90SFilipe Manana } 15711e701a32SChris Mason 15721a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15730b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15745c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15755c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15761a419d85SLi Zefan } 15771a419d85SLi Zefan 157860ccf82fSDiego Calleja ret = defrag_count; 1579940100a4SChris Mason 15804cb5300bSChris Mason out_ra: 15811e2ef46dSDavid Sterba if (do_compress) { 15825955102cSAl Viro inode_lock(inode); 1583eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15845955102cSAl Viro inode_unlock(inode); 1585633085c7SFilipe David Borba Manana } 15864cb5300bSChris Mason if (!file) 15874cb5300bSChris Mason kfree(ra); 15884cb5300bSChris Mason kfree(pages); 1589940100a4SChris Mason return ret; 1590f46b5a66SChristoph Hellwig } 1591f46b5a66SChristoph Hellwig 1592198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 159376dda93cSYan, Zheng void __user *arg) 1594f46b5a66SChristoph Hellwig { 15950b246afaSJeff Mahoney struct inode *inode = file_inode(file); 15960b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1597f46b5a66SChristoph Hellwig u64 new_size; 1598f46b5a66SChristoph Hellwig u64 old_size; 1599f46b5a66SChristoph Hellwig u64 devid = 1; 16000b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1601f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1602f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1603f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1604f46b5a66SChristoph Hellwig char *sizestr; 16059a40f122SGui Hecheng char *retptr; 1606f46b5a66SChristoph Hellwig char *devstr = NULL; 1607f46b5a66SChristoph Hellwig int ret = 0; 1608f46b5a66SChristoph Hellwig int mod = 0; 1609f46b5a66SChristoph Hellwig 1610e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1611e441d54dSChris Mason return -EPERM; 1612e441d54dSChris Mason 1613198605a8SMiao Xie ret = mnt_want_write_file(file); 1614198605a8SMiao Xie if (ret) 1615198605a8SMiao Xie return ret; 1616198605a8SMiao Xie 1617171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 161897547676SMiao Xie mnt_drop_write_file(file); 1619e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1620c9e9f97bSIlya Dryomov } 1621c9e9f97bSIlya Dryomov 1622dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1623c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1624c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1625c9e9f97bSIlya Dryomov goto out; 1626c9e9f97bSIlya Dryomov } 16275516e595SMark Fasheh 16285516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1629f46b5a66SChristoph Hellwig 1630f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1631f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1632f46b5a66SChristoph Hellwig if (devstr) { 1633f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1634f46b5a66SChristoph Hellwig *devstr = '\0'; 1635f46b5a66SChristoph Hellwig devstr = vol_args->name; 163658dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 163758dfae63SZhangZhen if (ret) 163858dfae63SZhangZhen goto out_free; 1639dfd79829SMiao Xie if (!devid) { 1640dfd79829SMiao Xie ret = -EINVAL; 1641dfd79829SMiao Xie goto out_free; 1642dfd79829SMiao Xie } 16430b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1644f46b5a66SChristoph Hellwig } 1645dba60f3fSMiao Xie 164609ba3bc9SAnand Jain device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true); 1647f46b5a66SChristoph Hellwig if (!device) { 16480b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1649c1c9ff7cSGeert Uytterhoeven devid); 1650dfd79829SMiao Xie ret = -ENODEV; 1651c9e9f97bSIlya Dryomov goto out_free; 1652f46b5a66SChristoph Hellwig } 1653dba60f3fSMiao Xie 1654ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16550b246afaSJeff Mahoney btrfs_info(fs_info, 1656efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1657c1c9ff7cSGeert Uytterhoeven devid); 1658dfd79829SMiao Xie ret = -EPERM; 16594e42ae1bSLiu Bo goto out_free; 16604e42ae1bSLiu Bo } 16614e42ae1bSLiu Bo 1662f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1663f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1664f46b5a66SChristoph Hellwig else { 1665f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1666f46b5a66SChristoph Hellwig mod = -1; 1667f46b5a66SChristoph Hellwig sizestr++; 1668f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1669f46b5a66SChristoph Hellwig mod = 1; 1670f46b5a66SChristoph Hellwig sizestr++; 1671f46b5a66SChristoph Hellwig } 16729a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16739a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1674f46b5a66SChristoph Hellwig ret = -EINVAL; 1675c9e9f97bSIlya Dryomov goto out_free; 1676f46b5a66SChristoph Hellwig } 1677f46b5a66SChristoph Hellwig } 1678f46b5a66SChristoph Hellwig 1679401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1680dfd79829SMiao Xie ret = -EPERM; 168163a212abSStefan Behrens goto out_free; 168263a212abSStefan Behrens } 168363a212abSStefan Behrens 16847cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1685f46b5a66SChristoph Hellwig 1686f46b5a66SChristoph Hellwig if (mod < 0) { 1687f46b5a66SChristoph Hellwig if (new_size > old_size) { 1688f46b5a66SChristoph Hellwig ret = -EINVAL; 1689c9e9f97bSIlya Dryomov goto out_free; 1690f46b5a66SChristoph Hellwig } 1691f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1692f46b5a66SChristoph Hellwig } else if (mod > 0) { 1693eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1694902c68a4SGui Hecheng ret = -ERANGE; 1695eb8052e0SWenliang Fan goto out_free; 1696eb8052e0SWenliang Fan } 1697f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1698f46b5a66SChristoph Hellwig } 1699f46b5a66SChristoph Hellwig 1700ee22184bSByongho Lee if (new_size < SZ_256M) { 1701f46b5a66SChristoph Hellwig ret = -EINVAL; 1702c9e9f97bSIlya Dryomov goto out_free; 1703f46b5a66SChristoph Hellwig } 1704f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1705f46b5a66SChristoph Hellwig ret = -EFBIG; 1706c9e9f97bSIlya Dryomov goto out_free; 1707f46b5a66SChristoph Hellwig } 1708f46b5a66SChristoph Hellwig 170947f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1710f46b5a66SChristoph Hellwig 17110b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1712c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1713f46b5a66SChristoph Hellwig 1714f46b5a66SChristoph Hellwig if (new_size > old_size) { 1715a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 171698d5dc13STsutomu Itoh if (IS_ERR(trans)) { 171798d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1718c9e9f97bSIlya Dryomov goto out_free; 171998d5dc13STsutomu Itoh } 1720f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17213a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1722ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1723f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17240253f40eSjeff.liu } /* equal, nothing need to do */ 1725f46b5a66SChristoph Hellwig 1726c9e9f97bSIlya Dryomov out_free: 1727f46b5a66SChristoph Hellwig kfree(vol_args); 1728c9e9f97bSIlya Dryomov out: 1729171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 173018f39c41SIlya Dryomov mnt_drop_write_file(file); 1731f46b5a66SChristoph Hellwig return ret; 1732f46b5a66SChristoph Hellwig } 1733f46b5a66SChristoph Hellwig 173472fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 173552f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17366f72c7e2SArne Jansen u64 *transid, bool readonly, 17378696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1738f46b5a66SChristoph Hellwig { 1739f46b5a66SChristoph Hellwig int namelen; 17403de4586cSChris Mason int ret = 0; 1741f46b5a66SChristoph Hellwig 1742325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1743325c50e3SJeff Mahoney return -ENOTDIR; 1744325c50e3SJeff Mahoney 1745a874a63eSLiu Bo ret = mnt_want_write_file(file); 1746a874a63eSLiu Bo if (ret) 1747a874a63eSLiu Bo goto out; 1748a874a63eSLiu Bo 174972fd032eSSage Weil namelen = strlen(name); 175072fd032eSSage Weil if (strchr(name, '/')) { 1751f46b5a66SChristoph Hellwig ret = -EINVAL; 1752a874a63eSLiu Bo goto out_drop_write; 1753f46b5a66SChristoph Hellwig } 1754f46b5a66SChristoph Hellwig 175516780cabSChris Mason if (name[0] == '.' && 175616780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 175716780cabSChris Mason ret = -EEXIST; 1758a874a63eSLiu Bo goto out_drop_write; 175916780cabSChris Mason } 176016780cabSChris Mason 17613de4586cSChris Mason if (subvol) { 176272fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17636f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1764cb8e7090SChristoph Hellwig } else { 17652903ff01SAl Viro struct fd src = fdget(fd); 17663de4586cSChris Mason struct inode *src_inode; 17672903ff01SAl Viro if (!src.file) { 17683de4586cSChris Mason ret = -EINVAL; 1769a874a63eSLiu Bo goto out_drop_write; 17703de4586cSChris Mason } 17713de4586cSChris Mason 1772496ad9aaSAl Viro src_inode = file_inode(src.file); 1773496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1774c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1775efe120a0SFrank Holton "Snapshot src from another FS"); 177623ad5b17SKusanagi Kouichi ret = -EXDEV; 1777d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1778d0242061SDavid Sterba /* 1779d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1780d0242061SDavid Sterba * are limited to own subvolumes only 1781d0242061SDavid Sterba */ 1782d0242061SDavid Sterba ret = -EPERM; 1783ecd18815SAl Viro } else { 178472fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 178572fd032eSSage Weil BTRFS_I(src_inode)->root, 17866f72c7e2SArne Jansen transid, readonly, inherit); 1787ecd18815SAl Viro } 17882903ff01SAl Viro fdput(src); 1789cb8e7090SChristoph Hellwig } 1790a874a63eSLiu Bo out_drop_write: 1791a874a63eSLiu Bo mnt_drop_write_file(file); 1792f46b5a66SChristoph Hellwig out: 179372fd032eSSage Weil return ret; 179472fd032eSSage Weil } 179572fd032eSSage Weil 179672fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1797fa0d2b9bSLi Zefan void __user *arg, int subvol) 179872fd032eSSage Weil { 1799fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 180072fd032eSSage Weil int ret; 180172fd032eSSage Weil 1802325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1803325c50e3SJeff Mahoney return -ENOTDIR; 1804325c50e3SJeff Mahoney 1805fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1806fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1807fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1808fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1809fa0d2b9bSLi Zefan 1810fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1811b83cc969SLi Zefan vol_args->fd, subvol, 18126f72c7e2SArne Jansen NULL, false, NULL); 1813fa0d2b9bSLi Zefan 1814fa0d2b9bSLi Zefan kfree(vol_args); 1815fa0d2b9bSLi Zefan return ret; 1816fa0d2b9bSLi Zefan } 1817fa0d2b9bSLi Zefan 1818fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1819fa0d2b9bSLi Zefan void __user *arg, int subvol) 1820fa0d2b9bSLi Zefan { 1821fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1822fa0d2b9bSLi Zefan int ret; 1823fdfb1e4fSLi Zefan u64 transid = 0; 1824fdfb1e4fSLi Zefan u64 *ptr = NULL; 1825b83cc969SLi Zefan bool readonly = false; 18266f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 182772fd032eSSage Weil 1828325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1829325c50e3SJeff Mahoney return -ENOTDIR; 1830325c50e3SJeff Mahoney 1831fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1832fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1833fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1834fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1835fdfb1e4fSLi Zefan 1836b83cc969SLi Zefan if (vol_args->flags & 18376f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18386f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1839b83cc969SLi Zefan ret = -EOPNOTSUPP; 1840c47ca32dSDan Carpenter goto free_args; 1841fdfb1e4fSLi Zefan } 1842fdfb1e4fSLi Zefan 1843ebc87351SNikolay Borisov if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1844ebc87351SNikolay Borisov struct inode *inode = file_inode(file); 1845ebc87351SNikolay Borisov struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1846ebc87351SNikolay Borisov 1847ebc87351SNikolay Borisov btrfs_warn(fs_info, 1848ebc87351SNikolay Borisov "SNAP_CREATE_V2 ioctl with CREATE_ASYNC is deprecated and will be removed in kernel 5.7"); 1849ebc87351SNikolay Borisov 1850fdfb1e4fSLi Zefan ptr = &transid; 1851ebc87351SNikolay Borisov } 1852b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1853b83cc969SLi Zefan readonly = true; 18546f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 185509cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18566f72c7e2SArne Jansen ret = -EINVAL; 1857c47ca32dSDan Carpenter goto free_args; 18586f72c7e2SArne Jansen } 18596f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18606f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18616f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1862c47ca32dSDan Carpenter goto free_args; 18636f72c7e2SArne Jansen } 18646f72c7e2SArne Jansen } 186575eaa0e2SSage Weil 1866fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18676f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18688696c533SMiao Xie readonly, inherit); 1869c47ca32dSDan Carpenter if (ret) 1870c47ca32dSDan Carpenter goto free_inherit; 187175eaa0e2SSage Weil 1872c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1873fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1874c47ca32dSDan Carpenter transid), 1875c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 187675eaa0e2SSage Weil ret = -EFAULT; 1877c47ca32dSDan Carpenter 1878c47ca32dSDan Carpenter free_inherit: 18796f72c7e2SArne Jansen kfree(inherit); 1880c47ca32dSDan Carpenter free_args: 1881c47ca32dSDan Carpenter kfree(vol_args); 1882f46b5a66SChristoph Hellwig return ret; 1883f46b5a66SChristoph Hellwig } 1884f46b5a66SChristoph Hellwig 18850caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18860caa102dSLi Zefan void __user *arg) 18870caa102dSLi Zefan { 1888496ad9aaSAl Viro struct inode *inode = file_inode(file); 18890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18900caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18910caa102dSLi Zefan int ret = 0; 18920caa102dSLi Zefan u64 flags = 0; 18930caa102dSLi Zefan 18944a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18950caa102dSLi Zefan return -EINVAL; 18960caa102dSLi Zefan 18970b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 18980caa102dSLi Zefan if (btrfs_root_readonly(root)) 18990caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 19000b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 19010caa102dSLi Zefan 19020caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 19030caa102dSLi Zefan ret = -EFAULT; 19040caa102dSLi Zefan 19050caa102dSLi Zefan return ret; 19060caa102dSLi Zefan } 19070caa102dSLi Zefan 19080caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19090caa102dSLi Zefan void __user *arg) 19100caa102dSLi Zefan { 1911496ad9aaSAl Viro struct inode *inode = file_inode(file); 19120b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19130caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19140caa102dSLi Zefan struct btrfs_trans_handle *trans; 19150caa102dSLi Zefan u64 root_flags; 19160caa102dSLi Zefan u64 flags; 19170caa102dSLi Zefan int ret = 0; 19180caa102dSLi Zefan 1919bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1920bd60ea0fSDavid Sterba return -EPERM; 1921bd60ea0fSDavid Sterba 1922b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1923b9ca0664SLiu Bo if (ret) 1924b9ca0664SLiu Bo goto out; 19250caa102dSLi Zefan 19264a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1927b9ca0664SLiu Bo ret = -EINVAL; 1928b9ca0664SLiu Bo goto out_drop_write; 1929b9ca0664SLiu Bo } 19300caa102dSLi Zefan 1931b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1932b9ca0664SLiu Bo ret = -EFAULT; 1933b9ca0664SLiu Bo goto out_drop_write; 1934b9ca0664SLiu Bo } 19350caa102dSLi Zefan 1936b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1937b9ca0664SLiu Bo ret = -EINVAL; 1938b9ca0664SLiu Bo goto out_drop_write; 1939b9ca0664SLiu Bo } 19400caa102dSLi Zefan 1941b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1942b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1943b9ca0664SLiu Bo goto out_drop_write; 1944b9ca0664SLiu Bo } 19450caa102dSLi Zefan 19460b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19470caa102dSLi Zefan 19480caa102dSLi Zefan /* nothing to do */ 19490caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1950b9ca0664SLiu Bo goto out_drop_sem; 19510caa102dSLi Zefan 19520caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19532c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19540caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19550caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19562c686537SDavid Sterba } else { 19572c686537SDavid Sterba /* 19582c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19592c686537SDavid Sterba * send 19602c686537SDavid Sterba */ 19612c686537SDavid Sterba spin_lock(&root->root_item_lock); 19622c686537SDavid Sterba if (root->send_in_progress == 0) { 19630caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19640caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19652c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19662c686537SDavid Sterba } else { 19672c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19680b246afaSJeff Mahoney btrfs_warn(fs_info, 19692c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19702c686537SDavid Sterba root->root_key.objectid); 19712c686537SDavid Sterba ret = -EPERM; 19722c686537SDavid Sterba goto out_drop_sem; 19732c686537SDavid Sterba } 19742c686537SDavid Sterba } 19750caa102dSLi Zefan 19760caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19770caa102dSLi Zefan if (IS_ERR(trans)) { 19780caa102dSLi Zefan ret = PTR_ERR(trans); 19790caa102dSLi Zefan goto out_reset; 19800caa102dSLi Zefan } 19810caa102dSLi Zefan 19820b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19830caa102dSLi Zefan &root->root_key, &root->root_item); 19849417ebc8SNikolay Borisov if (ret < 0) { 19859417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19869417ebc8SNikolay Borisov goto out_reset; 19879417ebc8SNikolay Borisov } 19880caa102dSLi Zefan 19899417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19909417ebc8SNikolay Borisov 19910caa102dSLi Zefan out_reset: 19920caa102dSLi Zefan if (ret) 19930caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1994b9ca0664SLiu Bo out_drop_sem: 19950b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1996b9ca0664SLiu Bo out_drop_write: 1997b9ca0664SLiu Bo mnt_drop_write_file(file); 1998b9ca0664SLiu Bo out: 19990caa102dSLi Zefan return ret; 20000caa102dSLi Zefan } 20010caa102dSLi Zefan 2002ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2003ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2004ac8e9819SChris Mason { 2005abc6e134SChris Mason struct btrfs_key test; 2006abc6e134SChris Mason int ret; 2007abc6e134SChris Mason 2008abc6e134SChris Mason test.objectid = sk->min_objectid; 2009abc6e134SChris Mason test.type = sk->min_type; 2010abc6e134SChris Mason test.offset = sk->min_offset; 2011abc6e134SChris Mason 2012abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2013abc6e134SChris Mason if (ret < 0) 2014ac8e9819SChris Mason return 0; 2015abc6e134SChris Mason 2016abc6e134SChris Mason test.objectid = sk->max_objectid; 2017abc6e134SChris Mason test.type = sk->max_type; 2018abc6e134SChris Mason test.offset = sk->max_offset; 2019abc6e134SChris Mason 2020abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2021abc6e134SChris Mason if (ret > 0) 2022ac8e9819SChris Mason return 0; 2023ac8e9819SChris Mason return 1; 2024ac8e9819SChris Mason } 2025ac8e9819SChris Mason 2026df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2027ac8e9819SChris Mason struct btrfs_key *key, 2028ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20299b6e817dSGerhard Heift size_t *buf_size, 2030ba346b35SGerhard Heift char __user *ubuf, 2031ac8e9819SChris Mason unsigned long *sk_offset, 2032ac8e9819SChris Mason int *num_found) 2033ac8e9819SChris Mason { 2034ac8e9819SChris Mason u64 found_transid; 2035ac8e9819SChris Mason struct extent_buffer *leaf; 2036ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2037dd81d459SNaohiro Aota struct btrfs_key test; 2038ac8e9819SChris Mason unsigned long item_off; 2039ac8e9819SChris Mason unsigned long item_len; 2040ac8e9819SChris Mason int nritems; 2041ac8e9819SChris Mason int i; 2042ac8e9819SChris Mason int slot; 2043ac8e9819SChris Mason int ret = 0; 2044ac8e9819SChris Mason 2045ac8e9819SChris Mason leaf = path->nodes[0]; 2046ac8e9819SChris Mason slot = path->slots[0]; 2047ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2048ac8e9819SChris Mason 2049ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2050ac8e9819SChris Mason i = nritems; 2051ac8e9819SChris Mason goto advance_key; 2052ac8e9819SChris Mason } 2053ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2054ac8e9819SChris Mason 2055ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2056ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2057ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2058ac8e9819SChris Mason 205903b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 206003b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 206103b71c6cSGabriel de Perthuis continue; 206203b71c6cSGabriel de Perthuis 20639b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20648f5f6178SGerhard Heift if (*num_found) { 2065ac8e9819SChris Mason ret = 1; 20668f5f6178SGerhard Heift goto out; 20678f5f6178SGerhard Heift } 20688f5f6178SGerhard Heift 20698f5f6178SGerhard Heift /* 20708f5f6178SGerhard Heift * return one empty item back for v1, which does not 20718f5f6178SGerhard Heift * handle -EOVERFLOW 20728f5f6178SGerhard Heift */ 20738f5f6178SGerhard Heift 20749b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2075ac8e9819SChris Mason item_len = 0; 20768f5f6178SGerhard Heift ret = -EOVERFLOW; 20778f5f6178SGerhard Heift } 2078ac8e9819SChris Mason 20799b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2080ac8e9819SChris Mason ret = 1; 208125c9bc2eSGerhard Heift goto out; 2082ac8e9819SChris Mason } 2083ac8e9819SChris Mason 2084ac8e9819SChris Mason sh.objectid = key->objectid; 2085ac8e9819SChris Mason sh.offset = key->offset; 2086ac8e9819SChris Mason sh.type = key->type; 2087ac8e9819SChris Mason sh.len = item_len; 2088ac8e9819SChris Mason sh.transid = found_transid; 2089ac8e9819SChris Mason 2090ac8e9819SChris Mason /* copy search result header */ 2091ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2092ba346b35SGerhard Heift ret = -EFAULT; 2093ba346b35SGerhard Heift goto out; 2094ba346b35SGerhard Heift } 2095ba346b35SGerhard Heift 2096ac8e9819SChris Mason *sk_offset += sizeof(sh); 2097ac8e9819SChris Mason 2098ac8e9819SChris Mason if (item_len) { 2099ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2100ac8e9819SChris Mason /* copy the