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> 311881fba8SOmar Sandoval #include <linux/sched/xacct.h> 32f46b5a66SChristoph Hellwig #include "ctree.h" 33f46b5a66SChristoph Hellwig #include "disk-io.h" 34949964c9SMarcos Paulo de Souza #include "export.h" 35f46b5a66SChristoph Hellwig #include "transaction.h" 36f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 37f46b5a66SChristoph Hellwig #include "print-tree.h" 38f46b5a66SChristoph Hellwig #include "volumes.h" 39925baeddSChris Mason #include "locking.h" 40d7728c96SJan Schmidt #include "backref.h" 41606686eeSJosef Bacik #include "rcu-string.h" 4231db9f7cSAlexander Block #include "send.h" 433f6bcfbdSStefan Behrens #include "dev-replace.h" 4463541927SFilipe David Borba Manana #include "props.h" 453b02a68aSJeff Mahoney #include "sysfs.h" 46fcebe456SJosef Bacik #include "qgroup.h" 471ec9a1aeSFilipe Manana #include "tree-log.h" 48ebb8765bSAnand Jain #include "compression.h" 498719aaaeSJosef Bacik #include "space-info.h" 5086736342SJosef Bacik #include "delalloc-space.h" 51aac0023cSJosef Bacik #include "block-group.h" 5222b398eeSQu Wenruo #include "subpage.h" 53f46b5a66SChristoph Hellwig 54abccd00fSHugo Mills #ifdef CONFIG_64BIT 55abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 56abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 57abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 58abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 59abccd00fSHugo Mills */ 60abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 61abccd00fSHugo Mills __u64 sec; 62abccd00fSHugo Mills __u32 nsec; 63abccd00fSHugo Mills } __attribute__ ((__packed__)); 64abccd00fSHugo Mills 65abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 66abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 67abccd00fSHugo Mills __u64 stransid; /* in */ 68abccd00fSHugo Mills __u64 rtransid; /* out */ 69abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 70abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 71abccd00fSHugo Mills __u64 flags; /* in */ 72abccd00fSHugo Mills __u64 reserved[16]; /* in */ 73abccd00fSHugo Mills } __attribute__ ((__packed__)); 74abccd00fSHugo Mills 75abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 76abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 77abccd00fSHugo Mills #endif 78abccd00fSHugo Mills 792351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 802351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 812351f431SJosef Bacik __s64 send_fd; /* in */ 822351f431SJosef Bacik __u64 clone_sources_count; /* in */ 832351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 842351f431SJosef Bacik __u64 parent_root; /* in */ 852351f431SJosef Bacik __u64 flags; /* in */ 86e77fbf99SDavid Sterba __u32 version; /* in */ 87e77fbf99SDavid Sterba __u8 reserved[28]; /* in */ 882351f431SJosef Bacik } __attribute__ ((__packed__)); 892351f431SJosef Bacik 902351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 912351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 921881fba8SOmar Sandoval 931881fba8SOmar Sandoval struct btrfs_ioctl_encoded_io_args_32 { 941881fba8SOmar Sandoval compat_uptr_t iov; 951881fba8SOmar Sandoval compat_ulong_t iovcnt; 961881fba8SOmar Sandoval __s64 offset; 971881fba8SOmar Sandoval __u64 flags; 981881fba8SOmar Sandoval __u64 len; 991881fba8SOmar Sandoval __u64 unencoded_len; 1001881fba8SOmar Sandoval __u64 unencoded_offset; 1011881fba8SOmar Sandoval __u32 compression; 1021881fba8SOmar Sandoval __u32 encryption; 1031881fba8SOmar Sandoval __u8 reserved[64]; 1041881fba8SOmar Sandoval }; 1051881fba8SOmar Sandoval 1061881fba8SOmar Sandoval #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \ 1071881fba8SOmar Sandoval struct btrfs_ioctl_encoded_io_args_32) 1087c0c7269SOmar Sandoval #define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \ 1097c0c7269SOmar Sandoval struct btrfs_ioctl_encoded_io_args_32) 1102351f431SJosef Bacik #endif 111abccd00fSHugo Mills 1126cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 1131905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 1141905a0f7SDavid Sterba unsigned int flags) 1156cbff00fSChristoph Hellwig { 1161905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 1176cbff00fSChristoph Hellwig return flags; 1181905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 1196cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1206cbff00fSChristoph Hellwig else 1216cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1226cbff00fSChristoph Hellwig } 123f46b5a66SChristoph Hellwig 1246cbff00fSChristoph Hellwig /* 125a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 126a157d4fdSDavid Sterba * ioctl. 1276cbff00fSChristoph Hellwig */ 12877eea05eSBoris Burkov static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode) 1296cbff00fSChristoph Hellwig { 1306cbff00fSChristoph Hellwig unsigned int iflags = 0; 13177eea05eSBoris Burkov u32 flags = binode->flags; 13214605409SBoris Burkov u32 ro_flags = binode->ro_flags; 1336cbff00fSChristoph Hellwig 1346cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1356cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1366cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1376cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1386cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1396cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1406cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1416cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1426cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1436cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1446cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1456cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 146d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 147d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 14814605409SBoris Burkov if (ro_flags & BTRFS_INODE_RO_VERITY) 14914605409SBoris Burkov iflags |= FS_VERITY_FL; 150d0092bddSLi Zefan 15113f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 152d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 15313f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 15413f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1556cbff00fSChristoph Hellwig 1566cbff00fSChristoph Hellwig return iflags; 1576cbff00fSChristoph Hellwig } 1586cbff00fSChristoph Hellwig 1596cbff00fSChristoph Hellwig /* 1606cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1616cbff00fSChristoph Hellwig */ 1627b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1636cbff00fSChristoph Hellwig { 1645c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1653cc79392SFilipe Manana unsigned int new_fl = 0; 1666cbff00fSChristoph Hellwig 1675c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1683cc79392SFilipe Manana new_fl |= S_SYNC; 1695c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1703cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1715c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1723cc79392SFilipe Manana new_fl |= S_APPEND; 1735c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1743cc79392SFilipe Manana new_fl |= S_NOATIME; 1755c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1763cc79392SFilipe Manana new_fl |= S_DIRSYNC; 17714605409SBoris Burkov if (binode->ro_flags & BTRFS_INODE_RO_VERITY) 17814605409SBoris Burkov new_fl |= S_VERITY; 1793cc79392SFilipe Manana 1803cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 18114605409SBoris Burkov S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC | 18214605409SBoris Burkov S_VERITY, new_fl); 1836cbff00fSChristoph Hellwig } 1846cbff00fSChristoph Hellwig 185f37c563bSDavid Sterba /* 186f37c563bSDavid Sterba * Check if @flags are a supported and valid set of FS_*_FL flags and that 187f37c563bSDavid Sterba * the old and new flags are not conflicting 188f37c563bSDavid Sterba */ 189f37c563bSDavid Sterba static int check_fsflags(unsigned int old_flags, unsigned int flags) 19075e7cb7fSLiu Bo { 19175e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 19275e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 19375e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 194e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 195e1e8fb6aSLi Zefan FS_NOCOW_FL)) 19675e7cb7fSLiu Bo return -EOPNOTSUPP; 19775e7cb7fSLiu Bo 198f37c563bSDavid Sterba /* COMPR and NOCOMP on new/old are valid */ 19975e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 20075e7cb7fSLiu Bo return -EINVAL; 20175e7cb7fSLiu Bo 202f37c563bSDavid Sterba if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL)) 203f37c563bSDavid Sterba return -EINVAL; 204f37c563bSDavid Sterba 205f37c563bSDavid Sterba /* NOCOW and compression options are mutually exclusive */ 206f37c563bSDavid Sterba if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL))) 207f37c563bSDavid Sterba return -EINVAL; 208f37c563bSDavid Sterba if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL))) 209f37c563bSDavid Sterba return -EINVAL; 210f37c563bSDavid Sterba 21175e7cb7fSLiu Bo return 0; 21275e7cb7fSLiu Bo } 21375e7cb7fSLiu Bo 214d206e9c9SNaohiro Aota static int check_fsflags_compatible(struct btrfs_fs_info *fs_info, 215d206e9c9SNaohiro Aota unsigned int flags) 216d206e9c9SNaohiro Aota { 217d206e9c9SNaohiro Aota if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL)) 218d206e9c9SNaohiro Aota return -EPERM; 219d206e9c9SNaohiro Aota 220d206e9c9SNaohiro Aota return 0; 221d206e9c9SNaohiro Aota } 222d206e9c9SNaohiro Aota 22397fc2977SMiklos Szeredi /* 22497fc2977SMiklos Szeredi * Set flags/xflags from the internal inode flags. The remaining items of 22597fc2977SMiklos Szeredi * fsxattr are zeroed. 22697fc2977SMiklos Szeredi */ 22797fc2977SMiklos Szeredi int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 2286cbff00fSChristoph Hellwig { 22997fc2977SMiklos Szeredi struct btrfs_inode *binode = BTRFS_I(d_inode(dentry)); 23097fc2977SMiklos Szeredi 23177eea05eSBoris Burkov fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode)); 23297fc2977SMiklos Szeredi return 0; 23397fc2977SMiklos Szeredi } 23497fc2977SMiklos Szeredi 23597fc2977SMiklos Szeredi int btrfs_fileattr_set(struct user_namespace *mnt_userns, 23697fc2977SMiklos Szeredi struct dentry *dentry, struct fileattr *fa) 23797fc2977SMiklos Szeredi { 23897fc2977SMiklos Szeredi struct inode *inode = d_inode(dentry); 2390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2405c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 2415c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 2426cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 2435aca2842SDarrick J. Wong unsigned int fsflags, old_fsflags; 2446cbff00fSChristoph Hellwig int ret; 245ff9fef55SAnand Jain const char *comp = NULL; 246f37c563bSDavid Sterba u32 binode_flags; 2476cbff00fSChristoph Hellwig 248b83cc969SLi Zefan if (btrfs_root_readonly(root)) 249b83cc969SLi Zefan return -EROFS; 250b83cc969SLi Zefan 25197fc2977SMiklos Szeredi if (fileattr_has_fsx(fa)) 25297fc2977SMiklos Szeredi return -EOPNOTSUPP; 2536cbff00fSChristoph Hellwig 25497fc2977SMiklos Szeredi fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags); 25577eea05eSBoris Burkov old_fsflags = btrfs_inode_flags_to_fsflags(binode); 25697fc2977SMiklos Szeredi ret = check_fsflags(old_fsflags, fsflags); 257e7848683SJan Kara if (ret) 258e7848683SJan Kara return ret; 259e7848683SJan Kara 260d206e9c9SNaohiro Aota ret = check_fsflags_compatible(fs_info, fsflags); 261d206e9c9SNaohiro Aota if (ret) 26297fc2977SMiklos Szeredi return ret; 263d206e9c9SNaohiro Aota 264f37c563bSDavid Sterba binode_flags = binode->flags; 2655c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 266d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_SYNC; 2676cbff00fSChristoph Hellwig else 268d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_SYNC; 2695c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 270d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_IMMUTABLE; 2716cbff00fSChristoph Hellwig else 272d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_IMMUTABLE; 2735c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 274d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_APPEND; 2756cbff00fSChristoph Hellwig else 276d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_APPEND; 2775c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 278d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODUMP; 2796cbff00fSChristoph Hellwig else 280d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODUMP; 2815c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 282d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOATIME; 2836cbff00fSChristoph Hellwig else 284d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOATIME; 28597fc2977SMiklos Szeredi 28697fc2977SMiklos Szeredi /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */ 28797fc2977SMiklos Szeredi if (!fa->flags_valid) { 28897fc2977SMiklos Szeredi /* 1 item for the inode */ 28997fc2977SMiklos Szeredi trans = btrfs_start_transaction(root, 1); 2909b8a233bSRitesh Harjani if (IS_ERR(trans)) 2919b8a233bSRitesh Harjani return PTR_ERR(trans); 29297fc2977SMiklos Szeredi goto update_flags; 29397fc2977SMiklos Szeredi } 29497fc2977SMiklos Szeredi 2955c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 296d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_DIRSYNC; 2976cbff00fSChristoph Hellwig else 298d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_DIRSYNC; 2995c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 30044e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 3017e97b8daSDavid Sterba /* 3027e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 3037e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 3047e97b8daSDavid Sterba * status of the file and will not set it. 3057e97b8daSDavid Sterba */ 3067e97b8daSDavid Sterba if (inode->i_size == 0) 307d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW | 308d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM; 3097e97b8daSDavid Sterba } else { 310d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW; 3117e97b8daSDavid Sterba } 3127e97b8daSDavid Sterba } else { 3137e97b8daSDavid Sterba /* 31401327610SNicholas D Steeves * Revert back under same assumptions as above 3157e97b8daSDavid Sterba */ 31644e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 3177e97b8daSDavid Sterba if (inode->i_size == 0) 318d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_NODATACOW | 319d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM); 3207e97b8daSDavid Sterba } else { 321d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODATACOW; 3227e97b8daSDavid Sterba } 3237e97b8daSDavid Sterba } 3246cbff00fSChristoph Hellwig 32575e7cb7fSLiu Bo /* 32675e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 32775e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 32875e7cb7fSLiu Bo * things smaller. 32975e7cb7fSLiu Bo */ 3305c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 331d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_COMPRESS; 332d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOCOMPRESS; 3335c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 33463541927SFilipe David Borba Manana 33597fc2977SMiklos Szeredi if (IS_SWAPFILE(inode)) 33697fc2977SMiklos Szeredi return -ETXTBSY; 337eede2bf3SOmar Sandoval 338d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_COMPRESS; 339d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOCOMPRESS; 34063541927SFilipe David Borba Manana 34193370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 34293370509SDavid Sterba if (!comp || comp[0] == 0) 34393370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 344ebcb904dSLi Zefan } else { 345d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 34675e7cb7fSLiu Bo } 3476cbff00fSChristoph Hellwig 348ff9fef55SAnand Jain /* 349ff9fef55SAnand Jain * 1 for inode item 350ff9fef55SAnand Jain * 2 for properties 351ff9fef55SAnand Jain */ 352ff9fef55SAnand Jain trans = btrfs_start_transaction(root, 3); 35397fc2977SMiklos Szeredi if (IS_ERR(trans)) 35497fc2977SMiklos Szeredi return PTR_ERR(trans); 3556cbff00fSChristoph Hellwig 356ff9fef55SAnand Jain if (comp) { 357ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, 358ff9fef55SAnand Jain strlen(comp), 0); 359ff9fef55SAnand Jain if (ret) { 360ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 361ff9fef55SAnand Jain goto out_end_trans; 362ff9fef55SAnand Jain } 363ff9fef55SAnand Jain } else { 364ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 365ff9fef55SAnand Jain 0, 0); 366ff9fef55SAnand Jain if (ret && ret != -ENODATA) { 367ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 368ff9fef55SAnand Jain goto out_end_trans; 369ff9fef55SAnand Jain } 370ff9fef55SAnand Jain } 371ff9fef55SAnand Jain 37297fc2977SMiklos Szeredi update_flags: 373d2b8fcfeSAnand Jain binode->flags = binode_flags; 3747b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3750c4d2d95SJosef Bacik inode_inc_iversion(inode); 376c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3779a56fcd1SNikolay Borisov ret = btrfs_update_inode(trans, root, BTRFS_I(inode)); 3786cbff00fSChristoph Hellwig 379ff9fef55SAnand Jain out_end_trans: 3803a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3812d4e6f6aSliubo return ret; 3826cbff00fSChristoph Hellwig } 3836cbff00fSChristoph Hellwig 3840d7ed32cSDavid Sterba /* 3850d7ed32cSDavid Sterba * Start exclusive operation @type, return true on success 3860d7ed32cSDavid Sterba */ 387c3e1f96cSGoldwyn Rodrigues bool btrfs_exclop_start(struct btrfs_fs_info *fs_info, 388c3e1f96cSGoldwyn Rodrigues enum btrfs_exclusive_operation type) 389c3e1f96cSGoldwyn Rodrigues { 3900d7ed32cSDavid Sterba bool ret = false; 3910d7ed32cSDavid Sterba 3920d7ed32cSDavid Sterba spin_lock(&fs_info->super_lock); 3930d7ed32cSDavid Sterba if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) { 3940d7ed32cSDavid Sterba fs_info->exclusive_operation = type; 3950d7ed32cSDavid Sterba ret = true; 3960d7ed32cSDavid Sterba } 3970d7ed32cSDavid Sterba spin_unlock(&fs_info->super_lock); 3980d7ed32cSDavid Sterba 3990d7ed32cSDavid Sterba return ret; 400c3e1f96cSGoldwyn Rodrigues } 401c3e1f96cSGoldwyn Rodrigues 402578bda9eSDavid Sterba /* 403578bda9eSDavid Sterba * Conditionally allow to enter the exclusive operation in case it's compatible 404578bda9eSDavid Sterba * with the running one. This must be paired with btrfs_exclop_start_unlock and 405578bda9eSDavid Sterba * btrfs_exclop_finish. 