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" 31949964c9SMarcos Paulo de Souza #include "export.h" 32f46b5a66SChristoph Hellwig #include "transaction.h" 33f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 34f46b5a66SChristoph Hellwig #include "print-tree.h" 35f46b5a66SChristoph Hellwig #include "volumes.h" 36925baeddSChris Mason #include "locking.h" 37581bb050SLi Zefan #include "inode-map.h" 38d7728c96SJan Schmidt #include "backref.h" 39606686eeSJosef Bacik #include "rcu-string.h" 4031db9f7cSAlexander Block #include "send.h" 413f6bcfbdSStefan Behrens #include "dev-replace.h" 4263541927SFilipe David Borba Manana #include "props.h" 433b02a68aSJeff Mahoney #include "sysfs.h" 44fcebe456SJosef Bacik #include "qgroup.h" 451ec9a1aeSFilipe Manana #include "tree-log.h" 46ebb8765bSAnand Jain #include "compression.h" 478719aaaeSJosef Bacik #include "space-info.h" 4886736342SJosef Bacik #include "delalloc-space.h" 49aac0023cSJosef Bacik #include "block-group.h" 50f46b5a66SChristoph Hellwig 51abccd00fSHugo Mills #ifdef CONFIG_64BIT 52abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 53abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 54abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 55abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 56abccd00fSHugo Mills */ 57abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 58abccd00fSHugo Mills __u64 sec; 59abccd00fSHugo Mills __u32 nsec; 60abccd00fSHugo Mills } __attribute__ ((__packed__)); 61abccd00fSHugo Mills 62abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 63abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 64abccd00fSHugo Mills __u64 stransid; /* in */ 65abccd00fSHugo Mills __u64 rtransid; /* out */ 66abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 67abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 68abccd00fSHugo Mills __u64 flags; /* in */ 69abccd00fSHugo Mills __u64 reserved[16]; /* in */ 70abccd00fSHugo Mills } __attribute__ ((__packed__)); 71abccd00fSHugo Mills 72abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 73abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 74abccd00fSHugo Mills #endif 75abccd00fSHugo Mills 762351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 772351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 782351f431SJosef Bacik __s64 send_fd; /* in */ 792351f431SJosef Bacik __u64 clone_sources_count; /* in */ 802351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 812351f431SJosef Bacik __u64 parent_root; /* in */ 822351f431SJosef Bacik __u64 flags; /* in */ 832351f431SJosef Bacik __u64 reserved[4]; /* in */ 842351f431SJosef Bacik } __attribute__ ((__packed__)); 852351f431SJosef Bacik 862351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 872351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 882351f431SJosef Bacik #endif 89abccd00fSHugo Mills 90416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 911c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 921c919a5eSMark Fasheh int no_time_update); 93416161dbSMark Fasheh 946cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 951905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 961905a0f7SDavid Sterba unsigned int flags) 976cbff00fSChristoph Hellwig { 981905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 996cbff00fSChristoph Hellwig return flags; 1001905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 1016cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1026cbff00fSChristoph Hellwig else 1036cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1046cbff00fSChristoph Hellwig } 105f46b5a66SChristoph Hellwig 1066cbff00fSChristoph Hellwig /* 107a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 108a157d4fdSDavid Sterba * ioctl. 1096cbff00fSChristoph Hellwig */ 110a157d4fdSDavid Sterba static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags) 1116cbff00fSChristoph Hellwig { 1126cbff00fSChristoph Hellwig unsigned int iflags = 0; 1136cbff00fSChristoph Hellwig 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1156cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1176cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1196cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1216cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1226cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1236cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1246cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1256cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 126d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 127d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 128d0092bddSLi Zefan 12913f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 130d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 13113f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13213f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1336cbff00fSChristoph Hellwig 1346cbff00fSChristoph Hellwig return iflags; 1356cbff00fSChristoph Hellwig } 1366cbff00fSChristoph Hellwig 1376cbff00fSChristoph Hellwig /* 1386cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1396cbff00fSChristoph Hellwig */ 1407b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1416cbff00fSChristoph Hellwig { 1425c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1433cc79392SFilipe Manana unsigned int new_fl = 0; 1446cbff00fSChristoph Hellwig 1455c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1463cc79392SFilipe Manana new_fl |= S_SYNC; 1475c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1483cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1495c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1503cc79392SFilipe Manana new_fl |= S_APPEND; 1515c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1523cc79392SFilipe Manana new_fl |= S_NOATIME; 1535c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1543cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1553cc79392SFilipe Manana 1563cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1573cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1583cc79392SFilipe Manana new_fl); 1596cbff00fSChristoph Hellwig } 1606cbff00fSChristoph Hellwig 1616cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1626cbff00fSChristoph Hellwig { 1635c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 1645c57b8b6SDavid Sterba unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags); 1656cbff00fSChristoph Hellwig 1666cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1676cbff00fSChristoph Hellwig return -EFAULT; 1686cbff00fSChristoph Hellwig return 0; 1696cbff00fSChristoph Hellwig } 1706cbff00fSChristoph Hellwig 1715ba76abfSDavid Sterba /* Check if @flags are a supported and valid set of FS_*_FL flags */ 1725ba76abfSDavid Sterba static int check_fsflags(unsigned int flags) 17375e7cb7fSLiu Bo { 17475e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17575e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17675e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 177e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 178e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17975e7cb7fSLiu Bo return -EOPNOTSUPP; 18075e7cb7fSLiu Bo 18175e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 18275e7cb7fSLiu Bo return -EINVAL; 18375e7cb7fSLiu Bo 18475e7cb7fSLiu Bo return 0; 18575e7cb7fSLiu Bo } 18675e7cb7fSLiu Bo 1876cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1886cbff00fSChristoph Hellwig { 189496ad9aaSAl Viro struct inode *inode = file_inode(file); 1900b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1915c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1925c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 1936cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1945aca2842SDarrick J. Wong unsigned int fsflags, old_fsflags; 1956cbff00fSChristoph Hellwig int ret; 196ff9fef55SAnand Jain const char *comp = NULL; 197d2b8fcfeSAnand Jain u32 binode_flags = binode->flags; 1986cbff00fSChristoph Hellwig 199bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 200bd60ea0fSDavid Sterba return -EPERM; 201bd60ea0fSDavid Sterba 202b83cc969SLi Zefan if (btrfs_root_readonly(root)) 203b83cc969SLi Zefan return -EROFS; 204b83cc969SLi Zefan 2055c57b8b6SDavid Sterba if (copy_from_user(&fsflags, arg, sizeof(fsflags))) 2066cbff00fSChristoph Hellwig return -EFAULT; 2076cbff00fSChristoph Hellwig 2085c57b8b6SDavid Sterba ret = check_fsflags(fsflags); 20975e7cb7fSLiu Bo if (ret) 21075e7cb7fSLiu Bo return ret; 2116cbff00fSChristoph Hellwig 212e7848683SJan Kara ret = mnt_want_write_file(file); 213e7848683SJan Kara if (ret) 214e7848683SJan Kara return ret; 215e7848683SJan Kara 2165955102cSAl Viro inode_lock(inode); 2176cbff00fSChristoph Hellwig 2185c57b8b6SDavid Sterba fsflags = btrfs_mask_fsflags_for_type(inode, fsflags); 2195aca2842SDarrick J. Wong old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags); 2205aca2842SDarrick J. Wong ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags); 2215aca2842SDarrick J. Wong if (ret) 2226cbff00fSChristoph Hellwig goto out_unlock; 2236cbff00fSChristoph Hellwig 2245c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 225d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_SYNC; 2266cbff00fSChristoph Hellwig else 227d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_SYNC; 2285c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 229d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_IMMUTABLE; 2306cbff00fSChristoph Hellwig else 231d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_IMMUTABLE; 2325c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 233d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_APPEND; 2346cbff00fSChristoph Hellwig else 235d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_APPEND; 2365c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 237d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODUMP; 2386cbff00fSChristoph Hellwig else 239d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODUMP; 2405c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 241d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOATIME; 2426cbff00fSChristoph Hellwig else 243d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOATIME; 2445c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 245d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_DIRSYNC; 2466cbff00fSChristoph Hellwig else 247d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_DIRSYNC; 2485c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 24944e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2507e97b8daSDavid Sterba /* 2517e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2527e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2537e97b8daSDavid Sterba * status of the file and will not set it. 2547e97b8daSDavid Sterba */ 2557e97b8daSDavid Sterba if (inode->i_size == 0) 256d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW | 257d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM; 2587e97b8daSDavid Sterba } else { 259d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW; 2607e97b8daSDavid Sterba } 2617e97b8daSDavid Sterba } else { 2627e97b8daSDavid Sterba /* 26301327610SNicholas D Steeves * Revert back under same assumptions as above 2647e97b8daSDavid Sterba */ 26544e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2667e97b8daSDavid Sterba if (inode->i_size == 0) 267d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_NODATACOW | 268d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM); 2697e97b8daSDavid Sterba } else { 270d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODATACOW; 2717e97b8daSDavid Sterba } 2727e97b8daSDavid Sterba } 2736cbff00fSChristoph Hellwig 27475e7cb7fSLiu Bo /* 27575e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 27675e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 27775e7cb7fSLiu Bo * things smaller. 27875e7cb7fSLiu Bo */ 2795c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 280d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_COMPRESS; 281d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOCOMPRESS; 2825c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 28363541927SFilipe David Borba Manana 284eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 285eede2bf3SOmar Sandoval ret = -ETXTBSY; 286eede2bf3SOmar Sandoval goto out_unlock; 287eede2bf3SOmar Sandoval } 288eede2bf3SOmar Sandoval 289d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_COMPRESS; 290d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOCOMPRESS; 29163541927SFilipe David Borba Manana 29293370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 29393370509SDavid Sterba if (!comp || comp[0] == 0) 29493370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 295ebcb904dSLi Zefan } else { 296d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 29775e7cb7fSLiu Bo } 2986cbff00fSChristoph Hellwig 299ff9fef55SAnand Jain /* 300ff9fef55SAnand Jain * 1 for inode item 301ff9fef55SAnand Jain * 2 for properties 302ff9fef55SAnand Jain */ 303ff9fef55SAnand Jain trans = btrfs_start_transaction(root, 3); 304f062abf0SLi Zefan if (IS_ERR(trans)) { 305f062abf0SLi Zefan ret = PTR_ERR(trans); 306d2b8fcfeSAnand Jain goto out_unlock; 307f062abf0SLi Zefan } 3086cbff00fSChristoph Hellwig 309ff9fef55SAnand Jain if (comp) { 310ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, 311ff9fef55SAnand Jain strlen(comp), 0); 312ff9fef55SAnand Jain if (ret) { 313ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 314ff9fef55SAnand Jain goto out_end_trans; 315ff9fef55SAnand Jain } 316ff9fef55SAnand Jain } else { 317ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 318ff9fef55SAnand Jain 0, 0); 319ff9fef55SAnand Jain if (ret && ret != -ENODATA) { 320ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 321ff9fef55SAnand Jain goto out_end_trans; 322ff9fef55SAnand Jain } 323ff9fef55SAnand Jain } 324ff9fef55SAnand Jain 325d2b8fcfeSAnand Jain binode->flags = binode_flags; 3267b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3270c4d2d95SJosef Bacik inode_inc_iversion(inode); 328c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3296cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3306cbff00fSChristoph Hellwig 331ff9fef55SAnand Jain out_end_trans: 3323a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3336cbff00fSChristoph Hellwig out_unlock: 3345955102cSAl Viro inode_unlock(inode); 335e7848683SJan Kara mnt_drop_write_file(file); 3362d4e6f6aSliubo return ret; 3376cbff00fSChristoph Hellwig } 3386cbff00fSChristoph Hellwig 33919f93b3cSDavid Sterba /* 34019f93b3cSDavid Sterba * Translate btrfs internal inode flags to xflags as expected by the 34119f93b3cSDavid Sterba * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are 34219f93b3cSDavid Sterba * silently dropped. 34319f93b3cSDavid Sterba */ 34419f93b3cSDavid Sterba static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags) 34519f93b3cSDavid Sterba { 34619f93b3cSDavid Sterba unsigned int xflags = 0; 34719f93b3cSDavid Sterba 34819f93b3cSDavid Sterba if (flags & BTRFS_INODE_APPEND) 34919f93b3cSDavid Sterba xflags |= FS_XFLAG_APPEND; 35019f93b3cSDavid Sterba if (flags & BTRFS_INODE_IMMUTABLE) 35119f93b3cSDavid Sterba xflags |= FS_XFLAG_IMMUTABLE; 35219f93b3cSDavid Sterba if (flags & BTRFS_INODE_NOATIME) 35319f93b3cSDavid Sterba xflags |= FS_XFLAG_NOATIME; 35419f93b3cSDavid Sterba if (flags & BTRFS_INODE_NODUMP) 35519f93b3cSDavid Sterba xflags |= FS_XFLAG_NODUMP; 35619f93b3cSDavid Sterba if (flags & BTRFS_INODE_SYNC) 35719f93b3cSDavid Sterba xflags |= FS_XFLAG_SYNC; 35819f93b3cSDavid Sterba 35919f93b3cSDavid Sterba return xflags; 36019f93b3cSDavid Sterba } 36119f93b3cSDavid Sterba 36219f93b3cSDavid Sterba /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */ 36319f93b3cSDavid Sterba static int check_xflags(unsigned int flags) 36419f93b3cSDavid Sterba { 36519f93b3cSDavid Sterba if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME | 36619f93b3cSDavid Sterba FS_XFLAG_NODUMP | FS_XFLAG_SYNC)) 36719f93b3cSDavid Sterba return -EOPNOTSUPP; 36819f93b3cSDavid Sterba return 0; 36919f93b3cSDavid Sterba } 37019f93b3cSDavid Sterba 371e4202ac9SDavid Sterba /* 372e4202ac9SDavid Sterba * Set the xflags from the internal inode flags. The remaining items of fsxattr 373e4202ac9SDavid Sterba * are zeroed. 374e4202ac9SDavid Sterba */ 375e4202ac9SDavid Sterba static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg) 376e4202ac9SDavid Sterba { 377e4202ac9SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 378e4202ac9SDavid Sterba struct fsxattr fa; 379e4202ac9SDavid Sterba 3807b0e492eSDarrick J. Wong simple_fill_fsxattr(&fa, btrfs_inode_flags_to_xflags(binode->flags)); 381e4202ac9SDavid Sterba if (copy_to_user(arg, &fa, sizeof(fa))) 382e4202ac9SDavid Sterba return -EFAULT; 383e4202ac9SDavid Sterba 384e4202ac9SDavid Sterba return 0; 385e4202ac9SDavid Sterba } 386e4202ac9SDavid Sterba 387025f2121SDavid Sterba static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg) 388025f2121SDavid Sterba { 389025f2121SDavid Sterba struct inode *inode = file_inode(file); 390025f2121SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 391025f2121SDavid Sterba struct btrfs_root *root = binode->root; 392025f2121SDavid Sterba struct btrfs_trans_handle *trans; 3937b0e492eSDarrick J. Wong struct fsxattr fa, old_fa; 394025f2121SDavid Sterba unsigned old_flags; 395025f2121SDavid Sterba unsigned old_i_flags; 396025f2121SDavid Sterba int ret = 0; 397025f2121SDavid Sterba 398025f2121SDavid Sterba if (!inode_owner_or_capable(inode)) 399025f2121SDavid Sterba return -EPERM; 400025f2121SDavid Sterba 401025f2121SDavid Sterba if (btrfs_root_readonly(root)) 402025f2121SDavid Sterba return -EROFS; 403025f2121SDavid Sterba 404025f2121SDavid Sterba if (copy_from_user(&fa, arg, sizeof(fa))) 405025f2121SDavid Sterba return -EFAULT; 406025f2121SDavid Sterba 407025f2121SDavid Sterba ret = check_xflags(fa.fsx_xflags); 408025f2121SDavid Sterba if (ret) 409025f2121SDavid Sterba return ret; 410025f2121SDavid Sterba 411025f2121SDavid Sterba if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0) 412025f2121SDavid Sterba return -EOPNOTSUPP; 413025f2121SDavid Sterba 414025f2121SDavid Sterba ret = mnt_want_write_file(file); 415025f2121SDavid Sterba if (ret) 416025f2121SDavid Sterba return ret; 417025f2121SDavid Sterba 418025f2121SDavid Sterba inode_lock(inode); 419025f2121SDavid Sterba 420025f2121SDavid Sterba old_flags = binode->flags; 421025f2121SDavid Sterba old_i_flags = inode->i_flags; 422025f2121SDavid Sterba 4237b0e492eSDarrick J. Wong simple_fill_fsxattr(&old_fa, 4247b0e492eSDarrick J. Wong btrfs_inode_flags_to_xflags(binode->flags)); 4257b0e492eSDarrick J. Wong ret = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa); 4267b0e492eSDarrick J. Wong if (ret) 427025f2121SDavid Sterba goto out_unlock; 428025f2121SDavid Sterba 429025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_SYNC) 430025f2121SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 431025f2121SDavid Sterba else 432025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 433025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE) 434025f2121SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 435025f2121SDavid Sterba else 436025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 437025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_APPEND) 438025f2121SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 439025f2121SDavid Sterba else 440025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 441025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NODUMP) 442025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 443025f2121SDavid Sterba else 444025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 445025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NOATIME) 446025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 447025f2121SDavid Sterba else 448025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 449025f2121SDavid Sterba 450025f2121SDavid Sterba /* 1 item for the inode */ 451025f2121SDavid Sterba trans = btrfs_start_transaction(root, 1); 452025f2121SDavid Sterba if (IS_ERR(trans)) { 453025f2121SDavid Sterba ret = PTR_ERR(trans); 454025f2121SDavid Sterba goto out_unlock; 455025f2121SDavid Sterba } 456025f2121SDavid Sterba 457025f2121SDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 458025f2121SDavid Sterba inode_inc_iversion(inode); 459025f2121SDavid Sterba inode->i_ctime = current_time(inode); 460025f2121SDavid Sterba ret = btrfs_update_inode(trans, root, inode); 461025f2121SDavid Sterba 462025f2121SDavid Sterba btrfs_end_transaction(trans); 463025f2121SDavid Sterba 464025f2121SDavid Sterba out_unlock: 465025f2121SDavid Sterba if (ret) { 466025f2121SDavid Sterba binode->flags = old_flags; 467025f2121SDavid Sterba inode->i_flags = old_i_flags; 468025f2121SDavid Sterba } 469025f2121SDavid Sterba 470025f2121SDavid Sterba inode_unlock(inode); 471025f2121SDavid Sterba mnt_drop_write_file(file); 472025f2121SDavid Sterba 473025f2121SDavid Sterba return ret; 474025f2121SDavid Sterba } 475025f2121SDavid Sterba 4766cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4776cbff00fSChristoph Hellwig { 478496ad9aaSAl Viro struct inode *inode = file_inode(file); 4796cbff00fSChristoph Hellwig 4806cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4816cbff00fSChristoph Hellwig } 482f46b5a66SChristoph Hellwig 483b929c1d8SMarcos Paulo de Souza static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, 484b929c1d8SMarcos Paulo de Souza void __user *arg) 485f7039b1dSLi Dongyang { 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 544e1f60a65SDavid Sterba int __pure 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; 6703619c94fSJosef Bacik new_root = btrfs_get_fs_root(fs_info, &key, true); 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); 68000246528SJosef Bacik btrfs_put_root(new_root); 681ce598979SMark Fasheh if (ret) { 682ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 68366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 684ce598979SMark Fasheh goto fail; 685ce598979SMark Fasheh } 686ce598979SMark Fasheh 687f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 688f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 689f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 690f32e48e9SChandan Rajendra 691f46b5a66SChristoph Hellwig /* 692f46b5a66SChristoph Hellwig * insert the directory item 693f46b5a66SChristoph Hellwig */ 694877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 69579787eaaSJeff Mahoney if (ret) { 69666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 69779787eaaSJeff Mahoney goto fail; 69879787eaaSJeff Mahoney } 6993de4586cSChris Mason 700684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 7013de4586cSChris Mason BTRFS_FT_DIR, index); 70279787eaaSJeff Mahoney if (ret) { 70366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 704f46b5a66SChristoph Hellwig goto fail; 70579787eaaSJeff Mahoney } 7060660b5afSChris Mason 7076ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 70852c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 709c7e54b51SJosef Bacik if (ret) { 710c7e54b51SJosef Bacik btrfs_abort_transaction(trans, ret); 711c7e54b51SJosef Bacik goto fail; 712c7e54b51SJosef Bacik } 71352c26179SYan Zheng 7146025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7154a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 716c7e54b51SJosef Bacik if (ret) { 717c7e54b51SJosef Bacik btrfs_abort_transaction(trans, ret); 718c7e54b51SJosef Bacik goto fail; 719c7e54b51SJosef Bacik } 7200660b5afSChris Mason 721cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7226bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 723dd5f9615SStefan Behrens if (ret) 72466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 725dd5f9615SStefan Behrens 726f46b5a66SChristoph Hellwig fail: 72749a3c4d9SDavid Sterba kfree(root_item); 728d5c12070SMiao Xie trans->block_rsv = NULL; 729d5c12070SMiao Xie trans->bytes_reserved = 0; 7307775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 731de6e8200SLiu Bo 73272fd032eSSage Weil if (async_transid) { 73372fd032eSSage Weil *async_transid = trans->transid; 7343a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 73500d71c9cSMiao Xie if (err) 7363a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73772fd032eSSage Weil } else { 7383a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73972fd032eSSage Weil } 740f46b5a66SChristoph Hellwig if (err && !ret) 741f46b5a66SChristoph Hellwig ret = err; 7421a65e24bSChris Mason 7435662344bSTsutomu Itoh if (!ret) { 7445662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 745de6e8200SLiu Bo if (IS_ERR(inode)) 746de6e8200SLiu Bo return PTR_ERR(inode); 7475662344bSTsutomu Itoh d_instantiate(dentry, inode); 7485662344bSTsutomu Itoh } 749f46b5a66SChristoph Hellwig return ret; 75049a3c4d9SDavid Sterba 75149a3c4d9SDavid Sterba fail_free: 75249a3c4d9SDavid Sterba kfree(root_item); 75349a3c4d9SDavid Sterba return ret; 754f46b5a66SChristoph Hellwig } 755f46b5a66SChristoph Hellwig 756e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 75761d7e4cbSDavid Sterba struct dentry *dentry, 758e9662f70SMiao Xie u64 *async_transid, bool readonly, 759e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 760f46b5a66SChristoph Hellwig { 7610b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7622e4bfab9SYan, Zheng struct inode *inode; 763f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 764f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7652e4bfab9SYan, Zheng int ret; 7668ecebf4dSRobbie Ko bool snapshot_force_cow = false; 767f46b5a66SChristoph Hellwig 76827cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 769f46b5a66SChristoph Hellwig return -EINVAL; 770f46b5a66SChristoph Hellwig 771eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 772eede2bf3SOmar Sandoval btrfs_warn(fs_info, 773eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 774eede2bf3SOmar Sandoval return -ETXTBSY; 775eede2bf3SOmar Sandoval } 776eede2bf3SOmar Sandoval 77723269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 778a1ee7362SDavid Sterba if (!pending_snapshot) 779a1ee7362SDavid Sterba return -ENOMEM; 780a1ee7362SDavid Sterba 781b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 78223269bf5SDavid Sterba GFP_KERNEL); 7838546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7848546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 785b0c0ea63SDavid Sterba ret = -ENOMEM; 786b0c0ea63SDavid Sterba goto free_pending; 787b0c0ea63SDavid Sterba } 788b0c0ea63SDavid Sterba 7898ecebf4dSRobbie Ko /* 7908ecebf4dSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 7918ecebf4dSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 7928ecebf4dSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 7938ecebf4dSRobbie Ko */ 794dcc3eb96SNikolay Borisov btrfs_drew_read_lock(&root->snapshot_lock); 7958257b2dcSMiao Xie 7963cd24c69SEthan Lien ret = btrfs_start_delalloc_snapshot(root); 7976a03843dSMiao Xie if (ret) 798a1ee7362SDavid Sterba goto dec_and_free; 7996a03843dSMiao Xie 8008ecebf4dSRobbie Ko /* 8018ecebf4dSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 8028ecebf4dSRobbie Ko * we force future writes to fallback to COW mode during snapshot 8038ecebf4dSRobbie Ko * creation. 8048ecebf4dSRobbie Ko */ 8058ecebf4dSRobbie Ko atomic_inc(&root->snapshot_force_cow); 8068ecebf4dSRobbie Ko snapshot_force_cow = true; 8078ecebf4dSRobbie Ko 8086374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 8096a03843dSMiao Xie 81066d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 81166d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 812d5c12070SMiao Xie /* 813d5c12070SMiao Xie * 1 - parent dir inode 814d5c12070SMiao Xie * 2 - dir entries 815d5c12070SMiao Xie * 1 - root item 816d5c12070SMiao Xie * 2 - root ref/backref 817d5c12070SMiao Xie * 1 - root of snapshot 818dd5f9615SStefan Behrens * 1 - UUID item 819d5c12070SMiao Xie */ 820d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 821dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 822ee3441b4SJeff Mahoney false); 823d5c12070SMiao Xie if (ret) 824a1ee7362SDavid Sterba goto dec_and_free; 825d5c12070SMiao Xie 826a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 827a22285a6SYan, Zheng pending_snapshot->root = root; 828b83cc969SLi Zefan pending_snapshot->readonly = readonly; 829e9662f70SMiao Xie pending_snapshot->dir = dir; 8308696c533SMiao Xie pending_snapshot->inherit = inherit; 831a22285a6SYan, Zheng 832d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 833a22285a6SYan, Zheng if (IS_ERR(trans)) { 834a22285a6SYan, Zheng ret = PTR_ERR(trans); 835a22285a6SYan, Zheng goto fail; 836a22285a6SYan, Zheng } 837a22285a6SYan, Zheng 8380b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 839a22285a6SYan, Zheng list_add(&pending_snapshot->list, 840a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8410b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 84272fd032eSSage Weil if (async_transid) { 84372fd032eSSage Weil *async_transid = trans->transid; 8443a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 84500d71c9cSMiao Xie if (ret) 8463a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 84772fd032eSSage Weil } else { 8483a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 84972fd032eSSage Weil } 850aec8030aSMiao Xie if (ret) 851c37b2b62SJosef Bacik goto fail; 852a22285a6SYan, Zheng 853a22285a6SYan, Zheng ret = pending_snapshot->error; 854f46b5a66SChristoph Hellwig if (ret) 8552e4bfab9SYan, Zheng goto fail; 856f46b5a66SChristoph Hellwig 857d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 858d3797308SChris Mason if (ret) 859d3797308SChris Mason goto fail; 860d3797308SChris Mason 8612b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8622e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8632e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8642e4bfab9SYan, Zheng goto fail; 8652e4bfab9SYan, Zheng } 8665662344bSTsutomu Itoh 8672e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8682e4bfab9SYan, Zheng ret = 0; 8692e4bfab9SYan, Zheng fail: 87000246528SJosef Bacik btrfs_put_root(pending_snapshot->snap); 8717775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 872a1ee7362SDavid Sterba dec_and_free: 8738ecebf4dSRobbie Ko if (snapshot_force_cow) 8748ecebf4dSRobbie Ko atomic_dec(&root->snapshot_force_cow); 875dcc3eb96SNikolay Borisov btrfs_drew_read_unlock(&root->snapshot_lock); 876dcc3eb96SNikolay Borisov 877b0c0ea63SDavid Sterba free_pending: 878b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8798546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 880a1ee7362SDavid Sterba kfree(pending_snapshot); 881a1ee7362SDavid Sterba 882f46b5a66SChristoph Hellwig return ret; 883f46b5a66SChristoph Hellwig } 884f46b5a66SChristoph Hellwig 8854260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8864260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8874260f7c7SSage Weil * whether the type of victim is right. 8884260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8894260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8904260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8914260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8924260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8934260f7c7SSage Weil * a. be owner of dir, or 8944260f7c7SSage Weil * b. be owner of victim, or 8954260f7c7SSage Weil * c. have CAP_FOWNER capability 89601327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8974260f7c7SSage Weil * links pointing to it. 8984260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8994260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 9004260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 9014260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 9024260f7c7SSage Weil * nfs_async_unlink(). 