item */ 2101ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2102ba346b35SGerhard Heift item_off, item_len)) { 2103ba346b35SGerhard Heift ret = -EFAULT; 2104ba346b35SGerhard Heift goto out; 2105ba346b35SGerhard Heift } 2106ba346b35SGerhard Heift 2107ac8e9819SChris Mason *sk_offset += item_len; 2108ac8e9819SChris Mason } 2109e2156867SHugo Mills (*num_found)++; 2110ac8e9819SChris Mason 21118f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21128f5f6178SGerhard Heift goto out; 21138f5f6178SGerhard Heift 211425c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 211525c9bc2eSGerhard Heift ret = 1; 211625c9bc2eSGerhard Heift goto out; 211725c9bc2eSGerhard Heift } 2118ac8e9819SChris Mason } 2119ac8e9819SChris Mason advance_key: 2120ac8e9819SChris Mason ret = 0; 2121dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2122dd81d459SNaohiro Aota test.type = sk->max_type; 2123dd81d459SNaohiro Aota test.offset = sk->max_offset; 2124dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2125dd81d459SNaohiro Aota ret = 1; 2126dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2127abc6e134SChris Mason key->offset++; 2128dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2129abc6e134SChris Mason key->offset = 0; 2130abc6e134SChris Mason key->type++; 2131dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2132abc6e134SChris Mason key->offset = 0; 2133abc6e134SChris Mason key->type = 0; 2134abc6e134SChris Mason key->objectid++; 2135abc6e134SChris Mason } else 2136abc6e134SChris Mason ret = 1; 213725c9bc2eSGerhard Heift out: 2138ba346b35SGerhard Heift /* 2139ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2140ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2141ba346b35SGerhard Heift * * all items were found 2142ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2143ba346b35SGerhard Heift * leaf 2144ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2145ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2146ba346b35SGerhard Heift */ 2147ac8e9819SChris Mason return ret; 2148ac8e9819SChris Mason } 2149ac8e9819SChris Mason 2150ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 215112544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21529b6e817dSGerhard Heift size_t *buf_size, 2153ba346b35SGerhard Heift char __user *ubuf) 2154ac8e9819SChris Mason { 21550b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2156ac8e9819SChris Mason struct btrfs_root *root; 2157ac8e9819SChris Mason struct btrfs_key key; 2158ac8e9819SChris Mason struct btrfs_path *path; 2159ac8e9819SChris Mason int ret; 2160ac8e9819SChris Mason int num_found = 0; 2161ac8e9819SChris Mason unsigned long sk_offset = 0; 2162ac8e9819SChris Mason 21639b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21649b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 216512544442SGerhard Heift return -EOVERFLOW; 21669b6e817dSGerhard Heift } 216712544442SGerhard Heift 2168ac8e9819SChris Mason path = btrfs_alloc_path(); 2169ac8e9819SChris Mason if (!path) 2170ac8e9819SChris Mason return -ENOMEM; 2171ac8e9819SChris Mason 2172ac8e9819SChris Mason if (sk->tree_id == 0) { 2173ac8e9819SChris Mason /* search the root of the inode that was passed */ 2174ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2175ac8e9819SChris Mason } else { 2176ac8e9819SChris Mason key.objectid = sk->tree_id; 2177ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2178ac8e9819SChris Mason key.offset = (u64)-1; 2179ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2180ac8e9819SChris Mason if (IS_ERR(root)) { 2181ac8e9819SChris Mason btrfs_free_path(path); 2182ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2183ac8e9819SChris Mason } 2184ac8e9819SChris Mason } 2185ac8e9819SChris Mason 2186ac8e9819SChris Mason key.objectid = sk->min_objectid; 2187ac8e9819SChris Mason key.type = sk->min_type; 2188ac8e9819SChris Mason key.offset = sk->min_offset; 2189ac8e9819SChris Mason 2190ac8e9819SChris Mason while (1) { 21916174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2192ac8e9819SChris Mason if (ret != 0) { 2193ac8e9819SChris Mason if (ret > 0) 2194ac8e9819SChris Mason ret = 0; 2195ac8e9819SChris Mason goto err; 2196ac8e9819SChris Mason } 2197df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2198ac8e9819SChris Mason &sk_offset, &num_found); 2199b3b4aa74SDavid Sterba btrfs_release_path(path); 220025c9bc2eSGerhard Heift if (ret) 2201ac8e9819SChris Mason break; 2202ac8e9819SChris Mason 2203ac8e9819SChris Mason } 22048f5f6178SGerhard Heift if (ret > 0) 2205ac8e9819SChris Mason ret = 0; 2206ac8e9819SChris Mason err: 2207ac8e9819SChris Mason sk->nr_items = num_found; 2208ac8e9819SChris Mason btrfs_free_path(path); 2209ac8e9819SChris Mason return ret; 2210ac8e9819SChris Mason } 2211ac8e9819SChris Mason 2212ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2213ac8e9819SChris Mason void __user *argp) 2214ac8e9819SChris Mason { 2215ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2216ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2217ac8e9819SChris Mason struct inode *inode; 2218ac8e9819SChris Mason int ret; 22199b6e817dSGerhard Heift size_t buf_size; 2220ac8e9819SChris Mason 2221ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2222ac8e9819SChris Mason return -EPERM; 2223ac8e9819SChris Mason 2224ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2225ac8e9819SChris Mason 2226ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2227ba346b35SGerhard Heift return -EFAULT; 2228ba346b35SGerhard Heift 2229ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2230ac8e9819SChris Mason 2231496ad9aaSAl Viro inode = file_inode(file); 2232ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22338f5f6178SGerhard Heift 22348f5f6178SGerhard Heift /* 22358f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22368f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22378f5f6178SGerhard Heift */ 22388f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22398f5f6178SGerhard Heift ret = 0; 22408f5f6178SGerhard Heift 2241ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2242ac8e9819SChris Mason ret = -EFAULT; 2243ac8e9819SChris Mason return ret; 2244ac8e9819SChris Mason } 2245ac8e9819SChris Mason 2246cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2247cc68a8a5SGerhard Heift void __user *argp) 2248cc68a8a5SGerhard Heift { 2249cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2250cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2251cc68a8a5SGerhard Heift struct inode *inode; 2252cc68a8a5SGerhard Heift int ret; 2253cc68a8a5SGerhard Heift size_t buf_size; 2254ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2255cc68a8a5SGerhard Heift 2256cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2257cc68a8a5SGerhard Heift return -EPERM; 2258cc68a8a5SGerhard Heift 2259cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2260cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2261cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2262cc68a8a5SGerhard Heift return -EFAULT; 2263cc68a8a5SGerhard Heift 2264cc68a8a5SGerhard Heift buf_size = args.buf_size; 2265cc68a8a5SGerhard Heift 2266cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2267cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2268cc68a8a5SGerhard Heift buf_size = buf_limit; 2269cc68a8a5SGerhard Heift 2270cc68a8a5SGerhard Heift inode = file_inode(file); 2271cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2272718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2273cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2274cc68a8a5SGerhard Heift ret = -EFAULT; 2275cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2276cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2277cc68a8a5SGerhard Heift ret = -EFAULT; 2278cc68a8a5SGerhard Heift 227976dda93cSYan, Zheng return ret; 228076dda93cSYan, Zheng } 228176dda93cSYan, Zheng 228298d377a0STARUISI Hiroaki /* 2283ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2284ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 228598d377a0STARUISI Hiroaki */ 228698d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 228798d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 228898d377a0STARUISI Hiroaki { 228998d377a0STARUISI Hiroaki struct btrfs_root *root; 229098d377a0STARUISI Hiroaki struct btrfs_key key; 2291ac8e9819SChris Mason char *ptr; 229298d377a0STARUISI Hiroaki int ret = -1; 229398d377a0STARUISI Hiroaki int slot; 229498d377a0STARUISI Hiroaki int len; 229598d377a0STARUISI Hiroaki int total_len = 0; 229698d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 229798d377a0STARUISI Hiroaki struct extent_buffer *l; 229898d377a0STARUISI Hiroaki struct btrfs_path *path; 229998d377a0STARUISI Hiroaki 230098d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 230198d377a0STARUISI Hiroaki name[0]='\0'; 230298d377a0STARUISI Hiroaki return 0; 230398d377a0STARUISI Hiroaki } 230498d377a0STARUISI Hiroaki 230598d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 230698d377a0STARUISI Hiroaki if (!path) 230798d377a0STARUISI Hiroaki return -ENOMEM; 230898d377a0STARUISI Hiroaki 2309c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 231098d377a0STARUISI Hiroaki 231198d377a0STARUISI Hiroaki key.objectid = tree_id; 231298d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 231398d377a0STARUISI Hiroaki key.offset = (u64)-1; 231498d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 231598d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2316ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 23178ad6fcabSChris Mason goto out; 231898d377a0STARUISI Hiroaki } 231998d377a0STARUISI Hiroaki 232098d377a0STARUISI Hiroaki key.objectid = dirid; 232198d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23228ad6fcabSChris Mason key.offset = (u64)-1; 232398d377a0STARUISI Hiroaki 232498d377a0STARUISI Hiroaki while (1) { 232598d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 232698d377a0STARUISI Hiroaki if (ret < 0) 232798d377a0STARUISI Hiroaki goto out; 232818674c6cSFilipe David Borba Manana else if (ret > 0) { 232918674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 233018674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 233118674c6cSFilipe David Borba Manana if (ret < 0) 233218674c6cSFilipe David Borba Manana goto out; 233318674c6cSFilipe David Borba Manana else if (ret > 0) { 2334ac8e9819SChris Mason ret = -ENOENT; 233598d377a0STARUISI Hiroaki goto out; 2336ac8e9819SChris Mason } 233718674c6cSFilipe David Borba Manana } 233818674c6cSFilipe David Borba Manana 233918674c6cSFilipe David Borba Manana l = path->nodes[0]; 234018674c6cSFilipe David Borba Manana slot = path->slots[0]; 234118674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 234298d377a0STARUISI Hiroaki 234398d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 234498d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 234598d377a0STARUISI Hiroaki ptr -= len + 1; 234698d377a0STARUISI Hiroaki total_len += len + 1; 2347a696cf35SFilipe David Borba Manana if (ptr < name) { 2348a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 234998d377a0STARUISI Hiroaki goto out; 2350a696cf35SFilipe David Borba Manana } 235198d377a0STARUISI Hiroaki 235298d377a0STARUISI Hiroaki *(ptr + len) = '/'; 235398d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 235498d377a0STARUISI Hiroaki 235598d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 235698d377a0STARUISI Hiroaki break; 235798d377a0STARUISI Hiroaki 2358b3b4aa74SDavid Sterba btrfs_release_path(path); 235998d377a0STARUISI Hiroaki key.objectid = key.offset; 23608ad6fcabSChris Mason key.offset = (u64)-1; 236198d377a0STARUISI Hiroaki dirid = key.objectid; 236298d377a0STARUISI Hiroaki } 236377906a50SLi Zefan memmove(name, ptr, total_len); 236498d377a0STARUISI Hiroaki name[total_len] = '\0'; 236598d377a0STARUISI Hiroaki ret = 0; 236698d377a0STARUISI Hiroaki out: 236798d377a0STARUISI Hiroaki btrfs_free_path(path); 2368ac8e9819SChris Mason return ret; 2369ac8e9819SChris Mason } 2370ac8e9819SChris Mason 237123d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 237223d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 237323d0b79dSTomohiro Misono { 237423d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 237523d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 237623d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 237723d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 237823d0b79dSTomohiro Misono u64 dirid = args->dirid; 237923d0b79dSTomohiro Misono unsigned long item_off; 238023d0b79dSTomohiro Misono unsigned long item_len; 238123d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 238223d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 238323d0b79dSTomohiro Misono struct btrfs_root *root; 238423d0b79dSTomohiro Misono struct btrfs_path *path; 238523d0b79dSTomohiro Misono struct btrfs_key key, key2; 238623d0b79dSTomohiro Misono struct extent_buffer *leaf; 238723d0b79dSTomohiro Misono struct inode *temp_inode; 238823d0b79dSTomohiro Misono char *ptr; 238923d0b79dSTomohiro Misono int slot; 239023d0b79dSTomohiro Misono int len; 239123d0b79dSTomohiro Misono int total_len = 0; 239223d0b79dSTomohiro Misono int ret; 239323d0b79dSTomohiro Misono 239423d0b79dSTomohiro Misono path = btrfs_alloc_path(); 239523d0b79dSTomohiro Misono if (!path) 239623d0b79dSTomohiro Misono return -ENOMEM; 239723d0b79dSTomohiro Misono 239823d0b79dSTomohiro Misono /* 239923d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 240023d0b79dSTomohiro Misono * construct the path in from the bottom up. 240123d0b79dSTomohiro Misono */ 240223d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 240323d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 240423d0b79dSTomohiro Misono 240523d0b79dSTomohiro Misono key.objectid = treeid; 240623d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 240723d0b79dSTomohiro Misono key.offset = (u64)-1; 240823d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 240923d0b79dSTomohiro Misono if (IS_ERR(root)) { 241023d0b79dSTomohiro Misono ret = PTR_ERR(root); 241123d0b79dSTomohiro Misono goto out; 241223d0b79dSTomohiro Misono } 241323d0b79dSTomohiro Misono 241423d0b79dSTomohiro Misono key.objectid = dirid; 241523d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 241623d0b79dSTomohiro Misono key.offset = (u64)-1; 241723d0b79dSTomohiro Misono while (1) { 241823d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 241923d0b79dSTomohiro Misono if (ret < 0) { 242023d0b79dSTomohiro Misono goto out; 242123d0b79dSTomohiro Misono } else if (ret > 0) { 242223d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 242323d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 242423d0b79dSTomohiro Misono if (ret < 0) { 242523d0b79dSTomohiro Misono goto out; 242623d0b79dSTomohiro Misono } else if (ret > 0) { 242723d0b79dSTomohiro Misono ret = -ENOENT; 242823d0b79dSTomohiro Misono goto out; 242923d0b79dSTomohiro Misono } 243023d0b79dSTomohiro Misono } 243123d0b79dSTomohiro Misono 243223d0b79dSTomohiro Misono leaf = path->nodes[0]; 243323d0b79dSTomohiro Misono slot = path->slots[0]; 243423d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 243523d0b79dSTomohiro Misono 243623d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 243723d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 243823d0b79dSTomohiro Misono ptr -= len + 1; 243923d0b79dSTomohiro Misono total_len += len + 1; 244023d0b79dSTomohiro Misono if (ptr < args->path) { 244123d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 244223d0b79dSTomohiro Misono goto out; 244323d0b79dSTomohiro Misono } 244423d0b79dSTomohiro Misono 244523d0b79dSTomohiro Misono *(ptr + len) = '/'; 244623d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 244723d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 244823d0b79dSTomohiro Misono 244923d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 245023d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 245123d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 245223d0b79dSTomohiro Misono if (ret < 0) { 245323d0b79dSTomohiro Misono goto out; 245423d0b79dSTomohiro Misono } else if (ret > 0) { 245523d0b79dSTomohiro Misono ret = -ENOENT; 245623d0b79dSTomohiro Misono goto out; 245723d0b79dSTomohiro Misono } 245823d0b79dSTomohiro Misono 245923d0b79dSTomohiro Misono leaf = path->nodes[0]; 246023d0b79dSTomohiro Misono slot = path->slots[0]; 246123d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 246223d0b79dSTomohiro Misono if (key2.objectid != dirid) { 246323d0b79dSTomohiro Misono ret = -ENOENT; 246423d0b79dSTomohiro Misono goto out; 246523d0b79dSTomohiro Misono } 246623d0b79dSTomohiro Misono 246723d0b79dSTomohiro Misono temp_inode = btrfs_iget(sb, &key2, root, NULL); 24683ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24693ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24703ca57bd6SMisono Tomohiro goto out; 24713ca57bd6SMisono Tomohiro } 247223d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 247323d0b79dSTomohiro Misono iput(temp_inode); 247423d0b79dSTomohiro Misono if (ret) { 247523d0b79dSTomohiro Misono ret = -EACCES; 247623d0b79dSTomohiro Misono goto out; 247723d0b79dSTomohiro Misono } 247823d0b79dSTomohiro Misono 247923d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 248023d0b79dSTomohiro Misono break; 248123d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 248223d0b79dSTomohiro Misono ret = -EACCES; 248323d0b79dSTomohiro Misono goto out; 248423d0b79dSTomohiro Misono } 248523d0b79dSTomohiro Misono 248623d0b79dSTomohiro Misono btrfs_release_path(path); 248723d0b79dSTomohiro Misono key.objectid = key.offset; 248823d0b79dSTomohiro Misono key.offset = (u64)-1; 248923d0b79dSTomohiro Misono dirid = key.objectid; 249023d0b79dSTomohiro Misono } 249123d0b79dSTomohiro Misono 249223d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 249323d0b79dSTomohiro Misono args->path[total_len] = '\0'; 249423d0b79dSTomohiro Misono btrfs_release_path(path); 249523d0b79dSTomohiro Misono } 249623d0b79dSTomohiro Misono 249723d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 249823d0b79dSTomohiro Misono root = fs_info->tree_root; 249923d0b79dSTomohiro Misono key.objectid = treeid; 250023d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 250123d0b79dSTomohiro Misono key.offset = args->treeid; 250223d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 250323d0b79dSTomohiro Misono if (ret < 0) { 250423d0b79dSTomohiro Misono goto out; 250523d0b79dSTomohiro Misono } else if (ret > 0) { 250623d0b79dSTomohiro Misono ret = -ENOENT; 250723d0b79dSTomohiro Misono goto out; 250823d0b79dSTomohiro Misono } 250923d0b79dSTomohiro Misono 251023d0b79dSTomohiro Misono leaf = path->nodes[0]; 251123d0b79dSTomohiro Misono slot = path->slots[0]; 251223d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 251323d0b79dSTomohiro Misono 251423d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 251523d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 251623d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 251723d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 251823d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 251923d0b79dSTomohiro Misono ret = -EINVAL; 252023d0b79dSTomohiro Misono goto out; 252123d0b79dSTomohiro Misono } 252223d0b79dSTomohiro Misono 252323d0b79dSTomohiro Misono /* Copy subvolume's name */ 252423d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 252523d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 252623d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 252723d0b79dSTomohiro Misono args->name[item_len] = 0; 252823d0b79dSTomohiro Misono 252923d0b79dSTomohiro Misono out: 253023d0b79dSTomohiro Misono btrfs_free_path(path); 253123d0b79dSTomohiro Misono return ret; 253223d0b79dSTomohiro Misono } 253323d0b79dSTomohiro Misono 2534ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2535ac8e9819SChris Mason void __user *argp) 2536ac8e9819SChris Mason { 2537ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2538ac8e9819SChris Mason struct inode *inode; 253901b810b8SDavid Sterba int ret = 0; 2540ac8e9819SChris Mason 25412354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25422354d08fSJulia Lawall if (IS_ERR(args)) 25432354d08fSJulia Lawall return PTR_ERR(args); 2544c2b96929SDan Carpenter 2545496ad9aaSAl Viro inode = file_inode(file); 2546ac8e9819SChris Mason 254701b810b8SDavid Sterba /* 254801b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 254901b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 255001b810b8SDavid Sterba */ 25511b53ac4dSChris Mason if (args->treeid == 0) 25521b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25531b53ac4dSChris Mason 255401b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 255501b810b8SDavid Sterba args->name[0] = 0; 255601b810b8SDavid Sterba goto out; 255701b810b8SDavid Sterba } 255801b810b8SDavid Sterba 255901b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 256001b810b8SDavid Sterba ret = -EPERM; 256101b810b8SDavid Sterba goto out; 256201b810b8SDavid Sterba } 256301b810b8SDavid Sterba 2564ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2565ac8e9819SChris Mason args->treeid, args->objectid, 2566ac8e9819SChris Mason args->name); 2567ac8e9819SChris Mason 256801b810b8SDavid Sterba out: 2569ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2570ac8e9819SChris Mason ret = -EFAULT; 2571ac8e9819SChris Mason 2572ac8e9819SChris Mason kfree(args); 257398d377a0STARUISI Hiroaki return ret; 257498d377a0STARUISI Hiroaki } 257598d377a0STARUISI Hiroaki 257623d0b79dSTomohiro Misono /* 257723d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 257823d0b79dSTomohiro Misono * 257923d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 258023d0b79dSTomohiro Misono * 258123d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 258223d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 258323d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 258423d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 258523d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 258623d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 258723d0b79dSTomohiro Misono */ 258823d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 258923d0b79dSTomohiro Misono { 259023d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 259123d0b79dSTomohiro Misono struct inode *inode; 259223d0b79dSTomohiro Misono int ret; 259323d0b79dSTomohiro Misono 259423d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 259523d0b79dSTomohiro Misono if (IS_ERR(args)) 259623d0b79dSTomohiro Misono return PTR_ERR(args); 259723d0b79dSTomohiro Misono 259823d0b79dSTomohiro Misono inode = file_inode(file); 259923d0b79dSTomohiro Misono 260023d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 260123d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 260223d0b79dSTomohiro Misono /* 260323d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 260423d0b79dSTomohiro Misono * called 260523d0b79dSTomohiro Misono */ 260623d0b79dSTomohiro Misono kfree(args); 260723d0b79dSTomohiro Misono return -EACCES; 260823d0b79dSTomohiro Misono } 260923d0b79dSTomohiro