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> 2997fc2977SMiklos Szeredi #include <linux/fileattr.h> 3014605409SBoris Burkov #include <linux/fsverity.h> 31f46b5a66SChristoph Hellwig #include "ctree.h" 32f46b5a66SChristoph Hellwig #include "disk-io.h" 33949964c9SMarcos Paulo de Souza #include "export.h" 34f46b5a66SChristoph Hellwig #include "transaction.h" 35f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 36f46b5a66SChristoph Hellwig #include "print-tree.h" 37f46b5a66SChristoph Hellwig #include "volumes.h" 38925baeddSChris Mason #include "locking.h" 39d7728c96SJan Schmidt #include "backref.h" 40606686eeSJosef Bacik #include "rcu-string.h" 4131db9f7cSAlexander Block #include "send.h" 423f6bcfbdSStefan Behrens #include "dev-replace.h" 4363541927SFilipe David Borba Manana #include "props.h" 443b02a68aSJeff Mahoney #include "sysfs.h" 45fcebe456SJosef Bacik #include "qgroup.h" 461ec9a1aeSFilipe Manana #include "tree-log.h" 47ebb8765bSAnand Jain #include "compression.h" 488719aaaeSJosef Bacik #include "space-info.h" 4986736342SJosef Bacik #include "delalloc-space.h" 50aac0023cSJosef Bacik #include "block-group.h" 5122b398eeSQu Wenruo #include "subpage.h" 52f46b5a66SChristoph Hellwig 53abccd00fSHugo Mills #ifdef CONFIG_64BIT 54abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 55abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 56abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 57abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 58abccd00fSHugo Mills */ 59abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 60abccd00fSHugo Mills __u64 sec; 61abccd00fSHugo Mills __u32 nsec; 62abccd00fSHugo Mills } __attribute__ ((__packed__)); 63abccd00fSHugo Mills 64abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 65abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 66abccd00fSHugo Mills __u64 stransid; /* in */ 67abccd00fSHugo Mills __u64 rtransid; /* out */ 68abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 69abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 70abccd00fSHugo Mills __u64 flags; /* in */ 71abccd00fSHugo Mills __u64 reserved[16]; /* in */ 72abccd00fSHugo Mills } __attribute__ ((__packed__)); 73abccd00fSHugo Mills 74abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 75abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 76abccd00fSHugo Mills #endif 77abccd00fSHugo Mills 782351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 792351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 802351f431SJosef Bacik __s64 send_fd; /* in */ 812351f431SJosef Bacik __u64 clone_sources_count; /* in */ 822351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 832351f431SJosef Bacik __u64 parent_root; /* in */ 842351f431SJosef Bacik __u64 flags; /* in */ 85e77fbf99SDavid Sterba __u32 version; /* in */ 86e77fbf99SDavid Sterba __u8 reserved[28]; /* in */ 872351f431SJosef Bacik } __attribute__ ((__packed__)); 882351f431SJosef Bacik 892351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 902351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 912351f431SJosef Bacik #endif 92abccd00fSHugo Mills 936cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 941905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 951905a0f7SDavid Sterba unsigned int flags) 966cbff00fSChristoph Hellwig { 971905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 986cbff00fSChristoph Hellwig return flags; 991905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 1006cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1016cbff00fSChristoph Hellwig else 1026cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1036cbff00fSChristoph Hellwig } 104f46b5a66SChristoph Hellwig 1056cbff00fSChristoph Hellwig /* 106a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 107a157d4fdSDavid Sterba * ioctl. 1086cbff00fSChristoph Hellwig */ 10977eea05eSBoris Burkov static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode) 1106cbff00fSChristoph Hellwig { 1116cbff00fSChristoph Hellwig unsigned int iflags = 0; 11277eea05eSBoris Burkov u32 flags = binode->flags; 11314605409SBoris Burkov u32 ro_flags = binode->ro_flags; 1146cbff00fSChristoph Hellwig 1156cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1166cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1176cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1186cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1196cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1206cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1216cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1226cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1236cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1246cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1256cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1266cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 127d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 128d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 12914605409SBoris Burkov if (ro_flags & BTRFS_INODE_RO_VERITY) 13014605409SBoris Burkov iflags |= FS_VERITY_FL; 131d0092bddSLi Zefan 13213f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 133d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 13413f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13513f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1366cbff00fSChristoph Hellwig 1376cbff00fSChristoph Hellwig return iflags; 1386cbff00fSChristoph Hellwig } 1396cbff00fSChristoph Hellwig 1406cbff00fSChristoph Hellwig /* 1416cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1426cbff00fSChristoph Hellwig */ 1437b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1446cbff00fSChristoph Hellwig { 1455c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1463cc79392SFilipe Manana unsigned int new_fl = 0; 1476cbff00fSChristoph Hellwig 1485c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1493cc79392SFilipe Manana new_fl |= S_SYNC; 1505c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1513cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1525c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1533cc79392SFilipe Manana new_fl |= S_APPEND; 1545c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1553cc79392SFilipe Manana new_fl |= S_NOATIME; 1565c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1573cc79392SFilipe Manana new_fl |= S_DIRSYNC; 15814605409SBoris Burkov if (binode->ro_flags & BTRFS_INODE_RO_VERITY) 15914605409SBoris Burkov new_fl |= S_VERITY; 1603cc79392SFilipe Manana 1613cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 16214605409SBoris Burkov S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC | 16314605409SBoris Burkov S_VERITY, new_fl); 1646cbff00fSChristoph Hellwig } 1656cbff00fSChristoph Hellwig 166f37c563bSDavid Sterba /* 167f37c563bSDavid Sterba * Check if @flags are a supported and valid set of FS_*_FL flags and that 168f37c563bSDavid Sterba * the old and new flags are not conflicting 169f37c563bSDavid Sterba */ 170f37c563bSDavid Sterba static int check_fsflags(unsigned int old_flags, unsigned int flags) 17175e7cb7fSLiu Bo { 17275e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17375e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17475e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 175e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 176e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17775e7cb7fSLiu Bo return -EOPNOTSUPP; 17875e7cb7fSLiu Bo 179f37c563bSDavid Sterba /* COMPR and NOCOMP on new/old are valid */ 18075e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 18175e7cb7fSLiu Bo return -EINVAL; 18275e7cb7fSLiu Bo 183f37c563bSDavid Sterba if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL)) 184f37c563bSDavid Sterba return -EINVAL; 185f37c563bSDavid Sterba 186f37c563bSDavid Sterba /* NOCOW and compression options are mutually exclusive */ 187f37c563bSDavid Sterba if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL))) 188f37c563bSDavid Sterba return -EINVAL; 189f37c563bSDavid Sterba if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL))) 190f37c563bSDavid Sterba return -EINVAL; 191f37c563bSDavid Sterba 19275e7cb7fSLiu Bo return 0; 19375e7cb7fSLiu Bo } 19475e7cb7fSLiu Bo 195d206e9c9SNaohiro Aota static int check_fsflags_compatible(struct btrfs_fs_info *fs_info, 196d206e9c9SNaohiro Aota unsigned int flags) 197d206e9c9SNaohiro Aota { 198d206e9c9SNaohiro Aota if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL)) 199d206e9c9SNaohiro Aota return -EPERM; 200d206e9c9SNaohiro Aota 201d206e9c9SNaohiro Aota return 0; 202d206e9c9SNaohiro Aota } 203d206e9c9SNaohiro Aota 20497fc2977SMiklos Szeredi /* 20597fc2977SMiklos Szeredi * Set flags/xflags from the internal inode flags. The remaining items of 20697fc2977SMiklos Szeredi * fsxattr are zeroed. 20797fc2977SMiklos Szeredi */ 20897fc2977SMiklos Szeredi int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 2096cbff00fSChristoph Hellwig { 21097fc2977SMiklos Szeredi struct btrfs_inode *binode = BTRFS_I(d_inode(dentry)); 21197fc2977SMiklos Szeredi 21277eea05eSBoris Burkov fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode)); 21397fc2977SMiklos Szeredi return 0; 21497fc2977SMiklos Szeredi } 21597fc2977SMiklos Szeredi 21697fc2977SMiklos Szeredi int btrfs_fileattr_set(struct user_namespace *mnt_userns, 21797fc2977SMiklos Szeredi struct dentry *dentry, struct fileattr *fa) 21897fc2977SMiklos Szeredi { 21997fc2977SMiklos Szeredi struct inode *inode = d_inode(dentry); 2200b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2215c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 2225c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 2236cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 2245aca2842SDarrick J. Wong unsigned int fsflags, old_fsflags; 2256cbff00fSChristoph Hellwig int ret; 226ff9fef55SAnand Jain const char *comp = NULL; 227f37c563bSDavid Sterba u32 binode_flags; 2286cbff00fSChristoph Hellwig 229b83cc969SLi Zefan if (btrfs_root_readonly(root)) 230b83cc969SLi Zefan return -EROFS; 231b83cc969SLi Zefan 23297fc2977SMiklos Szeredi if (fileattr_has_fsx(fa)) 23397fc2977SMiklos Szeredi return -EOPNOTSUPP; 2346cbff00fSChristoph Hellwig 23597fc2977SMiklos Szeredi fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags); 23677eea05eSBoris Burkov old_fsflags = btrfs_inode_flags_to_fsflags(binode); 23797fc2977SMiklos Szeredi ret = check_fsflags(old_fsflags, fsflags); 238e7848683SJan Kara if (ret) 239e7848683SJan Kara return ret; 240e7848683SJan Kara 241d206e9c9SNaohiro Aota ret = check_fsflags_compatible(fs_info, fsflags); 242d206e9c9SNaohiro Aota if (ret) 24397fc2977SMiklos Szeredi return ret; 244d206e9c9SNaohiro Aota 245f37c563bSDavid Sterba binode_flags = binode->flags; 2465c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 247d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_SYNC; 2486cbff00fSChristoph Hellwig else 249d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_SYNC; 2505c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 251d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_IMMUTABLE; 2526cbff00fSChristoph Hellwig else 253d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_IMMUTABLE; 2545c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 255d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_APPEND; 2566cbff00fSChristoph Hellwig else 257d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_APPEND; 2585c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 259d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODUMP; 2606cbff00fSChristoph Hellwig else 261d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODUMP; 2625c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 263d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOATIME; 2646cbff00fSChristoph Hellwig else 265d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOATIME; 26697fc2977SMiklos Szeredi 26797fc2977SMiklos Szeredi /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */ 26897fc2977SMiklos Szeredi if (!fa->flags_valid) { 26997fc2977SMiklos Szeredi /* 1 item for the inode */ 27097fc2977SMiklos Szeredi trans = btrfs_start_transaction(root, 1); 2719b8a233bSRitesh Harjani if (IS_ERR(trans)) 2729b8a233bSRitesh Harjani return PTR_ERR(trans); 27397fc2977SMiklos Szeredi goto update_flags; 27497fc2977SMiklos Szeredi } 27597fc2977SMiklos Szeredi 2765c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 277d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_DIRSYNC; 2786cbff00fSChristoph Hellwig else 279d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_DIRSYNC; 2805c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 28144e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2827e97b8daSDavid Sterba /* 2837e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2847e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2857e97b8daSDavid Sterba * status of the file and will not set it. 2867e97b8daSDavid Sterba */ 2877e97b8daSDavid Sterba if (inode->i_size == 0) 288d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW | 289d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM; 2907e97b8daSDavid Sterba } else { 291d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW; 2927e97b8daSDavid Sterba } 2937e97b8daSDavid Sterba } else { 2947e97b8daSDavid Sterba /* 29501327610SNicholas D Steeves * Revert back under same assumptions as above 2967e97b8daSDavid Sterba */ 29744e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2987e97b8daSDavid Sterba if (inode->i_size == 0) 299d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_NODATACOW | 300d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM); 3017e97b8daSDavid Sterba } else { 302d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODATACOW; 3037e97b8daSDavid Sterba } 3047e97b8daSDavid Sterba } 3056cbff00fSChristoph Hellwig 30675e7cb7fSLiu Bo /* 30775e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 30875e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 30975e7cb7fSLiu Bo * things smaller. 31075e7cb7fSLiu Bo */ 3115c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 312d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_COMPRESS; 313d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOCOMPRESS; 3145c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 31563541927SFilipe David Borba Manana 31697fc2977SMiklos Szeredi if (IS_SWAPFILE(inode)) 31797fc2977SMiklos Szeredi return -ETXTBSY; 318eede2bf3SOmar Sandoval 319d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_COMPRESS; 320d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOCOMPRESS; 32163541927SFilipe David Borba Manana 32293370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 32393370509SDavid Sterba if (!comp || comp[0] == 0) 32493370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 325ebcb904dSLi Zefan } else { 326d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 32775e7cb7fSLiu Bo } 3286cbff00fSChristoph Hellwig 329ff9fef55SAnand Jain /* 330ff9fef55SAnand Jain * 1 for inode item 331ff9fef55SAnand Jain * 2 for properties 332ff9fef55SAnand Jain */ 333ff9fef55SAnand Jain trans = btrfs_start_transaction(root, 3); 33497fc2977SMiklos Szeredi if (IS_ERR(trans)) 33597fc2977SMiklos Szeredi return PTR_ERR(trans); 3366cbff00fSChristoph Hellwig 337ff9fef55SAnand Jain if (comp) { 338ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, 339ff9fef55SAnand Jain strlen(comp), 0); 340ff9fef55SAnand Jain if (ret) { 341ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 342ff9fef55SAnand Jain goto out_end_trans; 343ff9fef55SAnand Jain } 344ff9fef55SAnand Jain } else { 345ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 346ff9fef55SAnand Jain 0, 0); 347ff9fef55SAnand Jain if (ret && ret != -ENODATA) { 348ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 349ff9fef55SAnand Jain goto out_end_trans; 350ff9fef55SAnand Jain } 351ff9fef55SAnand Jain } 352ff9fef55SAnand Jain 35397fc2977SMiklos Szeredi update_flags: 354d2b8fcfeSAnand Jain binode->flags = binode_flags; 3557b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3560c4d2d95SJosef Bacik inode_inc_iversion(inode); 357c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3589a56fcd1SNikolay Borisov ret = btrfs_update_inode(trans, root, BTRFS_I(inode)); 3596cbff00fSChristoph Hellwig 360ff9fef55SAnand Jain out_end_trans: 3613a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3622d4e6f6aSliubo return ret; 3636cbff00fSChristoph Hellwig } 3646cbff00fSChristoph Hellwig 3650d7ed32cSDavid Sterba /* 3660d7ed32cSDavid Sterba * Start exclusive operation @type, return true on success 3670d7ed32cSDavid Sterba */ 368c3e1f96cSGoldwyn Rodrigues bool btrfs_exclop_start(struct btrfs_fs_info *fs_info, 369c3e1f96cSGoldwyn Rodrigues enum btrfs_exclusive_operation type) 370c3e1f96cSGoldwyn Rodrigues { 3710d7ed32cSDavid Sterba bool ret = false; 3720d7ed32cSDavid Sterba 3730d7ed32cSDavid Sterba spin_lock(&fs_info->super_lock); 3740d7ed32cSDavid Sterba if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) { 3750d7ed32cSDavid Sterba fs_info->exclusive_operation = type; 3760d7ed32cSDavid Sterba ret = true; 3770d7ed32cSDavid Sterba } 3780d7ed32cSDavid Sterba spin_unlock(&fs_info->super_lock); 3790d7ed32cSDavid Sterba 3800d7ed32cSDavid Sterba return ret; 381c3e1f96cSGoldwyn Rodrigues } 382c3e1f96cSGoldwyn Rodrigues 383578bda9eSDavid Sterba /* 384578bda9eSDavid Sterba * Conditionally allow to enter the exclusive operation in case it's compatible 385578bda9eSDavid Sterba * with the running one. This must be paired with btrfs_exclop_start_unlock and 386578bda9eSDavid Sterba * btrfs_exclop_finish. 387578bda9eSDavid Sterba * 388578bda9eSDavid Sterba * Compatibility: 389578bda9eSDavid Sterba * - the same type is already running 390578bda9eSDavid Sterba * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller 391578bda9eSDavid Sterba * must check the condition first that would allow none -> @type 392578bda9eSDavid Sterba */ 393578bda9eSDavid Sterba bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info, 394578bda9eSDavid Sterba enum btrfs_exclusive_operation type) 395578bda9eSDavid Sterba { 396578bda9eSDavid Sterba spin_lock(&fs_info->super_lock); 397578bda9eSDavid Sterba if (fs_info->exclusive_operation == type) 398578bda9eSDavid Sterba return true; 399578bda9eSDavid Sterba 400578bda9eSDavid Sterba spin_unlock(&fs_info->super_lock); 401578bda9eSDavid Sterba return false; 402578bda9eSDavid Sterba } 403578bda9eSDavid Sterba 404578bda9eSDavid Sterba void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info) 405578bda9eSDavid Sterba { 406578bda9eSDavid Sterba spin_unlock(&fs_info->super_lock); 407578bda9eSDavid Sterba } 408578bda9eSDavid Sterba 409c3e1f96cSGoldwyn Rodrigues void btrfs_exclop_finish(struct btrfs_fs_info *fs_info) 410c3e1f96cSGoldwyn Rodrigues { 4110d7ed32cSDavid Sterba spin_lock(&fs_info->super_lock); 412c3e1f96cSGoldwyn Rodrigues WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE); 4130d7ed32cSDavid Sterba spin_unlock(&fs_info->super_lock); 41466a2823cSGoldwyn Rodrigues sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation"); 415c3e1f96cSGoldwyn Rodrigues } 416c3e1f96cSGoldwyn Rodrigues 417*efc0e69cSNikolay Borisov void btrfs_exclop_balance(struct btrfs_fs_info *fs_info, 418*efc0e69cSNikolay Borisov enum btrfs_exclusive_operation op) 419*efc0e69cSNikolay Borisov { 420*efc0e69cSNikolay Borisov switch (op) { 421*efc0e69cSNikolay Borisov case BTRFS_EXCLOP_BALANCE_PAUSED: 422*efc0e69cSNikolay Borisov spin_lock(&fs_info->super_lock); 423*efc0e69cSNikolay Borisov ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE || 424*efc0e69cSNikolay Borisov fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD); 425*efc0e69cSNikolay Borisov fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED; 426*efc0e69cSNikolay Borisov spin_unlock(&fs_info->super_lock); 427*efc0e69cSNikolay Borisov break; 428*efc0e69cSNikolay Borisov case BTRFS_EXCLOP_BALANCE: 429*efc0e69cSNikolay Borisov spin_lock(&fs_info->super_lock); 430*efc0e69cSNikolay Borisov ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED); 431*efc0e69cSNikolay Borisov fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE; 432*efc0e69cSNikolay Borisov spin_unlock(&fs_info->super_lock); 433*efc0e69cSNikolay Borisov break; 434*efc0e69cSNikolay Borisov default: 435*efc0e69cSNikolay Borisov btrfs_warn(fs_info, 436*efc0e69cSNikolay Borisov "invalid exclop balance operation %d requested", op); 437*efc0e69cSNikolay Borisov } 438*efc0e69cSNikolay Borisov } 439*efc0e69cSNikolay Borisov 4406cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4416cbff00fSChristoph Hellwig { 442496ad9aaSAl Viro struct inode *inode = file_inode(file); 4436cbff00fSChristoph Hellwig 4446cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4456cbff00fSChristoph Hellwig } 446f46b5a66SChristoph Hellwig 447b929c1d8SMarcos Paulo de Souza static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, 448b929c1d8SMarcos Paulo de Souza void __user *arg) 449f7039b1dSLi Dongyang { 450f7039b1dSLi Dongyang struct btrfs_device *device; 451f7039b1dSLi Dongyang struct request_queue *q; 452f7039b1dSLi Dongyang struct fstrim_range range; 453f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 454f7039b1dSLi Dongyang u64 num_devices = 0; 455f7039b1dSLi Dongyang int ret; 456f7039b1dSLi Dongyang 457f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 458f7039b1dSLi Dongyang return -EPERM; 459f7039b1dSLi Dongyang 460f35f06c3SFilipe Manana /* 4611cb3dc3fSNaohiro Aota * btrfs_trim_block_group() depends on space cache, which is not 4621cb3dc3fSNaohiro Aota * available in zoned filesystem. So, disallow fitrim on a zoned 4631cb3dc3fSNaohiro Aota * filesystem for now. 4641cb3dc3fSNaohiro Aota */ 4651cb3dc3fSNaohiro Aota if (btrfs_is_zoned(fs_info)) 4661cb3dc3fSNaohiro Aota return -EOPNOTSUPP; 4671cb3dc3fSNaohiro Aota 4681cb3dc3fSNaohiro Aota /* 469f35f06c3SFilipe Manana * If the fs is mounted with nologreplay, which requires it to be 470f35f06c3SFilipe Manana * mounted in RO mode as well, we can not allow discard on free space 471f35f06c3SFilipe Manana * inside block groups, because log trees refer to extents that are not 472f35f06c3SFilipe Manana * pinned in a block group's free space cache (pinning the extents is 473f35f06c3SFilipe Manana * precisely the first phase of replaying a log tree). 474f35f06c3SFilipe Manana */ 475f35f06c3SFilipe Manana if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 476f35f06c3SFilipe Manana return -EROFS; 477f35f06c3SFilipe Manana 4781f78160cSXiao Guangrong rcu_read_lock(); 4791f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 4801f78160cSXiao Guangrong dev_list) { 481f7039b1dSLi Dongyang if (!device->bdev) 482f7039b1dSLi Dongyang continue; 483f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 484f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 485f7039b1dSLi Dongyang num_devices++; 48650d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 487f7039b1dSLi Dongyang minlen); 488f7039b1dSLi Dongyang } 489f7039b1dSLi Dongyang } 4901f78160cSXiao Guangrong rcu_read_unlock(); 491f4c697e6SLukas Czerner 492f7039b1dSLi Dongyang if (!num_devices) 493f7039b1dSLi Dongyang return -EOPNOTSUPP; 494f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 495f7039b1dSLi Dongyang return -EFAULT; 4966ba9fc8eSQu Wenruo 4976ba9fc8eSQu Wenruo /* 4986ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 4996ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 5006ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 5016ba9fc8eSQu Wenruo */ 5026ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 503f4c697e6SLukas Czerner return -EINVAL; 504f7039b1dSLi Dongyang 505f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5062ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 507f7039b1dSLi Dongyang if (ret < 0) 508f7039b1dSLi Dongyang return ret; 509f7039b1dSLi Dongyang 510f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 511f7039b1dSLi Dongyang return -EFAULT; 512f7039b1dSLi Dongyang 513f7039b1dSLi Dongyang return 0; 514f7039b1dSLi Dongyang } 515f7039b1dSLi Dongyang 516e1f60a65SDavid Sterba int __pure btrfs_is_empty_uuid(u8 *uuid) 517dd5f9615SStefan Behrens { 51846e0f66aSChris Mason int i; 51946e0f66aSChris Mason 52046e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 52146e0f66aSChris Mason if (uuid[i]) 52246e0f66aSChris Mason return 0; 52346e0f66aSChris Mason } 52446e0f66aSChris Mason return 1; 525dd5f9615SStefan Behrens } 526dd5f9615SStefan Behrens 5274d4340c9SChristian Brauner static noinline int create_subvol(struct user_namespace *mnt_userns, 5284d4340c9SChristian Brauner struct inode *dir, struct dentry *dentry, 52952f75f4fSDavid Sterba const char *name, int namelen, 5308696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 531f46b5a66SChristoph Hellwig { 5320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 533f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 534f46b5a66SChristoph Hellwig struct btrfs_key key; 53549a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 536f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 537f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 538d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 53976dda93cSYan, Zheng struct btrfs_root *new_root; 540d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 54195582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5425662344bSTsutomu Itoh struct inode *inode; 543f46b5a66SChristoph Hellwig int ret; 544f46b5a66SChristoph Hellwig int err; 5452dfb1e43SQu Wenruo dev_t anon_dev = 0; 546f46b5a66SChristoph Hellwig u64 objectid; 5473de4586cSChris Mason u64 index = 0; 548f46b5a66SChristoph Hellwig 54949a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 55049a3c4d9SDavid Sterba if (!root_item) 55149a3c4d9SDavid Sterba return -ENOMEM; 55249a3c4d9SDavid Sterba 553543068a2SNikolay Borisov ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid); 5542fbe8c8aSAl Viro if (ret) 55549a3c4d9SDavid Sterba goto fail_free; 5566a912213SJosef Bacik 5572dfb1e43SQu Wenruo ret = get_anon_bdev(&anon_dev); 5582dfb1e43SQu Wenruo if (ret < 0) 5592dfb1e43SQu Wenruo goto fail_free; 5602dfb1e43SQu Wenruo 561e09fe2d2SQu Wenruo /* 562e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 56301327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 564e09fe2d2SQu Wenruo */ 56549a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 56649a3c4d9SDavid Sterba ret = -ENOSPC; 56749a3c4d9SDavid Sterba goto fail_free; 56849a3c4d9SDavid Sterba } 569e09fe2d2SQu Wenruo 570d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 5719ed74f2dSJosef Bacik /* 572d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 573d5c12070SMiao Xie * of create_snapshot(). 5749ed74f2dSJosef Bacik */ 575c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 576d5c12070SMiao Xie if (ret) 57749a3c4d9SDavid Sterba goto fail_free; 578f46b5a66SChristoph Hellwig 579d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 580d5c12070SMiao Xie if (IS_ERR(trans)) { 581d5c12070SMiao Xie ret = PTR_ERR(trans); 582e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &block_rsv); 58349a3c4d9SDavid Sterba goto fail_free; 584d5c12070SMiao Xie } 585d5c12070SMiao Xie trans->block_rsv = &block_rsv; 586d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 587f46b5a66SChristoph Hellwig 588a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 5896f72c7e2SArne Jansen if (ret) 5906f72c7e2SArne Jansen goto fail; 5916f72c7e2SArne Jansen 5929631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 5939631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 5948e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 5958e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 5968e8a1e31SJosef Bacik goto fail; 5978e8a1e31SJosef Bacik } 598f46b5a66SChristoph Hellwig 599f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 600f46b5a66SChristoph Hellwig 60149a3c4d9SDavid Sterba inode_item = &root_item->inode; 6023cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6033cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6043cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 605da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6060b246afaSJeff Mahoney fs_info->nodesize); 6073cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 608f46b5a66SChristoph Hellwig 60949a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 61049a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6113cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 61208fe4db1SLi Zefan 61349a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 61449a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 61549a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 61649a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 61749a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 61849a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 619f46b5a66SChristoph Hellwig 62049a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 62149a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 622807fc790SAndy Shevchenko generate_random_guid(root_item->uuid); 62349a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 62449a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 62549a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 62649a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 62749a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 628f46b5a66SChristoph Hellwig 629925baeddSChris Mason btrfs_tree_unlock(leaf); 630f46b5a66SChristoph Hellwig 63169948022SNikolay Borisov btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID); 632f46b5a66SChristoph Hellwig 633f46b5a66SChristoph Hellwig key.objectid = objectid; 6345d4f98a2SYan Zheng key.offset = 0; 635962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6360b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 63749a3c4d9SDavid Sterba root_item); 63867addf29SFilipe Manana if (ret) { 63967addf29SFilipe Manana /* 64067addf29SFilipe Manana * Since we don't abort the transaction in this case, free the 64167addf29SFilipe Manana * tree block so that we don't leak space and leave the 64267addf29SFilipe Manana * filesystem in an inconsistent state (an extent item in the 6437a163608SFilipe Manana * extent tree with a backreference for a root that does not 644212a58fdSFilipe Manana * exists). 