406578bda9eSDavid Sterba * 407578bda9eSDavid Sterba * Compatibility: 408578bda9eSDavid Sterba * - the same type is already running 409621a1ee1SNikolay Borisov * - when trying to add a device and balance has been paused 410578bda9eSDavid Sterba * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller 411578bda9eSDavid Sterba * must check the condition first that would allow none -> @type 412578bda9eSDavid Sterba */ 413578bda9eSDavid Sterba bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info, 414578bda9eSDavid Sterba enum btrfs_exclusive_operation type) 415578bda9eSDavid Sterba { 416578bda9eSDavid Sterba spin_lock(&fs_info->super_lock); 417621a1ee1SNikolay Borisov if (fs_info->exclusive_operation == type || 418621a1ee1SNikolay Borisov (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED && 419621a1ee1SNikolay Borisov type == BTRFS_EXCLOP_DEV_ADD)) 420578bda9eSDavid Sterba return true; 421578bda9eSDavid Sterba 422578bda9eSDavid Sterba spin_unlock(&fs_info->super_lock); 423578bda9eSDavid Sterba return false; 424578bda9eSDavid Sterba } 425578bda9eSDavid Sterba 426578bda9eSDavid Sterba void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info) 427578bda9eSDavid Sterba { 428578bda9eSDavid Sterba spin_unlock(&fs_info->super_lock); 429578bda9eSDavid Sterba } 430578bda9eSDavid Sterba 431c3e1f96cSGoldwyn Rodrigues void btrfs_exclop_finish(struct btrfs_fs_info *fs_info) 432c3e1f96cSGoldwyn Rodrigues { 4330d7ed32cSDavid Sterba spin_lock(&fs_info->super_lock); 434c3e1f96cSGoldwyn Rodrigues WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE); 4350d7ed32cSDavid Sterba spin_unlock(&fs_info->super_lock); 43666a2823cSGoldwyn Rodrigues sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation"); 437c3e1f96cSGoldwyn Rodrigues } 438c3e1f96cSGoldwyn Rodrigues 439efc0e69cSNikolay Borisov void btrfs_exclop_balance(struct btrfs_fs_info *fs_info, 440efc0e69cSNikolay Borisov enum btrfs_exclusive_operation op) 441efc0e69cSNikolay Borisov { 442efc0e69cSNikolay Borisov switch (op) { 443efc0e69cSNikolay Borisov case BTRFS_EXCLOP_BALANCE_PAUSED: 444efc0e69cSNikolay Borisov spin_lock(&fs_info->super_lock); 445efc0e69cSNikolay Borisov ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE || 446efc0e69cSNikolay Borisov fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD); 447efc0e69cSNikolay Borisov fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED; 448efc0e69cSNikolay Borisov spin_unlock(&fs_info->super_lock); 449efc0e69cSNikolay Borisov break; 450efc0e69cSNikolay Borisov case BTRFS_EXCLOP_BALANCE: 451efc0e69cSNikolay Borisov spin_lock(&fs_info->super_lock); 452efc0e69cSNikolay Borisov ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED); 453efc0e69cSNikolay Borisov fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE; 454efc0e69cSNikolay Borisov spin_unlock(&fs_info->super_lock); 455efc0e69cSNikolay Borisov break; 456efc0e69cSNikolay Borisov default: 457efc0e69cSNikolay Borisov btrfs_warn(fs_info, 458efc0e69cSNikolay Borisov "invalid exclop balance operation %d requested", op); 459efc0e69cSNikolay Borisov } 460efc0e69cSNikolay Borisov } 461efc0e69cSNikolay Borisov 462dc408ccdSSahil Kang static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg) 4636cbff00fSChristoph Hellwig { 4646cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4656cbff00fSChristoph Hellwig } 466f46b5a66SChristoph Hellwig 467b929c1d8SMarcos Paulo de Souza static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, 468b929c1d8SMarcos Paulo de Souza void __user *arg) 469f7039b1dSLi Dongyang { 470f7039b1dSLi Dongyang struct btrfs_device *device; 471f7039b1dSLi Dongyang struct fstrim_range range; 472f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 473f7039b1dSLi Dongyang u64 num_devices = 0; 474f7039b1dSLi Dongyang int ret; 475f7039b1dSLi Dongyang 476f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 477f7039b1dSLi Dongyang return -EPERM; 478f7039b1dSLi Dongyang 479f35f06c3SFilipe Manana /* 4801cb3dc3fSNaohiro Aota * btrfs_trim_block_group() depends on space cache, which is not 4811cb3dc3fSNaohiro Aota * available in zoned filesystem. So, disallow fitrim on a zoned 4821cb3dc3fSNaohiro Aota * filesystem for now. 4831cb3dc3fSNaohiro Aota */ 4841cb3dc3fSNaohiro Aota if (btrfs_is_zoned(fs_info)) 4851cb3dc3fSNaohiro Aota return -EOPNOTSUPP; 4861cb3dc3fSNaohiro Aota 4871cb3dc3fSNaohiro Aota /* 488f35f06c3SFilipe Manana * If the fs is mounted with nologreplay, which requires it to be 489f35f06c3SFilipe Manana * mounted in RO mode as well, we can not allow discard on free space 490f35f06c3SFilipe Manana * inside block groups, because log trees refer to extents that are not 491f35f06c3SFilipe Manana * pinned in a block group's free space cache (pinning the extents is 492f35f06c3SFilipe Manana * precisely the first phase of replaying a log tree). 493f35f06c3SFilipe Manana */ 494f35f06c3SFilipe Manana if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 495f35f06c3SFilipe Manana return -EROFS; 496f35f06c3SFilipe Manana 4971f78160cSXiao Guangrong rcu_read_lock(); 4981f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 4991f78160cSXiao Guangrong dev_list) { 5007b47ef52SChristoph Hellwig if (!device->bdev || !bdev_max_discard_sectors(device->bdev)) 501f7039b1dSLi Dongyang continue; 502f7039b1dSLi Dongyang num_devices++; 5037b47ef52SChristoph Hellwig minlen = min_t(u64, bdev_discard_granularity(device->bdev), 504f7039b1dSLi Dongyang minlen); 505f7039b1dSLi Dongyang } 5061f78160cSXiao Guangrong rcu_read_unlock(); 507f4c697e6SLukas Czerner 508f7039b1dSLi Dongyang if (!num_devices) 509f7039b1dSLi Dongyang return -EOPNOTSUPP; 510f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 511f7039b1dSLi Dongyang return -EFAULT; 5126ba9fc8eSQu Wenruo 5136ba9fc8eSQu Wenruo /* 5146ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 5156ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 5166ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 5176ba9fc8eSQu Wenruo */ 5186ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 519f4c697e6SLukas Czerner return -EINVAL; 520f7039b1dSLi Dongyang 521f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5222ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 523f7039b1dSLi Dongyang if (ret < 0) 524f7039b1dSLi Dongyang return ret; 525f7039b1dSLi Dongyang 526f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 527f7039b1dSLi Dongyang return -EFAULT; 528f7039b1dSLi Dongyang 529f7039b1dSLi Dongyang return 0; 530f7039b1dSLi Dongyang } 531f7039b1dSLi Dongyang 532e1f60a65SDavid Sterba int __pure btrfs_is_empty_uuid(u8 *uuid) 533dd5f9615SStefan Behrens { 53446e0f66aSChris Mason int i; 53546e0f66aSChris Mason 53646e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 53746e0f66aSChris Mason if (uuid[i]) 53846e0f66aSChris Mason return 0; 53946e0f66aSChris Mason } 54046e0f66aSChris Mason return 1; 541dd5f9615SStefan Behrens } 542dd5f9615SStefan Behrens 5433538d68dSOmar Sandoval /* 5443538d68dSOmar Sandoval * Calculate the number of transaction items to reserve for creating a subvolume 5453538d68dSOmar Sandoval * or snapshot, not including the inode, directory entries, or parent directory. 5463538d68dSOmar Sandoval */ 5473538d68dSOmar Sandoval static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit) 5483538d68dSOmar Sandoval { 5493538d68dSOmar Sandoval /* 5503538d68dSOmar Sandoval * 1 to add root block 5513538d68dSOmar Sandoval * 1 to add root item 5523538d68dSOmar Sandoval * 1 to add root ref 5533538d68dSOmar Sandoval * 1 to add root backref 5543538d68dSOmar Sandoval * 1 to add UUID item 5553538d68dSOmar Sandoval * 1 to add qgroup info 5563538d68dSOmar Sandoval * 1 to add qgroup limit 5573538d68dSOmar Sandoval * 5583538d68dSOmar Sandoval * Ideally the last two would only be accounted if qgroups are enabled, 5593538d68dSOmar Sandoval * but that can change between now and the time we would insert them. 5603538d68dSOmar Sandoval */ 5613538d68dSOmar Sandoval unsigned int num_items = 7; 5623538d68dSOmar Sandoval 5633538d68dSOmar Sandoval if (inherit) { 5643538d68dSOmar Sandoval /* 2 to add qgroup relations for each inherited qgroup */ 5653538d68dSOmar Sandoval num_items += 2 * inherit->num_qgroups; 5663538d68dSOmar Sandoval } 5673538d68dSOmar Sandoval return num_items; 5683538d68dSOmar Sandoval } 5693538d68dSOmar Sandoval 5704d4340c9SChristian Brauner static noinline int create_subvol(struct user_namespace *mnt_userns, 5714d4340c9SChristian Brauner struct inode *dir, struct dentry *dentry, 5728696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 573f46b5a66SChristoph Hellwig { 5740b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 575f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 576f46b5a66SChristoph Hellwig struct btrfs_key key; 57749a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 578f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 579f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 580d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 58176dda93cSYan, Zheng struct btrfs_root *new_root; 582d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 58395582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5843538d68dSOmar Sandoval struct btrfs_new_inode_args new_inode_args = { 5853538d68dSOmar Sandoval .dir = dir, 5863538d68dSOmar Sandoval .dentry = dentry, 5873538d68dSOmar Sandoval .subvol = true, 5883538d68dSOmar Sandoval }; 5893538d68dSOmar Sandoval unsigned int trans_num_items; 590f46b5a66SChristoph Hellwig int ret; 5912256e901SOmar Sandoval dev_t anon_dev; 592f46b5a66SChristoph Hellwig u64 objectid; 593f46b5a66SChristoph Hellwig 59449a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 59549a3c4d9SDavid Sterba if (!root_item) 59649a3c4d9SDavid Sterba return -ENOMEM; 59749a3c4d9SDavid Sterba 598543068a2SNikolay Borisov ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid); 5992fbe8c8aSAl Viro if (ret) 6002256e901SOmar Sandoval goto out_root_item; 6012dfb1e43SQu Wenruo 602e09fe2d2SQu Wenruo /* 603e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 60401327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 605e09fe2d2SQu Wenruo */ 60649a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 60749a3c4d9SDavid Sterba ret = -ENOSPC; 6082256e901SOmar Sandoval goto out_root_item; 60949a3c4d9SDavid Sterba } 610e09fe2d2SQu Wenruo 6112256e901SOmar Sandoval ret = get_anon_bdev(&anon_dev); 6122256e901SOmar Sandoval if (ret < 0) 6132256e901SOmar Sandoval goto out_root_item; 6142256e901SOmar Sandoval 6153538d68dSOmar Sandoval new_inode_args.inode = btrfs_new_subvol_inode(mnt_userns, dir); 6163538d68dSOmar Sandoval if (!new_inode_args.inode) { 617a1fd0c35SOmar Sandoval ret = -ENOMEM; 618a1fd0c35SOmar Sandoval goto out_anon_dev; 619a1fd0c35SOmar Sandoval } 6203538d68dSOmar Sandoval ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items); 621d5c12070SMiao Xie if (ret) 622a1fd0c35SOmar Sandoval goto out_inode; 6233538d68dSOmar Sandoval trans_num_items += create_subvol_num_items(inherit); 6243538d68dSOmar Sandoval 6253538d68dSOmar Sandoval btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 6263538d68dSOmar Sandoval ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 6273538d68dSOmar Sandoval trans_num_items, false); 6283538d68dSOmar Sandoval if (ret) 6293538d68dSOmar Sandoval goto out_new_inode_args; 630f46b5a66SChristoph Hellwig 631d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 632d5c12070SMiao Xie if (IS_ERR(trans)) { 633d5c12070SMiao Xie ret = PTR_ERR(trans); 634e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &block_rsv); 6353538d68dSOmar Sandoval goto out_new_inode_args; 636d5c12070SMiao Xie } 637d5c12070SMiao Xie trans->block_rsv = &block_rsv; 638d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 639f46b5a66SChristoph Hellwig 640a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 6416f72c7e2SArne Jansen if (ret) 6422256e901SOmar Sandoval goto out; 6436f72c7e2SArne Jansen 6449631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 6459631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 6468e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6478e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6482256e901SOmar Sandoval goto out; 6498e8a1e31SJosef Bacik } 650f46b5a66SChristoph Hellwig 651f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 652f46b5a66SChristoph Hellwig 65349a3c4d9SDavid Sterba inode_item = &root_item->inode; 6543cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6553cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6563cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 657da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6580b246afaSJeff Mahoney fs_info->nodesize); 6593cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 660f46b5a66SChristoph Hellwig 66149a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 66249a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6633cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 66408fe4db1SLi Zefan 66549a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 66649a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 66749a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 66849a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 66949a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 67049a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 671f46b5a66SChristoph Hellwig 67249a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 67349a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 674807fc790SAndy Shevchenko generate_random_guid(root_item->uuid); 67549a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 67649a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 67749a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 67849a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 67949a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 680f46b5a66SChristoph Hellwig 681925baeddSChris Mason btrfs_tree_unlock(leaf); 682f46b5a66SChristoph Hellwig 68369948022SNikolay Borisov btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID); 684f46b5a66SChristoph Hellwig 685f46b5a66SChristoph Hellwig key.objectid = objectid; 6865d4f98a2SYan Zheng key.offset = 0; 687962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6880b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 68949a3c4d9SDavid Sterba root_item); 69067addf29SFilipe Manana if (ret) { 69167addf29SFilipe Manana /* 69267addf29SFilipe Manana * Since we don't abort the transaction in this case, free the 69367addf29SFilipe Manana * tree block so that we don't leak space and leave the 69467addf29SFilipe Manana * filesystem in an inconsistent state (an extent item in the 6957a163608SFilipe Manana * extent tree with a backreference for a root that does not 696212a58fdSFilipe Manana * exists). 69767addf29SFilipe Manana */ 698212a58fdSFilipe Manana btrfs_tree_lock(leaf); 699212a58fdSFilipe Manana btrfs_clean_tree_block(leaf); 700212a58fdSFilipe Manana btrfs_tree_unlock(leaf); 7017a163608SFilipe Manana btrfs_free_tree_block(trans, objectid, leaf, 0, 1); 70267addf29SFilipe Manana free_extent_buffer(leaf); 7032256e901SOmar Sandoval goto out; 70467addf29SFilipe Manana } 70567addf29SFilipe Manana 70667addf29SFilipe Manana free_extent_buffer(leaf); 70767addf29SFilipe Manana leaf = NULL; 708f46b5a66SChristoph Hellwig 7092dfb1e43SQu Wenruo new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev); 71079787eaaSJeff Mahoney if (IS_ERR(new_root)) { 71179787eaaSJeff Mahoney ret = PTR_ERR(new_root); 71266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 7132256e901SOmar Sandoval goto out; 71479787eaaSJeff Mahoney } 7152256e901SOmar Sandoval /* anon_dev is owned by new_root now. */ 7162dfb1e43SQu Wenruo anon_dev = 0; 7173538d68dSOmar Sandoval BTRFS_I(new_inode_args.inode)->root = new_root; 7183538d68dSOmar Sandoval /* ... and new_root is owned by new_inode_args.inode now. */ 71976dda93cSYan, Zheng 720221581e4SJosef Bacik ret = btrfs_record_root_in_trans(trans, new_root); 721221581e4SJosef Bacik if (ret) { 722221581e4SJosef Bacik btrfs_abort_transaction(trans, ret); 7232256e901SOmar Sandoval goto out; 724c7e54b51SJosef Bacik } 7250660b5afSChris Mason 726cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7276bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 728caae78e0SOmar Sandoval if (ret) { 72966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 730caae78e0SOmar Sandoval goto out; 731caae78e0SOmar Sandoval } 732dd5f9615SStefan Behrens 733caae78e0SOmar Sandoval ret = btrfs_create_new_inode(trans, &new_inode_args); 734caae78e0SOmar Sandoval if (ret) { 735caae78e0SOmar Sandoval btrfs_abort_transaction(trans, ret); 736caae78e0SOmar Sandoval goto out; 737caae78e0SOmar Sandoval } 738caae78e0SOmar Sandoval 739caae78e0SOmar Sandoval d_instantiate_new(dentry, new_inode_args.inode); 740caae78e0SOmar Sandoval new_inode_args.inode = NULL; 741f46b5a66SChristoph Hellwig 7422256e901SOmar Sandoval out: 743d5c12070SMiao Xie trans->block_rsv = NULL; 744d5c12070SMiao Xie trans->bytes_reserved = 0; 745e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &block_rsv); 746de6e8200SLiu Bo 7471b58ae0eSFilipe Manana if (ret) 7481b58ae0eSFilipe Manana btrfs_end_transaction(trans); 7491b58ae0eSFilipe Manana else 7501b58ae0eSFilipe Manana ret = btrfs_commit_transaction(trans); 7513538d68dSOmar Sandoval out_new_inode_args: 7523538d68dSOmar Sandoval btrfs_new_inode_args_destroy(&new_inode_args); 753a1fd0c35SOmar Sandoval out_inode: 7543538d68dSOmar Sandoval iput(new_inode_args.inode); 7552256e901SOmar Sandoval out_anon_dev: 7562dfb1e43SQu Wenruo if (anon_dev) 7572dfb1e43SQu Wenruo free_anon_bdev(anon_dev); 7582256e901SOmar Sandoval out_root_item: 75949a3c4d9SDavid Sterba kfree(root_item); 76049a3c4d9SDavid Sterba return ret; 761f46b5a66SChristoph Hellwig } 762f46b5a66SChristoph Hellwig 763e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 7649babda9fSNikolay Borisov struct dentry *dentry, bool readonly, 765e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 766f46b5a66SChristoph Hellwig { 7670b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7682e4bfab9SYan, Zheng struct inode *inode; 769f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 7703538d68dSOmar Sandoval unsigned int trans_num_items; 771f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7722e4bfab9SYan, Zheng int ret; 773f46b5a66SChristoph Hellwig 774813febdbSJosef Bacik /* We do not support snapshotting right now. */ 775813febdbSJosef Bacik if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 776813febdbSJosef Bacik btrfs_warn(fs_info, 777813febdbSJosef Bacik "extent tree v2 doesn't support snapshotting yet"); 778813febdbSJosef Bacik return -EOPNOTSUPP; 779813febdbSJosef Bacik } 780813febdbSJosef Bacik 78192a7cc42SQu Wenruo if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) 782f46b5a66SChristoph Hellwig return -EINVAL; 783f46b5a66SChristoph Hellwig 784eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 785eede2bf3SOmar Sandoval btrfs_warn(fs_info, 786eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 787eede2bf3SOmar Sandoval return -ETXTBSY; 788eede2bf3SOmar Sandoval } 789eede2bf3SOmar Sandoval 79023269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 791a1ee7362SDavid