Misono 261023d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 261123d0b79dSTomohiro Misono 261223d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 261323d0b79dSTomohiro Misono ret = -EFAULT; 261423d0b79dSTomohiro Misono 261523d0b79dSTomohiro Misono kfree(args); 261623d0b79dSTomohiro Misono return ret; 261723d0b79dSTomohiro Misono } 261823d0b79dSTomohiro Misono 2619b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2620b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2621b64ec075STomohiro Misono { 2622b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2623b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2624b64ec075STomohiro Misono struct btrfs_root *root; 2625b64ec075STomohiro Misono struct btrfs_path *path; 2626b64ec075STomohiro Misono struct btrfs_key key; 2627b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2628b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2629b64ec075STomohiro Misono struct extent_buffer *leaf; 2630b64ec075STomohiro Misono unsigned long item_off; 2631b64ec075STomohiro Misono unsigned long item_len; 2632b64ec075STomohiro Misono struct inode *inode; 2633b64ec075STomohiro Misono int slot; 2634b64ec075STomohiro Misono int ret = 0; 2635b64ec075STomohiro Misono 2636b64ec075STomohiro Misono path = btrfs_alloc_path(); 2637b64ec075STomohiro Misono if (!path) 2638b64ec075STomohiro Misono return -ENOMEM; 2639b64ec075STomohiro Misono 2640b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2641b64ec075STomohiro Misono if (!subvol_info) { 2642b64ec075STomohiro Misono btrfs_free_path(path); 2643b64ec075STomohiro Misono return -ENOMEM; 2644b64ec075STomohiro Misono } 2645b64ec075STomohiro Misono 2646b64ec075STomohiro Misono inode = file_inode(file); 2647b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2648b64ec075STomohiro Misono 2649b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2650b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2651b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2652b64ec075STomohiro Misono key.offset = (u64)-1; 2653b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2654b64ec075STomohiro Misono if (IS_ERR(root)) { 2655b64ec075STomohiro Misono ret = PTR_ERR(root); 2656b64ec075STomohiro Misono goto out; 2657b64ec075STomohiro Misono } 2658b64ec075STomohiro Misono root_item = &root->root_item; 2659b64ec075STomohiro Misono 2660b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2661b64ec075STomohiro Misono 2662b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2663b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2664b64ec075STomohiro Misono 2665b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2666b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2667b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2668b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2669b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2670b64ec075STomohiro Misono 2671b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2672b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2673b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2674b64ec075STomohiro Misono 2675b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2676b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2677b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2678b64ec075STomohiro Misono 2679b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2680b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2681b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2682b64ec075STomohiro Misono 2683b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2684b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2685b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2686b64ec075STomohiro Misono 2687b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2688b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2689b64ec075STomohiro Misono root = fs_info->tree_root; 2690b64ec075STomohiro Misono 2691b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2692b64ec075STomohiro Misono key.offset = 0; 2693b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2694b64ec075STomohiro Misono if (ret < 0) { 2695b64ec075STomohiro Misono goto out; 2696b64ec075STomohiro Misono } else if (path->slots[0] >= 2697b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2698b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2699b64ec075STomohiro Misono if (ret < 0) { 2700b64ec075STomohiro Misono goto out; 2701b64ec075STomohiro Misono } else if (ret > 0) { 2702b64ec075STomohiro Misono ret = -EUCLEAN; 2703b64ec075STomohiro Misono goto out; 2704b64ec075STomohiro Misono } 2705b64ec075STomohiro Misono } 2706b64ec075STomohiro Misono 2707b64ec075STomohiro Misono leaf = path->nodes[0]; 2708b64ec075STomohiro Misono slot = path->slots[0]; 2709b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2710b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2711b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2712b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2713b64ec075STomohiro Misono 2714b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2715b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2716b64ec075STomohiro Misono 2717b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2718b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2719b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2720b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2721b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2722b64ec075STomohiro Misono item_off, item_len); 2723b64ec075STomohiro Misono } else { 2724b64ec075STomohiro Misono ret = -ENOENT; 2725b64ec075STomohiro Misono goto out; 2726b64ec075STomohiro Misono } 2727b64ec075STomohiro Misono } 2728b64ec075STomohiro Misono 2729b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2730b64ec075STomohiro Misono ret = -EFAULT; 2731b64ec075STomohiro Misono 2732b64ec075STomohiro Misono out: 2733b64ec075STomohiro Misono btrfs_free_path(path); 2734b64ec075STomohiro Misono kzfree(subvol_info); 2735b64ec075STomohiro Misono return ret; 2736b64ec075STomohiro Misono } 2737b64ec075STomohiro Misono 273842e4b520STomohiro Misono /* 273942e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 274042e4b520STomohiro Misono * except the subvolume name. 274142e4b520STomohiro Misono */ 274242e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 274342e4b520STomohiro Misono { 274442e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 274542e4b520STomohiro Misono struct btrfs_root_ref *rref; 274642e4b520STomohiro Misono struct btrfs_root *root; 274742e4b520STomohiro Misono struct btrfs_path *path; 274842e4b520STomohiro Misono struct btrfs_key key; 274942e4b520STomohiro Misono struct extent_buffer *leaf; 275042e4b520STomohiro Misono struct inode *inode; 275142e4b520STomohiro Misono u64 objectid; 275242e4b520STomohiro Misono int slot; 275342e4b520STomohiro Misono int ret; 275442e4b520STomohiro Misono u8 found; 275542e4b520STomohiro Misono 275642e4b520STomohiro Misono path = btrfs_alloc_path(); 275742e4b520STomohiro Misono if (!path) 275842e4b520STomohiro Misono return -ENOMEM; 275942e4b520STomohiro Misono 276042e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 276142e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 276242e4b520STomohiro Misono btrfs_free_path(path); 276342e4b520STomohiro Misono return PTR_ERR(rootrefs); 276442e4b520STomohiro Misono } 276542e4b520STomohiro Misono 276642e4b520STomohiro Misono inode = file_inode(file); 276742e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 276842e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 276942e4b520STomohiro Misono 277042e4b520STomohiro Misono key.objectid = objectid; 277142e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 277242e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 277342e4b520STomohiro Misono found = 0; 277442e4b520STomohiro Misono 277542e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 277642e4b520STomohiro Misono if (ret < 0) { 277742e4b520STomohiro Misono goto out; 277842e4b520STomohiro Misono } else if (path->slots[0] >= 277942e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 278042e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 278142e4b520STomohiro Misono if (ret < 0) { 278242e4b520STomohiro Misono goto out; 278342e4b520STomohiro Misono } else if (ret > 0) { 278442e4b520STomohiro Misono ret = -EUCLEAN; 278542e4b520STomohiro Misono goto out; 278642e4b520STomohiro Misono } 278742e4b520STomohiro Misono } 278842e4b520STomohiro Misono while (1) { 278942e4b520STomohiro Misono leaf = path->nodes[0]; 279042e4b520STomohiro Misono slot = path->slots[0]; 279142e4b520STomohiro Misono 279242e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 279342e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 279442e4b520STomohiro Misono ret = 0; 279542e4b520STomohiro Misono goto out; 279642e4b520STomohiro Misono } 279742e4b520STomohiro Misono 279842e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 279942e4b520STomohiro Misono ret = -EOVERFLOW; 280042e4b520STomohiro Misono goto out; 280142e4b520STomohiro Misono } 280242e4b520STomohiro Misono 280342e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 280442e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 280542e4b520STomohiro Misono rootrefs->rootref[found].dirid = 280642e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 280742e4b520STomohiro Misono found++; 280842e4b520STomohiro Misono 280942e4b520STomohiro Misono ret = btrfs_next_item(root, path); 281042e4b520STomohiro Misono if (ret < 0) { 281142e4b520STomohiro Misono goto out; 281242e4b520STomohiro Misono } else if (ret > 0) { 281342e4b520STomohiro Misono ret = -EUCLEAN; 281442e4b520STomohiro Misono goto out; 281542e4b520STomohiro Misono } 281642e4b520STomohiro Misono } 281742e4b520STomohiro Misono 281842e4b520STomohiro Misono out: 281942e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 282042e4b520STomohiro Misono rootrefs->num_items = found; 282142e4b520STomohiro Misono /* update min_treeid for next search */ 282242e4b520STomohiro Misono if (found) 282342e4b520STomohiro Misono rootrefs->min_treeid = 282442e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 282542e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 282642e4b520STomohiro Misono ret = -EFAULT; 282742e4b520STomohiro Misono } 282842e4b520STomohiro Misono 282942e4b520STomohiro Misono kfree(rootrefs); 283042e4b520STomohiro Misono btrfs_free_path(path); 283142e4b520STomohiro Misono 283242e4b520STomohiro Misono return ret; 283342e4b520STomohiro Misono } 283442e4b520STomohiro Misono 283576dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 283676dda93cSYan, Zheng void __user *arg) 283776dda93cSYan, Zheng { 283854563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 284076dda93cSYan, Zheng struct dentry *dentry; 28412b0143b5SDavid Howells struct inode *dir = d_inode(parent); 284276dda93cSYan, Zheng struct inode *inode; 284376dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 284476dda93cSYan, Zheng struct btrfs_root *dest = NULL; 284576dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 284676dda93cSYan, Zheng int namelen; 284776dda93cSYan, Zheng int err = 0; 284876dda93cSYan, Zheng 2849325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2850325c50e3SJeff Mahoney return -ENOTDIR; 2851325c50e3SJeff Mahoney 285276dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 285376dda93cSYan, Zheng if (IS_ERR(vol_args)) 285476dda93cSYan, Zheng return PTR_ERR(vol_args); 285576dda93cSYan, Zheng 285676dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 285776dda93cSYan, Zheng namelen = strlen(vol_args->name); 285876dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 285976dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 286076dda93cSYan, Zheng err = -EINVAL; 286176dda93cSYan, Zheng goto out; 286276dda93cSYan, Zheng } 286376dda93cSYan, Zheng 2864a561be71SAl Viro err = mnt_want_write_file(file); 286576dda93cSYan, Zheng if (err) 286676dda93cSYan, Zheng goto out; 286776dda93cSYan, Zheng 2868521e0546SDavid Sterba 286900235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 287000235411SAl Viro if (err == -EINTR) 287100235411SAl Viro goto out_drop_write; 287276dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 287376dda93cSYan, Zheng if (IS_ERR(dentry)) { 287476dda93cSYan, Zheng err = PTR_ERR(dentry); 287576dda93cSYan, Zheng goto out_unlock_dir; 287676dda93cSYan, Zheng } 287776dda93cSYan, Zheng 28782b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 287976dda93cSYan, Zheng err = -ENOENT; 288076dda93cSYan, Zheng goto out_dput; 288176dda93cSYan, Zheng } 288276dda93cSYan, Zheng 28832b0143b5SDavid Howells inode = d_inode(dentry); 28844260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28854260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28864260f7c7SSage Weil /* 28874260f7c7SSage Weil * Regular user. Only allow this with a special mount 28884260f7c7SSage Weil * option, when the user has write+exec access to the 28894260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28904260f7c7SSage Weil * allowed. 28914260f7c7SSage Weil * 28924260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28934260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28944260f7c7SSage Weil * otherwise be able to delete. 28954260f7c7SSage Weil * 28964260f7c7SSage Weil * Users who want to delete empty subvols should try 28974260f7c7SSage Weil * rmdir(2). 28984260f7c7SSage Weil */ 28994260f7c7SSage Weil err = -EPERM; 29000b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 29014260f7c7SSage Weil goto out_dput; 29024260f7c7SSage Weil 29034260f7c7SSage Weil /* 29044260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 29054260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 29064260f7c7SSage Weil * must be called on the dentry referencing the root 29074260f7c7SSage Weil * of the subvol, not a random directory contained 29084260f7c7SSage Weil * within it. 29094260f7c7SSage Weil */ 29104260f7c7SSage Weil err = -EINVAL; 29114260f7c7SSage Weil if (root == dest) 29124260f7c7SSage Weil goto out_dput; 29134260f7c7SSage Weil 29144260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 29154260f7c7SSage Weil if (err) 29164260f7c7SSage Weil goto out_dput; 29175c39da5bSMiao Xie } 29184260f7c7SSage Weil 29195c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 29204260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 29214260f7c7SSage Weil if (err) 29224260f7c7SSage Weil goto out_dput; 29234260f7c7SSage Weil 29244a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 292576dda93cSYan, Zheng err = -EINVAL; 292676dda93cSYan, Zheng goto out_dput; 292776dda93cSYan, Zheng } 292876dda93cSYan, Zheng 29295955102cSAl Viro inode_lock(inode); 2930f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 29315955102cSAl Viro inode_unlock(inode); 293246008d9dSAmir Goldstein if (!err) { 293346008d9dSAmir Goldstein fsnotify_rmdir(dir, dentry); 293476dda93cSYan, Zheng d_delete(dentry); 293546008d9dSAmir Goldstein } 2936fa6ac876SLiu Bo 293776dda93cSYan, Zheng out_dput: 293876dda93cSYan, Zheng dput(dentry); 293976dda93cSYan, Zheng out_unlock_dir: 29405955102cSAl Viro inode_unlock(dir); 294100235411SAl Viro out_drop_write: 29422a79f17eSAl Viro mnt_drop_write_file(file); 294376dda93cSYan, Zheng out: 294476dda93cSYan, Zheng kfree(vol_args); 294576dda93cSYan, Zheng return err; 294676dda93cSYan, Zheng } 294776dda93cSYan, Zheng 29481e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2949f46b5a66SChristoph Hellwig { 2950496ad9aaSAl Viro struct inode *inode = file_inode(file); 2951f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29521e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2953c146afadSYan Zheng int ret; 2954c146afadSYan Zheng 295525122d15SIlya Dryomov ret = mnt_want_write_file(file); 295625122d15SIlya Dryomov if (ret) 295725122d15SIlya Dryomov return ret; 2958b83cc969SLi Zefan 295925122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 296025122d15SIlya Dryomov ret = -EROFS; 296125122d15SIlya Dryomov goto out; 29625ac00addSStefan Behrens } 2963f46b5a66SChristoph Hellwig 2964f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2965f46b5a66SChristoph Hellwig case S_IFDIR: 2966e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2967e441d54dSChris Mason ret = -EPERM; 2968e441d54dSChris Mason goto out; 2969e441d54dSChris Mason } 2970de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2971f46b5a66SChristoph Hellwig break; 2972f46b5a66SChristoph Hellwig case S_IFREG: 2973616d374eSAdam Borowski /* 2974616d374eSAdam Borowski * Note that this does not check the file descriptor for write 2975616d374eSAdam Borowski * access. This prevents defragmenting executables that are 2976616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 2977616d374eSAdam Borowski */ 2978616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 2979616d374eSAdam Borowski inode_permission(inode, MAY_WRITE)) { 2980616d374eSAdam Borowski ret = -EPERM; 2981e441d54dSChris Mason goto out; 2982e441d54dSChris Mason } 29831e701a32SChris Mason 29841e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29851e701a32SChris Mason if (!range) { 29861e701a32SChris Mason ret = -ENOMEM; 29871e701a32SChris Mason goto out; 29881e701a32SChris Mason } 29891e701a32SChris Mason 29901e701a32SChris Mason if (argp) { 29911e701a32SChris Mason if (copy_from_user(range, argp, 29921e701a32SChris Mason sizeof(*range))) { 29931e701a32SChris Mason ret = -EFAULT; 29941e701a32SChris Mason kfree(range); 2995683be16eSDan Carpenter goto out; 29961e701a32SChris Mason } 29971e701a32SChris Mason /* compression requires us to start the IO */ 29981e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 29991e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 30001e701a32SChris Mason range->extent_thresh = (u32)-1; 30011e701a32SChris Mason } 30021e701a32SChris Mason } else { 30031e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 30041e701a32SChris Mason range->len = (u64)-1; 30051e701a32SChris Mason } 3006496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 30077c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 30084cb5300bSChris Mason if (ret > 0) 30094cb5300bSChris Mason ret = 0; 30101e701a32SChris Mason kfree(range); 3011f46b5a66SChristoph Hellwig break; 30128929ecfaSYan, Zheng default: 30138929ecfaSYan, Zheng ret = -EINVAL; 3014f46b5a66SChristoph Hellwig } 3015e441d54dSChris Mason out: 301625122d15SIlya Dryomov mnt_drop_write_file(file); 3017e441d54dSChris Mason return ret; 3018f46b5a66SChristoph Hellwig } 3019f46b5a66SChristoph Hellwig 30202ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3021f46b5a66SChristoph Hellwig { 3022f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3023f46b5a66SChristoph Hellwig int ret; 3024f46b5a66SChristoph Hellwig 3025e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3026e441d54dSChris Mason return -EPERM; 3027e441d54dSChris Mason 3028171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 3029e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3030c9e9f97bSIlya Dryomov 3031dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3032c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3033c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3034c9e9f97bSIlya Dryomov goto out; 3035c9e9f97bSIlya Dryomov } 3036f46b5a66SChristoph Hellwig 30375516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30382ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3039f46b5a66SChristoph Hellwig 304043d20761SAnand Jain if (!ret) 30410b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 304243d20761SAnand Jain 3043f46b5a66SChristoph Hellwig kfree(vol_args); 3044c9e9f97bSIlya Dryomov out: 3045171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3046f46b5a66SChristoph Hellwig return ret; 3047f46b5a66SChristoph Hellwig } 3048f46b5a66SChristoph Hellwig 30496b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3050f46b5a66SChristoph Hellwig { 30510b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30520b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30536b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3054f46b5a66SChristoph Hellwig int ret; 3055f46b5a66SChristoph Hellwig 3056e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3057e441d54dSChris Mason return -EPERM; 3058e441d54dSChris Mason 3059da24927bSMiao Xie ret = mnt_want_write_file(file); 3060da24927bSMiao Xie if (ret) 3061da24927bSMiao Xie return ret; 3062c146afadSYan Zheng 3063dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3064c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3065c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3066c47ca32dSDan Carpenter goto err_drop; 3067c9e9f97bSIlya Dryomov } 3068f46b5a66SChristoph Hellwig 30696b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3070fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3071fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3072fd4e994bSOmar Sandoval goto out; 3073fd4e994bSOmar Sandoval } 3074f46b5a66SChristoph Hellwig 3075171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3076183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3077183860f6SAnand Jain goto out; 3078183860f6SAnand Jain } 3079183860f6SAnand Jain 3080735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30812ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30826b526ed7SAnand Jain } else { 30836b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30842ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30856b526ed7SAnand Jain } 3086171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3087183860f6SAnand Jain 30886b526ed7SAnand Jain if (!ret) { 3089735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30900b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30916b526ed7SAnand Jain vol_args->devid); 30926b526ed7SAnand Jain else 30930b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30946b526ed7SAnand Jain vol_args->name); 30956b526ed7SAnand Jain } 3096183860f6SAnand Jain out: 3097183860f6SAnand Jain kfree(vol_args); 3098c47ca32dSDan Carpenter err_drop: 30994ac20c70SIlya Dryomov mnt_drop_write_file(file); 3100f46b5a66SChristoph Hellwig return ret; 3101f46b5a66SChristoph Hellwig } 3102f46b5a66SChristoph Hellwig 3103f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3104f46b5a66SChristoph Hellwig { 31050b246afaSJeff Mahoney struct inode *inode = file_inode(file); 31060b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3107f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3108f46b5a66SChristoph Hellwig int ret; 3109f46b5a66SChristoph Hellwig 3110f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3111f46b5a66SChristoph Hellwig return -EPERM; 3112f46b5a66SChristoph Hellwig 3113f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3114f46b5a66SChristoph Hellwig if (ret) 3115f46b5a66SChristoph Hellwig return ret; 3116f46b5a66SChristoph Hellwig 3117171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3118f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 311958d7bbf8SDavid Sterba goto out_drop_write; 312058d7bbf8SDavid Sterba } 312158d7bbf8SDavid