64567addf29SFilipe Manana */ 646212a58fdSFilipe Manana btrfs_tree_lock(leaf); 647212a58fdSFilipe Manana btrfs_clean_tree_block(leaf); 648212a58fdSFilipe Manana btrfs_tree_unlock(leaf); 6497a163608SFilipe Manana btrfs_free_tree_block(trans, objectid, leaf, 0, 1); 65067addf29SFilipe Manana free_extent_buffer(leaf); 651f46b5a66SChristoph Hellwig goto fail; 65267addf29SFilipe Manana } 65367addf29SFilipe Manana 65467addf29SFilipe Manana free_extent_buffer(leaf); 65567addf29SFilipe Manana leaf = NULL; 656f46b5a66SChristoph Hellwig 65776dda93cSYan, Zheng key.offset = (u64)-1; 6582dfb1e43SQu Wenruo new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev); 65979787eaaSJeff Mahoney if (IS_ERR(new_root)) { 6602dfb1e43SQu Wenruo free_anon_bdev(anon_dev); 66179787eaaSJeff Mahoney ret = PTR_ERR(new_root); 66266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 66379787eaaSJeff Mahoney goto fail; 66479787eaaSJeff Mahoney } 6652dfb1e43SQu Wenruo /* Freeing will be done in btrfs_put_root() of new_root */ 6662dfb1e43SQu Wenruo anon_dev = 0; 66776dda93cSYan, Zheng 668221581e4SJosef Bacik ret = btrfs_record_root_in_trans(trans, new_root); 669221581e4SJosef Bacik if (ret) { 670221581e4SJosef Bacik btrfs_put_root(new_root); 671221581e4SJosef Bacik btrfs_abort_transaction(trans, ret); 672221581e4SJosef Bacik goto fail; 673221581e4SJosef Bacik } 67476dda93cSYan, Zheng 6754d4340c9SChristian Brauner ret = btrfs_create_subvol_root(trans, new_root, root, mnt_userns); 67600246528SJosef Bacik btrfs_put_root(new_root); 677ce598979SMark Fasheh if (ret) { 678ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 67966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 680ce598979SMark Fasheh goto fail; 681ce598979SMark Fasheh } 682ce598979SMark Fasheh 683f46b5a66SChristoph Hellwig /* 684f46b5a66SChristoph Hellwig * insert the directory item 685f46b5a66SChristoph Hellwig */ 686877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 68779787eaaSJeff Mahoney if (ret) { 68866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 68979787eaaSJeff Mahoney goto fail; 69079787eaaSJeff Mahoney } 6913de4586cSChris Mason 692684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 6933de4586cSChris Mason BTRFS_FT_DIR, index); 69479787eaaSJeff Mahoney if (ret) { 69566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 696f46b5a66SChristoph Hellwig goto fail; 69779787eaaSJeff Mahoney } 6980660b5afSChris Mason 6996ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 7009a56fcd1SNikolay Borisov ret = btrfs_update_inode(trans, root, BTRFS_I(dir)); 701c7e54b51SJosef Bacik if (ret) { 702c7e54b51SJosef Bacik btrfs_abort_transaction(trans, ret); 703c7e54b51SJosef Bacik goto fail; 704c7e54b51SJosef Bacik } 70552c26179SYan Zheng 7066025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7074a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 708c7e54b51SJosef Bacik if (ret) { 709c7e54b51SJosef Bacik btrfs_abort_transaction(trans, ret); 710c7e54b51SJosef Bacik goto fail; 711c7e54b51SJosef Bacik } 7120660b5afSChris Mason 713cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7146bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 715dd5f9615SStefan Behrens if (ret) 71666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 717dd5f9615SStefan Behrens 718f46b5a66SChristoph Hellwig fail: 71949a3c4d9SDavid Sterba kfree(root_item); 720d5c12070SMiao Xie trans->block_rsv = NULL; 721d5c12070SMiao Xie trans->bytes_reserved = 0; 722e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &block_rsv); 723de6e8200SLiu Bo 7243a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 725f46b5a66SChristoph Hellwig if (err && !ret) 726f46b5a66SChristoph Hellwig ret = err; 7271a65e24bSChris Mason 7285662344bSTsutomu Itoh if (!ret) { 7295662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 730de6e8200SLiu Bo if (IS_ERR(inode)) 731de6e8200SLiu Bo return PTR_ERR(inode); 7325662344bSTsutomu Itoh d_instantiate(dentry, inode); 7335662344bSTsutomu Itoh } 734f46b5a66SChristoph Hellwig return ret; 73549a3c4d9SDavid Sterba 73649a3c4d9SDavid Sterba fail_free: 7372dfb1e43SQu Wenruo if (anon_dev) 7382dfb1e43SQu Wenruo free_anon_bdev(anon_dev); 73949a3c4d9SDavid Sterba kfree(root_item); 74049a3c4d9SDavid Sterba return ret; 741f46b5a66SChristoph Hellwig } 742f46b5a66SChristoph Hellwig 743e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 7449babda9fSNikolay Borisov struct dentry *dentry, bool readonly, 745e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 746f46b5a66SChristoph Hellwig { 7470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7482e4bfab9SYan, Zheng struct inode *inode; 749f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 750f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7512e4bfab9SYan, Zheng int ret; 752f46b5a66SChristoph Hellwig 75392a7cc42SQu Wenruo if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) 754f46b5a66SChristoph Hellwig return -EINVAL; 755f46b5a66SChristoph Hellwig 756eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 757eede2bf3SOmar Sandoval btrfs_warn(fs_info, 758eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 759eede2bf3SOmar Sandoval return -ETXTBSY; 760eede2bf3SOmar Sandoval } 761eede2bf3SOmar Sandoval 76223269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 763a1ee7362SDavid Sterba if (!pending_snapshot) 764a1ee7362SDavid Sterba return -ENOMEM; 765a1ee7362SDavid Sterba 7662dfb1e43SQu Wenruo ret = get_anon_bdev(&pending_snapshot->anon_dev); 7672dfb1e43SQu Wenruo if (ret < 0) 7682dfb1e43SQu Wenruo goto free_pending; 769b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 77023269bf5SDavid Sterba GFP_KERNEL); 7718546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7728546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 773b0c0ea63SDavid Sterba ret = -ENOMEM; 774b0c0ea63SDavid Sterba goto free_pending; 775b0c0ea63SDavid Sterba } 776b0c0ea63SDavid Sterba 77766d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 77866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 779d5c12070SMiao Xie /* 780d5c12070SMiao Xie * 1 - parent dir inode 781d5c12070SMiao Xie * 2 - dir entries 782d5c12070SMiao Xie * 1 - root item 783d5c12070SMiao Xie * 2 - root ref/backref 784d5c12070SMiao Xie * 1 - root of snapshot 785dd5f9615SStefan Behrens * 1 - UUID item 786d5c12070SMiao Xie */ 787d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 788dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 789ee3441b4SJeff Mahoney false); 790d5c12070SMiao Xie if (ret) 791c11fbb6eSRobbie Ko goto free_pending; 792d5c12070SMiao Xie 793a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 794a22285a6SYan, Zheng pending_snapshot->root = root; 795b83cc969SLi Zefan pending_snapshot->readonly = readonly; 796e9662f70SMiao Xie pending_snapshot->dir = dir; 7978696c533SMiao Xie pending_snapshot->inherit = inherit; 798a22285a6SYan, Zheng 799d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 800a22285a6SYan, Zheng if (IS_ERR(trans)) { 801a22285a6SYan, Zheng ret = PTR_ERR(trans); 802a22285a6SYan, Zheng goto fail; 803a22285a6SYan, Zheng } 804a22285a6SYan, Zheng 8050b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 806a22285a6SYan, Zheng list_add(&pending_snapshot->list, 807a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8080b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 8099babda9fSNikolay Borisov 8103a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 811aec8030aSMiao Xie if (ret) 812c37b2b62SJosef Bacik goto fail; 813a22285a6SYan, Zheng 814a22285a6SYan, Zheng ret = pending_snapshot->error; 815f46b5a66SChristoph Hellwig if (ret) 8162e4bfab9SYan, Zheng goto fail; 817f46b5a66SChristoph Hellwig 818d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 819d3797308SChris Mason if (ret) 820d3797308SChris Mason goto fail; 821d3797308SChris Mason 8222b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8232e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8242e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8252e4bfab9SYan, Zheng goto fail; 8262e4bfab9SYan, Zheng } 8275662344bSTsutomu Itoh 8282e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8292e4bfab9SYan, Zheng ret = 0; 8302dfb1e43SQu Wenruo pending_snapshot->anon_dev = 0; 8312e4bfab9SYan, Zheng fail: 8322dfb1e43SQu Wenruo /* Prevent double freeing of anon_dev */ 8332dfb1e43SQu Wenruo if (ret && pending_snapshot->snap) 8342dfb1e43SQu Wenruo pending_snapshot->snap->anon_dev = 0; 83500246528SJosef Bacik btrfs_put_root(pending_snapshot->snap); 836e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv); 837b0c0ea63SDavid Sterba free_pending: 8382dfb1e43SQu Wenruo if (pending_snapshot->anon_dev) 8392dfb1e43SQu Wenruo free_anon_bdev(pending_snapshot->anon_dev); 840b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8418546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 842a1ee7362SDavid Sterba kfree(pending_snapshot); 843a1ee7362SDavid Sterba 844f46b5a66SChristoph Hellwig return ret; 845f46b5a66SChristoph Hellwig } 846f46b5a66SChristoph Hellwig 8474260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8484260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8494260f7c7SSage Weil * whether the type of victim is right. 8504260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8514260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8524260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8534260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8544260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8554260f7c7SSage Weil * a. be owner of dir, or 8564260f7c7SSage Weil * b. be owner of victim, or 8574260f7c7SSage Weil * c. have CAP_FOWNER capability 85801327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8594260f7c7SSage Weil * links pointing to it. 8604260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8614260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8624260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8634260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8644260f7c7SSage Weil * nfs_async_unlink(). 8654260f7c7SSage Weil */ 8664260f7c7SSage Weil 867c4ed533bSChristian Brauner static int btrfs_may_delete(struct user_namespace *mnt_userns, 868c4ed533bSChristian Brauner struct inode *dir, struct dentry *victim, int isdir) 8694260f7c7SSage Weil { 8704260f7c7SSage Weil int error; 8714260f7c7SSage Weil 8722b0143b5SDavid Howells if (d_really_is_negative(victim)) 8734260f7c7SSage Weil return -ENOENT; 8744260f7c7SSage Weil 8752b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8764fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8774260f7c7SSage Weil 878c4ed533bSChristian Brauner error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); 8794260f7c7SSage Weil if (error) 8804260f7c7SSage Weil return error; 8814260f7c7SSage Weil if (IS_APPEND(dir)) 8824260f7c7SSage Weil return -EPERM; 883c4ed533bSChristian Brauner if (check_sticky(mnt_userns, dir, d_inode(victim)) || 884ba73d987SChristian Brauner IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) || 885ba73d987SChristian Brauner IS_SWAPFILE(d_inode(victim))) 8864260f7c7SSage Weil return -EPERM; 8874260f7c7SSage Weil if (isdir) { 888e36cb0b8SDavid Howells if (!d_is_dir(victim)) 8894260f7c7SSage Weil return -ENOTDIR; 8904260f7c7SSage Weil if (IS_ROOT(victim)) 8914260f7c7SSage Weil return -EBUSY; 892e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 8934260f7c7SSage Weil return -EISDIR; 8944260f7c7SSage Weil if (IS_DEADDIR(dir)) 8954260f7c7SSage Weil return -ENOENT; 8964260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 8974260f7c7SSage Weil return -EBUSY; 8984260f7c7SSage Weil return 0; 8994260f7c7SSage Weil } 9004260f7c7SSage Weil 901cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 9024d4340c9SChristian Brauner static inline int btrfs_may_create(struct user_namespace *mnt_userns, 9034d4340c9SChristian Brauner struct inode *dir, struct dentry *child) 904cb8e7090SChristoph Hellwig { 9052b0143b5SDavid Howells if (d_really_is_positive(child)) 906cb8e7090SChristoph Hellwig return -EEXIST; 907cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 908cb8e7090SChristoph Hellwig return -ENOENT; 9094d4340c9SChristian Brauner if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns)) 9105474bf40SChristian Brauner return -EOVERFLOW; 9114d4340c9SChristian Brauner return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); 912cb8e7090SChristoph Hellwig } 913cb8e7090SChristoph Hellwig 914cb8e7090SChristoph Hellwig /* 915cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 916cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 917cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 918cb8e7090SChristoph Hellwig */ 91992872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 9204d4340c9SChristian Brauner struct user_namespace *mnt_userns, 92152f75f4fSDavid Sterba const char *name, int namelen, 92272fd032eSSage Weil struct btrfs_root *snap_src, 9239babda9fSNikolay Borisov bool readonly, 9248696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 925cb8e7090SChristoph Hellwig { 9262b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9270b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 928cb8e7090SChristoph Hellwig struct dentry *dentry; 929cb8e7090SChristoph Hellwig int error; 930cb8e7090SChristoph Hellwig 93100235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 93200235411SAl Viro if (error == -EINTR) 93300235411SAl Viro return error; 934cb8e7090SChristoph Hellwig 9354d4340c9SChristian Brauner dentry = lookup_one(mnt_userns, name, parent->dentry, namelen); 936cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 937cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 938cb8e7090SChristoph Hellwig goto out_unlock; 939cb8e7090SChristoph Hellwig 9404d4340c9SChristian Brauner error = btrfs_may_create(mnt_userns, dir, dentry); 941cb8e7090SChristoph Hellwig if (error) 942a874a63eSLiu Bo goto out_dput; 943cb8e7090SChristoph Hellwig 9449c52057cSChris Mason /* 9459c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9469c52057cSChris Mason * check for them now when we can safely fail 9479c52057cSChris Mason */ 9489c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9499c52057cSChris Mason dir->i_ino, name, 9509c52057cSChris Mason namelen); 9519c52057cSChris Mason if (error) 9529c52057cSChris Mason goto out_dput; 9539c52057cSChris Mason 9540b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 95576dda93cSYan, Zheng 95676dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 95776dda93cSYan, Zheng goto out_up_read; 95876dda93cSYan, Zheng 9599babda9fSNikolay Borisov if (snap_src) 9609babda9fSNikolay Borisov error = create_snapshot(snap_src, dir, dentry, readonly, inherit); 9619babda9fSNikolay Borisov else 9624d4340c9SChristian Brauner error = create_subvol(mnt_userns, dir, dentry, name, namelen, inherit); 9639babda9fSNikolay Borisov 96476dda93cSYan, Zheng if (!error) 96576dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 96676dda93cSYan, Zheng out_up_read: 9670b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 968cb8e7090SChristoph Hellwig out_dput: 969cb8e7090SChristoph Hellwig dput(dentry); 970cb8e7090SChristoph Hellwig out_unlock: 97164708539SJosef Bacik btrfs_inode_unlock(dir, 0); 972cb8e7090SChristoph Hellwig return error; 973cb8e7090SChristoph Hellwig } 974cb8e7090SChristoph Hellwig 975c11fbb6eSRobbie Ko static noinline int btrfs_mksnapshot(const struct path *parent, 9764d4340c9SChristian Brauner struct user_namespace *mnt_userns, 977c11fbb6eSRobbie Ko const char *name, int namelen, 978c11fbb6eSRobbie Ko struct btrfs_root *root, 979c11fbb6eSRobbie Ko bool readonly, 980c11fbb6eSRobbie Ko struct btrfs_qgroup_inherit *inherit) 981c11fbb6eSRobbie Ko { 982c11fbb6eSRobbie Ko int ret; 983c11fbb6eSRobbie Ko bool snapshot_force_cow = false; 984c11fbb6eSRobbie Ko 985c11fbb6eSRobbie Ko /* 986c11fbb6eSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 987c11fbb6eSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 988c11fbb6eSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 989c11fbb6eSRobbie Ko */ 990c11fbb6eSRobbie Ko btrfs_drew_read_lock(&root->snapshot_lock); 991c11fbb6eSRobbie Ko 992f9baa501SFilipe Manana ret = btrfs_start_delalloc_snapshot(root, false); 993c11fbb6eSRobbie Ko if (ret) 994c11fbb6eSRobbie Ko goto out; 995c11fbb6eSRobbie Ko 996c11fbb6eSRobbie Ko /* 997c11fbb6eSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 998c11fbb6eSRobbie Ko * we force future writes to fallback to COW mode during snapshot 999c11fbb6eSRobbie Ko * creation. 1000c11fbb6eSRobbie Ko */ 1001c11fbb6eSRobbie Ko atomic_inc(&root->snapshot_force_cow); 1002c11fbb6eSRobbie Ko snapshot_force_cow = true; 1003c11fbb6eSRobbie Ko 1004c11fbb6eSRobbie Ko btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 1005c11fbb6eSRobbie Ko 10064d4340c9SChristian Brauner ret = btrfs_mksubvol(parent, mnt_userns, name, namelen, 1007c11fbb6eSRobbie Ko root, readonly, inherit); 1008c11fbb6eSRobbie Ko out: 1009c11fbb6eSRobbie Ko if (snapshot_force_cow) 1010c11fbb6eSRobbie Ko atomic_dec(&root->snapshot_force_cow); 1011c11fbb6eSRobbie Ko btrfs_drew_read_unlock(&root->snapshot_lock); 1012c11fbb6eSRobbie Ko return ret; 1013c11fbb6eSRobbie Ko } 1014c11fbb6eSRobbie Ko 1015e9eec721SQu Wenruo static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start, 1016e9eec721SQu Wenruo bool locked) 101717ce6ef8SLiu Bo { 101817ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1019940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10206c282eb4SLi Zefan struct extent_map *em; 102176068caeSQu Wenruo const u32 sectorsize = BTRFS_I(inode)->root->fs_info->sectorsize; 1022940100a4SChris Mason 1023940100a4SChris Mason /* 1024940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1025940100a4SChris Mason * the full extent lock 1026940100a4SChris Mason */ 1027940100a4SChris Mason read_lock(&em_tree->lock); 102876068caeSQu Wenruo em = lookup_extent_mapping(em_tree, start, sectorsize); 1029940100a4SChris Mason read_unlock(&em_tree->lock); 1030940100a4SChris Mason 1031940100a4SChris Mason if (!em) { 1032308d9800SFilipe Manana struct extent_state *cached = NULL; 103376068caeSQu Wenruo u64 end = start + sectorsize - 1; 1034308d9800SFilipe Manana 1035940100a4SChris Mason /* get the big lock and read metadata off disk */ 1036e9eec721SQu Wenruo if (!locked) 1037ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 103876068caeSQu Wenruo em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, sectorsize); 1039e9eec721SQu Wenruo if (!locked) 1040e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1041940100a4SChris Mason 10426cf8bfbfSDan Carpenter if (IS_ERR(em)) 10436c282eb4SLi Zefan return NULL; 1044940100a4SChris Mason } 1045940100a4SChris Mason 10466c282eb4SLi Zefan return em; 10476c282eb4SLi Zefan } 10486c282eb4SLi Zefan 1049e9eec721SQu Wenruo static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em, 1050e9eec721SQu Wenruo bool locked) 10516c282eb4SLi Zefan { 10526c282eb4SLi Zefan struct extent_map *next; 10536c282eb4SLi Zefan bool ret = true; 10546c282eb4SLi Zefan 10556c282eb4SLi Zefan /* this is the last extent */ 10566c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 10576c282eb4SLi Zefan return false; 10586c282eb4SLi Zefan 1059e9eec721SQu Wenruo next = defrag_lookup_extent(inode, em->start + em->len, locked); 1060e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1061e9512d72SChris Mason ret = false; 1062e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1063ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 10646c282eb4SLi Zefan ret = false; 10656c282eb4SLi Zefan 10666c282eb4SLi Zefan free_extent_map(next); 10676c282eb4SLi Zefan return ret; 10686c282eb4SLi Zefan } 10696c282eb4SLi Zefan 10704cb5300bSChris Mason /* 10715767b50cSQu Wenruo * Prepare one page to be defragged. 10725767b50cSQu Wenruo * 10735767b50cSQu Wenruo * This will ensure: 10745767b50cSQu Wenruo * 10755767b50cSQu Wenruo * - Returned page is locked and has been set up properly. 10765767b50cSQu Wenruo * - No ordered extent exists in the page. 10775767b50cSQu Wenruo * - The page is uptodate. 10785767b50cSQu Wenruo * 10795767b50cSQu Wenruo * NOTE: Caller should also wait for page writeback after the cluster is 10805767b50cSQu Wenruo * prepared, here we don't do writeback wait for each page. 10815767b50cSQu Wenruo */ 10825767b50cSQu Wenruo static struct page *defrag_prepare_one_page(struct btrfs_inode *inode, 10835767b50cSQu Wenruo pgoff_t index) 1084940100a4SChris Mason { 10855767b50cSQu Wenruo struct address_space *mapping = inode->vfs_inode.i_mapping; 10865767b50cSQu Wenruo gfp_t mask = btrfs_alloc_write_mask(mapping); 10875767b50cSQu Wenruo u64 page_start = (u64)index << PAGE_SHIFT; 10885767b50cSQu Wenruo u64 page_end = page_start + PAGE_SIZE - 1; 10894cb5300bSChris Mason struct extent_state *cached_state = NULL; 10904cb5300bSChris Mason struct page *page; 10915767b50cSQu Wenruo int ret; 10925767b50cSQu Wenruo 1093600a45e1SMiao Xie again: 10945767b50cSQu Wenruo page = find_or_create_page(mapping, index, mask); 10954cb5300bSChris Mason if (!page) 10965767b50cSQu Wenruo return ERR_PTR(-ENOMEM); 10975767b50cSQu Wenruo 109824bcb454SOmar Sandoval /* 109924bcb454SOmar Sandoval * Since we can defragment files opened read-only, we can encounter 110024bcb454SOmar Sandoval * transparent huge pages here (see CONFIG_READ_ONLY_THP_FOR_FS). We 110124bcb454SOmar Sandoval * can't do I/O using huge pages yet, so return an error for now. 110224bcb454SOmar Sandoval * Filesystem transparent huge pages are typically only used for 110324bcb454SOmar Sandoval * executables that explicitly enable them, so this isn't very 110424bcb454SOmar Sandoval * restrictive. 110524bcb454SOmar Sandoval */ 110624bcb454SOmar Sandoval if (PageCompound(page)) { 110724bcb454SOmar Sandoval unlock_page(page); 110824bcb454SOmar Sandoval put_page(page); 110924bcb454SOmar Sandoval return ERR_PTR(-ETXTBSY); 111024bcb454SOmar Sandoval } 11114cb5300bSChris Mason 111232443de3SQu Wenruo ret = set_page_extent_mapped(page); 111332443de3SQu Wenruo if (ret < 0) { 111432443de3SQu Wenruo unlock_page(page); 111532443de3SQu Wenruo put_page(page); 11165767b50cSQu Wenruo return ERR_PTR(ret); 111732443de3SQu Wenruo } 111832443de3SQu Wenruo 11195767b50cSQu Wenruo /* Wait for any existing ordered extent in the range */ 1120600a45e1SMiao Xie while (1) { 11215767b50cSQu Wenruo struct btrfs_ordered_extent *ordered; 11225767b50cSQu Wenruo 11235767b50cSQu Wenruo lock_extent_bits(&inode->io_tree, page_start, page_end, &cached_state); 11245767b50cSQu Wenruo ordered = btrfs_lookup_ordered_range(inode, page_start, PAGE_SIZE); 11255767b50cSQu Wenruo unlock_extent_cached(&inode->io_tree, page_start, page_end, 1126e43bbe5eSDavid Sterba &cached_state); 1127600a45e1SMiao Xie if (!ordered) 1128600a45e1SMiao Xie break; 1129600a45e1SMiao Xie 1130600a45e1SMiao Xie unlock_page(page); 1131c0a43603SNikolay Borisov btrfs_start_ordered_extent(ordered, 1); 1132600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1133600a45e1SMiao Xie lock_page(page); 11341f12bd06SLiu Bo /* 11355767b50cSQu Wenruo * We unlocked the page above, so we need check if it was 11365767b50cSQu Wenruo * released or not. 11371f12bd06SLiu Bo */ 11385767b50cSQu Wenruo if (page->mapping != mapping || !PagePrivate(page)) { 11391f12bd06SLiu Bo unlock_page(page); 114009cbfeafSKirill A. Shutemov put_page(page); 11411f12bd06SLiu Bo goto again; 11421f12bd06SLiu Bo } 1143600a45e1SMiao Xie } 1144600a45e1SMiao Xie 11455767b50cSQu Wenruo /* 11465767b50cSQu Wenruo * Now the page range has no ordered extent any more. Read the page to 11475767b50cSQu Wenruo * make it uptodate. 11485767b50cSQu Wenruo */ 11494cb5300bSChris Mason if (!PageUptodate(page)) { 11504cb5300bSChris Mason btrfs_readpage(NULL, page); 11514cb5300bSChris Mason lock_page(page); 11525767b50cSQu Wenruo if (page->mapping != mapping || !PagePrivate(page)) { 1153600a45e1SMiao Xie unlock_page(page); 115409cbfeafSKirill A. Shutemov put_page(page); 1155600a45e1SMiao Xie goto again; 1156600a45e1SMiao Xie } 11575767b50cSQu Wenruo if (!PageUptodate(page)) { 11585767b50cSQu Wenruo unlock_page(page); 11595767b50cSQu Wenruo put_page(page); 11605767b50cSQu Wenruo return ERR_PTR(-EIO); 11614cb5300bSChris Mason } 11625767b50cSQu Wenruo } 11635767b50cSQu Wenruo return page; 11645767b50cSQu Wenruo } 11654cb5300bSChris Mason 1166eb793cf8SQu Wenruo struct defrag_target_range { 1167eb793cf8SQu Wenruo struct list_head list; 1168eb793cf8SQu Wenruo u64 start; 1169eb793cf8SQu Wenruo u64 len; 1170eb793cf8SQu Wenruo }; 11714cb5300bSChris Mason 11724cb5300bSChris Mason /* 1173eb793cf8SQu Wenruo * Collect all valid target extents. 1174eb793cf8SQu Wenruo * 1175eb793cf8SQu Wenruo * @start: file offset to lookup 1176eb793cf8SQu Wenruo * @len: length to lookup 1177eb793cf8SQu Wenruo * @extent_thresh: file extent size threshold, any extent size >= this value 1178eb793cf8SQu Wenruo * will be ignored 1179eb793cf8SQu Wenruo * @newer_than: only defrag extents newer than this value 1180eb793cf8SQu Wenruo * @do_compress: whether the defrag is doing compression 1181eb793cf8SQu Wenruo * if true, @extent_thresh will be ignored and all regular 1182eb793cf8SQu Wenruo * file extents meeting @newer_than will be targets. 1183e9eec721SQu Wenruo * @locked: if the range has already held extent lock 1184eb793cf8SQu Wenruo * @target_list: list of targets file extents 11854cb5300bSChris Mason */ 1186eb793cf8SQu Wenruo static int defrag_collect_targets(struct btrfs_inode *inode, 1187eb793cf8SQu Wenruo u64 start, u64 len, u32 extent_thresh, 1188eb793cf8SQu Wenruo u64 newer_than, bool do_compress, 1189e9eec721SQu Wenruo bool locked, struct list_head *target_list) 1190eb793cf8SQu Wenruo { 1191eb793cf8SQu Wenruo u64 cur = start; 1192eb793cf8SQu Wenruo int ret = 0; 1193eb793cf8SQu Wenruo 1194eb793cf8SQu Wenruo while (cur < start + len) { 1195eb793cf8SQu Wenruo struct extent_map *em; 1196eb793cf8SQu Wenruo struct defrag_target_range *new; 1197eb793cf8SQu Wenruo bool next_mergeable = true; 1198eb793cf8SQu Wenruo u64 range_len; 1199eb793cf8SQu Wenruo 1200e9eec721SQu Wenruo em = defrag_lookup_extent(&inode->vfs_inode, cur, locked); 1201eb793cf8SQu Wenruo if (!em) 1202eb793cf8SQu Wenruo break; 1203eb793cf8SQu Wenruo 1204eb793cf8SQu Wenruo /* Skip hole/inline/preallocated extents */ 1205eb793cf8SQu Wenruo if (em->block_start >= EXTENT_MAP_LAST_BYTE || 1206eb793cf8SQu Wenruo test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) 1207eb793cf8SQu Wenruo goto next; 1208eb793cf8SQu Wenruo 1209eb793cf8SQu Wenruo /* Skip older extent */ 1210eb793cf8SQu Wenruo if (em->generation < newer_than) 1211eb793cf8SQu Wenruo goto next; 1212eb793cf8SQu Wenruo 1213eb793cf8SQu Wenruo /* 1214eb793cf8SQu Wenruo * For do_compress case, we want to compress all valid file 1215eb793cf8SQu Wenruo * extents, thus no @extent_thresh or mergeable check. 