Sterba if (!pending_snapshot) 792a1ee7362SDavid Sterba return -ENOMEM; 793a1ee7362SDavid Sterba 7942dfb1e43SQu Wenruo ret = get_anon_bdev(&pending_snapshot->anon_dev); 7952dfb1e43SQu Wenruo if (ret < 0) 7962dfb1e43SQu Wenruo goto free_pending; 797b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 79823269bf5SDavid Sterba GFP_KERNEL); 7998546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 8008546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 801b0c0ea63SDavid Sterba ret = -ENOMEM; 802b0c0ea63SDavid Sterba goto free_pending; 803b0c0ea63SDavid Sterba } 804b0c0ea63SDavid Sterba 80566d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 80666d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 807d5c12070SMiao Xie /* 8083538d68dSOmar Sandoval * 1 to add dir item 8093538d68dSOmar Sandoval * 1 to add dir index 8103538d68dSOmar Sandoval * 1 to update parent inode item 811d5c12070SMiao Xie */ 8123538d68dSOmar Sandoval trans_num_items = create_subvol_num_items(inherit) + 3; 813d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 8143538d68dSOmar Sandoval &pending_snapshot->block_rsv, 8153538d68dSOmar Sandoval trans_num_items, false); 816d5c12070SMiao Xie if (ret) 817c11fbb6eSRobbie Ko goto free_pending; 818d5c12070SMiao Xie 819a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 820a22285a6SYan, Zheng pending_snapshot->root = root; 821b83cc969SLi Zefan pending_snapshot->readonly = readonly; 822e9662f70SMiao Xie pending_snapshot->dir = dir; 8238696c533SMiao Xie pending_snapshot->inherit = inherit; 824a22285a6SYan, Zheng 825d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 826a22285a6SYan, Zheng if (IS_ERR(trans)) { 827a22285a6SYan, Zheng ret = PTR_ERR(trans); 828a22285a6SYan, Zheng goto fail; 829a22285a6SYan, Zheng } 830a22285a6SYan, Zheng 83128b21c55SFilipe Manana trans->pending_snapshot = pending_snapshot; 8329babda9fSNikolay Borisov 8333a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 834aec8030aSMiao Xie if (ret) 835c37b2b62SJosef Bacik goto fail; 836a22285a6SYan, Zheng 837a22285a6SYan, Zheng ret = pending_snapshot->error; 838f46b5a66SChristoph Hellwig if (ret) 8392e4bfab9SYan, Zheng goto fail; 840f46b5a66SChristoph Hellwig 841d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 842d3797308SChris Mason if (ret) 843d3797308SChris Mason goto fail; 844d3797308SChris Mason 8452b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8462e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8472e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8482e4bfab9SYan, Zheng goto fail; 8492e4bfab9SYan, Zheng } 8505662344bSTsutomu Itoh 8512e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8522e4bfab9SYan, Zheng ret = 0; 8532dfb1e43SQu Wenruo pending_snapshot->anon_dev = 0; 8542e4bfab9SYan, Zheng fail: 8552dfb1e43SQu Wenruo /* Prevent double freeing of anon_dev */ 8562dfb1e43SQu Wenruo if (ret && pending_snapshot->snap) 8572dfb1e43SQu Wenruo pending_snapshot->snap->anon_dev = 0; 85800246528SJosef Bacik btrfs_put_root(pending_snapshot->snap); 859e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv); 860b0c0ea63SDavid Sterba free_pending: 8612dfb1e43SQu Wenruo if (pending_snapshot->anon_dev) 8622dfb1e43SQu Wenruo free_anon_bdev(pending_snapshot->anon_dev); 863b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8648546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 865a1ee7362SDavid Sterba kfree(pending_snapshot); 866a1ee7362SDavid Sterba 867f46b5a66SChristoph Hellwig return ret; 868f46b5a66SChristoph Hellwig } 869f46b5a66SChristoph Hellwig 8704260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8714260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8724260f7c7SSage Weil * whether the type of victim is right. 8734260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8744260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8754260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8764260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8774260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8784260f7c7SSage Weil * a. be owner of dir, or 8794260f7c7SSage Weil * b. be owner of victim, or 8804260f7c7SSage Weil * c. have CAP_FOWNER capability 88101327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8824260f7c7SSage Weil * links pointing to it. 8834260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8844260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8854260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8864260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8874260f7c7SSage Weil * nfs_async_unlink(). 8884260f7c7SSage Weil */ 8894260f7c7SSage Weil 890c4ed533bSChristian Brauner static int btrfs_may_delete(struct user_namespace *mnt_userns, 891c4ed533bSChristian Brauner struct inode *dir, struct dentry *victim, int isdir) 8924260f7c7SSage Weil { 8934260f7c7SSage Weil int error; 8944260f7c7SSage Weil 8952b0143b5SDavid Howells if (d_really_is_negative(victim)) 8964260f7c7SSage Weil return -ENOENT; 8974260f7c7SSage Weil 8982b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8994fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 9004260f7c7SSage Weil 901c4ed533bSChristian Brauner error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); 9024260f7c7SSage Weil if (error) 9034260f7c7SSage Weil return error; 9044260f7c7SSage Weil if (IS_APPEND(dir)) 9054260f7c7SSage Weil return -EPERM; 906c4ed533bSChristian Brauner if (check_sticky(mnt_userns, dir, d_inode(victim)) || 907ba73d987SChristian Brauner IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) || 908ba73d987SChristian Brauner IS_SWAPFILE(d_inode(victim))) 9094260f7c7SSage Weil return -EPERM; 9104260f7c7SSage Weil if (isdir) { 911e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9124260f7c7SSage Weil return -ENOTDIR; 9134260f7c7SSage Weil if (IS_ROOT(victim)) 9144260f7c7SSage Weil return -EBUSY; 915e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9164260f7c7SSage Weil return -EISDIR; 9174260f7c7SSage Weil if (IS_DEADDIR(dir)) 9184260f7c7SSage Weil return -ENOENT; 9194260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9204260f7c7SSage Weil return -EBUSY; 9214260f7c7SSage Weil return 0; 9224260f7c7SSage Weil } 9234260f7c7SSage Weil 924cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 9254d4340c9SChristian Brauner static inline int btrfs_may_create(struct user_namespace *mnt_userns, 9264d4340c9SChristian Brauner struct inode *dir, struct dentry *child) 927cb8e7090SChristoph Hellwig { 9282b0143b5SDavid Howells if (d_really_is_positive(child)) 929cb8e7090SChristoph Hellwig return -EEXIST; 930cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 931cb8e7090SChristoph Hellwig return -ENOENT; 9324d4340c9SChristian Brauner if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns)) 9335474bf40SChristian Brauner return -EOVERFLOW; 9344d4340c9SChristian Brauner return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); 935cb8e7090SChristoph Hellwig } 936cb8e7090SChristoph Hellwig 937cb8e7090SChristoph Hellwig /* 938cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 939cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 940cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 941cb8e7090SChristoph Hellwig */ 94292872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 9434d4340c9SChristian Brauner struct user_namespace *mnt_userns, 94452f75f4fSDavid Sterba const char *name, int namelen, 94572fd032eSSage Weil struct btrfs_root *snap_src, 9469babda9fSNikolay Borisov bool readonly, 9478696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 948cb8e7090SChristoph Hellwig { 9492b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9500b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 951cb8e7090SChristoph Hellwig struct dentry *dentry; 952cb8e7090SChristoph Hellwig int error; 953cb8e7090SChristoph Hellwig 95400235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 95500235411SAl Viro if (error == -EINTR) 95600235411SAl Viro return error; 957cb8e7090SChristoph Hellwig 9584d4340c9SChristian Brauner dentry = lookup_one(mnt_userns, name, parent->dentry, namelen); 959cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 960cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 961cb8e7090SChristoph Hellwig goto out_unlock; 962cb8e7090SChristoph Hellwig 9634d4340c9SChristian Brauner error = btrfs_may_create(mnt_userns, dir, dentry); 964cb8e7090SChristoph Hellwig if (error) 965a874a63eSLiu Bo goto out_dput; 966cb8e7090SChristoph Hellwig 9679c52057cSChris Mason /* 9689c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9699c52057cSChris Mason * check for them now when we can safely fail 9709c52057cSChris Mason */ 9719c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9729c52057cSChris Mason dir->i_ino, name, 9739c52057cSChris Mason namelen); 9749c52057cSChris Mason if (error) 9759c52057cSChris Mason goto out_dput; 9769c52057cSChris Mason 9770b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 97876dda93cSYan, Zheng 97976dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 98076dda93cSYan, Zheng goto out_up_read; 98176dda93cSYan, Zheng 9829babda9fSNikolay Borisov if (snap_src) 9839babda9fSNikolay Borisov error = create_snapshot(snap_src, dir, dentry, readonly, inherit); 9849babda9fSNikolay Borisov else 98570dc55f4SOmar Sandoval error = create_subvol(mnt_userns, dir, dentry, inherit); 9869babda9fSNikolay Borisov 98776dda93cSYan, Zheng if (!error) 98876dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 98976dda93cSYan, Zheng out_up_read: 9900b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 991cb8e7090SChristoph Hellwig out_dput: 992cb8e7090SChristoph Hellwig dput(dentry); 993cb8e7090SChristoph Hellwig out_unlock: 99464708539SJosef Bacik btrfs_inode_unlock(dir, 0); 995cb8e7090SChristoph Hellwig return error; 996cb8e7090SChristoph Hellwig } 997cb8e7090SChristoph Hellwig 998c11fbb6eSRobbie Ko static noinline int btrfs_mksnapshot(const struct path *parent, 9994d4340c9SChristian Brauner struct user_namespace *mnt_userns, 1000c11fbb6eSRobbie Ko const char *name, int namelen, 1001c11fbb6eSRobbie Ko struct btrfs_root *root, 1002c11fbb6eSRobbie Ko bool readonly, 1003c11fbb6eSRobbie Ko struct btrfs_qgroup_inherit *inherit) 1004c11fbb6eSRobbie Ko { 1005c11fbb6eSRobbie Ko int ret; 1006c11fbb6eSRobbie Ko bool snapshot_force_cow = false; 1007c11fbb6eSRobbie Ko 1008c11fbb6eSRobbie Ko /* 1009c11fbb6eSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 1010c11fbb6eSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 1011c11fbb6eSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 1012c11fbb6eSRobbie Ko */ 1013c11fbb6eSRobbie Ko btrfs_drew_read_lock(&root->snapshot_lock); 1014c11fbb6eSRobbie Ko 1015f9baa501SFilipe Manana ret = btrfs_start_delalloc_snapshot(root, false); 1016c11fbb6eSRobbie Ko if (ret) 1017c11fbb6eSRobbie Ko goto out; 1018c11fbb6eSRobbie Ko 1019c11fbb6eSRobbie Ko /* 1020c11fbb6eSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 1021c11fbb6eSRobbie Ko * we force future writes to fallback to COW mode during snapshot 1022c11fbb6eSRobbie Ko * creation. 1023c11fbb6eSRobbie Ko */ 1024c11fbb6eSRobbie Ko atomic_inc(&root->snapshot_force_cow); 1025c11fbb6eSRobbie Ko snapshot_force_cow = true; 1026c11fbb6eSRobbie Ko 1027c11fbb6eSRobbie Ko btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 1028c11fbb6eSRobbie Ko 10294d4340c9SChristian Brauner ret = btrfs_mksubvol(parent, mnt_userns, name, namelen, 1030c11fbb6eSRobbie Ko root, readonly, inherit); 1031c11fbb6eSRobbie Ko out: 1032c11fbb6eSRobbie Ko if (snapshot_force_cow) 1033c11fbb6eSRobbie Ko atomic_dec(&root->snapshot_force_cow); 1034c11fbb6eSRobbie Ko btrfs_drew_read_unlock(&root->snapshot_lock); 1035c11fbb6eSRobbie Ko return ret; 1036c11fbb6eSRobbie Ko } 1037c11fbb6eSRobbie Ko 1038d5633b0dSQu Wenruo /* 1039d5633b0dSQu Wenruo * Defrag specific helper to get an extent map. 1040d5633b0dSQu Wenruo * 1041d5633b0dSQu Wenruo * Differences between this and btrfs_get_extent() are: 1042d5633b0dSQu Wenruo * 1043d5633b0dSQu Wenruo * - No extent_map will be added to inode->extent_tree 1044d5633b0dSQu Wenruo * To reduce memory usage in the long run. 1045d5633b0dSQu Wenruo * 1046d5633b0dSQu Wenruo * - Extra optimization to skip file extents older than @newer_than 1047d5633b0dSQu Wenruo * By using btrfs_search_forward() we can skip entire file ranges that 1048d5633b0dSQu Wenruo * have extents created in past transactions, because btrfs_search_forward() 1049d5633b0dSQu Wenruo * will not visit leaves and nodes with a generation smaller than given 1050d5633b0dSQu Wenruo * minimal generation threshold (@newer_than). 1051d5633b0dSQu Wenruo * 1052d5633b0dSQu Wenruo * Return valid em if we find a file extent matching the requirement. 1053d5633b0dSQu Wenruo * Return NULL if we can not find a file extent matching the requirement. 1054d5633b0dSQu Wenruo * 1055d5633b0dSQu Wenruo * Return ERR_PTR() for error. 1056d5633b0dSQu Wenruo */ 1057d5633b0dSQu Wenruo static struct extent_map *defrag_get_extent(struct btrfs_inode *inode, 1058d5633b0dSQu Wenruo u64 start, u64 newer_than) 1059d5633b0dSQu Wenruo { 1060d5633b0dSQu Wenruo struct btrfs_root *root = inode->root; 1061d5633b0dSQu Wenruo struct btrfs_file_extent_item *fi; 1062d5633b0dSQu Wenruo struct btrfs_path path = { 0 }; 1063d5633b0dSQu Wenruo struct extent_map *em; 1064d5633b0dSQu Wenruo struct btrfs_key key; 1065d5633b0dSQu Wenruo u64 ino = btrfs_ino(inode); 1066d5633b0dSQu Wenruo int ret; 1067d5633b0dSQu Wenruo 1068d5633b0dSQu Wenruo em = alloc_extent_map(); 1069d5633b0dSQu Wenruo if (!em) { 1070d5633b0dSQu Wenruo ret = -ENOMEM; 1071d5633b0dSQu Wenruo goto err; 1072d5633b0dSQu Wenruo } 1073d5633b0dSQu Wenruo 1074d5633b0dSQu Wenruo key.objectid = ino; 1075d5633b0dSQu Wenruo key.type = BTRFS_EXTENT_DATA_KEY; 1076d5633b0dSQu Wenruo key.offset = start; 1077d5633b0dSQu Wenruo 1078d5633b0dSQu Wenruo if (newer_than) { 1079d5633b0dSQu Wenruo ret = btrfs_search_forward(root, &key, &path, newer_than); 1080d5633b0dSQu Wenruo if (ret < 0) 1081d5633b0dSQu Wenruo goto err; 1082d5633b0dSQu Wenruo /* Can't find anything newer */ 1083d5633b0dSQu Wenruo if (ret > 0) 1084d5633b0dSQu Wenruo goto not_found; 1085d5633b0dSQu Wenruo } else { 1086d5633b0dSQu Wenruo ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0); 1087d5633b0dSQu Wenruo if (ret < 0) 1088d5633b0dSQu Wenruo goto err; 1089d5633b0dSQu Wenruo } 1090d5633b0dSQu Wenruo if (path.slots[0] >= btrfs_header_nritems(path.nodes[0])) { 1091d5633b0dSQu Wenruo /* 1092d5633b0dSQu Wenruo * If btrfs_search_slot() makes path to point beyond nritems, 1093d5633b0dSQu Wenruo * we should not have an empty leaf, as this inode must at 1094d5633b0dSQu Wenruo * least have its INODE_ITEM. 1095d5633b0dSQu Wenruo */ 1096d5633b0dSQu Wenruo ASSERT(btrfs_header_nritems(path.nodes[0])); 1097d5633b0dSQu Wenruo path.slots[0] = btrfs_header_nritems(path.nodes[0]) - 1; 1098d5633b0dSQu Wenruo } 1099d5633b0dSQu Wenruo btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]); 1100d5633b0dSQu Wenruo /* Perfect match, no need to go one slot back */ 1101d5633b0dSQu Wenruo if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY && 1102d5633b0dSQu Wenruo key.offset == start) 1103d5633b0dSQu Wenruo goto iterate; 1104d5633b0dSQu Wenruo 1105d5633b0dSQu Wenruo /* We didn't find a perfect match, needs to go one slot back */ 1106d5633b0dSQu Wenruo if (path.slots[0] > 0) { 1107d5633b0dSQu Wenruo btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]); 1108d5633b0dSQu Wenruo if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY) 1109d5633b0dSQu Wenruo path.slots[0]--; 1110d5633b0dSQu Wenruo } 1111d5633b0dSQu Wenruo 1112d5633b0dSQu Wenruo iterate: 1113d5633b0dSQu Wenruo /* Iterate through the path to find a file extent covering @start */ 1114d5633b0dSQu Wenruo while (true) { 1115d5633b0dSQu Wenruo u64 extent_end; 1116d5633b0dSQu Wenruo 1117d5633b0dSQu Wenruo if (path.slots[0] >= btrfs_header_nritems(path.nodes[0])) 1118d5633b0dSQu Wenruo goto next; 1119d5633b0dSQu Wenruo 1120d5633b0dSQu Wenruo btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]); 1121d5633b0dSQu Wenruo 1122d5633b0dSQu Wenruo /* 1123d5633b0dSQu Wenruo * We may go one slot back to INODE_REF/XATTR item, then 1124d5633b0dSQu Wenruo * need to go forward until we reach an EXTENT_DATA. 1125d5633b0dSQu Wenruo * But we should still has the correct ino as key.objectid. 1126d5633b0dSQu Wenruo */ 1127d5633b0dSQu Wenruo if (WARN_ON(key.objectid < ino) || key.type < BTRFS_EXTENT_DATA_KEY) 1128d5633b0dSQu Wenruo goto next; 1129d5633b0dSQu Wenruo 1130d5633b0dSQu Wenruo /* It's beyond our target range, definitely not extent found */ 1131d5633b0dSQu Wenruo if (key.objectid > ino || key.type > BTRFS_EXTENT_DATA_KEY) 1132d5633b0dSQu Wenruo goto not_found; 1133d5633b0dSQu Wenruo 1134d5633b0dSQu Wenruo /* 1135d5633b0dSQu Wenruo * | |<- File extent ->| 1136d5633b0dSQu Wenruo * \- start 1137d5633b0dSQu Wenruo * 1138d5633b0dSQu Wenruo * This means there is a hole between start and key.offset. 1139d5633b0dSQu Wenruo */ 1140d5633b0dSQu Wenruo if (key.offset > start) { 1141d5633b0dSQu Wenruo em->start = start; 1142d5633b0dSQu Wenruo em->orig_start = start; 1143d5633b0dSQu Wenruo em->block_start = EXTENT_MAP_HOLE; 1144d5633b0dSQu Wenruo em->len = key.offset - start; 1145d5633b0dSQu Wenruo break; 1146d5633b0dSQu Wenruo } 1147d5633b0dSQu Wenruo 1148d5633b0dSQu Wenruo fi = btrfs_item_ptr(path.nodes[0], path.slots[0], 1149d5633b0dSQu Wenruo struct btrfs_file_extent_item); 1150d5633b0dSQu Wenruo extent_end = btrfs_file_extent_end(&path); 1151d5633b0dSQu Wenruo 1152d5633b0dSQu Wenruo /* 1153d5633b0dSQu Wenruo * |<- file extent ->| | 1154d5633b0dSQu Wenruo * \- start 1155d5633b0dSQu Wenruo * 1156d5633b0dSQu Wenruo * We haven't reached start, search next slot. 1157d5633b0dSQu Wenruo */ 1158d5633b0dSQu Wenruo if (extent_end <= start) 1159d5633b0dSQu Wenruo goto next; 1160d5633b0dSQu Wenruo 1161d5633b0dSQu Wenruo /* Now this extent covers @start, convert it to em */ 1162d5633b0dSQu Wenruo btrfs_extent_item_to_extent_map(inode, &path, fi, false, em); 1163d5633b0dSQu Wenruo break; 1164d5633b0dSQu Wenruo next: 1165d5633b0dSQu Wenruo ret = btrfs_next_item(root, &path); 1166d5633b0dSQu Wenruo if (ret < 0) 1167d5633b0dSQu Wenruo goto err; 1168d5633b0dSQu Wenruo if (ret > 0) 1169d5633b0dSQu Wenruo goto not_found; 1170d5633b0dSQu Wenruo } 1171d5633b0dSQu Wenruo btrfs_release_path(&path); 1172d5633b0dSQu Wenruo return em; 1173d5633b0dSQu Wenruo 1174d5633b0dSQu Wenruo not_found: 1175d5633b0dSQu Wenruo btrfs_release_path(&path); 1176d5633b0dSQu Wenruo free_extent_map(em); 1177d5633b0dSQu Wenruo return NULL; 1178d5633b0dSQu Wenruo 1179d5633b0dSQu Wenruo err: 1180d5633b0dSQu Wenruo btrfs_release_path(&path); 1181d5633b0dSQu Wenruo free_extent_map(em); 1182d5633b0dSQu Wenruo return ERR_PTR(ret); 1183d5633b0dSQu Wenruo } 1184d5633b0dSQu Wenruo 1185e9eec721SQu Wenruo static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start, 1186d5633b0dSQu Wenruo u64 newer_than, bool locked) 118717ce6ef8SLiu Bo { 118817ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1189940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 11906c282eb4SLi Zefan struct extent_map *em; 119176068caeSQu Wenruo const u32 sectorsize = BTRFS_I(inode)->root->fs_info->sectorsize; 1192940100a4SChris Mason 1193940100a4SChris Mason /* 1194940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1195940100a4SChris Mason * the full extent lock 1196940100a4SChris Mason */ 1197940100a4SChris Mason read_lock(&em_tree->lock); 119876068caeSQu Wenruo em = lookup_extent_mapping(em_tree, start, sectorsize); 1199940100a4SChris Mason read_unlock(&em_tree->lock); 1200940100a4SChris Mason 1201199257a7SQu Wenruo /* 1202199257a7SQu Wenruo * We can get a merged extent, in that case, we need to re-search 1203199257a7SQu Wenruo * tree to get the original em for defrag. 