9034260f7c7SSage Weil */ 9044260f7c7SSage Weil 9054260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 9064260f7c7SSage Weil { 9074260f7c7SSage Weil int error; 9084260f7c7SSage Weil 9092b0143b5SDavid Howells if (d_really_is_negative(victim)) 9104260f7c7SSage Weil return -ENOENT; 9114260f7c7SSage Weil 9122b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 9134fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 9144260f7c7SSage Weil 9154260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 9164260f7c7SSage Weil if (error) 9174260f7c7SSage Weil return error; 9184260f7c7SSage Weil if (IS_APPEND(dir)) 9194260f7c7SSage Weil return -EPERM; 9202b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9212b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9224260f7c7SSage Weil return -EPERM; 9234260f7c7SSage Weil if (isdir) { 924e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9254260f7c7SSage Weil return -ENOTDIR; 9264260f7c7SSage Weil if (IS_ROOT(victim)) 9274260f7c7SSage Weil return -EBUSY; 928e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9294260f7c7SSage Weil return -EISDIR; 9304260f7c7SSage Weil if (IS_DEADDIR(dir)) 9314260f7c7SSage Weil return -ENOENT; 9324260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9334260f7c7SSage Weil return -EBUSY; 9344260f7c7SSage Weil return 0; 9354260f7c7SSage Weil } 9364260f7c7SSage Weil 937cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 938cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 939cb8e7090SChristoph Hellwig { 9402b0143b5SDavid Howells if (d_really_is_positive(child)) 941cb8e7090SChristoph Hellwig return -EEXIST; 942cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 943cb8e7090SChristoph Hellwig return -ENOENT; 944cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 945cb8e7090SChristoph Hellwig } 946cb8e7090SChristoph Hellwig 947cb8e7090SChristoph Hellwig /* 948cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 949cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 950cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 951cb8e7090SChristoph Hellwig */ 95292872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 95352f75f4fSDavid Sterba const char *name, int namelen, 95472fd032eSSage Weil struct btrfs_root *snap_src, 9556f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9568696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 957cb8e7090SChristoph Hellwig { 9582b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 960cb8e7090SChristoph Hellwig struct dentry *dentry; 961cb8e7090SChristoph Hellwig int error; 962cb8e7090SChristoph Hellwig 96300235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 96400235411SAl Viro if (error == -EINTR) 96500235411SAl Viro return error; 966cb8e7090SChristoph Hellwig 967cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 968cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 969cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 970cb8e7090SChristoph Hellwig goto out_unlock; 971cb8e7090SChristoph Hellwig 97276dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 973cb8e7090SChristoph Hellwig if (error) 974a874a63eSLiu Bo goto out_dput; 975cb8e7090SChristoph Hellwig 9769c52057cSChris Mason /* 9779c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9789c52057cSChris Mason * check for them now when we can safely fail 9799c52057cSChris Mason */ 9809c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9819c52057cSChris Mason dir->i_ino, name, 9829c52057cSChris Mason namelen); 9839c52057cSChris Mason if (error) 9849c52057cSChris Mason goto out_dput; 9859c52057cSChris Mason 9860b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 98776dda93cSYan, Zheng 98876dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 98976dda93cSYan, Zheng goto out_up_read; 99076dda93cSYan, Zheng 9913de4586cSChris Mason if (snap_src) { 99261d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9936f72c7e2SArne Jansen async_transid, readonly, inherit); 9943de4586cSChris Mason } else { 995d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 996d5c12070SMiao Xie async_transid, inherit); 9973de4586cSChris Mason } 99876dda93cSYan, Zheng if (!error) 99976dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 100076dda93cSYan, Zheng out_up_read: 10010b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 1002cb8e7090SChristoph Hellwig out_dput: 1003cb8e7090SChristoph Hellwig dput(dentry); 1004cb8e7090SChristoph Hellwig out_unlock: 10055955102cSAl Viro inode_unlock(dir); 1006cb8e7090SChristoph Hellwig return error; 1007cb8e7090SChristoph Hellwig } 1008cb8e7090SChristoph Hellwig 10094cb5300bSChris Mason /* 10104cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 10114cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 10124cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 10134cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 10144cb5300bSChris Mason * part of the file 10154cb5300bSChris Mason */ 1016aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 10174cb5300bSChris Mason { 10184cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10194cb5300bSChris Mason struct extent_map *em = NULL; 10204cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10214cb5300bSChris Mason u64 end; 10224cb5300bSChris Mason 10234cb5300bSChris Mason read_lock(&em_tree->lock); 102409cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10254cb5300bSChris Mason read_unlock(&em_tree->lock); 10264cb5300bSChris Mason 10274cb5300bSChris Mason if (em) { 10284cb5300bSChris Mason end = extent_map_end(em); 10294cb5300bSChris Mason free_extent_map(em); 10304cb5300bSChris Mason if (end - offset > thresh) 10314cb5300bSChris Mason return 0; 10324cb5300bSChris Mason } 10334cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10344cb5300bSChris Mason thresh /= 2; 10354cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10364cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10374cb5300bSChris Mason if (end >= thresh) 10384cb5300bSChris Mason return 0; 10394cb5300bSChris Mason return 1; 10404cb5300bSChris Mason } 10414cb5300bSChris Mason 10424cb5300bSChris Mason /* 10434cb5300bSChris Mason * helper function to walk through a file and find extents 10444cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10454cb5300bSChris Mason * 10464cb5300bSChris Mason * This is used by the defragging code to find new and small 10474cb5300bSChris Mason * extents 10484cb5300bSChris Mason */ 10494cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10504cb5300bSChris Mason struct inode *inode, u64 newer_than, 1051aab110abSDavid Sterba u64 *off, u32 thresh) 10524cb5300bSChris Mason { 10534cb5300bSChris Mason struct btrfs_path *path; 10544cb5300bSChris Mason struct btrfs_key min_key; 10554cb5300bSChris Mason struct extent_buffer *leaf; 10564cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10574cb5300bSChris Mason int type; 10584cb5300bSChris Mason int ret; 10594a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10604cb5300bSChris Mason 10614cb5300bSChris Mason path = btrfs_alloc_path(); 10624cb5300bSChris Mason if (!path) 10634cb5300bSChris Mason return -ENOMEM; 10644cb5300bSChris Mason 1065a4689d2bSDavid Sterba min_key.objectid = ino; 10664cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10674cb5300bSChris Mason min_key.offset = *off; 10684cb5300bSChris Mason 10694cb5300bSChris Mason while (1) { 10706174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10714cb5300bSChris Mason if (ret != 0) 10724cb5300bSChris Mason goto none; 1073f094c9bdSFilipe Manana process_slot: 1074a4689d2bSDavid Sterba if (min_key.objectid != ino) 10754cb5300bSChris Mason goto none; 10764cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10774cb5300bSChris Mason goto none; 10784cb5300bSChris Mason 10794cb5300bSChris Mason leaf = path->nodes[0]; 10804cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10814cb5300bSChris Mason struct btrfs_file_extent_item); 10824cb5300bSChris Mason 10834cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10844cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10854cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10864cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10874cb5300bSChris Mason *off = min_key.offset; 10884cb5300bSChris Mason btrfs_free_path(path); 10894cb5300bSChris Mason return 0; 10904cb5300bSChris Mason } 10914cb5300bSChris Mason 1092f094c9bdSFilipe Manana path->slots[0]++; 1093f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1094f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1095f094c9bdSFilipe Manana goto process_slot; 1096f094c9bdSFilipe Manana } 1097f094c9bdSFilipe Manana 10984cb5300bSChris Mason if (min_key.offset == (u64)-1) 10994cb5300bSChris Mason goto none; 11004cb5300bSChris Mason 11014cb5300bSChris Mason min_key.offset++; 11024cb5300bSChris Mason btrfs_release_path(path); 11034cb5300bSChris Mason } 11044cb5300bSChris Mason none: 11054cb5300bSChris Mason btrfs_free_path(path); 11064cb5300bSChris Mason return -ENOENT; 11074cb5300bSChris Mason } 11084cb5300bSChris Mason 11096c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 111017ce6ef8SLiu Bo { 111117ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1112940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 11136c282eb4SLi Zefan struct extent_map *em; 111409cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1115940100a4SChris Mason 1116940100a4SChris Mason /* 1117940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1118940100a4SChris Mason * the full extent lock 1119940100a4SChris Mason */ 1120940100a4SChris Mason read_lock(&em_tree->lock); 1121940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1122940100a4SChris Mason read_unlock(&em_tree->lock); 1123940100a4SChris Mason 1124940100a4SChris Mason if (!em) { 1125308d9800SFilipe Manana struct extent_state *cached = NULL; 1126308d9800SFilipe Manana u64 end = start + len - 1; 1127308d9800SFilipe Manana 1128940100a4SChris Mason /* get the big lock and read metadata off disk */ 1129ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 113039b07b5dSOmar Sandoval em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len); 1131e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1132940100a4SChris Mason 11336cf8bfbfSDan Carpenter if (IS_ERR(em)) 11346c282eb4SLi Zefan return NULL; 1135940100a4SChris Mason } 1136940100a4SChris Mason 11376c282eb4SLi Zefan return em; 11386c282eb4SLi Zefan } 11396c282eb4SLi Zefan 11406c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11416c282eb4SLi Zefan { 11426c282eb4SLi Zefan struct extent_map *next; 11436c282eb4SLi Zefan bool ret = true; 11446c282eb4SLi Zefan 11456c282eb4SLi Zefan /* this is the last extent */ 11466c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11476c282eb4SLi Zefan return false; 11486c282eb4SLi Zefan 11496c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1150e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1151e9512d72SChris Mason ret = false; 1152e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1153ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11546c282eb4SLi Zefan ret = false; 11556c282eb4SLi Zefan 11566c282eb4SLi Zefan free_extent_map(next); 11576c282eb4SLi Zefan return ret; 11586c282eb4SLi Zefan } 11596c282eb4SLi Zefan 1160aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1161a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1162a43a2111SAndrew Mahone int compress) 11636c282eb4SLi Zefan { 11646c282eb4SLi Zefan struct extent_map *em; 11656c282eb4SLi Zefan int ret = 1; 11666c282eb4SLi Zefan bool next_mergeable = true; 11674a3560c4SLiu Bo bool prev_mergeable = true; 11686c282eb4SLi Zefan 11696c282eb4SLi Zefan /* 11706c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11716c282eb4SLi Zefan * defragging it 11726c282eb4SLi Zefan */ 11736c282eb4SLi Zefan if (start < *defrag_end) 11746c282eb4SLi Zefan return 1; 11756c282eb4SLi Zefan 11766c282eb4SLi Zefan *skip = 0; 11776c282eb4SLi Zefan 11786c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11796c282eb4SLi Zefan if (!em) 11806c282eb4SLi Zefan return 0; 11816c282eb4SLi Zefan 1182940100a4SChris Mason /* this will cover holes, and inline extents */ 118317ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1184940100a4SChris Mason ret = 0; 118517ce6ef8SLiu Bo goto out; 118617ce6ef8SLiu Bo } 118717ce6ef8SLiu Bo 11884a3560c4SLiu Bo if (!*defrag_end) 11894a3560c4SLiu Bo prev_mergeable = false; 11904a3560c4SLiu Bo 11916c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1192940100a4SChris Mason /* 11936c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11946c282eb4SLi Zefan * real extent, don't bother defragging it 1195940100a4SChris Mason */ 1196a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11974a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1198940100a4SChris Mason ret = 0; 119917ce6ef8SLiu Bo out: 1200940100a4SChris Mason /* 1201940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1202940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1203940100a4SChris Mason * so that the next tiny extent will force a defrag. 1204940100a4SChris Mason * 1205940100a4SChris Mason * The end result of this is that tiny extents before a single big 1206940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1207940100a4SChris Mason */ 1208940100a4SChris Mason if (ret) { 1209940100a4SChris Mason *defrag_end = extent_map_end(em); 1210940100a4SChris Mason } else { 1211940100a4SChris Mason *last_len = 0; 1212940100a4SChris Mason *skip = extent_map_end(em); 1213940100a4SChris Mason *defrag_end = 0; 1214940100a4SChris Mason } 1215940100a4SChris Mason 1216940100a4SChris Mason free_extent_map(em); 1217940100a4SChris Mason return ret; 1218940100a4SChris Mason } 1219940100a4SChris Mason 12204cb5300bSChris Mason /* 12214cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12224cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12234cb5300bSChris Mason * to COW and defrag. 12244cb5300bSChris Mason * 12254cb5300bSChris Mason * It also makes sure the delalloc code has enough 12264cb5300bSChris Mason * dirty data to avoid making new small extents as part 12274cb5300bSChris Mason * of the defrag 12284cb5300bSChris Mason * 12294cb5300bSChris Mason * It's a good idea to start RA on this range 12304cb5300bSChris Mason * before calling this. 12314cb5300bSChris Mason */ 12324cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12334cb5300bSChris Mason struct page **pages, 12344cb5300bSChris Mason unsigned long start_index, 1235c41570c9SJustin Maggard unsigned long num_pages) 1236f46b5a66SChristoph Hellwig { 12374cb5300bSChris Mason unsigned long file_end; 12384cb5300bSChris Mason u64 isize = i_size_read(inode); 1239f46b5a66SChristoph Hellwig u64 page_start; 1240f46b5a66SChristoph Hellwig u64 page_end; 12411f12bd06SLiu Bo u64 page_cnt; 12424cb5300bSChris Mason int ret; 12434cb5300bSChris Mason int i; 12444cb5300bSChris Mason int i_done; 12454cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12464cb5300bSChris Mason struct extent_state *cached_state = NULL; 1247600a45e1SMiao Xie struct extent_io_tree *tree; 1248364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12493b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12504cb5300bSChris Mason 125109cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12521f12bd06SLiu Bo if (!isize || start_index > file_end) 12531f12bd06SLiu Bo return 0; 12541f12bd06SLiu Bo 12551f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12564cb5300bSChris Mason 1257364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 125809cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 125909cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12604cb5300bSChris Mason if (ret) 12614cb5300bSChris Mason return ret; 12624cb5300bSChris Mason i_done = 0; 1263600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12644cb5300bSChris Mason 12654cb5300bSChris Mason /* step one, lock all the pages */ 12661f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12674cb5300bSChris Mason struct page *page; 1268600a45e1SMiao Xie again: 1269a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12703b16a4e3SJosef Bacik start_index + i, mask); 12714cb5300bSChris Mason if (!page) 12724cb5300bSChris Mason break; 12734cb5300bSChris Mason 1274600a45e1SMiao Xie page_start = page_offset(page); 127509cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1276600a45e1SMiao Xie while (1) { 1277308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1278ff13db41SDavid Sterba &cached_state); 1279600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1280600a45e1SMiao Xie page_start); 1281308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1282e43bbe5eSDavid Sterba &cached_state); 1283600a45e1SMiao Xie if (!ordered) 1284600a45e1SMiao Xie break; 1285600a45e1SMiao Xie 1286600a45e1SMiao Xie unlock_page(page); 1287600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1288600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1289600a45e1SMiao Xie lock_page(page); 12901f12bd06SLiu Bo /* 12911f12bd06SLiu Bo * we unlocked the page above, so we need check if 12921f12bd06SLiu Bo * it was released or not. 12931f12bd06SLiu Bo */ 12941f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12951f12bd06SLiu Bo unlock_page(page); 129609cbfeafSKirill A. Shutemov put_page(page); 12971f12bd06SLiu Bo goto again; 12981f12bd06SLiu Bo } 1299600a45e1SMiao Xie } 1300600a45e1SMiao Xie 13014cb5300bSChris Mason if (!PageUptodate(page)) { 13024cb5300bSChris Mason btrfs_readpage(NULL, page); 13034cb5300bSChris Mason lock_page(page); 13044cb5300bSChris Mason if (!PageUptodate(page)) { 13054cb5300bSChris Mason unlock_page(page); 130609cbfeafSKirill A. Shutemov put_page(page); 13074cb5300bSChris Mason ret = -EIO; 13084cb5300bSChris Mason break; 13094cb5300bSChris Mason } 13104cb5300bSChris Mason } 1311600a45e1SMiao Xie 1312600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1313600a45e1SMiao Xie unlock_page(page); 131409cbfeafSKirill A. Shutemov put_page(page); 1315600a45e1SMiao Xie goto again; 1316600a45e1SMiao Xie } 1317600a45e1SMiao Xie 13184cb5300bSChris Mason pages[i] = page; 13194cb5300bSChris Mason i_done++; 13204cb5300bSChris Mason } 13214cb5300bSChris Mason if (!i_done || ret) 13224cb5300bSChris Mason goto out; 13234cb5300bSChris Mason 13241751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13254cb5300bSChris Mason goto out; 13264cb5300bSChris Mason 13274cb5300bSChris Mason /* 13284cb5300bSChris Mason * so now we have a nice long stream of locked 13294cb5300bSChris Mason * and up to date pages, lets wait on them 13304cb5300bSChris Mason */ 13314cb5300bSChris Mason for (i = 0; i < i_done; i++) 13324cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13334cb5300bSChris Mason 13344cb5300bSChris Mason page_start = page_offset(pages[0]); 133509cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13364cb5300bSChris Mason 13374cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1338ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13394cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 1340e182163dSOmar Sandoval page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | 1341e182163dSOmar Sandoval EXTENT_DEFRAG, 0, 0, &cached_state); 13424cb5300bSChris Mason 13431f12bd06SLiu Bo if (i_done != page_cnt) { 13449e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 134528c4a3e2SSu Yue btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); 13469e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1347bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 134809cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 134943b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13504cb5300bSChris Mason } 13514cb5300bSChris Mason 13524cb5300bSChris Mason 13539e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1354018ed4f7SDavid Sterba &cached_state); 13554cb5300bSChris Mason 13564cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1357e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13584cb5300bSChris Mason 13594cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13604cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13614cb5300bSChris Mason ClearPageChecked(pages[i]); 13624cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13634cb5300bSChris Mason set_page_dirty(pages[i]); 13644cb5300bSChris Mason unlock_page(pages[i]); 136509cbfeafSKirill A. Shutemov put_page(pages[i]); 13664cb5300bSChris Mason } 13678702ba93SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT); 1368364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13694cb5300bSChris Mason return i_done; 13704cb5300bSChris Mason out: 13714cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13724cb5300bSChris Mason unlock_page(pages[i]); 137309cbfeafSKirill A. Shutemov put_page(pages[i]); 13744cb5300bSChris Mason } 1375bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 137609cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 137743b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 13788702ba93SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT); 1379364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13804cb5300bSChris Mason return ret; 13814cb5300bSChris Mason 13824cb5300bSChris Mason } 13834cb5300bSChris Mason 13844cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13854cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13864cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13874cb5300bSChris Mason { 13880b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13894cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13904cb5300bSChris Mason struct file_ra_state *ra = NULL; 13914cb5300bSChris Mason unsigned long last_index; 1392151a31b2SLi Zefan u64 isize = i_size_read(inode); 1393940100a4SChris Mason u64 last_len = 0; 1394940100a4SChris Mason u64 skip = 0; 1395940100a4SChris Mason u64 defrag_end = 0; 13964cb5300bSChris Mason u64 newer_off = range->start; 1397f46b5a66SChristoph Hellwig unsigned long i; 1398008873eaSLi Zefan unsigned long ra_index = 0; 1399f46b5a66SChristoph Hellwig int ret; 14004cb5300bSChris Mason int defrag_count = 0; 14011a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1402aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 140309cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1404c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1405ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 14064cb5300bSChris Mason struct page **pages = NULL; 14071e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14084cb5300bSChris Mason 14090abd5b17SLiu Bo if (isize == 0) 14100abd5b17SLiu Bo return 0; 14110abd5b17SLiu Bo 14120abd5b17SLiu Bo if (range->start >= isize) 14130abd5b17SLiu Bo return -EINVAL; 14141a419d85SLi Zefan 14151e2ef46dSDavid Sterba if (do_compress) { 1416ce96b7ffSChengguang Xu if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES) 14171a419d85SLi Zefan return -EINVAL; 14181a419d85SLi Zefan if (range->compress_type) 14191a419d85SLi Zefan compress_type = range->compress_type; 14201a419d85SLi Zefan } 1421f46b5a66SChristoph Hellwig 14220abd5b17SLiu Bo if (extent_thresh == 0) 1423ee22184bSByongho Lee extent_thresh = SZ_256K; 1424f46b5a66SChristoph Hellwig 14254cb5300bSChris Mason /* 14260a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14270a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14280a52d108SDavid Sterba * we don't error out. 14294cb5300bSChris Mason */ 14304cb5300bSChris Mason if (!file) { 143163e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14320a52d108SDavid Sterba if (ra) 14334cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14344cb5300bSChris Mason } else { 14354cb5300bSChris Mason ra = &file->f_ra; 14364cb5300bSChris Mason } 14374cb5300bSChris Mason 143863e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14394cb5300bSChris Mason if (!pages) { 14404cb5300bSChris Mason ret = -ENOMEM; 14414cb5300bSChris Mason goto out_ra; 14424cb5300bSChris Mason } 14434cb5300bSChris Mason 14444cb5300bSChris Mason /* find the last page to defrag */ 14451e701a32SChris Mason if (range->start + range->len > range->start) { 1446151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 144709cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14481e701a32SChris Mason } else { 144909cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14501e701a32SChris Mason } 14511e701a32SChris Mason 14524cb5300bSChris Mason if (newer_than) { 14534cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1454ee22184bSByongho Lee &newer_off, SZ_64K); 14554cb5300bSChris Mason if (!ret) { 14564cb5300bSChris Mason range->start = newer_off; 14574cb5300bSChris Mason /* 14584cb5300bSChris Mason * we always align our defrag to help keep 14594cb5300bSChris Mason * the extents in the file evenly spaced 14604cb5300bSChris Mason */ 146109cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14624cb5300bSChris Mason } else 14634cb5300bSChris Mason goto out_ra; 14644cb5300bSChris Mason } else { 146509cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14664cb5300bSChris Mason } 14674cb5300bSChris Mason if (!max_to_defrag) 1468070034bdSchandan max_to_defrag = last_index - i + 1; 14694cb5300bSChris Mason 14702a0f7f57SLi Zefan /* 14712a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14722a0f7f57SLi Zefan * written sequentially. 14732a0f7f57SLi Zefan */ 14742a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14752a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14762a0f7f57SLi Zefan 1477f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 147809cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14794cb5300bSChris Mason /* 14804cb5300bSChris Mason * make sure we stop running if someone unmounts 14814cb5300bSChris Mason * the FS 14824cb5300bSChris Mason */ 14831751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14844cb5300bSChris Mason break; 14854cb5300bSChris Mason 14860b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14870b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1488210549ebSDavid Sterba ret = -EAGAIN; 1489210549ebSDavid Sterba break; 1490210549ebSDavid Sterba } 1491210549ebSDavid Sterba 149209cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14936c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14941e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1495940100a4SChris Mason unsigned long next; 1496940100a4SChris Mason /* 1497940100a4SChris Mason * the should_defrag function tells us how much to skip 1498940100a4SChris Mason * bump our counter by the suggested amount 1499940100a4SChris Mason */ 150009cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1501940100a4SChris Mason i = max(i + 1, next); 1502940100a4SChris Mason continue; 1503940100a4SChris Mason } 1504008873eaSLi Zefan 1505008873eaSLi Zefan if (!newer_than) { 150609cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 150709cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1508008873eaSLi Zefan cluster = min(cluster, max_cluster); 1509008873eaSLi Zefan } else { 1510008873eaSLi Zefan cluster = max_cluster; 1511008873eaSLi Zefan } 1512008873eaSLi Zefan 1513008873eaSLi Zefan if (i + cluster > ra_index) { 1514008873eaSLi Zefan ra_index = max(i, ra_index); 15150a52d108SDavid Sterba if (ra) 1516d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1517d3c0bab5SDavid Sterba file, ra_index, cluster); 1518e4826a5bSchandan ra_index += cluster; 1519008873eaSLi Zefan } 15204cb5300bSChris Mason 15215955102cSAl Viro inode_lock(inode); 1522eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1523eede2bf3SOmar Sandoval ret = -ETXTBSY; 1524eede2bf3SOmar Sandoval } else { 15251e2ef46dSDavid Sterba if (do_compress) 1526eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1527008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1528eede2bf3SOmar Sandoval } 1529ecb8bea8SLiu Bo if (ret < 0) { 15305955102cSAl Viro inode_unlock(inode); 15314cb5300bSChris Mason goto out_ra; 1532ecb8bea8SLiu Bo } 15334cb5300bSChris Mason 15344cb5300bSChris Mason defrag_count += ret; 1535d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15365955102cSAl Viro inode_unlock(inode); 15374cb5300bSChris Mason 15384cb5300bSChris Mason if (newer_than) { 15394cb5300bSChris Mason if (newer_off == (u64)-1) 15404cb5300bSChris Mason break; 15414cb5300bSChris Mason 1542e1f041e1SLiu Bo if (ret > 0) 1543e1f041e1SLiu Bo i += ret; 1544e1f041e1SLiu Bo 15454cb5300bSChris Mason newer_off = max(newer_off + 1, 154609cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15474cb5300bSChris Mason 1548ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1549ee22184bSByongho Lee &newer_off, SZ_64K); 15504cb5300bSChris Mason if (!ret) { 15514cb5300bSChris Mason range->start = newer_off; 155209cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15534cb5300bSChris Mason } else { 15544cb5300bSChris Mason break; 1555940100a4SChris Mason } 15564cb5300bSChris Mason } else { 1557008873eaSLi Zefan if (ret > 0) { 1558cbcc8326SLi Zefan i += ret; 155909cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1560008873eaSLi Zefan } else { 1561940100a4SChris Mason i++; 1562008873eaSLi Zefan last_len = 0; 1563008873eaSLi Zefan } 1564f46b5a66SChristoph Hellwig } 15654cb5300bSChris Mason } 1566f46b5a66SChristoph Hellwig 1567dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15681e701a32SChris Mason filemap_flush(inode->i_mapping); 1569dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1570dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1571dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1572dec8ef90SFilipe Manana } 15731e701a32SChris Mason 15741a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15750b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15765c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15775c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15781a419d85SLi Zefan } 15791a419d85SLi Zefan 158060ccf82fSDiego Calleja ret = defrag_count; 1581940100a4SChris Mason 15824cb5300bSChris Mason out_ra: 15831e2ef46dSDavid Sterba if (do_compress) { 15845955102cSAl Viro inode_lock(inode); 1585eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15865955102cSAl Viro inode_unlock(inode); 1587633085c7SFilipe David Borba Manana } 15884cb5300bSChris Mason if (!file) 15894cb5300bSChris Mason kfree(ra); 15904cb5300bSChris Mason kfree(pages); 1591940100a4SChris Mason return ret; 1592f46b5a66SChristoph Hellwig } 1593f46b5a66SChristoph Hellwig 1594198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 159576dda93cSYan, Zheng void __user *arg) 1596f46b5a66SChristoph Hellwig { 15970b246afaSJeff Mahoney struct inode *inode = file_inode(file); 15980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1599f46b5a66SChristoph Hellwig u64 new_size; 1600f46b5a66SChristoph Hellwig u64 old_size; 1601f46b5a66SChristoph Hellwig u64 devid = 1; 16020b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1603f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1604f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1605f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1606f46b5a66SChristoph Hellwig char *sizestr; 16079a40f122SGui Hecheng char *retptr; 1608f46b5a66SChristoph Hellwig char *devstr = NULL; 1609f46b5a66SChristoph Hellwig int ret = 0; 1610f46b5a66SChristoph Hellwig int mod = 0; 1611f46b5a66SChristoph Hellwig 1612e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1613e441d54dSChris Mason return -EPERM; 1614e441d54dSChris