Sterba 312258d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 312358d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 312458d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3125f46b5a66SChristoph Hellwig goto out; 3126f46b5a66SChristoph Hellwig } 3127f46b5a66SChristoph Hellwig 312858d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31292ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3130f46b5a66SChristoph Hellwig 3131f46b5a66SChristoph Hellwig if (!ret) 31320b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3133f46b5a66SChristoph Hellwig kfree(vol_args); 313458d7bbf8SDavid Sterba out: 3135171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 313658d7bbf8SDavid Sterba out_drop_write: 3137f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 313858d7bbf8SDavid Sterba 3139f46b5a66SChristoph Hellwig return ret; 3140f46b5a66SChristoph Hellwig } 3141f46b5a66SChristoph Hellwig 31422ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 31432ff7e61eSJeff Mahoney void __user *arg) 3144475f6387SJan Schmidt { 3145027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3146475f6387SJan Schmidt struct btrfs_device *device; 31470b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3148027ed2f0SLi Zefan int ret = 0; 3149475f6387SJan Schmidt 3150027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3151027ed2f0SLi Zefan if (!fi_args) 3152027ed2f0SLi Zefan return -ENOMEM; 3153027ed2f0SLi Zefan 3154d03262c7SDavid Sterba rcu_read_lock(); 3155027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3156475f6387SJan Schmidt 3157d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3158027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3159027ed2f0SLi Zefan fi_args->max_id = device->devid; 3160475f6387SJan Schmidt } 3161d03262c7SDavid Sterba rcu_read_unlock(); 3162475f6387SJan Schmidt 3163de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3164bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3165bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3166bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 316780a773fbSDavid Sterba 3168027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3169027ed2f0SLi Zefan ret = -EFAULT; 3170475f6387SJan Schmidt 3171027ed2f0SLi Zefan kfree(fi_args); 3172027ed2f0SLi Zefan return ret; 3173475f6387SJan Schmidt } 3174475f6387SJan Schmidt 31752ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31762ff7e61eSJeff Mahoney void __user *arg) 3177475f6387SJan Schmidt { 3178475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3179475f6387SJan Schmidt struct btrfs_device *dev; 3180475f6387SJan Schmidt int ret = 0; 3181475f6387SJan Schmidt char *s_uuid = NULL; 3182475f6387SJan Schmidt 3183475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3184475f6387SJan Schmidt if (IS_ERR(di_args)) 3185475f6387SJan Schmidt return PTR_ERR(di_args); 3186475f6387SJan Schmidt 3187dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3188475f6387SJan Schmidt s_uuid = di_args->uuid; 3189475f6387SJan Schmidt 3190c5593ca3SDavid Sterba rcu_read_lock(); 3191e4319cd9SAnand Jain dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid, 319209ba3bc9SAnand Jain NULL, true); 3193475f6387SJan Schmidt 3194475f6387SJan Schmidt if (!dev) { 3195475f6387SJan Schmidt ret = -ENODEV; 3196475f6387SJan Schmidt goto out; 3197475f6387SJan Schmidt } 3198475f6387SJan Schmidt 3199475f6387SJan Schmidt di_args->devid = dev->devid; 32007cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 32017cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3202475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3203a27202fbSJim Meyering if (dev->name) { 3204672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3205672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3206a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3207a27202fbSJim Meyering } else { 320899ba55adSStefan Behrens di_args->path[0] = '\0'; 3209a27202fbSJim Meyering } 3210475f6387SJan Schmidt 3211475f6387SJan Schmidt out: 3212c5593ca3SDavid Sterba rcu_read_unlock(); 3213475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3214475f6387SJan Schmidt ret = -EFAULT; 3215475f6387SJan Schmidt 3216475f6387SJan Schmidt kfree(di_args); 3217475f6387SJan Schmidt return ret; 3218475f6387SJan Schmidt } 3219475f6387SJan Schmidt 3220d8b55242SFilipe Manana static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3221d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3222d8b55242SFilipe Manana { 3223d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3224d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3225d8b55242SFilipe Manana } 3226d8b55242SFilipe Manana 3227d8b55242SFilipe Manana static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3228d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3229d8b55242SFilipe Manana { 3230d8b55242SFilipe Manana if (inode1 < inode2) { 3231d8b55242SFilipe Manana swap(inode1, inode2); 3232d8b55242SFilipe Manana swap(loff1, loff2); 3233d8b55242SFilipe Manana } else if (inode1 == inode2 && loff2 < loff1) { 3234d8b55242SFilipe Manana swap(loff1, loff2); 3235d8b55242SFilipe Manana } 3236d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3237d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3238d8b55242SFilipe Manana } 3239d8b55242SFilipe Manana 324057a50e25SFilipe Manana static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len, 324134a28e3dSFilipe Manana struct inode *dst, u64 dst_loff) 3242416161dbSMark Fasheh { 3243416161dbSMark Fasheh int ret; 32440efa9f48SMark Fasheh 3245de02b9f6SFilipe Manana /* 3246d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3247d8b55242SFilipe Manana * source range to serialize with relocation. 3248de02b9f6SFilipe Manana */ 3249d8b55242SFilipe Manana btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 325057a50e25SFilipe Manana ret = btrfs_clone(src, dst, loff, len, len, dst_loff, 1); 3251d8b55242SFilipe Manana btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 32523973909dSTimofey Titovets 32533973909dSTimofey Titovets return ret; 32543973909dSTimofey Titovets } 32553973909dSTimofey Titovets 3256b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3257b6728768STimofey Titovets 32583973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 32593973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 32603973909dSTimofey Titovets { 32613973909dSTimofey Titovets int ret; 3262b6728768STimofey Titovets u64 i, tail_len, chunk_count; 326362d54f3aSFilipe Manana struct btrfs_root *root_dst = BTRFS_I(dst)->root; 326462d54f3aSFilipe Manana 326562d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 326662d54f3aSFilipe Manana if (root_dst->send_in_progress) { 326762d54f3aSFilipe Manana btrfs_warn_rl(root_dst->fs_info, 326862d54f3aSFilipe Manana "cannot deduplicate to root %llu while send operations are using it (%d in progress)", 326962d54f3aSFilipe Manana root_dst->root_key.objectid, 327062d54f3aSFilipe Manana root_dst->send_in_progress); 327162d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 327262d54f3aSFilipe Manana return -EAGAIN; 327362d54f3aSFilipe Manana } 327462d54f3aSFilipe Manana root_dst->dedupe_in_progress++; 327562d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 32763973909dSTimofey Titovets 3277b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3278b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 327967b07bd4STimofey Titovets 3280b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3281b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 328234a28e3dSFilipe Manana dst, dst_loff); 3283b6728768STimofey Titovets if (ret) 328462d54f3aSFilipe Manana goto out; 3285b6728768STimofey Titovets 3286b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3287b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3288b6728768STimofey Titovets } 3289b6728768STimofey Titovets 3290b6728768STimofey Titovets if (tail_len > 0) 329167b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 329234a28e3dSFilipe Manana dst_loff); 329362d54f3aSFilipe Manana out: 329462d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 329562d54f3aSFilipe Manana root_dst->dedupe_in_progress--; 329662d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 3297416161dbSMark Fasheh 3298416161dbSMark Fasheh return ret; 3299416161dbSMark Fasheh } 3300416161dbSMark Fasheh 3301f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3302f82a9901SFilipe Manana struct inode *inode, 3303f82a9901SFilipe Manana u64 endoff, 3304f82a9901SFilipe Manana const u64 destoff, 33051c919a5eSMark Fasheh const u64 olen, 33061c919a5eSMark Fasheh int no_time_update) 3307f82a9901SFilipe Manana { 3308f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3309f82a9901SFilipe Manana int ret; 3310f82a9901SFilipe Manana 3311f82a9901SFilipe Manana inode_inc_iversion(inode); 33121c919a5eSMark Fasheh if (!no_time_update) 3313c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3314f82a9901SFilipe Manana /* 3315f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3316f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3317f82a9901SFilipe Manana */ 3318f82a9901SFilipe Manana if (endoff > destoff + olen) 3319f82a9901SFilipe Manana endoff = destoff + olen; 3320f82a9901SFilipe Manana if (endoff > inode->i_size) 33216ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3322f82a9901SFilipe Manana 3323f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3324f82a9901SFilipe Manana if (ret) { 332566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 33263a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3327f82a9901SFilipe Manana goto out; 3328f82a9901SFilipe Manana } 33293a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3330f82a9901SFilipe Manana out: 3331f82a9901SFilipe Manana return ret; 3332f82a9901SFilipe Manana } 3333f82a9901SFilipe Manana 33348039d87dSFilipe Manana /* 33358039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33368039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33378039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33388039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33398039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33408039d87dSFilipe Manana * 33418039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33428039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33438039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33448039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33458039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33468039d87dSFilipe Manana * i_mutex, which we are holding here). 33478039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33488039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33498039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33508039d87dSFilipe Manana * 33518039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33528039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33538039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33548039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33558039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33568039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33578039d87dSFilipe Manana * the former. 33588039d87dSFilipe Manana */ 33594a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 33608039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33618039d87dSFilipe Manana struct btrfs_path *path, 33628039d87dSFilipe Manana struct btrfs_key *new_key, 33638039d87dSFilipe Manana const u64 drop_start, 33648039d87dSFilipe Manana const u64 datal, 33658039d87dSFilipe Manana const u64 skip, 33668039d87dSFilipe Manana const u64 size, 33678039d87dSFilipe Manana char *inline_data) 33688039d87dSFilipe Manana { 33690b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 33708039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 33718039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 33720b246afaSJeff Mahoney fs_info->sectorsize); 33738039d87dSFilipe Manana int ret; 33748039d87dSFilipe Manana struct btrfs_key key; 33758039d87dSFilipe Manana 33768039d87dSFilipe Manana if (new_key->offset > 0) 33778039d87dSFilipe Manana return -EOPNOTSUPP; 33788039d87dSFilipe Manana 33794a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 33808039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 33818039d87dSFilipe Manana key.offset = 0; 33828039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 33838039d87dSFilipe Manana if (ret < 0) { 33848039d87dSFilipe Manana return ret; 33858039d87dSFilipe Manana } else if (ret > 0) { 33868039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 33878039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 33888039d87dSFilipe Manana if (ret < 0) 33898039d87dSFilipe Manana return ret; 33908039d87dSFilipe Manana else if (ret > 0) 33918039d87dSFilipe Manana goto copy_inline_extent; 33928039d87dSFilipe Manana } 33938039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 33944a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 33958039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 33968039d87dSFilipe Manana ASSERT(key.offset > 0); 33978039d87dSFilipe Manana return -EOPNOTSUPP; 33988039d87dSFilipe Manana } 33998039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34008039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34018039d87dSFilipe Manana u64 ext_len; 34028039d87dSFilipe Manana 34038039d87dSFilipe Manana /* 34048039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34058039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34068039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34078039d87dSFilipe Manana */ 34088039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34098039d87dSFilipe Manana struct btrfs_file_extent_item); 34108039d87dSFilipe Manana /* 34118039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34128039d87dSFilipe Manana * following it. 34138039d87dSFilipe Manana */ 34148039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34158039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34168039d87dSFilipe Manana goto copy_inline_extent; 34178039d87dSFilipe Manana 34188039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34198039d87dSFilipe Manana if (ext_len > aligned_end) 34208039d87dSFilipe Manana return -EOPNOTSUPP; 34218039d87dSFilipe Manana 34228039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34238039d87dSFilipe Manana if (ret < 0) { 34248039d87dSFilipe Manana return ret; 34258039d87dSFilipe Manana } else if (ret == 0) { 34268039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34278039d87dSFilipe Manana path->slots[0]); 34284a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34298039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34308039d87dSFilipe Manana return -EOPNOTSUPP; 34318039d87dSFilipe Manana } 34328039d87dSFilipe Manana } 34338039d87dSFilipe Manana 34348039d87dSFilipe Manana copy_inline_extent: 34358039d87dSFilipe Manana /* 34368039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34378039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34388039d87dSFilipe Manana */ 34398039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34408039d87dSFilipe Manana /* 34418039d87dSFilipe Manana * If the destination inode has an inline extent... 34428039d87dSFilipe Manana * This would require copying the data from the source inline 34438039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34448039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34458039d87dSFilipe Manana * or just one of them, which would require decompressing and 34468039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34478039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34488039d87dSFilipe Manana * inline extent). 34498039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34508039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34518039d87dSFilipe Manana */ 34528039d87dSFilipe Manana return -EOPNOTSUPP; 34538039d87dSFilipe Manana } 34548039d87dSFilipe Manana 34558039d87dSFilipe Manana btrfs_release_path(path); 34568039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34578039d87dSFilipe Manana if (ret) 34588039d87dSFilipe Manana return ret; 34598039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34608039d87dSFilipe Manana if (ret) 34618039d87dSFilipe Manana return ret; 34628039d87dSFilipe Manana 34638039d87dSFilipe Manana if (skip) { 34648039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 34658039d87dSFilipe Manana 34668039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 34678039d87dSFilipe Manana } 34688039d87dSFilipe Manana 34698039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 34708039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 34718039d87dSFilipe Manana path->slots[0]), 34728039d87dSFilipe Manana size); 34738039d87dSFilipe Manana inode_add_bytes(dst, datal); 3474690a5dbfSFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(dst)->runtime_flags); 34758039d87dSFilipe Manana 34768039d87dSFilipe Manana return 0; 34778039d87dSFilipe Manana } 34788039d87dSFilipe Manana 347932b7c687SMark Fasheh /** 348032b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 348132b7c687SMark Fasheh * 348232b7c687SMark Fasheh * @src: Inode to clone from 348332b7c687SMark Fasheh * @inode: Inode to clone to 348432b7c687SMark Fasheh * @off: Offset within source to start clone from 348532b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 34861c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 348732b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 34881c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 348932b7c687SMark Fasheh */ 349032b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3491f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 34921c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3493f46b5a66SChristoph Hellwig { 34940b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3495f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 349632b7c687SMark Fasheh struct btrfs_path *path = NULL; 3497f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 349832b7c687SMark Fasheh struct btrfs_trans_handle *trans; 349932b7c687SMark Fasheh char *buf = NULL; 3500ae01a0abSYan Zheng struct btrfs_key key; 3501f46b5a66SChristoph Hellwig u32 nritems; 3502f46b5a66SChristoph Hellwig int slot; 3503ae01a0abSYan Zheng int ret; 3504f82a9901SFilipe Manana const u64 len = olen_aligned; 3505f82a9901SFilipe Manana u64 last_dest_end = destoff; 3506ae01a0abSYan Zheng 3507ae01a0abSYan Zheng ret = -ENOMEM; 3508752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3509ae01a0abSYan Zheng if (!buf) 351032b7c687SMark Fasheh return ret; 3511ae01a0abSYan Zheng 3512ae01a0abSYan Zheng path = btrfs_alloc_path(); 3513ae01a0abSYan Zheng if (!path) { 351415351955SDavid Sterba kvfree(buf); 351532b7c687SMark Fasheh return ret; 3516ae01a0abSYan Zheng } 351732b7c687SMark Fasheh 3518e4058b54SDavid Sterba path->reada = READA_FORWARD; 3519c5c9cd4dSSage Weil /* clone data */ 35204a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3521ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35222c463823SFilipe Manana key.offset = off; 3523f46b5a66SChristoph Hellwig 3524f46b5a66SChristoph Hellwig while (1) { 3525de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3526b64119b5SFilipe Manana struct btrfs_file_extent_item *extent; 3527b64119b5SFilipe Manana int type; 3528b64119b5SFilipe Manana u32 size; 3529b64119b5SFilipe Manana struct btrfs_key new_key; 3530b64119b5SFilipe Manana u64 disko = 0, diskl = 0; 3531b64119b5SFilipe Manana u64 datao = 0, datal = 0; 3532b64119b5SFilipe Manana u8 comp; 3533b64119b5SFilipe Manana u64 drop_start; 3534df858e76SFilipe Manana 3535f46b5a66SChristoph Hellwig /* 3536f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3537f46b5a66SChristoph Hellwig * tree. 3538f46b5a66SChristoph Hellwig */ 3539e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3540362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3541362a20c5SDavid Sterba 0, 0); 3542f46b5a66SChristoph Hellwig if (ret < 0) 3543f46b5a66SChristoph Hellwig goto out; 35442c463823SFilipe Manana /* 35452c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35462c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35472c463823SFilipe Manana * it might overlap our target range, therefore process it. 35482c463823SFilipe Manana */ 35492c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35502c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35512c463823SFilipe Manana path->slots[0] - 1); 35522c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35532c463823SFilipe Manana path->slots[0]--; 35542c463823SFilipe Manana } 3555f46b5a66SChristoph Hellwig 3556ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3557e4355f34SFilipe David Borba Manana process_slot: 3558ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3559362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3560f46b5a66SChristoph Hellwig if (ret < 0) 3561f46b5a66SChristoph Hellwig goto out; 3562f46b5a66SChristoph Hellwig if (ret > 0) 3563f46b5a66SChristoph Hellwig break; 3564ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3565f46b5a66SChristoph Hellwig } 3566f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3567f46b5a66SChristoph Hellwig slot = path->slots[0]; 3568f46b5a66SChristoph Hellwig 3569ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3570962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 35714a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3572f46b5a66SChristoph Hellwig break; 3573f46b5a66SChristoph Hellwig 3574b64119b5SFilipe Manana ASSERT(key.type == BTRFS_EXTENT_DATA_KEY); 357531840ae1SZheng Yan 3576c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3577c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3578c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3579c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3580c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3581c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3582b64119b5SFilipe Manana disko = btrfs_file_extent_disk_bytenr(leaf, extent); 3583b64119b5SFilipe Manana diskl = btrfs_file_extent_disk_num_bytes(leaf, extent); 3584c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3585b64119b5SFilipe Manana datal = btrfs_file_extent_num_bytes(leaf, extent); 3586c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3587b64119b5SFilipe Manana /* Take upper bound, may be compressed */ 3588b64119b5SFilipe Manana datal = btrfs_file_extent_ram_bytes(leaf, extent); 3589c5c9cd4dSSage Weil } 359031840ae1SZheng Yan 35912c463823SFilipe Manana /* 3592b64119b5SFilipe Manana * The first search might have left us at an extent item that 3593b64119b5SFilipe Manana * ends before our target range's start, can happen if we have 3594b64119b5SFilipe Manana * holes and NO_HOLES feature enabled. 