1204199257a7SQu Wenruo * 1205199257a7SQu Wenruo * If @newer_than is 0 or em::generation < newer_than, we can trust 1206199257a7SQu Wenruo * this em, as either we don't care about the generation, or the 1207199257a7SQu Wenruo * merged extent map will be rejected anyway. 1208199257a7SQu Wenruo */ 1209199257a7SQu Wenruo if (em && test_bit(EXTENT_FLAG_MERGED, &em->flags) && 1210199257a7SQu Wenruo newer_than && em->generation >= newer_than) { 1211199257a7SQu Wenruo free_extent_map(em); 1212199257a7SQu Wenruo em = NULL; 1213199257a7SQu Wenruo } 1214199257a7SQu Wenruo 1215940100a4SChris Mason if (!em) { 1216308d9800SFilipe Manana struct extent_state *cached = NULL; 121776068caeSQu Wenruo u64 end = start + sectorsize - 1; 1218308d9800SFilipe Manana 1219940100a4SChris Mason /* get the big lock and read metadata off disk */ 1220e9eec721SQu Wenruo if (!locked) 1221570eb97bSJosef Bacik lock_extent(io_tree, start, end, &cached); 1222d5633b0dSQu Wenruo em = defrag_get_extent(BTRFS_I(inode), start, newer_than); 1223e9eec721SQu Wenruo if (!locked) 1224570eb97bSJosef Bacik unlock_extent(io_tree, start, end, &cached); 1225940100a4SChris Mason 12266cf8bfbfSDan Carpenter if (IS_ERR(em)) 12276c282eb4SLi Zefan return NULL; 1228940100a4SChris Mason } 1229940100a4SChris Mason 12306c282eb4SLi Zefan return em; 12316c282eb4SLi Zefan } 12326c282eb4SLi Zefan 1233d7601566SNaohiro Aota static u32 get_extent_max_capacity(const struct btrfs_fs_info *fs_info, 1234d7601566SNaohiro Aota const struct extent_map *em) 1235979b25c3SQu Wenruo { 1236979b25c3SQu Wenruo if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) 1237979b25c3SQu Wenruo return BTRFS_MAX_COMPRESSED; 1238d7601566SNaohiro Aota return fs_info->max_extent_size; 1239979b25c3SQu Wenruo } 1240979b25c3SQu Wenruo 1241e9eec721SQu Wenruo static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em, 124275a36a7dSQu Wenruo u32 extent_thresh, u64 newer_than, bool locked) 12436c282eb4SLi Zefan { 1244d7601566SNaohiro Aota struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 12456c282eb4SLi Zefan struct extent_map *next; 12467093f152SQu Wenruo bool ret = false; 12476c282eb4SLi Zefan 12486c282eb4SLi Zefan /* this is the last extent */ 12496c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 12506c282eb4SLi Zefan return false; 12516c282eb4SLi Zefan 1252d5633b0dSQu Wenruo /* 125375a36a7dSQu Wenruo * Here we need to pass @newer_then when checking the next extent, or 125475a36a7dSQu Wenruo * we will hit a case we mark current extent for defrag, but the next 125575a36a7dSQu Wenruo * one will not be a target. 125675a36a7dSQu Wenruo * This will just cause extra IO without really reducing the fragments. 1257d5633b0dSQu Wenruo */ 125875a36a7dSQu Wenruo next = defrag_lookup_extent(inode, em->start + em->len, newer_than, locked); 12597093f152SQu Wenruo /* No more em or hole */ 1260e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 12617093f152SQu Wenruo goto out; 12627093f152SQu Wenruo if (test_bit(EXTENT_FLAG_PREALLOC, &next->flags)) 12637093f152SQu Wenruo goto out; 1264979b25c3SQu Wenruo /* 1265979b25c3SQu Wenruo * If the next extent is at its max capacity, defragging current extent 1266979b25c3SQu Wenruo * makes no sense, as the total number of extents won't change. 1267979b25c3SQu Wenruo */ 1268d7601566SNaohiro Aota if (next->len >= get_extent_max_capacity(fs_info, em)) 1269979b25c3SQu Wenruo goto out; 127075a36a7dSQu Wenruo /* Skip older extent */ 127175a36a7dSQu Wenruo if (next->generation < newer_than) 127275a36a7dSQu Wenruo goto out; 127375a36a7dSQu Wenruo /* Also check extent size */ 127475a36a7dSQu Wenruo if (next->len >= extent_thresh) 127575a36a7dSQu Wenruo goto out; 127675a36a7dSQu Wenruo 12777093f152SQu Wenruo ret = true; 12787093f152SQu Wenruo out: 12796c282eb4SLi Zefan free_extent_map(next); 12806c282eb4SLi Zefan return ret; 12816c282eb4SLi Zefan } 12826c282eb4SLi Zefan 12834cb5300bSChris Mason /* 12845767b50cSQu Wenruo * Prepare one page to be defragged. 12855767b50cSQu Wenruo * 12865767b50cSQu Wenruo * This will ensure: 12875767b50cSQu Wenruo * 12885767b50cSQu Wenruo * - Returned page is locked and has been set up properly. 12895767b50cSQu Wenruo * - No ordered extent exists in the page. 12905767b50cSQu Wenruo * - The page is uptodate. 12915767b50cSQu Wenruo * 12925767b50cSQu Wenruo * NOTE: Caller should also wait for page writeback after the cluster is 12935767b50cSQu Wenruo * prepared, here we don't do writeback wait for each page. 12945767b50cSQu Wenruo */ 12955767b50cSQu Wenruo static struct page *defrag_prepare_one_page(struct btrfs_inode *inode, 12965767b50cSQu Wenruo pgoff_t index) 1297940100a4SChris Mason { 12985767b50cSQu Wenruo struct address_space *mapping = inode->vfs_inode.i_mapping; 12995767b50cSQu Wenruo gfp_t mask = btrfs_alloc_write_mask(mapping); 13005767b50cSQu Wenruo u64 page_start = (u64)index << PAGE_SHIFT; 13015767b50cSQu Wenruo u64 page_end = page_start + PAGE_SIZE - 1; 13024cb5300bSChris Mason struct extent_state *cached_state = NULL; 13034cb5300bSChris Mason struct page *page; 13045767b50cSQu Wenruo int ret; 13055767b50cSQu Wenruo 1306600a45e1SMiao Xie again: 13075767b50cSQu Wenruo page = find_or_create_page(mapping, index, mask); 13084cb5300bSChris Mason if (!page) 13095767b50cSQu Wenruo return ERR_PTR(-ENOMEM); 13105767b50cSQu Wenruo 131124bcb454SOmar Sandoval /* 131224bcb454SOmar Sandoval * Since we can defragment files opened read-only, we can encounter 131324bcb454SOmar Sandoval * transparent huge pages here (see CONFIG_READ_ONLY_THP_FOR_FS). We 131424bcb454SOmar Sandoval * can't do I/O using huge pages yet, so return an error for now. 131524bcb454SOmar Sandoval * Filesystem transparent huge pages are typically only used for 131624bcb454SOmar Sandoval * executables that explicitly enable them, so this isn't very 131724bcb454SOmar Sandoval * restrictive. 131824bcb454SOmar Sandoval */ 131924bcb454SOmar Sandoval if (PageCompound(page)) { 132024bcb454SOmar Sandoval unlock_page(page); 132124bcb454SOmar Sandoval put_page(page); 132224bcb454SOmar Sandoval return ERR_PTR(-ETXTBSY); 132324bcb454SOmar Sandoval } 13244cb5300bSChris Mason 132532443de3SQu Wenruo ret = set_page_extent_mapped(page); 132632443de3SQu Wenruo if (ret < 0) { 132732443de3SQu Wenruo unlock_page(page); 132832443de3SQu Wenruo put_page(page); 13295767b50cSQu Wenruo return ERR_PTR(ret); 133032443de3SQu Wenruo } 133132443de3SQu Wenruo 13325767b50cSQu Wenruo /* Wait for any existing ordered extent in the range */ 1333600a45e1SMiao Xie while (1) { 13345767b50cSQu Wenruo struct btrfs_ordered_extent *ordered; 13355767b50cSQu Wenruo 1336570eb97bSJosef Bacik lock_extent(&inode->io_tree, page_start, page_end, &cached_state); 13375767b50cSQu Wenruo ordered = btrfs_lookup_ordered_range(inode, page_start, PAGE_SIZE); 1338570eb97bSJosef Bacik unlock_extent(&inode->io_tree, page_start, page_end, 1339e43bbe5eSDavid Sterba &cached_state); 1340600a45e1SMiao Xie if (!ordered) 1341600a45e1SMiao Xie break; 1342600a45e1SMiao Xie 1343600a45e1SMiao Xie unlock_page(page); 1344c0a43603SNikolay Borisov btrfs_start_ordered_extent(ordered, 1); 1345600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1346600a45e1SMiao Xie lock_page(page); 13471f12bd06SLiu Bo /* 13485767b50cSQu Wenruo * We unlocked the page above, so we need check if it was 13495767b50cSQu Wenruo * released or not. 13501f12bd06SLiu Bo */ 13515767b50cSQu Wenruo if (page->mapping != mapping || !PagePrivate(page)) { 13521f12bd06SLiu Bo unlock_page(page); 135309cbfeafSKirill A. Shutemov put_page(page); 13541f12bd06SLiu Bo goto again; 13551f12bd06SLiu Bo } 1356600a45e1SMiao Xie } 1357600a45e1SMiao Xie 13585767b50cSQu Wenruo /* 13595767b50cSQu Wenruo * Now the page range has no ordered extent any more. Read the page to 13605767b50cSQu Wenruo * make it uptodate. 13615767b50cSQu Wenruo */ 13624cb5300bSChris Mason if (!PageUptodate(page)) { 1363fb12489bSMatthew Wilcox (Oracle) btrfs_read_folio(NULL, page_folio(page)); 13644cb5300bSChris Mason lock_page(page); 13655767b50cSQu Wenruo if (page->mapping != mapping || !PagePrivate(page)) { 1366600a45e1SMiao Xie unlock_page(page); 136709cbfeafSKirill A. Shutemov put_page(page); 1368600a45e1SMiao Xie goto again; 1369600a45e1SMiao Xie } 13705767b50cSQu Wenruo if (!PageUptodate(page)) { 13715767b50cSQu Wenruo unlock_page(page); 13725767b50cSQu Wenruo put_page(page); 13735767b50cSQu Wenruo return ERR_PTR(-EIO); 13744cb5300bSChris Mason } 13755767b50cSQu Wenruo } 13765767b50cSQu Wenruo return page; 13775767b50cSQu Wenruo } 13784cb5300bSChris Mason 1379eb793cf8SQu Wenruo struct defrag_target_range { 1380eb793cf8SQu Wenruo struct list_head list; 1381eb793cf8SQu Wenruo u64 start; 1382eb793cf8SQu Wenruo u64 len; 1383eb793cf8SQu Wenruo }; 13844cb5300bSChris Mason 13854cb5300bSChris Mason /* 1386eb793cf8SQu Wenruo * Collect all valid target extents. 1387eb793cf8SQu Wenruo * 1388eb793cf8SQu Wenruo * @start: file offset to lookup 1389eb793cf8SQu Wenruo * @len: length to lookup 1390eb793cf8SQu Wenruo * @extent_thresh: file extent size threshold, any extent size >= this value 1391eb793cf8SQu Wenruo * will be ignored 1392eb793cf8SQu Wenruo * @newer_than: only defrag extents newer than this value 1393eb793cf8SQu Wenruo * @do_compress: whether the defrag is doing compression 1394eb793cf8SQu Wenruo * if true, @extent_thresh will be ignored and all regular 1395eb793cf8SQu Wenruo * file extents meeting @newer_than will be targets. 1396e9eec721SQu Wenruo * @locked: if the range has already held extent lock 1397eb793cf8SQu Wenruo * @target_list: list of targets file extents 13984cb5300bSChris Mason */ 1399eb793cf8SQu Wenruo static int defrag_collect_targets(struct btrfs_inode *inode, 1400eb793cf8SQu Wenruo u64 start, u64 len, u32 extent_thresh, 1401eb793cf8SQu Wenruo u64 newer_than, bool do_compress, 1402966d879bSQu Wenruo bool locked, struct list_head *target_list, 1403966d879bSQu Wenruo u64 *last_scanned_ret) 1404eb793cf8SQu Wenruo { 1405d7601566SNaohiro Aota struct btrfs_fs_info *fs_info = inode->root->fs_info; 1406966d879bSQu Wenruo bool last_is_target = false; 1407eb793cf8SQu Wenruo u64 cur = start; 1408eb793cf8SQu Wenruo int ret = 0; 1409eb793cf8SQu Wenruo 1410eb793cf8SQu Wenruo while (cur < start + len) { 1411eb793cf8SQu Wenruo struct extent_map *em; 1412eb793cf8SQu Wenruo struct defrag_target_range *new; 1413eb793cf8SQu Wenruo bool next_mergeable = true; 1414eb793cf8SQu Wenruo u64 range_len; 1415eb793cf8SQu Wenruo 1416966d879bSQu Wenruo last_is_target = false; 1417d5633b0dSQu Wenruo em = defrag_lookup_extent(&inode->vfs_inode, cur, 1418d5633b0dSQu Wenruo newer_than, locked); 1419eb793cf8SQu Wenruo if (!em) 1420eb793cf8SQu Wenruo break; 1421eb793cf8SQu Wenruo 1422d8101a0cSQu Wenruo /* 1423d8101a0cSQu Wenruo * If the file extent is an inlined one, we may still want to 1424d8101a0cSQu Wenruo * defrag it (fallthrough) if it will cause a regular extent. 1425d8101a0cSQu Wenruo * This is for users who want to convert inline extents to 1426d8101a0cSQu Wenruo * regular ones through max_inline= mount option. 1427d8101a0cSQu Wenruo */ 1428d8101a0cSQu Wenruo if (em->block_start == EXTENT_MAP_INLINE && 1429d8101a0cSQu Wenruo em->len <= inode->root->fs_info->max_inline) 1430d8101a0cSQu Wenruo goto next; 1431d8101a0cSQu Wenruo 1432d8101a0cSQu Wenruo /* Skip hole/delalloc/preallocated extents */ 1433d8101a0cSQu Wenruo if (em->block_start == EXTENT_MAP_HOLE || 1434d8101a0cSQu Wenruo em->block_start == EXTENT_MAP_DELALLOC || 1435eb793cf8SQu Wenruo test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) 1436eb793cf8SQu Wenruo goto next; 1437eb793cf8SQu Wenruo 1438eb793cf8SQu Wenruo /* Skip older extent */ 1439eb793cf8SQu Wenruo if (em->generation < newer_than) 1440eb793cf8SQu Wenruo goto next; 1441eb793cf8SQu Wenruo 14420d1ffa22SQu Wenruo /* This em is under writeback, no need to defrag */ 14430d1ffa22SQu Wenruo if (em->generation == (u64)-1) 14440d1ffa22SQu Wenruo goto next; 14450d1ffa22SQu Wenruo 1446eb793cf8SQu Wenruo /* 14470cb5950fSFilipe Manana * Our start offset might be in the middle of an existing extent 14480cb5950fSFilipe Manana * map, so take that into account. 14490cb5950fSFilipe Manana */ 14500cb5950fSFilipe Manana range_len = em->len - (cur - em->start); 14510cb5950fSFilipe Manana /* 14520cb5950fSFilipe Manana * If this range of the extent map is already flagged for delalloc, 14530cb5950fSFilipe Manana * skip it, because: 14540cb5950fSFilipe Manana * 14550cb5950fSFilipe Manana * 1) We could deadlock later, when trying to reserve space for 14560cb5950fSFilipe Manana * delalloc, because in case we can't immediately reserve space 14570cb5950fSFilipe Manana * the flusher can start delalloc and wait for the respective 14580cb5950fSFilipe Manana * ordered extents to complete. The deadlock would happen 14590cb5950fSFilipe Manana * because we do the space reservation while holding the range 14600cb5950fSFilipe Manana * locked, and starting writeback, or finishing an ordered 14610cb5950fSFilipe Manana * extent, requires locking the range; 14620cb5950fSFilipe Manana * 14630cb5950fSFilipe Manana * 2) If there's delalloc there, it means there's dirty pages for 14640cb5950fSFilipe Manana * which writeback has not started yet (we clean the delalloc 14650cb5950fSFilipe Manana * flag when starting writeback and after creating an ordered 14660cb5950fSFilipe Manana * extent). If we mark pages in an adjacent range for defrag, 14670cb5950fSFilipe Manana * then we will have a larger contiguous range for delalloc, 14680cb5950fSFilipe Manana * very likely resulting in a larger extent after writeback is 14690cb5950fSFilipe Manana * triggered (except in a case of free space fragmentation). 14700cb5950fSFilipe Manana */ 14710cb5950fSFilipe Manana if (test_range_bit(&inode->io_tree, cur, cur + range_len - 1, 14720cb5950fSFilipe Manana EXTENT_DELALLOC, 0, NULL)) 14730cb5950fSFilipe Manana goto next; 14740cb5950fSFilipe Manana 14750cb5950fSFilipe Manana /* 1476eb793cf8SQu Wenruo * For do_compress case, we want to compress all valid file 1477eb793cf8SQu Wenruo * extents, thus no @extent_thresh or mergeable check. 1478eb793cf8SQu Wenruo */ 1479eb793cf8SQu Wenruo if (do_compress) 1480eb793cf8SQu Wenruo goto add; 1481eb793cf8SQu Wenruo 1482eb793cf8SQu Wenruo /* Skip too large extent */ 14830cb5950fSFilipe Manana if (range_len >= extent_thresh) 1484eb793cf8SQu Wenruo goto next; 1485eb793cf8SQu Wenruo 1486979b25c3SQu Wenruo /* 1487979b25c3SQu Wenruo * Skip extents already at its max capacity, this is mostly for 1488979b25c3SQu Wenruo * compressed extents, which max cap is only 128K. 1489979b25c3SQu Wenruo */ 1490d7601566SNaohiro Aota if (em->len >= get_extent_max_capacity(fs_info, em)) 1491979b25c3SQu Wenruo goto next; 1492979b25c3SQu Wenruo 1493d8101a0cSQu Wenruo /* 1494d8101a0cSQu Wenruo * Normally there are no more extents after an inline one, thus 1495d8101a0cSQu Wenruo * @next_mergeable will normally be false and not defragged. 1496d8101a0cSQu Wenruo * So if an inline extent passed all above checks, just add it 1497d8101a0cSQu Wenruo * for defrag, and be converted to regular extents. 1498d8101a0cSQu Wenruo */ 1499d8101a0cSQu Wenruo if (em->block_start == EXTENT_MAP_INLINE) 1500d8101a0cSQu Wenruo goto add; 1501d8101a0cSQu Wenruo 1502e9eec721SQu Wenruo next_mergeable = defrag_check_next_extent(&inode->vfs_inode, em, 150375a36a7dSQu Wenruo extent_thresh, newer_than, locked); 1504eb793cf8SQu Wenruo if (!next_mergeable) { 1505eb793cf8SQu Wenruo struct defrag_target_range *last; 1506eb793cf8SQu Wenruo 1507eb793cf8SQu Wenruo /* Empty target list, no way to merge with last entry */ 1508eb793cf8SQu Wenruo if (list_empty(target_list)) 1509eb793cf8SQu Wenruo goto next; 1510eb793cf8SQu Wenruo last = list_entry(target_list->prev, 1511eb793cf8SQu Wenruo struct defrag_target_range, list); 1512eb793cf8SQu Wenruo /* Not mergeable with last entry */ 1513eb793cf8SQu Wenruo if (last->start + last->len != cur) 1514eb793cf8SQu Wenruo goto next; 1515eb793cf8SQu Wenruo 1516eb793cf8SQu Wenruo /* Mergeable, fall through to add it to @target_list. */ 1517eb793cf8SQu Wenruo } 1518eb793cf8SQu Wenruo 1519eb793cf8SQu Wenruo add: 1520966d879bSQu Wenruo last_is_target = true; 1521eb793cf8SQu Wenruo range_len = min(extent_map_end(em), start + len) - cur; 1522eb793cf8SQu Wenruo /* 1523eb793cf8SQu Wenruo * This one is a good target, check if it can be merged into 1524eb793cf8SQu Wenruo * last range of the target list. 1525eb793cf8SQu Wenruo */ 1526eb793cf8SQu Wenruo if (!list_empty(target_list)) { 1527eb793cf8SQu Wenruo struct defrag_target_range *last; 1528eb793cf8SQu Wenruo 1529eb793cf8SQu Wenruo last = list_entry(target_list->prev, 1530eb793cf8SQu Wenruo struct defrag_target_range, list); 1531eb793cf8SQu Wenruo ASSERT(last->start + last->len <= cur); 1532eb793cf8SQu Wenruo if (last->start + last->len == cur) { 1533eb793cf8SQu Wenruo /* Mergeable, enlarge the last entry */ 1534eb793cf8SQu Wenruo last->len += range_len; 1535eb793cf8SQu Wenruo goto next; 1536eb793cf8SQu Wenruo } 1537eb793cf8SQu Wenruo /* Fall through to allocate a new entry */ 1538eb793cf8SQu Wenruo } 1539eb793cf8SQu Wenruo 1540eb793cf8SQu Wenruo /* Allocate new defrag_target_range */ 1541eb793cf8SQu Wenruo new = kmalloc(sizeof(*new), GFP_NOFS); 1542eb793cf8SQu Wenruo if (!new) { 1543eb793cf8SQu Wenruo free_extent_map(em); 1544eb793cf8SQu Wenruo ret = -ENOMEM; 1545eb793cf8SQu Wenruo break; 1546eb793cf8SQu Wenruo } 1547eb793cf8SQu Wenruo new->start = cur; 1548eb793cf8SQu Wenruo new->len = range_len; 1549eb793cf8SQu Wenruo list_add_tail(&new->list, target_list); 1550eb793cf8SQu Wenruo 1551eb793cf8SQu Wenruo next: 1552eb793cf8SQu Wenruo cur = extent_map_end(em); 1553eb793cf8SQu Wenruo free_extent_map(em); 1554eb793cf8SQu Wenruo } 1555eb793cf8SQu Wenruo if (ret < 0) { 1556eb793cf8SQu Wenruo struct defrag_target_range *entry; 1557eb793cf8SQu Wenruo struct defrag_target_range *tmp; 1558eb793cf8SQu Wenruo 1559eb793cf8SQu Wenruo list_for_each_entry_safe(entry, tmp, target_list, list) { 1560eb793cf8SQu Wenruo list_del_init(&entry->list); 1561eb793cf8SQu Wenruo kfree(entry); 1562eb793cf8SQu Wenruo } 1563eb793cf8SQu Wenruo } 1564966d879bSQu Wenruo if (!ret && last_scanned_ret) { 1565966d879bSQu Wenruo /* 1566966d879bSQu Wenruo * If the last extent is not a target, the caller can skip to 1567966d879bSQu Wenruo * the end of that extent. 