Mason 1615198605a8SMiao Xie ret = mnt_want_write_file(file); 1616198605a8SMiao Xie if (ret) 1617198605a8SMiao Xie return ret; 1618198605a8SMiao Xie 1619171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 162097547676SMiao Xie mnt_drop_write_file(file); 1621e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1622c9e9f97bSIlya Dryomov } 1623c9e9f97bSIlya Dryomov 1624dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1625c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1626c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1627c9e9f97bSIlya Dryomov goto out; 1628c9e9f97bSIlya Dryomov } 16295516e595SMark Fasheh 16305516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1631f46b5a66SChristoph Hellwig 1632f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1633f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1634f46b5a66SChristoph Hellwig if (devstr) { 1635f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1636f46b5a66SChristoph Hellwig *devstr = '\0'; 1637f46b5a66SChristoph Hellwig devstr = vol_args->name; 163858dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 163958dfae63SZhangZhen if (ret) 164058dfae63SZhangZhen goto out_free; 1641dfd79829SMiao Xie if (!devid) { 1642dfd79829SMiao Xie ret = -EINVAL; 1643dfd79829SMiao Xie goto out_free; 1644dfd79829SMiao Xie } 16450b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1646f46b5a66SChristoph Hellwig } 1647dba60f3fSMiao Xie 164809ba3bc9SAnand Jain device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true); 1649f46b5a66SChristoph Hellwig if (!device) { 16500b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1651c1c9ff7cSGeert Uytterhoeven devid); 1652dfd79829SMiao Xie ret = -ENODEV; 1653c9e9f97bSIlya Dryomov goto out_free; 1654f46b5a66SChristoph Hellwig } 1655dba60f3fSMiao Xie 1656ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16570b246afaSJeff Mahoney btrfs_info(fs_info, 1658efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1659c1c9ff7cSGeert Uytterhoeven devid); 1660dfd79829SMiao Xie ret = -EPERM; 16614e42ae1bSLiu Bo goto out_free; 16624e42ae1bSLiu Bo } 16634e42ae1bSLiu Bo 1664f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1665f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1666f46b5a66SChristoph Hellwig else { 1667f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1668f46b5a66SChristoph Hellwig mod = -1; 1669f46b5a66SChristoph Hellwig sizestr++; 1670f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1671f46b5a66SChristoph Hellwig mod = 1; 1672f46b5a66SChristoph Hellwig sizestr++; 1673f46b5a66SChristoph Hellwig } 16749a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16759a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1676f46b5a66SChristoph Hellwig ret = -EINVAL; 1677c9e9f97bSIlya Dryomov goto out_free; 1678f46b5a66SChristoph Hellwig } 1679f46b5a66SChristoph Hellwig } 1680f46b5a66SChristoph Hellwig 1681401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1682dfd79829SMiao Xie ret = -EPERM; 168363a212abSStefan Behrens goto out_free; 168463a212abSStefan Behrens } 168563a212abSStefan Behrens 16867cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1687f46b5a66SChristoph Hellwig 1688f46b5a66SChristoph Hellwig if (mod < 0) { 1689f46b5a66SChristoph Hellwig if (new_size > old_size) { 1690f46b5a66SChristoph Hellwig ret = -EINVAL; 1691c9e9f97bSIlya Dryomov goto out_free; 1692f46b5a66SChristoph Hellwig } 1693f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1694f46b5a66SChristoph Hellwig } else if (mod > 0) { 1695eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1696902c68a4SGui Hecheng ret = -ERANGE; 1697eb8052e0SWenliang Fan goto out_free; 1698eb8052e0SWenliang Fan } 1699f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1700f46b5a66SChristoph Hellwig } 1701f46b5a66SChristoph Hellwig 1702ee22184bSByongho Lee if (new_size < SZ_256M) { 1703f46b5a66SChristoph Hellwig ret = -EINVAL; 1704c9e9f97bSIlya Dryomov goto out_free; 1705f46b5a66SChristoph Hellwig } 1706f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1707f46b5a66SChristoph Hellwig ret = -EFBIG; 1708c9e9f97bSIlya Dryomov goto out_free; 1709f46b5a66SChristoph Hellwig } 1710f46b5a66SChristoph Hellwig 171147f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1712f46b5a66SChristoph Hellwig 1713f46b5a66SChristoph Hellwig if (new_size > old_size) { 1714a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 171598d5dc13STsutomu Itoh if (IS_ERR(trans)) { 171698d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1717c9e9f97bSIlya Dryomov goto out_free; 171898d5dc13STsutomu Itoh } 1719f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17203a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1721ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1722f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17230253f40eSjeff.liu } /* equal, nothing need to do */ 1724f46b5a66SChristoph Hellwig 1725faf8f7b9SMarcos Paulo de Souza if (ret == 0 && new_size != old_size) 1726faf8f7b9SMarcos Paulo de Souza btrfs_info_in_rcu(fs_info, 1727faf8f7b9SMarcos Paulo de Souza "resize device %s (devid %llu) from %llu to %llu", 1728faf8f7b9SMarcos Paulo de Souza rcu_str_deref(device->name), device->devid, 1729faf8f7b9SMarcos Paulo de Souza old_size, new_size); 1730c9e9f97bSIlya Dryomov out_free: 1731f46b5a66SChristoph Hellwig kfree(vol_args); 1732c9e9f97bSIlya Dryomov out: 1733171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 173418f39c41SIlya Dryomov mnt_drop_write_file(file); 1735f46b5a66SChristoph Hellwig return ret; 1736f46b5a66SChristoph Hellwig } 1737f46b5a66SChristoph Hellwig 173872fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 173952f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17406f72c7e2SArne Jansen u64 *transid, bool readonly, 17418696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1742f46b5a66SChristoph Hellwig { 1743f46b5a66SChristoph Hellwig int namelen; 17443de4586cSChris Mason int ret = 0; 1745f46b5a66SChristoph Hellwig 1746325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1747325c50e3SJeff Mahoney return -ENOTDIR; 1748325c50e3SJeff Mahoney 1749a874a63eSLiu Bo ret = mnt_want_write_file(file); 1750a874a63eSLiu Bo if (ret) 1751a874a63eSLiu Bo goto out; 1752a874a63eSLiu Bo 175372fd032eSSage Weil namelen = strlen(name); 175472fd032eSSage Weil if (strchr(name, '/')) { 1755f46b5a66SChristoph Hellwig ret = -EINVAL; 1756a874a63eSLiu Bo goto out_drop_write; 1757f46b5a66SChristoph Hellwig } 1758f46b5a66SChristoph Hellwig 175916780cabSChris Mason if (name[0] == '.' && 176016780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 176116780cabSChris Mason ret = -EEXIST; 1762a874a63eSLiu Bo goto out_drop_write; 176316780cabSChris Mason } 176416780cabSChris Mason 17653de4586cSChris Mason if (subvol) { 176672fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17676f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1768cb8e7090SChristoph Hellwig } else { 17692903ff01SAl Viro struct fd src = fdget(fd); 17703de4586cSChris Mason struct inode *src_inode; 17712903ff01SAl Viro if (!src.file) { 17723de4586cSChris Mason ret = -EINVAL; 1773a874a63eSLiu Bo goto out_drop_write; 17743de4586cSChris Mason } 17753de4586cSChris Mason 1776496ad9aaSAl Viro src_inode = file_inode(src.file); 1777496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1778c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1779efe120a0SFrank Holton "Snapshot src from another FS"); 178023ad5b17SKusanagi Kouichi ret = -EXDEV; 1781d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1782d0242061SDavid Sterba /* 1783d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1784d0242061SDavid Sterba * are limited to own subvolumes only 1785d0242061SDavid Sterba */ 1786d0242061SDavid Sterba ret = -EPERM; 1787ecd18815SAl Viro } else { 178872fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 178972fd032eSSage Weil BTRFS_I(src_inode)->root, 17906f72c7e2SArne Jansen transid, readonly, inherit); 1791ecd18815SAl Viro } 17922903ff01SAl Viro fdput(src); 1793cb8e7090SChristoph Hellwig } 1794a874a63eSLiu Bo out_drop_write: 1795a874a63eSLiu Bo mnt_drop_write_file(file); 1796f46b5a66SChristoph Hellwig out: 179772fd032eSSage Weil return ret; 179872fd032eSSage Weil } 179972fd032eSSage Weil 180072fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1801fa0d2b9bSLi Zefan void __user *arg, int subvol) 180272fd032eSSage Weil { 1803fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 180472fd032eSSage Weil int ret; 180572fd032eSSage Weil 1806325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1807325c50e3SJeff Mahoney return -ENOTDIR; 1808325c50e3SJeff Mahoney 1809fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1810fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1811fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1812fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1813fa0d2b9bSLi Zefan 1814fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1815b83cc969SLi Zefan vol_args->fd, subvol, 18166f72c7e2SArne Jansen NULL, false, NULL); 1817fa0d2b9bSLi Zefan 1818fa0d2b9bSLi Zefan kfree(vol_args); 1819fa0d2b9bSLi Zefan return ret; 1820fa0d2b9bSLi Zefan } 1821fa0d2b9bSLi Zefan 1822fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1823fa0d2b9bSLi Zefan void __user *arg, int subvol) 1824fa0d2b9bSLi Zefan { 1825fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1826fa0d2b9bSLi Zefan int ret; 1827fdfb1e4fSLi Zefan u64 transid = 0; 1828fdfb1e4fSLi Zefan u64 *ptr = NULL; 1829b83cc969SLi Zefan bool readonly = false; 18306f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 183172fd032eSSage Weil 1832325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1833325c50e3SJeff Mahoney return -ENOTDIR; 1834325c50e3SJeff Mahoney 1835fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1836fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1837fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1838fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1839fdfb1e4fSLi Zefan 1840673990dbSDavid Sterba if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) { 1841b83cc969SLi Zefan ret = -EOPNOTSUPP; 1842c47ca32dSDan Carpenter goto free_args; 1843fdfb1e4fSLi Zefan } 1844fdfb1e4fSLi Zefan 1845ebc87351SNikolay Borisov if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1846ebc87351SNikolay Borisov struct inode *inode = file_inode(file); 1847ebc87351SNikolay Borisov struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1848ebc87351SNikolay Borisov 1849ebc87351SNikolay Borisov btrfs_warn(fs_info, 1850ebc87351SNikolay Borisov "SNAP_CREATE_V2 ioctl with CREATE_ASYNC is deprecated and will be removed in kernel 5.7"); 1851ebc87351SNikolay Borisov 1852fdfb1e4fSLi Zefan ptr = &transid; 1853ebc87351SNikolay Borisov } 1854b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1855b83cc969SLi Zefan readonly = true; 18566f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 185709cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18586f72c7e2SArne Jansen ret = -EINVAL; 1859c47ca32dSDan Carpenter goto free_args; 18606f72c7e2SArne Jansen } 18616f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18626f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18636f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1864c47ca32dSDan Carpenter goto free_args; 18656f72c7e2SArne Jansen } 18666f72c7e2SArne Jansen } 186775eaa0e2SSage Weil 1868fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18696f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18708696c533SMiao Xie readonly, inherit); 1871c47ca32dSDan Carpenter if (ret) 1872c47ca32dSDan Carpenter goto free_inherit; 187375eaa0e2SSage Weil 1874c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1875fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1876c47ca32dSDan Carpenter transid), 1877c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 187875eaa0e2SSage Weil ret = -EFAULT; 1879c47ca32dSDan Carpenter 1880c47ca32dSDan Carpenter free_inherit: 18816f72c7e2SArne Jansen kfree(inherit); 1882c47ca32dSDan Carpenter free_args: 1883c47ca32dSDan Carpenter kfree(vol_args); 1884f46b5a66SChristoph Hellwig return ret; 1885f46b5a66SChristoph Hellwig } 1886f46b5a66SChristoph Hellwig 18870caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18880caa102dSLi Zefan void __user *arg) 18890caa102dSLi Zefan { 1890496ad9aaSAl Viro struct inode *inode = file_inode(file); 18910b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18920caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18930caa102dSLi Zefan int ret = 0; 18940caa102dSLi Zefan u64 flags = 0; 18950caa102dSLi Zefan 18964a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18970caa102dSLi Zefan return -EINVAL; 18980caa102dSLi Zefan 18990b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 19000caa102dSLi Zefan if (btrfs_root_readonly(root)) 19010caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 19020b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 19030caa102dSLi Zefan 19040caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 19050caa102dSLi Zefan ret = -EFAULT; 19060caa102dSLi Zefan 19070caa102dSLi Zefan return ret; 19080caa102dSLi Zefan } 19090caa102dSLi Zefan 19100caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19110caa102dSLi Zefan void __user *arg) 19120caa102dSLi Zefan { 1913496ad9aaSAl Viro struct inode *inode = file_inode(file); 19140b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19150caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19160caa102dSLi Zefan struct btrfs_trans_handle *trans; 19170caa102dSLi Zefan u64 root_flags; 19180caa102dSLi Zefan u64 flags; 19190caa102dSLi Zefan int ret = 0; 19200caa102dSLi Zefan 1921bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1922bd60ea0fSDavid Sterba return -EPERM; 1923bd60ea0fSDavid Sterba 1924b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1925b9ca0664SLiu Bo if (ret) 1926b9ca0664SLiu Bo goto out; 19270caa102dSLi Zefan 19284a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1929b9ca0664SLiu Bo ret = -EINVAL; 1930b9ca0664SLiu Bo goto out_drop_write; 1931b9ca0664SLiu Bo } 19320caa102dSLi Zefan 1933b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1934b9ca0664SLiu Bo ret = -EFAULT; 1935b9ca0664SLiu Bo goto out_drop_write; 1936b9ca0664SLiu Bo } 19370caa102dSLi Zefan 1938b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1939b9ca0664SLiu Bo ret = -EINVAL; 1940b9ca0664SLiu Bo goto out_drop_write; 1941b9ca0664SLiu Bo } 19420caa102dSLi Zefan 1943b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1944b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1945b9ca0664SLiu Bo goto out_drop_write; 1946b9ca0664SLiu Bo } 19470caa102dSLi Zefan 19480b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19490caa102dSLi Zefan 19500caa102dSLi Zefan /* nothing to do */ 19510caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1952b9ca0664SLiu Bo goto out_drop_sem; 19530caa102dSLi Zefan 19540caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19552c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19560caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19570caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19582c686537SDavid Sterba } else { 19592c686537SDavid Sterba /* 19602c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19612c686537SDavid Sterba * send 19622c686537SDavid Sterba */ 19632c686537SDavid Sterba spin_lock(&root->root_item_lock); 19642c686537SDavid Sterba if (root->send_in_progress == 0) { 19650caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19660caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19672c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19682c686537SDavid Sterba } else { 19692c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19700b246afaSJeff Mahoney btrfs_warn(fs_info, 19712c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19722c686537SDavid Sterba root->root_key.objectid); 19732c686537SDavid Sterba ret = -EPERM; 19742c686537SDavid Sterba goto out_drop_sem; 19752c686537SDavid Sterba } 19762c686537SDavid Sterba } 19770caa102dSLi Zefan 19780caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19790caa102dSLi Zefan if (IS_ERR(trans)) { 19800caa102dSLi Zefan ret = PTR_ERR(trans); 19810caa102dSLi Zefan goto out_reset; 19820caa102dSLi Zefan } 19830caa102dSLi Zefan 19840b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19850caa102dSLi Zefan &root->root_key, &root->root_item); 19869417ebc8SNikolay Borisov if (ret < 0) { 19879417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19889417ebc8SNikolay Borisov goto out_reset; 19899417ebc8SNikolay Borisov } 19900caa102dSLi Zefan 19919417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19929417ebc8SNikolay Borisov 19930caa102dSLi Zefan out_reset: 19940caa102dSLi Zefan if (ret) 19950caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1996b9ca0664SLiu Bo out_drop_sem: 19970b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1998b9ca0664SLiu Bo out_drop_write: 1999b9ca0664SLiu Bo mnt_drop_write_file(file); 2000b9ca0664SLiu Bo out: 20010caa102dSLi Zefan return ret; 20020caa102dSLi Zefan } 20030caa102dSLi Zefan 2004ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2005ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2006ac8e9819SChris Mason { 2007abc6e134SChris Mason struct btrfs_key test; 2008abc6e134SChris Mason int ret; 2009abc6e134SChris Mason 2010abc6e134SChris Mason test.objectid = sk->min_objectid; 2011abc6e134SChris Mason test.type = sk->min_type; 2012abc6e134SChris Mason test.offset = sk->min_offset; 2013abc6e134SChris Mason 2014abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2015abc6e134SChris Mason if (ret < 0) 2016ac8e9819SChris Mason return 0; 2017abc6e134SChris Mason 2018abc6e134SChris Mason test.objectid = sk->max_objectid; 2019abc6e134SChris Mason test.type = sk->max_type; 2020abc6e134SChris Mason test.offset = sk->max_offset; 2021abc6e134SChris Mason 2022abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2023abc6e134SChris Mason if (ret > 0) 2024ac8e9819SChris Mason return 0; 2025ac8e9819SChris Mason return 1; 2026ac8e9819SChris Mason } 2027ac8e9819SChris Mason 2028df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2029ac8e9819SChris Mason struct btrfs_key *key, 2030ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20319b6e817dSGerhard Heift size_t *buf_size, 2032ba346b35SGerhard Heift char __user *ubuf, 2033ac8e9819SChris Mason unsigned long *sk_offset, 2034ac8e9819SChris Mason int *num_found) 2035ac8e9819SChris Mason { 2036ac8e9819SChris Mason u64 found_transid; 2037ac8e9819SChris Mason struct extent_buffer *leaf; 2038ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2039dd81d459SNaohiro Aota struct btrfs_key test; 2040ac8e9819SChris Mason unsigned long item_off; 2041ac8e9819SChris Mason unsigned long item_len; 2042ac8e9819SChris Mason int nritems; 2043ac8e9819SChris Mason int i; 2044ac8e9819SChris Mason int slot; 2045ac8e9819SChris Mason int ret = 0; 2046ac8e9819SChris Mason 2047ac8e9819SChris Mason leaf = path->nodes[0]; 2048ac8e9819SChris Mason slot = path->slots[0]; 2049ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2050ac8e9819SChris Mason 2051ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2052ac8e9819SChris Mason i = nritems; 2053ac8e9819SChris Mason goto advance_key; 2054ac8e9819SChris Mason } 2055ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2056ac8e9819SChris Mason 2057ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2058ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2059ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2060ac8e9819SChris Mason 206103b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 206203b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 206303b71c6cSGabriel de Perthuis continue; 206403b71c6cSGabriel de Perthuis 20659b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20668f5f6178SGerhard Heift if (*num_found) { 2067ac8e9819SChris Mason ret = 1; 20688f5f6178SGerhard Heift goto out; 20698f5f6178SGerhard Heift } 20708f5f6178SGerhard Heift 20718f5f6178SGerhard Heift /* 20728f5f6178SGerhard Heift * return one empty item back for v1, which does not 20738f5f6178SGerhard Heift * handle -EOVERFLOW 20748f5f6178SGerhard Heift */ 20758f5f6178SGerhard Heift 20769b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2077ac8e9819SChris Mason item_len = 0; 20788f5f6178SGerhard Heift ret = -EOVERFLOW; 20798f5f6178SGerhard Heift } 2080ac8e9819SChris Mason 20819b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2082ac8e9819SChris Mason ret = 1; 208325c9bc2eSGerhard Heift goto out; 2084ac8e9819SChris Mason } 2085ac8e9819SChris Mason 2086ac8e9819SChris Mason sh.objectid = key->objectid; 2087ac8e9819SChris Mason sh.offset = key->offset; 2088ac8e9819SChris Mason sh.type = key->type; 2089ac8e9819SChris Mason sh.len = item_len; 2090ac8e9819SChris Mason sh.transid = found_transid; 2091ac8e9819SChris Mason 2092ac8e9819SChris Mason /* copy search result header */ 2093ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2094ba346b35SGerhard Heift ret = -EFAULT; 2095ba346b35SGerhard Heift goto out; 2096ba346b35SGerhard Heift } 2097ba346b35SGerhard Heift 2098ac8e9819SChris Mason *sk_offset += sizeof(sh); 2099ac8e9819SChris Mason 2100ac8e9819SChris Mason if (item_len) { 2101ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2102ac8e9819SChris Mason /* copy the item */ 2103ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2104ba346b35SGerhard Heift item_off, item_len)) { 2105ba346b35SGerhard Heift ret = -EFAULT; 2106ba346b35SGerhard Heift goto out; 2107ba346b35SGerhard Heift } 2108ba346b35SGerhard Heift 2109ac8e9819SChris Mason *sk_offset += item_len; 2110ac8e9819SChris Mason } 2111e2156867SHugo Mills (*num_found)++; 2112ac8e9819SChris Mason 21138f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21148f5f6178SGerhard Heift goto out; 21158f5f6178SGerhard Heift 211625c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 211725c9bc2eSGerhard Heift ret = 1; 211825c9bc2eSGerhard Heift goto out; 211925c9bc2eSGerhard Heift } 2120ac8e9819SChris Mason } 2121ac8e9819SChris Mason advance_key: 2122ac8e9819SChris Mason ret = 0; 2123dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2124dd81d459SNaohiro Aota test.type = sk->max_type; 2125dd81d459SNaohiro Aota test.offset = sk->max_offset; 2126dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2127dd81d459SNaohiro Aota ret = 1; 2128dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2129abc6e134SChris Mason key->offset++; 2130dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2131abc6e134SChris Mason key->offset = 0; 2132abc6e134SChris Mason key->type++; 2133dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2134abc6e134SChris Mason key->offset = 0; 2135abc6e134SChris Mason key->type = 0; 2136abc6e134SChris Mason key->objectid++; 2137abc6e134SChris Mason } else 2138abc6e134SChris Mason ret = 1; 213925c9bc2eSGerhard Heift out: 2140ba346b35SGerhard Heift /* 2141ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2142ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2143ba346b35SGerhard Heift * * all items were found 2144ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2145ba346b35SGerhard Heift * leaf 2146ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2147ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2148ba346b35SGerhard Heift */ 2149ac8e9819SChris Mason return ret; 2150ac8e9819SChris Mason } 2151ac8e9819SChris Mason 2152ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 215312544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21549b6e817dSGerhard Heift size_t *buf_size, 2155ba346b35SGerhard Heift char __user *ubuf) 2156ac8e9819SChris Mason { 21570b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2158ac8e9819SChris Mason struct btrfs_root *root; 2159ac8e9819SChris Mason struct btrfs_key key; 2160ac8e9819SChris Mason struct btrfs_path *path; 2161ac8e9819SChris Mason int ret; 2162ac8e9819SChris Mason int num_found = 0; 2163ac8e9819SChris Mason unsigned long sk_offset = 0; 2164ac8e9819SChris Mason 21659b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21669b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 216712544442SGerhard Heift return -EOVERFLOW; 21689b6e817dSGerhard Heift } 216912544442SGerhard Heift 2170ac8e9819SChris Mason path = btrfs_alloc_path(); 2171ac8e9819SChris Mason if (!path) 2172ac8e9819SChris Mason return -ENOMEM; 2173ac8e9819SChris Mason 2174ac8e9819SChris Mason if (sk->tree_id == 0) { 2175ac8e9819SChris Mason /* search the root of the inode that was passed */ 217600246528SJosef Bacik root = btrfs_grab_root(BTRFS_I(inode)->root); 2177ac8e9819SChris Mason } else { 2178ac8e9819SChris Mason key.objectid = sk->tree_id; 2179ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2180ac8e9819SChris Mason key.offset = (u64)-1; 21813619c94fSJosef Bacik root = btrfs_get_fs_root(info, &key, true); 2182ac8e9819SChris Mason if (IS_ERR(root)) { 2183ac8e9819SChris Mason btrfs_free_path(path); 2184ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2185ac8e9819SChris Mason } 2186ac8e9819SChris Mason } 2187ac8e9819SChris Mason 2188ac8e9819SChris Mason key.objectid = sk->min_objectid; 2189ac8e9819SChris Mason key.type = sk->min_type; 2190ac8e9819SChris Mason key.offset = sk->min_offset; 2191ac8e9819SChris Mason 2192ac8e9819SChris Mason while (1) { 21936174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2194ac8e9819SChris Mason if (ret != 0) { 2195ac8e9819SChris Mason if (ret > 0) 2196ac8e9819SChris Mason ret = 0; 2197ac8e9819SChris Mason goto err; 2198ac8e9819SChris Mason } 2199df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2200ac8e9819SChris Mason &sk_offset, &num_found); 2201b3b4aa74SDavid Sterba btrfs_release_path(path); 220225c9bc2eSGerhard Heift if (ret) 2203ac8e9819SChris Mason break; 2204ac8e9819SChris Mason 2205ac8e9819SChris Mason } 22068f5f6178SGerhard Heift if (ret > 0) 2207ac8e9819SChris Mason ret = 0; 2208ac8e9819SChris Mason err: 2209ac8e9819SChris Mason sk->nr_items = num_found; 221000246528SJosef Bacik btrfs_put_root(root); 2211ac8e9819SChris Mason btrfs_free_path(path); 2212ac8e9819SChris Mason return ret; 2213ac8e9819SChris Mason } 2214ac8e9819SChris Mason 2215ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2216ac8e9819SChris Mason void __user *argp) 2217ac8e9819SChris Mason { 2218ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2219ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2220ac8e9819SChris Mason struct inode *inode; 2221ac8e9819SChris Mason int ret; 22229b6e817dSGerhard Heift size_t buf_size; 2223ac8e9819SChris Mason 2224ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2225ac8e9819SChris Mason return -EPERM; 2226ac8e9819SChris Mason 2227ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2228ac8e9819SChris Mason 2229ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2230ba346b35SGerhard Heift return -EFAULT; 2231ba346b35SGerhard Heift 2232ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2233ac8e9819SChris Mason 2234496ad9aaSAl Viro inode = file_inode(file); 2235ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22368f5f6178SGerhard Heift 22378f5f6178SGerhard Heift /* 22388f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22398f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22408f5f6178SGerhard Heift */ 22418f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22428f5f6178SGerhard Heift ret = 0; 22438f5f6178SGerhard Heift 2244ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2245ac8e9819SChris Mason ret = -EFAULT; 2246ac8e9819SChris Mason return ret; 2247ac8e9819SChris Mason } 2248ac8e9819SChris Mason 2249cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2250cc68a8a5SGerhard Heift void __user *argp) 2251cc68a8a5SGerhard Heift { 2252cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2253cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2254cc68a8a5SGerhard Heift struct inode *inode; 2255cc68a8a5SGerhard Heift int ret; 2256cc68a8a5SGerhard Heift size_t buf_size; 2257ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2258cc68a8a5SGerhard Heift 2259cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2260cc68a8a5SGerhard Heift return -EPERM; 2261cc68a8a5SGerhard Heift 2262cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2263cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2264cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2265cc68a8a5SGerhard Heift return -EFAULT; 2266cc68a8a5SGerhard Heift 2267cc68a8a5SGerhard Heift buf_size = args.buf_size; 2268cc68a8a5SGerhard Heift 2269cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2270cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2271cc68a8a5SGerhard Heift buf_size = buf_limit; 2272cc68a8a5SGerhard Heift 2273cc68a8a5SGerhard Heift inode = file_inode(file); 2274cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2275718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2276cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2277cc68a8a5SGerhard Heift ret = -EFAULT; 2278cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2279cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2280cc68a8a5SGerhard Heift ret = -EFAULT; 2281cc68a8a5SGerhard Heift 228276dda93cSYan, Zheng return ret; 228376dda93cSYan, Zheng } 228476dda93cSYan, Zheng 228598d377a0STARUISI Hiroaki /* 2286ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2287ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 228898d377a0STARUISI Hiroaki */ 228998d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 229098d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 229198d377a0STARUISI Hiroaki { 229298d377a0STARUISI Hiroaki struct btrfs_root *root; 229398d377a0STARUISI Hiroaki struct btrfs_key key; 2294ac8e9819SChris Mason char *ptr; 229598d377a0STARUISI Hiroaki int ret = -1; 229698d377a0STARUISI Hiroaki int slot; 229798d377a0STARUISI Hiroaki int len; 229898d377a0STARUISI Hiroaki int total_len = 0; 229998d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 230098d377a0STARUISI Hiroaki struct extent_buffer *l; 230198d377a0STARUISI Hiroaki struct btrfs_path *path; 230298d377a0STARUISI Hiroaki 230398d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 230498d377a0STARUISI Hiroaki name[0]='\0'; 230598d377a0STARUISI Hiroaki return 0; 230698d377a0STARUISI Hiroaki } 230798d377a0STARUISI Hiroaki 230898d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 230998d377a0STARUISI Hiroaki if (!path) 231098d377a0STARUISI Hiroaki return -ENOMEM; 231198d377a0STARUISI Hiroaki 2312c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 231398d377a0STARUISI Hiroaki 231498d377a0STARUISI Hiroaki key.objectid = tree_id; 231598d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 231698d377a0STARUISI Hiroaki key.offset = (u64)-1; 23173619c94fSJosef Bacik root = btrfs_get_fs_root(info, &key, true); 231898d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2319ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 232088234012SJosef Bacik root = NULL; 232188234012SJosef Bacik goto out; 232288234012SJosef Bacik } 232398d377a0STARUISI Hiroaki 232498d377a0STARUISI Hiroaki key.objectid = dirid; 232598d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23268ad6fcabSChris Mason key.offset = (u64)-1; 232798d377a0STARUISI Hiroaki 232898d377a0STARUISI Hiroaki while (1) { 232998d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 233098d377a0STARUISI Hiroaki if (ret < 0) 233198d377a0STARUISI Hiroaki goto out; 233218674c6cSFilipe David Borba Manana else if (ret > 0) { 233318674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 233418674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 233518674c6cSFilipe David