35952c463823SFilipe Manana */ 35962c463823SFilipe Manana if (key.offset + datal <= off) { 3597e4355f34SFilipe David Borba Manana path->slots[0]++; 3598e4355f34SFilipe David Borba Manana goto process_slot; 35992c463823SFilipe Manana } else if (key.offset >= off + len) { 36002c463823SFilipe Manana break; 3601e4355f34SFilipe David Borba Manana } 3602df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3603e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3604b64119b5SFilipe Manana read_extent_buffer(leaf, buf, btrfs_item_ptr_offset(leaf, slot), 3605e4355f34SFilipe David Borba Manana size); 3606e4355f34SFilipe David Borba Manana 3607e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3608e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3609c5c9cd4dSSage Weil 361031840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36114a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36124d728ec7SLi Zefan if (off <= key.offset) 3613c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36144d728ec7SLi Zefan else 36154d728ec7SLi Zefan new_key.offset = destoff; 3616c5c9cd4dSSage Weil 3617b6f3409bSSage Weil /* 3618b64119b5SFilipe Manana * Deal with a hole that doesn't have an extent item that 3619b64119b5SFilipe Manana * represents it (NO_HOLES feature enabled). 3620b64119b5SFilipe Manana * This hole is either in the middle of the cloning range or at 3621b64119b5SFilipe Manana * the beginning (fully overlaps it or partially overlaps it). 3622f82a9901SFilipe Manana */ 3623f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3624f82a9901SFilipe Manana drop_start = last_dest_end; 3625f82a9901SFilipe Manana else 3626f82a9901SFilipe Manana drop_start = new_key.offset; 3627f82a9901SFilipe Manana 3628c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3629c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3630690a5dbfSFilipe Manana struct btrfs_clone_extent_info clone_info; 3631690a5dbfSFilipe Manana 3632d72c0842SLi Zefan /* 3633d72c0842SLi Zefan * a | --- range to clone ---| b 3634d72c0842SLi Zefan * | ------------- extent ------------- | 3635d72c0842SLi Zefan */ 3636d72c0842SLi Zefan 3637b64119b5SFilipe Manana /* Subtract range b */ 3638d72c0842SLi Zefan if (key.offset + datal > off + len) 3639d72c0842SLi Zefan datal = off + len - key.offset; 3640d72c0842SLi Zefan 3641b64119b5SFilipe Manana /* Subtract range a */ 3642a22285a6SYan, Zheng if (off > key.offset) { 3643a22285a6SYan, Zheng datao += off - key.offset; 3644a22285a6SYan, Zheng datal -= off - key.offset; 3645a22285a6SYan, Zheng } 3646a22285a6SYan, Zheng 3647690a5dbfSFilipe Manana clone_info.disk_offset = disko; 3648690a5dbfSFilipe Manana clone_info.disk_len = diskl; 3649690a5dbfSFilipe Manana clone_info.data_offset = datao; 3650690a5dbfSFilipe Manana clone_info.data_len = datal; 3651690a5dbfSFilipe Manana clone_info.file_offset = new_key.offset; 3652690a5dbfSFilipe Manana clone_info.extent_buf = buf; 3653690a5dbfSFilipe Manana clone_info.item_size = size; 3654690a5dbfSFilipe Manana ret = btrfs_punch_hole_range(inode, path, 3655f82a9901SFilipe Manana drop_start, 3656690a5dbfSFilipe Manana new_key.offset + datal - 1, 3657690a5dbfSFilipe Manana &clone_info, &trans); 3658690a5dbfSFilipe Manana if (ret) 365979787eaaSJeff Mahoney goto out; 3660c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3661c5c9cd4dSSage Weil u64 skip = 0; 3662c5c9cd4dSSage Weil u64 trim = 0; 3663ed958762SFilipe Manana 3664c5c9cd4dSSage Weil if (off > key.offset) { 3665c5c9cd4dSSage Weil skip = off - key.offset; 3666c5c9cd4dSSage Weil new_key.offset += skip; 366731840ae1SZheng Yan } 3668d397712bSChris Mason 3669c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3670c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3671d397712bSChris Mason 3672c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3673c5c9cd4dSSage Weil ret = -EINVAL; 3674c5c9cd4dSSage Weil goto out; 367531840ae1SZheng Yan } 3676c5c9cd4dSSage Weil size -= skip + trim; 3677c5c9cd4dSSage Weil datal -= skip + trim; 3678a22285a6SYan, Zheng 3679690a5dbfSFilipe Manana /* 3680b64119b5SFilipe Manana * If our extent is inline, we know we will drop or 3681b64119b5SFilipe Manana * adjust at most 1 extent item in the destination root. 3682690a5dbfSFilipe Manana * 3683b64119b5SFilipe Manana * 1 - adjusting old extent (we may have to split it) 3684690a5dbfSFilipe Manana * 1 - add new extent 3685690a5dbfSFilipe Manana * 1 - inode update 3686690a5dbfSFilipe Manana */ 3687690a5dbfSFilipe Manana trans = btrfs_start_transaction(root, 3); 3688690a5dbfSFilipe Manana if (IS_ERR(trans)) { 3689690a5dbfSFilipe Manana ret = PTR_ERR(trans); 3690690a5dbfSFilipe Manana goto out; 3691690a5dbfSFilipe Manana } 3692690a5dbfSFilipe Manana 3693b64119b5SFilipe Manana ret = clone_copy_inline_extent(inode, trans, path, 3694b64119b5SFilipe Manana &new_key, drop_start, 3695b64119b5SFilipe Manana datal, skip, size, buf); 369679787eaaSJeff Mahoney if (ret) { 36973f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 3698b64119b5SFilipe Manana btrfs_abort_transaction(trans, ret); 36993a45bb20SJeff Mahoney btrfs_end_transaction(trans); 370079787eaaSJeff Mahoney goto out; 370179787eaaSJeff Mahoney } 3702c5c9cd4dSSage Weil } 3703c5c9cd4dSSage Weil 3704b3b4aa74SDavid Sterba btrfs_release_path(path); 3705c5c9cd4dSSage Weil 370662e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 37070b246afaSJeff Mahoney fs_info->sectorsize); 3708b64119b5SFilipe Manana ret = clone_finish_inode_update(trans, inode, last_dest_end, 3709b64119b5SFilipe Manana destoff, olen, no_time_update); 3710f82a9901SFilipe Manana if (ret) 371179787eaaSJeff Mahoney goto out; 37122c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 37132c463823SFilipe Manana break; 3714b64119b5SFilipe Manana 3715b3b4aa74SDavid Sterba btrfs_release_path(path); 3716df858e76SFilipe Manana key.offset = next_key_min_offset; 371769ae5e44SWang Xiaoguang 371869ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 371969ae5e44SWang Xiaoguang ret = -EINTR; 372069ae5e44SWang Xiaoguang goto out; 372169ae5e44SWang Xiaoguang } 3722ae01a0abSYan Zheng } 3723f46b5a66SChristoph Hellwig ret = 0; 372432b7c687SMark Fasheh 3725f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3726690a5dbfSFilipe Manana struct btrfs_clone_extent_info clone_info = { 0 }; 3727f82a9901SFilipe Manana /* 3728f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3729f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3730f82a9901SFilipe Manana */ 3731b3b4aa74SDavid Sterba btrfs_release_path(path); 3732690a5dbfSFilipe Manana path->leave_spinning = 0; 3733f82a9901SFilipe Manana 3734f82a9901SFilipe Manana /* 3735690a5dbfSFilipe Manana * We are dealing with a hole and our clone_info already has a 3736690a5dbfSFilipe Manana * disk_offset of 0, we only need to fill the data length and 3737690a5dbfSFilipe Manana * file offset. 3738f82a9901SFilipe Manana */ 3739690a5dbfSFilipe Manana clone_info.data_len = destoff + len - last_dest_end; 3740690a5dbfSFilipe Manana clone_info.file_offset = last_dest_end; 3741690a5dbfSFilipe Manana ret = btrfs_punch_hole_range(inode, path, 3742690a5dbfSFilipe Manana last_dest_end, destoff + len - 1, 3743690a5dbfSFilipe Manana &clone_info, &trans); 3744690a5dbfSFilipe Manana if (ret) 3745f82a9901SFilipe Manana goto out; 3746690a5dbfSFilipe Manana 3747f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 37481c919a5eSMark Fasheh destoff, olen, no_time_update); 3749f82a9901SFilipe Manana } 3750f82a9901SFilipe Manana 3751f46b5a66SChristoph Hellwig out: 375232b7c687SMark Fasheh btrfs_free_path(path); 375315351955SDavid Sterba kvfree(buf); 375432b7c687SMark Fasheh return ret; 375532b7c687SMark Fasheh } 375632b7c687SMark Fasheh 37573db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 375832b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 375932b7c687SMark Fasheh { 376054563d41SAl Viro struct inode *inode = file_inode(file); 37613db11b2eSZach Brown struct inode *src = file_inode(file_src); 37620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 376332b7c687SMark Fasheh int ret; 376432b7c687SMark Fasheh u64 len = olen; 37650b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 376632b7c687SMark Fasheh 376732b7c687SMark Fasheh /* 376832b7c687SMark Fasheh * TODO: 376932b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 377032b7c687SMark Fasheh * decompress into destination's address_space (the file offset 377132b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 377232b7c687SMark Fasheh * otherwise reinsert) a subrange. 377300fdf13aSLiu Bo * 377400fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 377500fdf13aSLiu Bo * be either compressed or non-compressed. 377632b7c687SMark Fasheh */ 377732b7c687SMark Fasheh 3778ac765f83SFilipe Manana /* 377934a28e3dSFilipe Manana * VFS's generic_remap_file_range_prep() protects us from cloning the 378034a28e3dSFilipe Manana * eof block into the middle of a file, which would result in corruption 378134a28e3dSFilipe Manana * if the file size is not blocksize aligned. So we don't need to check 378234a28e3dSFilipe Manana * for that case here. 3783ac765f83SFilipe Manana */ 378434a28e3dSFilipe Manana if (off + len == src->i_size) 378532b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 378632b7c687SMark Fasheh 378732b7c687SMark Fasheh if (destoff > inode->i_size) { 3788f7fa1107SFilipe Manana const u64 wb_start = ALIGN_DOWN(inode->i_size, bs); 3789f7fa1107SFilipe Manana 379032b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 379132b7c687SMark Fasheh if (ret) 379234a28e3dSFilipe Manana return ret; 3793f7fa1107SFilipe Manana /* 3794f7fa1107SFilipe Manana * We may have truncated the last block if the inode's size is 3795f7fa1107SFilipe Manana * not sector size aligned, so we need to wait for writeback to 3796f7fa1107SFilipe Manana * complete before proceeding further, otherwise we can race 3797f7fa1107SFilipe Manana * with cloning and attempt to increment a reference to an 3798f7fa1107SFilipe Manana * extent that no longer exists (writeback completed right after 3799f7fa1107SFilipe Manana * we found the previous extent covering eof and before we 3800f7fa1107SFilipe Manana * attempted to increment its reference count). 3801f7fa1107SFilipe Manana */ 3802f7fa1107SFilipe Manana ret = btrfs_wait_ordered_range(inode, wb_start, 3803f7fa1107SFilipe Manana destoff - wb_start); 3804f7fa1107SFilipe Manana if (ret) 3805f7fa1107SFilipe Manana return ret; 380632b7c687SMark Fasheh } 380732b7c687SMark Fasheh 3808c125b8bfSFilipe Manana /* 3809d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3810d8b55242SFilipe Manana * source range to serialize with relocation. 3811c125b8bfSFilipe Manana */ 3812d8b55242SFilipe Manana btrfs_double_extent_lock(src, off, inode, destoff, len); 38131c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 3814d8b55242SFilipe Manana btrfs_double_extent_unlock(src, off, inode, destoff, len); 3815c125b8bfSFilipe Manana /* 3816c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3817c125b8bfSFilipe Manana * data immediately and not the previous data. 3818c125b8bfSFilipe Manana */ 381965bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 382009cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 382109cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 382234a28e3dSFilipe Manana 382334a28e3dSFilipe Manana return ret; 382434a28e3dSFilipe Manana } 382534a28e3dSFilipe Manana 382634a28e3dSFilipe Manana static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in, 382734a28e3dSFilipe Manana struct file *file_out, loff_t pos_out, 382834a28e3dSFilipe Manana loff_t *len, unsigned int remap_flags) 382934a28e3dSFilipe Manana { 383034a28e3dSFilipe Manana struct inode *inode_in = file_inode(file_in); 383134a28e3dSFilipe Manana struct inode *inode_out = file_inode(file_out); 383234a28e3dSFilipe Manana u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize; 383334a28e3dSFilipe Manana bool same_inode = inode_out == inode_in; 383434a28e3dSFilipe Manana u64 wb_len; 383534a28e3dSFilipe Manana int ret; 383634a28e3dSFilipe Manana 383734a28e3dSFilipe Manana if (!(remap_flags & REMAP_FILE_DEDUP)) { 383834a28e3dSFilipe Manana struct btrfs_root *root_out = BTRFS_I(inode_out)->root; 383934a28e3dSFilipe Manana 384034a28e3dSFilipe Manana if (btrfs_root_readonly(root_out)) 384134a28e3dSFilipe Manana return -EROFS; 384234a28e3dSFilipe Manana 384334a28e3dSFilipe Manana if (file_in->f_path.mnt != file_out->f_path.mnt || 384434a28e3dSFilipe Manana inode_in->i_sb != inode_out->i_sb) 384534a28e3dSFilipe Manana return -EXDEV; 384634a28e3dSFilipe Manana } 384734a28e3dSFilipe Manana 3848500710d3SFilipe Manana /* don't make the dst file partly checksummed */ 3849500710d3SFilipe Manana if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) != 3850500710d3SFilipe Manana (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) { 38517984ae52SGoldwyn Rodrigues return -EINVAL; 3852500710d3SFilipe Manana } 3853500710d3SFilipe Manana 385434a28e3dSFilipe Manana /* 385534a28e3dSFilipe Manana * Now that the inodes are locked, we need to start writeback ourselves 385634a28e3dSFilipe Manana * and can not rely on the writeback from the VFS's generic helper 385734a28e3dSFilipe Manana * generic_remap_file_range_prep() because: 385834a28e3dSFilipe Manana * 385934a28e3dSFilipe Manana * 1) For compression we must call filemap_fdatawrite_range() range 386034a28e3dSFilipe Manana * twice (btrfs_fdatawrite_range() does it for us), and the generic 386134a28e3dSFilipe Manana * helper only calls it once; 386234a28e3dSFilipe Manana * 386334a28e3dSFilipe Manana * 2) filemap_fdatawrite_range(), called by the generic helper only 386434a28e3dSFilipe Manana * waits for the writeback to complete, i.e. for IO to be done, and 386534a28e3dSFilipe Manana * not for the ordered extents to complete. We need to wait for them 386634a28e3dSFilipe Manana * to complete so that new file extent items are in the fs tree. 386734a28e3dSFilipe Manana */ 386834a28e3dSFilipe Manana if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP)) 386934a28e3dSFilipe Manana wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs); 387034a28e3dSFilipe Manana else 387134a28e3dSFilipe Manana wb_len = ALIGN(*len, bs); 387234a28e3dSFilipe Manana 387334a28e3dSFilipe Manana /* 387434a28e3dSFilipe Manana * Since we don't lock ranges, wait for ongoing lockless dio writes (as 387534a28e3dSFilipe Manana * any in progress could create its ordered extents after we wait for 387634a28e3dSFilipe Manana * existing ordered extents below). 387734a28e3dSFilipe Manana */ 387834a28e3dSFilipe Manana inode_dio_wait(inode_in); 387934a28e3dSFilipe Manana if (!same_inode) 388034a28e3dSFilipe Manana inode_dio_wait(inode_out); 388134a28e3dSFilipe Manana 3882a94d1d0cSQu Wenruo /* 3883a94d1d0cSQu Wenruo * Workaround to make sure NOCOW buffered write reach disk as NOCOW. 3884a94d1d0cSQu Wenruo * 3885a94d1d0cSQu Wenruo * Btrfs' back references do not have a block level granularity, they 3886a94d1d0cSQu Wenruo * work at the whole extent level. 3887a94d1d0cSQu Wenruo * NOCOW buffered write without data space reserved may not be able 3888a94d1d0cSQu Wenruo * to fall back to CoW due to lack of data space, thus could cause 3889a94d1d0cSQu Wenruo * data loss. 3890a94d1d0cSQu Wenruo * 3891a94d1d0cSQu Wenruo * Here we take a shortcut by flushing the whole inode, so that all 3892a94d1d0cSQu Wenruo * nocow write should reach disk as nocow before we increase the 3893a94d1d0cSQu Wenruo * reference of the extent. We could do better by only flushing NOCOW 3894a94d1d0cSQu Wenruo * data, but that needs extra accounting. 3895a94d1d0cSQu Wenruo * 3896a94d1d0cSQu Wenruo * Also we don't need to check ASYNC_EXTENT, as async extent will be 3897a94d1d0cSQu Wenruo * CoWed anyway, not affecting nocow part. 3898a94d1d0cSQu Wenruo */ 3899a94d1d0cSQu Wenruo ret = filemap_flush(inode_in->i_mapping); 3900a94d1d0cSQu Wenruo if (ret < 0) 3901a94d1d0cSQu Wenruo return ret; 3902a94d1d0cSQu Wenruo 390334a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs), 390434a28e3dSFilipe Manana wb_len); 390534a28e3dSFilipe Manana if (ret < 0) 39067984ae52SGoldwyn Rodrigues return ret; 390734a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs), 390834a28e3dSFilipe Manana wb_len); 390934a28e3dSFilipe Manana if (ret < 0) 39103db11b2eSZach Brown return ret; 39117984ae52SGoldwyn Rodrigues 39127984ae52SGoldwyn Rodrigues return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out, 39137984ae52SGoldwyn Rodrigues len, remap_flags); 39143db11b2eSZach Brown } 39153db11b2eSZach Brown 391642ec3d4cSDarrick J. Wong loff_t btrfs_remap_file_range(struct file *src_file, loff_t off, 391742ec3d4cSDarrick J. Wong struct file *dst_file, loff_t destoff, loff_t len, 39182e5dfc99SDarrick J. Wong unsigned int remap_flags) 39193db11b2eSZach Brown { 392034a28e3dSFilipe Manana struct inode *src_inode = file_inode(src_file); 392134a28e3dSFilipe Manana struct inode *dst_inode = file_inode(dst_file); 392234a28e3dSFilipe Manana bool same_inode = dst_inode == src_inode; 392342ec3d4cSDarrick J. Wong int ret; 392442ec3d4cSDarrick J. Wong 39252e5dfc99SDarrick J. Wong if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY)) 39262e5dfc99SDarrick J. Wong return -EINVAL; 39272e5dfc99SDarrick J. Wong 39287984ae52SGoldwyn Rodrigues if (same_inode) 39297984ae52SGoldwyn Rodrigues inode_lock(src_inode); 39307984ae52SGoldwyn Rodrigues else 39317984ae52SGoldwyn Rodrigues lock_two_nondirectories(src_inode, dst_inode); 39327984ae52SGoldwyn Rodrigues 393334a28e3dSFilipe Manana ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff, 393434a28e3dSFilipe Manana &len, remap_flags); 393534a28e3dSFilipe Manana if (ret < 0 || len == 0) 39367984ae52SGoldwyn Rodrigues goto out_unlock; 39372e5dfc99SDarrick J. Wong 393834a28e3dSFilipe Manana if (remap_flags & REMAP_FILE_DEDUP) 393934a28e3dSFilipe Manana ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff); 394034a28e3dSFilipe Manana else 394142ec3d4cSDarrick J. Wong ret = btrfs_clone_files(dst_file, src_file, off, len, destoff); 394234a28e3dSFilipe Manana 39437984ae52SGoldwyn Rodrigues out_unlock: 394434a28e3dSFilipe Manana if (same_inode) 394534a28e3dSFilipe Manana inode_unlock(src_inode); 394634a28e3dSFilipe Manana else 39474ea748e1SFilipe Manana unlock_two_nondirectories(src_inode, dst_inode); 394834a28e3dSFilipe Manana 394942ec3d4cSDarrick J. Wong return ret < 0 ? ret : len; 3950c5c9cd4dSSage Weil } 3951c5c9cd4dSSage Weil 39526ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 39536ef5ed0dSJosef Bacik { 3954496ad9aaSAl Viro struct inode *inode = file_inode(file); 39550b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 39566ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 39576ef5ed0dSJosef Bacik struct btrfs_root *new_root; 39586ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 39596ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 39606ef5ed0dSJosef Bacik struct btrfs_path *path; 39616ef5ed0dSJosef Bacik struct btrfs_key location; 39626ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 39636ef5ed0dSJosef Bacik u64 objectid = 0; 39646ef5ed0dSJosef Bacik u64 dir_id; 39653c04ce01SMiao Xie int ret; 39666ef5ed0dSJosef Bacik 39676ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 39686ef5ed0dSJosef Bacik return -EPERM; 39696ef5ed0dSJosef Bacik 39703c04ce01SMiao Xie ret = mnt_want_write_file(file); 39713c04ce01SMiao Xie if (ret) 39723c04ce01SMiao Xie return ret; 39733c04ce01SMiao Xie 39743c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 39753c04ce01SMiao Xie ret = -EFAULT; 39763c04ce01SMiao Xie goto out; 39773c04ce01SMiao Xie } 39786ef5ed0dSJosef Bacik 39796ef5ed0dSJosef Bacik if (!objectid) 39801cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 39816ef5ed0dSJosef Bacik 39826ef5ed0dSJosef Bacik location.objectid = objectid; 39836ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 39846ef5ed0dSJosef Bacik location.offset = (u64)-1; 39856ef5ed0dSJosef Bacik 39860b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 39873c04ce01SMiao Xie if (IS_ERR(new_root)) { 39883c04ce01SMiao Xie ret = PTR_ERR(new_root); 39893c04ce01SMiao Xie goto out; 39903c04ce01SMiao Xie } 39914fd786e6SMisono Tomohiro if (!is_fstree(new_root->root_key.objectid)) { 39926d6d2829Ssatoru takeuchi ret = -ENOENT; 39936d6d2829Ssatoru takeuchi goto out; 39946d6d2829Ssatoru takeuchi } 39956ef5ed0dSJosef Bacik 39966ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 39973c04ce01SMiao Xie if (!path) { 39983c04ce01SMiao Xie ret = -ENOMEM; 39993c04ce01SMiao Xie goto out; 40003c04ce01SMiao Xie } 40016ef5ed0dSJosef Bacik path->leave_spinning = 1; 40026ef5ed0dSJosef Bacik 40036ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 400498d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40056ef5ed0dSJosef Bacik btrfs_free_path(path); 40063c04ce01SMiao Xie ret = PTR_ERR(trans); 40073c04ce01SMiao Xie goto out; 40086ef5ed0dSJosef Bacik } 40096ef5ed0dSJosef Bacik 40100b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 40110b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 40126ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4013cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40146ef5ed0dSJosef Bacik btrfs_free_path(path); 40153a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40160b246afaSJeff Mahoney btrfs_err(fs_info, 40175d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 40183c04ce01SMiao Xie ret = -ENOENT; 40193c04ce01SMiao Xie goto out; 40206ef5ed0dSJosef Bacik } 40216ef5ed0dSJosef Bacik 40226ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 40236ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 40246ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 40256ef5ed0dSJosef Bacik btrfs_free_path(path); 40266ef5ed0dSJosef Bacik 40270b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 40283a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40293c04ce01SMiao Xie out: 40303c04ce01SMiao Xie mnt_drop_write_file(file); 40313c04ce01SMiao Xie return ret; 40326ef5ed0dSJosef Bacik } 40336ef5ed0dSJosef Bacik 4034c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4035bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4036bf5fc093SJosef Bacik { 4037bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4038bf5fc093SJosef Bacik 4039bf5fc093SJosef Bacik space->total_bytes = 0; 4040bf5fc093SJosef Bacik space->used_bytes = 0; 4041bf5fc093SJosef Bacik space->flags = 0; 4042bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4043bf5fc093SJosef Bacik space->flags = block_group->flags; 4044bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4045bf5fc093SJosef Bacik space->used_bytes += 4046bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4047bf5fc093SJosef Bacik } 4048bf5fc093SJosef Bacik } 4049bf5fc093SJosef Bacik 40502ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 40512ff7e61eSJeff Mahoney void __user *arg) 40521406e432SJosef Bacik { 40531406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 40541406e432SJosef Bacik struct btrfs_ioctl_space_info space; 40551406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 40567fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 405713f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 40581406e432SJosef Bacik struct btrfs_space_info *info; 4059315d8e98SColin Ian King static const u64 types[] = { 