1568966d879bSQu Wenruo * Otherwise, we can only go the end of the specified range. 1569966d879bSQu Wenruo */ 1570966d879bSQu Wenruo if (!last_is_target) 1571966d879bSQu Wenruo *last_scanned_ret = max(cur, *last_scanned_ret); 1572966d879bSQu Wenruo else 1573966d879bSQu Wenruo *last_scanned_ret = max(start + len, *last_scanned_ret); 1574966d879bSQu Wenruo } 1575eb793cf8SQu Wenruo return ret; 1576eb793cf8SQu Wenruo } 1577eb793cf8SQu Wenruo 157822b398eeSQu Wenruo #define CLUSTER_SIZE (SZ_256K) 1579a55e65b8SDavid Sterba static_assert(IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE)); 158022b398eeSQu Wenruo 158122b398eeSQu Wenruo /* 158222b398eeSQu Wenruo * Defrag one contiguous target range. 158322b398eeSQu Wenruo * 158422b398eeSQu Wenruo * @inode: target inode 158522b398eeSQu Wenruo * @target: target range to defrag 158622b398eeSQu Wenruo * @pages: locked pages covering the defrag range 158722b398eeSQu Wenruo * @nr_pages: number of locked pages 158822b398eeSQu Wenruo * 158922b398eeSQu Wenruo * Caller should ensure: 159022b398eeSQu Wenruo * 159122b398eeSQu Wenruo * - Pages are prepared 159222b398eeSQu Wenruo * Pages should be locked, no ordered extent in the pages range, 159322b398eeSQu Wenruo * no writeback. 159422b398eeSQu Wenruo * 159522b398eeSQu Wenruo * - Extent bits are locked 159622b398eeSQu Wenruo */ 159722b398eeSQu Wenruo static int defrag_one_locked_target(struct btrfs_inode *inode, 159822b398eeSQu Wenruo struct defrag_target_range *target, 159922b398eeSQu Wenruo struct page **pages, int nr_pages, 160022b398eeSQu Wenruo struct extent_state **cached_state) 160122b398eeSQu Wenruo { 160222b398eeSQu Wenruo struct btrfs_fs_info *fs_info = inode->root->fs_info; 160322b398eeSQu Wenruo struct extent_changeset *data_reserved = NULL; 160422b398eeSQu Wenruo const u64 start = target->start; 160522b398eeSQu Wenruo const u64 len = target->len; 160622b398eeSQu Wenruo unsigned long last_index = (start + len - 1) >> PAGE_SHIFT; 160722b398eeSQu Wenruo unsigned long start_index = start >> PAGE_SHIFT; 160822b398eeSQu Wenruo unsigned long first_index = page_index(pages[0]); 160922b398eeSQu Wenruo int ret = 0; 161022b398eeSQu Wenruo int i; 161122b398eeSQu Wenruo 161222b398eeSQu Wenruo ASSERT(last_index - first_index + 1 <= nr_pages); 161322b398eeSQu Wenruo 161422b398eeSQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, start, len); 161522b398eeSQu Wenruo if (ret < 0) 161622b398eeSQu Wenruo return ret; 161722b398eeSQu Wenruo clear_extent_bit(&inode->io_tree, start, start + len - 1, 161822b398eeSQu Wenruo EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | 1619bd015294SJosef Bacik EXTENT_DEFRAG, cached_state); 162022b398eeSQu Wenruo set_extent_defrag(&inode->io_tree, start, start + len - 1, cached_state); 162122b398eeSQu Wenruo 162222b398eeSQu Wenruo /* Update the page status */ 162322b398eeSQu Wenruo for (i = start_index - first_index; i <= last_index - first_index; i++) { 162422b398eeSQu Wenruo ClearPageChecked(pages[i]); 162522b398eeSQu Wenruo btrfs_page_clamp_set_dirty(fs_info, pages[i], start, len); 162622b398eeSQu Wenruo } 162722b398eeSQu Wenruo btrfs_delalloc_release_extents(inode, len); 162822b398eeSQu Wenruo extent_changeset_free(data_reserved); 162922b398eeSQu Wenruo 163022b398eeSQu Wenruo return ret; 163122b398eeSQu Wenruo } 163222b398eeSQu Wenruo 1633e9eec721SQu Wenruo static int defrag_one_range(struct btrfs_inode *inode, u64 start, u32 len, 1634966d879bSQu Wenruo u32 extent_thresh, u64 newer_than, bool do_compress, 1635966d879bSQu Wenruo u64 *last_scanned_ret) 1636e9eec721SQu Wenruo { 1637e9eec721SQu Wenruo struct extent_state *cached_state = NULL; 1638e9eec721SQu Wenruo struct defrag_target_range *entry; 1639e9eec721SQu Wenruo struct defrag_target_range *tmp; 1640e9eec721SQu Wenruo LIST_HEAD(target_list); 1641e9eec721SQu Wenruo struct page **pages; 1642e9eec721SQu Wenruo const u32 sectorsize = inode->root->fs_info->sectorsize; 1643e9eec721SQu Wenruo u64 last_index = (start + len - 1) >> PAGE_SHIFT; 1644e9eec721SQu Wenruo u64 start_index = start >> PAGE_SHIFT; 1645e9eec721SQu Wenruo unsigned int nr_pages = last_index - start_index + 1; 1646e9eec721SQu Wenruo int ret = 0; 1647e9eec721SQu Wenruo int i; 1648e9eec721SQu Wenruo 1649e9eec721SQu Wenruo ASSERT(nr_pages <= CLUSTER_SIZE / PAGE_SIZE); 1650e9eec721SQu Wenruo ASSERT(IS_ALIGNED(start, sectorsize) && IS_ALIGNED(len, sectorsize)); 1651e9eec721SQu Wenruo 1652e9eec721SQu Wenruo pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS); 1653e9eec721SQu Wenruo if (!pages) 1654e9eec721SQu Wenruo return -ENOMEM; 1655e9eec721SQu Wenruo 1656e9eec721SQu Wenruo /* Prepare all pages */ 1657e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) { 1658e9eec721SQu Wenruo pages[i] = defrag_prepare_one_page(inode, start_index + i); 1659e9eec721SQu Wenruo if (IS_ERR(pages[i])) { 1660e9eec721SQu Wenruo ret = PTR_ERR(pages[i]); 1661e9eec721SQu Wenruo pages[i] = NULL; 1662e9eec721SQu Wenruo goto free_pages; 1663e9eec721SQu Wenruo } 1664e9eec721SQu Wenruo } 1665e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) 16664cb5300bSChris Mason wait_on_page_writeback(pages[i]); 16674cb5300bSChris Mason 1668e9eec721SQu Wenruo /* Lock the pages range */ 1669570eb97bSJosef Bacik lock_extent(&inode->io_tree, start_index << PAGE_SHIFT, 1670e9eec721SQu Wenruo (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, 1671e9eec721SQu Wenruo &cached_state); 1672e9eec721SQu Wenruo /* 1673e9eec721SQu Wenruo * Now we have a consistent view about the extent map, re-check 1674e9eec721SQu Wenruo * which range really needs to be defragged. 1675e9eec721SQu Wenruo * 1676e9eec721SQu Wenruo * And this time we have extent locked already, pass @locked = true 1677e9eec721SQu Wenruo * so that we won't relock the extent range and cause deadlock. 1678e9eec721SQu Wenruo */ 1679e9eec721SQu Wenruo ret = defrag_collect_targets(inode, start, len, extent_thresh, 1680e9eec721SQu Wenruo newer_than, do_compress, true, 1681966d879bSQu Wenruo &target_list, last_scanned_ret); 1682e9eec721SQu Wenruo if (ret < 0) 1683e9eec721SQu Wenruo goto unlock_extent; 16844cb5300bSChris Mason 1685e9eec721SQu Wenruo list_for_each_entry(entry, &target_list, list) { 1686e9eec721SQu Wenruo ret = defrag_one_locked_target(inode, entry, pages, nr_pages, 1687e9eec721SQu Wenruo &cached_state); 1688e9eec721SQu Wenruo if (ret < 0) 1689e9eec721SQu Wenruo break; 1690e9eec721SQu Wenruo } 1691e9eec721SQu Wenruo 1692e9eec721SQu Wenruo list_for_each_entry_safe(entry, tmp, &target_list, list) { 1693e9eec721SQu Wenruo list_del_init(&entry->list); 1694e9eec721SQu Wenruo kfree(entry); 1695e9eec721SQu Wenruo } 1696e9eec721SQu Wenruo unlock_extent: 1697570eb97bSJosef Bacik unlock_extent(&inode->io_tree, start_index << PAGE_SHIFT, 1698e9eec721SQu Wenruo (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, 1699e9eec721SQu Wenruo &cached_state); 1700e9eec721SQu Wenruo free_pages: 1701e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) { 1702e9eec721SQu Wenruo if (pages[i]) { 1703e9eec721SQu Wenruo unlock_page(pages[i]); 1704e9eec721SQu Wenruo put_page(pages[i]); 1705e9eec721SQu Wenruo } 1706e9eec721SQu Wenruo } 1707e9eec721SQu Wenruo kfree(pages); 1708e9eec721SQu Wenruo return ret; 1709e9eec721SQu Wenruo } 1710e9eec721SQu Wenruo 1711b18c3ab2SQu Wenruo static int defrag_one_cluster(struct btrfs_inode *inode, 1712b18c3ab2SQu Wenruo struct file_ra_state *ra, 1713b18c3ab2SQu Wenruo u64 start, u32 len, u32 extent_thresh, 1714b18c3ab2SQu Wenruo u64 newer_than, bool do_compress, 1715b18c3ab2SQu Wenruo unsigned long *sectors_defragged, 1716966d879bSQu Wenruo unsigned long max_sectors, 1717966d879bSQu Wenruo u64 *last_scanned_ret) 1718b18c3ab2SQu Wenruo { 1719b18c3ab2SQu Wenruo const u32 sectorsize = inode->root->fs_info->sectorsize; 1720b18c3ab2SQu Wenruo struct defrag_target_range *entry; 1721b18c3ab2SQu Wenruo struct defrag_target_range *tmp; 1722b18c3ab2SQu Wenruo LIST_HEAD(target_list); 1723b18c3ab2SQu Wenruo int ret; 1724b18c3ab2SQu Wenruo 1725b18c3ab2SQu Wenruo ret = defrag_collect_targets(inode, start, len, extent_thresh, 1726b18c3ab2SQu Wenruo newer_than, do_compress, false, 1727966d879bSQu Wenruo &target_list, NULL); 1728b18c3ab2SQu Wenruo if (ret < 0) 1729b18c3ab2SQu Wenruo goto out; 1730b18c3ab2SQu Wenruo 1731b18c3ab2SQu Wenruo list_for_each_entry(entry, &target_list, list) { 1732b18c3ab2SQu Wenruo u32 range_len = entry->len; 1733b18c3ab2SQu Wenruo 1734484167daSQu Wenruo /* Reached or beyond the limit */ 1735c080b414SQu Wenruo if (max_sectors && *sectors_defragged >= max_sectors) { 1736c080b414SQu Wenruo ret = 1; 1737b18c3ab2SQu Wenruo break; 1738c080b414SQu Wenruo } 1739b18c3ab2SQu Wenruo 1740b18c3ab2SQu Wenruo if (max_sectors) 1741b18c3ab2SQu Wenruo range_len = min_t(u32, range_len, 1742b18c3ab2SQu Wenruo (max_sectors - *sectors_defragged) * sectorsize); 1743b18c3ab2SQu Wenruo 1744966d879bSQu Wenruo /* 1745966d879bSQu Wenruo * If defrag_one_range() has updated last_scanned_ret, 1746966d879bSQu Wenruo * our range may already be invalid (e.g. hole punched). 1747966d879bSQu Wenruo * Skip if our range is before last_scanned_ret, as there is 1748966d879bSQu Wenruo * no need to defrag the range anymore. 1749966d879bSQu Wenruo */ 1750966d879bSQu Wenruo if (entry->start + range_len <= *last_scanned_ret) 1751966d879bSQu Wenruo continue; 1752966d879bSQu Wenruo 1753b18c3ab2SQu Wenruo if (ra) 1754b18c3ab2SQu Wenruo page_cache_sync_readahead(inode->vfs_inode.i_mapping, 1755b18c3ab2SQu Wenruo ra, NULL, entry->start >> PAGE_SHIFT, 1756b18c3ab2SQu Wenruo ((entry->start + range_len - 1) >> PAGE_SHIFT) - 1757b18c3ab2SQu Wenruo (entry->start >> PAGE_SHIFT) + 1); 1758b18c3ab2SQu Wenruo /* 1759b18c3ab2SQu Wenruo * Here we may not defrag any range if holes are punched before 1760b18c3ab2SQu Wenruo * we locked the pages. 1761b18c3ab2SQu Wenruo * But that's fine, it only affects the @sectors_defragged 1762b18c3ab2SQu Wenruo * accounting. 1763b18c3ab2SQu Wenruo */ 1764b18c3ab2SQu Wenruo ret = defrag_one_range(inode, entry->start, range_len, 1765966d879bSQu Wenruo extent_thresh, newer_than, do_compress, 1766966d879bSQu Wenruo last_scanned_ret); 1767b18c3ab2SQu Wenruo if (ret < 0) 1768b18c3ab2SQu Wenruo break; 1769484167daSQu Wenruo *sectors_defragged += range_len >> 1770484167daSQu Wenruo inode->root->fs_info->sectorsize_bits; 1771b18c3ab2SQu Wenruo } 1772b18c3ab2SQu Wenruo out: 1773b18c3ab2SQu Wenruo list_for_each_entry_safe(entry, tmp, &target_list, list) { 1774b18c3ab2SQu Wenruo list_del_init(&entry->list); 1775b18c3ab2SQu Wenruo kfree(entry); 1776b18c3ab2SQu Wenruo } 1777966d879bSQu Wenruo if (ret >= 0) 1778966d879bSQu Wenruo *last_scanned_ret = max(*last_scanned_ret, start + len); 1779b18c3ab2SQu Wenruo return ret; 1780b18c3ab2SQu Wenruo } 17817f458a38SFilipe Manana 17827f458a38SFilipe Manana /* 17831ccc2e8aSQu Wenruo * Entry point to file defragmentation. 17841ccc2e8aSQu Wenruo * 17851ccc2e8aSQu Wenruo * @inode: inode to be defragged 17861ccc2e8aSQu Wenruo * @ra: readahead state (can be NUL) 17871ccc2e8aSQu Wenruo * @range: defrag options including range and flags 17881ccc2e8aSQu Wenruo * @newer_than: minimum transid to defrag 17891ccc2e8aSQu Wenruo * @max_to_defrag: max number of sectors to be defragged, if 0, the whole inode 17901ccc2e8aSQu Wenruo * will be defragged. 1791484167daSQu Wenruo * 1792484167daSQu Wenruo * Return <0 for error. 1793c080b414SQu Wenruo * Return >=0 for the number of sectors defragged, and range->start will be updated 1794c080b414SQu Wenruo * to indicate the file offset where next defrag should be started at. 1795c080b414SQu Wenruo * (Mostly for autodefrag, which sets @max_to_defrag thus we may exit early without 1796c080b414SQu Wenruo * defragging all the range). 17977f458a38SFilipe Manana */ 17981ccc2e8aSQu Wenruo int btrfs_defrag_file(struct inode *inode, struct file_ra_state *ra, 17994cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 18004cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 18014cb5300bSChris Mason { 18020b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18037b508037SQu Wenruo unsigned long sectors_defragged = 0; 1804151a31b2SLi Zefan u64 isize = i_size_read(inode); 18057b508037SQu Wenruo u64 cur; 18067b508037SQu Wenruo u64 last_byte; 18071e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 18081ccc2e8aSQu Wenruo bool ra_allocated = false; 18097b508037SQu Wenruo int compress_type = BTRFS_COMPRESS_ZLIB; 18107b508037SQu Wenruo int ret = 0; 18117b508037SQu Wenruo u32 extent_thresh = range->extent_thresh; 181227cdfde1SFilipe Manana pgoff_t start_index; 18134cb5300bSChris Mason 18140abd5b17SLiu Bo if (isize == 0) 18150abd5b17SLiu Bo return 0; 18160abd5b17SLiu Bo 18170abd5b17SLiu Bo if (range->start >= isize) 18180abd5b17SLiu Bo return -EINVAL; 18191a419d85SLi Zefan 18201e2ef46dSDavid Sterba if (do_compress) { 1821ce96b7ffSChengguang Xu if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES) 18221a419d85SLi Zefan return -EINVAL; 18231a419d85SLi Zefan if (range->compress_type) 18241a419d85SLi Zefan compress_type = range->compress_type; 18251a419d85SLi Zefan } 1826f46b5a66SChristoph Hellwig 18270abd5b17SLiu Bo if (extent_thresh == 0) 1828ee22184bSByongho Lee extent_thresh = SZ_256K; 1829f46b5a66SChristoph Hellwig 18307b508037SQu Wenruo if (range->start + range->len > range->start) { 18317b508037SQu Wenruo /* Got a specific range */ 18326b34cd8eSFilipe Manana last_byte = min(isize, range->start + range->len); 18337b508037SQu Wenruo } else { 18347b508037SQu Wenruo /* Defrag until file end */ 18356b34cd8eSFilipe Manana last_byte = isize; 18367b508037SQu Wenruo } 18377b508037SQu Wenruo 18386b34cd8eSFilipe Manana /* Align the range */ 18396b34cd8eSFilipe Manana cur = round_down(range->start, fs_info->sectorsize); 18406b34cd8eSFilipe Manana last_byte = round_up(last_byte, fs_info->sectorsize) - 1; 18416b34cd8eSFilipe Manana 18424cb5300bSChris Mason /* 18431ccc2e8aSQu Wenruo * If we were not given a ra, allocate a readahead context. As 18440a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 18450a52d108SDavid Sterba * we don't error out. 18464cb5300bSChris Mason */ 18471ccc2e8aSQu Wenruo if (!ra) { 18481ccc2e8aSQu Wenruo ra_allocated = true; 184963e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 18500a52d108SDavid Sterba if (ra) 18514cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 18524cb5300bSChris Mason } 18534cb5300bSChris Mason 185427cdfde1SFilipe Manana /* 185527cdfde1SFilipe Manana * Make writeback start from the beginning of the range, so that the 185627cdfde1SFilipe Manana * defrag range can be written sequentially. 185727cdfde1SFilipe Manana */ 185827cdfde1SFilipe Manana start_index = cur >> PAGE_SHIFT; 185927cdfde1SFilipe Manana if (start_index < inode->i_mapping->writeback_index) 186027cdfde1SFilipe Manana inode->i_mapping->writeback_index = start_index; 186127cdfde1SFilipe Manana 18627b508037SQu Wenruo while (cur < last_byte) { 18633c9d31c7SFilipe Manana const unsigned long prev_sectors_defragged = sectors_defragged; 1864966d879bSQu Wenruo u64 last_scanned = cur; 18657b508037SQu Wenruo u64 cluster_end; 18661e701a32SChris Mason 1867b767c2fcSFilipe Manana if (btrfs_defrag_cancelled(fs_info)) { 1868b767c2fcSFilipe Manana ret = -EAGAIN; 1869b767c2fcSFilipe Manana break; 1870b767c2fcSFilipe Manana } 1871b767c2fcSFilipe Manana 18727b508037SQu Wenruo /* We want the cluster end at page boundary when possible */ 18737b508037SQu Wenruo cluster_end = (((cur >> PAGE_SHIFT) + 18747b508037SQu Wenruo (SZ_256K >> PAGE_SHIFT)) << PAGE_SHIFT) - 1; 18757b508037SQu Wenruo cluster_end = min(cluster_end, last_byte); 18764cb5300bSChris Mason 187764708539SJosef Bacik btrfs_inode_lock(inode, 0); 1878eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1879eede2bf3SOmar Sandoval ret = -ETXTBSY; 18807b508037SQu Wenruo btrfs_inode_unlock(inode, 0); 18817b508037SQu Wenruo break; 18827b508037SQu Wenruo } 18837b508037SQu Wenruo if (!(inode->i_sb->s_flags & SB_ACTIVE)) { 18847b508037SQu Wenruo btrfs_inode_unlock(inode, 0); 18857b508037SQu Wenruo break; 18867b508037SQu Wenruo } 18871e2ef46dSDavid Sterba if (do_compress) 1888eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 18897b508037SQu Wenruo ret = defrag_one_cluster(BTRFS_I(inode), ra, cur, 18907b508037SQu Wenruo cluster_end + 1 - cur, extent_thresh, 1891966d879bSQu Wenruo newer_than, do_compress, §ors_defragged, 1892966d879bSQu Wenruo max_to_defrag, &last_scanned); 18933c9d31c7SFilipe Manana 18943c9d31c7SFilipe Manana if (sectors_defragged > prev_sectors_defragged) 18953c9d31c7SFilipe Manana balance_dirty_pages_ratelimited(inode->i_mapping); 18963c9d31c7SFilipe Manana 189764708539SJosef Bacik btrfs_inode_unlock(inode, 0); 18987b508037SQu Wenruo if (ret < 0) 18994cb5300bSChris Mason break; 1900966d879bSQu Wenruo cur = max(cluster_end + 1, last_scanned); 1901c080b414SQu Wenruo if (ret > 0) { 1902c080b414SQu Wenruo ret = 0; 1903c080b414SQu Wenruo break; 1904c080b414SQu Wenruo } 1905ea0eba69SQu Wenruo cond_resched(); 19064cb5300bSChris Mason } 1907f46b5a66SChristoph Hellwig 19087b508037SQu Wenruo if (ra_allocated) 19097b508037SQu Wenruo kfree(ra); 1910c080b414SQu Wenruo /* 1911c080b414SQu Wenruo * Update range.start for autodefrag, this will indicate where to start 1912c080b414SQu Wenruo * in next run. 1913c080b414SQu Wenruo */ 1914c080b414SQu Wenruo range->start = cur; 19157b508037SQu Wenruo if (sectors_defragged) { 19167b508037SQu Wenruo /* 19177b508037SQu Wenruo * We have defragged some sectors, for compression case they 19187b508037SQu Wenruo * need to be written back immediately. 19197b508037SQu Wenruo */ 19207b508037SQu Wenruo if (range->flags & BTRFS_DEFRAG_RANGE_START_IO) { 19211e701a32SChris Mason filemap_flush(inode->i_mapping); 1922dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1923dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1924dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1925dec8ef90SFilipe Manana } 19267b508037SQu Wenruo if (range->compress_type == BTRFS_COMPRESS_LZO) 19270b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 19287b508037SQu Wenruo else if (range->compress_type == BTRFS_COMPRESS_ZSTD) 19295c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 19307b508037SQu Wenruo ret = sectors_defragged; 19311a419d85SLi Zefan } 19321e2ef46dSDavid Sterba if (do_compress) { 193364708539SJosef Bacik btrfs_inode_lock(inode, 0); 1934eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 193564708539SJosef Bacik btrfs_inode_unlock(inode, 0); 1936633085c7SFilipe David Borba Manana } 1937940100a4SChris Mason return ret; 1938f46b5a66SChristoph Hellwig } 1939f46b5a66SChristoph Hellwig 194017aaa434SDavid Sterba /* 194117aaa434SDavid Sterba * Try to start exclusive operation @type or cancel it if it's running. 