Borba Manana if (ret < 0) 233618674c6cSFilipe David Borba Manana goto out; 233718674c6cSFilipe David Borba Manana else if (ret > 0) { 2338ac8e9819SChris Mason ret = -ENOENT; 233998d377a0STARUISI Hiroaki goto out; 2340ac8e9819SChris Mason } 234118674c6cSFilipe David Borba Manana } 234218674c6cSFilipe David Borba Manana 234318674c6cSFilipe David Borba Manana l = path->nodes[0]; 234418674c6cSFilipe David Borba Manana slot = path->slots[0]; 234518674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 234698d377a0STARUISI Hiroaki 234798d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 234898d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 234998d377a0STARUISI Hiroaki ptr -= len + 1; 235098d377a0STARUISI Hiroaki total_len += len + 1; 2351a696cf35SFilipe David Borba Manana if (ptr < name) { 2352a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 235398d377a0STARUISI Hiroaki goto out; 2354a696cf35SFilipe David Borba Manana } 235598d377a0STARUISI Hiroaki 235698d377a0STARUISI Hiroaki *(ptr + len) = '/'; 235798d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 235898d377a0STARUISI Hiroaki 235998d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 236098d377a0STARUISI Hiroaki break; 236198d377a0STARUISI Hiroaki 2362b3b4aa74SDavid Sterba btrfs_release_path(path); 236398d377a0STARUISI Hiroaki key.objectid = key.offset; 23648ad6fcabSChris Mason key.offset = (u64)-1; 236598d377a0STARUISI Hiroaki dirid = key.objectid; 236698d377a0STARUISI Hiroaki } 236777906a50SLi Zefan memmove(name, ptr, total_len); 236898d377a0STARUISI Hiroaki name[total_len] = '\0'; 236998d377a0STARUISI Hiroaki ret = 0; 237098d377a0STARUISI Hiroaki out: 237100246528SJosef Bacik btrfs_put_root(root); 237298d377a0STARUISI Hiroaki btrfs_free_path(path); 2373ac8e9819SChris Mason return ret; 2374ac8e9819SChris Mason } 2375ac8e9819SChris Mason 237623d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 237723d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 237823d0b79dSTomohiro Misono { 237923d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 238023d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 238123d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 238223d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 238323d0b79dSTomohiro Misono u64 dirid = args->dirid; 238423d0b79dSTomohiro Misono unsigned long item_off; 238523d0b79dSTomohiro Misono unsigned long item_len; 238623d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 238723d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 2388b8a49ae1SJosef Bacik struct btrfs_root *root = NULL; 238923d0b79dSTomohiro Misono struct btrfs_path *path; 239023d0b79dSTomohiro Misono struct btrfs_key key, key2; 239123d0b79dSTomohiro Misono struct extent_buffer *leaf; 239223d0b79dSTomohiro Misono struct inode *temp_inode; 239323d0b79dSTomohiro Misono char *ptr; 239423d0b79dSTomohiro Misono int slot; 239523d0b79dSTomohiro Misono int len; 239623d0b79dSTomohiro Misono int total_len = 0; 239723d0b79dSTomohiro Misono int ret; 239823d0b79dSTomohiro Misono 239923d0b79dSTomohiro Misono path = btrfs_alloc_path(); 240023d0b79dSTomohiro Misono if (!path) 240123d0b79dSTomohiro Misono return -ENOMEM; 240223d0b79dSTomohiro Misono 240323d0b79dSTomohiro Misono /* 240423d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 240523d0b79dSTomohiro Misono * construct the path in from the bottom up. 240623d0b79dSTomohiro Misono */ 240723d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 240823d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 240923d0b79dSTomohiro Misono 241023d0b79dSTomohiro Misono key.objectid = treeid; 241123d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 241223d0b79dSTomohiro Misono key.offset = (u64)-1; 24133619c94fSJosef Bacik root = btrfs_get_fs_root(fs_info, &key, true); 241423d0b79dSTomohiro Misono if (IS_ERR(root)) { 241523d0b79dSTomohiro Misono ret = PTR_ERR(root); 241623d0b79dSTomohiro Misono goto out; 241723d0b79dSTomohiro Misono } 241823d0b79dSTomohiro Misono 241923d0b79dSTomohiro Misono key.objectid = dirid; 242023d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 242123d0b79dSTomohiro Misono key.offset = (u64)-1; 242223d0b79dSTomohiro Misono while (1) { 242323d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 242423d0b79dSTomohiro Misono if (ret < 0) { 2425b8a49ae1SJosef Bacik goto out_put; 242623d0b79dSTomohiro Misono } else if (ret > 0) { 242723d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 242823d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 242923d0b79dSTomohiro Misono if (ret < 0) { 2430b8a49ae1SJosef Bacik goto out_put; 243123d0b79dSTomohiro Misono } else if (ret > 0) { 243223d0b79dSTomohiro Misono ret = -ENOENT; 2433b8a49ae1SJosef Bacik goto out_put; 243423d0b79dSTomohiro Misono } 243523d0b79dSTomohiro Misono } 243623d0b79dSTomohiro Misono 243723d0b79dSTomohiro Misono leaf = path->nodes[0]; 243823d0b79dSTomohiro Misono slot = path->slots[0]; 243923d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 244023d0b79dSTomohiro Misono 244123d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 244223d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 244323d0b79dSTomohiro Misono ptr -= len + 1; 244423d0b79dSTomohiro Misono total_len += len + 1; 244523d0b79dSTomohiro Misono if (ptr < args->path) { 244623d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 2447b8a49ae1SJosef Bacik goto out_put; 244823d0b79dSTomohiro Misono } 244923d0b79dSTomohiro Misono 245023d0b79dSTomohiro Misono *(ptr + len) = '/'; 245123d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 245223d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 245323d0b79dSTomohiro Misono 245423d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 245523d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 245623d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 245723d0b79dSTomohiro Misono if (ret < 0) { 2458b8a49ae1SJosef Bacik goto out_put; 245923d0b79dSTomohiro Misono } else if (ret > 0) { 246023d0b79dSTomohiro Misono ret = -ENOENT; 2461b8a49ae1SJosef Bacik goto out_put; 246223d0b79dSTomohiro Misono } 246323d0b79dSTomohiro Misono 246423d0b79dSTomohiro Misono leaf = path->nodes[0]; 246523d0b79dSTomohiro Misono slot = path->slots[0]; 246623d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 246723d0b79dSTomohiro Misono if (key2.objectid != dirid) { 246823d0b79dSTomohiro Misono ret = -ENOENT; 2469b8a49ae1SJosef Bacik goto out_put; 247023d0b79dSTomohiro Misono } 247123d0b79dSTomohiro Misono 24724c66e0d4SDavid Sterba temp_inode = btrfs_iget(sb, &key2, root); 24733ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24743ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 2475b8a49ae1SJosef Bacik goto out_put; 24763ca57bd6SMisono Tomohiro } 247723d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 247823d0b79dSTomohiro Misono iput(temp_inode); 247923d0b79dSTomohiro Misono if (ret) { 248023d0b79dSTomohiro Misono ret = -EACCES; 2481b8a49ae1SJosef Bacik goto out_put; 248223d0b79dSTomohiro Misono } 248323d0b79dSTomohiro Misono 248423d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 248523d0b79dSTomohiro Misono break; 248623d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 248723d0b79dSTomohiro Misono ret = -EACCES; 2488b8a49ae1SJosef Bacik goto out_put; 248923d0b79dSTomohiro Misono } 249023d0b79dSTomohiro Misono 249123d0b79dSTomohiro Misono btrfs_release_path(path); 249223d0b79dSTomohiro Misono key.objectid = key.offset; 249323d0b79dSTomohiro Misono key.offset = (u64)-1; 249423d0b79dSTomohiro Misono dirid = key.objectid; 249523d0b79dSTomohiro Misono } 249623d0b79dSTomohiro Misono 249723d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 249823d0b79dSTomohiro Misono args->path[total_len] = '\0'; 249900246528SJosef Bacik btrfs_put_root(root); 2500b8a49ae1SJosef Bacik root = NULL; 250123d0b79dSTomohiro Misono btrfs_release_path(path); 250223d0b79dSTomohiro Misono } 250323d0b79dSTomohiro Misono 250423d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 250523d0b79dSTomohiro Misono key.objectid = treeid; 250623d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 250723d0b79dSTomohiro Misono key.offset = args->treeid; 2508b8a49ae1SJosef Bacik ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 250923d0b79dSTomohiro Misono if (ret < 0) { 251023d0b79dSTomohiro Misono goto out; 251123d0b79dSTomohiro Misono } else if (ret > 0) { 251223d0b79dSTomohiro Misono ret = -ENOENT; 251323d0b79dSTomohiro Misono goto out; 251423d0b79dSTomohiro Misono } 251523d0b79dSTomohiro Misono 251623d0b79dSTomohiro Misono leaf = path->nodes[0]; 251723d0b79dSTomohiro Misono slot = path->slots[0]; 251823d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 251923d0b79dSTomohiro Misono 252023d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 252123d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 252223d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 252323d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 252423d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 252523d0b79dSTomohiro Misono ret = -EINVAL; 252623d0b79dSTomohiro Misono goto out; 252723d0b79dSTomohiro Misono } 252823d0b79dSTomohiro Misono 252923d0b79dSTomohiro Misono /* Copy subvolume's name */ 253023d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 253123d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 253223d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 253323d0b79dSTomohiro Misono args->name[item_len] = 0; 253423d0b79dSTomohiro Misono 2535b8a49ae1SJosef Bacik out_put: 253600246528SJosef Bacik btrfs_put_root(root); 253723d0b79dSTomohiro Misono out: 253823d0b79dSTomohiro Misono btrfs_free_path(path); 253923d0b79dSTomohiro Misono return ret; 254023d0b79dSTomohiro Misono } 254123d0b79dSTomohiro Misono 2542ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2543ac8e9819SChris Mason void __user *argp) 2544ac8e9819SChris Mason { 2545ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2546ac8e9819SChris Mason struct inode *inode; 254701b810b8SDavid Sterba int ret = 0; 2548ac8e9819SChris Mason 25492354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25502354d08fSJulia Lawall if (IS_ERR(args)) 25512354d08fSJulia Lawall return PTR_ERR(args); 2552c2b96929SDan Carpenter 2553496ad9aaSAl Viro inode = file_inode(file); 2554ac8e9819SChris Mason 255501b810b8SDavid Sterba /* 255601b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 255701b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 255801b810b8SDavid Sterba */ 25591b53ac4dSChris Mason if (args->treeid == 0) 25601b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25611b53ac4dSChris Mason 256201b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 256301b810b8SDavid Sterba args->name[0] = 0; 256401b810b8SDavid Sterba goto out; 256501b810b8SDavid Sterba } 256601b810b8SDavid Sterba 256701b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 256801b810b8SDavid Sterba ret = -EPERM; 256901b810b8SDavid Sterba goto out; 257001b810b8SDavid Sterba } 257101b810b8SDavid Sterba 2572ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2573ac8e9819SChris Mason args->treeid, args->objectid, 2574ac8e9819SChris Mason args->name); 2575ac8e9819SChris Mason 257601b810b8SDavid Sterba out: 2577ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2578ac8e9819SChris Mason ret = -EFAULT; 2579ac8e9819SChris Mason 2580ac8e9819SChris Mason kfree(args); 258198d377a0STARUISI Hiroaki return ret; 258298d377a0STARUISI Hiroaki } 258398d377a0STARUISI Hiroaki 258423d0b79dSTomohiro Misono /* 258523d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 258623d0b79dSTomohiro Misono * 258723d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 258823d0b79dSTomohiro Misono * 258923d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 259023d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 259123d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 259223d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 259323d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 259423d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 259523d0b79dSTomohiro Misono */ 259623d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 259723d0b79dSTomohiro Misono { 259823d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 259923d0b79dSTomohiro Misono struct inode *inode; 260023d0b79dSTomohiro Misono int ret; 260123d0b79dSTomohiro Misono 260223d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 260323d0b79dSTomohiro Misono if (IS_ERR(args)) 260423d0b79dSTomohiro Misono return PTR_ERR(args); 260523d0b79dSTomohiro Misono 260623d0b79dSTomohiro Misono inode = file_inode(file); 260723d0b79dSTomohiro Misono 260823d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 260923d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 261023d0b79dSTomohiro Misono /* 261123d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 261223d0b79dSTomohiro Misono * called 261323d0b79dSTomohiro Misono */ 261423d0b79dSTomohiro Misono kfree(args); 261523d0b79dSTomohiro Misono return -EACCES; 261623d0b79dSTomohiro Misono } 261723d0b79dSTomohiro Misono 261823d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 261923d0b79dSTomohiro Misono 262023d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 262123d0b79dSTomohiro Misono ret = -EFAULT; 262223d0b79dSTomohiro Misono 262323d0b79dSTomohiro Misono kfree(args); 262423d0b79dSTomohiro Misono return ret; 262523d0b79dSTomohiro Misono } 262623d0b79dSTomohiro Misono 2627b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2628b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2629b64ec075STomohiro Misono { 2630b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2631b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2632b64ec075STomohiro Misono struct btrfs_root *root; 2633b64ec075STomohiro Misono struct btrfs_path *path; 2634b64ec075STomohiro Misono struct btrfs_key key; 2635b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2636b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2637b64ec075STomohiro Misono struct extent_buffer *leaf; 2638b64ec075STomohiro Misono unsigned long item_off; 2639b64ec075STomohiro Misono unsigned long item_len; 2640b64ec075STomohiro Misono struct inode *inode; 2641b64ec075STomohiro Misono int slot; 2642b64ec075STomohiro Misono int ret = 0; 2643b64ec075STomohiro Misono 2644b64ec075STomohiro Misono path = btrfs_alloc_path(); 2645b64ec075STomohiro Misono if (!path) 2646b64ec075STomohiro Misono return -ENOMEM; 2647b64ec075STomohiro Misono 2648b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2649b64ec075STomohiro Misono if (!subvol_info) { 2650b64ec075STomohiro Misono btrfs_free_path(path); 2651b64ec075STomohiro Misono return -ENOMEM; 2652b64ec075STomohiro Misono } 2653b64ec075STomohiro Misono 2654b64ec075STomohiro Misono inode = file_inode(file); 2655b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2656b64ec075STomohiro Misono 2657b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2658b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2659b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2660b64ec075STomohiro Misono key.offset = (u64)-1; 26613619c94fSJosef Bacik root = btrfs_get_fs_root(fs_info, &key, true); 2662b64ec075STomohiro Misono if (IS_ERR(root)) { 2663b64ec075STomohiro Misono ret = PTR_ERR(root); 266404734e84SJosef Bacik goto out_free; 266504734e84SJosef Bacik } 2666b64ec075STomohiro Misono root_item = &root->root_item; 2667b64ec075STomohiro Misono 2668b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2669b64ec075STomohiro Misono 2670b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2671b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2672b64ec075STomohiro Misono 2673b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2674b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2675b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2676b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2677b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2678b64ec075STomohiro Misono 2679b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2680b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2681b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2682b64ec075STomohiro Misono 2683b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2684b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2685b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2686b64ec075STomohiro Misono 2687b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2688b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2689b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2690b64ec075STomohiro Misono 2691b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2692b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2693b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2694b64ec075STomohiro Misono 2695b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2696b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2697b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2698b64ec075STomohiro Misono key.offset = 0; 269904734e84SJosef Bacik ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 2700b64ec075STomohiro Misono if (ret < 0) { 2701b64ec075STomohiro Misono goto out; 2702b64ec075STomohiro Misono } else if (path->slots[0] >= 2703b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 270404734e84SJosef Bacik ret = btrfs_next_leaf(fs_info->tree_root, path); 2705b64ec075STomohiro Misono if (ret < 0) { 2706b64ec075STomohiro Misono goto out; 2707b64ec075STomohiro Misono } else if (ret > 0) { 2708b64ec075STomohiro Misono ret = -EUCLEAN; 2709b64ec075STomohiro Misono goto out; 2710b64ec075STomohiro Misono } 2711b64ec075STomohiro Misono } 2712b64ec075STomohiro Misono 2713b64ec075STomohiro Misono leaf = path->nodes[0]; 2714b64ec075STomohiro Misono slot = path->slots[0]; 2715b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2716b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2717b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2718b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2719b64ec075STomohiro Misono 2720b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2721b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2722b64ec075STomohiro Misono 2723b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2724b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2725b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2726b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2727b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2728b64ec075STomohiro Misono item_off, item_len); 2729b64ec075STomohiro Misono } else { 2730b64ec075STomohiro Misono ret = -ENOENT; 2731b64ec075STomohiro Misono goto out; 2732b64ec075STomohiro Misono } 2733b64ec075STomohiro Misono } 2734b64ec075STomohiro Misono 2735b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2736b64ec075STomohiro Misono ret = -EFAULT; 2737b64ec075STomohiro Misono 2738b64ec075STomohiro Misono out: 273900246528SJosef Bacik btrfs_put_root(root); 274004734e84SJosef Bacik out_free: 2741b64ec075STomohiro Misono btrfs_free_path(path); 2742b64ec075STomohiro Misono kzfree(subvol_info); 2743b64ec075STomohiro Misono return ret; 2744b64ec075STomohiro Misono } 2745b64ec075STomohiro Misono 274642e4b520STomohiro Misono /* 274742e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 274842e4b520STomohiro Misono * except the subvolume name. 274942e4b520STomohiro Misono */ 275042e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 275142e4b520STomohiro Misono { 275242e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 275342e4b520STomohiro Misono struct btrfs_root_ref *rref; 275442e4b520STomohiro Misono struct btrfs_root *root; 275542e4b520STomohiro Misono struct btrfs_path *path; 275642e4b520STomohiro Misono struct btrfs_key key; 275742e4b520STomohiro Misono struct extent_buffer *leaf; 275842e4b520STomohiro Misono struct inode *inode; 275942e4b520STomohiro Misono u64 objectid; 276042e4b520STomohiro Misono int slot; 276142e4b520STomohiro Misono int ret; 276242e4b520STomohiro Misono u8 found; 276342e4b520STomohiro Misono 276442e4b520STomohiro Misono path = btrfs_alloc_path(); 276542e4b520STomohiro Misono if (!path) 276642e4b520STomohiro Misono return -ENOMEM; 276742e4b520STomohiro Misono 276842e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 276942e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 277042e4b520STomohiro Misono btrfs_free_path(path); 277142e4b520STomohiro Misono return PTR_ERR(rootrefs); 277242e4b520STomohiro Misono } 277342e4b520STomohiro Misono 277442e4b520STomohiro Misono inode = file_inode(file); 277542e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 277642e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 277742e4b520STomohiro Misono 277842e4b520STomohiro Misono key.objectid = objectid; 277942e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 278042e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 278142e4b520STomohiro Misono found = 0; 278242e4b520STomohiro Misono 278342e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 278442e4b520STomohiro Misono if (ret < 0) { 278542e4b520STomohiro Misono goto out; 278642e4b520STomohiro Misono } else if (path->slots[0] >= 278742e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 278842e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 278942e4b520STomohiro Misono if (ret < 0) { 279042e4b520STomohiro Misono goto out; 279142e4b520STomohiro Misono } else if (ret > 0) { 279242e4b520STomohiro Misono ret = -EUCLEAN; 279342e4b520STomohiro Misono goto out; 279442e4b520STomohiro Misono } 279542e4b520STomohiro Misono } 279642e4b520STomohiro Misono while (1) { 279742e4b520STomohiro Misono leaf = path->nodes[0]; 279842e4b520STomohiro Misono slot = path->slots[0]; 279942e4b520STomohiro Misono 280042e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 280142e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 280242e4b520STomohiro Misono ret = 0; 280342e4b520STomohiro Misono goto out; 280442e4b520STomohiro Misono } 280542e4b520STomohiro Misono 280642e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 280742e4b520STomohiro Misono ret = -EOVERFLOW; 280842e4b520STomohiro Misono goto out; 280942e4b520STomohiro Misono } 281042e4b520STomohiro Misono 281142e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 281242e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 281342e4b520STomohiro Misono rootrefs->rootref[found].dirid = 281442e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 281542e4b520STomohiro Misono found++; 281642e4b520STomohiro Misono 281742e4b520STomohiro Misono ret = btrfs_next_item(root, path); 281842e4b520STomohiro Misono if (ret < 0) { 281942e4b520STomohiro Misono goto out; 282042e4b520STomohiro Misono } else if (ret > 0) { 282142e4b520STomohiro Misono ret = -EUCLEAN; 282242e4b520STomohiro Misono goto out; 282342e4b520STomohiro Misono } 282442e4b520STomohiro Misono } 282542e4b520STomohiro Misono 282642e4b520STomohiro Misono out: 282742e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 282842e4b520STomohiro Misono rootrefs->num_items = found; 282942e4b520STomohiro Misono /* update min_treeid for next search */ 283042e4b520STomohiro Misono if (found) 283142e4b520STomohiro Misono rootrefs->min_treeid = 283242e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 283342e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 283442e4b520STomohiro Misono ret = -EFAULT; 283542e4b520STomohiro Misono } 283642e4b520STomohiro Misono 283742e4b520STomohiro Misono kfree(rootrefs); 283842e4b520STomohiro Misono btrfs_free_path(path); 283942e4b520STomohiro Misono 284042e4b520STomohiro Misono return ret; 284142e4b520STomohiro Misono } 284242e4b520STomohiro Misono 284376dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 2844949964c9SMarcos Paulo de Souza void __user *arg, 2845949964c9SMarcos Paulo de Souza bool destroy_v2) 284676dda93cSYan, Zheng { 284754563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28480b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 284976dda93cSYan, Zheng struct dentry *dentry; 28502b0143b5SDavid Howells struct inode *dir = d_inode(parent); 285176dda93cSYan, Zheng struct inode *inode; 285276dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 285376dda93cSYan, Zheng struct btrfs_root *dest = NULL; 2854949964c9SMarcos Paulo de Souza struct btrfs_ioctl_vol_args *vol_args = NULL; 2855949964c9SMarcos Paulo de Souza struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL; 2856949964c9SMarcos Paulo de Souza char *subvol_name, *subvol_name_ptr = NULL; 2857949964c9SMarcos Paulo de Souza int subvol_namelen; 285876dda93cSYan, Zheng int err = 0; 2859949964c9SMarcos Paulo de Souza bool destroy_parent = false; 286076dda93cSYan, Zheng 2861949964c9SMarcos Paulo de Souza if (destroy_v2) { 2862949964c9SMarcos Paulo de Souza vol_args2 = memdup_user(arg, sizeof(*vol_args2)); 2863949964c9SMarcos Paulo de Souza if (IS_ERR(vol_args2)) 2864949964c9SMarcos Paulo de Souza return PTR_ERR(vol_args2); 2865325c50e3SJeff Mahoney 2866949964c9SMarcos Paulo de Souza if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) { 2867949964c9SMarcos Paulo de Souza err = -EOPNOTSUPP; 2868949964c9SMarcos Paulo de Souza goto out; 2869949964c9SMarcos Paulo de Souza } 287076dda93cSYan, Zheng 2871949964c9SMarcos Paulo de Souza /* 2872949964c9SMarcos Paulo de Souza * If SPEC_BY_ID is not set, we are looking for the subvolume by 2873949964c9SMarcos Paulo de Souza * name, same as v1 currently does. 2874949964c9SMarcos Paulo de Souza */ 2875949964c9SMarcos Paulo de Souza if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { 2876949964c9SMarcos Paulo de Souza vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0; 2877949964c9SMarcos Paulo de Souza subvol_name = vol_args2->name; 2878949964c9SMarcos Paulo de Souza 2879949964c9SMarcos Paulo de Souza err = mnt_want_write_file(file); 2880949964c9SMarcos Paulo de Souza if (err) 2881949964c9SMarcos Paulo de Souza goto out; 2882949964c9SMarcos Paulo de Souza } else { 2883949964c9SMarcos Paulo de Souza if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) { 288476dda93cSYan, Zheng err = -EINVAL; 288576dda93cSYan, Zheng goto out; 288676dda93cSYan, Zheng } 288776dda93cSYan, Zheng 2888a561be71SAl Viro err = mnt_want_write_file(file); 288976dda93cSYan, Zheng if (err) 289076dda93cSYan, Zheng goto out; 289176dda93cSYan, Zheng 2892949964c9SMarcos Paulo de Souza dentry = btrfs_get_dentry(fs_info->sb, 2893949964c9SMarcos Paulo de Souza BTRFS_FIRST_FREE_OBJECTID, 2894949964c9SMarcos Paulo de Souza vol_args2->subvolid, 0, 0); 2895949964c9SMarcos Paulo de Souza if (IS_ERR(dentry)) { 2896949964c9SMarcos Paulo de Souza err = PTR_ERR(dentry); 2897949964c9SMarcos Paulo de Souza goto out_drop_write; 2898949964c9SMarcos Paulo de Souza } 2899949964c9SMarcos Paulo de Souza 2900949964c9SMarcos Paulo de Souza /* 2901949964c9SMarcos Paulo de Souza * Change the default parent since the subvolume being 2902949964c9SMarcos Paulo de Souza * deleted can be outside of the current mount point. 2903949964c9SMarcos Paulo de Souza */ 2904949964c9SMarcos Paulo de Souza parent = btrfs_get_parent(dentry); 2905949964c9SMarcos Paulo de Souza 2906949964c9SMarcos Paulo de Souza /* 2907949964c9SMarcos Paulo de Souza * At this point dentry->d_name can point to '/' if the 2908949964c9SMarcos Paulo de Souza * subvolume we want to destroy is outsite of the 2909949964c9SMarcos Paulo de Souza * current mount point, so we need to release the 2910949964c9SMarcos Paulo de Souza * current dentry and execute the lookup to return a new 2911949964c9SMarcos Paulo de Souza * one with ->d_name pointing to the 2912949964c9SMarcos Paulo de Souza * <mount point>/subvol_name. 2913949964c9SMarcos Paulo de Souza */ 2914949964c9SMarcos Paulo de Souza dput(dentry); 2915949964c9SMarcos Paulo de Souza if (IS_ERR(parent)) { 2916949964c9SMarcos Paulo de Souza err = PTR_ERR(parent); 2917949964c9SMarcos Paulo de Souza goto out_drop_write; 2918949964c9SMarcos Paulo de Souza } 2919949964c9SMarcos Paulo de Souza dir = d_inode(parent); 2920949964c9SMarcos Paulo de Souza 2921949964c9SMarcos Paulo de Souza /* 2922949964c9SMarcos Paulo de Souza * If v2 was used with SPEC_BY_ID, a new parent was 2923949964c9SMarcos Paulo de Souza * allocated since the subvolume can be outside of the 2924949964c9SMarcos Paulo de Souza * current mount point. Later on we need to release this 2925949964c9SMarcos Paulo de Souza * new parent dentry. 2926949964c9SMarcos Paulo de Souza */ 2927949964c9SMarcos Paulo de Souza destroy_parent = true; 2928949964c9SMarcos Paulo de Souza 2929949964c9SMarcos Paulo de Souza subvol_name_ptr = btrfs_get_subvol_name_from_objectid( 2930949964c9SMarcos Paulo de Souza fs_info, vol_args2->subvolid); 2931949964c9SMarcos Paulo de Souza if (IS_ERR(subvol_name_ptr)) { 2932949964c9SMarcos Paulo de Souza err = PTR_ERR(subvol_name_ptr); 2933949964c9SMarcos Paulo de Souza goto free_parent; 2934949964c9SMarcos Paulo de Souza } 2935949964c9SMarcos Paulo de Souza /* subvol_name_ptr is already NULL termined */ 2936949964c9SMarcos Paulo de Souza subvol_name = (char *)kbasename(subvol_name_ptr); 2937949964c9SMarcos Paulo de Souza } 2938949964c9SMarcos Paulo de Souza } else { 2939949964c9SMarcos Paulo de Souza vol_args = memdup_user(arg, sizeof(*vol_args)); 2940949964c9SMarcos Paulo de Souza if (IS_ERR(vol_args)) 2941949964c9SMarcos Paulo de Souza return PTR_ERR(vol_args); 2942949964c9SMarcos Paulo de Souza 2943949964c9SMarcos Paulo de Souza vol_args->name[BTRFS_PATH_NAME_MAX] = 0; 2944949964c9SMarcos Paulo de Souza subvol_name = vol_args->name; 2945949964c9SMarcos Paulo de Souza 2946949964c9SMarcos Paulo de Souza err = mnt_want_write_file(file); 2947949964c9SMarcos Paulo de Souza if (err) 2948949964c9SMarcos Paulo de Souza goto out; 2949949964c9SMarcos Paulo de Souza } 2950949964c9SMarcos Paulo de Souza 2951949964c9SMarcos Paulo de Souza subvol_namelen = strlen(subvol_name); 2952949964c9SMarcos Paulo de Souza 2953949964c9SMarcos Paulo de Souza if (strchr(subvol_name, '/') || 2954949964c9SMarcos Paulo de Souza strncmp(subvol_name, "..", subvol_namelen) == 0) { 2955949964c9SMarcos Paulo de Souza err = -EINVAL; 2956949964c9SMarcos Paulo de Souza goto free_subvol_name; 2957949964c9SMarcos Paulo de Souza } 2958949964c9SMarcos Paulo de Souza 2959949964c9SMarcos Paulo de Souza if (!S_ISDIR(dir->i_mode)) { 2960949964c9SMarcos Paulo de Souza err = -ENOTDIR; 2961949964c9SMarcos Paulo de Souza goto free_subvol_name; 2962949964c9SMarcos Paulo de Souza } 2963521e0546SDavid Sterba 296400235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 296500235411SAl Viro if (err == -EINTR) 2966949964c9SMarcos Paulo de Souza goto free_subvol_name; 2967949964c9SMarcos Paulo de Souza dentry = lookup_one_len(subvol_name, parent, subvol_namelen); 296876dda93cSYan, Zheng if (IS_ERR(dentry)) { 296976dda93cSYan, Zheng err = PTR_ERR(dentry); 297076dda93cSYan, Zheng goto out_unlock_dir; 297176dda93cSYan, Zheng } 297276dda93cSYan, Zheng 29732b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 297476dda93cSYan, Zheng err = -ENOENT; 297576dda93cSYan, Zheng goto out_dput; 297676dda93cSYan, Zheng } 297776dda93cSYan, Zheng 29782b0143b5SDavid Howells inode = d_inode(dentry); 29794260f7c7SSage Weil dest = BTRFS_I(inode)->root; 29804260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 29814260f7c7SSage Weil /* 29824260f7c7SSage Weil * Regular user. Only allow this with a special mount 29834260f7c7SSage Weil * option, when the user has write+exec access to the 29844260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 29854260f7c7SSage Weil * allowed. 