1216eb793cf8SQu Wenruo */ 1217eb793cf8SQu Wenruo if (do_compress) 1218eb793cf8SQu Wenruo goto add; 1219eb793cf8SQu Wenruo 1220eb793cf8SQu Wenruo /* Skip too large extent */ 1221eb793cf8SQu Wenruo if (em->len >= extent_thresh) 1222eb793cf8SQu Wenruo goto next; 1223eb793cf8SQu Wenruo 1224e9eec721SQu Wenruo next_mergeable = defrag_check_next_extent(&inode->vfs_inode, em, 1225e9eec721SQu Wenruo locked); 1226eb793cf8SQu Wenruo if (!next_mergeable) { 1227eb793cf8SQu Wenruo struct defrag_target_range *last; 1228eb793cf8SQu Wenruo 1229eb793cf8SQu Wenruo /* Empty target list, no way to merge with last entry */ 1230eb793cf8SQu Wenruo if (list_empty(target_list)) 1231eb793cf8SQu Wenruo goto next; 1232eb793cf8SQu Wenruo last = list_entry(target_list->prev, 1233eb793cf8SQu Wenruo struct defrag_target_range, list); 1234eb793cf8SQu Wenruo /* Not mergeable with last entry */ 1235eb793cf8SQu Wenruo if (last->start + last->len != cur) 1236eb793cf8SQu Wenruo goto next; 1237eb793cf8SQu Wenruo 1238eb793cf8SQu Wenruo /* Mergeable, fall through to add it to @target_list. */ 1239eb793cf8SQu Wenruo } 1240eb793cf8SQu Wenruo 1241eb793cf8SQu Wenruo add: 1242eb793cf8SQu Wenruo range_len = min(extent_map_end(em), start + len) - cur; 1243eb793cf8SQu Wenruo /* 1244eb793cf8SQu Wenruo * This one is a good target, check if it can be merged into 1245eb793cf8SQu Wenruo * last range of the target list. 1246eb793cf8SQu Wenruo */ 1247eb793cf8SQu Wenruo if (!list_empty(target_list)) { 1248eb793cf8SQu Wenruo struct defrag_target_range *last; 1249eb793cf8SQu Wenruo 1250eb793cf8SQu Wenruo last = list_entry(target_list->prev, 1251eb793cf8SQu Wenruo struct defrag_target_range, list); 1252eb793cf8SQu Wenruo ASSERT(last->start + last->len <= cur); 1253eb793cf8SQu Wenruo if (last->start + last->len == cur) { 1254eb793cf8SQu Wenruo /* Mergeable, enlarge the last entry */ 1255eb793cf8SQu Wenruo last->len += range_len; 1256eb793cf8SQu Wenruo goto next; 1257eb793cf8SQu Wenruo } 1258eb793cf8SQu Wenruo /* Fall through to allocate a new entry */ 1259eb793cf8SQu Wenruo } 1260eb793cf8SQu Wenruo 1261eb793cf8SQu Wenruo /* Allocate new defrag_target_range */ 1262eb793cf8SQu Wenruo new = kmalloc(sizeof(*new), GFP_NOFS); 1263eb793cf8SQu Wenruo if (!new) { 1264eb793cf8SQu Wenruo free_extent_map(em); 1265eb793cf8SQu Wenruo ret = -ENOMEM; 1266eb793cf8SQu Wenruo break; 1267eb793cf8SQu Wenruo } 1268eb793cf8SQu Wenruo new->start = cur; 1269eb793cf8SQu Wenruo new->len = range_len; 1270eb793cf8SQu Wenruo list_add_tail(&new->list, target_list); 1271eb793cf8SQu Wenruo 1272eb793cf8SQu Wenruo next: 1273eb793cf8SQu Wenruo cur = extent_map_end(em); 1274eb793cf8SQu Wenruo free_extent_map(em); 1275eb793cf8SQu Wenruo } 1276eb793cf8SQu Wenruo if (ret < 0) { 1277eb793cf8SQu Wenruo struct defrag_target_range *entry; 1278eb793cf8SQu Wenruo struct defrag_target_range *tmp; 1279eb793cf8SQu Wenruo 1280eb793cf8SQu Wenruo list_for_each_entry_safe(entry, tmp, target_list, list) { 1281eb793cf8SQu Wenruo list_del_init(&entry->list); 1282eb793cf8SQu Wenruo kfree(entry); 1283eb793cf8SQu Wenruo } 1284eb793cf8SQu Wenruo } 1285eb793cf8SQu Wenruo return ret; 1286eb793cf8SQu Wenruo } 1287eb793cf8SQu Wenruo 128822b398eeSQu Wenruo #define CLUSTER_SIZE (SZ_256K) 128922b398eeSQu Wenruo 129022b398eeSQu Wenruo /* 129122b398eeSQu Wenruo * Defrag one contiguous target range. 129222b398eeSQu Wenruo * 129322b398eeSQu Wenruo * @inode: target inode 129422b398eeSQu Wenruo * @target: target range to defrag 129522b398eeSQu Wenruo * @pages: locked pages covering the defrag range 129622b398eeSQu Wenruo * @nr_pages: number of locked pages 129722b398eeSQu Wenruo * 129822b398eeSQu Wenruo * Caller should ensure: 129922b398eeSQu Wenruo * 130022b398eeSQu Wenruo * - Pages are prepared 130122b398eeSQu Wenruo * Pages should be locked, no ordered extent in the pages range, 130222b398eeSQu Wenruo * no writeback. 130322b398eeSQu Wenruo * 130422b398eeSQu Wenruo * - Extent bits are locked 130522b398eeSQu Wenruo */ 130622b398eeSQu Wenruo static int defrag_one_locked_target(struct btrfs_inode *inode, 130722b398eeSQu Wenruo struct defrag_target_range *target, 130822b398eeSQu Wenruo struct page **pages, int nr_pages, 130922b398eeSQu Wenruo struct extent_state **cached_state) 131022b398eeSQu Wenruo { 131122b398eeSQu Wenruo struct btrfs_fs_info *fs_info = inode->root->fs_info; 131222b398eeSQu Wenruo struct extent_changeset *data_reserved = NULL; 131322b398eeSQu Wenruo const u64 start = target->start; 131422b398eeSQu Wenruo const u64 len = target->len; 131522b398eeSQu Wenruo unsigned long last_index = (start + len - 1) >> PAGE_SHIFT; 131622b398eeSQu Wenruo unsigned long start_index = start >> PAGE_SHIFT; 131722b398eeSQu Wenruo unsigned long first_index = page_index(pages[0]); 131822b398eeSQu Wenruo int ret = 0; 131922b398eeSQu Wenruo int i; 132022b398eeSQu Wenruo 132122b398eeSQu Wenruo ASSERT(last_index - first_index + 1 <= nr_pages); 132222b398eeSQu Wenruo 132322b398eeSQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, start, len); 132422b398eeSQu Wenruo if (ret < 0) 132522b398eeSQu Wenruo return ret; 132622b398eeSQu Wenruo clear_extent_bit(&inode->io_tree, start, start + len - 1, 132722b398eeSQu Wenruo EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | 132822b398eeSQu Wenruo EXTENT_DEFRAG, 0, 0, cached_state); 132922b398eeSQu Wenruo set_extent_defrag(&inode->io_tree, start, start + len - 1, cached_state); 133022b398eeSQu Wenruo 133122b398eeSQu Wenruo /* Update the page status */ 133222b398eeSQu Wenruo for (i = start_index - first_index; i <= last_index - first_index; i++) { 133322b398eeSQu Wenruo ClearPageChecked(pages[i]); 133422b398eeSQu Wenruo btrfs_page_clamp_set_dirty(fs_info, pages[i], start, len); 133522b398eeSQu Wenruo } 133622b398eeSQu Wenruo btrfs_delalloc_release_extents(inode, len); 133722b398eeSQu Wenruo extent_changeset_free(data_reserved); 133822b398eeSQu Wenruo 133922b398eeSQu Wenruo return ret; 134022b398eeSQu Wenruo } 134122b398eeSQu Wenruo 1342e9eec721SQu Wenruo static int defrag_one_range(struct btrfs_inode *inode, u64 start, u32 len, 1343e9eec721SQu Wenruo u32 extent_thresh, u64 newer_than, bool do_compress) 1344e9eec721SQu Wenruo { 1345e9eec721SQu Wenruo struct extent_state *cached_state = NULL; 1346e9eec721SQu Wenruo struct defrag_target_range *entry; 1347e9eec721SQu Wenruo struct defrag_target_range *tmp; 1348e9eec721SQu Wenruo LIST_HEAD(target_list); 1349e9eec721SQu Wenruo struct page **pages; 1350e9eec721SQu Wenruo const u32 sectorsize = inode->root->fs_info->sectorsize; 1351e9eec721SQu Wenruo u64 last_index = (start + len - 1) >> PAGE_SHIFT; 1352e9eec721SQu Wenruo u64 start_index = start >> PAGE_SHIFT; 1353e9eec721SQu Wenruo unsigned int nr_pages = last_index - start_index + 1; 1354e9eec721SQu Wenruo int ret = 0; 1355e9eec721SQu Wenruo int i; 1356e9eec721SQu Wenruo 1357e9eec721SQu Wenruo ASSERT(nr_pages <= CLUSTER_SIZE / PAGE_SIZE); 1358e9eec721SQu Wenruo ASSERT(IS_ALIGNED(start, sectorsize) && IS_ALIGNED(len, sectorsize)); 1359e9eec721SQu Wenruo 1360e9eec721SQu Wenruo pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS); 1361e9eec721SQu Wenruo if (!pages) 1362e9eec721SQu Wenruo return -ENOMEM; 1363e9eec721SQu Wenruo 1364e9eec721SQu Wenruo /* Prepare all pages */ 1365e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) { 1366e9eec721SQu Wenruo pages[i] = defrag_prepare_one_page(inode, start_index + i); 1367e9eec721SQu Wenruo if (IS_ERR(pages[i])) { 1368e9eec721SQu Wenruo ret = PTR_ERR(pages[i]); 1369e9eec721SQu Wenruo pages[i] = NULL; 1370e9eec721SQu Wenruo goto free_pages; 1371e9eec721SQu Wenruo } 1372e9eec721SQu Wenruo } 1373e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) 13744cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13754cb5300bSChris Mason 1376e9eec721SQu Wenruo /* Lock the pages range */ 1377e9eec721SQu Wenruo lock_extent_bits(&inode->io_tree, start_index << PAGE_SHIFT, 1378e9eec721SQu Wenruo (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, 1379e9eec721SQu Wenruo &cached_state); 1380e9eec721SQu Wenruo /* 1381e9eec721SQu Wenruo * Now we have a consistent view about the extent map, re-check 1382e9eec721SQu Wenruo * which range really needs to be defragged. 1383e9eec721SQu Wenruo * 1384e9eec721SQu Wenruo * And this time we have extent locked already, pass @locked = true 1385e9eec721SQu Wenruo * so that we won't relock the extent range and cause deadlock. 1386e9eec721SQu Wenruo */ 1387e9eec721SQu Wenruo ret = defrag_collect_targets(inode, start, len, extent_thresh, 1388e9eec721SQu Wenruo newer_than, do_compress, true, 1389e9eec721SQu Wenruo &target_list); 1390e9eec721SQu Wenruo if (ret < 0) 1391e9eec721SQu Wenruo goto unlock_extent; 13924cb5300bSChris Mason 1393e9eec721SQu Wenruo list_for_each_entry(entry, &target_list, list) { 1394e9eec721SQu Wenruo ret = defrag_one_locked_target(inode, entry, pages, nr_pages, 1395e9eec721SQu Wenruo &cached_state); 1396e9eec721SQu Wenruo if (ret < 0) 1397e9eec721SQu Wenruo break; 1398e9eec721SQu Wenruo } 1399e9eec721SQu Wenruo 1400e9eec721SQu Wenruo list_for_each_entry_safe(entry, tmp, &target_list, list) { 1401e9eec721SQu Wenruo list_del_init(&entry->list); 1402e9eec721SQu Wenruo kfree(entry); 1403e9eec721SQu Wenruo } 1404e9eec721SQu Wenruo unlock_extent: 1405e9eec721SQu Wenruo unlock_extent_cached(&inode->io_tree, start_index << PAGE_SHIFT, 1406e9eec721SQu Wenruo (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, 1407e9eec721SQu Wenruo &cached_state); 1408e9eec721SQu Wenruo free_pages: 1409e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) { 1410e9eec721SQu Wenruo if (pages[i]) { 1411e9eec721SQu Wenruo unlock_page(pages[i]); 1412e9eec721SQu Wenruo put_page(pages[i]); 1413e9eec721SQu Wenruo } 1414e9eec721SQu Wenruo } 1415e9eec721SQu Wenruo kfree(pages); 1416e9eec721SQu Wenruo return ret; 1417e9eec721SQu Wenruo } 1418e9eec721SQu Wenruo 1419b18c3ab2SQu Wenruo static int defrag_one_cluster(struct btrfs_inode *inode, 1420b18c3ab2SQu Wenruo struct file_ra_state *ra, 1421b18c3ab2SQu Wenruo u64 start, u32 len, u32 extent_thresh, 1422b18c3ab2SQu Wenruo u64 newer_than, bool do_compress, 1423b18c3ab2SQu Wenruo unsigned long *sectors_defragged, 1424b18c3ab2SQu Wenruo unsigned long max_sectors) 1425b18c3ab2SQu Wenruo { 1426b18c3ab2SQu Wenruo const u32 sectorsize = inode->root->fs_info->sectorsize; 1427b18c3ab2SQu Wenruo struct defrag_target_range *entry; 1428b18c3ab2SQu Wenruo struct defrag_target_range *tmp; 1429b18c3ab2SQu Wenruo LIST_HEAD(target_list); 1430b18c3ab2SQu Wenruo int ret; 1431b18c3ab2SQu Wenruo 1432b18c3ab2SQu Wenruo BUILD_BUG_ON(!IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE)); 1433b18c3ab2SQu Wenruo ret = defrag_collect_targets(inode, start, len, extent_thresh, 1434b18c3ab2SQu Wenruo newer_than, do_compress, false, 1435b18c3ab2SQu Wenruo &target_list); 1436b18c3ab2SQu Wenruo if (ret < 0) 1437b18c3ab2SQu Wenruo goto out; 1438b18c3ab2SQu Wenruo 1439b18c3ab2SQu Wenruo list_for_each_entry(entry, &target_list, list) { 1440b18c3ab2SQu Wenruo u32 range_len = entry->len; 1441b18c3ab2SQu Wenruo 1442b18c3ab2SQu Wenruo /* Reached the limit */ 1443b18c3ab2SQu Wenruo if (max_sectors && max_sectors == *sectors_defragged) 1444b18c3ab2SQu Wenruo break; 1445b18c3ab2SQu Wenruo 1446b18c3ab2SQu Wenruo if (max_sectors) 1447b18c3ab2SQu Wenruo range_len = min_t(u32, range_len, 1448b18c3ab2SQu Wenruo (max_sectors - *sectors_defragged) * sectorsize); 1449b18c3ab2SQu Wenruo 1450b18c3ab2SQu Wenruo if (ra) 1451b18c3ab2SQu Wenruo page_cache_sync_readahead(inode->vfs_inode.i_mapping, 1452b18c3ab2SQu Wenruo ra, NULL, entry->start >> PAGE_SHIFT, 1453b18c3ab2SQu Wenruo ((entry->start + range_len - 1) >> PAGE_SHIFT) - 1454b18c3ab2SQu Wenruo (entry->start >> PAGE_SHIFT) + 1); 1455b18c3ab2SQu Wenruo /* 1456b18c3ab2SQu Wenruo * Here we may not defrag any range if holes are punched before 1457b18c3ab2SQu Wenruo * we locked the pages. 1458b18c3ab2SQu Wenruo * But that's fine, it only affects the @sectors_defragged 1459b18c3ab2SQu Wenruo * accounting. 1460b18c3ab2SQu Wenruo */ 1461b18c3ab2SQu Wenruo ret = defrag_one_range(inode, entry->start, range_len, 1462b18c3ab2SQu Wenruo extent_thresh, newer_than, do_compress); 1463b18c3ab2SQu Wenruo if (ret < 0) 1464b18c3ab2SQu Wenruo break; 1465b18c3ab2SQu Wenruo *sectors_defragged += range_len; 1466b18c3ab2SQu Wenruo } 1467b18c3ab2SQu Wenruo out: 1468b18c3ab2SQu Wenruo list_for_each_entry_safe(entry, tmp, &target_list, list) { 1469b18c3ab2SQu Wenruo list_del_init(&entry->list); 1470b18c3ab2SQu Wenruo kfree(entry); 1471b18c3ab2SQu Wenruo } 1472b18c3ab2SQu Wenruo return ret; 1473b18c3ab2SQu Wenruo } 14747f458a38SFilipe Manana 14757f458a38SFilipe Manana /* 14761ccc2e8aSQu Wenruo * Entry point to file defragmentation. 14771ccc2e8aSQu Wenruo * 14781ccc2e8aSQu Wenruo * @inode: inode to be defragged 14791ccc2e8aSQu Wenruo * @ra: readahead state (can be NUL) 14801ccc2e8aSQu Wenruo * @range: defrag options including range and flags 14811ccc2e8aSQu Wenruo * @newer_than: minimum transid to defrag 14821ccc2e8aSQu Wenruo * @max_to_defrag: max number of sectors to be defragged, if 0, the whole inode 14831ccc2e8aSQu Wenruo * will be defragged. 14847f458a38SFilipe Manana */ 14851ccc2e8aSQu Wenruo int btrfs_defrag_file(struct inode *inode, struct file_ra_state *ra, 14864cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 14874cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 14884cb5300bSChris Mason { 14890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 14907b508037SQu Wenruo unsigned long sectors_defragged = 0; 1491151a31b2SLi Zefan u64 isize = i_size_read(inode); 14927b508037SQu Wenruo u64 cur; 14937b508037SQu Wenruo u64 last_byte; 14941e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14951ccc2e8aSQu Wenruo bool ra_allocated = false; 14967b508037SQu Wenruo int compress_type = BTRFS_COMPRESS_ZLIB; 14977b508037SQu Wenruo int ret = 0; 14987b508037SQu Wenruo u32 extent_thresh = range->extent_thresh; 14994cb5300bSChris Mason 15000abd5b17SLiu Bo if (isize == 0) 15010abd5b17SLiu Bo return 0; 15020abd5b17SLiu Bo 15030abd5b17SLiu Bo if (range->start >= isize) 15040abd5b17SLiu Bo return -EINVAL; 15051a419d85SLi Zefan 15061e2ef46dSDavid Sterba if (do_compress) { 1507ce96b7ffSChengguang Xu if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES) 15081a419d85SLi Zefan return -EINVAL; 15091a419d85SLi Zefan if (range->compress_type) 15101a419d85SLi Zefan compress_type = range->compress_type; 15111a419d85SLi Zefan } 1512f46b5a66SChristoph Hellwig 15130abd5b17SLiu Bo if (extent_thresh == 0) 1514ee22184bSByongho Lee extent_thresh = SZ_256K; 1515f46b5a66SChristoph Hellwig 15167b508037SQu Wenruo if (range->start + range->len > range->start) { 15177b508037SQu Wenruo /* Got a specific range */ 15187b508037SQu Wenruo last_byte = min(isize, range->start + range->len) - 1; 15197b508037SQu Wenruo } else { 15207b508037SQu Wenruo /* Defrag until file end */ 15217b508037SQu Wenruo last_byte = isize - 1; 15227b508037SQu Wenruo } 15237b508037SQu Wenruo 15244cb5300bSChris Mason /* 15251ccc2e8aSQu Wenruo * If we were not given a ra, allocate a readahead context. As 15260a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 15270a52d108SDavid Sterba * we don't error out. 15284cb5300bSChris Mason */ 15291ccc2e8aSQu Wenruo if (!ra) { 15301ccc2e8aSQu Wenruo ra_allocated = true; 153163e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 15320a52d108SDavid Sterba if (ra) 15334cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 15344cb5300bSChris Mason } 15354cb5300bSChris Mason 15367b508037SQu Wenruo /* Align the range */ 15377b508037SQu Wenruo cur = round_down(range->start, fs_info->sectorsize); 15387b508037SQu Wenruo last_byte = round_up(last_byte, fs_info->sectorsize) - 1; 15394cb5300bSChris Mason 15407b508037SQu Wenruo while (cur < last_byte) { 15417b508037SQu Wenruo u64 cluster_end; 15421e701a32SChris Mason 15437b508037SQu Wenruo /* The cluster size 256K should always be page aligned */ 15447b508037SQu Wenruo BUILD_BUG_ON(!IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE)); 15454cb5300bSChris Mason 15467b508037SQu Wenruo /* We want the cluster end at page boundary when possible */ 15477b508037SQu Wenruo cluster_end = (((cur >> PAGE_SHIFT) + 15487b508037SQu Wenruo (SZ_256K >> PAGE_SHIFT)) << PAGE_SHIFT) - 1; 15497b508037SQu Wenruo cluster_end = min(cluster_end, last_byte); 15504cb5300bSChris Mason 155164708539SJosef Bacik btrfs_inode_lock(inode, 0); 1552eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1553eede2bf3SOmar Sandoval ret = -ETXTBSY; 15547b508037SQu Wenruo btrfs_inode_unlock(inode, 0); 15557b508037SQu Wenruo break; 15567b508037SQu Wenruo } 15577b508037SQu Wenruo if (!(inode->i_sb->s_flags & SB_ACTIVE)) { 15587b508037SQu Wenruo btrfs_inode_unlock(inode, 0); 15597b508037SQu Wenruo break; 15607b508037SQu Wenruo } 15611e2ef46dSDavid Sterba if (do_compress) 1562eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 15637b508037SQu Wenruo ret = defrag_one_cluster(BTRFS_I(inode), ra, cur, 15647b508037SQu Wenruo cluster_end + 1 - cur, extent_thresh, 15657b508037SQu Wenruo newer_than, do_compress, 15667b508037SQu Wenruo §ors_defragged, max_to_defrag); 156764708539SJosef Bacik btrfs_inode_unlock(inode, 0); 15687b508037SQu Wenruo if (ret < 0) 15694cb5300bSChris Mason break; 15707b508037SQu Wenruo cur = cluster_end + 1; 15714cb5300bSChris Mason } 1572f46b5a66SChristoph Hellwig 15737b508037SQu Wenruo if (ra_allocated) 15747b508037SQu Wenruo kfree(ra); 15757b508037SQu Wenruo if (sectors_defragged) { 15767b508037SQu Wenruo /* 15777b508037SQu Wenruo * We have defragged some sectors, for compression case they 15787b508037SQu Wenruo * need to be written back immediately. 15797b508037SQu Wenruo */ 15807b508037SQu Wenruo if (range->flags & BTRFS_DEFRAG_RANGE_START_IO) { 15811e701a32SChris Mason filemap_flush(inode->i_mapping); 1582dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1583dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1584dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1585dec8ef90SFilipe Manana } 15867b508037SQu Wenruo if (range->compress_type == BTRFS_COMPRESS_LZO) 15870b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15887b508037SQu Wenruo else if (range->compress_type == BTRFS_COMPRESS_ZSTD) 15895c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15907b508037SQu Wenruo ret = sectors_defragged; 15911a419d85SLi Zefan } 15921e2ef46dSDavid Sterba if (do_compress) { 159364708539SJosef Bacik btrfs_inode_lock(inode, 0); 1594eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 159564708539SJosef Bacik btrfs_inode_unlock(inode, 0); 1596633085c7SFilipe David Borba Manana } 1597940100a4SChris Mason return ret; 1598f46b5a66SChristoph Hellwig } 1599f46b5a66SChristoph Hellwig 160017aaa434SDavid Sterba /* 160117aaa434SDavid Sterba * Try to start exclusive operation @type or cancel it if it's running. 160217aaa434SDavid Sterba * 160317aaa434SDavid Sterba * Return: 160417aaa434SDavid Sterba * 0 - normal mode, newly claimed op started 160517aaa434SDavid Sterba * >0 - normal mode, something else is running, 160617aaa434SDavid Sterba * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space 160717aaa434SDavid Sterba * ECANCELED - cancel mode, successful cancel 160817aaa434SDavid Sterba * ENOTCONN - cancel mode, operation not running anymore 160917aaa434SDavid Sterba */ 161017aaa434SDavid Sterba static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info, 161117aaa434SDavid Sterba enum btrfs_exclusive_operation type, bool cancel) 161217aaa434SDavid Sterba { 161317aaa434SDavid Sterba if (!cancel) { 161417aaa434SDavid Sterba /* Start normal op */ 161517aaa434SDavid Sterba if (!btrfs_exclop_start(fs_info, type)) 161617aaa434SDavid Sterba return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 161717aaa434SDavid Sterba /* Exclusive operation is now claimed */ 161817aaa434SDavid Sterba return 0; 161917aaa434SDavid Sterba } 162017aaa434SDavid Sterba 162117aaa434SDavid Sterba /* Cancel running op */ 162217aaa434SDavid Sterba if (btrfs_exclop_start_try_lock(fs_info, type)) { 162317aaa434SDavid Sterba /* 162417aaa434SDavid Sterba * This blocks any exclop finish from setting it to NONE, so we 162517aaa434SDavid Sterba * request cancellation. Either it runs and we will wait for it, 162617aaa434SDavid Sterba * or it has finished and no waiting will happen. 162717aaa434SDavid Sterba */ 162817aaa434SDavid Sterba atomic_inc(&fs_info->reloc_cancel_req); 162917aaa434SDavid Sterba btrfs_exclop_start_unlock(fs_info); 163017aaa434SDavid Sterba 163117aaa434SDavid Sterba if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) 163217aaa434SDavid Sterba wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING, 163317aaa434SDavid Sterba TASK_INTERRUPTIBLE); 163417aaa434SDavid Sterba 163517aaa434SDavid Sterba return -ECANCELED; 163617aaa434SDavid Sterba } 163717aaa434SDavid Sterba 163817aaa434SDavid Sterba /* Something else is running or none */ 163917aaa434SDavid Sterba return -ENOTCONN; 164017aaa434SDavid Sterba } 164117aaa434SDavid Sterba 1642198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 164376dda93cSYan, Zheng void __user *arg) 1644f46b5a66SChristoph Hellwig { 1645562d7b15SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 16460b246afaSJeff Mahoney struct inode *inode = file_inode(file); 16470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1648f46b5a66SChristoph Hellwig u64 new_size; 1649f46b5a66SChristoph Hellwig u64 old_size; 1650f46b5a66SChristoph Hellwig u64 devid = 1; 16510b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1652f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1653f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1654f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1655f46b5a66SChristoph Hellwig char *sizestr; 16569a40f122SGui Hecheng char *retptr; 1657f46b5a66SChristoph Hellwig char *devstr = NULL; 1658f46b5a66SChristoph Hellwig int ret = 0; 1659f46b5a66SChristoph Hellwig int mod = 0; 1660bb059a37SDavid Sterba bool cancel; 1661f46b5a66SChristoph Hellwig 1662e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1663e441d54dSChris Mason return -EPERM; 1664e441d54dSChris Mason 1665198605a8SMiao Xie ret = mnt_want_write_file(file); 1666198605a8SMiao Xie if (ret) 1667198605a8SMiao Xie return ret; 1668198605a8SMiao Xie 1669bb059a37SDavid Sterba /* 1670bb059a37SDavid Sterba * Read the arguments before checking exclusivity to be able to 1671bb059a37SDavid Sterba * distinguish regular resize and cancel 1672bb059a37SDavid Sterba */ 1673dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1674c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1675c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1676bb059a37SDavid Sterba goto out_drop; 1677c9e9f97bSIlya Dryomov } 16785516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1679f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1680bb059a37SDavid Sterba cancel = (strcmp("cancel", sizestr) == 0); 1681bb059a37SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel); 1682bb059a37SDavid Sterba if (ret) 1683bb059a37SDavid Sterba goto out_free; 1684bb059a37SDavid Sterba /* Exclusive operation is now claimed */ 1685bb059a37SDavid Sterba 1686f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1687f46b5a66SChristoph Hellwig if (devstr) { 1688f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1689f46b5a66SChristoph Hellwig *devstr = '\0'; 1690f46b5a66SChristoph Hellwig devstr = vol_args->name; 169158dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 169258dfae63SZhangZhen if (ret) 1693bb059a37SDavid Sterba goto out_finish; 1694dfd79829SMiao Xie if (!devid) { 1695dfd79829SMiao Xie ret = -EINVAL; 1696bb059a37SDavid Sterba goto out_finish; 1697dfd79829SMiao Xie } 16980b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1699f46b5a66SChristoph Hellwig } 1700dba60f3fSMiao Xie 1701562d7b15SJosef Bacik args.devid = devid; 1702562d7b15SJosef Bacik device = btrfs_find_device(fs_info->fs_devices, &args); 1703f46b5a66SChristoph Hellwig if (!device) { 17040b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1705c1c9ff7cSGeert Uytterhoeven devid); 1706dfd79829SMiao Xie ret = -ENODEV; 1707bb059a37SDavid Sterba goto out_finish; 1708f46b5a66SChristoph Hellwig } 1709dba60f3fSMiao Xie 1710ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 17110b246afaSJeff Mahoney btrfs_info(fs_info, 1712efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1713c1c9ff7cSGeert Uytterhoeven devid); 1714dfd79829SMiao Xie ret = -EPERM; 1715bb059a37SDavid Sterba goto out_finish; 17164e42ae1bSLiu Bo } 17174e42ae1bSLiu Bo 1718f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1719cda00ebaSChristoph Hellwig new_size = bdev_nr_bytes(device->bdev); 1720f46b5a66SChristoph Hellwig else { 1721f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1722f46b5a66SChristoph Hellwig mod = -1; 1723f46b5a66SChristoph Hellwig sizestr++; 1724f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1725f46b5a66SChristoph Hellwig mod = 1; 1726f46b5a66SChristoph Hellwig sizestr++; 1727f46b5a66SChristoph Hellwig } 17289a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 17299a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1730f46b5a66SChristoph Hellwig ret = -EINVAL; 1731bb059a37SDavid Sterba goto out_finish; 1732f46b5a66SChristoph Hellwig } 1733f46b5a66SChristoph Hellwig } 1734f46b5a66SChristoph Hellwig 1735401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1736dfd79829SMiao Xie ret = -EPERM; 1737bb059a37SDavid Sterba goto out_finish; 173863a212abSStefan Behrens } 173963a212abSStefan Behrens 17407cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1741f46b5a66SChristoph Hellwig 1742f46b5a66SChristoph Hellwig if (mod < 0) { 1743f46b5a66SChristoph Hellwig if (new_size > old_size) { 1744f46b5a66SChristoph Hellwig ret = -EINVAL; 1745bb059a37SDavid Sterba goto out_finish; 1746f46b5a66SChristoph Hellwig } 1747f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1748f46b5a66SChristoph Hellwig } else if (mod > 0) { 1749eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1750902c68a4SGui Hecheng ret = -ERANGE; 1751bb059a37SDavid Sterba goto out_finish; 1752eb8052e0SWenliang Fan } 1753f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1754f46b5a66SChristoph Hellwig } 1755f46b5a66SChristoph Hellwig 1756ee22184bSByongho Lee if (new_size < SZ_256M) { 1757f46b5a66SChristoph Hellwig ret = -EINVAL; 1758bb059a37SDavid Sterba goto out_finish; 1759f46b5a66SChristoph Hellwig } 1760cda00ebaSChristoph Hellwig if (new_size > bdev_nr_bytes(device->bdev)) { 1761f46b5a66SChristoph Hellwig ret = -EFBIG; 1762bb059a37SDavid Sterba goto out_finish; 1763f46b5a66SChristoph Hellwig } 1764f46b5a66SChristoph Hellwig 176547f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1766f46b5a66SChristoph Hellwig 1767f46b5a66SChristoph Hellwig if (new_size > old_size) { 1768a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 176998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 177098d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1771bb059a37SDavid Sterba goto out_finish; 177298d5dc13STsutomu Itoh } 1773f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17743a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1775ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1776f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17770253f40eSjeff.liu } /* equal, nothing need to do */ 1778f46b5a66SChristoph Hellwig 1779faf8f7b9SMarcos Paulo de Souza if (ret == 0 && new_size != old_size) 1780faf8f7b9SMarcos Paulo de Souza btrfs_info_in_rcu(fs_info, 1781faf8f7b9SMarcos Paulo de Souza "resize device %s (devid %llu) from %llu to %llu", 1782faf8f7b9SMarcos Paulo de Souza rcu_str_deref(device->name), device->devid, 1783faf8f7b9SMarcos Paulo de Souza old_size, new_size); 1784bb059a37SDavid Sterba out_finish: 1785bb059a37SDavid Sterba btrfs_exclop_finish(fs_info); 1786c9e9f97bSIlya Dryomov out_free: 1787f46b5a66SChristoph Hellwig kfree(vol_args); 1788bb059a37SDavid Sterba out_drop: 178918f39c41SIlya Dryomov mnt_drop_write_file(file); 1790f46b5a66SChristoph Hellwig return ret; 1791f46b5a66SChristoph Hellwig } 1792f46b5a66SChristoph Hellwig 17935d54c67eSNikolay Borisov static noinline int __btrfs_ioctl_snap_create(struct file *file, 17944d4340c9SChristian Brauner struct user_namespace *mnt_userns, 179552f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17965d54c67eSNikolay Borisov bool readonly, 17978696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1798f46b5a66SChristoph Hellwig { 1799f46b5a66SChristoph Hellwig int namelen; 18003de4586cSChris Mason int ret = 0; 1801f46b5a66SChristoph Hellwig 1802325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1803325c50e3SJeff Mahoney return -ENOTDIR; 1804325c50e3SJeff Mahoney 1805a874a63eSLiu Bo ret = mnt_want_write_file(file); 1806a874a63eSLiu Bo if (ret) 1807a874a63eSLiu Bo goto out; 1808a874a63eSLiu Bo 180972fd032eSSage Weil namelen = strlen(name); 181072fd032eSSage Weil if (strchr(name, '/')) { 1811f46b5a66SChristoph Hellwig ret = -EINVAL; 1812a874a63eSLiu Bo goto out_drop_write; 1813f46b5a66SChristoph Hellwig } 1814f46b5a66SChristoph Hellwig 181516780cabSChris Mason if (name[0] == '.' && 181616780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 181716780cabSChris Mason ret = -EEXIST; 1818a874a63eSLiu Bo goto out_drop_write; 181916780cabSChris Mason } 182016780cabSChris Mason 18213de4586cSChris Mason if (subvol) { 18224d4340c9SChristian Brauner ret = btrfs_mksubvol(&file->f_path, mnt_userns, name, 18234d4340c9SChristian Brauner namelen, NULL, readonly, inherit); 1824cb8e7090SChristoph Hellwig } else { 18252903ff01SAl Viro struct fd src = fdget(fd); 18263de4586cSChris Mason struct inode *src_inode; 18272903ff01SAl Viro if (!src.file) { 18283de4586cSChris Mason ret = -EINVAL; 1829a874a63eSLiu Bo goto out_drop_write; 18303de4586cSChris Mason } 18313de4586cSChris Mason 1832496ad9aaSAl Viro src_inode = file_inode(src.file); 1833496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1834c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1835efe120a0SFrank Holton "Snapshot src from another FS"); 183623ad5b17SKusanagi Kouichi ret = -EXDEV; 18374d4340c9SChristian Brauner } else if (!inode_owner_or_capable(mnt_userns, src_inode)) { 1838d0242061SDavid Sterba /* 1839d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1840d0242061SDavid Sterba * are limited to own subvolumes only 1841d0242061SDavid Sterba */ 1842d0242061SDavid Sterba ret = -EPERM; 1843ecd18815SAl Viro } else { 18444d4340c9SChristian Brauner ret = btrfs_mksnapshot(&file->f_path, mnt_userns, 18454d4340c9SChristian Brauner name, namelen, 184672fd032eSSage Weil BTRFS_I(src_inode)->root, 18475d54c67eSNikolay Borisov readonly, inherit); 1848ecd18815SAl Viro } 18492903ff01SAl Viro fdput(src); 1850cb8e7090SChristoph Hellwig } 1851a874a63eSLiu Bo out_drop_write: 1852a874a63eSLiu Bo mnt_drop_write_file(file); 1853f46b5a66SChristoph Hellwig out: 185472fd032eSSage Weil return ret; 185572fd032eSSage Weil } 185672fd032eSSage Weil 