194217aaa434SDavid Sterba * 194317aaa434SDavid Sterba * Return: 194417aaa434SDavid Sterba * 0 - normal mode, newly claimed op started 194517aaa434SDavid Sterba * >0 - normal mode, something else is running, 194617aaa434SDavid Sterba * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space 194717aaa434SDavid Sterba * ECANCELED - cancel mode, successful cancel 194817aaa434SDavid Sterba * ENOTCONN - cancel mode, operation not running anymore 194917aaa434SDavid Sterba */ 195017aaa434SDavid Sterba static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info, 195117aaa434SDavid Sterba enum btrfs_exclusive_operation type, bool cancel) 195217aaa434SDavid Sterba { 195317aaa434SDavid Sterba if (!cancel) { 195417aaa434SDavid Sterba /* Start normal op */ 195517aaa434SDavid Sterba if (!btrfs_exclop_start(fs_info, type)) 195617aaa434SDavid Sterba return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 195717aaa434SDavid Sterba /* Exclusive operation is now claimed */ 195817aaa434SDavid Sterba return 0; 195917aaa434SDavid Sterba } 196017aaa434SDavid Sterba 196117aaa434SDavid Sterba /* Cancel running op */ 196217aaa434SDavid Sterba if (btrfs_exclop_start_try_lock(fs_info, type)) { 196317aaa434SDavid Sterba /* 196417aaa434SDavid Sterba * This blocks any exclop finish from setting it to NONE, so we 196517aaa434SDavid Sterba * request cancellation. Either it runs and we will wait for it, 196617aaa434SDavid Sterba * or it has finished and no waiting will happen. 196717aaa434SDavid Sterba */ 196817aaa434SDavid Sterba atomic_inc(&fs_info->reloc_cancel_req); 196917aaa434SDavid Sterba btrfs_exclop_start_unlock(fs_info); 197017aaa434SDavid Sterba 197117aaa434SDavid Sterba if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) 197217aaa434SDavid Sterba wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING, 197317aaa434SDavid Sterba TASK_INTERRUPTIBLE); 197417aaa434SDavid Sterba 197517aaa434SDavid Sterba return -ECANCELED; 197617aaa434SDavid Sterba } 197717aaa434SDavid Sterba 197817aaa434SDavid Sterba /* Something else is running or none */ 197917aaa434SDavid Sterba return -ENOTCONN; 198017aaa434SDavid Sterba } 198117aaa434SDavid Sterba 1982198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 198376dda93cSYan, Zheng void __user *arg) 1984f46b5a66SChristoph Hellwig { 1985562d7b15SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 19860b246afaSJeff Mahoney struct inode *inode = file_inode(file); 19870b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1988f46b5a66SChristoph Hellwig u64 new_size; 1989f46b5a66SChristoph Hellwig u64 old_size; 1990f46b5a66SChristoph Hellwig u64 devid = 1; 19910b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1992f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1993f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1994f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1995f46b5a66SChristoph Hellwig char *sizestr; 19969a40f122SGui Hecheng char *retptr; 1997f46b5a66SChristoph Hellwig char *devstr = NULL; 1998f46b5a66SChristoph Hellwig int ret = 0; 1999f46b5a66SChristoph Hellwig int mod = 0; 2000bb059a37SDavid Sterba bool cancel; 2001f46b5a66SChristoph Hellwig 2002e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2003e441d54dSChris Mason return -EPERM; 2004e441d54dSChris Mason 2005198605a8SMiao Xie ret = mnt_want_write_file(file); 2006198605a8SMiao Xie if (ret) 2007198605a8SMiao Xie return ret; 2008198605a8SMiao Xie 2009bb059a37SDavid Sterba /* 2010bb059a37SDavid Sterba * Read the arguments before checking exclusivity to be able to 2011bb059a37SDavid Sterba * distinguish regular resize and cancel 2012bb059a37SDavid Sterba */ 2013dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2014c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2015c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2016bb059a37SDavid Sterba goto out_drop; 2017c9e9f97bSIlya Dryomov } 20185516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2019f46b5a66SChristoph Hellwig sizestr = vol_args->name; 2020bb059a37SDavid Sterba cancel = (strcmp("cancel", sizestr) == 0); 2021bb059a37SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel); 2022bb059a37SDavid Sterba if (ret) 2023bb059a37SDavid Sterba goto out_free; 2024bb059a37SDavid Sterba /* Exclusive operation is now claimed */ 2025bb059a37SDavid Sterba 2026f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 2027f46b5a66SChristoph Hellwig if (devstr) { 2028f46b5a66SChristoph Hellwig sizestr = devstr + 1; 2029f46b5a66SChristoph Hellwig *devstr = '\0'; 2030f46b5a66SChristoph Hellwig devstr = vol_args->name; 203158dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 203258dfae63SZhangZhen if (ret) 2033bb059a37SDavid Sterba goto out_finish; 2034dfd79829SMiao Xie if (!devid) { 2035dfd79829SMiao Xie ret = -EINVAL; 2036bb059a37SDavid Sterba goto out_finish; 2037dfd79829SMiao Xie } 20380b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 2039f46b5a66SChristoph Hellwig } 2040dba60f3fSMiao Xie 2041562d7b15SJosef Bacik args.devid = devid; 2042562d7b15SJosef Bacik device = btrfs_find_device(fs_info->fs_devices, &args); 2043f46b5a66SChristoph Hellwig if (!device) { 20440b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 2045c1c9ff7cSGeert Uytterhoeven devid); 2046dfd79829SMiao Xie ret = -ENODEV; 2047bb059a37SDavid Sterba goto out_finish; 2048f46b5a66SChristoph Hellwig } 2049dba60f3fSMiao Xie 2050ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 20510b246afaSJeff Mahoney btrfs_info(fs_info, 2052efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 2053c1c9ff7cSGeert Uytterhoeven devid); 2054dfd79829SMiao Xie ret = -EPERM; 2055bb059a37SDavid Sterba goto out_finish; 20564e42ae1bSLiu Bo } 20574e42ae1bSLiu Bo 2058f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 2059cda00ebaSChristoph Hellwig new_size = bdev_nr_bytes(device->bdev); 2060f46b5a66SChristoph Hellwig else { 2061f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 2062f46b5a66SChristoph Hellwig mod = -1; 2063f46b5a66SChristoph Hellwig sizestr++; 2064f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 2065f46b5a66SChristoph Hellwig mod = 1; 2066f46b5a66SChristoph Hellwig sizestr++; 2067f46b5a66SChristoph Hellwig } 20689a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 20699a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 2070f46b5a66SChristoph Hellwig ret = -EINVAL; 2071bb059a37SDavid Sterba goto out_finish; 2072f46b5a66SChristoph Hellwig } 2073f46b5a66SChristoph Hellwig } 2074f46b5a66SChristoph Hellwig 2075401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 2076dfd79829SMiao Xie ret = -EPERM; 2077bb059a37SDavid Sterba goto out_finish; 207863a212abSStefan Behrens } 207963a212abSStefan Behrens 20807cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 2081f46b5a66SChristoph Hellwig 2082f46b5a66SChristoph Hellwig if (mod < 0) { 2083f46b5a66SChristoph Hellwig if (new_size > old_size) { 2084f46b5a66SChristoph Hellwig ret = -EINVAL; 2085bb059a37SDavid Sterba goto out_finish; 2086f46b5a66SChristoph Hellwig } 2087f46b5a66SChristoph Hellwig new_size = old_size - new_size; 2088f46b5a66SChristoph Hellwig } else if (mod > 0) { 2089eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 2090902c68a4SGui Hecheng ret = -ERANGE; 2091bb059a37SDavid Sterba goto out_finish; 2092eb8052e0SWenliang Fan } 2093f46b5a66SChristoph Hellwig new_size = old_size + new_size; 2094f46b5a66SChristoph Hellwig } 2095f46b5a66SChristoph Hellwig 2096ee22184bSByongho Lee if (new_size < SZ_256M) { 2097f46b5a66SChristoph Hellwig ret = -EINVAL; 2098bb059a37SDavid Sterba goto out_finish; 2099f46b5a66SChristoph Hellwig } 2100cda00ebaSChristoph Hellwig if (new_size > bdev_nr_bytes(device->bdev)) { 2101f46b5a66SChristoph Hellwig ret = -EFBIG; 2102bb059a37SDavid Sterba goto out_finish; 2103f46b5a66SChristoph Hellwig } 2104f46b5a66SChristoph Hellwig 210547f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 2106f46b5a66SChristoph Hellwig 2107f46b5a66SChristoph Hellwig if (new_size > old_size) { 2108a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 210998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 211098d5dc13STsutomu Itoh ret = PTR_ERR(trans); 2111bb059a37SDavid Sterba goto out_finish; 211298d5dc13STsutomu Itoh } 2113f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 21143a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 2115ece7d20eSMike Fleetwood } else if (new_size < old_size) { 2116f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 21170253f40eSjeff.liu } /* equal, nothing need to do */ 2118f46b5a66SChristoph Hellwig 2119faf8f7b9SMarcos Paulo de Souza if (ret == 0 && new_size != old_size) 2120faf8f7b9SMarcos Paulo de Souza btrfs_info_in_rcu(fs_info, 2121faf8f7b9SMarcos Paulo de Souza "resize device %s (devid %llu) from %llu to %llu", 2122faf8f7b9SMarcos Paulo de Souza rcu_str_deref(device->name), device->devid, 2123faf8f7b9SMarcos Paulo de Souza old_size, new_size); 2124bb059a37SDavid Sterba out_finish: 2125bb059a37SDavid Sterba btrfs_exclop_finish(fs_info); 2126c9e9f97bSIlya Dryomov out_free: 2127f46b5a66SChristoph Hellwig kfree(vol_args); 2128bb059a37SDavid Sterba out_drop: 212918f39c41SIlya Dryomov mnt_drop_write_file(file); 2130f46b5a66SChristoph Hellwig return ret; 2131f46b5a66SChristoph Hellwig } 2132f46b5a66SChristoph Hellwig 21335d54c67eSNikolay Borisov static noinline int __btrfs_ioctl_snap_create(struct file *file, 21344d4340c9SChristian Brauner struct user_namespace *mnt_userns, 213552f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 21365d54c67eSNikolay Borisov bool readonly, 21378696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 2138f46b5a66SChristoph Hellwig { 2139f46b5a66SChristoph Hellwig int namelen; 21403de4586cSChris Mason int ret = 0; 2141f46b5a66SChristoph Hellwig 2142325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 2143325c50e3SJeff Mahoney return -ENOTDIR; 2144325c50e3SJeff Mahoney 2145a874a63eSLiu Bo ret = mnt_want_write_file(file); 2146a874a63eSLiu Bo if (ret) 2147a874a63eSLiu Bo goto out; 2148a874a63eSLiu Bo 214972fd032eSSage Weil namelen = strlen(name); 215072fd032eSSage Weil if (strchr(name, '/')) { 2151f46b5a66SChristoph Hellwig ret = -EINVAL; 2152a874a63eSLiu Bo goto out_drop_write; 2153f46b5a66SChristoph Hellwig } 2154f46b5a66SChristoph Hellwig 215516780cabSChris Mason if (name[0] == '.' && 215616780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 215716780cabSChris Mason ret = -EEXIST; 2158a874a63eSLiu Bo goto out_drop_write; 215916780cabSChris Mason } 216016780cabSChris Mason 21613de4586cSChris Mason if (subvol) { 21624d4340c9SChristian Brauner ret = btrfs_mksubvol(&file->f_path, mnt_userns, name, 21634d4340c9SChristian Brauner namelen, NULL, readonly, inherit); 2164cb8e7090SChristoph Hellwig } else { 21652903ff01SAl Viro struct fd src = fdget(fd); 21663de4586cSChris Mason struct inode *src_inode; 21672903ff01SAl Viro if (!src.file) { 21683de4586cSChris Mason ret = -EINVAL; 2169a874a63eSLiu Bo goto out_drop_write; 21703de4586cSChris Mason } 21713de4586cSChris Mason 2172496ad9aaSAl Viro src_inode = file_inode(src.file); 2173496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 2174c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 2175efe120a0SFrank Holton "Snapshot src from another FS"); 217623ad5b17SKusanagi Kouichi ret = -EXDEV; 21774d4340c9SChristian Brauner } else if (!inode_owner_or_capable(mnt_userns, src_inode)) { 2178d0242061SDavid Sterba /* 2179d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 2180d0242061SDavid Sterba * are limited to own subvolumes only 2181d0242061SDavid Sterba */ 2182d0242061SDavid Sterba ret = -EPERM; 2183ecd18815SAl Viro } else { 21844d4340c9SChristian Brauner ret = btrfs_mksnapshot(&file->f_path, mnt_userns, 21854d4340c9SChristian Brauner name, namelen, 218672fd032eSSage Weil BTRFS_I(src_inode)->root, 21875d54c67eSNikolay Borisov readonly, inherit); 2188ecd18815SAl Viro } 21892903ff01SAl Viro fdput(src); 2190cb8e7090SChristoph Hellwig } 2191a874a63eSLiu Bo out_drop_write: 2192a874a63eSLiu Bo mnt_drop_write_file(file); 2193f46b5a66SChristoph Hellwig out: 219472fd032eSSage Weil return ret; 219572fd032eSSage Weil } 219672fd032eSSage Weil 219772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 2198fa0d2b9bSLi Zefan void __user *arg, int subvol) 219972fd032eSSage Weil { 2200fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 220172fd032eSSage Weil int ret; 220272fd032eSSage Weil 2203325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 2204325c50e3SJeff Mahoney return -ENOTDIR; 2205325c50e3SJeff Mahoney 2206fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2207fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 2208fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 2209fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2210fa0d2b9bSLi Zefan 22114d4340c9SChristian Brauner ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), 22124d4340c9SChristian Brauner vol_args->name, vol_args->fd, subvol, 22134d4340c9SChristian Brauner false, NULL); 2214fa0d2b9bSLi Zefan 2215fa0d2b9bSLi Zefan kfree(vol_args); 2216fa0d2b9bSLi Zefan return ret; 2217fa0d2b9bSLi Zefan } 2218fa0d2b9bSLi Zefan 2219fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 2220fa0d2b9bSLi Zefan void __user *arg, int subvol) 2221fa0d2b9bSLi Zefan { 2222fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 2223fa0d2b9bSLi Zefan int ret; 2224b83cc969SLi Zefan bool readonly = false; 22256f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 222672fd032eSSage Weil 2227325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 2228325c50e3SJeff Mahoney return -ENOTDIR; 2229325c50e3SJeff Mahoney 2230fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2231fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 2232fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 2233fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 2234fdfb1e4fSLi Zefan 2235673990dbSDavid Sterba if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) { 2236b83cc969SLi Zefan ret = -EOPNOTSUPP; 2237c47ca32dSDan Carpenter goto free_args; 2238fdfb1e4fSLi Zefan } 2239fdfb1e4fSLi Zefan 2240b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 2241b83cc969SLi Zefan readonly = true; 22426f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 22435011c5a6SDan Carpenter u64 nums; 22445011c5a6SDan Carpenter 22455011c5a6SDan Carpenter if (vol_args->size < sizeof(*inherit) || 22465011c5a6SDan Carpenter vol_args->size > PAGE_SIZE) { 22476f72c7e2SArne Jansen ret = -EINVAL; 2248c47ca32dSDan Carpenter goto free_args; 22496f72c7e2SArne Jansen } 22506f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 22516f72c7e2SArne Jansen if (IS_ERR(inherit)) { 22526f72c7e2SArne Jansen ret = PTR_ERR(inherit); 2253c47ca32dSDan Carpenter goto free_args; 22546f72c7e2SArne Jansen } 22555011c5a6SDan Carpenter 22565011c5a6SDan Carpenter if (inherit->num_qgroups > PAGE_SIZE || 22575011c5a6SDan Carpenter inherit->num_ref_copies > PAGE_SIZE || 22585011c5a6SDan Carpenter inherit->num_excl_copies > PAGE_SIZE) { 22595011c5a6SDan Carpenter ret = -EINVAL; 22605011c5a6SDan Carpenter goto free_inherit; 22615011c5a6SDan Carpenter } 22625011c5a6SDan Carpenter 22635011c5a6SDan Carpenter nums = inherit->num_qgroups + 2 * inherit->num_ref_copies + 22645011c5a6SDan Carpenter 2 * inherit->num_excl_copies; 22655011c5a6SDan Carpenter if (vol_args->size != struct_size(inherit, qgroups, nums)) { 22665011c5a6SDan Carpenter ret = -EINVAL; 22675011c5a6SDan Carpenter goto free_inherit; 22685011c5a6SDan Carpenter } 22696f72c7e2SArne Jansen } 227075eaa0e2SSage Weil 22714d4340c9SChristian Brauner ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), 22724d4340c9SChristian Brauner vol_args->name, vol_args->fd, subvol, 22734d4340c9SChristian Brauner readonly, inherit); 2274c47ca32dSDan Carpenter if (ret) 2275c47ca32dSDan Carpenter goto free_inherit; 2276c47ca32dSDan Carpenter free_inherit: 22776f72c7e2SArne Jansen kfree(inherit); 2278c47ca32dSDan Carpenter free_args: 2279c47ca32dSDan Carpenter kfree(vol_args); 2280f46b5a66SChristoph Hellwig return ret; 2281f46b5a66SChristoph Hellwig } 2282f46b5a66SChristoph Hellwig 22839ad12305SSahil Kang static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode, 22840caa102dSLi Zefan void __user *arg) 22850caa102dSLi Zefan { 22860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 22870caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 22880caa102dSLi Zefan int ret = 0; 22890caa102dSLi Zefan u64 flags = 0; 22900caa102dSLi Zefan 22914a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 22920caa102dSLi Zefan return -EINVAL; 22930caa102dSLi Zefan 22940b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 22950caa102dSLi Zefan if (btrfs_root_readonly(root)) 22960caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 22970b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 22980caa102dSLi Zefan 22990caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 23000caa102dSLi Zefan ret = -EFAULT; 23010caa102dSLi Zefan 23020caa102dSLi Zefan return ret; 23030caa102dSLi Zefan } 23040caa102dSLi Zefan 23050caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 23060caa102dSLi Zefan void __user *arg) 23070caa102dSLi Zefan { 2308496ad9aaSAl Viro struct inode *inode = file_inode(file); 23090b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 23100caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 23110caa102dSLi Zefan struct btrfs_trans_handle *trans; 23120caa102dSLi Zefan u64 root_flags; 23130caa102dSLi Zefan u64 flags; 23140caa102dSLi Zefan int ret = 0; 23150caa102dSLi Zefan 231639e1674fSChristian Brauner if (!inode_owner_or_capable(file_mnt_user_ns(file), inode)) 2317bd60ea0fSDavid Sterba return -EPERM; 2318bd60ea0fSDavid Sterba 2319b9ca0664SLiu Bo ret = mnt_want_write_file(file); 2320b9ca0664SLiu Bo if (ret) 2321b9ca0664SLiu Bo goto out; 23220caa102dSLi Zefan 23234a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 2324b9ca0664SLiu Bo ret = -EINVAL; 2325b9ca0664SLiu Bo goto out_drop_write; 2326b9ca0664SLiu Bo } 23270caa102dSLi Zefan 2328b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 2329b9ca0664SLiu Bo ret = -EFAULT; 2330b9ca0664SLiu Bo goto out_drop_write; 2331b9ca0664SLiu Bo } 23320caa102dSLi Zefan 2333b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 2334b9ca0664SLiu Bo ret = -EOPNOTSUPP; 2335b9ca0664SLiu Bo goto out_drop_write; 2336b9ca0664SLiu Bo } 23370caa102dSLi Zefan 