4060315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4061bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4062bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4063315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4064315d8e98SColin Ian King }; 4065bf5fc093SJosef Bacik int num_types = 4; 40667fde62bfSChris Mason int alloc_size; 40671406e432SJosef Bacik int ret = 0; 406851788b1bSDan Rosenberg u64 slot_count = 0; 4069bf5fc093SJosef Bacik int i, c; 40701406e432SJosef Bacik 40711406e432SJosef Bacik if (copy_from_user(&space_args, 40721406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 40731406e432SJosef Bacik sizeof(space_args))) 40741406e432SJosef Bacik return -EFAULT; 40751406e432SJosef Bacik 4076bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4077bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4078bf5fc093SJosef Bacik 4079bf5fc093SJosef Bacik info = NULL; 40807fde62bfSChris Mason rcu_read_lock(); 40810b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4082bf5fc093SJosef Bacik list) { 4083bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4084bf5fc093SJosef Bacik info = tmp; 4085bf5fc093SJosef Bacik break; 4086bf5fc093SJosef Bacik } 4087bf5fc093SJosef Bacik } 40887fde62bfSChris Mason rcu_read_unlock(); 40891406e432SJosef Bacik 4090bf5fc093SJosef Bacik if (!info) 4091bf5fc093SJosef Bacik continue; 4092bf5fc093SJosef Bacik 4093bf5fc093SJosef Bacik down_read(&info->groups_sem); 4094bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4095bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4096bf5fc093SJosef Bacik slot_count++; 4097bf5fc093SJosef Bacik } 4098bf5fc093SJosef Bacik up_read(&info->groups_sem); 4099bf5fc093SJosef Bacik } 4100bf5fc093SJosef Bacik 410136523e95SDavid Sterba /* 410236523e95SDavid Sterba * Global block reserve, exported as a space_info 410336523e95SDavid Sterba */ 410436523e95SDavid Sterba slot_count++; 410536523e95SDavid Sterba 41067fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 41077fde62bfSChris Mason if (space_args.space_slots == 0) { 41087fde62bfSChris Mason space_args.total_spaces = slot_count; 41097fde62bfSChris Mason goto out; 41107fde62bfSChris Mason } 4111bf5fc093SJosef Bacik 411251788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4113bf5fc093SJosef Bacik 41147fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4115bf5fc093SJosef Bacik 41167fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41177fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41187fde62bfSChris Mason */ 411909cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 41207fde62bfSChris Mason return -ENOMEM; 41217fde62bfSChris Mason 41227fde62bfSChris Mason space_args.total_spaces = 0; 41238d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 41247fde62bfSChris Mason if (!dest) 41257fde62bfSChris Mason return -ENOMEM; 41267fde62bfSChris Mason dest_orig = dest; 41277fde62bfSChris Mason 41287fde62bfSChris Mason /* now we have a buffer to copy into */ 4129bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4130bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4131bf5fc093SJosef Bacik 413251788b1bSDan Rosenberg if (!slot_count) 413351788b1bSDan Rosenberg break; 413451788b1bSDan Rosenberg 4135bf5fc093SJosef Bacik info = NULL; 41361406e432SJosef Bacik rcu_read_lock(); 41370b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4138bf5fc093SJosef Bacik list) { 4139bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4140bf5fc093SJosef Bacik info = tmp; 41417fde62bfSChris Mason break; 4142bf5fc093SJosef Bacik } 4143bf5fc093SJosef Bacik } 4144bf5fc093SJosef Bacik rcu_read_unlock(); 41457fde62bfSChris Mason 4146bf5fc093SJosef Bacik if (!info) 4147bf5fc093SJosef Bacik continue; 4148bf5fc093SJosef Bacik down_read(&info->groups_sem); 4149bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4150bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4151c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4152c065f5b1SSu Yue &space); 41537fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 41541406e432SJosef Bacik dest++; 41551406e432SJosef Bacik space_args.total_spaces++; 415651788b1bSDan Rosenberg slot_count--; 41571406e432SJosef Bacik } 415851788b1bSDan Rosenberg if (!slot_count) 415951788b1bSDan Rosenberg break; 4160bf5fc093SJosef Bacik } 4161bf5fc093SJosef Bacik up_read(&info->groups_sem); 4162bf5fc093SJosef Bacik } 41631406e432SJosef Bacik 416436523e95SDavid Sterba /* 416536523e95SDavid Sterba * Add global block reserve 416636523e95SDavid Sterba */ 416736523e95SDavid Sterba if (slot_count) { 41680b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 416936523e95SDavid Sterba 417036523e95SDavid Sterba spin_lock(&block_rsv->lock); 417136523e95SDavid Sterba space.total_bytes = block_rsv->size; 417236523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 417336523e95SDavid Sterba spin_unlock(&block_rsv->lock); 417436523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 417536523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 417636523e95SDavid Sterba space_args.total_spaces++; 417736523e95SDavid Sterba } 417836523e95SDavid Sterba 41792eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 41807fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 41817fde62bfSChris Mason 41827fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 41837fde62bfSChris Mason ret = -EFAULT; 41847fde62bfSChris Mason 41857fde62bfSChris Mason kfree(dest_orig); 41867fde62bfSChris Mason out: 41877fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 41881406e432SJosef Bacik ret = -EFAULT; 41891406e432SJosef Bacik 41901406e432SJosef Bacik return ret; 41911406e432SJosef Bacik } 41921406e432SJosef Bacik 41939a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 41949a8c28beSMiao Xie void __user *argp) 419546204592SSage Weil { 419646204592SSage Weil struct btrfs_trans_handle *trans; 419746204592SSage Weil u64 transid; 4198db5b493aSTsutomu Itoh int ret; 419946204592SSage Weil 4200ebc87351SNikolay Borisov btrfs_warn(root->fs_info, 4201ebc87351SNikolay Borisov "START_SYNC ioctl is deprecated and will be removed in kernel 5.7"); 4202ebc87351SNikolay Borisov 4203d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4204ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4205ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 420698d5dc13STsutomu Itoh return PTR_ERR(trans); 4207ff7c1d33SMiao Xie 4208ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4209ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4210ff7c1d33SMiao Xie goto out; 4211ff7c1d33SMiao Xie } 421246204592SSage Weil transid = trans->transid; 42133a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 42148b2b2d3cSTsutomu Itoh if (ret) { 42153a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4216db5b493aSTsutomu Itoh return ret; 42178b2b2d3cSTsutomu Itoh } 4218ff7c1d33SMiao Xie out: 421946204592SSage Weil if (argp) 422046204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 422146204592SSage Weil return -EFAULT; 422246204592SSage Weil return 0; 422346204592SSage Weil } 422446204592SSage Weil 42252ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 42269a8c28beSMiao Xie void __user *argp) 422746204592SSage Weil { 422846204592SSage Weil u64 transid; 422946204592SSage Weil 4230ebc87351SNikolay Borisov btrfs_warn(fs_info, 4231ebc87351SNikolay Borisov "WAIT_SYNC ioctl is deprecated and will be removed in kernel 5.7"); 4232ebc87351SNikolay Borisov 423346204592SSage Weil if (argp) { 423446204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 423546204592SSage Weil return -EFAULT; 423646204592SSage Weil } else { 423746204592SSage Weil transid = 0; /* current trans */ 423846204592SSage Weil } 42392ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 424046204592SSage Weil } 424146204592SSage Weil 4242b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4243475f6387SJan Schmidt { 42440b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4245475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4246b8e95489SMiao Xie int ret; 4247475f6387SJan Schmidt 4248475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4249475f6387SJan Schmidt return -EPERM; 4250475f6387SJan Schmidt 4251475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4252475f6387SJan Schmidt if (IS_ERR(sa)) 4253475f6387SJan Schmidt return PTR_ERR(sa); 4254475f6387SJan Schmidt 4255b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4256b8e95489SMiao Xie ret = mnt_want_write_file(file); 4257b8e95489SMiao Xie if (ret) 4258b8e95489SMiao Xie goto out; 4259b8e95489SMiao Xie } 4260b8e95489SMiao Xie 42610b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 426263a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 426363a212abSStefan Behrens 0); 4264475f6387SJan Schmidt 426506fe39abSFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4266475f6387SJan Schmidt ret = -EFAULT; 4267475f6387SJan Schmidt 4268b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4269b8e95489SMiao Xie mnt_drop_write_file(file); 4270b8e95489SMiao Xie out: 4271475f6387SJan Schmidt kfree(sa); 4272475f6387SJan Schmidt return ret; 4273475f6387SJan Schmidt } 4274475f6387SJan Schmidt 42752ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4276475f6387SJan Schmidt { 4277475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4278475f6387SJan Schmidt return -EPERM; 4279475f6387SJan Schmidt 42802ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4281475f6387SJan Schmidt } 4282475f6387SJan Schmidt 42832ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4284475f6387SJan Schmidt void __user *arg) 4285475f6387SJan Schmidt { 4286475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4287475f6387SJan Schmidt int ret; 4288475f6387SJan Schmidt 4289475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4290475f6387SJan Schmidt return -EPERM; 4291475f6387SJan Schmidt 4292475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4293475f6387SJan Schmidt if (IS_ERR(sa)) 4294475f6387SJan Schmidt return PTR_ERR(sa); 4295475f6387SJan Schmidt 42962ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4297475f6387SJan Schmidt 42984fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4299475f6387SJan Schmidt ret = -EFAULT; 4300475f6387SJan Schmidt 4301475f6387SJan Schmidt kfree(sa); 4302475f6387SJan Schmidt return ret; 4303475f6387SJan Schmidt } 4304475f6387SJan Schmidt 43052ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4306b27f7c0cSDavid Sterba void __user *arg) 4307c11d2c23SStefan Behrens { 4308c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4309c11d2c23SStefan Behrens int ret; 4310c11d2c23SStefan Behrens 4311c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4312c11d2c23SStefan Behrens if (IS_ERR(sa)) 4313c11d2c23SStefan Behrens return PTR_ERR(sa); 4314c11d2c23SStefan Behrens 4315b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4316b27f7c0cSDavid Sterba kfree(sa); 4317b27f7c0cSDavid Sterba return -EPERM; 4318b27f7c0cSDavid Sterba } 4319b27f7c0cSDavid Sterba 43202ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4321c11d2c23SStefan Behrens 4322eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4323c11d2c23SStefan Behrens ret = -EFAULT; 4324c11d2c23SStefan Behrens 4325c11d2c23SStefan Behrens kfree(sa); 4326c11d2c23SStefan Behrens return ret; 4327c11d2c23SStefan Behrens } 4328c11d2c23SStefan Behrens 43292ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 43302ff7e61eSJeff Mahoney void __user *arg) 43313f6bcfbdSStefan Behrens { 43323f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 43333f6bcfbdSStefan Behrens int ret; 43343f6bcfbdSStefan Behrens 43353f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 43363f6bcfbdSStefan Behrens return -EPERM; 43373f6bcfbdSStefan Behrens 43383f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 43393f6bcfbdSStefan Behrens if (IS_ERR(p)) 43403f6bcfbdSStefan Behrens return PTR_ERR(p); 43413f6bcfbdSStefan Behrens 43423f6bcfbdSStefan Behrens switch (p->cmd) { 43433f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4344bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4345adfa97cbSIlya Dryomov ret = -EROFS; 4346adfa97cbSIlya Dryomov goto out; 4347adfa97cbSIlya Dryomov } 4348171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4349e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 43503f6bcfbdSStefan Behrens } else { 43512ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4352171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 43533f6bcfbdSStefan Behrens } 43543f6bcfbdSStefan Behrens break; 43553f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 43560b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 43573f6bcfbdSStefan Behrens ret = 0; 43583f6bcfbdSStefan Behrens break; 43593f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 436017d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 436197282031SAnand Jain ret = 0; 43623f6bcfbdSStefan Behrens break; 43633f6bcfbdSStefan Behrens default: 43643f6bcfbdSStefan Behrens ret = -EINVAL; 43653f6bcfbdSStefan Behrens break; 43663f6bcfbdSStefan Behrens } 43673f6bcfbdSStefan Behrens 4368d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 43693f6bcfbdSStefan Behrens ret = -EFAULT; 4370adfa97cbSIlya Dryomov out: 43713f6bcfbdSStefan Behrens kfree(p); 43723f6bcfbdSStefan Behrens return ret; 43733f6bcfbdSStefan Behrens } 43743f6bcfbdSStefan Behrens 4375d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4376d7728c96SJan Schmidt { 4377d7728c96SJan Schmidt int ret = 0; 4378d7728c96SJan Schmidt int i; 4379740c3d22SChris Mason u64 rel_ptr; 4380d7728c96SJan Schmidt int size; 4381806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4382d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4383d7728c96SJan Schmidt struct btrfs_path *path; 4384d7728c96SJan Schmidt 438582b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4386d7728c96SJan Schmidt return -EPERM; 4387d7728c96SJan Schmidt 4388d7728c96SJan Schmidt path = btrfs_alloc_path(); 4389d7728c96SJan Schmidt if (!path) { 4390d7728c96SJan Schmidt ret = -ENOMEM; 4391d7728c96SJan Schmidt goto out; 4392d7728c96SJan Schmidt } 4393d7728c96SJan Schmidt 4394d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4395d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4396d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4397d7728c96SJan Schmidt ipa = NULL; 4398d7728c96SJan Schmidt goto out; 4399d7728c96SJan Schmidt } 4400d7728c96SJan Schmidt 4401d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4402d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4403d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4404d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4405d7728c96SJan Schmidt ipath = NULL; 4406d7728c96SJan Schmidt goto out; 4407d7728c96SJan Schmidt } 4408d7728c96SJan Schmidt 4409d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4410d7728c96SJan Schmidt if (ret < 0) 4411d7728c96SJan Schmidt goto out; 4412d7728c96SJan Schmidt 4413d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4414745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4415745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4416740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4417d7728c96SJan Schmidt } 4418d7728c96SJan Schmidt 4419718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4420718dc5faSOmar Sandoval ipath->fspath, size); 4421d7728c96SJan Schmidt if (ret) { 4422d7728c96SJan Schmidt ret = -EFAULT; 4423d7728c96SJan Schmidt goto out; 4424d7728c96SJan Schmidt } 4425d7728c96SJan Schmidt 4426d7728c96SJan Schmidt out: 4427d7728c96SJan Schmidt btrfs_free_path(path); 4428d7728c96SJan Schmidt free_ipath(ipath); 4429d7728c96SJan Schmidt kfree(ipa); 4430d7728c96SJan Schmidt 4431d7728c96SJan Schmidt return ret; 4432d7728c96SJan Schmidt } 4433d7728c96SJan Schmidt 4434d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4435d7728c96SJan Schmidt { 4436d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4437d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4438d7728c96SJan Schmidt 4439d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4440d7728c96SJan Schmidt inodes->bytes_left -= c; 4441d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4442d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4443d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4444d7728c96SJan Schmidt inodes->elem_cnt += 3; 4445d7728c96SJan Schmidt } else { 4446d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4447d7728c96SJan Schmidt inodes->bytes_left = 0; 4448d7728c96SJan Schmidt inodes->elem_missed += 3; 4449d7728c96SJan Schmidt } 4450d7728c96SJan Schmidt 4451d7728c96SJan Schmidt return 0; 4452d7728c96SJan Schmidt } 4453d7728c96SJan Schmidt 44542ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4455d24a67b2SZygo Blaxell void __user *arg, int version) 4456d7728c96SJan Schmidt { 4457d7728c96SJan Schmidt int ret = 0; 4458d7728c96SJan Schmidt int size; 4459d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4460d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4461d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4462d24a67b2SZygo Blaxell bool ignore_offset; 4463d7728c96SJan Schmidt 4464d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4465d7728c96SJan Schmidt return -EPERM; 4466d7728c96SJan Schmidt 4467d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 44687b9ea627SShailendra Verma if (IS_ERR(loi)) 44697b9ea627SShailendra Verma return PTR_ERR(loi); 4470d7728c96SJan Schmidt 4471d24a67b2SZygo Blaxell if (version == 1) { 4472d24a67b2SZygo Blaxell ignore_offset = false; 4473b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4474d24a67b2SZygo Blaxell } else { 4475d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4476d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4477d24a67b2SZygo Blaxell ret = -EINVAL; 4478d24a67b2SZygo Blaxell goto out_loi; 4479d24a67b2SZygo Blaxell } 4480d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4481d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4482d24a67b2SZygo Blaxell ret = -EINVAL; 4483d24a67b2SZygo Blaxell goto out_loi; 4484d24a67b2SZygo Blaxell } 4485d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4486b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4487d24a67b2SZygo Blaxell } 4488d24a67b2SZygo Blaxell 4489d7728c96SJan Schmidt path = btrfs_alloc_path(); 4490d7728c96SJan Schmidt if (!path) { 4491d7728c96SJan Schmidt ret = -ENOMEM; 4492d7728c96SJan Schmidt goto out; 4493d7728c96SJan Schmidt } 4494d7728c96SJan Schmidt 4495d7728c96SJan Schmidt inodes = init_data_container(size); 4496d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4497d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4498d7728c96SJan Schmidt inodes = NULL; 4499d7728c96SJan Schmidt goto out; 4500d7728c96SJan Schmidt } 4501d7728c96SJan Schmidt 45022ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4503d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4504df031f07SLiu Bo if (ret == -EINVAL) 4505d7728c96SJan Schmidt ret = -ENOENT; 4506d7728c96SJan Schmidt if (ret < 0) 4507d7728c96SJan Schmidt goto out; 4508d7728c96SJan Schmidt 4509718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4510718dc5faSOmar Sandoval size); 4511d7728c96SJan Schmidt if (ret) 4512d7728c96SJan Schmidt ret = -EFAULT; 4513d7728c96SJan Schmidt 4514d7728c96SJan Schmidt out: 4515d7728c96SJan Schmidt btrfs_free_path(path); 4516f54de068SDavid Sterba kvfree(inodes); 4517d24a67b2SZygo Blaxell out_loi: 4518d7728c96SJan Schmidt kfree(loi); 4519d7728c96SJan Schmidt 4520d7728c96SJan Schmidt return ret; 4521d7728c96SJan Schmidt } 4522d7728c96SJan Schmidt 4523008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4524c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4525c9e9f97bSIlya Dryomov { 4526c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4527c9e9f97bSIlya Dryomov 4528c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4529c9e9f97bSIlya Dryomov 45303009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4531837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4532837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4533837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4534a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4535a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4536837d5b6eSIlya Dryomov 4537c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4538c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4539c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 454019a39dceSIlya Dryomov 454119a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 454219a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 454319a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4544c9e9f97bSIlya Dryomov } 4545c9e9f97bSIlya Dryomov 45469ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4547c9e9f97bSIlya Dryomov { 4548496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4549c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4550c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4551c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4552ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4553c9e9f97bSIlya Dryomov int ret; 4554c9e9f97bSIlya Dryomov 4555c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4556c9e9f97bSIlya Dryomov return -EPERM; 4557c9e9f97bSIlya Dryomov 4558e54bfa31SLiu Bo ret = mnt_want_write_file(file); 45599ba1f6e4SLiu Bo if (ret) 45609ba1f6e4SLiu Bo return ret; 45619ba1f6e4SLiu Bo 4562ed0fb78fSIlya Dryomov again: 4563171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4564c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4565ed0fb78fSIlya Dryomov need_unlock = true; 4566ed0fb78fSIlya Dryomov goto locked; 4567ed0fb78fSIlya Dryomov } 4568ed0fb78fSIlya Dryomov 4569ed0fb78fSIlya Dryomov /* 457001327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4571ed0fb78fSIlya Dryomov * (1) some other op is running 4572ed0fb78fSIlya Dryomov * (2) balance is running 4573ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4574ed0fb78fSIlya Dryomov */ 4575ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4576ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4577ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 45783009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4579ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4580dccdb07bSDavid Sterba /* 4581dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4582dccdb07bSDavid Sterba * we'll reexamine the status again. 