185772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1858fa0d2b9bSLi Zefan void __user *arg, int subvol) 185972fd032eSSage Weil { 1860fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 186172fd032eSSage Weil int ret; 186272fd032eSSage Weil 1863325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1864325c50e3SJeff Mahoney return -ENOTDIR; 1865325c50e3SJeff Mahoney 1866fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1867fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1868fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1869fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1870fa0d2b9bSLi Zefan 18714d4340c9SChristian Brauner ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), 18724d4340c9SChristian Brauner vol_args->name, vol_args->fd, subvol, 18734d4340c9SChristian Brauner false, NULL); 1874fa0d2b9bSLi Zefan 1875fa0d2b9bSLi Zefan kfree(vol_args); 1876fa0d2b9bSLi Zefan return ret; 1877fa0d2b9bSLi Zefan } 1878fa0d2b9bSLi Zefan 1879fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1880fa0d2b9bSLi Zefan void __user *arg, int subvol) 1881fa0d2b9bSLi Zefan { 1882fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1883fa0d2b9bSLi Zefan int ret; 1884b83cc969SLi Zefan bool readonly = false; 18856f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 188672fd032eSSage Weil 1887325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1888325c50e3SJeff Mahoney return -ENOTDIR; 1889325c50e3SJeff Mahoney 1890fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1891fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1892fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1893fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1894fdfb1e4fSLi Zefan 1895673990dbSDavid Sterba if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) { 1896b83cc969SLi Zefan ret = -EOPNOTSUPP; 1897c47ca32dSDan Carpenter goto free_args; 1898fdfb1e4fSLi Zefan } 1899fdfb1e4fSLi Zefan 1900b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1901b83cc969SLi Zefan readonly = true; 19026f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 19035011c5a6SDan Carpenter u64 nums; 19045011c5a6SDan Carpenter 19055011c5a6SDan Carpenter if (vol_args->size < sizeof(*inherit) || 19065011c5a6SDan Carpenter vol_args->size > PAGE_SIZE) { 19076f72c7e2SArne Jansen ret = -EINVAL; 1908c47ca32dSDan Carpenter goto free_args; 19096f72c7e2SArne Jansen } 19106f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 19116f72c7e2SArne Jansen if (IS_ERR(inherit)) { 19126f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1913c47ca32dSDan Carpenter goto free_args; 19146f72c7e2SArne Jansen } 19155011c5a6SDan Carpenter 19165011c5a6SDan Carpenter if (inherit->num_qgroups > PAGE_SIZE || 19175011c5a6SDan Carpenter inherit->num_ref_copies > PAGE_SIZE || 19185011c5a6SDan Carpenter inherit->num_excl_copies > PAGE_SIZE) { 19195011c5a6SDan Carpenter ret = -EINVAL; 19205011c5a6SDan Carpenter goto free_inherit; 19215011c5a6SDan Carpenter } 19225011c5a6SDan Carpenter 19235011c5a6SDan Carpenter nums = inherit->num_qgroups + 2 * inherit->num_ref_copies + 19245011c5a6SDan Carpenter 2 * inherit->num_excl_copies; 19255011c5a6SDan Carpenter if (vol_args->size != struct_size(inherit, qgroups, nums)) { 19265011c5a6SDan Carpenter ret = -EINVAL; 19275011c5a6SDan Carpenter goto free_inherit; 19285011c5a6SDan Carpenter } 19296f72c7e2SArne Jansen } 193075eaa0e2SSage Weil 19314d4340c9SChristian Brauner ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), 19324d4340c9SChristian Brauner vol_args->name, vol_args->fd, subvol, 19334d4340c9SChristian Brauner readonly, inherit); 1934c47ca32dSDan Carpenter if (ret) 1935c47ca32dSDan Carpenter goto free_inherit; 1936c47ca32dSDan Carpenter free_inherit: 19376f72c7e2SArne Jansen kfree(inherit); 1938c47ca32dSDan Carpenter free_args: 1939c47ca32dSDan Carpenter kfree(vol_args); 1940f46b5a66SChristoph Hellwig return ret; 1941f46b5a66SChristoph Hellwig } 1942f46b5a66SChristoph Hellwig 19430caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 19440caa102dSLi Zefan void __user *arg) 19450caa102dSLi Zefan { 1946496ad9aaSAl Viro struct inode *inode = file_inode(file); 19470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19480caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19490caa102dSLi Zefan int ret = 0; 19500caa102dSLi Zefan u64 flags = 0; 19510caa102dSLi Zefan 19524a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 19530caa102dSLi Zefan return -EINVAL; 19540caa102dSLi Zefan 19550b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 19560caa102dSLi Zefan if (btrfs_root_readonly(root)) 19570caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 19580b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 19590caa102dSLi Zefan 19600caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 19610caa102dSLi Zefan ret = -EFAULT; 19620caa102dSLi Zefan 19630caa102dSLi Zefan return ret; 19640caa102dSLi Zefan } 19650caa102dSLi Zefan 19660caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19670caa102dSLi Zefan void __user *arg) 19680caa102dSLi Zefan { 1969496ad9aaSAl Viro struct inode *inode = file_inode(file); 19700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19710caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19720caa102dSLi Zefan struct btrfs_trans_handle *trans; 19730caa102dSLi Zefan u64 root_flags; 19740caa102dSLi Zefan u64 flags; 19750caa102dSLi Zefan int ret = 0; 19760caa102dSLi Zefan 197739e1674fSChristian Brauner if (!inode_owner_or_capable(file_mnt_user_ns(file), inode)) 1978bd60ea0fSDavid Sterba return -EPERM; 1979bd60ea0fSDavid Sterba 1980b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1981b9ca0664SLiu Bo if (ret) 1982b9ca0664SLiu Bo goto out; 19830caa102dSLi Zefan 19844a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1985b9ca0664SLiu Bo ret = -EINVAL; 1986b9ca0664SLiu Bo goto out_drop_write; 1987b9ca0664SLiu Bo } 19880caa102dSLi Zefan 1989b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1990b9ca0664SLiu Bo ret = -EFAULT; 1991b9ca0664SLiu Bo goto out_drop_write; 1992b9ca0664SLiu Bo } 19930caa102dSLi Zefan 1994b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1995b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1996b9ca0664SLiu Bo goto out_drop_write; 1997b9ca0664SLiu Bo } 19980caa102dSLi Zefan 19990b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 20000caa102dSLi Zefan 20010caa102dSLi Zefan /* nothing to do */ 20020caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 2003b9ca0664SLiu Bo goto out_drop_sem; 20040caa102dSLi Zefan 20050caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 20062c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 20070caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 20080caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 20092c686537SDavid Sterba } else { 20102c686537SDavid Sterba /* 20112c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 20122c686537SDavid Sterba * send 20132c686537SDavid Sterba */ 20142c686537SDavid Sterba spin_lock(&root->root_item_lock); 20152c686537SDavid Sterba if (root->send_in_progress == 0) { 20160caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 20170caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 20182c686537SDavid Sterba spin_unlock(&root->root_item_lock); 20192c686537SDavid Sterba } else { 20202c686537SDavid Sterba spin_unlock(&root->root_item_lock); 20210b246afaSJeff Mahoney btrfs_warn(fs_info, 20222c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 20232c686537SDavid Sterba root->root_key.objectid); 20242c686537SDavid Sterba ret = -EPERM; 20252c686537SDavid Sterba goto out_drop_sem; 20262c686537SDavid Sterba } 20272c686537SDavid Sterba } 20280caa102dSLi Zefan 20290caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 20300caa102dSLi Zefan if (IS_ERR(trans)) { 20310caa102dSLi Zefan ret = PTR_ERR(trans); 20320caa102dSLi Zefan goto out_reset; 20330caa102dSLi Zefan } 20340caa102dSLi Zefan 20350b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 20360caa102dSLi Zefan &root->root_key, &root->root_item); 20379417ebc8SNikolay Borisov if (ret < 0) { 20389417ebc8SNikolay Borisov btrfs_end_transaction(trans); 20399417ebc8SNikolay Borisov goto out_reset; 20409417ebc8SNikolay Borisov } 20410caa102dSLi Zefan 20429417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 20439417ebc8SNikolay Borisov 20440caa102dSLi Zefan out_reset: 20450caa102dSLi Zefan if (ret) 20460caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 2047b9ca0664SLiu Bo out_drop_sem: 20480b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 2049b9ca0664SLiu Bo out_drop_write: 2050b9ca0664SLiu Bo mnt_drop_write_file(file); 2051b9ca0664SLiu Bo out: 20520caa102dSLi Zefan return ret; 20530caa102dSLi Zefan } 20540caa102dSLi Zefan 2055ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2056ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2057ac8e9819SChris Mason { 2058abc6e134SChris Mason struct btrfs_key test; 2059abc6e134SChris Mason int ret; 2060abc6e134SChris Mason 2061abc6e134SChris Mason test.objectid = sk->min_objectid; 2062abc6e134SChris Mason test.type = sk->min_type; 2063abc6e134SChris Mason test.offset = sk->min_offset; 2064abc6e134SChris Mason 2065abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2066abc6e134SChris Mason if (ret < 0) 2067ac8e9819SChris Mason return 0; 2068abc6e134SChris Mason 2069abc6e134SChris Mason test.objectid = sk->max_objectid; 2070abc6e134SChris Mason test.type = sk->max_type; 2071abc6e134SChris Mason test.offset = sk->max_offset; 2072abc6e134SChris Mason 2073abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2074abc6e134SChris Mason if (ret > 0) 2075ac8e9819SChris Mason return 0; 2076ac8e9819SChris Mason return 1; 2077ac8e9819SChris Mason } 2078ac8e9819SChris Mason 2079df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2080ac8e9819SChris Mason struct btrfs_key *key, 2081ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20829b6e817dSGerhard Heift size_t *buf_size, 2083ba346b35SGerhard Heift char __user *ubuf, 2084ac8e9819SChris Mason unsigned long *sk_offset, 2085ac8e9819SChris Mason int *num_found) 2086ac8e9819SChris Mason { 2087ac8e9819SChris Mason u64 found_transid; 2088ac8e9819SChris Mason struct extent_buffer *leaf; 2089ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2090dd81d459SNaohiro Aota struct btrfs_key test; 2091ac8e9819SChris Mason unsigned long item_off; 2092ac8e9819SChris Mason unsigned long item_len; 2093ac8e9819SChris Mason int nritems; 2094ac8e9819SChris Mason int i; 2095ac8e9819SChris Mason int slot; 2096ac8e9819SChris Mason int ret = 0; 2097ac8e9819SChris Mason 2098ac8e9819SChris Mason leaf = path->nodes[0]; 2099ac8e9819SChris Mason slot = path->slots[0]; 2100ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2101ac8e9819SChris Mason 2102ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2103ac8e9819SChris Mason i = nritems; 2104ac8e9819SChris Mason goto advance_key; 2105ac8e9819SChris Mason } 2106ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2107ac8e9819SChris Mason 2108ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2109ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 21103212fa14SJosef Bacik item_len = btrfs_item_size(leaf, i); 2111ac8e9819SChris Mason 211203b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 211303b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 211403b71c6cSGabriel de Perthuis continue; 211503b71c6cSGabriel de Perthuis 21169b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 21178f5f6178SGerhard Heift if (*num_found) { 2118ac8e9819SChris Mason ret = 1; 21198f5f6178SGerhard Heift goto out; 21208f5f6178SGerhard Heift } 21218f5f6178SGerhard Heift 21228f5f6178SGerhard Heift /* 21238f5f6178SGerhard Heift * return one empty item back for v1, which does not 21248f5f6178SGerhard Heift * handle -EOVERFLOW 21258f5f6178SGerhard Heift */ 21268f5f6178SGerhard Heift 21279b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2128ac8e9819SChris Mason item_len = 0; 21298f5f6178SGerhard Heift ret = -EOVERFLOW; 21308f5f6178SGerhard Heift } 2131ac8e9819SChris Mason 21329b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2133ac8e9819SChris Mason ret = 1; 213425c9bc2eSGerhard Heift goto out; 2135ac8e9819SChris Mason } 2136ac8e9819SChris Mason 2137ac8e9819SChris Mason sh.objectid = key->objectid; 2138ac8e9819SChris Mason sh.offset = key->offset; 2139ac8e9819SChris Mason sh.type = key->type; 2140ac8e9819SChris Mason sh.len = item_len; 2141ac8e9819SChris Mason sh.transid = found_transid; 2142ac8e9819SChris Mason 2143a48b73ecSJosef Bacik /* 2144a48b73ecSJosef Bacik * Copy search result header. If we fault then loop again so we 2145a48b73ecSJosef Bacik * can fault in the pages and -EFAULT there if there's a 2146a48b73ecSJosef Bacik * problem. Otherwise we'll fault and then copy the buffer in 2147a48b73ecSJosef Bacik * properly this next time through 2148a48b73ecSJosef Bacik */ 2149a48b73ecSJosef Bacik if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) { 2150a48b73ecSJosef Bacik ret = 0; 2151ba346b35SGerhard Heift goto out; 2152ba346b35SGerhard Heift } 2153ba346b35SGerhard Heift 2154ac8e9819SChris Mason *sk_offset += sizeof(sh); 2155ac8e9819SChris Mason 2156ac8e9819SChris Mason if (item_len) { 2157ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2158a48b73ecSJosef Bacik /* 2159a48b73ecSJosef Bacik * Copy the item, same behavior as above, but reset the 2160a48b73ecSJosef Bacik * * sk_offset so we copy the full thing again. 2161a48b73ecSJosef Bacik */ 2162a48b73ecSJosef Bacik if (read_extent_buffer_to_user_nofault(leaf, up, 2163ba346b35SGerhard Heift item_off, item_len)) { 2164a48b73ecSJosef Bacik ret = 0; 2165a48b73ecSJosef Bacik *sk_offset -= sizeof(sh); 2166ba346b35SGerhard Heift goto out; 2167ba346b35SGerhard Heift } 2168ba346b35SGerhard Heift 2169ac8e9819SChris Mason *sk_offset += item_len; 2170ac8e9819SChris Mason } 2171e2156867SHugo Mills (*num_found)++; 2172ac8e9819SChris Mason 21738f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21748f5f6178SGerhard Heift goto out; 21758f5f6178SGerhard Heift 217625c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 217725c9bc2eSGerhard Heift ret = 1; 217825c9bc2eSGerhard Heift goto out; 217925c9bc2eSGerhard Heift } 2180ac8e9819SChris Mason } 2181ac8e9819SChris Mason advance_key: 2182ac8e9819SChris Mason ret = 0; 2183dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2184dd81d459SNaohiro Aota test.type = sk->max_type; 2185dd81d459SNaohiro Aota test.offset = sk->max_offset; 2186dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2187dd81d459SNaohiro Aota ret = 1; 2188dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2189abc6e134SChris Mason key->offset++; 2190dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2191abc6e134SChris Mason key->offset = 0; 2192abc6e134SChris Mason key->type++; 2193dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2194abc6e134SChris Mason key->offset = 0; 2195abc6e134SChris Mason key->type = 0; 2196abc6e134SChris Mason key->objectid++; 2197abc6e134SChris Mason } else 2198abc6e134SChris Mason ret = 1; 219925c9bc2eSGerhard Heift out: 2200ba346b35SGerhard Heift /* 2201ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2202ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2203ba346b35SGerhard Heift * * all items were found 2204ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2205ba346b35SGerhard Heift * leaf 2206ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2207ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2208ba346b35SGerhard Heift */ 2209ac8e9819SChris Mason return ret; 2210ac8e9819SChris Mason } 2211ac8e9819SChris Mason 2212ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 221312544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 22149b6e817dSGerhard Heift size_t *buf_size, 2215ba346b35SGerhard Heift char __user *ubuf) 2216ac8e9819SChris Mason { 22170b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2218ac8e9819SChris Mason struct btrfs_root *root; 2219ac8e9819SChris Mason struct btrfs_key key; 2220ac8e9819SChris Mason struct btrfs_path *path; 2221ac8e9819SChris Mason int ret; 2222ac8e9819SChris Mason int num_found = 0; 2223ac8e9819SChris Mason unsigned long sk_offset = 0; 2224ac8e9819SChris Mason 22259b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 22269b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 222712544442SGerhard Heift return -EOVERFLOW; 22289b6e817dSGerhard Heift } 222912544442SGerhard Heift 2230ac8e9819SChris Mason path = btrfs_alloc_path(); 2231ac8e9819SChris Mason if (!path) 2232ac8e9819SChris Mason return -ENOMEM; 2233ac8e9819SChris Mason 2234ac8e9819SChris Mason if (sk->tree_id == 0) { 2235ac8e9819SChris Mason /* search the root of the inode that was passed */ 223600246528SJosef Bacik root = btrfs_grab_root(BTRFS_I(inode)->root); 2237ac8e9819SChris Mason } else { 223856e9357aSDavid Sterba root = btrfs_get_fs_root(info, sk->tree_id, true); 2239ac8e9819SChris Mason if (IS_ERR(root)) { 2240ac8e9819SChris Mason btrfs_free_path(path); 2241ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2242ac8e9819SChris Mason } 2243ac8e9819SChris Mason } 2244ac8e9819SChris Mason 2245ac8e9819SChris Mason key.objectid = sk->min_objectid; 2246ac8e9819SChris Mason key.type = sk->min_type; 2247ac8e9819SChris Mason key.offset = sk->min_offset; 2248ac8e9819SChris Mason 2249ac8e9819SChris Mason while (1) { 2250bb523b40SAndreas Gruenbacher ret = -EFAULT; 2251bb523b40SAndreas Gruenbacher if (fault_in_writeable(ubuf + sk_offset, *buf_size - sk_offset)) 2252a48b73ecSJosef Bacik break; 2253a48b73ecSJosef Bacik 22546174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2255ac8e9819SChris Mason if (ret != 0) { 2256ac8e9819SChris Mason if (ret > 0) 2257ac8e9819SChris Mason ret = 0; 2258ac8e9819SChris Mason goto err; 2259ac8e9819SChris Mason } 2260df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2261ac8e9819SChris Mason &sk_offset, &num_found); 2262b3b4aa74SDavid Sterba btrfs_release_path(path); 226325c9bc2eSGerhard Heift if (ret) 2264ac8e9819SChris Mason break; 2265ac8e9819SChris Mason 2266ac8e9819SChris Mason } 22678f5f6178SGerhard Heift if (ret > 0) 2268ac8e9819SChris Mason ret = 0; 2269ac8e9819SChris Mason err: 2270ac8e9819SChris Mason sk->nr_items = num_found; 227100246528SJosef Bacik btrfs_put_root(root); 2272ac8e9819SChris Mason btrfs_free_path(path); 2273ac8e9819SChris Mason return ret; 2274ac8e9819SChris Mason } 2275ac8e9819SChris Mason 2276ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2277ac8e9819SChris Mason void __user *argp) 2278ac8e9819SChris Mason { 2279ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2280ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2281ac8e9819SChris Mason struct inode *inode; 2282ac8e9819SChris Mason int ret; 22839b6e817dSGerhard Heift size_t buf_size; 2284ac8e9819SChris Mason 2285ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2286ac8e9819SChris Mason return -EPERM; 2287ac8e9819SChris Mason 2288ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2289ac8e9819SChris Mason 2290ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2291ba346b35SGerhard Heift return -EFAULT; 2292ba346b35SGerhard Heift 2293ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2294ac8e9819SChris Mason 2295496ad9aaSAl Viro inode = file_inode(file); 2296ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22978f5f6178SGerhard Heift 22988f5f6178SGerhard Heift /* 22998f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 23008f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 23018f5f6178SGerhard Heift */ 23028f5f6178SGerhard Heift if (ret == -EOVERFLOW) 23038f5f6178SGerhard Heift ret = 0; 23048f5f6178SGerhard Heift 2305ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2306ac8e9819SChris Mason ret = -EFAULT; 2307ac8e9819SChris Mason return ret; 2308ac8e9819SChris Mason } 2309ac8e9819SChris Mason 2310cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2311cc68a8a5SGerhard Heift void __user *argp) 2312cc68a8a5SGerhard Heift { 2313cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2314cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2315cc68a8a5SGerhard Heift struct inode *inode; 2316cc68a8a5SGerhard Heift int ret; 2317cc68a8a5SGerhard Heift size_t buf_size; 2318ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2319cc68a8a5SGerhard Heift 2320cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2321cc68a8a5SGerhard Heift return -EPERM; 2322cc68a8a5SGerhard Heift 2323cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2324cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2325cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2326cc68a8a5SGerhard Heift return -EFAULT; 2327cc68a8a5SGerhard Heift 2328cc68a8a5SGerhard Heift buf_size = args.buf_size; 2329cc68a8a5SGerhard Heift 2330cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2331cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2332cc68a8a5SGerhard Heift buf_size = buf_limit; 2333cc68a8a5SGerhard Heift 2334cc68a8a5SGerhard Heift inode = file_inode(file); 2335cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2336718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2337cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2338cc68a8a5SGerhard Heift ret = -EFAULT; 2339cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2340cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2341cc68a8a5SGerhard Heift ret = -EFAULT; 2342cc68a8a5SGerhard Heift 234376dda93cSYan, Zheng return ret; 234476dda93cSYan, Zheng } 234576dda93cSYan, Zheng 234698d377a0STARUISI Hiroaki /* 2347ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2348ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 234998d377a0STARUISI Hiroaki */ 235098d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 235198d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 235298d377a0STARUISI Hiroaki { 235398d377a0STARUISI Hiroaki struct btrfs_root *root; 235498d377a0STARUISI Hiroaki struct btrfs_key key; 2355ac8e9819SChris Mason char *ptr; 235698d377a0STARUISI Hiroaki int ret = -1; 235798d377a0STARUISI Hiroaki int slot; 235898d377a0STARUISI Hiroaki int len; 235998d377a0STARUISI Hiroaki int total_len = 0; 236098d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 236198d377a0STARUISI Hiroaki struct extent_buffer *l; 236298d377a0STARUISI Hiroaki struct btrfs_path *path; 236398d377a0STARUISI Hiroaki 236498d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 236598d377a0STARUISI Hiroaki name[0]='\0'; 236698d377a0STARUISI Hiroaki return 0; 236798d377a0STARUISI Hiroaki } 236898d377a0STARUISI Hiroaki 236998d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 237098d377a0STARUISI Hiroaki if (!path) 237198d377a0STARUISI Hiroaki return -ENOMEM; 237298d377a0STARUISI Hiroaki 2373c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 237498d377a0STARUISI Hiroaki 237556e9357aSDavid Sterba root = btrfs_get_fs_root(info, tree_id, true); 237698d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2377ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 237888234012SJosef Bacik root = NULL; 237988234012SJosef Bacik goto out; 238088234012SJosef Bacik } 238198d377a0STARUISI Hiroaki 238298d377a0STARUISI Hiroaki key.objectid = dirid; 238398d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23848ad6fcabSChris Mason key.offset = (u64)-1; 238598d377a0STARUISI Hiroaki 238698d377a0STARUISI Hiroaki while (1) { 23870ff40a91SMarcos Paulo de Souza ret = btrfs_search_backwards(root, &key, path); 238818674c6cSFilipe David Borba Manana if (ret < 0) 238918674c6cSFilipe David Borba Manana goto out; 239018674c6cSFilipe David Borba Manana else if (ret > 0) { 2391ac8e9819SChris Mason ret = -ENOENT; 239298d377a0STARUISI Hiroaki goto out; 2393ac8e9819SChris Mason } 239418674c6cSFilipe David Borba Manana 239518674c6cSFilipe David Borba Manana l = path->nodes[0]; 239618674c6cSFilipe David Borba Manana slot = path->slots[0]; 239798d377a0STARUISI Hiroaki 239898d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 239998d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 240098d377a0STARUISI Hiroaki ptr -= len + 1; 240198d377a0STARUISI Hiroaki total_len += len + 1; 2402a696cf35SFilipe David Borba Manana if (ptr < name) { 2403a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 240498d377a0STARUISI Hiroaki goto out; 2405a696cf35SFilipe David Borba Manana } 240698d377a0STARUISI Hiroaki 240798d377a0STARUISI Hiroaki *(ptr + len) = '/'; 240898d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 240998d377a0STARUISI Hiroaki 241098d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 241198d377a0STARUISI Hiroaki break; 241298d377a0STARUISI Hiroaki 2413b3b4aa74SDavid Sterba btrfs_release_path(path); 241498d377a0STARUISI Hiroaki key.objectid = key.offset; 24158ad6fcabSChris Mason key.offset = (u64)-1; 241698d377a0STARUISI Hiroaki dirid = key.objectid; 241798d377a0STARUISI Hiroaki } 241877906a50SLi Zefan memmove(name, ptr, total_len); 241998d377a0STARUISI Hiroaki name[total_len] = '\0'; 242098d377a0STARUISI Hiroaki ret = 0; 242198d377a0STARUISI Hiroaki out: 242200246528SJosef Bacik btrfs_put_root(root); 242398d377a0STARUISI Hiroaki btrfs_free_path(path); 2424ac8e9819SChris Mason return ret; 2425ac8e9819SChris Mason } 2426ac8e9819SChris Mason 24276623d9a0SChristian Brauner static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns, 24286623d9a0SChristian Brauner struct inode *inode, 242923d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 243023d0b79dSTomohiro Misono { 243123d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 243223d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 243323d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 243423d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 243523d0b79dSTomohiro Misono u64 dirid = args->dirid; 243623d0b79dSTomohiro Misono unsigned long item_off; 243723d0b79dSTomohiro Misono unsigned long item_len; 243823d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 243923d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 2440b8a49ae1SJosef Bacik struct btrfs_root *root = NULL; 244123d0b79dSTomohiro Misono struct btrfs_path *path; 244223d0b79dSTomohiro Misono struct btrfs_key key, key2; 244323d0b79dSTomohiro Misono struct extent_buffer *leaf; 244423d0b79dSTomohiro Misono struct inode *temp_inode; 244523d0b79dSTomohiro Misono char *ptr; 244623d0b79dSTomohiro Misono int slot; 244723d0b79dSTomohiro Misono int len; 244823d0b79dSTomohiro Misono int total_len = 0; 244923d0b79dSTomohiro Misono int ret; 245023d0b79dSTomohiro Misono 245123d0b79dSTomohiro Misono path = btrfs_alloc_path(); 245223d0b79dSTomohiro Misono if (!path) 245323d0b79dSTomohiro Misono return -ENOMEM; 245423d0b79dSTomohiro Misono 245523d0b79dSTomohiro Misono /* 245623d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 245723d0b79dSTomohiro Misono * construct the path in from the bottom up. 245823d0b79dSTomohiro Misono */ 245923d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 246023d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 246123d0b79dSTomohiro Misono 246256e9357aSDavid Sterba root = btrfs_get_fs_root(fs_info, treeid, true); 246323d0b79dSTomohiro Misono if (IS_ERR(root)) { 246423d0b79dSTomohiro Misono ret = PTR_ERR(root); 246523d0b79dSTomohiro Misono goto out; 246623d0b79dSTomohiro Misono } 246723d0b79dSTomohiro Misono 246823d0b79dSTomohiro Misono key.objectid = dirid; 246923d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 247023d0b79dSTomohiro Misono key.offset = (u64)-1; 247123d0b79dSTomohiro Misono while (1) { 24720ff40a91SMarcos Paulo de Souza ret = btrfs_search_backwards(root, &key, path); 24730ff40a91SMarcos Paulo de Souza if (ret < 0) 2474b8a49ae1SJosef Bacik goto out_put; 24750ff40a91SMarcos Paulo de Souza else if (ret > 0) { 247623d0b79dSTomohiro Misono ret = -ENOENT; 2477b8a49ae1SJosef Bacik goto out_put; 247823d0b79dSTomohiro Misono } 247923d0b79dSTomohiro Misono 248023d0b79dSTomohiro Misono leaf = path->nodes[0]; 248123d0b79dSTomohiro Misono slot = path->slots[0]; 248223d0b79dSTomohiro Misono 248323d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 248423d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 248523d0b79dSTomohiro Misono ptr -= len + 1; 248623d0b79dSTomohiro Misono total_len += len + 1; 248723d0b79dSTomohiro Misono if (ptr < args->path) { 248823d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 2489b8a49ae1SJosef Bacik goto out_put; 249023d0b79dSTomohiro Misono } 249123d0b79dSTomohiro Misono 249223d0b79dSTomohiro Misono *(ptr + len) = '/'; 249323d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 249423d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 249523d0b79dSTomohiro Misono 249623d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 249723d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 249823d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 249923d0b79dSTomohiro Misono if (ret < 0) { 2500b8a49ae1SJosef Bacik goto out_put; 250123d0b79dSTomohiro Misono } else if (ret > 0) { 250223d0b79dSTomohiro Misono ret = -ENOENT; 2503b8a49ae1SJosef Bacik goto out_put; 250423d0b79dSTomohiro Misono } 250523d0b79dSTomohiro Misono 250623d0b79dSTomohiro Misono leaf = path->nodes[0]; 250723d0b79dSTomohiro Misono slot = path->slots[0]; 250823d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 250923d0b79dSTomohiro Misono if (key2.objectid != dirid) { 251023d0b79dSTomohiro Misono ret = -ENOENT; 2511b8a49ae1SJosef