29864260f7c7SSage Weil * 29874260f7c7SSage Weil * Note that this is _not_ check that the subvol is 29884260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 29894260f7c7SSage Weil * otherwise be able to delete. 29904260f7c7SSage Weil * 29914260f7c7SSage Weil * Users who want to delete empty subvols should try 29924260f7c7SSage Weil * rmdir(2). 29934260f7c7SSage Weil */ 29944260f7c7SSage Weil err = -EPERM; 29950b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 29964260f7c7SSage Weil goto out_dput; 29974260f7c7SSage Weil 29984260f7c7SSage Weil /* 29994260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 30004260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 30014260f7c7SSage Weil * must be called on the dentry referencing the root 30024260f7c7SSage Weil * of the subvol, not a random directory contained 30034260f7c7SSage Weil * within it. 30044260f7c7SSage Weil */ 30054260f7c7SSage Weil err = -EINVAL; 30064260f7c7SSage Weil if (root == dest) 30074260f7c7SSage Weil goto out_dput; 30084260f7c7SSage Weil 30094260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 30104260f7c7SSage Weil if (err) 30114260f7c7SSage Weil goto out_dput; 30125c39da5bSMiao Xie } 30134260f7c7SSage Weil 30145c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 30154260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 30164260f7c7SSage Weil if (err) 30174260f7c7SSage Weil goto out_dput; 30184260f7c7SSage Weil 30194a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 302076dda93cSYan, Zheng err = -EINVAL; 302176dda93cSYan, Zheng goto out_dput; 302276dda93cSYan, Zheng } 302376dda93cSYan, Zheng 30245955102cSAl Viro inode_lock(inode); 3025f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 30265955102cSAl Viro inode_unlock(inode); 302746008d9dSAmir Goldstein if (!err) { 302846008d9dSAmir Goldstein fsnotify_rmdir(dir, dentry); 302976dda93cSYan, Zheng d_delete(dentry); 303046008d9dSAmir Goldstein } 3031fa6ac876SLiu Bo 303276dda93cSYan, Zheng out_dput: 303376dda93cSYan, Zheng dput(dentry); 303476dda93cSYan, Zheng out_unlock_dir: 30355955102cSAl Viro inode_unlock(dir); 3036949964c9SMarcos Paulo de Souza free_subvol_name: 3037949964c9SMarcos Paulo de Souza kfree(subvol_name_ptr); 3038949964c9SMarcos Paulo de Souza free_parent: 3039949964c9SMarcos Paulo de Souza if (destroy_parent) 3040949964c9SMarcos Paulo de Souza dput(parent); 304100235411SAl Viro out_drop_write: 30422a79f17eSAl Viro mnt_drop_write_file(file); 304376dda93cSYan, Zheng out: 3044949964c9SMarcos Paulo de Souza kfree(vol_args2); 304576dda93cSYan, Zheng kfree(vol_args); 304676dda93cSYan, Zheng return err; 304776dda93cSYan, Zheng } 304876dda93cSYan, Zheng 30491e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 3050f46b5a66SChristoph Hellwig { 3051496ad9aaSAl Viro struct inode *inode = file_inode(file); 3052f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 30531e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 3054c146afadSYan Zheng int ret; 3055c146afadSYan Zheng 305625122d15SIlya Dryomov ret = mnt_want_write_file(file); 305725122d15SIlya Dryomov if (ret) 305825122d15SIlya Dryomov return ret; 3059b83cc969SLi Zefan 306025122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 306125122d15SIlya Dryomov ret = -EROFS; 306225122d15SIlya Dryomov goto out; 30635ac00addSStefan Behrens } 3064f46b5a66SChristoph Hellwig 3065f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 3066f46b5a66SChristoph Hellwig case S_IFDIR: 3067e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 3068e441d54dSChris Mason ret = -EPERM; 3069e441d54dSChris Mason goto out; 3070e441d54dSChris Mason } 3071de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 3072f46b5a66SChristoph Hellwig break; 3073f46b5a66SChristoph Hellwig case S_IFREG: 3074616d374eSAdam Borowski /* 3075616d374eSAdam Borowski * Note that this does not check the file descriptor for write 3076616d374eSAdam Borowski * access. This prevents defragmenting executables that are 3077616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 3078616d374eSAdam Borowski */ 3079616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 3080616d374eSAdam Borowski inode_permission(inode, MAY_WRITE)) { 3081616d374eSAdam Borowski ret = -EPERM; 3082e441d54dSChris Mason goto out; 3083e441d54dSChris Mason } 30841e701a32SChris Mason 30851e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 30861e701a32SChris Mason if (!range) { 30871e701a32SChris Mason ret = -ENOMEM; 30881e701a32SChris Mason goto out; 30891e701a32SChris Mason } 30901e701a32SChris Mason 30911e701a32SChris Mason if (argp) { 30921e701a32SChris Mason if (copy_from_user(range, argp, 30931e701a32SChris Mason sizeof(*range))) { 30941e701a32SChris Mason ret = -EFAULT; 30951e701a32SChris Mason kfree(range); 3096683be16eSDan Carpenter goto out; 30971e701a32SChris Mason } 30981e701a32SChris Mason /* compression requires us to start the IO */ 30991e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 31001e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 31011e701a32SChris Mason range->extent_thresh = (u32)-1; 31021e701a32SChris Mason } 31031e701a32SChris Mason } else { 31041e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 31051e701a32SChris Mason range->len = (u64)-1; 31061e701a32SChris Mason } 3107496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 31087c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 31094cb5300bSChris Mason if (ret > 0) 31104cb5300bSChris Mason ret = 0; 31111e701a32SChris Mason kfree(range); 3112f46b5a66SChristoph Hellwig break; 31138929ecfaSYan, Zheng default: 31148929ecfaSYan, Zheng ret = -EINVAL; 3115f46b5a66SChristoph Hellwig } 3116e441d54dSChris Mason out: 311725122d15SIlya Dryomov mnt_drop_write_file(file); 3118e441d54dSChris Mason return ret; 3119f46b5a66SChristoph Hellwig } 3120f46b5a66SChristoph Hellwig 31212ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3122f46b5a66SChristoph Hellwig { 3123f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3124f46b5a66SChristoph Hellwig int ret; 3125f46b5a66SChristoph Hellwig 3126e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3127e441d54dSChris Mason return -EPERM; 3128e441d54dSChris Mason 3129171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 3130e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3131c9e9f97bSIlya Dryomov 3132dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3133c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3134c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3135c9e9f97bSIlya Dryomov goto out; 3136c9e9f97bSIlya Dryomov } 3137f46b5a66SChristoph Hellwig 31385516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31392ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3140f46b5a66SChristoph Hellwig 314143d20761SAnand Jain if (!ret) 31420b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 314343d20761SAnand Jain 3144f46b5a66SChristoph Hellwig kfree(vol_args); 3145c9e9f97bSIlya Dryomov out: 3146171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3147f46b5a66SChristoph Hellwig return ret; 3148f46b5a66SChristoph Hellwig } 3149f46b5a66SChristoph Hellwig 31506b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3151f46b5a66SChristoph Hellwig { 31520b246afaSJeff Mahoney struct inode *inode = file_inode(file); 31530b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 31546b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3155f46b5a66SChristoph Hellwig int ret; 3156f46b5a66SChristoph Hellwig 3157e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3158e441d54dSChris Mason return -EPERM; 3159e441d54dSChris Mason 3160da24927bSMiao Xie ret = mnt_want_write_file(file); 3161da24927bSMiao Xie if (ret) 3162da24927bSMiao Xie return ret; 3163c146afadSYan Zheng 3164dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3165c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3166c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3167c47ca32dSDan Carpenter goto err_drop; 3168c9e9f97bSIlya Dryomov } 3169f46b5a66SChristoph Hellwig 3170748449cdSDavid Sterba if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { 3171fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3172fd4e994bSOmar Sandoval goto out; 3173fd4e994bSOmar Sandoval } 3174f46b5a66SChristoph Hellwig 3175171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3176183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3177183860f6SAnand Jain goto out; 3178183860f6SAnand Jain } 3179183860f6SAnand Jain 3180735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 31812ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 31826b526ed7SAnand Jain } else { 31836b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 31842ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 31856b526ed7SAnand Jain } 3186171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3187183860f6SAnand Jain 31886b526ed7SAnand Jain if (!ret) { 3189735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 31900b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 31916b526ed7SAnand Jain vol_args->devid); 31926b526ed7SAnand Jain else 31930b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 31946b526ed7SAnand Jain vol_args->name); 31956b526ed7SAnand Jain } 3196183860f6SAnand Jain out: 3197183860f6SAnand Jain kfree(vol_args); 3198c47ca32dSDan Carpenter err_drop: 31994ac20c70SIlya Dryomov mnt_drop_write_file(file); 3200f46b5a66SChristoph Hellwig return ret; 3201f46b5a66SChristoph Hellwig } 3202f46b5a66SChristoph Hellwig 3203f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3204f46b5a66SChristoph Hellwig { 32050b246afaSJeff Mahoney struct inode *inode = file_inode(file); 32060b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3207f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3208f46b5a66SChristoph Hellwig int ret; 3209f46b5a66SChristoph Hellwig 3210f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3211f46b5a66SChristoph Hellwig return -EPERM; 3212f46b5a66SChristoph Hellwig 3213f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3214f46b5a66SChristoph Hellwig if (ret) 3215f46b5a66SChristoph Hellwig return ret; 3216f46b5a66SChristoph Hellwig 3217171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3218f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 321958d7bbf8SDavid Sterba goto out_drop_write; 322058d7bbf8SDavid Sterba } 322158d7bbf8SDavid Sterba 322258d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 322358d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 322458d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3225f46b5a66SChristoph Hellwig goto out; 3226f46b5a66SChristoph Hellwig } 3227f46b5a66SChristoph Hellwig 322858d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 32292ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3230f46b5a66SChristoph Hellwig 3231f46b5a66SChristoph Hellwig if (!ret) 32320b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3233f46b5a66SChristoph Hellwig kfree(vol_args); 323458d7bbf8SDavid Sterba out: 3235171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 323658d7bbf8SDavid Sterba out_drop_write: 3237f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 323858d7bbf8SDavid Sterba 3239f46b5a66SChristoph Hellwig return ret; 3240f46b5a66SChristoph Hellwig } 3241f46b5a66SChristoph Hellwig 32422ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 32432ff7e61eSJeff Mahoney void __user *arg) 3244475f6387SJan Schmidt { 3245027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3246475f6387SJan Schmidt struct btrfs_device *device; 32470b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3248027ed2f0SLi Zefan int ret = 0; 3249475f6387SJan Schmidt 3250027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3251027ed2f0SLi Zefan if (!fi_args) 3252027ed2f0SLi Zefan return -ENOMEM; 3253027ed2f0SLi Zefan 3254d03262c7SDavid Sterba rcu_read_lock(); 3255027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3256475f6387SJan Schmidt 3257d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3258027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3259027ed2f0SLi Zefan fi_args->max_id = device->devid; 3260475f6387SJan Schmidt } 3261d03262c7SDavid Sterba rcu_read_unlock(); 3262475f6387SJan Schmidt 3263de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3264bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3265bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3266bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 326780a773fbSDavid Sterba 3268027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3269027ed2f0SLi Zefan ret = -EFAULT; 3270475f6387SJan Schmidt 3271027ed2f0SLi Zefan kfree(fi_args); 3272027ed2f0SLi Zefan return ret; 3273475f6387SJan Schmidt } 3274475f6387SJan Schmidt 32752ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 32762ff7e61eSJeff Mahoney void __user *arg) 3277475f6387SJan Schmidt { 3278475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3279475f6387SJan Schmidt struct btrfs_device *dev; 3280475f6387SJan Schmidt int ret = 0; 3281475f6387SJan Schmidt char *s_uuid = NULL; 3282475f6387SJan Schmidt 3283475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3284475f6387SJan Schmidt if (IS_ERR(di_args)) 3285475f6387SJan Schmidt return PTR_ERR(di_args); 3286475f6387SJan Schmidt 3287dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3288475f6387SJan Schmidt s_uuid = di_args->uuid; 3289475f6387SJan Schmidt 3290c5593ca3SDavid Sterba rcu_read_lock(); 3291e4319cd9SAnand Jain dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid, 329209ba3bc9SAnand Jain NULL, true); 3293475f6387SJan Schmidt 3294475f6387SJan Schmidt if (!dev) { 3295475f6387SJan Schmidt ret = -ENODEV; 3296475f6387SJan Schmidt goto out; 3297475f6387SJan Schmidt } 3298475f6387SJan Schmidt 3299475f6387SJan Schmidt di_args->devid = dev->devid; 33007cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 33017cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3302475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3303a27202fbSJim Meyering if (dev->name) { 3304672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3305672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3306a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3307a27202fbSJim Meyering } else { 330899ba55adSStefan Behrens di_args->path[0] = '\0'; 3309a27202fbSJim Meyering } 3310475f6387SJan Schmidt 3311475f6387SJan Schmidt out: 3312c5593ca3SDavid Sterba rcu_read_unlock(); 3313475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3314475f6387SJan Schmidt ret = -EFAULT; 3315475f6387SJan Schmidt 3316475f6387SJan Schmidt kfree(di_args); 3317475f6387SJan Schmidt return ret; 3318475f6387SJan Schmidt } 3319475f6387SJan Schmidt 3320d8b55242SFilipe Manana static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3321d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3322d8b55242SFilipe Manana { 3323d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3324d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3325d8b55242SFilipe Manana } 3326d8b55242SFilipe Manana 3327d8b55242SFilipe Manana static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3328d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3329d8b55242SFilipe Manana { 3330d8b55242SFilipe Manana if (inode1 < inode2) { 3331d8b55242SFilipe Manana swap(inode1, inode2); 3332d8b55242SFilipe Manana swap(loff1, loff2); 3333d8b55242SFilipe Manana } else if (inode1 == inode2 && loff2 < loff1) { 3334d8b55242SFilipe Manana swap(loff1, loff2); 3335d8b55242SFilipe Manana } 3336d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3337d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3338d8b55242SFilipe Manana } 3339d8b55242SFilipe Manana 334057a50e25SFilipe Manana static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len, 334134a28e3dSFilipe Manana struct inode *dst, u64 dst_loff) 3342416161dbSMark Fasheh { 3343831d2fa2SFilipe Manana const u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 3344416161dbSMark Fasheh int ret; 33450efa9f48SMark Fasheh 3346de02b9f6SFilipe Manana /* 3347d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3348d8b55242SFilipe Manana * source range to serialize with relocation. 3349de02b9f6SFilipe Manana */ 3350d8b55242SFilipe Manana btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 3351831d2fa2SFilipe Manana ret = btrfs_clone(src, dst, loff, len, ALIGN(len, bs), dst_loff, 1); 3352d8b55242SFilipe Manana btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 33533973909dSTimofey Titovets 33543973909dSTimofey Titovets return ret; 33553973909dSTimofey Titovets } 33563973909dSTimofey Titovets 3357b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3358b6728768STimofey Titovets 33593973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 33603973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 33613973909dSTimofey Titovets { 33623973909dSTimofey Titovets int ret; 3363b6728768STimofey Titovets u64 i, tail_len, chunk_count; 336462d54f3aSFilipe Manana struct btrfs_root *root_dst = BTRFS_I(dst)->root; 336562d54f3aSFilipe Manana 336662d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 336762d54f3aSFilipe Manana if (root_dst->send_in_progress) { 336862d54f3aSFilipe Manana btrfs_warn_rl(root_dst->fs_info, 336962d54f3aSFilipe Manana "cannot deduplicate to root %llu while send operations are using it (%d in progress)", 337062d54f3aSFilipe Manana root_dst->root_key.objectid, 337162d54f3aSFilipe Manana root_dst->send_in_progress); 337262d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 337362d54f3aSFilipe Manana return -EAGAIN; 337462d54f3aSFilipe Manana } 337562d54f3aSFilipe Manana root_dst->dedupe_in_progress++; 337662d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 33773973909dSTimofey Titovets 3378b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3379b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 338067b07bd4STimofey Titovets 3381b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3382b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 338334a28e3dSFilipe Manana dst, dst_loff); 3384b6728768STimofey Titovets if (ret) 338562d54f3aSFilipe Manana goto out; 3386b6728768STimofey Titovets 3387b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3388b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3389b6728768STimofey Titovets } 3390b6728768STimofey Titovets 3391b6728768STimofey Titovets if (tail_len > 0) 339267b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 339334a28e3dSFilipe Manana dst_loff); 339462d54f3aSFilipe Manana out: 339562d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 339662d54f3aSFilipe Manana root_dst->dedupe_in_progress--; 339762d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 3398416161dbSMark Fasheh 3399416161dbSMark Fasheh return ret; 3400416161dbSMark Fasheh } 3401416161dbSMark Fasheh 3402f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3403f82a9901SFilipe Manana struct inode *inode, 3404f82a9901SFilipe Manana u64 endoff, 3405f82a9901SFilipe Manana const u64 destoff, 34061c919a5eSMark Fasheh const u64 olen, 34071c919a5eSMark Fasheh int no_time_update) 3408f82a9901SFilipe Manana { 3409f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3410f82a9901SFilipe Manana int ret; 3411f82a9901SFilipe Manana 3412f82a9901SFilipe Manana inode_inc_iversion(inode); 34131c919a5eSMark Fasheh if (!no_time_update) 3414c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3415f82a9901SFilipe Manana /* 3416f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3417f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3418f82a9901SFilipe Manana */ 3419f82a9901SFilipe Manana if (endoff > destoff + olen) 3420f82a9901SFilipe Manana endoff = destoff + olen; 3421790a1d44SJosef Bacik if (endoff > inode->i_size) { 3422790a1d44SJosef Bacik i_size_write(inode, endoff); 3423d923afe9SJosef Bacik btrfs_inode_safe_disk_i_size_write(inode, 0); 3424790a1d44SJosef Bacik } 3425f82a9901SFilipe Manana 3426f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3427f82a9901SFilipe Manana if (ret) { 342866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 34293a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3430f82a9901SFilipe Manana goto out; 3431f82a9901SFilipe Manana } 34323a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3433f82a9901SFilipe Manana out: 3434f82a9901SFilipe Manana return ret; 3435f82a9901SFilipe Manana } 3436f82a9901SFilipe Manana 34378039d87dSFilipe Manana /* 34388039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 34398039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 34408039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 34418039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 34428039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 34438039d87dSFilipe Manana * 34448039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 34458039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 34468039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 34478039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 34488039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 34498039d87dSFilipe Manana * i_mutex, which we are holding here). 34508039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 34518039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 34528039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 34538039d87dSFilipe Manana * 34548039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 34558039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 34568039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 34578039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 34588039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 34598039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 34608039d87dSFilipe Manana * the former. 34618039d87dSFilipe Manana */ 34624a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 34638039d87dSFilipe Manana struct btrfs_trans_handle *trans, 34648039d87dSFilipe Manana struct btrfs_path *path, 34658039d87dSFilipe Manana struct btrfs_key *new_key, 34668039d87dSFilipe Manana const u64 drop_start, 34678039d87dSFilipe Manana const u64 datal, 34688039d87dSFilipe Manana const u64 skip, 34698039d87dSFilipe Manana const u64 size, 34708039d87dSFilipe Manana char *inline_data) 34718039d87dSFilipe Manana { 34720b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 34738039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 34748039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 34750b246afaSJeff Mahoney fs_info->sectorsize); 34768039d87dSFilipe Manana int ret; 34778039d87dSFilipe Manana struct btrfs_key key; 34788039d87dSFilipe Manana 34798039d87dSFilipe Manana if (new_key->offset > 0) 34808039d87dSFilipe Manana return -EOPNOTSUPP; 34818039d87dSFilipe Manana 34824a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 34838039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 34848039d87dSFilipe Manana key.offset = 0; 34858039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 34868039d87dSFilipe Manana if (ret < 0) { 34878039d87dSFilipe Manana return ret; 34888039d87dSFilipe Manana } else if (ret > 0) { 34898039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 34908039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 34918039d87dSFilipe Manana if (ret < 0) 34928039d87dSFilipe Manana return ret; 34938039d87dSFilipe Manana else if (ret > 0) 34948039d87dSFilipe Manana goto copy_inline_extent; 34958039d87dSFilipe Manana } 34968039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 34974a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34988039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 34998039d87dSFilipe Manana ASSERT(key.offset > 0); 35008039d87dSFilipe Manana return -EOPNOTSUPP; 35018039d87dSFilipe Manana } 35028039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 35038039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 35048039d87dSFilipe Manana u64 ext_len; 35058039d87dSFilipe Manana 35068039d87dSFilipe Manana /* 35078039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 35088039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 35098039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 35108039d87dSFilipe Manana */ 35118039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 35128039d87dSFilipe Manana struct btrfs_file_extent_item); 35138039d87dSFilipe Manana /* 35148039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 35158039d87dSFilipe Manana * following it. 35168039d87dSFilipe Manana */ 35178039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 35188039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 35198039d87dSFilipe Manana goto copy_inline_extent; 35208039d87dSFilipe Manana 35218039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 35228039d87dSFilipe Manana if (ext_len > aligned_end) 35238039d87dSFilipe Manana return -EOPNOTSUPP; 35248039d87dSFilipe Manana 35258039d87dSFilipe Manana ret = btrfs_next_item(root, path); 35268039d87dSFilipe Manana if (ret < 0) { 35278039d87dSFilipe Manana return ret; 35288039d87dSFilipe Manana } else if (ret == 0) { 35298039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35308039d87dSFilipe Manana path->slots[0]); 35314a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 35328039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 35338039d87dSFilipe Manana return -EOPNOTSUPP; 35348039d87dSFilipe Manana } 35358039d87dSFilipe Manana } 35368039d87dSFilipe Manana 35378039d87dSFilipe Manana copy_inline_extent: 35388039d87dSFilipe Manana /* 35398039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 35408039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 35418039d87dSFilipe Manana */ 35428039d87dSFilipe Manana if (i_size_read(dst) > datal) { 35438039d87dSFilipe Manana /* 35448039d87dSFilipe Manana * If the destination inode has an inline extent... 35458039d87dSFilipe Manana * This would require copying the data from the source inline 35468039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 35478039d87dSFilipe Manana * But this is really complex, both extents can be compressed 35488039d87dSFilipe Manana * or just one of them, which would require decompressing and 35498039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 35508039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 35518039d87dSFilipe Manana * inline extent). 35528039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 35538039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 35548039d87dSFilipe Manana */ 35558039d87dSFilipe Manana return -EOPNOTSUPP; 35568039d87dSFilipe Manana } 35578039d87dSFilipe Manana 35588039d87dSFilipe Manana btrfs_release_path(path); 35598039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 35608039d87dSFilipe Manana if (ret) 35618039d87dSFilipe Manana return ret; 35628039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 35638039d87dSFilipe Manana if (ret) 35648039d87dSFilipe Manana return ret; 35658039d87dSFilipe Manana 35668039d87dSFilipe Manana if (skip) { 35678039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 35688039d87dSFilipe Manana 35698039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 35708039d87dSFilipe Manana } 35718039d87dSFilipe Manana 35728039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 35738039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 35748039d87dSFilipe Manana path->slots[0]), 35758039d87dSFilipe Manana size); 35768039d87dSFilipe Manana inode_add_bytes(dst, datal); 3577690a5dbfSFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(dst)->runtime_flags); 35788039d87dSFilipe Manana 35798039d87dSFilipe Manana return 0; 35808039d87dSFilipe Manana } 35818039d87dSFilipe Manana 358232b7c687SMark Fasheh /** 358332b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 358432b7c687SMark Fasheh * 358532b7c687SMark Fasheh * @src: Inode to clone from 358632b7c687SMark Fasheh * @inode: Inode to clone to 358732b7c687SMark Fasheh * @off: Offset within source to start clone from 358832b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 35891c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 359032b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 35911c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 359232b7c687SMark Fasheh */ 359332b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3594f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 35951c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3596f46b5a66SChristoph Hellwig { 35970b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3598f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 359932b7c687SMark Fasheh struct btrfs_path *path = NULL; 3600f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 360132b7c687SMark Fasheh struct btrfs_trans_handle *trans; 360232b7c687SMark Fasheh char *buf = NULL; 3603ae01a0abSYan Zheng struct btrfs_key key; 3604f46b5a66SChristoph Hellwig u32 nritems; 3605f46b5a66SChristoph Hellwig int slot; 3606ae01a0abSYan Zheng int ret; 3607f82a9901SFilipe Manana const u64 len = olen_aligned; 3608f82a9901SFilipe Manana u64 last_dest_end = destoff; 3609ae01a0abSYan Zheng 3610ae01a0abSYan Zheng ret = -ENOMEM; 3611752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3612ae01a0abSYan Zheng if (!buf) 361332b7c687SMark Fasheh return ret; 3614ae01a0abSYan Zheng 3615ae01a0abSYan Zheng path = btrfs_alloc_path(); 3616ae01a0abSYan Zheng if (!path) { 361715351955SDavid Sterba kvfree(buf); 361832b7c687SMark Fasheh return ret; 3619ae01a0abSYan Zheng } 362032b7c687SMark Fasheh 3621e4058b54SDavid Sterba path->reada = READA_FORWARD; 3622c5c9cd4dSSage Weil /* clone data */ 36234a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3624ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 36252c463823SFilipe Manana key.offset = off; 3626f46b5a66SChristoph Hellwig 3627f46b5a66SChristoph Hellwig while (1) { 3628de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3629b64119b5SFilipe Manana struct btrfs_file_extent_item *extent; 3630b64119b5SFilipe Manana int type; 3631b64119b5SFilipe Manana u32 size; 3632b64119b5SFilipe Manana struct btrfs_key new_key; 3633b64119b5SFilipe Manana u64 disko = 0, diskl = 0; 3634b64119b5SFilipe Manana u64 datao = 0, datal = 0; 3635b64119b5SFilipe Manana u8 comp; 3636b64119b5SFilipe Manana u64 drop_start; 3637df858e76SFilipe Manana 3638f46b5a66SChristoph Hellwig /* 3639f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3640f46b5a66SChristoph Hellwig * tree. 3641f46b5a66SChristoph Hellwig */ 3642e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3643362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3644362a20c5SDavid Sterba 0, 0); 3645f46b5a66SChristoph Hellwig if (ret < 0) 3646f46b5a66SChristoph Hellwig goto out; 36472c463823SFilipe Manana /* 36482c463823SFilipe Manana * First search, if no extent item that starts at offset off was 36492c463823SFilipe Manana * found but the previous item is an extent item, it's possible 36502c463823SFilipe Manana * it might overlap our target range, therefore process it. 36512c463823SFilipe Manana */ 36522c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 36532c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 36542c463823SFilipe Manana path->slots[0] - 1); 36552c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 36562c463823SFilipe Manana path->slots[0]--; 36572c463823SFilipe Manana } 3658f46b5a66SChristoph Hellwig 3659ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3660e4355f34SFilipe David Borba Manana process_slot: 3661ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3662362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3663f46b5a66SChristoph Hellwig if (ret < 0) 3664f46b5a66SChristoph Hellwig goto out; 3665f46b5a66SChristoph Hellwig if (ret > 0) 3666f46b5a66SChristoph Hellwig break; 3667ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3668f46b5a66SChristoph Hellwig } 3669f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3670f46b5a66SChristoph Hellwig slot = path->slots[0]; 3671f46b5a66SChristoph Hellwig 3672ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3673962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 36744a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3675f46b5a66SChristoph Hellwig break; 3676f46b5a66SChristoph Hellwig 3677b64119b5SFilipe Manana ASSERT(key.type == BTRFS_EXTENT_DATA_KEY); 367831840ae1SZheng Yan 3679c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3680c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3681c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3682c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3683c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3684c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3685b64119b5SFilipe Manana disko = btrfs_file_extent_disk_bytenr(leaf, extent); 3686b64119b5SFilipe Manana diskl = btrfs_file_extent_disk_num_bytes(leaf, extent); 3687c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3688b64119b5SFilipe Manana datal = btrfs_file_extent_num_bytes(leaf, extent); 3689c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3690b64119b5SFilipe Manana /* Take upper bound, may be compressed */ 3691b64119b5SFilipe Manana datal = btrfs_file_extent_ram_bytes(leaf, extent); 3692c5c9cd4dSSage Weil } 369331840ae1SZheng Yan 36942c463823SFilipe Manana /* 3695b64119b5SFilipe Manana * The first search might have left us at an extent item that 3696b64119b5SFilipe Manana * ends before our target range's start, can happen if we have 3697b64119b5SFilipe Manana * holes and NO_HOLES feature enabled. 