4583dccdb07bSDavid Sterba */ 4584ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4585ed0fb78fSIlya Dryomov 4586ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 45873009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4588ed0fb78fSIlya Dryomov /* this is (3) */ 4589ed0fb78fSIlya Dryomov need_unlock = false; 4590ed0fb78fSIlya Dryomov goto locked; 4591ed0fb78fSIlya Dryomov } 4592ed0fb78fSIlya Dryomov 4593ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4594ed0fb78fSIlya Dryomov goto again; 4595ed0fb78fSIlya Dryomov } else { 4596ed0fb78fSIlya Dryomov /* this is (2) */ 4597ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4598ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4599ed0fb78fSIlya Dryomov goto out; 4600ed0fb78fSIlya Dryomov } 4601ed0fb78fSIlya Dryomov } else { 4602ed0fb78fSIlya Dryomov /* this is (1) */ 4603ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4604e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4605ed0fb78fSIlya Dryomov goto out; 4606ed0fb78fSIlya Dryomov } 4607ed0fb78fSIlya Dryomov 4608ed0fb78fSIlya Dryomov locked: 4609171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4610c9e9f97bSIlya Dryomov 4611c9e9f97bSIlya Dryomov if (arg) { 4612c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4613c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4614c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4615ed0fb78fSIlya Dryomov goto out_unlock; 4616c9e9f97bSIlya Dryomov } 4617de322263SIlya Dryomov 4618de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4619de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4620de322263SIlya Dryomov ret = -ENOTCONN; 4621de322263SIlya Dryomov goto out_bargs; 4622de322263SIlya Dryomov } 4623de322263SIlya Dryomov 4624de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4625de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4626de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4627de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4628de322263SIlya Dryomov 4629de322263SIlya Dryomov goto do_balance; 4630de322263SIlya Dryomov } 4631c9e9f97bSIlya Dryomov } else { 4632c9e9f97bSIlya Dryomov bargs = NULL; 4633c9e9f97bSIlya Dryomov } 4634c9e9f97bSIlya Dryomov 4635ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4636837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4637837d5b6eSIlya Dryomov goto out_bargs; 4638837d5b6eSIlya Dryomov } 4639837d5b6eSIlya Dryomov 46408d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4641c9e9f97bSIlya Dryomov if (!bctl) { 4642c9e9f97bSIlya Dryomov ret = -ENOMEM; 4643c9e9f97bSIlya Dryomov goto out_bargs; 4644c9e9f97bSIlya Dryomov } 4645c9e9f97bSIlya Dryomov 4646c9e9f97bSIlya Dryomov if (arg) { 4647c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4648c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4649c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4650c9e9f97bSIlya Dryomov 4651c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4652f43ffb60SIlya Dryomov } else { 4653f43ffb60SIlya Dryomov /* balance everything - no filters */ 4654f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4655c9e9f97bSIlya Dryomov } 4656c9e9f97bSIlya Dryomov 46578eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 46588eb93459SDavid Sterba ret = -EINVAL; 46590f89abf5SChristian Engelmayer goto out_bctl; 46608eb93459SDavid Sterba } 46618eb93459SDavid Sterba 4662de322263SIlya Dryomov do_balance: 4663c9e9f97bSIlya Dryomov /* 4664149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4665149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4666149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4667149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4668c9e9f97bSIlya Dryomov */ 4669ed0fb78fSIlya Dryomov need_unlock = false; 4670ed0fb78fSIlya Dryomov 46716fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 46720f89abf5SChristian Engelmayer bctl = NULL; 4673ed0fb78fSIlya Dryomov 4674d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4675c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4676c9e9f97bSIlya Dryomov ret = -EFAULT; 4677c9e9f97bSIlya Dryomov } 4678c9e9f97bSIlya Dryomov 46790f89abf5SChristian Engelmayer out_bctl: 46800f89abf5SChristian Engelmayer kfree(bctl); 4681c9e9f97bSIlya Dryomov out_bargs: 4682c9e9f97bSIlya Dryomov kfree(bargs); 4683ed0fb78fSIlya Dryomov out_unlock: 4684c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4685ed0fb78fSIlya Dryomov if (need_unlock) 4686171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4687ed0fb78fSIlya Dryomov out: 4688e54bfa31SLiu Bo mnt_drop_write_file(file); 4689c9e9f97bSIlya Dryomov return ret; 4690c9e9f97bSIlya Dryomov } 4691c9e9f97bSIlya Dryomov 46922ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4693837d5b6eSIlya Dryomov { 4694837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4695837d5b6eSIlya Dryomov return -EPERM; 4696837d5b6eSIlya Dryomov 4697837d5b6eSIlya Dryomov switch (cmd) { 4698837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 46990b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4700a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47010b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4702837d5b6eSIlya Dryomov } 4703837d5b6eSIlya Dryomov 4704837d5b6eSIlya Dryomov return -EINVAL; 4705837d5b6eSIlya Dryomov } 4706837d5b6eSIlya Dryomov 47072ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 470819a39dceSIlya Dryomov void __user *arg) 470919a39dceSIlya Dryomov { 471019a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 471119a39dceSIlya Dryomov int ret = 0; 471219a39dceSIlya Dryomov 471319a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 471419a39dceSIlya Dryomov return -EPERM; 471519a39dceSIlya Dryomov 471619a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 471719a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 471819a39dceSIlya Dryomov ret = -ENOTCONN; 471919a39dceSIlya Dryomov goto out; 472019a39dceSIlya Dryomov } 472119a39dceSIlya Dryomov 47228d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 472319a39dceSIlya Dryomov if (!bargs) { 472419a39dceSIlya Dryomov ret = -ENOMEM; 472519a39dceSIlya Dryomov goto out; 472619a39dceSIlya Dryomov } 472719a39dceSIlya Dryomov 4728008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 472919a39dceSIlya Dryomov 473019a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 473119a39dceSIlya Dryomov ret = -EFAULT; 473219a39dceSIlya Dryomov 473319a39dceSIlya Dryomov kfree(bargs); 473419a39dceSIlya Dryomov out: 473519a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 473619a39dceSIlya Dryomov return ret; 473719a39dceSIlya Dryomov } 473819a39dceSIlya Dryomov 4739905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 47405d13a37bSArne Jansen { 47410b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47420b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47435d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 47445d13a37bSArne Jansen int ret; 47455d13a37bSArne Jansen 47465d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47475d13a37bSArne Jansen return -EPERM; 47485d13a37bSArne Jansen 4749905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4750905b0ddaSMiao Xie if (ret) 4751905b0ddaSMiao Xie return ret; 47525d13a37bSArne Jansen 47535d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4754905b0ddaSMiao Xie if (IS_ERR(sa)) { 4755905b0ddaSMiao Xie ret = PTR_ERR(sa); 4756905b0ddaSMiao Xie goto drop_write; 4757905b0ddaSMiao Xie } 47585d13a37bSArne Jansen 47590b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 47605d13a37bSArne Jansen 47615d13a37bSArne Jansen switch (sa->cmd) { 47625d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4763340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 47645d13a37bSArne Jansen break; 47655d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4766340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 47675d13a37bSArne Jansen break; 47685d13a37bSArne Jansen default: 47695d13a37bSArne Jansen ret = -EINVAL; 47705d13a37bSArne Jansen break; 47715d13a37bSArne Jansen } 47725d13a37bSArne Jansen 47735d13a37bSArne Jansen kfree(sa); 47740b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4775905b0ddaSMiao Xie drop_write: 4776905b0ddaSMiao Xie mnt_drop_write_file(file); 47775d13a37bSArne Jansen return ret; 47785d13a37bSArne Jansen } 47795d13a37bSArne Jansen 4780905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 47815d13a37bSArne Jansen { 47820b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47830b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47840b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 47855d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 47865d13a37bSArne Jansen struct btrfs_trans_handle *trans; 47875d13a37bSArne Jansen int ret; 47885d13a37bSArne Jansen int err; 47895d13a37bSArne Jansen 47905d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47915d13a37bSArne Jansen return -EPERM; 47925d13a37bSArne Jansen 4793905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4794905b0ddaSMiao Xie if (ret) 4795905b0ddaSMiao Xie return ret; 47965d13a37bSArne Jansen 47975d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4798905b0ddaSMiao Xie if (IS_ERR(sa)) { 4799905b0ddaSMiao Xie ret = PTR_ERR(sa); 4800905b0ddaSMiao Xie goto drop_write; 4801905b0ddaSMiao Xie } 48025d13a37bSArne Jansen 48035d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48045d13a37bSArne Jansen if (IS_ERR(trans)) { 48055d13a37bSArne Jansen ret = PTR_ERR(trans); 48065d13a37bSArne Jansen goto out; 48075d13a37bSArne Jansen } 48085d13a37bSArne Jansen 48095d13a37bSArne Jansen if (sa->assign) { 48109f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 48115d13a37bSArne Jansen } else { 481239616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 48135d13a37bSArne Jansen } 48145d13a37bSArne Jansen 4815e082f563SQu Wenruo /* update qgroup status and info */ 4816280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4817e082f563SQu Wenruo if (err < 0) 48180b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4819ad8403dfSAnand Jain "failed to update qgroup status and info"); 48203a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48215d13a37bSArne Jansen if (err && !ret) 48225d13a37bSArne Jansen ret = err; 48235d13a37bSArne Jansen 48245d13a37bSArne Jansen out: 48255d13a37bSArne Jansen kfree(sa); 4826905b0ddaSMiao Xie drop_write: 4827905b0ddaSMiao Xie mnt_drop_write_file(file); 48285d13a37bSArne Jansen return ret; 48295d13a37bSArne Jansen } 48305d13a37bSArne Jansen 4831905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 48325d13a37bSArne Jansen { 48330b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48340b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48355d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 48365d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48375d13a37bSArne Jansen int ret; 48385d13a37bSArne Jansen int err; 48395d13a37bSArne Jansen 48405d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48415d13a37bSArne Jansen return -EPERM; 48425d13a37bSArne Jansen 4843905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4844905b0ddaSMiao Xie if (ret) 4845905b0ddaSMiao Xie return ret; 48465d13a37bSArne Jansen 48475d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4848905b0ddaSMiao Xie if (IS_ERR(sa)) { 4849905b0ddaSMiao Xie ret = PTR_ERR(sa); 4850905b0ddaSMiao Xie goto drop_write; 4851905b0ddaSMiao Xie } 48525d13a37bSArne Jansen 4853d86e56cfSMiao Xie if (!sa->qgroupid) { 4854d86e56cfSMiao Xie ret = -EINVAL; 4855d86e56cfSMiao Xie goto out; 4856d86e56cfSMiao Xie } 4857d86e56cfSMiao Xie 48585d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48595d13a37bSArne Jansen if (IS_ERR(trans)) { 48605d13a37bSArne Jansen ret = PTR_ERR(trans); 48615d13a37bSArne Jansen goto out; 48625d13a37bSArne Jansen } 48635d13a37bSArne Jansen 48645d13a37bSArne Jansen if (sa->create) { 486549a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 48665d13a37bSArne Jansen } else { 48673efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 48685d13a37bSArne Jansen } 48695d13a37bSArne Jansen 48703a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48715d13a37bSArne Jansen if (err && !ret) 48725d13a37bSArne Jansen ret = err; 48735d13a37bSArne Jansen 48745d13a37bSArne Jansen out: 48755d13a37bSArne Jansen kfree(sa); 4876905b0ddaSMiao Xie drop_write: 4877905b0ddaSMiao Xie mnt_drop_write_file(file); 48785d13a37bSArne Jansen return ret; 48795d13a37bSArne Jansen } 48805d13a37bSArne Jansen 4881905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 48825d13a37bSArne Jansen { 48830b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48840b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48855d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 48865d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48875d13a37bSArne Jansen int ret; 48885d13a37bSArne Jansen int err; 48895d13a37bSArne Jansen u64 qgroupid; 48905d13a37bSArne Jansen 48915d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48925d13a37bSArne Jansen return -EPERM; 48935d13a37bSArne Jansen 4894905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4895905b0ddaSMiao Xie if (ret) 4896905b0ddaSMiao Xie return ret; 48975d13a37bSArne Jansen 48985d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4899905b0ddaSMiao Xie if (IS_ERR(sa)) { 4900905b0ddaSMiao Xie ret = PTR_ERR(sa); 4901905b0ddaSMiao Xie goto drop_write; 4902905b0ddaSMiao Xie } 49035d13a37bSArne Jansen 49045d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49055d13a37bSArne Jansen if (IS_ERR(trans)) { 49065d13a37bSArne Jansen ret = PTR_ERR(trans); 49075d13a37bSArne Jansen goto out; 49085d13a37bSArne Jansen } 49095d13a37bSArne Jansen 49105d13a37bSArne Jansen qgroupid = sa->qgroupid; 49115d13a37bSArne Jansen if (!qgroupid) { 49125d13a37bSArne Jansen /* take the current subvol as qgroup */ 49135d13a37bSArne Jansen qgroupid = root->root_key.objectid; 49145d13a37bSArne Jansen } 49155d13a37bSArne Jansen 4916f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 49175d13a37bSArne Jansen 49183a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49195d13a37bSArne Jansen if (err && !ret) 49205d13a37bSArne Jansen ret = err; 49215d13a37bSArne Jansen 49225d13a37bSArne Jansen out: 49235d13a37bSArne Jansen kfree(sa); 4924905b0ddaSMiao Xie drop_write: 4925905b0ddaSMiao Xie mnt_drop_write_file(file); 49265d13a37bSArne Jansen return ret; 49275d13a37bSArne Jansen } 49285d13a37bSArne Jansen 49292f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 49302f232036SJan Schmidt { 49310b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49332f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49342f232036SJan Schmidt int ret; 49352f232036SJan Schmidt 49362f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49372f232036SJan Schmidt return -EPERM; 49382f232036SJan Schmidt 49392f232036SJan Schmidt ret = mnt_want_write_file(file); 49402f232036SJan Schmidt if (ret) 49412f232036SJan Schmidt return ret; 49422f232036SJan Schmidt 49432f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 49442f232036SJan Schmidt if (IS_ERR(qsa)) { 49452f232036SJan Schmidt ret = PTR_ERR(qsa); 49462f232036SJan Schmidt goto drop_write; 49472f232036SJan Schmidt } 49482f232036SJan Schmidt 49492f232036SJan Schmidt if (qsa->flags) { 49502f232036SJan Schmidt ret = -EINVAL; 49512f232036SJan Schmidt goto out; 49522f232036SJan Schmidt } 49532f232036SJan Schmidt 49540b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 49552f232036SJan Schmidt 49562f232036SJan Schmidt out: 49572f232036SJan Schmidt kfree(qsa); 49582f232036SJan Schmidt drop_write: 49592f232036SJan Schmidt mnt_drop_write_file(file); 49602f232036SJan Schmidt return ret; 49612f232036SJan Schmidt } 49622f232036SJan Schmidt 49632f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 49642f232036SJan Schmidt { 49650b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49660b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49672f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49682f232036SJan Schmidt int ret = 0; 49692f232036SJan Schmidt 49702f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49712f232036SJan Schmidt return -EPERM; 49722f232036SJan Schmidt 49738d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 49742f232036SJan Schmidt if (!qsa) 49752f232036SJan Schmidt return -ENOMEM; 49762f232036SJan Schmidt 49770b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 49782f232036SJan Schmidt qsa->flags = 1; 49790b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 49802f232036SJan Schmidt } 49812f232036SJan Schmidt 49822f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 49832f232036SJan Schmidt ret = -EFAULT; 49842f232036SJan Schmidt 49852f232036SJan Schmidt kfree(qsa); 49862f232036SJan Schmidt return ret; 49872f232036SJan Schmidt } 49882f232036SJan Schmidt 498957254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 499057254b6eSJan Schmidt { 49910b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49920b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 499357254b6eSJan Schmidt 499457254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 499557254b6eSJan Schmidt return -EPERM; 499657254b6eSJan Schmidt 49970b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 499857254b6eSJan Schmidt } 499957254b6eSJan Schmidt 5000abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5001abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50028ea05e3aSAlexander Block { 5003496ad9aaSAl Viro struct inode *inode = file_inode(file); 50040b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50058ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50068ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50078ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 500895582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 50098ea05e3aSAlexander Block int ret = 0; 5010dd5f9615SStefan Behrens int received_uuid_changed; 50118ea05e3aSAlexander Block 5012bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5013bd60ea0fSDavid Sterba return -EPERM; 5014bd60ea0fSDavid Sterba 50158ea05e3aSAlexander Block ret = mnt_want_write_file(file); 50168ea05e3aSAlexander Block if (ret < 0) 50178ea05e3aSAlexander Block return ret; 50188ea05e3aSAlexander Block 50190b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 50208ea05e3aSAlexander Block 50214a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 50228ea05e3aSAlexander Block ret = -EINVAL; 50238ea05e3aSAlexander Block goto out; 50248ea05e3aSAlexander Block } 50258ea05e3aSAlexander Block 50268ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 50278ea05e3aSAlexander Block ret = -EROFS; 50288ea05e3aSAlexander Block goto out; 50298ea05e3aSAlexander Block } 50308ea05e3aSAlexander Block 5031dd5f9615SStefan Behrens /* 5032dd5f9615SStefan Behrens * 1 - root item 5033dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5034dd5f9615SStefan Behrens */ 5035dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 50368ea05e3aSAlexander Block if (IS_ERR(trans)) { 50378ea05e3aSAlexander Block ret = PTR_ERR(trans); 50388ea05e3aSAlexander Block trans = NULL; 50398ea05e3aSAlexander Block goto out; 50408ea05e3aSAlexander Block } 50418ea05e3aSAlexander Block 50428ea05e3aSAlexander Block sa->rtransid = trans->transid; 50438ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 50448ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 50458ea05e3aSAlexander Block 5046dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5047dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5048dd5f9615SStefan Behrens if (received_uuid_changed && 5049d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5050d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5051dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5052dd5f9615SStefan Behrens root->root_key.objectid); 5053d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5054d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5055d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5056d87ff758SNikolay Borisov goto out; 5057d87ff758SNikolay Borisov } 5058d87ff758SNikolay Borisov } 50598ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 50608ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 50618ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 50623cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 50633cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 50643cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 50653cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 50668ea05e3aSAlexander Block 50670b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 50688ea05e3aSAlexander Block &root->root_key, &root->root_item); 50698ea05e3aSAlexander Block if (ret < 0) { 50703a45bb20SJeff Mahoney btrfs_end_transaction(trans); 50718ea05e3aSAlexander Block goto out; 5072dd5f9615SStefan Behrens } 5073dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5074cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5075dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5076dd5f9615SStefan Behrens root->root_key.objectid); 5077dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 507866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5079efd38150SNikolay Borisov btrfs_end_transaction(trans); 5080dd5f9615SStefan Behrens goto out; 5081dd5f9615SStefan Behrens } 5082dd5f9615SStefan Behrens } 50833a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5084abccd00fSHugo Mills out: 50850b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5086abccd00fSHugo Mills mnt_drop_write_file(file); 5087abccd00fSHugo Mills return ret; 5088abccd00fSHugo Mills } 5089abccd00fSHugo Mills 5090abccd00fSHugo Mills #ifdef CONFIG_64BIT 5091abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5092abccd00fSHugo Mills void __user *arg) 5093abccd00fSHugo Mills { 5094abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5095abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5096abccd00fSHugo Mills int ret = 0; 5097abccd00fSHugo Mills 5098abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 50997b9ea627SShailendra Verma if (IS_ERR(args32)) 51007b9ea627SShailendra Verma return PTR_ERR(args32); 5101abccd00fSHugo Mills 51028d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 510384dbeb87SDan Carpenter if (!args64) { 510484dbeb87SDan Carpenter ret = -ENOMEM; 5105abccd00fSHugo Mills goto out; 5106abccd00fSHugo Mills } 5107abccd00fSHugo Mills 5108abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5109abccd00fSHugo Mills args64->stransid = args32->stransid; 5110abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5111abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5112abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5113abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5114abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5115abccd00fSHugo Mills args64->flags = args32->flags; 5116abccd00fSHugo Mills 5117abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5118abccd00fSHugo Mills if (ret) 5119abccd00fSHugo Mills goto out; 5120abccd00fSHugo Mills 5121abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5122abccd00fSHugo Mills args32->stransid = args64->stransid; 5123abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5124abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5125abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5126abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5127abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5128abccd00fSHugo Mills args32->flags = args64->flags; 5129abccd00fSHugo Mills 5130abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5131abccd00fSHugo Mills if (ret) 5132abccd00fSHugo Mills ret = -EFAULT; 5133abccd00fSHugo Mills 5134abccd00fSHugo Mills out: 5135abccd00fSHugo Mills kfree(args32); 5136abccd00fSHugo Mills kfree(args64); 5137abccd00fSHugo Mills return ret; 