Bacik goto out_put; 251223d0b79dSTomohiro Misono } 251323d0b79dSTomohiro Misono 25140202e83fSDavid Sterba temp_inode = btrfs_iget(sb, key2.objectid, root); 25153ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 25163ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 2517b8a49ae1SJosef Bacik goto out_put; 25183ca57bd6SMisono Tomohiro } 25196623d9a0SChristian Brauner ret = inode_permission(mnt_userns, temp_inode, 252047291baaSChristian Brauner MAY_READ | MAY_EXEC); 252123d0b79dSTomohiro Misono iput(temp_inode); 252223d0b79dSTomohiro Misono if (ret) { 252323d0b79dSTomohiro Misono ret = -EACCES; 2524b8a49ae1SJosef Bacik goto out_put; 252523d0b79dSTomohiro Misono } 252623d0b79dSTomohiro Misono 252723d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 252823d0b79dSTomohiro Misono break; 252923d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 253023d0b79dSTomohiro Misono ret = -EACCES; 2531b8a49ae1SJosef Bacik goto out_put; 253223d0b79dSTomohiro Misono } 253323d0b79dSTomohiro Misono 253423d0b79dSTomohiro Misono btrfs_release_path(path); 253523d0b79dSTomohiro Misono key.objectid = key.offset; 253623d0b79dSTomohiro Misono key.offset = (u64)-1; 253723d0b79dSTomohiro Misono dirid = key.objectid; 253823d0b79dSTomohiro Misono } 253923d0b79dSTomohiro Misono 254023d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 254123d0b79dSTomohiro Misono args->path[total_len] = '\0'; 254200246528SJosef Bacik btrfs_put_root(root); 2543b8a49ae1SJosef Bacik root = NULL; 254423d0b79dSTomohiro Misono btrfs_release_path(path); 254523d0b79dSTomohiro Misono } 254623d0b79dSTomohiro Misono 254723d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 254823d0b79dSTomohiro Misono key.objectid = treeid; 254923d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 255023d0b79dSTomohiro Misono key.offset = args->treeid; 2551b8a49ae1SJosef Bacik ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 255223d0b79dSTomohiro Misono if (ret < 0) { 255323d0b79dSTomohiro Misono goto out; 255423d0b79dSTomohiro Misono } else if (ret > 0) { 255523d0b79dSTomohiro Misono ret = -ENOENT; 255623d0b79dSTomohiro Misono goto out; 255723d0b79dSTomohiro Misono } 255823d0b79dSTomohiro Misono 255923d0b79dSTomohiro Misono leaf = path->nodes[0]; 256023d0b79dSTomohiro Misono slot = path->slots[0]; 256123d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 256223d0b79dSTomohiro Misono 256323d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 25643212fa14SJosef Bacik item_len = btrfs_item_size(leaf, slot); 256523d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 256623d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 256723d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 256823d0b79dSTomohiro Misono ret = -EINVAL; 256923d0b79dSTomohiro Misono goto out; 257023d0b79dSTomohiro Misono } 257123d0b79dSTomohiro Misono 257223d0b79dSTomohiro Misono /* Copy subvolume's name */ 257323d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 257423d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 257523d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 257623d0b79dSTomohiro Misono args->name[item_len] = 0; 257723d0b79dSTomohiro Misono 2578b8a49ae1SJosef Bacik out_put: 257900246528SJosef Bacik btrfs_put_root(root); 258023d0b79dSTomohiro Misono out: 258123d0b79dSTomohiro Misono btrfs_free_path(path); 258223d0b79dSTomohiro Misono return ret; 258323d0b79dSTomohiro Misono } 258423d0b79dSTomohiro Misono 2585ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2586ac8e9819SChris Mason void __user *argp) 2587ac8e9819SChris Mason { 2588ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2589ac8e9819SChris Mason struct inode *inode; 259001b810b8SDavid Sterba int ret = 0; 2591ac8e9819SChris Mason 25922354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25932354d08fSJulia Lawall if (IS_ERR(args)) 25942354d08fSJulia Lawall return PTR_ERR(args); 2595c2b96929SDan Carpenter 2596496ad9aaSAl Viro inode = file_inode(file); 2597ac8e9819SChris Mason 259801b810b8SDavid Sterba /* 259901b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 260001b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 260101b810b8SDavid Sterba */ 26021b53ac4dSChris Mason if (args->treeid == 0) 26031b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 26041b53ac4dSChris Mason 260501b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 260601b810b8SDavid Sterba args->name[0] = 0; 260701b810b8SDavid Sterba goto out; 260801b810b8SDavid Sterba } 260901b810b8SDavid Sterba 261001b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 261101b810b8SDavid Sterba ret = -EPERM; 261201b810b8SDavid Sterba goto out; 261301b810b8SDavid Sterba } 261401b810b8SDavid Sterba 2615ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2616ac8e9819SChris Mason args->treeid, args->objectid, 2617ac8e9819SChris Mason args->name); 2618ac8e9819SChris Mason 261901b810b8SDavid Sterba out: 2620ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2621ac8e9819SChris Mason ret = -EFAULT; 2622ac8e9819SChris Mason 2623ac8e9819SChris Mason kfree(args); 262498d377a0STARUISI Hiroaki return ret; 262598d377a0STARUISI Hiroaki } 262698d377a0STARUISI Hiroaki 262723d0b79dSTomohiro Misono /* 262823d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 262923d0b79dSTomohiro Misono * 263023d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 263123d0b79dSTomohiro Misono * 263223d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 263323d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 263423d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 263523d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 263623d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 263723d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 263823d0b79dSTomohiro Misono */ 263923d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 264023d0b79dSTomohiro Misono { 264123d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 264223d0b79dSTomohiro Misono struct inode *inode; 264323d0b79dSTomohiro Misono int ret; 264423d0b79dSTomohiro Misono 264523d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 264623d0b79dSTomohiro Misono if (IS_ERR(args)) 264723d0b79dSTomohiro Misono return PTR_ERR(args); 264823d0b79dSTomohiro Misono 264923d0b79dSTomohiro Misono inode = file_inode(file); 265023d0b79dSTomohiro Misono 265123d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 265223d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 265323d0b79dSTomohiro Misono /* 265423d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 265523d0b79dSTomohiro Misono * called 265623d0b79dSTomohiro Misono */ 265723d0b79dSTomohiro Misono kfree(args); 265823d0b79dSTomohiro Misono return -EACCES; 265923d0b79dSTomohiro Misono } 266023d0b79dSTomohiro Misono 26616623d9a0SChristian Brauner ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args); 266223d0b79dSTomohiro Misono 266323d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 266423d0b79dSTomohiro Misono ret = -EFAULT; 266523d0b79dSTomohiro Misono 266623d0b79dSTomohiro Misono kfree(args); 266723d0b79dSTomohiro Misono return ret; 266823d0b79dSTomohiro Misono } 266923d0b79dSTomohiro Misono 2670b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2671b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2672b64ec075STomohiro Misono { 2673b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2674b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2675b64ec075STomohiro Misono struct btrfs_root *root; 2676b64ec075STomohiro Misono struct btrfs_path *path; 2677b64ec075STomohiro Misono struct btrfs_key key; 2678b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2679b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2680b64ec075STomohiro Misono struct extent_buffer *leaf; 2681b64ec075STomohiro Misono unsigned long item_off; 2682b64ec075STomohiro Misono unsigned long item_len; 2683b64ec075STomohiro Misono struct inode *inode; 2684b64ec075STomohiro Misono int slot; 2685b64ec075STomohiro Misono int ret = 0; 2686b64ec075STomohiro Misono 2687b64ec075STomohiro Misono path = btrfs_alloc_path(); 2688b64ec075STomohiro Misono if (!path) 2689b64ec075STomohiro Misono return -ENOMEM; 2690b64ec075STomohiro Misono 2691b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2692b64ec075STomohiro Misono if (!subvol_info) { 2693b64ec075STomohiro Misono btrfs_free_path(path); 2694b64ec075STomohiro Misono return -ENOMEM; 2695b64ec075STomohiro Misono } 2696b64ec075STomohiro Misono 2697b64ec075STomohiro Misono inode = file_inode(file); 2698b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2699b64ec075STomohiro Misono 2700b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2701b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 270256e9357aSDavid Sterba root = btrfs_get_fs_root(fs_info, key.objectid, true); 2703b64ec075STomohiro Misono if (IS_ERR(root)) { 2704b64ec075STomohiro Misono ret = PTR_ERR(root); 270504734e84SJosef Bacik goto out_free; 270604734e84SJosef Bacik } 2707b64ec075STomohiro Misono root_item = &root->root_item; 2708b64ec075STomohiro Misono 2709b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2710b64ec075STomohiro Misono 2711b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2712b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2713b64ec075STomohiro Misono 2714b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2715b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2716b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2717b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2718b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2719b64ec075STomohiro Misono 2720b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2721b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2722b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2723b64ec075STomohiro Misono 2724b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2725b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2726b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2727b64ec075STomohiro Misono 2728b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2729b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2730b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2731b64ec075STomohiro Misono 2732b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2733b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2734b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2735b64ec075STomohiro Misono 2736b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2737b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2738b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2739b64ec075STomohiro Misono key.offset = 0; 274004734e84SJosef Bacik ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 2741b64ec075STomohiro Misono if (ret < 0) { 2742b64ec075STomohiro Misono goto out; 2743b64ec075STomohiro Misono } else if (path->slots[0] >= 2744b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 274504734e84SJosef Bacik ret = btrfs_next_leaf(fs_info->tree_root, path); 2746b64ec075STomohiro Misono if (ret < 0) { 2747b64ec075STomohiro Misono goto out; 2748b64ec075STomohiro Misono } else if (ret > 0) { 2749b64ec075STomohiro Misono ret = -EUCLEAN; 2750b64ec075STomohiro Misono goto out; 2751b64ec075STomohiro Misono } 2752b64ec075STomohiro Misono } 2753b64ec075STomohiro Misono 2754b64ec075STomohiro Misono leaf = path->nodes[0]; 2755b64ec075STomohiro Misono slot = path->slots[0]; 2756b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2757b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2758b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2759b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2760b64ec075STomohiro Misono 2761b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2762b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2763b64ec075STomohiro Misono 2764b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2765b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 27663212fa14SJosef Bacik item_len = btrfs_item_size(leaf, slot) 2767b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2768b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2769b64ec075STomohiro Misono item_off, item_len); 2770b64ec075STomohiro Misono } else { 2771b64ec075STomohiro Misono ret = -ENOENT; 2772b64ec075STomohiro Misono goto out; 2773b64ec075STomohiro Misono } 2774b64ec075STomohiro Misono } 2775b64ec075STomohiro Misono 2776b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2777b64ec075STomohiro Misono ret = -EFAULT; 2778b64ec075STomohiro Misono 2779b64ec075STomohiro Misono out: 278000246528SJosef Bacik btrfs_put_root(root); 278104734e84SJosef Bacik out_free: 2782b64ec075STomohiro Misono btrfs_free_path(path); 2783b091f7feSWaiman Long kfree(subvol_info); 2784b64ec075STomohiro Misono return ret; 2785b64ec075STomohiro Misono } 2786b64ec075STomohiro Misono 278742e4b520STomohiro Misono /* 278842e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 278942e4b520STomohiro Misono * except the subvolume name. 279042e4b520STomohiro Misono */ 279142e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 279242e4b520STomohiro Misono { 279342e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 279442e4b520STomohiro Misono struct btrfs_root_ref *rref; 279542e4b520STomohiro Misono struct btrfs_root *root; 279642e4b520STomohiro Misono struct btrfs_path *path; 279742e4b520STomohiro Misono struct btrfs_key key; 279842e4b520STomohiro Misono struct extent_buffer *leaf; 279942e4b520STomohiro Misono struct inode *inode; 280042e4b520STomohiro Misono u64 objectid; 280142e4b520STomohiro Misono int slot; 280242e4b520STomohiro Misono int ret; 280342e4b520STomohiro Misono u8 found; 280442e4b520STomohiro Misono 280542e4b520STomohiro Misono path = btrfs_alloc_path(); 280642e4b520STomohiro Misono if (!path) 280742e4b520STomohiro Misono return -ENOMEM; 280842e4b520STomohiro Misono 280942e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 281042e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 281142e4b520STomohiro Misono btrfs_free_path(path); 281242e4b520STomohiro Misono return PTR_ERR(rootrefs); 281342e4b520STomohiro Misono } 281442e4b520STomohiro Misono 281542e4b520STomohiro Misono inode = file_inode(file); 281642e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 281742e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 281842e4b520STomohiro Misono 281942e4b520STomohiro Misono key.objectid = objectid; 282042e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 282142e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 282242e4b520STomohiro Misono found = 0; 282342e4b520STomohiro Misono 282442e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 282542e4b520STomohiro Misono if (ret < 0) { 282642e4b520STomohiro Misono goto out; 282742e4b520STomohiro Misono } else if (path->slots[0] >= 282842e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 282942e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 283042e4b520STomohiro Misono if (ret < 0) { 283142e4b520STomohiro Misono goto out; 283242e4b520STomohiro Misono } else if (ret > 0) { 283342e4b520STomohiro Misono ret = -EUCLEAN; 283442e4b520STomohiro Misono goto out; 283542e4b520STomohiro Misono } 283642e4b520STomohiro Misono } 283742e4b520STomohiro Misono while (1) { 283842e4b520STomohiro Misono leaf = path->nodes[0]; 283942e4b520STomohiro Misono slot = path->slots[0]; 284042e4b520STomohiro Misono 284142e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 284242e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 284342e4b520STomohiro Misono ret = 0; 284442e4b520STomohiro Misono goto out; 284542e4b520STomohiro Misono } 284642e4b520STomohiro Misono 284742e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 284842e4b520STomohiro Misono ret = -EOVERFLOW; 284942e4b520STomohiro Misono goto out; 285042e4b520STomohiro Misono } 285142e4b520STomohiro Misono 285242e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 285342e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 285442e4b520STomohiro Misono rootrefs->rootref[found].dirid = 285542e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 285642e4b520STomohiro Misono found++; 285742e4b520STomohiro Misono 285842e4b520STomohiro Misono ret = btrfs_next_item(root, path); 285942e4b520STomohiro Misono if (ret < 0) { 286042e4b520STomohiro Misono goto out; 286142e4b520STomohiro Misono } else if (ret > 0) { 286242e4b520STomohiro Misono ret = -EUCLEAN; 286342e4b520STomohiro Misono goto out; 286442e4b520STomohiro Misono } 286542e4b520STomohiro Misono } 286642e4b520STomohiro Misono 286742e4b520STomohiro Misono out: 286842e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 286942e4b520STomohiro Misono rootrefs->num_items = found; 287042e4b520STomohiro Misono /* update min_treeid for next search */ 287142e4b520STomohiro Misono if (found) 287242e4b520STomohiro Misono rootrefs->min_treeid = 287342e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 287442e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 287542e4b520STomohiro Misono ret = -EFAULT; 287642e4b520STomohiro Misono } 287742e4b520STomohiro Misono 287842e4b520STomohiro Misono kfree(rootrefs); 287942e4b520STomohiro Misono btrfs_free_path(path); 288042e4b520STomohiro Misono 288142e4b520STomohiro Misono return ret; 288242e4b520STomohiro Misono } 288342e4b520STomohiro Misono 288476dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 2885949964c9SMarcos Paulo de Souza void __user *arg, 2886949964c9SMarcos Paulo de Souza bool destroy_v2) 288776dda93cSYan, Zheng { 288854563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 289076dda93cSYan, Zheng struct dentry *dentry; 28912b0143b5SDavid Howells struct inode *dir = d_inode(parent); 289276dda93cSYan, Zheng struct inode *inode; 289376dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 289476dda93cSYan, Zheng struct btrfs_root *dest = NULL; 2895949964c9SMarcos Paulo de Souza struct btrfs_ioctl_vol_args *vol_args = NULL; 2896949964c9SMarcos Paulo de Souza struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL; 2897c4ed533bSChristian Brauner struct user_namespace *mnt_userns = file_mnt_user_ns(file); 2898949964c9SMarcos Paulo de Souza char *subvol_name, *subvol_name_ptr = NULL; 2899949964c9SMarcos Paulo de Souza int subvol_namelen; 290076dda93cSYan, Zheng int err = 0; 2901949964c9SMarcos Paulo de Souza bool destroy_parent = false; 290276dda93cSYan, Zheng 2903949964c9SMarcos Paulo de Souza if (destroy_v2) { 2904949964c9SMarcos Paulo de Souza vol_args2 = memdup_user(arg, sizeof(*vol_args2)); 2905949964c9SMarcos Paulo de Souza if (IS_ERR(vol_args2)) 2906949964c9SMarcos Paulo de Souza return PTR_ERR(vol_args2); 2907325c50e3SJeff Mahoney 2908949964c9SMarcos Paulo de Souza if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) { 2909949964c9SMarcos Paulo de Souza err = -EOPNOTSUPP; 2910949964c9SMarcos Paulo de Souza goto out; 2911949964c9SMarcos Paulo de Souza } 291276dda93cSYan, Zheng 2913949964c9SMarcos Paulo de Souza /* 2914949964c9SMarcos Paulo de Souza * If SPEC_BY_ID is not set, we are looking for the subvolume by 2915949964c9SMarcos Paulo de Souza * name, same as v1 currently does. 2916949964c9SMarcos Paulo de Souza */ 2917949964c9SMarcos Paulo de Souza if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { 2918949964c9SMarcos Paulo de Souza vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0; 2919949964c9SMarcos Paulo de Souza subvol_name = vol_args2->name; 2920949964c9SMarcos Paulo de Souza 2921949964c9SMarcos Paulo de Souza err = mnt_want_write_file(file); 2922949964c9SMarcos Paulo de Souza if (err) 2923949964c9SMarcos Paulo de Souza goto out; 2924949964c9SMarcos Paulo de Souza } else { 2925aabb34e7SChristian Brauner struct inode *old_dir; 2926c4ed533bSChristian Brauner 2927949964c9SMarcos Paulo de Souza if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) { 292876dda93cSYan, Zheng err = -EINVAL; 292976dda93cSYan, Zheng goto out; 293076dda93cSYan, Zheng } 293176dda93cSYan, Zheng 2932a561be71SAl Viro err = mnt_want_write_file(file); 293376dda93cSYan, Zheng if (err) 293476dda93cSYan, Zheng goto out; 293576dda93cSYan, Zheng 2936949964c9SMarcos Paulo de Souza dentry = btrfs_get_dentry(fs_info->sb, 2937949964c9SMarcos Paulo de Souza BTRFS_FIRST_FREE_OBJECTID, 2938949964c9SMarcos Paulo de Souza vol_args2->subvolid, 0, 0); 2939949964c9SMarcos Paulo de Souza if (IS_ERR(dentry)) { 2940949964c9SMarcos Paulo de Souza err = PTR_ERR(dentry); 2941949964c9SMarcos Paulo de Souza goto out_drop_write; 2942949964c9SMarcos Paulo de Souza } 2943949964c9SMarcos Paulo de Souza 2944949964c9SMarcos Paulo de Souza /* 2945949964c9SMarcos Paulo de Souza * Change the default parent since the subvolume being 2946949964c9SMarcos Paulo de Souza * deleted can be outside of the current mount point. 2947949964c9SMarcos Paulo de Souza */ 2948949964c9SMarcos Paulo de Souza parent = btrfs_get_parent(dentry); 2949949964c9SMarcos Paulo de Souza 2950949964c9SMarcos Paulo de Souza /* 2951949964c9SMarcos Paulo de Souza * At this point dentry->d_name can point to '/' if the 2952949964c9SMarcos Paulo de Souza * subvolume we want to destroy is outsite of the 2953949964c9SMarcos Paulo de Souza * current mount point, so we need to release the 2954949964c9SMarcos Paulo de Souza * current dentry and execute the lookup to return a new 2955949964c9SMarcos Paulo de Souza * one with ->d_name pointing to the 2956949964c9SMarcos Paulo de Souza * <mount point>/subvol_name. 2957949964c9SMarcos Paulo de Souza */ 2958949964c9SMarcos Paulo de Souza dput(dentry); 2959949964c9SMarcos Paulo de Souza if (IS_ERR(parent)) { 2960949964c9SMarcos Paulo de Souza err = PTR_ERR(parent); 2961949964c9SMarcos Paulo de Souza goto out_drop_write; 2962949964c9SMarcos Paulo de Souza } 2963aabb34e7SChristian Brauner old_dir = dir; 2964949964c9SMarcos Paulo de Souza dir = d_inode(parent); 2965949964c9SMarcos Paulo de Souza 2966949964c9SMarcos Paulo de Souza /* 2967949964c9SMarcos Paulo de Souza * If v2 was used with SPEC_BY_ID, a new parent was 2968949964c9SMarcos Paulo de Souza * allocated since the subvolume can be outside of the 2969949964c9SMarcos Paulo de Souza * current mount point. Later on we need to release this 2970949964c9SMarcos Paulo de Souza * new parent dentry. 2971949964c9SMarcos Paulo de Souza */ 2972949964c9SMarcos Paulo de Souza destroy_parent = true; 2973949964c9SMarcos Paulo de Souza 2974aabb34e7SChristian Brauner /* 2975aabb34e7SChristian Brauner * On idmapped mounts, deletion via subvolid is 2976aabb34e7SChristian Brauner * restricted to subvolumes that are immediate 2977aabb34e7SChristian Brauner * ancestors of the inode referenced by the file 2978aabb34e7SChristian Brauner * descriptor in the ioctl. Otherwise the idmapping 2979aabb34e7SChristian Brauner * could potentially be abused to delete subvolumes 2980aabb34e7SChristian Brauner * anywhere in the filesystem the user wouldn't be able 2981aabb34e7SChristian Brauner * to delete without an idmapped mount. 2982aabb34e7SChristian Brauner */ 2983aabb34e7SChristian Brauner if (old_dir != dir && mnt_userns != &init_user_ns) { 2984aabb34e7SChristian Brauner err = -EOPNOTSUPP; 2985aabb34e7SChristian Brauner goto free_parent; 2986aabb34e7SChristian Brauner } 2987aabb34e7SChristian Brauner 2988949964c9SMarcos Paulo de Souza subvol_name_ptr = btrfs_get_subvol_name_from_objectid( 2989949964c9SMarcos Paulo de Souza fs_info, vol_args2->subvolid); 2990949964c9SMarcos Paulo de Souza if (IS_ERR(subvol_name_ptr)) { 2991949964c9SMarcos Paulo de Souza err = PTR_ERR(subvol_name_ptr); 2992949964c9SMarcos Paulo de Souza goto free_parent; 2993949964c9SMarcos Paulo de Souza } 29941a9fd417SDavid Sterba /* subvol_name_ptr is already nul terminated */ 2995949964c9SMarcos Paulo de Souza subvol_name = (char *)kbasename(subvol_name_ptr); 2996949964c9SMarcos Paulo de Souza } 2997949964c9SMarcos Paulo de Souza } else { 2998949964c9SMarcos Paulo de Souza vol_args = memdup_user(arg, sizeof(*vol_args)); 2999949964c9SMarcos Paulo de Souza if (IS_ERR(vol_args)) 3000949964c9SMarcos Paulo de Souza return PTR_ERR(vol_args); 3001949964c9SMarcos Paulo de Souza 3002949964c9SMarcos Paulo de Souza vol_args->name[BTRFS_PATH_NAME_MAX] = 0; 3003949964c9SMarcos Paulo de Souza subvol_name = vol_args->name; 3004949964c9SMarcos Paulo de Souza 3005949964c9SMarcos Paulo de Souza err = mnt_want_write_file(file); 3006949964c9SMarcos Paulo de Souza if (err) 3007949964c9SMarcos Paulo de Souza goto out; 3008949964c9SMarcos Paulo de Souza } 3009949964c9SMarcos Paulo de Souza 3010949964c9SMarcos Paulo de Souza subvol_namelen = strlen(subvol_name); 3011949964c9SMarcos Paulo de Souza 3012949964c9SMarcos Paulo de Souza if (strchr(subvol_name, '/') || 3013949964c9SMarcos Paulo de Souza strncmp(subvol_name, "..", subvol_namelen) == 0) { 3014949964c9SMarcos Paulo de Souza err = -EINVAL; 3015949964c9SMarcos Paulo de Souza goto free_subvol_name; 3016949964c9SMarcos Paulo de Souza } 3017949964c9SMarcos Paulo de Souza 3018949964c9SMarcos Paulo de Souza if (!S_ISDIR(dir->i_mode)) { 3019949964c9SMarcos Paulo de Souza err = -ENOTDIR; 3020949964c9SMarcos Paulo de Souza goto free_subvol_name; 3021949964c9SMarcos Paulo de Souza } 3022521e0546SDavid Sterba 302300235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 302400235411SAl Viro if (err == -EINTR) 3025949964c9SMarcos Paulo de Souza goto free_subvol_name; 3026c4ed533bSChristian Brauner dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen); 302776dda93cSYan, Zheng if (IS_ERR(dentry)) { 302876dda93cSYan, Zheng err = PTR_ERR(dentry); 302976dda93cSYan, Zheng goto out_unlock_dir; 303076dda93cSYan, Zheng } 303176dda93cSYan, Zheng 30322b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 303376dda93cSYan, Zheng err = -ENOENT; 303476dda93cSYan, Zheng goto out_dput; 303576dda93cSYan, Zheng } 303676dda93cSYan, Zheng 30372b0143b5SDavid Howells inode = d_inode(dentry); 30384260f7c7SSage Weil dest = BTRFS_I(inode)->root; 30394260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 30404260f7c7SSage Weil /* 30414260f7c7SSage Weil * Regular user. Only allow this with a special mount 30424260f7c7SSage Weil * option, when the user has write+exec access to the 30434260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 30444260f7c7SSage Weil * allowed. 30454260f7c7SSage Weil * 30464260f7c7SSage Weil * Note that this is _not_ check that the subvol is 30474260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 30484260f7c7SSage Weil * otherwise be able to delete. 30494260f7c7SSage Weil * 30504260f7c7SSage Weil * Users who want to delete empty subvols should try 30514260f7c7SSage Weil * rmdir(2). 