36982c463823SFilipe Manana */ 36992c463823SFilipe Manana if (key.offset + datal <= off) { 3700e4355f34SFilipe David Borba Manana path->slots[0]++; 3701e4355f34SFilipe David Borba Manana goto process_slot; 37022c463823SFilipe Manana } else if (key.offset >= off + len) { 37032c463823SFilipe Manana break; 3704e4355f34SFilipe David Borba Manana } 3705df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3706e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3707b64119b5SFilipe Manana read_extent_buffer(leaf, buf, btrfs_item_ptr_offset(leaf, slot), 3708e4355f34SFilipe David Borba Manana size); 3709e4355f34SFilipe David Borba Manana 3710e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3711e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3712c5c9cd4dSSage Weil 371331840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 37144a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 37154d728ec7SLi Zefan if (off <= key.offset) 3716c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 37174d728ec7SLi Zefan else 37184d728ec7SLi Zefan new_key.offset = destoff; 3719c5c9cd4dSSage Weil 3720b6f3409bSSage Weil /* 3721b64119b5SFilipe Manana * Deal with a hole that doesn't have an extent item that 3722b64119b5SFilipe Manana * represents it (NO_HOLES feature enabled). 3723b64119b5SFilipe Manana * This hole is either in the middle of the cloning range or at 3724b64119b5SFilipe Manana * the beginning (fully overlaps it or partially overlaps it). 3725f82a9901SFilipe Manana */ 3726f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3727f82a9901SFilipe Manana drop_start = last_dest_end; 3728f82a9901SFilipe Manana else 3729f82a9901SFilipe Manana drop_start = new_key.offset; 3730f82a9901SFilipe Manana 3731c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3732c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3733690a5dbfSFilipe Manana struct btrfs_clone_extent_info clone_info; 3734690a5dbfSFilipe Manana 3735d72c0842SLi Zefan /* 3736d72c0842SLi Zefan * a | --- range to clone ---| b 3737d72c0842SLi Zefan * | ------------- extent ------------- | 3738d72c0842SLi Zefan */ 3739d72c0842SLi Zefan 3740b64119b5SFilipe Manana /* Subtract range b */ 3741d72c0842SLi Zefan if (key.offset + datal > off + len) 3742d72c0842SLi Zefan datal = off + len - key.offset; 3743d72c0842SLi Zefan 3744b64119b5SFilipe Manana /* Subtract range a */ 3745a22285a6SYan, Zheng if (off > key.offset) { 3746a22285a6SYan, Zheng datao += off - key.offset; 3747a22285a6SYan, Zheng datal -= off - key.offset; 3748a22285a6SYan, Zheng } 3749a22285a6SYan, Zheng 3750690a5dbfSFilipe Manana clone_info.disk_offset = disko; 3751690a5dbfSFilipe Manana clone_info.disk_len = diskl; 3752690a5dbfSFilipe Manana clone_info.data_offset = datao; 3753690a5dbfSFilipe Manana clone_info.data_len = datal; 3754690a5dbfSFilipe Manana clone_info.file_offset = new_key.offset; 3755690a5dbfSFilipe Manana clone_info.extent_buf = buf; 3756690a5dbfSFilipe Manana clone_info.item_size = size; 3757690a5dbfSFilipe Manana ret = btrfs_punch_hole_range(inode, path, 3758f82a9901SFilipe Manana drop_start, 3759690a5dbfSFilipe Manana new_key.offset + datal - 1, 3760690a5dbfSFilipe Manana &clone_info, &trans); 3761690a5dbfSFilipe Manana if (ret) 376279787eaaSJeff Mahoney goto out; 3763c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3764c5c9cd4dSSage Weil u64 skip = 0; 3765c5c9cd4dSSage Weil u64 trim = 0; 3766ed958762SFilipe Manana 3767c5c9cd4dSSage Weil if (off > key.offset) { 3768c5c9cd4dSSage Weil skip = off - key.offset; 3769c5c9cd4dSSage Weil new_key.offset += skip; 377031840ae1SZheng Yan } 3771d397712bSChris Mason 3772c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3773c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3774d397712bSChris Mason 3775c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3776c5c9cd4dSSage Weil ret = -EINVAL; 3777c5c9cd4dSSage Weil goto out; 377831840ae1SZheng Yan } 3779c5c9cd4dSSage Weil size -= skip + trim; 3780c5c9cd4dSSage Weil datal -= skip + trim; 3781a22285a6SYan, Zheng 3782690a5dbfSFilipe Manana /* 3783b64119b5SFilipe Manana * If our extent is inline, we know we will drop or 3784b64119b5SFilipe Manana * adjust at most 1 extent item in the destination root. 3785690a5dbfSFilipe Manana * 3786b64119b5SFilipe Manana * 1 - adjusting old extent (we may have to split it) 3787690a5dbfSFilipe Manana * 1 - add new extent 3788690a5dbfSFilipe Manana * 1 - inode update 3789690a5dbfSFilipe Manana */ 3790690a5dbfSFilipe Manana trans = btrfs_start_transaction(root, 3); 3791690a5dbfSFilipe Manana if (IS_ERR(trans)) { 3792690a5dbfSFilipe Manana ret = PTR_ERR(trans); 3793690a5dbfSFilipe Manana goto out; 3794690a5dbfSFilipe Manana } 3795690a5dbfSFilipe Manana 3796b64119b5SFilipe Manana ret = clone_copy_inline_extent(inode, trans, path, 3797b64119b5SFilipe Manana &new_key, drop_start, 3798b64119b5SFilipe Manana datal, skip, size, buf); 379979787eaaSJeff Mahoney if (ret) { 38003f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 3801b64119b5SFilipe Manana btrfs_abort_transaction(trans, ret); 38023a45bb20SJeff Mahoney btrfs_end_transaction(trans); 380379787eaaSJeff Mahoney goto out; 380479787eaaSJeff Mahoney } 3805c5c9cd4dSSage Weil } 3806c5c9cd4dSSage Weil 3807b3b4aa74SDavid Sterba btrfs_release_path(path); 3808c5c9cd4dSSage Weil 380962e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 38100b246afaSJeff Mahoney fs_info->sectorsize); 3811b64119b5SFilipe Manana ret = clone_finish_inode_update(trans, inode, last_dest_end, 3812b64119b5SFilipe Manana destoff, olen, no_time_update); 3813f82a9901SFilipe Manana if (ret) 381479787eaaSJeff Mahoney goto out; 38152c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 38162c463823SFilipe Manana break; 3817b64119b5SFilipe Manana 3818b3b4aa74SDavid Sterba btrfs_release_path(path); 3819df858e76SFilipe Manana key.offset = next_key_min_offset; 382069ae5e44SWang Xiaoguang 382169ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 382269ae5e44SWang Xiaoguang ret = -EINTR; 382369ae5e44SWang Xiaoguang goto out; 382469ae5e44SWang Xiaoguang } 3825ae01a0abSYan Zheng } 3826f46b5a66SChristoph Hellwig ret = 0; 382732b7c687SMark Fasheh 3828f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3829f82a9901SFilipe Manana /* 3830147271e3SFilipe Manana * We have an implicit hole that fully or partially overlaps our 3831147271e3SFilipe Manana * cloning range at its end. This means that we either have the 3832147271e3SFilipe Manana * NO_HOLES feature enabled or the implicit hole happened due to 3833147271e3SFilipe Manana * mixing buffered and direct IO writes against this file. 3834f82a9901SFilipe Manana */ 3835b3b4aa74SDavid Sterba btrfs_release_path(path); 3836690a5dbfSFilipe Manana path->leave_spinning = 0; 3837f82a9901SFilipe Manana 3838690a5dbfSFilipe Manana ret = btrfs_punch_hole_range(inode, path, 3839690a5dbfSFilipe Manana last_dest_end, destoff + len - 1, 3840147271e3SFilipe Manana NULL, &trans); 3841690a5dbfSFilipe Manana if (ret) 3842f82a9901SFilipe Manana goto out; 3843690a5dbfSFilipe Manana 3844f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38451c919a5eSMark Fasheh destoff, olen, no_time_update); 3846f82a9901SFilipe Manana } 3847f82a9901SFilipe Manana 3848f46b5a66SChristoph Hellwig out: 384932b7c687SMark Fasheh btrfs_free_path(path); 385015351955SDavid Sterba kvfree(buf); 385132b7c687SMark Fasheh return ret; 385232b7c687SMark Fasheh } 385332b7c687SMark Fasheh 38543db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 385532b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 385632b7c687SMark Fasheh { 385754563d41SAl Viro struct inode *inode = file_inode(file); 38583db11b2eSZach Brown struct inode *src = file_inode(file_src); 38590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 386032b7c687SMark Fasheh int ret; 386132b7c687SMark Fasheh u64 len = olen; 38620b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 386332b7c687SMark Fasheh 386432b7c687SMark Fasheh /* 386532b7c687SMark Fasheh * TODO: 386632b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 386732b7c687SMark Fasheh * decompress into destination's address_space (the file offset 386832b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 386932b7c687SMark Fasheh * otherwise reinsert) a subrange. 387000fdf13aSLiu Bo * 387100fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 387200fdf13aSLiu Bo * be either compressed or non-compressed. 387332b7c687SMark Fasheh */ 387432b7c687SMark Fasheh 3875ac765f83SFilipe Manana /* 387634a28e3dSFilipe Manana * VFS's generic_remap_file_range_prep() protects us from cloning the 387734a28e3dSFilipe Manana * eof block into the middle of a file, which would result in corruption 387834a28e3dSFilipe Manana * if the file size is not blocksize aligned. So we don't need to check 387934a28e3dSFilipe Manana * for that case here. 3880ac765f83SFilipe Manana */ 388134a28e3dSFilipe Manana if (off + len == src->i_size) 388232b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 388332b7c687SMark Fasheh 388432b7c687SMark Fasheh if (destoff > inode->i_size) { 3885f7fa1107SFilipe Manana const u64 wb_start = ALIGN_DOWN(inode->i_size, bs); 3886f7fa1107SFilipe Manana 388732b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 388832b7c687SMark Fasheh if (ret) 388934a28e3dSFilipe Manana return ret; 3890f7fa1107SFilipe Manana /* 3891f7fa1107SFilipe Manana * We may have truncated the last block if the inode's size is 3892f7fa1107SFilipe Manana * not sector size aligned, so we need to wait for writeback to 3893f7fa1107SFilipe Manana * complete before proceeding further, otherwise we can race 3894f7fa1107SFilipe Manana * with cloning and attempt to increment a reference to an 3895f7fa1107SFilipe Manana * extent that no longer exists (writeback completed right after 3896f7fa1107SFilipe Manana * we found the previous extent covering eof and before we 3897f7fa1107SFilipe Manana * attempted to increment its reference count). 3898f7fa1107SFilipe Manana */ 3899f7fa1107SFilipe Manana ret = btrfs_wait_ordered_range(inode, wb_start, 3900f7fa1107SFilipe Manana destoff - wb_start); 3901f7fa1107SFilipe Manana if (ret) 3902f7fa1107SFilipe Manana return ret; 390332b7c687SMark Fasheh } 390432b7c687SMark Fasheh 3905c125b8bfSFilipe Manana /* 3906d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3907d8b55242SFilipe Manana * source range to serialize with relocation. 3908c125b8bfSFilipe Manana */ 3909d8b55242SFilipe Manana btrfs_double_extent_lock(src, off, inode, destoff, len); 39101c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 3911d8b55242SFilipe Manana btrfs_double_extent_unlock(src, off, inode, destoff, len); 3912c125b8bfSFilipe Manana /* 3913c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3914c125b8bfSFilipe Manana * data immediately and not the previous data. 3915c125b8bfSFilipe Manana */ 391665bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 391709cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 391809cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 391934a28e3dSFilipe Manana 392034a28e3dSFilipe Manana return ret; 392134a28e3dSFilipe Manana } 392234a28e3dSFilipe Manana 392334a28e3dSFilipe Manana static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in, 392434a28e3dSFilipe Manana struct file *file_out, loff_t pos_out, 392534a28e3dSFilipe Manana loff_t *len, unsigned int remap_flags) 392634a28e3dSFilipe Manana { 392734a28e3dSFilipe Manana struct inode *inode_in = file_inode(file_in); 392834a28e3dSFilipe Manana struct inode *inode_out = file_inode(file_out); 392934a28e3dSFilipe Manana u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize; 393034a28e3dSFilipe Manana bool same_inode = inode_out == inode_in; 393134a28e3dSFilipe Manana u64 wb_len; 393234a28e3dSFilipe Manana int ret; 393334a28e3dSFilipe Manana 393434a28e3dSFilipe Manana if (!(remap_flags & REMAP_FILE_DEDUP)) { 393534a28e3dSFilipe Manana struct btrfs_root *root_out = BTRFS_I(inode_out)->root; 393634a28e3dSFilipe Manana 393734a28e3dSFilipe Manana if (btrfs_root_readonly(root_out)) 393834a28e3dSFilipe Manana return -EROFS; 393934a28e3dSFilipe Manana 394034a28e3dSFilipe Manana if (file_in->f_path.mnt != file_out->f_path.mnt || 394134a28e3dSFilipe Manana inode_in->i_sb != inode_out->i_sb) 394234a28e3dSFilipe Manana return -EXDEV; 394334a28e3dSFilipe Manana } 394434a28e3dSFilipe Manana 3945500710d3SFilipe Manana /* don't make the dst file partly checksummed */ 3946500710d3SFilipe Manana if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) != 3947500710d3SFilipe Manana (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) { 39487984ae52SGoldwyn Rodrigues return -EINVAL; 3949500710d3SFilipe Manana } 3950500710d3SFilipe Manana 395134a28e3dSFilipe Manana /* 395234a28e3dSFilipe Manana * Now that the inodes are locked, we need to start writeback ourselves 395334a28e3dSFilipe Manana * and can not rely on the writeback from the VFS's generic helper 395434a28e3dSFilipe Manana * generic_remap_file_range_prep() because: 395534a28e3dSFilipe Manana * 395634a28e3dSFilipe Manana * 1) For compression we must call filemap_fdatawrite_range() range 395734a28e3dSFilipe Manana * twice (btrfs_fdatawrite_range() does it for us), and the generic 395834a28e3dSFilipe Manana * helper only calls it once; 395934a28e3dSFilipe Manana * 396034a28e3dSFilipe Manana * 2) filemap_fdatawrite_range(), called by the generic helper only 396134a28e3dSFilipe Manana * waits for the writeback to complete, i.e. for IO to be done, and 396234a28e3dSFilipe Manana * not for the ordered extents to complete. We need to wait for them 396334a28e3dSFilipe Manana * to complete so that new file extent items are in the fs tree. 396434a28e3dSFilipe Manana */ 396534a28e3dSFilipe Manana if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP)) 396634a28e3dSFilipe Manana wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs); 396734a28e3dSFilipe Manana else 396834a28e3dSFilipe Manana wb_len = ALIGN(*len, bs); 396934a28e3dSFilipe Manana 397034a28e3dSFilipe Manana /* 397134a28e3dSFilipe Manana * Since we don't lock ranges, wait for ongoing lockless dio writes (as 397234a28e3dSFilipe Manana * any in progress could create its ordered extents after we wait for 397334a28e3dSFilipe Manana * existing ordered extents below). 397434a28e3dSFilipe Manana */ 397534a28e3dSFilipe Manana inode_dio_wait(inode_in); 397634a28e3dSFilipe Manana if (!same_inode) 397734a28e3dSFilipe Manana inode_dio_wait(inode_out); 397834a28e3dSFilipe Manana 3979a94d1d0cSQu Wenruo /* 3980a94d1d0cSQu Wenruo * Workaround to make sure NOCOW buffered write reach disk as NOCOW. 3981a94d1d0cSQu Wenruo * 3982a94d1d0cSQu Wenruo * Btrfs' back references do not have a block level granularity, they 3983a94d1d0cSQu Wenruo * work at the whole extent level. 3984a94d1d0cSQu Wenruo * NOCOW buffered write without data space reserved may not be able 3985a94d1d0cSQu Wenruo * to fall back to CoW due to lack of data space, thus could cause 3986a94d1d0cSQu Wenruo * data loss. 3987a94d1d0cSQu Wenruo * 3988a94d1d0cSQu Wenruo * Here we take a shortcut by flushing the whole inode, so that all 3989a94d1d0cSQu Wenruo * nocow write should reach disk as nocow before we increase the 3990a94d1d0cSQu Wenruo * reference of the extent. We could do better by only flushing NOCOW 3991a94d1d0cSQu Wenruo * data, but that needs extra accounting. 3992a94d1d0cSQu Wenruo * 3993a94d1d0cSQu Wenruo * Also we don't need to check ASYNC_EXTENT, as async extent will be 3994a94d1d0cSQu Wenruo * CoWed anyway, not affecting nocow part. 3995a94d1d0cSQu Wenruo */ 3996a94d1d0cSQu Wenruo ret = filemap_flush(inode_in->i_mapping); 3997a94d1d0cSQu Wenruo if (ret < 0) 3998a94d1d0cSQu Wenruo return ret; 3999a94d1d0cSQu Wenruo 400034a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs), 400134a28e3dSFilipe Manana wb_len); 400234a28e3dSFilipe Manana if (ret < 0) 40037984ae52SGoldwyn Rodrigues return ret; 400434a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs), 400534a28e3dSFilipe Manana wb_len); 400634a28e3dSFilipe Manana if (ret < 0) 40073db11b2eSZach Brown return ret; 40087984ae52SGoldwyn Rodrigues 40097984ae52SGoldwyn Rodrigues return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out, 40107984ae52SGoldwyn Rodrigues len, remap_flags); 40113db11b2eSZach Brown } 40123db11b2eSZach Brown 401342ec3d4cSDarrick J. Wong loff_t btrfs_remap_file_range(struct file *src_file, loff_t off, 401442ec3d4cSDarrick J. Wong struct file *dst_file, loff_t destoff, loff_t len, 40152e5dfc99SDarrick J. Wong unsigned int remap_flags) 40163db11b2eSZach Brown { 401734a28e3dSFilipe Manana struct inode *src_inode = file_inode(src_file); 401834a28e3dSFilipe Manana struct inode *dst_inode = file_inode(dst_file); 401934a28e3dSFilipe Manana bool same_inode = dst_inode == src_inode; 402042ec3d4cSDarrick J. Wong int ret; 402142ec3d4cSDarrick J. Wong 40222e5dfc99SDarrick J. Wong if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY)) 40232e5dfc99SDarrick J. Wong return -EINVAL; 40242e5dfc99SDarrick J. Wong 40257984ae52SGoldwyn Rodrigues if (same_inode) 40267984ae52SGoldwyn Rodrigues inode_lock(src_inode); 40277984ae52SGoldwyn Rodrigues else 40287984ae52SGoldwyn Rodrigues lock_two_nondirectories(src_inode, dst_inode); 40297984ae52SGoldwyn Rodrigues 403034a28e3dSFilipe Manana ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff, 403134a28e3dSFilipe Manana &len, remap_flags); 403234a28e3dSFilipe Manana if (ret < 0 || len == 0) 40337984ae52SGoldwyn Rodrigues goto out_unlock; 40342e5dfc99SDarrick J. Wong 403534a28e3dSFilipe Manana if (remap_flags & REMAP_FILE_DEDUP) 403634a28e3dSFilipe Manana ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff); 403734a28e3dSFilipe Manana else 403842ec3d4cSDarrick J. Wong ret = btrfs_clone_files(dst_file, src_file, off, len, destoff); 403934a28e3dSFilipe Manana 40407984ae52SGoldwyn Rodrigues out_unlock: 404134a28e3dSFilipe Manana if (same_inode) 404234a28e3dSFilipe Manana inode_unlock(src_inode); 404334a28e3dSFilipe Manana else 40444ea748e1SFilipe Manana unlock_two_nondirectories(src_inode, dst_inode); 404534a28e3dSFilipe Manana 404642ec3d4cSDarrick J. Wong return ret < 0 ? ret : len; 4047c5c9cd4dSSage Weil } 4048c5c9cd4dSSage Weil 40496ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40506ef5ed0dSJosef Bacik { 4051496ad9aaSAl Viro struct inode *inode = file_inode(file); 40520b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 40536ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40546ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40556ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40566ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40572a2b5d62SJosef Bacik struct btrfs_path *path = NULL; 40586ef5ed0dSJosef Bacik struct btrfs_key location; 40596ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40606ef5ed0dSJosef Bacik u64 objectid = 0; 40616ef5ed0dSJosef Bacik u64 dir_id; 40623c04ce01SMiao Xie int ret; 40636ef5ed0dSJosef Bacik 40646ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40656ef5ed0dSJosef Bacik return -EPERM; 40666ef5ed0dSJosef Bacik 40673c04ce01SMiao Xie ret = mnt_want_write_file(file); 40683c04ce01SMiao Xie if (ret) 40693c04ce01SMiao Xie return ret; 40703c04ce01SMiao Xie 40713c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40723c04ce01SMiao Xie ret = -EFAULT; 40733c04ce01SMiao Xie goto out; 40743c04ce01SMiao Xie } 40756ef5ed0dSJosef Bacik 40766ef5ed0dSJosef Bacik if (!objectid) 40771cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40786ef5ed0dSJosef Bacik 40796ef5ed0dSJosef Bacik location.objectid = objectid; 40806ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40816ef5ed0dSJosef Bacik location.offset = (u64)-1; 40826ef5ed0dSJosef Bacik 40833619c94fSJosef Bacik new_root = btrfs_get_fs_root(fs_info, &location, true); 40843c04ce01SMiao Xie if (IS_ERR(new_root)) { 40853c04ce01SMiao Xie ret = PTR_ERR(new_root); 40863c04ce01SMiao Xie goto out; 40873c04ce01SMiao Xie } 40882a2b5d62SJosef Bacik if (!is_fstree(new_root->root_key.objectid)) { 40892a2b5d62SJosef Bacik ret = -ENOENT; 40902a2b5d62SJosef Bacik goto out_free; 40912a2b5d62SJosef Bacik } 40926ef5ed0dSJosef Bacik 40936ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40943c04ce01SMiao Xie if (!path) { 40953c04ce01SMiao Xie ret = -ENOMEM; 40962a2b5d62SJosef Bacik goto out_free; 40973c04ce01SMiao Xie } 40986ef5ed0dSJosef Bacik path->leave_spinning = 1; 40996ef5ed0dSJosef Bacik 41006ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 410198d5dc13STsutomu Itoh if (IS_ERR(trans)) { 41023c04ce01SMiao Xie ret = PTR_ERR(trans); 41032a2b5d62SJosef Bacik goto out_free; 41046ef5ed0dSJosef Bacik } 41056ef5ed0dSJosef Bacik 41060b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 41070b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 41086ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4109cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 41102a2b5d62SJosef Bacik btrfs_release_path(path); 41113a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41120b246afaSJeff Mahoney btrfs_err(fs_info, 41135d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 41143c04ce01SMiao Xie ret = -ENOENT; 41152a2b5d62SJosef Bacik goto out_free; 41166ef5ed0dSJosef Bacik } 41176ef5ed0dSJosef Bacik 41186ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 41196ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 41206ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 41212a2b5d62SJosef Bacik btrfs_release_path(path); 41226ef5ed0dSJosef Bacik 41230b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 41243a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41252a2b5d62SJosef Bacik out_free: 412600246528SJosef Bacik btrfs_put_root(new_root); 41272a2b5d62SJosef Bacik btrfs_free_path(path); 41283c04ce01SMiao Xie out: 41293c04ce01SMiao Xie mnt_drop_write_file(file); 41303c04ce01SMiao Xie return ret; 41316ef5ed0dSJosef Bacik } 41326ef5ed0dSJosef Bacik 4133c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4134bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4135bf5fc093SJosef Bacik { 413632da5386SDavid Sterba struct btrfs_block_group *block_group; 4137bf5fc093SJosef Bacik 4138bf5fc093SJosef Bacik space->total_bytes = 0; 4139bf5fc093SJosef Bacik space->used_bytes = 0; 4140bf5fc093SJosef Bacik space->flags = 0; 4141bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4142bf5fc093SJosef Bacik space->flags = block_group->flags; 4143b3470b5dSDavid Sterba space->total_bytes += block_group->length; 4144bf38be65SDavid Sterba space->used_bytes += block_group->used; 4145bf5fc093SJosef Bacik } 4146bf5fc093SJosef Bacik } 4147bf5fc093SJosef Bacik 41482ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 41492ff7e61eSJeff Mahoney void __user *arg) 41501406e432SJosef Bacik { 41511406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41521406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41531406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41547fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 415513f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41561406e432SJosef Bacik struct btrfs_space_info *info; 4157315d8e98SColin Ian King static const u64 types[] = { 4158315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4159bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4160bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4161315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4162315d8e98SColin Ian King }; 4163bf5fc093SJosef Bacik int num_types = 4; 41647fde62bfSChris Mason int alloc_size; 41651406e432SJosef Bacik int ret = 0; 416651788b1bSDan Rosenberg u64 slot_count = 0; 4167bf5fc093SJosef Bacik int i, c; 41681406e432SJosef Bacik 41691406e432SJosef Bacik if (copy_from_user(&space_args, 41701406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41711406e432SJosef Bacik sizeof(space_args))) 41721406e432SJosef Bacik return -EFAULT; 41731406e432SJosef Bacik 4174bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4175bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4176bf5fc093SJosef Bacik 4177bf5fc093SJosef Bacik info = NULL; 41787fde62bfSChris Mason rcu_read_lock(); 41790b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4180bf5fc093SJosef Bacik list) { 4181bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4182bf5fc093SJosef Bacik info = tmp; 4183bf5fc093SJosef Bacik break; 4184bf5fc093SJosef Bacik } 4185bf5fc093SJosef Bacik } 41867fde62bfSChris Mason rcu_read_unlock(); 41871406e432SJosef Bacik 4188bf5fc093SJosef Bacik if (!info) 4189bf5fc093SJosef Bacik continue; 4190bf5fc093SJosef Bacik 4191bf5fc093SJosef Bacik down_read(&info->groups_sem); 4192bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4193bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4194bf5fc093SJosef Bacik slot_count++; 4195bf5fc093SJosef Bacik } 4196bf5fc093SJosef Bacik up_read(&info->groups_sem); 4197bf5fc093SJosef Bacik } 4198bf5fc093SJosef Bacik 419936523e95SDavid Sterba /* 420036523e95SDavid Sterba * Global block reserve, exported as a space_info 420136523e95SDavid Sterba */ 420236523e95SDavid Sterba slot_count++; 420336523e95SDavid Sterba 42047fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 42057fde62bfSChris Mason if (space_args.space_slots == 0) { 42067fde62bfSChris Mason space_args.total_spaces = slot_count; 42077fde62bfSChris Mason goto out; 42087fde62bfSChris Mason } 4209bf5fc093SJosef Bacik 421051788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4211bf5fc093SJosef Bacik 42127fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4213bf5fc093SJosef Bacik 42147fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 42157fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 42167fde62bfSChris Mason */ 421709cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 42187fde62bfSChris Mason return -ENOMEM; 42197fde62bfSChris Mason 42207fde62bfSChris Mason space_args.total_spaces = 0; 42218d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 42227fde62bfSChris Mason if (!dest) 42237fde62bfSChris Mason return -ENOMEM; 42247fde62bfSChris Mason dest_orig = dest; 42257fde62bfSChris Mason 42267fde62bfSChris Mason /* now we have a buffer to copy into */ 4227bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4228bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4229bf5fc093SJosef Bacik 423051788b1bSDan Rosenberg if (!slot_count) 423151788b1bSDan Rosenberg break; 423251788b1bSDan Rosenberg 4233bf5fc093SJosef Bacik info = NULL; 42341406e432SJosef Bacik rcu_read_lock(); 42350b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4236bf5fc093SJosef Bacik list) { 4237bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4238bf5fc093SJosef Bacik info = tmp; 42397fde62bfSChris Mason break; 4240bf5fc093SJosef Bacik } 4241bf5fc093SJosef Bacik } 4242bf5fc093SJosef Bacik rcu_read_unlock(); 42437fde62bfSChris Mason 4244bf5fc093SJosef Bacik if (!info) 4245bf5fc093SJosef Bacik continue; 4246bf5fc093SJosef Bacik down_read(&info->groups_sem); 4247bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4248bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4249c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4250c065f5b1SSu Yue &space); 42517fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42521406e432SJosef Bacik dest++; 42531406e432SJosef Bacik space_args.total_spaces++; 425451788b1bSDan Rosenberg slot_count--; 42551406e432SJosef Bacik } 425651788b1bSDan Rosenberg if (!slot_count) 425751788b1bSDan Rosenberg break; 4258bf5fc093SJosef Bacik } 4259bf5fc093SJosef Bacik up_read(&info->groups_sem); 4260bf5fc093SJosef Bacik } 42611406e432SJosef Bacik 426236523e95SDavid Sterba /* 426336523e95SDavid Sterba * Add global block reserve 426436523e95SDavid Sterba */ 426536523e95SDavid Sterba if (slot_count) { 42660b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 426736523e95SDavid Sterba 426836523e95SDavid Sterba spin_lock(&block_rsv->lock); 426936523e95SDavid Sterba space.total_bytes = block_rsv->size; 427036523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 427136523e95SDavid Sterba spin_unlock(&block_rsv->lock); 427236523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 427336523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 427436523e95SDavid Sterba space_args.total_spaces++; 427536523e95SDavid Sterba } 427636523e95SDavid Sterba 42772eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42787fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42797fde62bfSChris Mason 42807fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42817fde62bfSChris Mason ret = -EFAULT; 42827fde62bfSChris Mason 42837fde62bfSChris Mason kfree(dest_orig); 42847fde62bfSChris Mason out: 42857fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42861406e432SJosef Bacik ret = -EFAULT; 42871406e432SJosef Bacik 42881406e432SJosef Bacik return ret; 42891406e432SJosef Bacik } 42901406e432SJosef Bacik 42919a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42929a8c28beSMiao Xie void __user *argp) 429346204592SSage Weil { 429446204592SSage Weil struct btrfs_trans_handle *trans; 429546204592SSage Weil u64 transid; 4296db5b493aSTsutomu Itoh int ret; 429746204592SSage Weil 4298d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4299ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4300ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 430198d5dc13STsutomu Itoh return PTR_ERR(trans); 4302ff7c1d33SMiao Xie 4303ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4304ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4305ff7c1d33SMiao Xie goto out; 4306ff7c1d33SMiao Xie } 430746204592SSage Weil transid = trans->transid; 43083a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 43098b2b2d3cSTsutomu Itoh if (ret) { 43103a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4311db5b493aSTsutomu Itoh return ret; 43128b2b2d3cSTsutomu Itoh } 4313ff7c1d33SMiao Xie out: 431446204592SSage Weil if (argp) 431546204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 431646204592SSage Weil return -EFAULT; 431746204592SSage Weil return 0; 431846204592SSage Weil } 431946204592SSage Weil 43202ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 43219a8c28beSMiao Xie void __user *argp) 432246204592SSage Weil { 432346204592SSage Weil u64 transid; 432446204592SSage Weil 432546204592SSage Weil if (argp) { 432646204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 432746204592SSage Weil return -EFAULT; 432846204592SSage Weil } else { 432946204592SSage Weil transid = 0; /* current trans */ 433046204592SSage Weil } 43312ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 433246204592SSage Weil } 433346204592SSage Weil 4334b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4335475f6387SJan Schmidt { 43360b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4337475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4338b8e95489SMiao Xie int ret; 4339475f6387SJan Schmidt 4340475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4341475f6387SJan Schmidt return -EPERM; 4342475f6387SJan Schmidt 4343475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4344475f6387SJan Schmidt if (IS_ERR(sa)) 4345475f6387SJan Schmidt return PTR_ERR(sa); 4346475f6387SJan Schmidt 4347b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4348b8e95489SMiao Xie ret = mnt_want_write_file(file); 4349b8e95489SMiao Xie if (ret) 4350b8e95489SMiao Xie goto out; 4351b8e95489SMiao Xie } 4352b8e95489SMiao Xie 43530b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 435463a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 435563a212abSStefan Behrens 0); 4356475f6387SJan Schmidt 43575afe6ce7SFilipe Manana /* 43585afe6ce7SFilipe Manana * Copy scrub args to user space even if btrfs_scrub_dev() returned an 43595afe6ce7SFilipe Manana * error. This is important as it allows user space to know how much 43605afe6ce7SFilipe Manana * progress scrub has done. For example, if scrub is canceled we get 43615afe6ce7SFilipe Manana * -ECANCELED from btrfs_scrub_dev() and return that error back to user 43625afe6ce7SFilipe Manana * space. Later user space can inspect the progress from the structure 43635afe6ce7SFilipe Manana * btrfs_ioctl_scrub_args and resume scrub from where it left off 43645afe6ce7SFilipe Manana * previously (btrfs-progs does this). 