23380b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 23390caa102dSLi Zefan 23400caa102dSLi Zefan /* nothing to do */ 23410caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 2342b9ca0664SLiu Bo goto out_drop_sem; 23430caa102dSLi Zefan 23440caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 23452c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 23460caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 23470caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 23482c686537SDavid Sterba } else { 23492c686537SDavid Sterba /* 23502c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 23512c686537SDavid Sterba * send 23522c686537SDavid Sterba */ 23532c686537SDavid Sterba spin_lock(&root->root_item_lock); 23542c686537SDavid Sterba if (root->send_in_progress == 0) { 23550caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 23560caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 23572c686537SDavid Sterba spin_unlock(&root->root_item_lock); 23582c686537SDavid Sterba } else { 23592c686537SDavid Sterba spin_unlock(&root->root_item_lock); 23600b246afaSJeff Mahoney btrfs_warn(fs_info, 23612c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 23622c686537SDavid Sterba root->root_key.objectid); 23632c686537SDavid Sterba ret = -EPERM; 23642c686537SDavid Sterba goto out_drop_sem; 23652c686537SDavid Sterba } 23662c686537SDavid Sterba } 23670caa102dSLi Zefan 23680caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 23690caa102dSLi Zefan if (IS_ERR(trans)) { 23700caa102dSLi Zefan ret = PTR_ERR(trans); 23710caa102dSLi Zefan goto out_reset; 23720caa102dSLi Zefan } 23730caa102dSLi Zefan 23740b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 23750caa102dSLi Zefan &root->root_key, &root->root_item); 23769417ebc8SNikolay Borisov if (ret < 0) { 23779417ebc8SNikolay Borisov btrfs_end_transaction(trans); 23789417ebc8SNikolay Borisov goto out_reset; 23799417ebc8SNikolay Borisov } 23800caa102dSLi Zefan 23819417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 23829417ebc8SNikolay Borisov 23830caa102dSLi Zefan out_reset: 23840caa102dSLi Zefan if (ret) 23850caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 2386b9ca0664SLiu Bo out_drop_sem: 23870b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 2388b9ca0664SLiu Bo out_drop_write: 2389b9ca0664SLiu Bo mnt_drop_write_file(file); 2390b9ca0664SLiu Bo out: 23910caa102dSLi Zefan return ret; 23920caa102dSLi Zefan } 23930caa102dSLi Zefan 2394ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2395ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2396ac8e9819SChris Mason { 2397abc6e134SChris Mason struct btrfs_key test; 2398abc6e134SChris Mason int ret; 2399abc6e134SChris Mason 2400abc6e134SChris Mason test.objectid = sk->min_objectid; 2401abc6e134SChris Mason test.type = sk->min_type; 2402abc6e134SChris Mason test.offset = sk->min_offset; 2403abc6e134SChris Mason 2404abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2405abc6e134SChris Mason if (ret < 0) 2406ac8e9819SChris Mason return 0; 2407abc6e134SChris Mason 2408abc6e134SChris Mason test.objectid = sk->max_objectid; 2409abc6e134SChris Mason test.type = sk->max_type; 2410abc6e134SChris Mason test.offset = sk->max_offset; 2411abc6e134SChris Mason 2412abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2413abc6e134SChris Mason if (ret > 0) 2414ac8e9819SChris Mason return 0; 2415ac8e9819SChris Mason return 1; 2416ac8e9819SChris Mason } 2417ac8e9819SChris Mason 2418df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2419ac8e9819SChris Mason struct btrfs_key *key, 2420ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 24219b6e817dSGerhard Heift size_t *buf_size, 2422ba346b35SGerhard Heift char __user *ubuf, 2423ac8e9819SChris Mason unsigned long *sk_offset, 2424ac8e9819SChris Mason int *num_found) 2425ac8e9819SChris Mason { 2426ac8e9819SChris Mason u64 found_transid; 2427ac8e9819SChris Mason struct extent_buffer *leaf; 2428ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2429dd81d459SNaohiro Aota struct btrfs_key test; 2430ac8e9819SChris Mason unsigned long item_off; 2431ac8e9819SChris Mason unsigned long item_len; 2432ac8e9819SChris Mason int nritems; 2433ac8e9819SChris Mason int i; 2434ac8e9819SChris Mason int slot; 2435ac8e9819SChris Mason int ret = 0; 2436ac8e9819SChris Mason 2437ac8e9819SChris Mason leaf = path->nodes[0]; 2438ac8e9819SChris Mason slot = path->slots[0]; 2439ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2440ac8e9819SChris Mason 2441ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2442ac8e9819SChris Mason i = nritems; 2443ac8e9819SChris Mason goto advance_key; 2444ac8e9819SChris Mason } 2445ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2446ac8e9819SChris Mason 2447ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2448ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 24493212fa14SJosef Bacik item_len = btrfs_item_size(leaf, i); 2450ac8e9819SChris Mason 245103b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 245203b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 245303b71c6cSGabriel de Perthuis continue; 245403b71c6cSGabriel de Perthuis 24559b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 24568f5f6178SGerhard Heift if (*num_found) { 2457ac8e9819SChris Mason ret = 1; 24588f5f6178SGerhard Heift goto out; 24598f5f6178SGerhard Heift } 24608f5f6178SGerhard Heift 24618f5f6178SGerhard Heift /* 24628f5f6178SGerhard Heift * return one empty item back for v1, which does not 24638f5f6178SGerhard Heift * handle -EOVERFLOW 24648f5f6178SGerhard Heift */ 24658f5f6178SGerhard Heift 24669b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2467ac8e9819SChris Mason item_len = 0; 24688f5f6178SGerhard Heift ret = -EOVERFLOW; 24698f5f6178SGerhard Heift } 2470ac8e9819SChris Mason 24719b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2472ac8e9819SChris Mason ret = 1; 247325c9bc2eSGerhard Heift goto out; 2474ac8e9819SChris Mason } 2475ac8e9819SChris Mason 2476ac8e9819SChris Mason sh.objectid = key->objectid; 2477ac8e9819SChris Mason sh.offset = key->offset; 2478ac8e9819SChris Mason sh.type = key->type; 2479ac8e9819SChris Mason sh.len = item_len; 2480ac8e9819SChris Mason sh.transid = found_transid; 2481ac8e9819SChris Mason 2482a48b73ecSJosef Bacik /* 2483a48b73ecSJosef Bacik * Copy search result header. If we fault then loop again so we 2484a48b73ecSJosef Bacik * can fault in the pages and -EFAULT there if there's a 2485a48b73ecSJosef Bacik * problem. Otherwise we'll fault and then copy the buffer in 2486a48b73ecSJosef Bacik * properly this next time through 2487a48b73ecSJosef Bacik */ 2488a48b73ecSJosef Bacik if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) { 2489a48b73ecSJosef Bacik ret = 0; 2490ba346b35SGerhard Heift goto out; 2491ba346b35SGerhard Heift } 2492ba346b35SGerhard Heift 2493ac8e9819SChris Mason *sk_offset += sizeof(sh); 2494ac8e9819SChris Mason 2495ac8e9819SChris Mason if (item_len) { 2496ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2497a48b73ecSJosef Bacik /* 2498a48b73ecSJosef Bacik * Copy the item, same behavior as above, but reset the 2499a48b73ecSJosef Bacik * * sk_offset so we copy the full thing again. 2500a48b73ecSJosef Bacik */ 2501a48b73ecSJosef Bacik if (read_extent_buffer_to_user_nofault(leaf, up, 2502ba346b35SGerhard Heift item_off, item_len)) { 2503a48b73ecSJosef Bacik ret = 0; 2504a48b73ecSJosef Bacik *sk_offset -= sizeof(sh); 2505ba346b35SGerhard Heift goto out; 2506ba346b35SGerhard Heift } 2507ba346b35SGerhard Heift 2508ac8e9819SChris Mason *sk_offset += item_len; 2509ac8e9819SChris Mason } 2510e2156867SHugo Mills (*num_found)++; 2511ac8e9819SChris Mason 25128f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 25138f5f6178SGerhard Heift goto out; 25148f5f6178SGerhard Heift 251525c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 251625c9bc2eSGerhard Heift ret = 1; 251725c9bc2eSGerhard Heift goto out; 251825c9bc2eSGerhard Heift } 2519ac8e9819SChris Mason } 2520ac8e9819SChris Mason advance_key: 2521ac8e9819SChris Mason ret = 0; 2522dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2523dd81d459SNaohiro Aota test.type = sk->max_type; 2524dd81d459SNaohiro Aota test.offset = sk->max_offset; 2525dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2526dd81d459SNaohiro Aota ret = 1; 2527dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2528abc6e134SChris Mason key->offset++; 2529dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2530abc6e134SChris Mason key->offset = 0; 2531abc6e134SChris Mason key->type++; 2532dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2533abc6e134SChris Mason key->offset = 0; 2534abc6e134SChris Mason key->type = 0; 2535abc6e134SChris Mason key->objectid++; 2536abc6e134SChris Mason } else 2537abc6e134SChris Mason ret = 1; 253825c9bc2eSGerhard Heift out: 2539ba346b35SGerhard Heift /* 2540ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2541ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2542ba346b35SGerhard Heift * * all items were found 2543ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2544ba346b35SGerhard Heift * leaf 2545ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2546ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2547ba346b35SGerhard Heift */ 2548ac8e9819SChris Mason return ret; 2549ac8e9819SChris Mason } 2550ac8e9819SChris Mason 2551ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 255212544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 25539b6e817dSGerhard Heift size_t *buf_size, 2554ba346b35SGerhard Heift char __user *ubuf) 2555ac8e9819SChris Mason { 25560b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2557ac8e9819SChris Mason struct btrfs_root *root; 2558ac8e9819SChris Mason struct btrfs_key key; 2559ac8e9819SChris Mason struct btrfs_path *path; 2560ac8e9819SChris Mason int ret; 2561ac8e9819SChris Mason int num_found = 0; 2562ac8e9819SChris Mason unsigned long sk_offset = 0; 2563ac8e9819SChris Mason 25649b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 25659b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 256612544442SGerhard Heift return -EOVERFLOW; 25679b6e817dSGerhard Heift } 256812544442SGerhard Heift 2569ac8e9819SChris Mason path = btrfs_alloc_path(); 2570ac8e9819SChris Mason if (!path) 2571ac8e9819SChris Mason return -ENOMEM; 2572ac8e9819SChris Mason 2573ac8e9819SChris Mason if (sk->tree_id == 0) { 2574ac8e9819SChris Mason /* search the root of the inode that was passed */ 257500246528SJosef Bacik root = btrfs_grab_root(BTRFS_I(inode)->root); 2576ac8e9819SChris Mason } else { 257756e9357aSDavid Sterba root = btrfs_get_fs_root(info, sk->tree_id, true); 2578ac8e9819SChris Mason if (IS_ERR(root)) { 2579ac8e9819SChris Mason btrfs_free_path(path); 2580ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2581ac8e9819SChris Mason } 2582ac8e9819SChris Mason } 2583ac8e9819SChris Mason 2584ac8e9819SChris Mason key.objectid = sk->min_objectid; 2585ac8e9819SChris Mason key.type = sk->min_type; 2586ac8e9819SChris Mason key.offset = sk->min_offset; 2587ac8e9819SChris Mason 2588ac8e9819SChris Mason while (1) { 2589bb523b40SAndreas Gruenbacher ret = -EFAULT; 259018788e34SCatalin Marinas /* 259118788e34SCatalin Marinas * Ensure that the whole user buffer is faulted in at sub-page 259218788e34SCatalin Marinas * granularity, otherwise the loop may live-lock. 259318788e34SCatalin Marinas */ 259418788e34SCatalin Marinas if (fault_in_subpage_writeable(ubuf + sk_offset, 259518788e34SCatalin Marinas *buf_size - sk_offset)) 2596a48b73ecSJosef Bacik break; 2597a48b73ecSJosef Bacik 25986174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2599ac8e9819SChris Mason if (ret != 0) { 2600ac8e9819SChris Mason if (ret > 0) 2601ac8e9819SChris Mason ret = 0; 2602ac8e9819SChris Mason goto err; 2603ac8e9819SChris Mason } 2604df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2605ac8e9819SChris Mason &sk_offset, &num_found); 2606b3b4aa74SDavid Sterba btrfs_release_path(path); 260725c9bc2eSGerhard Heift if (ret) 2608ac8e9819SChris Mason break; 2609ac8e9819SChris Mason 2610ac8e9819SChris Mason } 26118f5f6178SGerhard Heift if (ret > 0) 2612ac8e9819SChris Mason ret = 0; 2613ac8e9819SChris Mason err: 2614ac8e9819SChris Mason sk->nr_items = num_found; 261500246528SJosef Bacik btrfs_put_root(root); 2616ac8e9819SChris Mason btrfs_free_path(path); 2617ac8e9819SChris Mason return ret; 2618ac8e9819SChris Mason } 2619ac8e9819SChris Mason 26209ad12305SSahil Kang static noinline int btrfs_ioctl_tree_search(struct inode *inode, 2621ac8e9819SChris Mason void __user *argp) 2622ac8e9819SChris Mason { 26230d031dc4SYu Zhe struct btrfs_ioctl_search_args __user *uargs = argp; 2624ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2625ac8e9819SChris Mason int ret; 26269b6e817dSGerhard Heift size_t buf_size; 2627ac8e9819SChris Mason 2628ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2629ac8e9819SChris Mason return -EPERM; 2630ac8e9819SChris Mason 2631ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2632ba346b35SGerhard Heift return -EFAULT; 2633ba346b35SGerhard Heift 2634ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2635ac8e9819SChris Mason 2636ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 26378f5f6178SGerhard Heift 26388f5f6178SGerhard Heift /* 26398f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 26408f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 26418f5f6178SGerhard Heift */ 26428f5f6178SGerhard Heift if (ret == -EOVERFLOW) 26438f5f6178SGerhard Heift ret = 0; 26448f5f6178SGerhard Heift 2645ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2646ac8e9819SChris Mason ret = -EFAULT; 2647ac8e9819SChris Mason return ret; 2648ac8e9819SChris Mason } 2649ac8e9819SChris Mason 26509ad12305SSahil Kang static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode, 2651cc68a8a5SGerhard Heift void __user *argp) 2652cc68a8a5SGerhard Heift { 26530d031dc4SYu Zhe struct btrfs_ioctl_search_args_v2 __user *uarg = argp; 2654cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2655cc68a8a5SGerhard Heift int ret; 2656cc68a8a5SGerhard Heift size_t buf_size; 2657ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2658cc68a8a5SGerhard Heift 2659cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2660cc68a8a5SGerhard Heift return -EPERM; 2661cc68a8a5SGerhard Heift 2662cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2663cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2664cc68a8a5SGerhard Heift return -EFAULT; 2665cc68a8a5SGerhard Heift 2666cc68a8a5SGerhard Heift buf_size = args.buf_size; 2667cc68a8a5SGerhard Heift 2668cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2669cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2670cc68a8a5SGerhard Heift buf_size = buf_limit; 2671cc68a8a5SGerhard Heift 2672cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2673718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2674cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2675cc68a8a5SGerhard Heift ret = -EFAULT; 2676cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2677cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2678cc68a8a5SGerhard Heift ret = -EFAULT; 2679cc68a8a5SGerhard Heift 268076dda93cSYan, Zheng return ret; 268176dda93cSYan, Zheng } 268276dda93cSYan, Zheng 268398d377a0STARUISI Hiroaki /* 2684ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2685ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 268698d377a0STARUISI Hiroaki */ 268798d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 268898d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 268998d377a0STARUISI Hiroaki { 269098d377a0STARUISI Hiroaki struct btrfs_root *root; 269198d377a0STARUISI Hiroaki struct btrfs_key key; 2692ac8e9819SChris Mason char *ptr; 269398d377a0STARUISI Hiroaki int ret = -1; 269498d377a0STARUISI Hiroaki int slot; 269598d377a0STARUISI Hiroaki int len; 269698d377a0STARUISI Hiroaki int total_len = 0; 269798d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 269898d377a0STARUISI Hiroaki struct extent_buffer *l; 269998d377a0STARUISI Hiroaki struct btrfs_path *path; 270098d377a0STARUISI Hiroaki 270198d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 270298d377a0STARUISI Hiroaki name[0]='\0'; 270398d377a0STARUISI Hiroaki return 0; 270498d377a0STARUISI Hiroaki } 270598d377a0STARUISI Hiroaki 270698d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 270798d377a0STARUISI Hiroaki if (!path) 270898d377a0STARUISI Hiroaki return -ENOMEM; 270998d377a0STARUISI Hiroaki 2710c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 271198d377a0STARUISI Hiroaki 271256e9357aSDavid Sterba root = btrfs_get_fs_root(info, tree_id, true); 271398d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2714ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 271588234012SJosef Bacik root = NULL; 271688234012SJosef Bacik goto out; 271788234012SJosef Bacik } 271898d377a0STARUISI Hiroaki 271998d377a0STARUISI Hiroaki key.objectid = dirid; 272098d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 27218ad6fcabSChris Mason key.offset = (u64)-1; 272298d377a0STARUISI Hiroaki 272398d377a0STARUISI Hiroaki while (1) { 27240ff40a91SMarcos Paulo de Souza ret = btrfs_search_backwards(root, &key, path); 272518674c6cSFilipe David Borba Manana if (ret < 0) 272618674c6cSFilipe David Borba Manana goto out; 272718674c6cSFilipe David Borba Manana else if (ret > 0) { 2728ac8e9819SChris Mason ret = -ENOENT; 272998d377a0STARUISI Hiroaki goto out; 2730ac8e9819SChris Mason } 273118674c6cSFilipe David Borba Manana 273218674c6cSFilipe David Borba Manana l = path->nodes[0]; 273318674c6cSFilipe David Borba Manana slot = path->slots[0]; 273498d377a0STARUISI Hiroaki 273598d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 273698d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 273798d377a0STARUISI Hiroaki ptr -= len + 1; 273898d377a0STARUISI Hiroaki total_len += len + 1; 2739a696cf35SFilipe David Borba Manana if (ptr < name) { 2740a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 274198d377a0STARUISI Hiroaki goto out; 2742a696cf35SFilipe David Borba Manana } 274398d377a0STARUISI Hiroaki 274498d377a0STARUISI Hiroaki *(ptr + len) = '/'; 274598d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 274698d377a0STARUISI Hiroaki 274798d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 274898d377a0STARUISI Hiroaki break; 274998d377a0STARUISI Hiroaki 2750b3b4aa74SDavid Sterba btrfs_release_path(path); 275198d377a0STARUISI Hiroaki key.objectid = key.offset; 27528ad6fcabSChris Mason key.offset = (u64)-1; 275398d377a0STARUISI Hiroaki dirid = key.objectid; 275498d377a0STARUISI Hiroaki } 275577906a50SLi Zefan memmove(name, ptr, total_len); 275698d377a0STARUISI Hiroaki name[total_len] = '\0'; 275798d377a0STARUISI Hiroaki ret = 0; 275898d377a0STARUISI Hiroaki out: 275900246528SJosef Bacik btrfs_put_root(root); 