30524260f7c7SSage Weil */ 30534260f7c7SSage Weil err = -EPERM; 30540b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 30554260f7c7SSage Weil goto out_dput; 30564260f7c7SSage Weil 30574260f7c7SSage Weil /* 30584260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 30594260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 30604260f7c7SSage Weil * must be called on the dentry referencing the root 30614260f7c7SSage Weil * of the subvol, not a random directory contained 30624260f7c7SSage Weil * within it. 30634260f7c7SSage Weil */ 30644260f7c7SSage Weil err = -EINVAL; 30654260f7c7SSage Weil if (root == dest) 30664260f7c7SSage Weil goto out_dput; 30674260f7c7SSage Weil 3068c4ed533bSChristian Brauner err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC); 30694260f7c7SSage Weil if (err) 30704260f7c7SSage Weil goto out_dput; 30715c39da5bSMiao Xie } 30724260f7c7SSage Weil 30735c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 3074c4ed533bSChristian Brauner err = btrfs_may_delete(mnt_userns, dir, dentry, 1); 30754260f7c7SSage Weil if (err) 30764260f7c7SSage Weil goto out_dput; 30774260f7c7SSage Weil 30784a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 307976dda93cSYan, Zheng err = -EINVAL; 308076dda93cSYan, Zheng goto out_dput; 308176dda93cSYan, Zheng } 308276dda93cSYan, Zheng 308364708539SJosef Bacik btrfs_inode_lock(inode, 0); 3084f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 308564708539SJosef Bacik btrfs_inode_unlock(inode, 0); 308646008d9dSAmir Goldstein if (!err) { 308746008d9dSAmir Goldstein fsnotify_rmdir(dir, dentry); 308876dda93cSYan, Zheng d_delete(dentry); 308946008d9dSAmir Goldstein } 3090fa6ac876SLiu Bo 309176dda93cSYan, Zheng out_dput: 309276dda93cSYan, Zheng dput(dentry); 309376dda93cSYan, Zheng out_unlock_dir: 309464708539SJosef Bacik btrfs_inode_unlock(dir, 0); 3095949964c9SMarcos Paulo de Souza free_subvol_name: 3096949964c9SMarcos Paulo de Souza kfree(subvol_name_ptr); 3097949964c9SMarcos Paulo de Souza free_parent: 3098949964c9SMarcos Paulo de Souza if (destroy_parent) 3099949964c9SMarcos Paulo de Souza dput(parent); 310000235411SAl Viro out_drop_write: 31012a79f17eSAl Viro mnt_drop_write_file(file); 310276dda93cSYan, Zheng out: 3103949964c9SMarcos Paulo de Souza kfree(vol_args2); 310476dda93cSYan, Zheng kfree(vol_args); 310576dda93cSYan, Zheng return err; 310676dda93cSYan, Zheng } 310776dda93cSYan, Zheng 31081e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 3109f46b5a66SChristoph Hellwig { 3110496ad9aaSAl Viro struct inode *inode = file_inode(file); 3111f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3112c853a578SGoldwyn Rodrigues struct btrfs_ioctl_defrag_range_args range = {0}; 3113c146afadSYan Zheng int ret; 3114c146afadSYan Zheng 311525122d15SIlya Dryomov ret = mnt_want_write_file(file); 311625122d15SIlya Dryomov if (ret) 311725122d15SIlya Dryomov return ret; 3118b83cc969SLi Zefan 311925122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 312025122d15SIlya Dryomov ret = -EROFS; 312125122d15SIlya Dryomov goto out; 31225ac00addSStefan Behrens } 3123f46b5a66SChristoph Hellwig 3124f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 3125f46b5a66SChristoph Hellwig case S_IFDIR: 3126e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 3127e441d54dSChris Mason ret = -EPERM; 3128e441d54dSChris Mason goto out; 3129e441d54dSChris Mason } 3130de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 3131f46b5a66SChristoph Hellwig break; 3132f46b5a66SChristoph Hellwig case S_IFREG: 3133616d374eSAdam Borowski /* 3134616d374eSAdam Borowski * Note that this does not check the file descriptor for write 3135616d374eSAdam Borowski * access. This prevents defragmenting executables that are 3136616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 3137616d374eSAdam Borowski */ 3138616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 313947291baaSChristian Brauner inode_permission(&init_user_ns, inode, MAY_WRITE)) { 3140616d374eSAdam Borowski ret = -EPERM; 3141e441d54dSChris Mason goto out; 3142e441d54dSChris Mason } 31431e701a32SChris Mason 31441e701a32SChris Mason if (argp) { 3145c853a578SGoldwyn Rodrigues if (copy_from_user(&range, argp, sizeof(range))) { 31461e701a32SChris Mason ret = -EFAULT; 3147683be16eSDan Carpenter goto out; 31481e701a32SChris Mason } 31491e701a32SChris Mason /* compression requires us to start the IO */ 3150c853a578SGoldwyn Rodrigues if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 3151c853a578SGoldwyn Rodrigues range.flags |= BTRFS_DEFRAG_RANGE_START_IO; 3152c853a578SGoldwyn Rodrigues range.extent_thresh = (u32)-1; 31531e701a32SChris Mason } 31541e701a32SChris Mason } else { 31551e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 3156c853a578SGoldwyn Rodrigues range.len = (u64)-1; 31571e701a32SChris Mason } 31581ccc2e8aSQu Wenruo ret = btrfs_defrag_file(file_inode(file), &file->f_ra, 3159c853a578SGoldwyn Rodrigues &range, BTRFS_OLDEST_GENERATION, 0); 31604cb5300bSChris Mason if (ret > 0) 31614cb5300bSChris Mason ret = 0; 3162f46b5a66SChristoph Hellwig break; 31638929ecfaSYan, Zheng default: 31648929ecfaSYan, Zheng ret = -EINVAL; 3165f46b5a66SChristoph Hellwig } 3166e441d54dSChris Mason out: 316725122d15SIlya Dryomov mnt_drop_write_file(file); 3168e441d54dSChris Mason return ret; 3169f46b5a66SChristoph Hellwig } 3170f46b5a66SChristoph Hellwig 31712ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3172f46b5a66SChristoph Hellwig { 3173f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3174f46b5a66SChristoph Hellwig int ret; 3175f46b5a66SChristoph Hellwig 3176e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3177e441d54dSChris Mason return -EPERM; 3178e441d54dSChris Mason 3179c3e1f96cSGoldwyn Rodrigues if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) 3180e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3181c9e9f97bSIlya Dryomov 3182dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3183c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3184c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3185c9e9f97bSIlya Dryomov goto out; 3186c9e9f97bSIlya Dryomov } 3187f46b5a66SChristoph Hellwig 31885516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31892ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3190f46b5a66SChristoph Hellwig 319143d20761SAnand Jain if (!ret) 31920b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 319343d20761SAnand Jain 3194f46b5a66SChristoph Hellwig kfree(vol_args); 3195c9e9f97bSIlya Dryomov out: 3196c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 3197f46b5a66SChristoph Hellwig return ret; 3198f46b5a66SChristoph Hellwig } 3199f46b5a66SChristoph Hellwig 32006b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3201f46b5a66SChristoph Hellwig { 32021a15eb72SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 32030b246afaSJeff Mahoney struct inode *inode = file_inode(file); 32040b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 32056b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 32063fa421deSJosef Bacik struct block_device *bdev = NULL; 32073fa421deSJosef Bacik fmode_t mode; 3208f46b5a66SChristoph Hellwig int ret; 320967ae34b6SDavid Sterba bool cancel = false; 3210f46b5a66SChristoph Hellwig 3211e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3212e441d54dSChris Mason return -EPERM; 3213e441d54dSChris Mason 3214dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3215d815b3f2SDan Carpenter if (IS_ERR(vol_args)) 3216d815b3f2SDan Carpenter return PTR_ERR(vol_args); 3217f46b5a66SChristoph Hellwig 3218748449cdSDavid Sterba if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { 3219fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3220fd4e994bSOmar Sandoval goto out; 3221fd4e994bSOmar Sandoval } 32221a15eb72SJosef Bacik 32236b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 32241a15eb72SJosef Bacik if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 32251a15eb72SJosef Bacik args.devid = vol_args->devid; 32261a15eb72SJosef Bacik } else if (!strcmp("cancel", vol_args->name)) { 322767ae34b6SDavid Sterba cancel = true; 32281a15eb72SJosef Bacik } else { 32291a15eb72SJosef Bacik ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 32301a15eb72SJosef Bacik if (ret) 32311a15eb72SJosef Bacik goto out; 32321a15eb72SJosef Bacik } 32331a15eb72SJosef Bacik 32341a15eb72SJosef Bacik ret = mnt_want_write_file(file); 32351a15eb72SJosef Bacik if (ret) 32361a15eb72SJosef Bacik goto out; 323767ae34b6SDavid Sterba 323867ae34b6SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 323967ae34b6SDavid Sterba cancel); 324067ae34b6SDavid Sterba if (ret) 32411a15eb72SJosef Bacik goto err_drop; 324267ae34b6SDavid Sterba 32431a15eb72SJosef Bacik /* Exclusive operation is now claimed */ 32441a15eb72SJosef Bacik ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); 324567ae34b6SDavid Sterba 3246c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 3247183860f6SAnand Jain 32486b526ed7SAnand Jain if (!ret) { 3249735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 32500b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 32516b526ed7SAnand Jain vol_args->devid); 32526b526ed7SAnand Jain else 32530b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 32546b526ed7SAnand Jain vol_args->name); 32556b526ed7SAnand Jain } 3256c47ca32dSDan Carpenter err_drop: 32574ac20c70SIlya Dryomov mnt_drop_write_file(file); 32583fa421deSJosef Bacik if (bdev) 32593fa421deSJosef Bacik blkdev_put(bdev, mode); 32601a15eb72SJosef Bacik out: 32611a15eb72SJosef Bacik btrfs_put_dev_args_from_path(&args); 32621a15eb72SJosef Bacik kfree(vol_args); 3263f46b5a66SChristoph Hellwig return ret; 3264f46b5a66SChristoph Hellwig } 3265f46b5a66SChristoph Hellwig 3266f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3267f46b5a66SChristoph Hellwig { 32681a15eb72SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 32690b246afaSJeff Mahoney struct inode *inode = file_inode(file); 32700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3271f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 32723fa421deSJosef Bacik struct block_device *bdev = NULL; 32733fa421deSJosef Bacik fmode_t mode; 3274f46b5a66SChristoph Hellwig int ret; 327567ae34b6SDavid Sterba bool cancel; 3276f46b5a66SChristoph Hellwig 3277f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3278f46b5a66SChristoph Hellwig return -EPERM; 3279f46b5a66SChristoph Hellwig 32801a15eb72SJosef Bacik vol_args = memdup_user(arg, sizeof(*vol_args)); 32811a15eb72SJosef Bacik if (IS_ERR(vol_args)) 32821a15eb72SJosef Bacik return PTR_ERR(vol_args); 32831a15eb72SJosef Bacik 32841a15eb72SJosef Bacik vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 32851a15eb72SJosef Bacik if (!strcmp("cancel", vol_args->name)) { 32861a15eb72SJosef Bacik cancel = true; 32871a15eb72SJosef Bacik } else { 32881a15eb72SJosef Bacik ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 32891a15eb72SJosef Bacik if (ret) 32901a15eb72SJosef Bacik goto out; 32911a15eb72SJosef Bacik } 32921a15eb72SJosef Bacik 3293f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3294f46b5a66SChristoph Hellwig if (ret) 32951a15eb72SJosef Bacik goto out; 3296f46b5a66SChristoph Hellwig 329767ae34b6SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 329867ae34b6SDavid Sterba cancel); 329967ae34b6SDavid Sterba if (ret == 0) { 33001a15eb72SJosef Bacik ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); 3301f46b5a66SChristoph Hellwig if (!ret) 33020b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3303c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 330467ae34b6SDavid Sterba } 330567ae34b6SDavid Sterba 3306f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 33073fa421deSJosef Bacik if (bdev) 33083fa421deSJosef Bacik blkdev_put(bdev, mode); 33091a15eb72SJosef Bacik out: 33101a15eb72SJosef Bacik btrfs_put_dev_args_from_path(&args); 33111a15eb72SJosef Bacik kfree(vol_args); 3312f46b5a66SChristoph Hellwig return ret; 3313f46b5a66SChristoph Hellwig } 3314f46b5a66SChristoph Hellwig 33152ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 33162ff7e61eSJeff Mahoney void __user *arg) 3317475f6387SJan Schmidt { 3318027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3319475f6387SJan Schmidt struct btrfs_device *device; 33200b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3321137c5418SJohannes Thumshirn u64 flags_in; 3322027ed2f0SLi Zefan int ret = 0; 3323475f6387SJan Schmidt 3324137c5418SJohannes Thumshirn fi_args = memdup_user(arg, sizeof(*fi_args)); 3325137c5418SJohannes Thumshirn if (IS_ERR(fi_args)) 3326137c5418SJohannes Thumshirn return PTR_ERR(fi_args); 3327137c5418SJohannes Thumshirn 3328137c5418SJohannes Thumshirn flags_in = fi_args->flags; 3329137c5418SJohannes Thumshirn memset(fi_args, 0, sizeof(*fi_args)); 3330027ed2f0SLi Zefan 3331d03262c7SDavid Sterba rcu_read_lock(); 3332027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3333475f6387SJan Schmidt 3334d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3335027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3336027ed2f0SLi Zefan fi_args->max_id = device->devid; 3337475f6387SJan Schmidt } 3338d03262c7SDavid Sterba rcu_read_unlock(); 3339475f6387SJan Schmidt 3340de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3341bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3342bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3343bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 334480a773fbSDavid Sterba 3345137c5418SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) { 3346137c5418SJohannes Thumshirn fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy); 3347137c5418SJohannes Thumshirn fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy); 3348137c5418SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO; 3349137c5418SJohannes Thumshirn } 3350137c5418SJohannes Thumshirn 33510fb408a5SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) { 33520fb408a5SJohannes Thumshirn fi_args->generation = fs_info->generation; 33530fb408a5SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION; 33540fb408a5SJohannes Thumshirn } 33550fb408a5SJohannes Thumshirn 335649bac897SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) { 335749bac897SJohannes Thumshirn memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid, 335849bac897SJohannes Thumshirn sizeof(fi_args->metadata_uuid)); 335949bac897SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID; 336049bac897SJohannes Thumshirn } 336149bac897SJohannes Thumshirn 3362027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3363027ed2f0SLi Zefan ret = -EFAULT; 3364475f6387SJan Schmidt 3365027ed2f0SLi Zefan kfree(fi_args); 3366027ed2f0SLi Zefan return ret; 3367475f6387SJan Schmidt } 3368475f6387SJan Schmidt 33692ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 33702ff7e61eSJeff Mahoney void __user *arg) 3371475f6387SJan Schmidt { 3372562d7b15SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 3373475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3374475f6387SJan Schmidt struct btrfs_device *dev; 3375475f6387SJan Schmidt int ret = 0; 3376475f6387SJan Schmidt 3377475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3378475f6387SJan Schmidt if (IS_ERR(di_args)) 3379475f6387SJan Schmidt return PTR_ERR(di_args); 3380475f6387SJan Schmidt 3381562d7b15SJosef Bacik args.devid = di_args->devid; 3382dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3383562d7b15SJosef Bacik args.uuid = di_args->uuid; 3384475f6387SJan Schmidt 3385c5593ca3SDavid Sterba rcu_read_lock(); 3386562d7b15SJosef Bacik dev = btrfs_find_device(fs_info->fs_devices, &args); 3387475f6387SJan Schmidt if (!dev) { 3388475f6387SJan Schmidt ret = -ENODEV; 3389475f6387SJan Schmidt goto out; 3390475f6387SJan Schmidt } 3391475f6387SJan Schmidt 3392475f6387SJan Schmidt di_args->devid = dev->devid; 33937cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 33947cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3395475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3396a27202fbSJim Meyering if (dev->name) { 3397672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3398672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3399a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3400a27202fbSJim Meyering } else { 340199ba55adSStefan Behrens di_args->path[0] = '\0'; 3402a27202fbSJim Meyering } 3403475f6387SJan Schmidt 3404475f6387SJan Schmidt out: 3405c5593ca3SDavid Sterba rcu_read_unlock(); 3406475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3407475f6387SJan Schmidt ret = -EFAULT; 3408475f6387SJan Schmidt 3409475f6387SJan Schmidt kfree(di_args); 3410475f6387SJan Schmidt return ret; 3411475f6387SJan Schmidt } 3412475f6387SJan Schmidt 34136ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 34146ef5ed0dSJosef Bacik { 3415496ad9aaSAl Viro struct inode *inode = file_inode(file); 34160b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 34176ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 34186ef5ed0dSJosef Bacik struct btrfs_root *new_root; 34196ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 34206ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 34212a2b5d62SJosef Bacik struct btrfs_path *path = NULL; 34226ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 34236ef5ed0dSJosef Bacik u64 objectid = 0; 34246ef5ed0dSJosef Bacik u64 dir_id; 34253c04ce01SMiao Xie int ret; 34266ef5ed0dSJosef Bacik 34276ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 34286ef5ed0dSJosef Bacik return -EPERM; 34296ef5ed0dSJosef Bacik 34303c04ce01SMiao Xie ret = mnt_want_write_file(file); 34313c04ce01SMiao Xie if (ret) 34323c04ce01SMiao Xie return ret; 34333c04ce01SMiao Xie 34343c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 34353c04ce01SMiao Xie ret = -EFAULT; 34363c04ce01SMiao Xie goto out; 34373c04ce01SMiao Xie } 34386ef5ed0dSJosef Bacik 34396ef5ed0dSJosef Bacik if (!objectid) 34401cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 34416ef5ed0dSJosef Bacik 344256e9357aSDavid Sterba new_root = btrfs_get_fs_root(fs_info, objectid, true); 34433c04ce01SMiao Xie if (IS_ERR(new_root)) { 34443c04ce01SMiao Xie ret = PTR_ERR(new_root); 34453c04ce01SMiao Xie goto out; 34463c04ce01SMiao Xie } 34472a2b5d62SJosef Bacik if (!is_fstree(new_root->root_key.objectid)) { 34482a2b5d62SJosef Bacik ret = -ENOENT; 34492a2b5d62SJosef Bacik goto out_free; 34502a2b5d62SJosef Bacik } 34516ef5ed0dSJosef Bacik 34526ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 34533c04ce01SMiao Xie if (!path) { 34543c04ce01SMiao Xie ret = -ENOMEM; 34552a2b5d62SJosef Bacik goto out_free; 34563c04ce01SMiao Xie } 34576ef5ed0dSJosef Bacik 34586ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 345998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 34603c04ce01SMiao Xie ret = PTR_ERR(trans); 34612a2b5d62SJosef Bacik goto out_free; 34626ef5ed0dSJosef Bacik } 34636ef5ed0dSJosef Bacik 34640b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 34650b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 34666ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 3467cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 34682a2b5d62SJosef Bacik btrfs_release_path(path); 34693a45bb20SJeff Mahoney btrfs_end_transaction(trans); 34700b246afaSJeff Mahoney btrfs_err(fs_info, 34715d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 34723c04ce01SMiao Xie ret = -ENOENT; 34732a2b5d62SJosef Bacik goto out_free; 34746ef5ed0dSJosef Bacik } 34756ef5ed0dSJosef Bacik 34766ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 34776ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 34786ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 34792a2b5d62SJosef Bacik btrfs_release_path(path); 34806ef5ed0dSJosef Bacik 34810b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 34823a45bb20SJeff Mahoney btrfs_end_transaction(trans); 34832a2b5d62SJosef Bacik out_free: 348400246528SJosef Bacik btrfs_put_root(new_root); 34852a2b5d62SJosef Bacik btrfs_free_path(path); 34863c04ce01SMiao Xie out: 34873c04ce01SMiao Xie mnt_drop_write_file(file); 34883c04ce01SMiao Xie return ret; 34896ef5ed0dSJosef Bacik } 34906ef5ed0dSJosef Bacik 3491c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 3492bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 3493bf5fc093SJosef Bacik { 349432da5386SDavid Sterba struct btrfs_block_group *block_group; 3495bf5fc093SJosef Bacik 3496bf5fc093SJosef Bacik space->total_bytes = 0; 3497bf5fc093SJosef Bacik space->used_bytes = 0; 3498bf5fc093SJosef Bacik space->flags = 0; 3499bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 3500bf5fc093SJosef Bacik space->flags = block_group->flags; 3501b3470b5dSDavid Sterba space->total_bytes += block_group->length; 3502bf38be65SDavid Sterba space->used_bytes += block_group->used; 3503bf5fc093SJosef Bacik } 3504bf5fc093SJosef Bacik } 3505bf5fc093SJosef Bacik 35062ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 35072ff7e61eSJeff Mahoney void __user *arg) 35081406e432SJosef Bacik { 35091406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 35101406e432SJosef Bacik struct btrfs_ioctl_space_info space; 35111406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 35127fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 351313f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 35141406e432SJosef Bacik struct btrfs_space_info *info; 3515315d8e98SColin Ian King static const u64 types[] = { 3516315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 3517bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 3518bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 3519315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 3520315d8e98SColin Ian King }; 3521bf5fc093SJosef Bacik int num_types = 4; 35227fde62bfSChris Mason int alloc_size; 35231406e432SJosef Bacik int ret = 0; 352451788b1bSDan Rosenberg u64 slot_count = 0; 3525bf5fc093SJosef Bacik int i, c; 35261406e432SJosef Bacik 35271406e432SJosef Bacik if (copy_from_user(&space_args, 35281406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 35291406e432SJosef Bacik sizeof(space_args))) 35301406e432SJosef Bacik return -EFAULT; 35311406e432SJosef Bacik 3532bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3533bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3534bf5fc093SJosef Bacik 3535bf5fc093SJosef Bacik info = NULL; 353672804905SJosef Bacik list_for_each_entry(tmp, &fs_info->space_info, list) { 3537bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3538bf5fc093SJosef Bacik info = tmp; 3539bf5fc093SJosef Bacik break; 3540bf5fc093SJosef Bacik } 3541bf5fc093SJosef Bacik } 35421406e432SJosef Bacik 3543bf5fc093SJosef Bacik if (!info) 3544bf5fc093SJosef Bacik continue; 3545bf5fc093SJosef Bacik 3546bf5fc093SJosef Bacik down_read(&info->groups_sem); 3547bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3548bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 3549bf5fc093SJosef Bacik slot_count++; 3550bf5fc093SJosef Bacik } 3551bf5fc093SJosef Bacik up_read(&info->groups_sem); 3552bf5fc093SJosef Bacik } 3553bf5fc093SJosef Bacik 355436523e95SDavid Sterba /* 355536523e95SDavid Sterba * Global block reserve, exported as a space_info 355636523e95SDavid Sterba */ 355736523e95SDavid Sterba slot_count++; 355836523e95SDavid Sterba 35597fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 35607fde62bfSChris Mason if (space_args.space_slots == 0) { 35617fde62bfSChris Mason space_args.total_spaces = slot_count; 35627fde62bfSChris Mason goto out; 35637fde62bfSChris Mason } 3564bf5fc093SJosef Bacik 356551788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 3566bf5fc093SJosef Bacik 35677fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 3568bf5fc093SJosef Bacik 35697fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 35707fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 35717fde62bfSChris Mason */ 357209cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 35737fde62bfSChris Mason return -ENOMEM; 35747fde62bfSChris Mason 35757fde62bfSChris Mason space_args.total_spaces = 0; 35768d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 35777fde62bfSChris Mason if (!dest) 35787fde62bfSChris Mason return -ENOMEM; 35797fde62bfSChris Mason dest_orig = dest; 35807fde62bfSChris Mason 35817fde62bfSChris Mason /* now we have a buffer to copy into */ 3582bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3583bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3584bf5fc093SJosef Bacik 358551788b1bSDan Rosenberg if (!slot_count) 358651788b1bSDan Rosenberg break; 358751788b1bSDan Rosenberg 3588bf5fc093SJosef Bacik info = NULL; 358972804905SJosef Bacik list_for_each_entry(tmp, &fs_info->space_info, list) { 3590bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3591bf5fc093SJosef Bacik info = tmp; 35927fde62bfSChris Mason break; 3593bf5fc093SJosef Bacik } 3594bf5fc093SJosef Bacik } 35957fde62bfSChris Mason 3596bf5fc093SJosef Bacik if (!info) 3597bf5fc093SJosef Bacik continue; 3598bf5fc093SJosef Bacik down_read(&info->groups_sem); 3599bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3600bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 3601c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 3602c065f5b1SSu Yue &space); 36037fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 36041406e432SJosef Bacik dest++; 36051406e432SJosef Bacik space_args.total_spaces++; 360651788b1bSDan Rosenberg slot_count--; 36071406e432SJosef Bacik } 360851788b1bSDan Rosenberg if (!slot_count) 360951788b1bSDan Rosenberg break; 3610bf5fc093SJosef Bacik } 3611bf5fc093SJosef Bacik up_read(&info->groups_sem); 3612bf5fc093SJosef Bacik } 36131406e432SJosef Bacik 361436523e95SDavid Sterba /* 361536523e95SDavid Sterba * Add global block reserve 361636523e95SDavid Sterba */ 361736523e95SDavid Sterba if (slot_count) { 36180b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 361936523e95SDavid Sterba 362036523e95SDavid Sterba spin_lock(&block_rsv->lock); 362136523e95SDavid Sterba space.total_bytes = block_rsv->size; 362236523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 362336523e95SDavid Sterba spin_unlock(&block_rsv->lock); 362436523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 362536523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 362636523e95SDavid Sterba space_args.total_spaces++; 362736523e95SDavid Sterba } 362836523e95SDavid Sterba 36292eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 36307fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 36317fde62bfSChris Mason 36327fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 36337fde62bfSChris Mason ret = -EFAULT; 36347fde62bfSChris Mason 36357fde62bfSChris Mason kfree(dest_orig); 36367fde62bfSChris Mason out: 36377fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 36381406e432SJosef Bacik ret = -EFAULT; 36391406e432SJosef Bacik 36401406e432SJosef Bacik return ret; 36411406e432SJosef Bacik } 36421406e432SJosef Bacik 36439a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 36449a8c28beSMiao Xie void __user *argp) 364546204592SSage Weil { 364646204592SSage Weil struct btrfs_trans_handle *trans; 364746204592SSage Weil u64 transid; 364846204592SSage Weil 3649d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 3650ff7c1d33SMiao Xie if (IS_ERR(trans)) { 3651ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 365298d5dc13STsutomu Itoh return PTR_ERR(trans); 3653ff7c1d33SMiao Xie 3654ff7c1d33SMiao Xie /* No running transaction, don't bother */ 3655ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 3656ff7c1d33SMiao Xie goto out; 3657ff7c1d33SMiao Xie } 365846204592SSage Weil transid = trans->transid; 3659fdfbf020SJosef Bacik btrfs_commit_transaction_async(trans); 3660ff7c1d33SMiao Xie out: 366146204592SSage Weil if (argp) 366246204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 366346204592SSage Weil return -EFAULT; 366446204592SSage Weil return 0; 366546204592SSage Weil } 366646204592SSage Weil 36672ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 36689a8c28beSMiao Xie void __user *argp) 366946204592SSage Weil { 367046204592SSage Weil u64 transid; 367146204592SSage Weil 367246204592SSage Weil if (argp) { 367346204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 367446204592SSage Weil return -EFAULT; 367546204592SSage Weil } else { 367646204592SSage Weil transid = 0; /* current trans */ 367746204592SSage Weil } 36782ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 367946204592SSage Weil } 368046204592SSage Weil 3681b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 3682475f6387SJan Schmidt { 36830b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 3684475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 3685b8e95489SMiao Xie int ret; 3686475f6387SJan Schmidt 3687475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3688475f6387SJan Schmidt return -EPERM; 3689475f6387SJan Schmidt 3690475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 3691475f6387SJan Schmidt if (IS_ERR(sa)) 3692475f6387SJan Schmidt return PTR_ERR(sa); 3693475f6387SJan Schmidt 3694b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 3695b8e95489SMiao Xie ret = mnt_want_write_file(file); 3696b8e95489SMiao Xie if (ret) 3697b8e95489SMiao Xie goto out; 3698b8e95489SMiao Xie } 3699b8e95489SMiao Xie 37000b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 370163a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 370263a212abSStefan Behrens 0); 3703475f6387SJan Schmidt 37045afe6ce7SFilipe Manana /* 37055afe6ce7SFilipe Manana * Copy scrub args to user space even if btrfs_scrub_dev() returned an 37065afe6ce7SFilipe Manana * error. This is important as it allows user space to know how much 37075afe6ce7SFilipe Manana * progress scrub has done. For example, if scrub is canceled we get 37085afe6ce7SFilipe Manana * -ECANCELED from btrfs_scrub_dev() and return that error back to user 37095afe6ce7SFilipe Manana * space. Later user space can inspect the progress from the structure 37105afe6ce7SFilipe Manana * btrfs_ioctl_scrub_args and resume scrub from where it left off 37115afe6ce7SFilipe Manana * previously (btrfs-progs does this). 