43655afe6ce7SFilipe Manana * If we fail to copy the btrfs_ioctl_scrub_args structure to user space 43665afe6ce7SFilipe Manana * then return -EFAULT to signal the structure was not copied or it may 43675afe6ce7SFilipe Manana * be corrupt and unreliable due to a partial copy. 43685afe6ce7SFilipe Manana */ 43695afe6ce7SFilipe Manana if (copy_to_user(arg, sa, sizeof(*sa))) 4370475f6387SJan Schmidt ret = -EFAULT; 4371475f6387SJan Schmidt 4372b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4373b8e95489SMiao Xie mnt_drop_write_file(file); 4374b8e95489SMiao Xie out: 4375475f6387SJan Schmidt kfree(sa); 4376475f6387SJan Schmidt return ret; 4377475f6387SJan Schmidt } 4378475f6387SJan Schmidt 43792ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4380475f6387SJan Schmidt { 4381475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4382475f6387SJan Schmidt return -EPERM; 4383475f6387SJan Schmidt 43842ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4385475f6387SJan Schmidt } 4386475f6387SJan Schmidt 43872ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4388475f6387SJan Schmidt void __user *arg) 4389475f6387SJan Schmidt { 4390475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4391475f6387SJan Schmidt int ret; 4392475f6387SJan Schmidt 4393475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4394475f6387SJan Schmidt return -EPERM; 4395475f6387SJan Schmidt 4396475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4397475f6387SJan Schmidt if (IS_ERR(sa)) 4398475f6387SJan Schmidt return PTR_ERR(sa); 4399475f6387SJan Schmidt 44002ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4401475f6387SJan Schmidt 44024fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4403475f6387SJan Schmidt ret = -EFAULT; 4404475f6387SJan Schmidt 4405475f6387SJan Schmidt kfree(sa); 4406475f6387SJan Schmidt return ret; 4407475f6387SJan Schmidt } 4408475f6387SJan Schmidt 44092ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4410b27f7c0cSDavid Sterba void __user *arg) 4411c11d2c23SStefan Behrens { 4412c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4413c11d2c23SStefan Behrens int ret; 4414c11d2c23SStefan Behrens 4415c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4416c11d2c23SStefan Behrens if (IS_ERR(sa)) 4417c11d2c23SStefan Behrens return PTR_ERR(sa); 4418c11d2c23SStefan Behrens 4419b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4420b27f7c0cSDavid Sterba kfree(sa); 4421b27f7c0cSDavid Sterba return -EPERM; 4422b27f7c0cSDavid Sterba } 4423b27f7c0cSDavid Sterba 44242ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4425c11d2c23SStefan Behrens 4426eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4427c11d2c23SStefan Behrens ret = -EFAULT; 4428c11d2c23SStefan Behrens 4429c11d2c23SStefan Behrens kfree(sa); 4430c11d2c23SStefan Behrens return ret; 4431c11d2c23SStefan Behrens } 4432c11d2c23SStefan Behrens 44332ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 44342ff7e61eSJeff Mahoney void __user *arg) 44353f6bcfbdSStefan Behrens { 44363f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 44373f6bcfbdSStefan Behrens int ret; 44383f6bcfbdSStefan Behrens 44393f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44403f6bcfbdSStefan Behrens return -EPERM; 44413f6bcfbdSStefan Behrens 44423f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44433f6bcfbdSStefan Behrens if (IS_ERR(p)) 44443f6bcfbdSStefan Behrens return PTR_ERR(p); 44453f6bcfbdSStefan Behrens 44463f6bcfbdSStefan Behrens switch (p->cmd) { 44473f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4448bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4449adfa97cbSIlya Dryomov ret = -EROFS; 4450adfa97cbSIlya Dryomov goto out; 4451adfa97cbSIlya Dryomov } 4452171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4453e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44543f6bcfbdSStefan Behrens } else { 44552ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4456171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 44573f6bcfbdSStefan Behrens } 44583f6bcfbdSStefan Behrens break; 44593f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44600b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 44613f6bcfbdSStefan Behrens ret = 0; 44623f6bcfbdSStefan Behrens break; 44633f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 446417d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 446597282031SAnand Jain ret = 0; 44663f6bcfbdSStefan Behrens break; 44673f6bcfbdSStefan Behrens default: 44683f6bcfbdSStefan Behrens ret = -EINVAL; 44693f6bcfbdSStefan Behrens break; 44703f6bcfbdSStefan Behrens } 44713f6bcfbdSStefan Behrens 4472d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 44733f6bcfbdSStefan Behrens ret = -EFAULT; 4474adfa97cbSIlya Dryomov out: 44753f6bcfbdSStefan Behrens kfree(p); 44763f6bcfbdSStefan Behrens return ret; 44773f6bcfbdSStefan Behrens } 44783f6bcfbdSStefan Behrens 4479d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4480d7728c96SJan Schmidt { 4481d7728c96SJan Schmidt int ret = 0; 4482d7728c96SJan Schmidt int i; 4483740c3d22SChris Mason u64 rel_ptr; 4484d7728c96SJan Schmidt int size; 4485806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4486d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4487d7728c96SJan Schmidt struct btrfs_path *path; 4488d7728c96SJan Schmidt 448982b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4490d7728c96SJan Schmidt return -EPERM; 4491d7728c96SJan Schmidt 4492d7728c96SJan Schmidt path = btrfs_alloc_path(); 4493d7728c96SJan Schmidt if (!path) { 4494d7728c96SJan Schmidt ret = -ENOMEM; 4495d7728c96SJan Schmidt goto out; 4496d7728c96SJan Schmidt } 4497d7728c96SJan Schmidt 4498d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4499d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4500d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4501d7728c96SJan Schmidt ipa = NULL; 4502d7728c96SJan Schmidt goto out; 4503d7728c96SJan Schmidt } 4504d7728c96SJan Schmidt 4505d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4506d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4507d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4508d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4509d7728c96SJan Schmidt ipath = NULL; 4510d7728c96SJan Schmidt goto out; 4511d7728c96SJan Schmidt } 4512d7728c96SJan Schmidt 4513d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4514d7728c96SJan Schmidt if (ret < 0) 4515d7728c96SJan Schmidt goto out; 4516d7728c96SJan Schmidt 4517d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4518745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4519745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4520740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4521d7728c96SJan Schmidt } 4522d7728c96SJan Schmidt 4523718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4524718dc5faSOmar Sandoval ipath->fspath, size); 4525d7728c96SJan Schmidt if (ret) { 4526d7728c96SJan Schmidt ret = -EFAULT; 4527d7728c96SJan Schmidt goto out; 4528d7728c96SJan Schmidt } 4529d7728c96SJan Schmidt 4530d7728c96SJan Schmidt out: 4531d7728c96SJan Schmidt btrfs_free_path(path); 4532d7728c96SJan Schmidt free_ipath(ipath); 4533d7728c96SJan Schmidt kfree(ipa); 4534d7728c96SJan Schmidt 4535d7728c96SJan Schmidt return ret; 4536d7728c96SJan Schmidt } 4537d7728c96SJan Schmidt 4538d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4539d7728c96SJan Schmidt { 4540d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4541d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4542d7728c96SJan Schmidt 4543d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4544d7728c96SJan Schmidt inodes->bytes_left -= c; 4545d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4546d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4547d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4548d7728c96SJan Schmidt inodes->elem_cnt += 3; 4549d7728c96SJan Schmidt } else { 4550d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4551d7728c96SJan Schmidt inodes->bytes_left = 0; 4552d7728c96SJan Schmidt inodes->elem_missed += 3; 4553d7728c96SJan Schmidt } 4554d7728c96SJan Schmidt 4555d7728c96SJan Schmidt return 0; 4556d7728c96SJan Schmidt } 4557d7728c96SJan Schmidt 45582ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4559d24a67b2SZygo Blaxell void __user *arg, int version) 4560d7728c96SJan Schmidt { 4561d7728c96SJan Schmidt int ret = 0; 4562d7728c96SJan Schmidt int size; 4563d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4564d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4565d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4566d24a67b2SZygo Blaxell bool ignore_offset; 4567d7728c96SJan Schmidt 4568d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4569d7728c96SJan Schmidt return -EPERM; 4570d7728c96SJan Schmidt 4571d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 45727b9ea627SShailendra Verma if (IS_ERR(loi)) 45737b9ea627SShailendra Verma return PTR_ERR(loi); 4574d7728c96SJan Schmidt 4575d24a67b2SZygo Blaxell if (version == 1) { 4576d24a67b2SZygo Blaxell ignore_offset = false; 4577b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4578d24a67b2SZygo Blaxell } else { 4579d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4580d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4581d24a67b2SZygo Blaxell ret = -EINVAL; 4582d24a67b2SZygo Blaxell goto out_loi; 4583d24a67b2SZygo Blaxell } 4584d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4585d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4586d24a67b2SZygo Blaxell ret = -EINVAL; 4587d24a67b2SZygo Blaxell goto out_loi; 4588d24a67b2SZygo Blaxell } 4589d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4590b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4591d24a67b2SZygo Blaxell } 4592d24a67b2SZygo Blaxell 4593d7728c96SJan Schmidt path = btrfs_alloc_path(); 4594d7728c96SJan Schmidt if (!path) { 4595d7728c96SJan Schmidt ret = -ENOMEM; 4596d7728c96SJan Schmidt goto out; 4597d7728c96SJan Schmidt } 4598d7728c96SJan Schmidt 4599d7728c96SJan Schmidt inodes = init_data_container(size); 4600d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4601d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4602d7728c96SJan Schmidt inodes = NULL; 4603d7728c96SJan Schmidt goto out; 4604d7728c96SJan Schmidt } 4605d7728c96SJan Schmidt 46062ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4607d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4608df031f07SLiu Bo if (ret == -EINVAL) 4609d7728c96SJan Schmidt ret = -ENOENT; 4610d7728c96SJan Schmidt if (ret < 0) 4611d7728c96SJan Schmidt goto out; 4612d7728c96SJan Schmidt 4613718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4614718dc5faSOmar Sandoval size); 4615d7728c96SJan Schmidt if (ret) 4616d7728c96SJan Schmidt ret = -EFAULT; 4617d7728c96SJan Schmidt 4618d7728c96SJan Schmidt out: 4619d7728c96SJan Schmidt btrfs_free_path(path); 4620f54de068SDavid Sterba kvfree(inodes); 4621d24a67b2SZygo Blaxell out_loi: 4622d7728c96SJan Schmidt kfree(loi); 4623d7728c96SJan Schmidt 4624d7728c96SJan Schmidt return ret; 4625d7728c96SJan Schmidt } 4626d7728c96SJan Schmidt 4627008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4628c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4629c9e9f97bSIlya Dryomov { 4630c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4631c9e9f97bSIlya Dryomov 4632c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4633c9e9f97bSIlya Dryomov 46343009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4635837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4636837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4637837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4638a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4639a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4640837d5b6eSIlya Dryomov 4641c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4642c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4643c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 464419a39dceSIlya Dryomov 464519a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 464619a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 464719a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4648c9e9f97bSIlya Dryomov } 4649c9e9f97bSIlya Dryomov 46509ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4651c9e9f97bSIlya Dryomov { 4652496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4653c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4654c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4655c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4656ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4657c9e9f97bSIlya Dryomov int ret; 4658c9e9f97bSIlya Dryomov 4659c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4660c9e9f97bSIlya Dryomov return -EPERM; 4661c9e9f97bSIlya Dryomov 4662e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46639ba1f6e4SLiu Bo if (ret) 46649ba1f6e4SLiu Bo return ret; 46659ba1f6e4SLiu Bo 4666ed0fb78fSIlya Dryomov again: 4667171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4668c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4669ed0fb78fSIlya Dryomov need_unlock = true; 4670ed0fb78fSIlya Dryomov goto locked; 4671ed0fb78fSIlya Dryomov } 4672ed0fb78fSIlya Dryomov 4673ed0fb78fSIlya Dryomov /* 467401327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4675ed0fb78fSIlya Dryomov * (1) some other op is running 4676ed0fb78fSIlya Dryomov * (2) balance is running 4677ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4678ed0fb78fSIlya Dryomov */ 4679ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4680ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4681ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 46823009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4683ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4684dccdb07bSDavid Sterba /* 4685dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4686dccdb07bSDavid Sterba * we'll reexamine the status again. 4687dccdb07bSDavid Sterba */ 4688ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4689ed0fb78fSIlya Dryomov 4690ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 46913009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4692ed0fb78fSIlya Dryomov /* this is (3) */ 4693ed0fb78fSIlya Dryomov need_unlock = false; 4694ed0fb78fSIlya Dryomov goto locked; 4695ed0fb78fSIlya Dryomov } 4696ed0fb78fSIlya Dryomov 4697ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4698ed0fb78fSIlya Dryomov goto again; 4699ed0fb78fSIlya Dryomov } else { 4700ed0fb78fSIlya Dryomov /* this is (2) */ 4701ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4702ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4703ed0fb78fSIlya Dryomov goto out; 4704ed0fb78fSIlya Dryomov } 4705ed0fb78fSIlya Dryomov } else { 4706ed0fb78fSIlya Dryomov /* this is (1) */ 4707ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4708e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4709ed0fb78fSIlya Dryomov goto out; 4710ed0fb78fSIlya Dryomov } 4711ed0fb78fSIlya Dryomov 4712ed0fb78fSIlya Dryomov locked: 4713171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4714c9e9f97bSIlya Dryomov 4715c9e9f97bSIlya Dryomov if (arg) { 4716c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4717c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4718c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4719ed0fb78fSIlya Dryomov goto out_unlock; 4720c9e9f97bSIlya Dryomov } 4721de322263SIlya Dryomov 4722de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4723de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4724de322263SIlya Dryomov ret = -ENOTCONN; 4725de322263SIlya Dryomov goto out_bargs; 4726de322263SIlya Dryomov } 4727de322263SIlya Dryomov 4728de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4729de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4730de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4731de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4732de322263SIlya Dryomov 4733de322263SIlya Dryomov goto do_balance; 4734de322263SIlya Dryomov } 4735c9e9f97bSIlya Dryomov } else { 4736c9e9f97bSIlya Dryomov bargs = NULL; 4737c9e9f97bSIlya Dryomov } 4738c9e9f97bSIlya Dryomov 4739ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4740837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4741837d5b6eSIlya Dryomov goto out_bargs; 4742837d5b6eSIlya Dryomov } 4743837d5b6eSIlya Dryomov 47448d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4745c9e9f97bSIlya Dryomov if (!bctl) { 4746c9e9f97bSIlya Dryomov ret = -ENOMEM; 4747c9e9f97bSIlya Dryomov goto out_bargs; 4748c9e9f97bSIlya Dryomov } 4749c9e9f97bSIlya Dryomov 4750c9e9f97bSIlya Dryomov if (arg) { 4751c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4752c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4753c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4754c9e9f97bSIlya Dryomov 4755c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4756f43ffb60SIlya Dryomov } else { 4757f43ffb60SIlya Dryomov /* balance everything - no filters */ 4758f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4759c9e9f97bSIlya Dryomov } 4760c9e9f97bSIlya Dryomov 47618eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47628eb93459SDavid Sterba ret = -EINVAL; 47630f89abf5SChristian Engelmayer goto out_bctl; 47648eb93459SDavid Sterba } 47658eb93459SDavid Sterba 4766de322263SIlya Dryomov do_balance: 4767c9e9f97bSIlya Dryomov /* 4768149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4769149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4770149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4771149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4772c9e9f97bSIlya Dryomov */ 4773ed0fb78fSIlya Dryomov need_unlock = false; 4774ed0fb78fSIlya Dryomov 47756fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 47760f89abf5SChristian Engelmayer bctl = NULL; 4777ed0fb78fSIlya Dryomov 4778d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4779c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4780c9e9f97bSIlya Dryomov ret = -EFAULT; 4781c9e9f97bSIlya Dryomov } 4782c9e9f97bSIlya Dryomov 47830f89abf5SChristian Engelmayer out_bctl: 47840f89abf5SChristian Engelmayer kfree(bctl); 4785c9e9f97bSIlya Dryomov out_bargs: 4786c9e9f97bSIlya Dryomov kfree(bargs); 4787ed0fb78fSIlya Dryomov out_unlock: 4788c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4789ed0fb78fSIlya Dryomov if (need_unlock) 4790171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4791ed0fb78fSIlya Dryomov out: 4792e54bfa31SLiu Bo mnt_drop_write_file(file); 4793c9e9f97bSIlya Dryomov return ret; 4794c9e9f97bSIlya Dryomov } 4795c9e9f97bSIlya Dryomov 47962ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4797837d5b6eSIlya Dryomov { 4798837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4799837d5b6eSIlya Dryomov return -EPERM; 4800837d5b6eSIlya Dryomov 4801837d5b6eSIlya Dryomov switch (cmd) { 4802837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 48030b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4804a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 48050b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4806837d5b6eSIlya Dryomov } 4807837d5b6eSIlya Dryomov 4808837d5b6eSIlya Dryomov return -EINVAL; 4809837d5b6eSIlya Dryomov } 4810837d5b6eSIlya Dryomov 48112ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 481219a39dceSIlya Dryomov void __user *arg) 481319a39dceSIlya Dryomov { 481419a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 481519a39dceSIlya Dryomov int ret = 0; 481619a39dceSIlya Dryomov 481719a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 481819a39dceSIlya Dryomov return -EPERM; 481919a39dceSIlya Dryomov 482019a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 482119a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 482219a39dceSIlya Dryomov ret = -ENOTCONN; 482319a39dceSIlya Dryomov goto out; 482419a39dceSIlya Dryomov } 482519a39dceSIlya Dryomov 48268d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 482719a39dceSIlya Dryomov if (!bargs) { 482819a39dceSIlya Dryomov ret = -ENOMEM; 482919a39dceSIlya Dryomov goto out; 483019a39dceSIlya Dryomov } 483119a39dceSIlya Dryomov 4832008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 483319a39dceSIlya Dryomov 483419a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 483519a39dceSIlya Dryomov ret = -EFAULT; 483619a39dceSIlya Dryomov 483719a39dceSIlya Dryomov kfree(bargs); 483819a39dceSIlya Dryomov out: 483919a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 484019a39dceSIlya Dryomov return ret; 484119a39dceSIlya Dryomov } 484219a39dceSIlya Dryomov 4843905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 48445d13a37bSArne Jansen { 48450b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48460b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48475d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 48485d13a37bSArne Jansen int ret; 48495d13a37bSArne Jansen 48505d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48515d13a37bSArne Jansen return -EPERM; 48525d13a37bSArne Jansen 4853905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4854905b0ddaSMiao Xie if (ret) 4855905b0ddaSMiao Xie return ret; 48565d13a37bSArne Jansen 48575d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4858905b0ddaSMiao Xie if (IS_ERR(sa)) { 4859905b0ddaSMiao Xie ret = PTR_ERR(sa); 4860905b0ddaSMiao Xie goto drop_write; 4861905b0ddaSMiao Xie } 48625d13a37bSArne Jansen 48630b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48645d13a37bSArne Jansen 48655d13a37bSArne Jansen switch (sa->cmd) { 48665d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4867340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 48685d13a37bSArne Jansen break; 48695d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4870340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 48715d13a37bSArne Jansen break; 48725d13a37bSArne Jansen default: 48735d13a37bSArne Jansen ret = -EINVAL; 48745d13a37bSArne Jansen break; 48755d13a37bSArne Jansen } 48765d13a37bSArne Jansen 48775d13a37bSArne Jansen kfree(sa); 48780b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4879905b0ddaSMiao Xie drop_write: 4880905b0ddaSMiao Xie mnt_drop_write_file(file); 48815d13a37bSArne Jansen return ret; 48825d13a37bSArne Jansen } 48835d13a37bSArne Jansen 4884905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48855d13a37bSArne Jansen { 48860b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48870b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48880b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48895d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48905d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48915d13a37bSArne Jansen int ret; 48925d13a37bSArne Jansen int err; 48935d13a37bSArne Jansen 48945d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48955d13a37bSArne Jansen return -EPERM; 48965d13a37bSArne Jansen 4897905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4898905b0ddaSMiao Xie if (ret) 4899905b0ddaSMiao Xie return ret; 49005d13a37bSArne Jansen 49015d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4902905b0ddaSMiao Xie if (IS_ERR(sa)) { 4903905b0ddaSMiao Xie ret = PTR_ERR(sa); 4904905b0ddaSMiao Xie goto drop_write; 4905905b0ddaSMiao Xie } 49065d13a37bSArne Jansen 49075d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49085d13a37bSArne Jansen if (IS_ERR(trans)) { 49095d13a37bSArne Jansen ret = PTR_ERR(trans); 49105d13a37bSArne Jansen goto out; 49115d13a37bSArne Jansen } 49125d13a37bSArne Jansen 49135d13a37bSArne Jansen if (sa->assign) { 49149f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 49155d13a37bSArne Jansen } else { 491639616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 49175d13a37bSArne Jansen } 49185d13a37bSArne Jansen 4919e082f563SQu Wenruo /* update qgroup status and info */ 4920280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4921e082f563SQu Wenruo if (err < 0) 49220b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4923ad8403dfSAnand Jain "failed to update qgroup status and info"); 49243a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49255d13a37bSArne Jansen if (err && !ret) 49265d13a37bSArne Jansen ret = err; 49275d13a37bSArne Jansen 49285d13a37bSArne Jansen out: 49295d13a37bSArne Jansen kfree(sa); 4930905b0ddaSMiao Xie drop_write: 4931905b0ddaSMiao Xie mnt_drop_write_file(file); 49325d13a37bSArne Jansen return ret; 49335d13a37bSArne Jansen } 49345d13a37bSArne Jansen 4935905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 49365d13a37bSArne Jansen { 49370b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49380b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49395d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49405d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49415d13a37bSArne Jansen int ret; 49425d13a37bSArne Jansen int err; 49435d13a37bSArne Jansen 49445d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49455d13a37bSArne Jansen return -EPERM; 49465d13a37bSArne Jansen 4947905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4948905b0ddaSMiao Xie if (ret) 4949905b0ddaSMiao Xie return ret; 49505d13a37bSArne Jansen 49515d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4952905b0ddaSMiao Xie if (IS_ERR(sa)) { 4953905b0ddaSMiao Xie ret = PTR_ERR(sa); 4954905b0ddaSMiao Xie goto drop_write; 4955905b0ddaSMiao Xie } 49565d13a37bSArne Jansen 4957d86e56cfSMiao Xie if (!sa->qgroupid) { 4958d86e56cfSMiao Xie ret = -EINVAL; 4959d86e56cfSMiao Xie goto out; 4960d86e56cfSMiao Xie } 4961d86e56cfSMiao Xie 49625d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49635d13a37bSArne Jansen if (IS_ERR(trans)) { 49645d13a37bSArne Jansen ret = PTR_ERR(trans); 49655d13a37bSArne Jansen goto out; 49665d13a37bSArne Jansen } 49675d13a37bSArne Jansen 49685d13a37bSArne Jansen if (sa->create) { 496949a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 49705d13a37bSArne Jansen } else { 49713efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 49725d13a37bSArne Jansen } 49735d13a37bSArne Jansen 49743a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49755d13a37bSArne Jansen if (err && !ret) 49765d13a37bSArne Jansen ret = err; 49775d13a37bSArne Jansen 49785d13a37bSArne Jansen out: 49795d13a37bSArne Jansen kfree(sa); 4980905b0ddaSMiao Xie drop_write: 4981905b0ddaSMiao Xie mnt_drop_write_file(file); 49825d13a37bSArne Jansen return ret; 49835d13a37bSArne Jansen } 49845d13a37bSArne Jansen 4985905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49865d13a37bSArne Jansen { 49870b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49880b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49895d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49905d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49915d13a37bSArne Jansen int ret; 49925d13a37bSArne Jansen int err; 49935d13a37bSArne Jansen u64 qgroupid; 49945d13a37bSArne Jansen 49955d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49965d13a37bSArne Jansen return -EPERM; 49975d13a37bSArne Jansen 4998905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4999905b0ddaSMiao Xie if (ret) 5000905b0ddaSMiao Xie return ret; 50015d13a37bSArne Jansen 50025d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5003905b0ddaSMiao Xie if (IS_ERR(sa)) { 5004905b0ddaSMiao Xie ret = PTR_ERR(sa); 5005905b0ddaSMiao Xie goto drop_write; 5006905b0ddaSMiao Xie } 50075d13a37bSArne Jansen 50085d13a37bSArne Jansen trans = btrfs_join_transaction(root); 50095d13a37bSArne Jansen if (IS_ERR(trans)) { 50105d13a37bSArne Jansen ret = PTR_ERR(trans); 50115d13a37bSArne Jansen goto out; 50125d13a37bSArne Jansen } 50135d13a37bSArne Jansen 50145d13a37bSArne Jansen qgroupid = sa->qgroupid; 50155d13a37bSArne Jansen if (!qgroupid) { 50165d13a37bSArne Jansen /* take the current subvol as qgroup */ 50175d13a37bSArne Jansen qgroupid = root->root_key.objectid; 50185d13a37bSArne Jansen } 50195d13a37bSArne Jansen 5020f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 50215d13a37bSArne Jansen 50223a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 50235d13a37bSArne Jansen if (err && !ret) 50245d13a37bSArne Jansen ret = err; 50255d13a37bSArne Jansen 50265d13a37bSArne Jansen out: 50275d13a37bSArne Jansen kfree(sa); 5028905b0ddaSMiao Xie drop_write: 5029905b0ddaSMiao Xie mnt_drop_write_file(file); 50305d13a37bSArne Jansen return ret; 50315d13a37bSArne Jansen } 50325d13a37bSArne Jansen 50332f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 50342f232036SJan Schmidt { 50350b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50360b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50372f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50382f232036SJan Schmidt int ret; 50392f232036SJan Schmidt 50402f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50412f232036SJan Schmidt return -EPERM; 50422f232036SJan Schmidt 50432f232036SJan Schmidt ret = mnt_want_write_file(file); 50442f232036SJan Schmidt if (ret) 50452f232036SJan Schmidt return ret; 50462f232036SJan Schmidt 50472f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50482f232036SJan Schmidt if (IS_ERR(qsa)) { 50492f232036SJan Schmidt ret = PTR_ERR(qsa); 50502f232036SJan Schmidt goto drop_write; 50512f232036SJan Schmidt } 50522f232036SJan