5138abccd00fSHugo Mills } 5139abccd00fSHugo Mills #endif 5140abccd00fSHugo Mills 5141abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5142abccd00fSHugo Mills void __user *arg) 5143abccd00fSHugo Mills { 5144abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5145abccd00fSHugo Mills int ret = 0; 5146abccd00fSHugo Mills 5147abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 51487b9ea627SShailendra Verma if (IS_ERR(sa)) 51497b9ea627SShailendra Verma return PTR_ERR(sa); 5150abccd00fSHugo Mills 5151abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5152abccd00fSHugo Mills 5153abccd00fSHugo Mills if (ret) 5154abccd00fSHugo Mills goto out; 5155abccd00fSHugo Mills 51568ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 51578ea05e3aSAlexander Block if (ret) 51588ea05e3aSAlexander Block ret = -EFAULT; 51598ea05e3aSAlexander Block 51608ea05e3aSAlexander Block out: 51618ea05e3aSAlexander Block kfree(sa); 51628ea05e3aSAlexander Block return ret; 51638ea05e3aSAlexander Block } 51648ea05e3aSAlexander Block 5165867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5166867ab667Sjeff.liu { 51670b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51680b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5169a1b83ac5SAnand Jain size_t len; 5170867ab667Sjeff.liu int ret; 5171a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5172a1b83ac5SAnand Jain 51730b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 51740b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 51750b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5176a1b83ac5SAnand Jain 5177a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5178867ab667Sjeff.liu 5179867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 51800b246afaSJeff Mahoney btrfs_warn(fs_info, 51810b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 51820b246afaSJeff Mahoney --len); 5183867ab667Sjeff.liu } 5184867ab667Sjeff.liu 5185867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5186867ab667Sjeff.liu 5187867ab667Sjeff.liu return ret ? -EFAULT : 0; 5188867ab667Sjeff.liu } 5189867ab667Sjeff.liu 5190a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5191a8bfd4abSjeff.liu { 51920b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51940b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51950b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5196a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5197a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5198a8bfd4abSjeff.liu int ret; 5199a8bfd4abSjeff.liu 5200a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5201a8bfd4abSjeff.liu return -EPERM; 5202a8bfd4abSjeff.liu 5203a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5204a8bfd4abSjeff.liu return -EFAULT; 5205a8bfd4abSjeff.liu 5206a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 52070b246afaSJeff Mahoney btrfs_err(fs_info, 52085d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5209a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5210a8bfd4abSjeff.liu return -EINVAL; 5211a8bfd4abSjeff.liu } 5212a8bfd4abSjeff.liu 5213a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5214a8bfd4abSjeff.liu if (ret) 5215a8bfd4abSjeff.liu return ret; 5216a8bfd4abSjeff.liu 5217a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5218a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5219a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5220a8bfd4abSjeff.liu goto out_unlock; 5221a8bfd4abSjeff.liu } 5222a8bfd4abSjeff.liu 52230b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5224a8bfd4abSjeff.liu strcpy(super_block->label, label); 52250b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 52263a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5227a8bfd4abSjeff.liu 5228a8bfd4abSjeff.liu out_unlock: 5229a8bfd4abSjeff.liu mnt_drop_write_file(file); 5230a8bfd4abSjeff.liu return ret; 5231a8bfd4abSjeff.liu } 5232a8bfd4abSjeff.liu 52332eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 52342eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 52352eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 52362eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 52372eaa055fSJeff Mahoney 5238d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 52392eaa055fSJeff Mahoney { 52404d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 52412eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 52422eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 52432eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 52442eaa055fSJeff Mahoney }; 52452eaa055fSJeff Mahoney 52462eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52472eaa055fSJeff Mahoney return -EFAULT; 52482eaa055fSJeff Mahoney 52492eaa055fSJeff Mahoney return 0; 52502eaa055fSJeff Mahoney } 52512eaa055fSJeff Mahoney 52522eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 52532eaa055fSJeff Mahoney { 52540b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52550b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52560b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 52572eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 52582eaa055fSJeff Mahoney 52592eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 52602eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 52612eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 52622eaa055fSJeff Mahoney 52632eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52642eaa055fSJeff Mahoney return -EFAULT; 52652eaa055fSJeff Mahoney 52662eaa055fSJeff Mahoney return 0; 52672eaa055fSJeff Mahoney } 52682eaa055fSJeff Mahoney 52692ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 52703b02a68aSJeff Mahoney enum btrfs_feature_set set, 52712eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 52722eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 52732eaa055fSJeff Mahoney { 5274f10152bcSDavid Sterba const char *type = btrfs_feature_set_name(set); 52753b02a68aSJeff Mahoney char *names; 52762eaa055fSJeff Mahoney u64 disallowed, unsupported; 52772eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 52782eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 52792eaa055fSJeff Mahoney 52802eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 52812eaa055fSJeff Mahoney if (unsupported) { 52823b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 52833b02a68aSJeff Mahoney if (names) { 52840b246afaSJeff Mahoney btrfs_warn(fs_info, 52853b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 52863b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 52873b02a68aSJeff Mahoney kfree(names); 52883b02a68aSJeff Mahoney } else 52890b246afaSJeff Mahoney btrfs_warn(fs_info, 52902eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 52912eaa055fSJeff Mahoney type, unsupported); 52922eaa055fSJeff Mahoney return -EOPNOTSUPP; 52932eaa055fSJeff Mahoney } 52942eaa055fSJeff Mahoney 52952eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 52962eaa055fSJeff Mahoney if (disallowed) { 52973b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 52983b02a68aSJeff Mahoney if (names) { 52990b246afaSJeff Mahoney btrfs_warn(fs_info, 53003b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53013b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53023b02a68aSJeff Mahoney kfree(names); 53033b02a68aSJeff Mahoney } else 53040b246afaSJeff Mahoney btrfs_warn(fs_info, 53052eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53062eaa055fSJeff Mahoney type, disallowed); 53072eaa055fSJeff Mahoney return -EPERM; 53082eaa055fSJeff Mahoney } 53092eaa055fSJeff Mahoney 53102eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 53112eaa055fSJeff Mahoney if (disallowed) { 53123b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53133b02a68aSJeff Mahoney if (names) { 53140b246afaSJeff Mahoney btrfs_warn(fs_info, 53153b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 53163b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53173b02a68aSJeff Mahoney kfree(names); 53183b02a68aSJeff Mahoney } else 53190b246afaSJeff Mahoney btrfs_warn(fs_info, 53202eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 53212eaa055fSJeff Mahoney type, disallowed); 53222eaa055fSJeff Mahoney return -EPERM; 53232eaa055fSJeff Mahoney } 53242eaa055fSJeff Mahoney 53252eaa055fSJeff Mahoney return 0; 53262eaa055fSJeff Mahoney } 53272eaa055fSJeff Mahoney 53282ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 53292ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 53302eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 53312eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 53322eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 53332eaa055fSJeff Mahoney 53342eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 53352eaa055fSJeff Mahoney { 53360b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53370b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53380b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 53390b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53402eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 53412eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 53422eaa055fSJeff Mahoney u64 newflags; 53432eaa055fSJeff Mahoney int ret; 53442eaa055fSJeff Mahoney 53452eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 53462eaa055fSJeff Mahoney return -EPERM; 53472eaa055fSJeff Mahoney 53482eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 53492eaa055fSJeff Mahoney return -EFAULT; 53502eaa055fSJeff Mahoney 53512eaa055fSJeff Mahoney /* Nothing to do */ 53522eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 53532eaa055fSJeff Mahoney !flags[0].incompat_flags) 53542eaa055fSJeff Mahoney return 0; 53552eaa055fSJeff Mahoney 53562ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 53572eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 53582eaa055fSJeff Mahoney if (ret) 53592eaa055fSJeff Mahoney return ret; 53602eaa055fSJeff Mahoney 53612ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 53622eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 53632eaa055fSJeff Mahoney if (ret) 53642eaa055fSJeff Mahoney return ret; 53652eaa055fSJeff Mahoney 53662ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 53672eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 53682eaa055fSJeff Mahoney if (ret) 53692eaa055fSJeff Mahoney return ret; 53702eaa055fSJeff Mahoney 53717ab19625SDavid Sterba ret = mnt_want_write_file(file); 53727ab19625SDavid Sterba if (ret) 53737ab19625SDavid Sterba return ret; 53747ab19625SDavid Sterba 53758051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 53767ab19625SDavid Sterba if (IS_ERR(trans)) { 53777ab19625SDavid Sterba ret = PTR_ERR(trans); 53787ab19625SDavid Sterba goto out_drop_write; 53797ab19625SDavid Sterba } 53802eaa055fSJeff Mahoney 53810b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 53822eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 53832eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 53842eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 53852eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 53862eaa055fSJeff Mahoney 53872eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 53882eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 53892eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 53902eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 53912eaa055fSJeff Mahoney 53922eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 53932eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 53942eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 53952eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 53960b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53972eaa055fSJeff Mahoney 53983a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 53997ab19625SDavid Sterba out_drop_write: 54007ab19625SDavid Sterba mnt_drop_write_file(file); 54017ab19625SDavid Sterba 54027ab19625SDavid Sterba return ret; 54032eaa055fSJeff Mahoney } 54042eaa055fSJeff Mahoney 54052351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 54062351f431SJosef Bacik { 54072351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 54082351f431SJosef Bacik int ret; 54092351f431SJosef Bacik 54102351f431SJosef Bacik if (compat) { 54112351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 54122351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 54132351f431SJosef Bacik 54142351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 54152351f431SJosef Bacik if (ret) 54162351f431SJosef Bacik return -EFAULT; 54172351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 54182351f431SJosef Bacik if (!arg) 54192351f431SJosef Bacik return -ENOMEM; 54202351f431SJosef Bacik arg->send_fd = args32.send_fd; 54212351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 54222351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 54232351f431SJosef Bacik arg->parent_root = args32.parent_root; 54242351f431SJosef Bacik arg->flags = args32.flags; 54252351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 54262351f431SJosef Bacik sizeof(args32.reserved)); 54272351f431SJosef Bacik #else 54282351f431SJosef Bacik return -ENOTTY; 54292351f431SJosef Bacik #endif 54302351f431SJosef Bacik } else { 54312351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 54322351f431SJosef Bacik if (IS_ERR(arg)) 54332351f431SJosef Bacik return PTR_ERR(arg); 54342351f431SJosef Bacik } 54352351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 54362351f431SJosef Bacik kfree(arg); 54372351f431SJosef Bacik return ret; 54382351f431SJosef Bacik } 54392351f431SJosef Bacik 5440f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5441f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5442f46b5a66SChristoph Hellwig { 54430b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54440b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54450b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54464bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5447f46b5a66SChristoph Hellwig 5448f46b5a66SChristoph Hellwig switch (cmd) { 54496cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 54506cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 54516cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 54526cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 54536cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 54546cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 545540cf931fSEric Sandeen case FS_IOC_GETFSLABEL: 545640cf931fSEric Sandeen return btrfs_ioctl_get_fslabel(file, argp); 545740cf931fSEric Sandeen case FS_IOC_SETFSLABEL: 545840cf931fSEric Sandeen return btrfs_ioctl_set_fslabel(file, argp); 5459f7039b1dSLi Dongyang case FITRIM: 5460f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5461f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5462fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5463fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5464fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 54653de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5466fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 54676f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 54686f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 546976dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 547076dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 54710caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 54720caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 54730caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 54740caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 54756ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 54766ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5477f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 54781e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 54791e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 54801e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5481f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5482198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5483f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 54842ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5485f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5486da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 54876b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 54886b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5489475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 54902ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5491475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 54922ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5493f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 54949ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5495ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5496ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5497cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5498cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5499ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5500ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5501d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5502d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5503d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5504d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5505d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5506d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 55071406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55082ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 55099b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55109b199859SFilipe David Borba Manana int ret; 55119b199859SFilipe David Borba Manana 551282b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 55139b199859SFilipe David Borba Manana if (ret) 55149b199859SFilipe David Borba Manana return ret; 55150b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 55162fad4e83SDavid Sterba /* 55172fad4e83SDavid Sterba * The transaction thread may want to do more work, 551801327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 55192fad4e83SDavid Sterba * processing uncleaned subvols. 55202fad4e83SDavid Sterba */ 55210b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 55229b199859SFilipe David Borba Manana return ret; 55239b199859SFilipe David Borba Manana } 552446204592SSage Weil case BTRFS_IOC_START_SYNC: 55259a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 552646204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55272ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5528475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5529b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5530475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 55312ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5532475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 55332ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5534c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55359ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5536837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 55372ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 553819a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 55392ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 55408ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55418ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5542abccd00fSHugo Mills #ifdef CONFIG_64BIT 5543abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5544abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5545abccd00fSHugo Mills #endif 554631db9f7cSAlexander Block case BTRFS_IOC_SEND: 55472351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 55482351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55492351f431SJosef Bacik case BTRFS_IOC_SEND_32: 55502351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 55512351f431SJosef Bacik #endif 5552c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 55532ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 55545d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5555905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55565d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5557905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55585d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5559905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55605d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5561905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 55622f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 55632f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 55642f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 55652f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 556657254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 556757254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 55683f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 55692ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 55702eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5571d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 55722eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 55732eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 55742eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 55752eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5576e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5577e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5578025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5579025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5580b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5581b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 558242e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 558342e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 558423d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 558523d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5586f46b5a66SChristoph Hellwig } 5587f46b5a66SChristoph Hellwig 5588f46b5a66SChristoph Hellwig return -ENOTTY; 5589f46b5a66SChristoph Hellwig } 55904c63c245SLuke Dashjr 55914c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 55924c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 55934c63c245SLuke Dashjr { 55942a362249SJeff Mahoney /* 55952a362249SJeff Mahoney * These all access 32-bit values anyway so no further 55962a362249SJeff Mahoney * handling is necessary. 55972a362249SJeff Mahoney */ 55984c63c245SLuke Dashjr switch (cmd) { 55994c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56004c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56014c63c245SLuke Dashjr break; 56024c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56034c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56044c63c245SLuke Dashjr break; 56054c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56064c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56074c63c245SLuke Dashjr break; 56084c63c245SLuke Dashjr } 56094c63c245SLuke Dashjr 56104c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56114c63c245SLuke Dashjr } 56124c63c245SLuke Dashjr #endif 5613