37125afe6ce7SFilipe Manana * If we fail to copy the btrfs_ioctl_scrub_args structure to user space 37135afe6ce7SFilipe Manana * then return -EFAULT to signal the structure was not copied or it may 37145afe6ce7SFilipe Manana * be corrupt and unreliable due to a partial copy. 37155afe6ce7SFilipe Manana */ 37165afe6ce7SFilipe Manana if (copy_to_user(arg, sa, sizeof(*sa))) 3717475f6387SJan Schmidt ret = -EFAULT; 3718475f6387SJan Schmidt 3719b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 3720b8e95489SMiao Xie mnt_drop_write_file(file); 3721b8e95489SMiao Xie out: 3722475f6387SJan Schmidt kfree(sa); 3723475f6387SJan Schmidt return ret; 3724475f6387SJan Schmidt } 3725475f6387SJan Schmidt 37262ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 3727475f6387SJan Schmidt { 3728475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3729475f6387SJan Schmidt return -EPERM; 3730475f6387SJan Schmidt 37312ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 3732475f6387SJan Schmidt } 3733475f6387SJan Schmidt 37342ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 3735475f6387SJan Schmidt void __user *arg) 3736475f6387SJan Schmidt { 3737475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 3738475f6387SJan Schmidt int ret; 3739475f6387SJan Schmidt 3740475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3741475f6387SJan Schmidt return -EPERM; 3742475f6387SJan Schmidt 3743475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 3744475f6387SJan Schmidt if (IS_ERR(sa)) 3745475f6387SJan Schmidt return PTR_ERR(sa); 3746475f6387SJan Schmidt 37472ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 3748475f6387SJan Schmidt 37494fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 3750475f6387SJan Schmidt ret = -EFAULT; 3751475f6387SJan Schmidt 3752475f6387SJan Schmidt kfree(sa); 3753475f6387SJan Schmidt return ret; 3754475f6387SJan Schmidt } 3755475f6387SJan Schmidt 37562ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 3757b27f7c0cSDavid Sterba void __user *arg) 3758c11d2c23SStefan Behrens { 3759c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 3760c11d2c23SStefan Behrens int ret; 3761c11d2c23SStefan Behrens 3762c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 3763c11d2c23SStefan Behrens if (IS_ERR(sa)) 3764c11d2c23SStefan Behrens return PTR_ERR(sa); 3765c11d2c23SStefan Behrens 3766b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 3767b27f7c0cSDavid Sterba kfree(sa); 3768b27f7c0cSDavid Sterba return -EPERM; 3769b27f7c0cSDavid Sterba } 3770b27f7c0cSDavid Sterba 37712ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 3772c11d2c23SStefan Behrens 3773eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 3774c11d2c23SStefan Behrens ret = -EFAULT; 3775c11d2c23SStefan Behrens 3776c11d2c23SStefan Behrens kfree(sa); 3777c11d2c23SStefan Behrens return ret; 3778c11d2c23SStefan Behrens } 3779c11d2c23SStefan Behrens 37802ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 37812ff7e61eSJeff Mahoney void __user *arg) 37823f6bcfbdSStefan Behrens { 37833f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 37843f6bcfbdSStefan Behrens int ret; 37853f6bcfbdSStefan Behrens 37863f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 37873f6bcfbdSStefan Behrens return -EPERM; 37883f6bcfbdSStefan Behrens 37893f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 37903f6bcfbdSStefan Behrens if (IS_ERR(p)) 37913f6bcfbdSStefan Behrens return PTR_ERR(p); 37923f6bcfbdSStefan Behrens 37933f6bcfbdSStefan Behrens switch (p->cmd) { 37943f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 3795bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 3796adfa97cbSIlya Dryomov ret = -EROFS; 3797adfa97cbSIlya Dryomov goto out; 3798adfa97cbSIlya Dryomov } 3799c3e1f96cSGoldwyn Rodrigues if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) { 3800e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 38013f6bcfbdSStefan Behrens } else { 38022ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 3803c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 38043f6bcfbdSStefan Behrens } 38053f6bcfbdSStefan Behrens break; 38063f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 38070b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 38083f6bcfbdSStefan Behrens ret = 0; 38093f6bcfbdSStefan Behrens break; 38103f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 381117d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 381297282031SAnand Jain ret = 0; 38133f6bcfbdSStefan Behrens break; 38143f6bcfbdSStefan Behrens default: 38153f6bcfbdSStefan Behrens ret = -EINVAL; 38163f6bcfbdSStefan Behrens break; 38173f6bcfbdSStefan Behrens } 38183f6bcfbdSStefan Behrens 3819d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 38203f6bcfbdSStefan Behrens ret = -EFAULT; 3821adfa97cbSIlya Dryomov out: 38223f6bcfbdSStefan Behrens kfree(p); 38233f6bcfbdSStefan Behrens return ret; 38243f6bcfbdSStefan Behrens } 38253f6bcfbdSStefan Behrens 3826d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 3827d7728c96SJan Schmidt { 3828d7728c96SJan Schmidt int ret = 0; 3829d7728c96SJan Schmidt int i; 3830740c3d22SChris Mason u64 rel_ptr; 3831d7728c96SJan Schmidt int size; 3832806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 3833d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 3834d7728c96SJan Schmidt struct btrfs_path *path; 3835d7728c96SJan Schmidt 383682b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 3837d7728c96SJan Schmidt return -EPERM; 3838d7728c96SJan Schmidt 3839d7728c96SJan Schmidt path = btrfs_alloc_path(); 3840d7728c96SJan Schmidt if (!path) { 3841d7728c96SJan Schmidt ret = -ENOMEM; 3842d7728c96SJan Schmidt goto out; 3843d7728c96SJan Schmidt } 3844d7728c96SJan Schmidt 3845d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 3846d7728c96SJan Schmidt if (IS_ERR(ipa)) { 3847d7728c96SJan Schmidt ret = PTR_ERR(ipa); 3848d7728c96SJan Schmidt ipa = NULL; 3849d7728c96SJan Schmidt goto out; 3850d7728c96SJan Schmidt } 3851d7728c96SJan Schmidt 3852d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 3853d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 3854d7728c96SJan Schmidt if (IS_ERR(ipath)) { 3855d7728c96SJan Schmidt ret = PTR_ERR(ipath); 3856d7728c96SJan Schmidt ipath = NULL; 3857d7728c96SJan Schmidt goto out; 3858d7728c96SJan Schmidt } 3859d7728c96SJan Schmidt 3860d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 3861d7728c96SJan Schmidt if (ret < 0) 3862d7728c96SJan Schmidt goto out; 3863d7728c96SJan Schmidt 3864d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 3865745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 3866745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 3867740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 3868d7728c96SJan Schmidt } 3869d7728c96SJan Schmidt 3870718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 3871718dc5faSOmar Sandoval ipath->fspath, size); 3872d7728c96SJan Schmidt if (ret) { 3873d7728c96SJan Schmidt ret = -EFAULT; 3874d7728c96SJan Schmidt goto out; 3875d7728c96SJan Schmidt } 3876d7728c96SJan Schmidt 3877d7728c96SJan Schmidt out: 3878d7728c96SJan Schmidt btrfs_free_path(path); 3879d7728c96SJan Schmidt free_ipath(ipath); 3880d7728c96SJan Schmidt kfree(ipa); 3881d7728c96SJan Schmidt 3882d7728c96SJan Schmidt return ret; 3883d7728c96SJan Schmidt } 3884d7728c96SJan Schmidt 3885d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 3886d7728c96SJan Schmidt { 3887d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 3888d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 3889d7728c96SJan Schmidt 3890d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 3891d7728c96SJan Schmidt inodes->bytes_left -= c; 3892d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 3893d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 3894d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 3895d7728c96SJan Schmidt inodes->elem_cnt += 3; 3896d7728c96SJan Schmidt } else { 3897d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 3898d7728c96SJan Schmidt inodes->bytes_left = 0; 3899d7728c96SJan Schmidt inodes->elem_missed += 3; 3900d7728c96SJan Schmidt } 3901d7728c96SJan Schmidt 3902d7728c96SJan Schmidt return 0; 3903d7728c96SJan Schmidt } 3904d7728c96SJan Schmidt 39052ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 3906d24a67b2SZygo Blaxell void __user *arg, int version) 3907d7728c96SJan Schmidt { 3908d7728c96SJan Schmidt int ret = 0; 3909d7728c96SJan Schmidt int size; 3910d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 3911d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 3912d7728c96SJan Schmidt struct btrfs_path *path = NULL; 3913d24a67b2SZygo Blaxell bool ignore_offset; 3914d7728c96SJan Schmidt 3915d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3916d7728c96SJan Schmidt return -EPERM; 3917d7728c96SJan Schmidt 3918d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 39197b9ea627SShailendra Verma if (IS_ERR(loi)) 39207b9ea627SShailendra Verma return PTR_ERR(loi); 3921d7728c96SJan Schmidt 3922d24a67b2SZygo Blaxell if (version == 1) { 3923d24a67b2SZygo Blaxell ignore_offset = false; 3924b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 3925d24a67b2SZygo Blaxell } else { 3926d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 3927d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 3928d24a67b2SZygo Blaxell ret = -EINVAL; 3929d24a67b2SZygo Blaxell goto out_loi; 3930d24a67b2SZygo Blaxell } 3931d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 3932d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 3933d24a67b2SZygo Blaxell ret = -EINVAL; 3934d24a67b2SZygo Blaxell goto out_loi; 3935d24a67b2SZygo Blaxell } 3936d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 3937b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 3938d24a67b2SZygo Blaxell } 3939d24a67b2SZygo Blaxell 3940d7728c96SJan Schmidt path = btrfs_alloc_path(); 3941d7728c96SJan Schmidt if (!path) { 3942d7728c96SJan Schmidt ret = -ENOMEM; 3943d7728c96SJan Schmidt goto out; 3944d7728c96SJan Schmidt } 3945d7728c96SJan Schmidt 3946d7728c96SJan Schmidt inodes = init_data_container(size); 3947d7728c96SJan Schmidt if (IS_ERR(inodes)) { 3948d7728c96SJan Schmidt ret = PTR_ERR(inodes); 3949d7728c96SJan Schmidt inodes = NULL; 3950d7728c96SJan Schmidt goto out; 3951d7728c96SJan Schmidt } 3952d7728c96SJan Schmidt 39532ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 3954d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 3955df031f07SLiu Bo if (ret == -EINVAL) 3956d7728c96SJan Schmidt ret = -ENOENT; 3957d7728c96SJan Schmidt if (ret < 0) 3958d7728c96SJan Schmidt goto out; 3959d7728c96SJan Schmidt 3960718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 3961718dc5faSOmar Sandoval size); 3962d7728c96SJan Schmidt if (ret) 3963d7728c96SJan Schmidt ret = -EFAULT; 3964d7728c96SJan Schmidt 3965d7728c96SJan Schmidt out: 3966d7728c96SJan Schmidt btrfs_free_path(path); 3967f54de068SDavid Sterba kvfree(inodes); 3968d24a67b2SZygo Blaxell out_loi: 3969d7728c96SJan Schmidt kfree(loi); 3970d7728c96SJan Schmidt 3971d7728c96SJan Schmidt return ret; 3972d7728c96SJan Schmidt } 3973d7728c96SJan Schmidt 3974008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 3975c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 3976c9e9f97bSIlya Dryomov { 3977c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 3978c9e9f97bSIlya Dryomov 3979c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 3980c9e9f97bSIlya Dryomov 39813009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 3982837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 3983837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 3984837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 3985a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 3986a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 3987837d5b6eSIlya Dryomov 3988c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 3989c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 3990c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 399119a39dceSIlya Dryomov 399219a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 399319a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 399419a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 3995c9e9f97bSIlya Dryomov } 3996c9e9f97bSIlya Dryomov 39979ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 3998c9e9f97bSIlya Dryomov { 3999496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4000c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4001c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4002c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4003ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4004c9e9f97bSIlya Dryomov int ret; 4005c9e9f97bSIlya Dryomov 40066c405b24SNikolay Borisov if (!arg) 40076c405b24SNikolay Borisov btrfs_warn(fs_info, 40086c405b24SNikolay Borisov "IOC_BALANCE ioctl (v1) is deprecated and will be removed in kernel 5.18"); 40096c405b24SNikolay Borisov 4010c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4011c9e9f97bSIlya Dryomov return -EPERM; 4012c9e9f97bSIlya Dryomov 4013e54bfa31SLiu Bo ret = mnt_want_write_file(file); 40149ba1f6e4SLiu Bo if (ret) 40159ba1f6e4SLiu Bo return ret; 40169ba1f6e4SLiu Bo 4017ed0fb78fSIlya Dryomov again: 4018c3e1f96cSGoldwyn Rodrigues if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { 4019c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4020ed0fb78fSIlya Dryomov need_unlock = true; 4021ed0fb78fSIlya Dryomov goto locked; 4022ed0fb78fSIlya Dryomov } 4023ed0fb78fSIlya Dryomov 4024ed0fb78fSIlya Dryomov /* 402501327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4026ed0fb78fSIlya Dryomov * (1) some other op is running 4027ed0fb78fSIlya Dryomov * (2) balance is running 4028ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4029ed0fb78fSIlya Dryomov */ 4030ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4031ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4032ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 40333009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4034ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4035dccdb07bSDavid Sterba /* 4036dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4037dccdb07bSDavid Sterba * we'll reexamine the status again. 4038dccdb07bSDavid Sterba */ 4039ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4040ed0fb78fSIlya Dryomov 4041ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 40423009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4043ed0fb78fSIlya Dryomov /* this is (3) */ 4044ed0fb78fSIlya Dryomov need_unlock = false; 4045ed0fb78fSIlya Dryomov goto locked; 4046ed0fb78fSIlya Dryomov } 4047ed0fb78fSIlya Dryomov 4048ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4049ed0fb78fSIlya Dryomov goto again; 4050ed0fb78fSIlya Dryomov } else { 4051ed0fb78fSIlya Dryomov /* this is (2) */ 4052ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4053ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4054ed0fb78fSIlya Dryomov goto out; 4055ed0fb78fSIlya Dryomov } 4056ed0fb78fSIlya Dryomov } else { 4057ed0fb78fSIlya Dryomov /* this is (1) */ 4058ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4059e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4060ed0fb78fSIlya Dryomov goto out; 4061ed0fb78fSIlya Dryomov } 4062ed0fb78fSIlya Dryomov 4063ed0fb78fSIlya Dryomov locked: 4064c9e9f97bSIlya Dryomov 4065c9e9f97bSIlya Dryomov if (arg) { 4066c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4067c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4068c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4069ed0fb78fSIlya Dryomov goto out_unlock; 4070c9e9f97bSIlya Dryomov } 4071de322263SIlya Dryomov 4072de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4073de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4074de322263SIlya Dryomov ret = -ENOTCONN; 4075de322263SIlya Dryomov goto out_bargs; 4076de322263SIlya Dryomov } 4077de322263SIlya Dryomov 4078de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4079de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4080de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4081de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4082*efc0e69cSNikolay Borisov btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE); 4083de322263SIlya Dryomov 4084de322263SIlya Dryomov goto do_balance; 4085de322263SIlya Dryomov } 4086c9e9f97bSIlya Dryomov } else { 4087c9e9f97bSIlya Dryomov bargs = NULL; 4088c9e9f97bSIlya Dryomov } 4089c9e9f97bSIlya Dryomov 4090ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4091837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4092837d5b6eSIlya Dryomov goto out_bargs; 4093837d5b6eSIlya Dryomov } 4094837d5b6eSIlya Dryomov 40958d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4096c9e9f97bSIlya Dryomov if (!bctl) { 4097c9e9f97bSIlya Dryomov ret = -ENOMEM; 4098c9e9f97bSIlya Dryomov goto out_bargs; 4099c9e9f97bSIlya Dryomov } 4100c9e9f97bSIlya Dryomov 4101c9e9f97bSIlya Dryomov if (arg) { 4102c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4103c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4104c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4105c9e9f97bSIlya Dryomov 4106c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4107f43ffb60SIlya Dryomov } else { 4108f43ffb60SIlya Dryomov /* balance everything - no filters */ 4109f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4110c9e9f97bSIlya Dryomov } 4111c9e9f97bSIlya Dryomov 41128eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 41138eb93459SDavid Sterba ret = -EINVAL; 41140f89abf5SChristian Engelmayer goto out_bctl; 41158eb93459SDavid Sterba } 41168eb93459SDavid Sterba 4117de322263SIlya Dryomov do_balance: 4118c9e9f97bSIlya Dryomov /* 4119c3e1f96cSGoldwyn Rodrigues * Ownership of bctl and exclusive operation goes to btrfs_balance. 4120c3e1f96cSGoldwyn Rodrigues * bctl is freed in reset_balance_state, or, if restriper was paused 4121c3e1f96cSGoldwyn Rodrigues * all the way until unmount, in free_fs_info. The flag should be 4122c3e1f96cSGoldwyn Rodrigues * cleared after reset_balance_state. 4123c9e9f97bSIlya Dryomov */ 4124ed0fb78fSIlya Dryomov need_unlock = false; 4125ed0fb78fSIlya Dryomov 41266fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 41270f89abf5SChristian Engelmayer bctl = NULL; 4128ed0fb78fSIlya Dryomov 4129d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4130c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4131c9e9f97bSIlya Dryomov ret = -EFAULT; 4132c9e9f97bSIlya Dryomov } 4133c9e9f97bSIlya Dryomov 41340f89abf5SChristian Engelmayer out_bctl: 41350f89abf5SChristian Engelmayer kfree(bctl); 4136c9e9f97bSIlya Dryomov out_bargs: 4137c9e9f97bSIlya Dryomov kfree(bargs); 4138ed0fb78fSIlya Dryomov out_unlock: 4139c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4140ed0fb78fSIlya Dryomov if (need_unlock) 4141c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 4142ed0fb78fSIlya Dryomov out: 4143e54bfa31SLiu Bo mnt_drop_write_file(file); 4144c9e9f97bSIlya Dryomov return ret; 4145c9e9f97bSIlya Dryomov } 4146c9e9f97bSIlya Dryomov 41472ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4148837d5b6eSIlya Dryomov { 4149837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4150837d5b6eSIlya Dryomov return -EPERM; 4151837d5b6eSIlya Dryomov 4152837d5b6eSIlya Dryomov switch (cmd) { 4153837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 41540b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4155a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 41560b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4157837d5b6eSIlya Dryomov } 4158837d5b6eSIlya Dryomov 4159837d5b6eSIlya Dryomov return -EINVAL; 4160837d5b6eSIlya Dryomov } 4161837d5b6eSIlya Dryomov 41622ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 416319a39dceSIlya Dryomov void __user *arg) 416419a39dceSIlya Dryomov { 416519a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 416619a39dceSIlya Dryomov int ret = 0; 416719a39dceSIlya Dryomov 416819a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 416919a39dceSIlya Dryomov return -EPERM; 417019a39dceSIlya Dryomov 417119a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 417219a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 417319a39dceSIlya Dryomov ret = -ENOTCONN; 417419a39dceSIlya Dryomov goto out; 417519a39dceSIlya Dryomov } 417619a39dceSIlya Dryomov 41778d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 417819a39dceSIlya Dryomov if (!bargs) { 417919a39dceSIlya Dryomov ret = -ENOMEM; 418019a39dceSIlya Dryomov goto out; 418119a39dceSIlya Dryomov } 418219a39dceSIlya Dryomov 4183008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 418419a39dceSIlya Dryomov 418519a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 418619a39dceSIlya Dryomov ret = -EFAULT; 418719a39dceSIlya Dryomov 418819a39dceSIlya Dryomov kfree(bargs); 418919a39dceSIlya Dryomov out: 419019a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 419119a39dceSIlya Dryomov return ret; 419219a39dceSIlya Dryomov } 419319a39dceSIlya Dryomov 4194905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 41955d13a37bSArne Jansen { 41960b246afaSJeff Mahoney struct inode *inode = file_inode(file); 41970b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 41985d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 41995d13a37bSArne Jansen int ret; 42005d13a37bSArne Jansen 42015d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 42025d13a37bSArne Jansen return -EPERM; 42035d13a37bSArne Jansen 4204905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4205905b0ddaSMiao Xie if (ret) 4206905b0ddaSMiao Xie return ret; 42075d13a37bSArne Jansen 42085d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4209905b0ddaSMiao Xie if (IS_ERR(sa)) { 4210905b0ddaSMiao Xie ret = PTR_ERR(sa); 4211905b0ddaSMiao Xie goto drop_write; 4212905b0ddaSMiao Xie } 42135d13a37bSArne Jansen 42140b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 42155d13a37bSArne Jansen 42165d13a37bSArne Jansen switch (sa->cmd) { 42175d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4218340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 42195d13a37bSArne Jansen break; 42205d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4221340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 42225d13a37bSArne Jansen break; 42235d13a37bSArne Jansen default: 42245d13a37bSArne Jansen ret = -EINVAL; 42255d13a37bSArne Jansen break; 42265d13a37bSArne Jansen } 42275d13a37bSArne Jansen 42285d13a37bSArne Jansen kfree(sa); 42290b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4230905b0ddaSMiao Xie drop_write: 4231905b0ddaSMiao Xie mnt_drop_write_file(file); 42325d13a37bSArne Jansen return ret; 42335d13a37bSArne Jansen } 42345d13a37bSArne Jansen 4235905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 42365d13a37bSArne Jansen { 42370b246afaSJeff Mahoney struct inode *inode = file_inode(file); 42380b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 42390b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 42405d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 42415d13a37bSArne Jansen struct btrfs_trans_handle *trans; 42425d13a37bSArne Jansen int ret; 42435d13a37bSArne Jansen int err; 42445d13a37bSArne Jansen 42455d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 42465d13a37bSArne Jansen return -EPERM; 42475d13a37bSArne Jansen 4248905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4249905b0ddaSMiao Xie if (ret) 4250905b0ddaSMiao Xie return ret; 42515d13a37bSArne Jansen 42525d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4253905b0ddaSMiao Xie if (IS_ERR(sa)) { 4254905b0ddaSMiao Xie ret = PTR_ERR(sa); 4255905b0ddaSMiao Xie goto drop_write; 4256905b0ddaSMiao Xie } 42575d13a37bSArne Jansen 42585d13a37bSArne Jansen trans = btrfs_join_transaction(root); 42595d13a37bSArne Jansen if (IS_ERR(trans)) { 42605d13a37bSArne Jansen ret = PTR_ERR(trans); 42615d13a37bSArne Jansen goto out; 42625d13a37bSArne Jansen } 42635d13a37bSArne Jansen 42645d13a37bSArne Jansen if (sa->assign) { 42659f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 42665d13a37bSArne Jansen } else { 426739616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 42685d13a37bSArne Jansen } 42695d13a37bSArne Jansen 4270e082f563SQu Wenruo /* update qgroup status and info */ 4271280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4272e082f563SQu Wenruo if (err < 0) 42730b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4274ad8403dfSAnand Jain "failed to update qgroup status and info"); 42753a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 42765d13a37bSArne Jansen if (err && !ret) 42775d13a37bSArne Jansen ret = err; 42785d13a37bSArne Jansen 42795d13a37bSArne Jansen out: 42805d13a37bSArne Jansen kfree(sa); 4281905b0ddaSMiao Xie drop_write: 4282905b0ddaSMiao Xie mnt_drop_write_file(file); 42835d13a37bSArne Jansen return ret; 42845d13a37bSArne Jansen } 42855d13a37bSArne Jansen 4286905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 42875d13a37bSArne Jansen { 42880b246afaSJeff Mahoney struct inode *inode = file_inode(file); 42890b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 42905d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 42915d13a37bSArne Jansen struct btrfs_trans_handle *trans; 42925d13a37bSArne Jansen int ret; 42935d13a37bSArne Jansen int err; 42945d13a37bSArne Jansen 42955d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 42965d13a37bSArne Jansen return -EPERM; 42975d13a37bSArne Jansen 4298905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4299905b0ddaSMiao Xie if (ret) 4300905b0ddaSMiao Xie return ret; 43015d13a37bSArne Jansen 43025d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4303905b0ddaSMiao Xie if (IS_ERR(sa)) { 4304905b0ddaSMiao Xie ret = PTR_ERR(sa); 4305905b0ddaSMiao Xie goto drop_write; 4306905b0ddaSMiao Xie } 43075d13a37bSArne Jansen 4308d86e56cfSMiao Xie if (!sa->qgroupid) { 4309d86e56cfSMiao Xie ret = -EINVAL; 4310d86e56cfSMiao Xie goto out; 4311d86e56cfSMiao Xie } 4312d86e56cfSMiao Xie 43135d13a37bSArne Jansen trans = btrfs_join_transaction(root); 43145d13a37bSArne Jansen if (IS_ERR(trans)) { 43155d13a37bSArne Jansen ret = PTR_ERR(trans); 43165d13a37bSArne Jansen goto out; 43175d13a37bSArne Jansen } 43185d13a37bSArne Jansen 43195d13a37bSArne Jansen if (sa->create) { 432049a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 43215d13a37bSArne Jansen } else { 43223efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 43235d13a37bSArne Jansen } 43245d13a37bSArne Jansen 43253a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 43265d13a37bSArne Jansen if (err && !ret) 43275d13a37bSArne Jansen ret = err; 43285d13a37bSArne Jansen 43295d13a37bSArne Jansen out: 43305d13a37bSArne Jansen kfree(sa); 4331905b0ddaSMiao Xie drop_write: 4332905b0ddaSMiao Xie mnt_drop_write_file(file); 43335d13a37bSArne Jansen return ret; 43345d13a37bSArne Jansen } 43355d13a37bSArne Jansen 4336905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 43375d13a37bSArne Jansen { 43380b246afaSJeff Mahoney struct inode *inode = file_inode(file); 43390b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 43405d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 43415d13a37bSArne Jansen struct btrfs_trans_handle *trans; 43425d13a37bSArne Jansen int ret; 43435d13a37bSArne Jansen int err; 43445d13a37bSArne Jansen u64 qgroupid; 43455d13a37bSArne Jansen 43465d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 43475d13a37bSArne Jansen return -EPERM; 43485d13a37bSArne Jansen 4349905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4350905b0ddaSMiao Xie if (ret) 4351905b0ddaSMiao Xie return ret; 43525d13a37bSArne Jansen 43535d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4354905b0ddaSMiao Xie if (IS_ERR(sa)) { 4355905b0ddaSMiao Xie ret = PTR_ERR(sa); 4356905b0ddaSMiao Xie goto drop_write; 4357905b0ddaSMiao Xie } 43585d13a37bSArne Jansen 43595d13a37bSArne Jansen trans = btrfs_join_transaction(root); 43605d13a37bSArne Jansen if (IS_ERR(trans)) { 43615d13a37bSArne Jansen ret = PTR_ERR(trans); 43625d13a37bSArne Jansen goto out; 43635d13a37bSArne Jansen } 43645d13a37bSArne Jansen 43655d13a37bSArne Jansen qgroupid = sa->qgroupid; 43665d13a37bSArne Jansen if (!qgroupid) { 43675d13a37bSArne Jansen /* take the current subvol as qgroup */ 43685d13a37bSArne Jansen qgroupid = root->root_key.objectid; 43695d13a37bSArne Jansen } 43705d13a37bSArne Jansen 4371f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 43725d13a37bSArne Jansen 43733a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 43745d13a37bSArne Jansen if (err && !ret) 43755d13a37bSArne Jansen ret = err; 43765d13a37bSArne Jansen 43775d13a37bSArne Jansen out: 43785d13a37bSArne Jansen kfree(sa); 4379905b0ddaSMiao Xie drop_write: 4380905b0ddaSMiao Xie mnt_drop_write_file(file); 43815d13a37bSArne Jansen return ret; 43825d13a37bSArne Jansen } 43835d13a37bSArne Jansen 43842f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 43852f232036SJan Schmidt { 43860b246afaSJeff Mahoney struct inode *inode = file_inode(file); 43870b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 43882f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 43892f232036SJan Schmidt int ret; 43902f232036SJan Schmidt 43912f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 43922f232036SJan Schmidt return -EPERM; 43932f232036SJan Schmidt 43942f232036SJan Schmidt ret = mnt_want_write_file(file); 43952f232036SJan Schmidt if (ret) 43962f232036SJan Schmidt return ret; 43972f232036SJan Schmidt 43982f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 43992f232036SJan Schmidt if (IS_ERR(qsa)) { 44002f232036SJan Schmidt ret = PTR_ERR(qsa); 44012f232036SJan Schmidt goto drop_write; 44022f232036SJan Schmidt } 44032f232036SJan Schmidt 44042f232036SJan Schmidt if (qsa->flags) { 44052f232036SJan Schmidt ret = -EINVAL; 44062f232036SJan Schmidt goto out; 44072f232036SJan Schmidt } 44082f232036SJan Schmidt 44090b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 44102f232036SJan Schmidt 44112f232036SJan Schmidt out: 44122f232036SJan Schmidt kfree(qsa); 44132f232036SJan Schmidt drop_write: 44142f232036SJan Schmidt mnt_drop_write_file(file); 44152f232036SJan Schmidt return ret; 44162f232036SJan Schmidt } 44172f232036SJan Schmidt 4418b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, 4419b929c1d8SMarcos Paulo de Souza void __user *arg) 44202f232036SJan Schmidt { 44210afb603aSGoldwyn Rodrigues struct btrfs_ioctl_quota_rescan_args qsa = {0}; 44222f232036SJan Schmidt 44232f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 44242f232036SJan Schmidt return -EPERM; 44252f232036SJan Schmidt 44260b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 44270afb603aSGoldwyn Rodrigues qsa.flags = 1; 44280afb603aSGoldwyn Rodrigues qsa.progress = fs_info->qgroup_rescan_progress.objectid; 44292f232036SJan Schmidt } 44302f232036SJan Schmidt 44310afb603aSGoldwyn Rodrigues if (copy_to_user(arg, &qsa, sizeof(qsa))) 4432991a3daeSAnand Jain return -EFAULT; 44332f232036SJan Schmidt 4434991a3daeSAnand Jain return 0; 44352f232036SJan Schmidt } 44362f232036SJan Schmidt 4437b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info, 4438b929c1d8SMarcos Paulo de Souza void __user *arg) 443957254b6eSJan Schmidt { 444057254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 444157254b6eSJan Schmidt return -EPERM; 444257254b6eSJan Schmidt 44430b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 444457254b6eSJan Schmidt } 444557254b6eSJan Schmidt 4446abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 4447e4fed17aSChristian Brauner