Schmidt 50532f232036SJan Schmidt if (qsa->flags) { 50542f232036SJan Schmidt ret = -EINVAL; 50552f232036SJan Schmidt goto out; 50562f232036SJan Schmidt } 50572f232036SJan Schmidt 50580b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 50592f232036SJan Schmidt 50602f232036SJan Schmidt out: 50612f232036SJan Schmidt kfree(qsa); 50622f232036SJan Schmidt drop_write: 50632f232036SJan Schmidt mnt_drop_write_file(file); 50642f232036SJan Schmidt return ret; 50652f232036SJan Schmidt } 50662f232036SJan Schmidt 5067b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, 5068b929c1d8SMarcos Paulo de Souza void __user *arg) 50692f232036SJan Schmidt { 50702f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50712f232036SJan Schmidt int ret = 0; 50722f232036SJan Schmidt 50732f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50742f232036SJan Schmidt return -EPERM; 50752f232036SJan Schmidt 50768d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 50772f232036SJan Schmidt if (!qsa) 50782f232036SJan Schmidt return -ENOMEM; 50792f232036SJan Schmidt 50800b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50812f232036SJan Schmidt qsa->flags = 1; 50820b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 50832f232036SJan Schmidt } 50842f232036SJan Schmidt 50852f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50862f232036SJan Schmidt ret = -EFAULT; 50872f232036SJan Schmidt 50882f232036SJan Schmidt kfree(qsa); 50892f232036SJan Schmidt return ret; 50902f232036SJan Schmidt } 50912f232036SJan Schmidt 5092b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info, 5093b929c1d8SMarcos Paulo de Souza void __user *arg) 509457254b6eSJan Schmidt { 509557254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 509657254b6eSJan Schmidt return -EPERM; 509757254b6eSJan Schmidt 50980b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 509957254b6eSJan Schmidt } 510057254b6eSJan Schmidt 5101abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5102abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 51038ea05e3aSAlexander Block { 5104496ad9aaSAl Viro struct inode *inode = file_inode(file); 51050b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51068ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 51078ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 51088ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 510995582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 51108ea05e3aSAlexander Block int ret = 0; 5111dd5f9615SStefan Behrens int received_uuid_changed; 51128ea05e3aSAlexander Block 5113bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5114bd60ea0fSDavid Sterba return -EPERM; 5115bd60ea0fSDavid Sterba 51168ea05e3aSAlexander Block ret = mnt_want_write_file(file); 51178ea05e3aSAlexander Block if (ret < 0) 51188ea05e3aSAlexander Block return ret; 51198ea05e3aSAlexander Block 51200b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51218ea05e3aSAlexander Block 51224a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 51238ea05e3aSAlexander Block ret = -EINVAL; 51248ea05e3aSAlexander Block goto out; 51258ea05e3aSAlexander Block } 51268ea05e3aSAlexander Block 51278ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 51288ea05e3aSAlexander Block ret = -EROFS; 51298ea05e3aSAlexander Block goto out; 51308ea05e3aSAlexander Block } 51318ea05e3aSAlexander Block 5132dd5f9615SStefan Behrens /* 5133dd5f9615SStefan Behrens * 1 - root item 5134dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5135dd5f9615SStefan Behrens */ 5136dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51378ea05e3aSAlexander Block if (IS_ERR(trans)) { 51388ea05e3aSAlexander Block ret = PTR_ERR(trans); 51398ea05e3aSAlexander Block trans = NULL; 51408ea05e3aSAlexander Block goto out; 51418ea05e3aSAlexander Block } 51428ea05e3aSAlexander Block 51438ea05e3aSAlexander Block sa->rtransid = trans->transid; 51448ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51458ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51468ea05e3aSAlexander Block 5147dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5148dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5149dd5f9615SStefan Behrens if (received_uuid_changed && 5150d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5151d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5152dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5153dd5f9615SStefan Behrens root->root_key.objectid); 5154d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5155d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5156d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5157d87ff758SNikolay Borisov goto out; 5158d87ff758SNikolay Borisov } 5159d87ff758SNikolay Borisov } 51608ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51618ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51628ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51633cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51643cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51653cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51663cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51678ea05e3aSAlexander Block 51680b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 51698ea05e3aSAlexander Block &root->root_key, &root->root_item); 51708ea05e3aSAlexander Block if (ret < 0) { 51713a45bb20SJeff Mahoney btrfs_end_transaction(trans); 51728ea05e3aSAlexander Block goto out; 5173dd5f9615SStefan Behrens } 5174dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5175cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5176dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5177dd5f9615SStefan Behrens root->root_key.objectid); 5178dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 517966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5180efd38150SNikolay Borisov btrfs_end_transaction(trans); 5181dd5f9615SStefan Behrens goto out; 5182dd5f9615SStefan Behrens } 5183dd5f9615SStefan Behrens } 51843a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5185abccd00fSHugo Mills out: 51860b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5187abccd00fSHugo Mills mnt_drop_write_file(file); 5188abccd00fSHugo Mills return ret; 5189abccd00fSHugo Mills } 5190abccd00fSHugo Mills 5191abccd00fSHugo Mills #ifdef CONFIG_64BIT 5192abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5193abccd00fSHugo Mills void __user *arg) 5194abccd00fSHugo Mills { 5195abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5196abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5197abccd00fSHugo Mills int ret = 0; 5198abccd00fSHugo Mills 5199abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 52007b9ea627SShailendra Verma if (IS_ERR(args32)) 52017b9ea627SShailendra Verma return PTR_ERR(args32); 5202abccd00fSHugo Mills 52038d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 520484dbeb87SDan Carpenter if (!args64) { 520584dbeb87SDan Carpenter ret = -ENOMEM; 5206abccd00fSHugo Mills goto out; 5207abccd00fSHugo Mills } 5208abccd00fSHugo Mills 5209abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5210abccd00fSHugo Mills args64->stransid = args32->stransid; 5211abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5212abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5213abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5214abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5215abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5216abccd00fSHugo Mills args64->flags = args32->flags; 5217abccd00fSHugo Mills 5218abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5219abccd00fSHugo Mills if (ret) 5220abccd00fSHugo Mills goto out; 5221abccd00fSHugo Mills 5222abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5223abccd00fSHugo Mills args32->stransid = args64->stransid; 5224abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5225abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5226abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5227abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5228abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5229abccd00fSHugo Mills args32->flags = args64->flags; 5230abccd00fSHugo Mills 5231abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5232abccd00fSHugo Mills if (ret) 5233abccd00fSHugo Mills ret = -EFAULT; 5234abccd00fSHugo Mills 5235abccd00fSHugo Mills out: 5236abccd00fSHugo Mills kfree(args32); 5237abccd00fSHugo Mills kfree(args64); 5238abccd00fSHugo Mills return ret; 5239abccd00fSHugo Mills } 5240abccd00fSHugo Mills #endif 5241abccd00fSHugo Mills 5242abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5243abccd00fSHugo Mills void __user *arg) 5244abccd00fSHugo Mills { 5245abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5246abccd00fSHugo Mills int ret = 0; 5247abccd00fSHugo Mills 5248abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 52497b9ea627SShailendra Verma if (IS_ERR(sa)) 52507b9ea627SShailendra Verma return PTR_ERR(sa); 5251abccd00fSHugo Mills 5252abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5253abccd00fSHugo Mills 5254abccd00fSHugo Mills if (ret) 5255abccd00fSHugo Mills goto out; 5256abccd00fSHugo Mills 52578ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52588ea05e3aSAlexander Block if (ret) 52598ea05e3aSAlexander Block ret = -EFAULT; 52608ea05e3aSAlexander Block 52618ea05e3aSAlexander Block out: 52628ea05e3aSAlexander Block kfree(sa); 52638ea05e3aSAlexander Block return ret; 52648ea05e3aSAlexander Block } 52658ea05e3aSAlexander Block 5266b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, 5267b929c1d8SMarcos Paulo de Souza void __user *arg) 5268867ab667Sjeff.liu { 5269a1b83ac5SAnand Jain size_t len; 5270867ab667Sjeff.liu int ret; 5271a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5272a1b83ac5SAnand Jain 52730b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 52740b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 52750b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5276a1b83ac5SAnand Jain 5277a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5278867ab667Sjeff.liu 5279867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 52800b246afaSJeff Mahoney btrfs_warn(fs_info, 52810b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 52820b246afaSJeff Mahoney --len); 5283867ab667Sjeff.liu } 5284867ab667Sjeff.liu 5285867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5286867ab667Sjeff.liu 5287867ab667Sjeff.liu return ret ? -EFAULT : 0; 5288867ab667Sjeff.liu } 5289867ab667Sjeff.liu 5290a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5291a8bfd4abSjeff.liu { 52920b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52940b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52950b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5296a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5297a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5298a8bfd4abSjeff.liu int ret; 5299a8bfd4abSjeff.liu 5300a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5301a8bfd4abSjeff.liu return -EPERM; 5302a8bfd4abSjeff.liu 5303a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5304a8bfd4abSjeff.liu return -EFAULT; 5305a8bfd4abSjeff.liu 5306a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 53070b246afaSJeff Mahoney btrfs_err(fs_info, 53085d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5309a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5310a8bfd4abSjeff.liu return -EINVAL; 5311a8bfd4abSjeff.liu } 5312a8bfd4abSjeff.liu 5313a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5314a8bfd4abSjeff.liu if (ret) 5315a8bfd4abSjeff.liu return ret; 5316a8bfd4abSjeff.liu 5317a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5318a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5319a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5320a8bfd4abSjeff.liu goto out_unlock; 5321a8bfd4abSjeff.liu } 5322a8bfd4abSjeff.liu 53230b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5324a8bfd4abSjeff.liu strcpy(super_block->label, label); 53250b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53263a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5327a8bfd4abSjeff.liu 5328a8bfd4abSjeff.liu out_unlock: 5329a8bfd4abSjeff.liu mnt_drop_write_file(file); 5330a8bfd4abSjeff.liu return ret; 5331a8bfd4abSjeff.liu } 5332a8bfd4abSjeff.liu 53332eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 53342eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 53352eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53362eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53372eaa055fSJeff Mahoney 5338d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53392eaa055fSJeff Mahoney { 53404d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53412eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53422eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53432eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53442eaa055fSJeff Mahoney }; 53452eaa055fSJeff Mahoney 53462eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53472eaa055fSJeff Mahoney return -EFAULT; 53482eaa055fSJeff Mahoney 53492eaa055fSJeff Mahoney return 0; 53502eaa055fSJeff Mahoney } 53512eaa055fSJeff Mahoney 5352b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, 5353b929c1d8SMarcos Paulo de Souza void __user *arg) 53542eaa055fSJeff Mahoney { 53550b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53562eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53572eaa055fSJeff Mahoney 53582eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53592eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53602eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53612eaa055fSJeff Mahoney 53622eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53632eaa055fSJeff Mahoney return -EFAULT; 53642eaa055fSJeff Mahoney 53652eaa055fSJeff Mahoney return 0; 53662eaa055fSJeff Mahoney } 53672eaa055fSJeff Mahoney 53682ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 53693b02a68aSJeff Mahoney enum btrfs_feature_set set, 53702eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53712eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53722eaa055fSJeff Mahoney { 5373f10152bcSDavid Sterba const char *type = btrfs_feature_set_name(set); 53743b02a68aSJeff Mahoney char *names; 53752eaa055fSJeff Mahoney u64 disallowed, unsupported; 53762eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53772eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53782eaa055fSJeff Mahoney 53792eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53802eaa055fSJeff Mahoney if (unsupported) { 53813b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53823b02a68aSJeff Mahoney if (names) { 53830b246afaSJeff Mahoney btrfs_warn(fs_info, 53843b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53853b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53863b02a68aSJeff Mahoney kfree(names); 53873b02a68aSJeff Mahoney } else 53880b246afaSJeff Mahoney btrfs_warn(fs_info, 53892eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53902eaa055fSJeff Mahoney type, unsupported); 53912eaa055fSJeff Mahoney return -EOPNOTSUPP; 53922eaa055fSJeff Mahoney } 53932eaa055fSJeff Mahoney 53942eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53952eaa055fSJeff Mahoney if (disallowed) { 53963b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53973b02a68aSJeff Mahoney if (names) { 53980b246afaSJeff Mahoney btrfs_warn(fs_info, 53993b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 54003b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54013b02a68aSJeff Mahoney kfree(names); 54023b02a68aSJeff Mahoney } else 54030b246afaSJeff Mahoney btrfs_warn(fs_info, 54042eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 54052eaa055fSJeff Mahoney type, disallowed); 54062eaa055fSJeff Mahoney return -EPERM; 54072eaa055fSJeff Mahoney } 54082eaa055fSJeff Mahoney 54092eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 54102eaa055fSJeff Mahoney if (disallowed) { 54113b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54123b02a68aSJeff Mahoney if (names) { 54130b246afaSJeff Mahoney btrfs_warn(fs_info, 54143b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 54153b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54163b02a68aSJeff Mahoney kfree(names); 54173b02a68aSJeff Mahoney } else 54180b246afaSJeff Mahoney btrfs_warn(fs_info, 54192eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 54202eaa055fSJeff Mahoney type, disallowed); 54212eaa055fSJeff Mahoney return -EPERM; 54222eaa055fSJeff Mahoney } 54232eaa055fSJeff Mahoney 54242eaa055fSJeff Mahoney return 0; 54252eaa055fSJeff Mahoney } 54262eaa055fSJeff Mahoney 54272ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 54282ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 54292eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 54302eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 54312eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 54322eaa055fSJeff Mahoney 54332eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 54342eaa055fSJeff Mahoney { 54350b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54360b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54370b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54380b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 54392eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54402eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54412eaa055fSJeff Mahoney u64 newflags; 54422eaa055fSJeff Mahoney int ret; 54432eaa055fSJeff Mahoney 54442eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54452eaa055fSJeff Mahoney return -EPERM; 54462eaa055fSJeff Mahoney 54472eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54482eaa055fSJeff Mahoney return -EFAULT; 54492eaa055fSJeff Mahoney 54502eaa055fSJeff Mahoney /* Nothing to do */ 54512eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54522eaa055fSJeff Mahoney !flags[0].incompat_flags) 54532eaa055fSJeff Mahoney return 0; 54542eaa055fSJeff Mahoney 54552ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 54562eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54572eaa055fSJeff Mahoney if (ret) 54582eaa055fSJeff Mahoney return ret; 54592eaa055fSJeff Mahoney 54602ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 54612eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54622eaa055fSJeff Mahoney if (ret) 54632eaa055fSJeff Mahoney return ret; 54642eaa055fSJeff Mahoney 54652ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 54662eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54672eaa055fSJeff Mahoney if (ret) 54682eaa055fSJeff Mahoney return ret; 54692eaa055fSJeff Mahoney 54707ab19625SDavid Sterba ret = mnt_want_write_file(file); 54717ab19625SDavid Sterba if (ret) 54727ab19625SDavid Sterba return ret; 54737ab19625SDavid Sterba 54748051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54757ab19625SDavid Sterba if (IS_ERR(trans)) { 54767ab19625SDavid Sterba ret = PTR_ERR(trans); 54777ab19625SDavid Sterba goto out_drop_write; 54787ab19625SDavid Sterba } 54792eaa055fSJeff Mahoney 54800b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 54812eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54822eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54832eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54842eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54852eaa055fSJeff Mahoney 54862eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54872eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54882eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54892eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54902eaa055fSJeff Mahoney 54912eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54922eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54932eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54942eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54950b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 54962eaa055fSJeff Mahoney 54973a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 54987ab19625SDavid Sterba out_drop_write: 54997ab19625SDavid Sterba mnt_drop_write_file(file); 55007ab19625SDavid Sterba 55017ab19625SDavid Sterba return ret; 55022eaa055fSJeff Mahoney } 55032eaa055fSJeff Mahoney 55042351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 55052351f431SJosef Bacik { 55062351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 55072351f431SJosef Bacik int ret; 55082351f431SJosef Bacik 55092351f431SJosef Bacik if (compat) { 55102351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55112351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 55122351f431SJosef Bacik 55132351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 55142351f431SJosef Bacik if (ret) 55152351f431SJosef Bacik return -EFAULT; 55162351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 55172351f431SJosef Bacik if (!arg) 55182351f431SJosef Bacik return -ENOMEM; 55192351f431SJosef Bacik arg->send_fd = args32.send_fd; 55202351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 55212351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 55222351f431SJosef Bacik arg->parent_root = args32.parent_root; 55232351f431SJosef Bacik arg->flags = args32.flags; 55242351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 55252351f431SJosef Bacik sizeof(args32.reserved)); 55262351f431SJosef Bacik #else 55272351f431SJosef Bacik return -ENOTTY; 55282351f431SJosef Bacik #endif 55292351f431SJosef Bacik } else { 55302351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 55312351f431SJosef Bacik if (IS_ERR(arg)) 55322351f431SJosef Bacik return PTR_ERR(arg); 55332351f431SJosef Bacik } 55342351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 55352351f431SJosef Bacik kfree(arg); 55362351f431SJosef Bacik return ret; 55372351f431SJosef Bacik } 55382351f431SJosef Bacik 5539f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5540f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5541f46b5a66SChristoph Hellwig { 55420b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55430b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55440b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55454bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5546f46b5a66SChristoph Hellwig 5547f46b5a66SChristoph Hellwig switch (cmd) { 55486cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 55496cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 55506cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 55516cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 55526cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 55536cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 555440cf931fSEric Sandeen case FS_IOC_GETFSLABEL: 5555b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_fslabel(fs_info, argp); 555640cf931fSEric Sandeen case FS_IOC_SETFSLABEL: 555740cf931fSEric Sandeen return btrfs_ioctl_set_fslabel(file, argp); 5558f7039b1dSLi Dongyang case FITRIM: 5559b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_fitrim(fs_info, argp); 5560f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5561fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5562fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5563fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 55643de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5565fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 55666f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 55676f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 556876dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 5569949964c9SMarcos Paulo de Souza return btrfs_ioctl_snap_destroy(file, argp, false); 5570949964c9SMarcos Paulo de Souza case BTRFS_IOC_SNAP_DESTROY_V2: 5571949964c9SMarcos Paulo de Souza return btrfs_ioctl_snap_destroy(file, argp, true); 55720caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 55730caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 55740caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 55750caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55766ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55776ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5578f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55791e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55801e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55811e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5582f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5583198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5584f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55852ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5586f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5587da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55886b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55896b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5590475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 55912ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5592475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 55932ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5594f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55959ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5596ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5597ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5598cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5599cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5600ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5601ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5602d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5603d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5604d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5605d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5606d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5607d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 56081406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 56092ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 56109b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 56119b199859SFilipe David Borba Manana int ret; 56129b199859SFilipe David Borba Manana 561382b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 56149b199859SFilipe David Borba Manana if (ret) 56159b199859SFilipe David Borba Manana return ret; 56160b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 56172fad4e83SDavid Sterba /* 56182fad4e83SDavid Sterba * The transaction thread may want to do more work, 561901327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 56202fad4e83SDavid Sterba * processing uncleaned subvols. 56212fad4e83SDavid Sterba */ 56220b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 56239b199859SFilipe David Borba Manana return ret; 56249b199859SFilipe David Borba Manana } 562546204592SSage Weil case BTRFS_IOC_START_SYNC: 56269a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 562746204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 56282ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5629475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5630b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5631475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 56322ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5633475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 56342ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5635c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 56369ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5637837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 56382ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 563919a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 56402ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 56418ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 56428ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5643abccd00fSHugo Mills #ifdef CONFIG_64BIT 5644abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5645abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5646abccd00fSHugo Mills #endif 564731db9f7cSAlexander Block case BTRFS_IOC_SEND: 56482351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 56492351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 56502351f431SJosef Bacik case BTRFS_IOC_SEND_32: 56512351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 56522351f431SJosef Bacik #endif 5653c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 56542ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 56555d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5656905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 56575d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5658905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 56595d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5660905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 56615d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5662905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 56632f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 56642f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 56652f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 5666b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_status(fs_info, argp); 566757254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 5668b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_wait(fs_info, argp); 56693f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 56702ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 56712eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5672d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 56732eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 5674b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_features(fs_info, argp); 56752eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56762eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5677e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5678e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5679025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5680025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5681b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5682b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 568342e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 568442e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 568523d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 568623d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5687f46b5a66SChristoph Hellwig } 5688f46b5a66SChristoph Hellwig 5689f46b5a66SChristoph Hellwig return -ENOTTY; 5690f46b5a66SChristoph Hellwig } 56914c63c245SLuke Dashjr 56924c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56934c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56944c63c245SLuke Dashjr { 56952a362249SJeff Mahoney /* 56962a362249SJeff Mahoney * These all access 32-bit values anyway so no further 56972a362249SJeff Mahoney * handling is necessary. 56982a362249SJeff Mahoney */ 56994c63c245SLuke Dashjr switch (cmd) { 57004c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 57014c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 57024c63c245SLuke Dashjr break; 57034c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 57044c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 57054c63c245SLuke Dashjr break; 57064c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 57074c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 57084c63c245SLuke Dashjr break; 57094c63c245SLuke Dashjr } 57104c63c245SLuke Dashjr 57114c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 57124c63c245SLuke Dashjr } 57134c63c245SLuke Dashjr #endif 5714