struct user_namespace *mnt_userns, 4448abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 44498ea05e3aSAlexander Block { 4450496ad9aaSAl Viro struct inode *inode = file_inode(file); 44510b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 44528ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 44538ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 44548ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 445595582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 44568ea05e3aSAlexander Block int ret = 0; 4457dd5f9615SStefan Behrens int received_uuid_changed; 44588ea05e3aSAlexander Block 4459e4fed17aSChristian Brauner if (!inode_owner_or_capable(mnt_userns, inode)) 4460bd60ea0fSDavid Sterba return -EPERM; 4461bd60ea0fSDavid Sterba 44628ea05e3aSAlexander Block ret = mnt_want_write_file(file); 44638ea05e3aSAlexander Block if (ret < 0) 44648ea05e3aSAlexander Block return ret; 44658ea05e3aSAlexander Block 44660b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 44678ea05e3aSAlexander Block 44684a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 44698ea05e3aSAlexander Block ret = -EINVAL; 44708ea05e3aSAlexander Block goto out; 44718ea05e3aSAlexander Block } 44728ea05e3aSAlexander Block 44738ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 44748ea05e3aSAlexander Block ret = -EROFS; 44758ea05e3aSAlexander Block goto out; 44768ea05e3aSAlexander Block } 44778ea05e3aSAlexander Block 4478dd5f9615SStefan Behrens /* 4479dd5f9615SStefan Behrens * 1 - root item 4480dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 4481dd5f9615SStefan Behrens */ 4482dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 44838ea05e3aSAlexander Block if (IS_ERR(trans)) { 44848ea05e3aSAlexander Block ret = PTR_ERR(trans); 44858ea05e3aSAlexander Block trans = NULL; 44868ea05e3aSAlexander Block goto out; 44878ea05e3aSAlexander Block } 44888ea05e3aSAlexander Block 44898ea05e3aSAlexander Block sa->rtransid = trans->transid; 44908ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 44918ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 44928ea05e3aSAlexander Block 4493dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 4494dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 4495dd5f9615SStefan Behrens if (received_uuid_changed && 4496d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 4497d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 4498dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4499dd5f9615SStefan Behrens root->root_key.objectid); 4500d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 4501d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 4502d87ff758SNikolay Borisov btrfs_end_transaction(trans); 4503d87ff758SNikolay Borisov goto out; 4504d87ff758SNikolay Borisov } 4505d87ff758SNikolay Borisov } 45068ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 45078ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 45088ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 45093cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 45103cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 45113cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 45123cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 45138ea05e3aSAlexander Block 45140b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 45158ea05e3aSAlexander Block &root->root_key, &root->root_item); 45168ea05e3aSAlexander Block if (ret < 0) { 45173a45bb20SJeff Mahoney btrfs_end_transaction(trans); 45188ea05e3aSAlexander Block goto out; 4519dd5f9615SStefan Behrens } 4520dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 4521cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 4522dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4523dd5f9615SStefan Behrens root->root_key.objectid); 4524dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 452566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 4526efd38150SNikolay Borisov btrfs_end_transaction(trans); 4527dd5f9615SStefan Behrens goto out; 4528dd5f9615SStefan Behrens } 4529dd5f9615SStefan Behrens } 45303a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 4531abccd00fSHugo Mills out: 45320b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4533abccd00fSHugo Mills mnt_drop_write_file(file); 4534abccd00fSHugo Mills return ret; 4535abccd00fSHugo Mills } 4536abccd00fSHugo Mills 4537abccd00fSHugo Mills #ifdef CONFIG_64BIT 4538abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 4539abccd00fSHugo Mills void __user *arg) 4540abccd00fSHugo Mills { 4541abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 4542abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 4543abccd00fSHugo Mills int ret = 0; 4544abccd00fSHugo Mills 4545abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 45467b9ea627SShailendra Verma if (IS_ERR(args32)) 45477b9ea627SShailendra Verma return PTR_ERR(args32); 4548abccd00fSHugo Mills 45498d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 455084dbeb87SDan Carpenter if (!args64) { 455184dbeb87SDan Carpenter ret = -ENOMEM; 4552abccd00fSHugo Mills goto out; 4553abccd00fSHugo Mills } 4554abccd00fSHugo Mills 4555abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 4556abccd00fSHugo Mills args64->stransid = args32->stransid; 4557abccd00fSHugo Mills args64->rtransid = args32->rtransid; 4558abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 4559abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 4560abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 4561abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 4562abccd00fSHugo Mills args64->flags = args32->flags; 4563abccd00fSHugo Mills 4564e4fed17aSChristian Brauner ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64); 4565abccd00fSHugo Mills if (ret) 4566abccd00fSHugo Mills goto out; 4567abccd00fSHugo Mills 4568abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 4569abccd00fSHugo Mills args32->stransid = args64->stransid; 4570abccd00fSHugo Mills args32->rtransid = args64->rtransid; 4571abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 4572abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 4573abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 4574abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 4575abccd00fSHugo Mills args32->flags = args64->flags; 4576abccd00fSHugo Mills 4577abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 4578abccd00fSHugo Mills if (ret) 4579abccd00fSHugo Mills ret = -EFAULT; 4580abccd00fSHugo Mills 4581abccd00fSHugo Mills out: 4582abccd00fSHugo Mills kfree(args32); 4583abccd00fSHugo Mills kfree(args64); 4584abccd00fSHugo Mills return ret; 4585abccd00fSHugo Mills } 4586abccd00fSHugo Mills #endif 4587abccd00fSHugo Mills 4588abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 4589abccd00fSHugo Mills void __user *arg) 4590abccd00fSHugo Mills { 4591abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 4592abccd00fSHugo Mills int ret = 0; 4593abccd00fSHugo Mills 4594abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 45957b9ea627SShailendra Verma if (IS_ERR(sa)) 45967b9ea627SShailendra Verma return PTR_ERR(sa); 4597abccd00fSHugo Mills 4598e4fed17aSChristian Brauner ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa); 4599abccd00fSHugo Mills 4600abccd00fSHugo Mills if (ret) 4601abccd00fSHugo Mills goto out; 4602abccd00fSHugo Mills 46038ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 46048ea05e3aSAlexander Block if (ret) 46058ea05e3aSAlexander Block ret = -EFAULT; 46068ea05e3aSAlexander Block 46078ea05e3aSAlexander Block out: 46088ea05e3aSAlexander Block kfree(sa); 46098ea05e3aSAlexander Block return ret; 46108ea05e3aSAlexander Block } 46118ea05e3aSAlexander Block 4612b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, 4613b929c1d8SMarcos Paulo de Souza void __user *arg) 4614867ab667Sjeff.liu { 4615a1b83ac5SAnand Jain size_t len; 4616867ab667Sjeff.liu int ret; 4617a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 4618a1b83ac5SAnand Jain 46190b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 46200b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 46210b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 4622a1b83ac5SAnand Jain 4623a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 4624867ab667Sjeff.liu 4625867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 46260b246afaSJeff Mahoney btrfs_warn(fs_info, 46270b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 46280b246afaSJeff Mahoney --len); 4629867ab667Sjeff.liu } 4630867ab667Sjeff.liu 4631867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 4632867ab667Sjeff.liu 4633867ab667Sjeff.liu return ret ? -EFAULT : 0; 4634867ab667Sjeff.liu } 4635867ab667Sjeff.liu 4636a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 4637a8bfd4abSjeff.liu { 46380b246afaSJeff Mahoney struct inode *inode = file_inode(file); 46390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 46400b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 46410b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 4642a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 4643a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 4644a8bfd4abSjeff.liu int ret; 4645a8bfd4abSjeff.liu 4646a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 4647a8bfd4abSjeff.liu return -EPERM; 4648a8bfd4abSjeff.liu 4649a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 4650a8bfd4abSjeff.liu return -EFAULT; 4651a8bfd4abSjeff.liu 4652a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 46530b246afaSJeff Mahoney btrfs_err(fs_info, 46545d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 4655a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 4656a8bfd4abSjeff.liu return -EINVAL; 4657a8bfd4abSjeff.liu } 4658a8bfd4abSjeff.liu 4659a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 4660a8bfd4abSjeff.liu if (ret) 4661a8bfd4abSjeff.liu return ret; 4662a8bfd4abSjeff.liu 4663a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 4664a8bfd4abSjeff.liu if (IS_ERR(trans)) { 4665a8bfd4abSjeff.liu ret = PTR_ERR(trans); 4666a8bfd4abSjeff.liu goto out_unlock; 4667a8bfd4abSjeff.liu } 4668a8bfd4abSjeff.liu 46690b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 4670a8bfd4abSjeff.liu strcpy(super_block->label, label); 46710b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 46723a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 4673a8bfd4abSjeff.liu 4674a8bfd4abSjeff.liu out_unlock: 4675a8bfd4abSjeff.liu mnt_drop_write_file(file); 4676a8bfd4abSjeff.liu return ret; 4677a8bfd4abSjeff.liu } 4678a8bfd4abSjeff.liu 46792eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 46802eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 46812eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 46822eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 46832eaa055fSJeff Mahoney 4684d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 46852eaa055fSJeff Mahoney { 46864d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 46872eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 46882eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 46892eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 46902eaa055fSJeff Mahoney }; 46912eaa055fSJeff Mahoney 46922eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 46932eaa055fSJeff Mahoney return -EFAULT; 46942eaa055fSJeff Mahoney 46952eaa055fSJeff Mahoney return 0; 46962eaa055fSJeff Mahoney } 46972eaa055fSJeff Mahoney 4698b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, 4699b929c1d8SMarcos Paulo de Souza void __user *arg) 47002eaa055fSJeff Mahoney { 47010b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 47022eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 47032eaa055fSJeff Mahoney 47042eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 47052eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 47062eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 47072eaa055fSJeff Mahoney 47082eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 47092eaa055fSJeff Mahoney return -EFAULT; 47102eaa055fSJeff Mahoney 47112eaa055fSJeff Mahoney return 0; 47122eaa055fSJeff Mahoney } 47132eaa055fSJeff Mahoney 47142ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 47153b02a68aSJeff Mahoney enum btrfs_feature_set set, 47162eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 47172eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 47182eaa055fSJeff Mahoney { 4719f10152bcSDavid Sterba const char *type = btrfs_feature_set_name(set); 47203b02a68aSJeff Mahoney char *names; 47212eaa055fSJeff Mahoney u64 disallowed, unsupported; 47222eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 47232eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 47242eaa055fSJeff Mahoney 47252eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 47262eaa055fSJeff Mahoney if (unsupported) { 47273b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 47283b02a68aSJeff Mahoney if (names) { 47290b246afaSJeff Mahoney btrfs_warn(fs_info, 47303b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 47313b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 47323b02a68aSJeff Mahoney kfree(names); 47333b02a68aSJeff Mahoney } else 47340b246afaSJeff Mahoney btrfs_warn(fs_info, 47352eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 47362eaa055fSJeff Mahoney type, unsupported); 47372eaa055fSJeff Mahoney return -EOPNOTSUPP; 47382eaa055fSJeff Mahoney } 47392eaa055fSJeff Mahoney 47402eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 47412eaa055fSJeff Mahoney if (disallowed) { 47423b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 47433b02a68aSJeff Mahoney if (names) { 47440b246afaSJeff Mahoney btrfs_warn(fs_info, 47453b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 47463b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 47473b02a68aSJeff Mahoney kfree(names); 47483b02a68aSJeff Mahoney } else 47490b246afaSJeff Mahoney btrfs_warn(fs_info, 47502eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 47512eaa055fSJeff Mahoney type, disallowed); 47522eaa055fSJeff Mahoney return -EPERM; 47532eaa055fSJeff Mahoney } 47542eaa055fSJeff Mahoney 47552eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 47562eaa055fSJeff Mahoney if (disallowed) { 47573b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 47583b02a68aSJeff Mahoney if (names) { 47590b246afaSJeff Mahoney btrfs_warn(fs_info, 47603b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 47613b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 47623b02a68aSJeff Mahoney kfree(names); 47633b02a68aSJeff Mahoney } else 47640b246afaSJeff Mahoney btrfs_warn(fs_info, 47652eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 47662eaa055fSJeff Mahoney type, disallowed); 47672eaa055fSJeff Mahoney return -EPERM; 47682eaa055fSJeff Mahoney } 47692eaa055fSJeff Mahoney 47702eaa055fSJeff Mahoney return 0; 47712eaa055fSJeff Mahoney } 47722eaa055fSJeff Mahoney 47732ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 47742ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 47752eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 47762eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 47772eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 47782eaa055fSJeff Mahoney 47792eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 47802eaa055fSJeff Mahoney { 47810b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47820b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47830b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 47840b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 47852eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 47862eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 47872eaa055fSJeff Mahoney u64 newflags; 47882eaa055fSJeff Mahoney int ret; 47892eaa055fSJeff Mahoney 47902eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 47912eaa055fSJeff Mahoney return -EPERM; 47922eaa055fSJeff Mahoney 47932eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 47942eaa055fSJeff Mahoney return -EFAULT; 47952eaa055fSJeff Mahoney 47962eaa055fSJeff Mahoney /* Nothing to do */ 47972eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 47982eaa055fSJeff Mahoney !flags[0].incompat_flags) 47992eaa055fSJeff Mahoney return 0; 48002eaa055fSJeff Mahoney 48012ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 48022eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 48032eaa055fSJeff Mahoney if (ret) 48042eaa055fSJeff Mahoney return ret; 48052eaa055fSJeff Mahoney 48062ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 48072eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 48082eaa055fSJeff Mahoney if (ret) 48092eaa055fSJeff Mahoney return ret; 48102eaa055fSJeff Mahoney 48112ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 48122eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 48132eaa055fSJeff Mahoney if (ret) 48142eaa055fSJeff Mahoney return ret; 48152eaa055fSJeff Mahoney 48167ab19625SDavid Sterba ret = mnt_want_write_file(file); 48177ab19625SDavid Sterba if (ret) 48187ab19625SDavid Sterba return ret; 48197ab19625SDavid Sterba 48208051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 48217ab19625SDavid Sterba if (IS_ERR(trans)) { 48227ab19625SDavid Sterba ret = PTR_ERR(trans); 48237ab19625SDavid Sterba goto out_drop_write; 48247ab19625SDavid Sterba } 48252eaa055fSJeff Mahoney 48260b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 48272eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 48282eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 48292eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 48302eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 48312eaa055fSJeff Mahoney 48322eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 48332eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 48342eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 48352eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 48362eaa055fSJeff Mahoney 48372eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 48382eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 48392eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 48402eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 48410b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 48422eaa055fSJeff Mahoney 48433a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 48447ab19625SDavid Sterba out_drop_write: 48457ab19625SDavid Sterba mnt_drop_write_file(file); 48467ab19625SDavid Sterba 48477ab19625SDavid Sterba return ret; 48482eaa055fSJeff Mahoney } 48492eaa055fSJeff Mahoney 48502351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 48512351f431SJosef Bacik { 48522351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 48532351f431SJosef Bacik int ret; 48542351f431SJosef Bacik 48552351f431SJosef Bacik if (compat) { 48562351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 48572351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 48582351f431SJosef Bacik 48592351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 48602351f431SJosef Bacik if (ret) 48612351f431SJosef Bacik return -EFAULT; 48622351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 48632351f431SJosef Bacik if (!arg) 48642351f431SJosef Bacik return -ENOMEM; 48652351f431SJosef Bacik arg->send_fd = args32.send_fd; 48662351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 48672351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 48682351f431SJosef Bacik arg->parent_root = args32.parent_root; 48692351f431SJosef Bacik arg->flags = args32.flags; 48702351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 48712351f431SJosef Bacik sizeof(args32.reserved)); 48722351f431SJosef Bacik #else 48732351f431SJosef Bacik return -ENOTTY; 48742351f431SJosef Bacik #endif 48752351f431SJosef Bacik } else { 48762351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 48772351f431SJosef Bacik if (IS_ERR(arg)) 48782351f431SJosef Bacik return PTR_ERR(arg); 48792351f431SJosef Bacik } 48802351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 48812351f431SJosef Bacik kfree(arg); 48822351f431SJosef Bacik return ret; 48832351f431SJosef Bacik } 48842351f431SJosef Bacik 4885f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 4886f46b5a66SChristoph Hellwig cmd, unsigned long arg) 4887f46b5a66SChristoph Hellwig { 48880b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48900b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48914bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 4892f46b5a66SChristoph Hellwig 4893f46b5a66SChristoph Hellwig switch (cmd) { 48946cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 48956cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 489640cf931fSEric Sandeen case FS_IOC_GETFSLABEL: 4897b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_fslabel(fs_info, argp); 489840cf931fSEric Sandeen case FS_IOC_SETFSLABEL: 489940cf931fSEric Sandeen return btrfs_ioctl_set_fslabel(file, argp); 4900f7039b1dSLi Dongyang case FITRIM: 4901b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_fitrim(fs_info, argp); 4902f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 4903fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 4904fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 4905fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 49063de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 4907fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 49086f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 49096f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 491076dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 4911949964c9SMarcos Paulo de Souza return btrfs_ioctl_snap_destroy(file, argp, false); 4912949964c9SMarcos Paulo de Souza case BTRFS_IOC_SNAP_DESTROY_V2: 4913949964c9SMarcos Paulo de Souza return btrfs_ioctl_snap_destroy(file, argp, true); 49140caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 49150caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 49160caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 49170caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 49186ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 49196ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 4920f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 49211e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 49221e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 49231e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 4924f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 4925198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 4926f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 49272ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 4928f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 4929da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 49306b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 49316b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 4932475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 49332ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 4934475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 49352ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 4936f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 49379ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 4938ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 4939ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 4940cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 4941cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 4942ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 4943ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 4944d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 4945d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 4946d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 4947d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 4948d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 4949d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 49501406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 49512ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 49529b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 49539b199859SFilipe David Borba Manana int ret; 49549b199859SFilipe David Borba Manana 49559db4dc24SNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false); 49569b199859SFilipe David Borba Manana if (ret) 49579b199859SFilipe David Borba Manana return ret; 49580b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 49592fad4e83SDavid Sterba /* 49602fad4e83SDavid Sterba * The transaction thread may want to do more work, 496101327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 49622fad4e83SDavid Sterba * processing uncleaned subvols. 49632fad4e83SDavid Sterba */ 49640b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 49659b199859SFilipe David Borba Manana return ret; 49669b199859SFilipe David Borba Manana } 496746204592SSage Weil case BTRFS_IOC_START_SYNC: 49689a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 496946204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 49702ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 4971475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 4972b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 4973475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 49742ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 4975475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 49762ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 4977c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 49789ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 4979837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 49802ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 498119a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 49822ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 49838ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 49848ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 4985abccd00fSHugo Mills #ifdef CONFIG_64BIT 4986abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 4987abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 4988abccd00fSHugo Mills #endif 498931db9f7cSAlexander Block case BTRFS_IOC_SEND: 49902351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 49912351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 49922351f431SJosef Bacik case BTRFS_IOC_SEND_32: 49932351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 49942351f431SJosef Bacik #endif 4995c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 49962ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 49975d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 4998905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 49995d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5000905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 50015d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5002905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 50035d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5004905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 50052f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 50062f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 50072f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 5008b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_status(fs_info, argp); 500957254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 5010b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_wait(fs_info, argp); 50113f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 50122ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 50132eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5014d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 50152eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 5016b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_features(fs_info, argp); 50172eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 50182eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5019b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5020b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 502142e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 502242e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 502323d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 502423d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 502514605409SBoris Burkov case FS_IOC_ENABLE_VERITY: 502614605409SBoris Burkov return fsverity_ioctl_enable(file, (const void __user *)argp); 502714605409SBoris Burkov case FS_IOC_MEASURE_VERITY: 502814605409SBoris Burkov return fsverity_ioctl_measure(file, argp); 5029f46b5a66SChristoph Hellwig } 5030f46b5a66SChristoph Hellwig 5031f46b5a66SChristoph Hellwig return -ENOTTY; 5032f46b5a66SChristoph Hellwig } 50334c63c245SLuke Dashjr 50344c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 50354c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 50364c63c245SLuke Dashjr { 50372a362249SJeff Mahoney /* 50382a362249SJeff Mahoney * These all access 32-bit values anyway so no further 50392a362249SJeff Mahoney * handling is necessary. 50402a362249SJeff Mahoney */ 50414c63c245SLuke Dashjr switch (cmd) { 50424c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 50434c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 50444c63c245SLuke Dashjr break; 50454c63c245SLuke Dashjr } 50464c63c245SLuke Dashjr 50474c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 50484c63c245SLuke Dashjr } 50494c63c245SLuke Dashjr #endif 5050