276098d377a0STARUISI Hiroaki btrfs_free_path(path); 2761ac8e9819SChris Mason return ret; 2762ac8e9819SChris Mason } 2763ac8e9819SChris Mason 27646623d9a0SChristian Brauner static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns, 27656623d9a0SChristian Brauner struct inode *inode, 276623d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 276723d0b79dSTomohiro Misono { 276823d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 276923d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 277023d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 277123d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 277223d0b79dSTomohiro Misono u64 dirid = args->dirid; 277323d0b79dSTomohiro Misono unsigned long item_off; 277423d0b79dSTomohiro Misono unsigned long item_len; 277523d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 277623d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 2777b8a49ae1SJosef Bacik struct btrfs_root *root = NULL; 277823d0b79dSTomohiro Misono struct btrfs_path *path; 277923d0b79dSTomohiro Misono struct btrfs_key key, key2; 278023d0b79dSTomohiro Misono struct extent_buffer *leaf; 278123d0b79dSTomohiro Misono struct inode *temp_inode; 278223d0b79dSTomohiro Misono char *ptr; 278323d0b79dSTomohiro Misono int slot; 278423d0b79dSTomohiro Misono int len; 278523d0b79dSTomohiro Misono int total_len = 0; 278623d0b79dSTomohiro Misono int ret; 278723d0b79dSTomohiro Misono 278823d0b79dSTomohiro Misono path = btrfs_alloc_path(); 278923d0b79dSTomohiro Misono if (!path) 279023d0b79dSTomohiro Misono return -ENOMEM; 279123d0b79dSTomohiro Misono 279223d0b79dSTomohiro Misono /* 279323d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 279423d0b79dSTomohiro Misono * construct the path in from the bottom up. 279523d0b79dSTomohiro Misono */ 279623d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 279723d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 279823d0b79dSTomohiro Misono 279956e9357aSDavid Sterba root = btrfs_get_fs_root(fs_info, treeid, true); 280023d0b79dSTomohiro Misono if (IS_ERR(root)) { 280123d0b79dSTomohiro Misono ret = PTR_ERR(root); 280223d0b79dSTomohiro Misono goto out; 280323d0b79dSTomohiro Misono } 280423d0b79dSTomohiro Misono 280523d0b79dSTomohiro Misono key.objectid = dirid; 280623d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 280723d0b79dSTomohiro Misono key.offset = (u64)-1; 280823d0b79dSTomohiro Misono while (1) { 28090ff40a91SMarcos Paulo de Souza ret = btrfs_search_backwards(root, &key, path); 28100ff40a91SMarcos Paulo de Souza if (ret < 0) 2811b8a49ae1SJosef Bacik goto out_put; 28120ff40a91SMarcos Paulo de Souza else if (ret > 0) { 281323d0b79dSTomohiro Misono ret = -ENOENT; 2814b8a49ae1SJosef Bacik goto out_put; 281523d0b79dSTomohiro Misono } 281623d0b79dSTomohiro Misono 281723d0b79dSTomohiro Misono leaf = path->nodes[0]; 281823d0b79dSTomohiro Misono slot = path->slots[0]; 281923d0b79dSTomohiro Misono 282023d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 282123d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 282223d0b79dSTomohiro Misono ptr -= len + 1; 282323d0b79dSTomohiro Misono total_len += len + 1; 282423d0b79dSTomohiro Misono if (ptr < args->path) { 282523d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 2826b8a49ae1SJosef Bacik goto out_put; 282723d0b79dSTomohiro Misono } 282823d0b79dSTomohiro Misono 282923d0b79dSTomohiro Misono *(ptr + len) = '/'; 283023d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 283123d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 283223d0b79dSTomohiro Misono 283323d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 283423d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 283523d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 283623d0b79dSTomohiro Misono if (ret < 0) { 2837b8a49ae1SJosef Bacik goto out_put; 283823d0b79dSTomohiro Misono } else if (ret > 0) { 283923d0b79dSTomohiro Misono ret = -ENOENT; 2840b8a49ae1SJosef Bacik goto out_put; 284123d0b79dSTomohiro Misono } 284223d0b79dSTomohiro Misono 284323d0b79dSTomohiro Misono leaf = path->nodes[0]; 284423d0b79dSTomohiro Misono slot = path->slots[0]; 284523d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 284623d0b79dSTomohiro Misono if (key2.objectid != dirid) { 284723d0b79dSTomohiro Misono ret = -ENOENT; 2848b8a49ae1SJosef Bacik goto out_put; 284923d0b79dSTomohiro Misono } 285023d0b79dSTomohiro Misono 28510202e83fSDavid Sterba temp_inode = btrfs_iget(sb, key2.objectid, root); 28523ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 28533ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 2854b8a49ae1SJosef Bacik goto out_put; 28553ca57bd6SMisono Tomohiro } 28566623d9a0SChristian Brauner ret = inode_permission(mnt_userns, temp_inode, 285747291baaSChristian Brauner MAY_READ | MAY_EXEC); 285823d0b79dSTomohiro Misono iput(temp_inode); 285923d0b79dSTomohiro Misono if (ret) { 286023d0b79dSTomohiro Misono ret = -EACCES; 2861b8a49ae1SJosef Bacik goto out_put; 286223d0b79dSTomohiro Misono } 286323d0b79dSTomohiro Misono 286423d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 286523d0b79dSTomohiro Misono break; 286623d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 286723d0b79dSTomohiro Misono ret = -EACCES; 2868b8a49ae1SJosef Bacik goto out_put; 286923d0b79dSTomohiro Misono } 287023d0b79dSTomohiro Misono 287123d0b79dSTomohiro Misono btrfs_release_path(path); 287223d0b79dSTomohiro Misono key.objectid = key.offset; 287323d0b79dSTomohiro Misono key.offset = (u64)-1; 287423d0b79dSTomohiro Misono dirid = key.objectid; 287523d0b79dSTomohiro Misono } 287623d0b79dSTomohiro Misono 287723d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 287823d0b79dSTomohiro Misono args->path[total_len] = '\0'; 287900246528SJosef Bacik btrfs_put_root(root); 2880b8a49ae1SJosef Bacik root = NULL; 288123d0b79dSTomohiro Misono btrfs_release_path(path); 288223d0b79dSTomohiro Misono } 288323d0b79dSTomohiro Misono 288423d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 288523d0b79dSTomohiro Misono key.objectid = treeid; 288623d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 288723d0b79dSTomohiro Misono key.offset = args->treeid; 2888b8a49ae1SJosef Bacik ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 288923d0b79dSTomohiro Misono if (ret < 0) { 289023d0b79dSTomohiro Misono goto out; 289123d0b79dSTomohiro Misono } else if (ret > 0) { 289223d0b79dSTomohiro Misono ret = -ENOENT; 289323d0b79dSTomohiro Misono goto out; 289423d0b79dSTomohiro Misono } 289523d0b79dSTomohiro Misono 289623d0b79dSTomohiro Misono leaf = path->nodes[0]; 289723d0b79dSTomohiro Misono slot = path->slots[0]; 289823d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 289923d0b79dSTomohiro Misono 290023d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 29013212fa14SJosef Bacik item_len = btrfs_item_size(leaf, slot); 290223d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 290323d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 290423d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 290523d0b79dSTomohiro Misono ret = -EINVAL; 290623d0b79dSTomohiro Misono goto out; 290723d0b79dSTomohiro Misono } 290823d0b79dSTomohiro Misono 290923d0b79dSTomohiro Misono /* Copy subvolume's name */ 291023d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 291123d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 291223d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 291323d0b79dSTomohiro Misono args->name[item_len] = 0; 291423d0b79dSTomohiro Misono 2915b8a49ae1SJosef Bacik out_put: 291600246528SJosef Bacik btrfs_put_root(root); 291723d0b79dSTomohiro Misono out: 291823d0b79dSTomohiro Misono btrfs_free_path(path); 291923d0b79dSTomohiro Misono return ret; 292023d0b79dSTomohiro Misono } 292123d0b79dSTomohiro Misono 2922dc408ccdSSahil Kang static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root, 2923ac8e9819SChris Mason void __user *argp) 2924ac8e9819SChris Mason { 2925ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 292601b810b8SDavid Sterba int ret = 0; 2927ac8e9819SChris Mason 29282354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 29292354d08fSJulia Lawall if (IS_ERR(args)) 29302354d08fSJulia Lawall return PTR_ERR(args); 2931c2b96929SDan Carpenter 293201b810b8SDavid Sterba /* 293301b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 293401b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 293501b810b8SDavid Sterba */ 29361b53ac4dSChris Mason if (args->treeid == 0) 2937dc408ccdSSahil Kang args->treeid = root->root_key.objectid; 29381b53ac4dSChris Mason 293901b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 294001b810b8SDavid Sterba args->name[0] = 0; 294101b810b8SDavid Sterba goto out; 294201b810b8SDavid Sterba } 294301b810b8SDavid Sterba 294401b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 294501b810b8SDavid Sterba ret = -EPERM; 294601b810b8SDavid Sterba goto out; 294701b810b8SDavid Sterba } 294801b810b8SDavid Sterba 2949dc408ccdSSahil Kang ret = btrfs_search_path_in_tree(root->fs_info, 2950ac8e9819SChris Mason args->treeid, args->objectid, 2951ac8e9819SChris Mason args->name); 2952ac8e9819SChris Mason 295301b810b8SDavid Sterba out: 2954ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2955ac8e9819SChris Mason ret = -EFAULT; 2956ac8e9819SChris Mason 2957ac8e9819SChris Mason kfree(args); 295898d377a0STARUISI Hiroaki return ret; 295998d377a0STARUISI Hiroaki } 296098d377a0STARUISI Hiroaki 296123d0b79dSTomohiro Misono /* 296223d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 296323d0b79dSTomohiro Misono * 296423d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 296523d0b79dSTomohiro Misono * 296623d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 296723d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 296823d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 296923d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 297023d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 297123d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 297223d0b79dSTomohiro Misono */ 297323d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 297423d0b79dSTomohiro Misono { 297523d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 297623d0b79dSTomohiro Misono struct inode *inode; 297723d0b79dSTomohiro Misono int ret; 297823d0b79dSTomohiro Misono 297923d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 298023d0b79dSTomohiro Misono if (IS_ERR(args)) 298123d0b79dSTomohiro Misono return PTR_ERR(args); 298223d0b79dSTomohiro Misono 298323d0b79dSTomohiro Misono inode = file_inode(file); 298423d0b79dSTomohiro Misono 298523d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 298623d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 298723d0b79dSTomohiro Misono /* 298823d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 298923d0b79dSTomohiro Misono * called 299023d0b79dSTomohiro Misono */ 299123d0b79dSTomohiro Misono kfree(args); 299223d0b79dSTomohiro Misono return -EACCES; 299323d0b79dSTomohiro Misono } 299423d0b79dSTomohiro Misono 29956623d9a0SChristian Brauner ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args); 299623d0b79dSTomohiro Misono 299723d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 299823d0b79dSTomohiro Misono ret = -EFAULT; 299923d0b79dSTomohiro Misono 300023d0b79dSTomohiro Misono kfree(args); 300123d0b79dSTomohiro Misono return ret; 300223d0b79dSTomohiro Misono } 300323d0b79dSTomohiro Misono 3004b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 30059ad12305SSahil Kang static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp) 3006b64ec075STomohiro Misono { 3007b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 3008b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 3009b64ec075STomohiro Misono struct btrfs_root *root; 3010b64ec075STomohiro Misono struct btrfs_path *path; 3011b64ec075STomohiro Misono struct btrfs_key key; 3012b64ec075STomohiro Misono struct btrfs_root_item *root_item; 3013b64ec075STomohiro Misono struct btrfs_root_ref *rref; 3014b64ec075STomohiro Misono struct extent_buffer *leaf; 3015b64ec075STomohiro Misono unsigned long item_off; 3016b64ec075STomohiro Misono unsigned long item_len; 3017b64ec075STomohiro Misono int slot; 3018b64ec075STomohiro Misono int ret = 0; 3019b64ec075STomohiro Misono 3020b64ec075STomohiro Misono path = btrfs_alloc_path(); 3021b64ec075STomohiro Misono if (!path) 3022b64ec075STomohiro Misono return -ENOMEM; 3023b64ec075STomohiro Misono 3024b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 3025b64ec075STomohiro Misono if (!subvol_info) { 3026b64ec075STomohiro Misono btrfs_free_path(path); 3027b64ec075STomohiro Misono return -ENOMEM; 3028b64ec075STomohiro Misono } 3029b64ec075STomohiro Misono 3030b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 3031b64ec075STomohiro Misono 3032b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 3033b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 303456e9357aSDavid Sterba root = btrfs_get_fs_root(fs_info, key.objectid, true); 3035b64ec075STomohiro Misono if (IS_ERR(root)) { 3036b64ec075STomohiro Misono ret = PTR_ERR(root); 303704734e84SJosef Bacik goto out_free; 303804734e84SJosef Bacik } 3039b64ec075STomohiro Misono root_item = &root->root_item; 3040b64ec075STomohiro Misono 3041b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 3042b64ec075STomohiro Misono 3043b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 3044b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 3045b64ec075STomohiro Misono 3046b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 3047b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 3048b64ec075STomohiro Misono BTRFS_UUID_SIZE); 3049b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 3050b64ec075STomohiro Misono BTRFS_UUID_SIZE); 3051b64ec075STomohiro Misono 3052b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 3053b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 3054b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 3055b64ec075STomohiro Misono 3056b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 3057b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 3058b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 3059b64ec075STomohiro Misono 3060b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 3061b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 3062b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 3063b64ec075STomohiro Misono 3064b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 3065b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 3066b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 3067b64ec075STomohiro Misono 3068b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 3069b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 3070b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 3071b64ec075STomohiro Misono key.offset = 0; 307204734e84SJosef Bacik ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 3073b64ec075STomohiro Misono if (ret < 0) { 3074b64ec075STomohiro Misono goto out; 3075b64ec075STomohiro Misono } else if (path->slots[0] >= 3076b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 307704734e84SJosef Bacik ret = btrfs_next_leaf(fs_info->tree_root, path); 3078b64ec075STomohiro Misono if (ret < 0) { 3079b64ec075STomohiro Misono goto out; 3080b64ec075STomohiro Misono } else if (ret > 0) { 3081b64ec075STomohiro Misono ret = -EUCLEAN; 3082b64ec075STomohiro Misono goto out; 3083b64ec075STomohiro Misono } 3084b64ec075STomohiro Misono } 3085b64ec075STomohiro Misono 3086b64ec075STomohiro Misono leaf = path->nodes[0]; 3087b64ec075STomohiro Misono slot = path->slots[0]; 3088b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 3089b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 3090b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 3091b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 3092b64ec075STomohiro Misono 3093b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 3094b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 3095b64ec075STomohiro Misono 3096b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 3097b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 30983212fa14SJosef Bacik item_len = btrfs_item_size(leaf, slot) 3099b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 3100b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 3101b64ec075STomohiro Misono item_off, item_len); 3102b64ec075STomohiro Misono } else { 3103b64ec075STomohiro Misono ret = -ENOENT; 3104b64ec075STomohiro Misono goto out; 3105b64ec075STomohiro Misono } 3106b64ec075STomohiro Misono } 3107b64ec075STomohiro Misono 3108b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 3109b64ec075STomohiro Misono ret = -EFAULT; 3110b64ec075STomohiro Misono 3111b64ec075STomohiro Misono out: 311200246528SJosef Bacik btrfs_put_root(root); 311304734e84SJosef Bacik out_free: 3114b64ec075STomohiro Misono btrfs_free_path(path); 3115b091f7feSWaiman Long kfree(subvol_info); 3116b64ec075STomohiro Misono return ret; 3117b64ec075STomohiro Misono } 3118b64ec075STomohiro Misono 311942e4b520STomohiro Misono /* 312042e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 312142e4b520STomohiro Misono * except the subvolume name. 312242e4b520STomohiro Misono */ 31239ad12305SSahil Kang static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root, 3124dc408ccdSSahil Kang void __user *argp) 312542e4b520STomohiro Misono { 312642e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 312742e4b520STomohiro Misono struct btrfs_root_ref *rref; 312842e4b520STomohiro Misono struct btrfs_path *path; 312942e4b520STomohiro Misono struct btrfs_key key; 313042e4b520STomohiro Misono struct extent_buffer *leaf; 313142e4b520STomohiro Misono u64 objectid; 313242e4b520STomohiro Misono int slot; 313342e4b520STomohiro Misono int ret; 313442e4b520STomohiro Misono u8 found; 313542e4b520STomohiro Misono 313642e4b520STomohiro Misono path = btrfs_alloc_path(); 313742e4b520STomohiro Misono if (!path) 313842e4b520STomohiro Misono return -ENOMEM; 313942e4b520STomohiro Misono 314042e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 314142e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 314242e4b520STomohiro Misono btrfs_free_path(path); 314342e4b520STomohiro Misono return PTR_ERR(rootrefs); 314442e4b520STomohiro Misono } 314542e4b520STomohiro Misono 31469ad12305SSahil Kang objectid = root->root_key.objectid; 314742e4b520STomohiro Misono key.objectid = objectid; 314842e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 314942e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 315042e4b520STomohiro Misono found = 0; 315142e4b520STomohiro Misono 31529ad12305SSahil Kang root = root->fs_info->tree_root; 315342e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 315442e4b520STomohiro Misono if (ret < 0) { 315542e4b520STomohiro Misono goto out; 315642e4b520STomohiro Misono } else if (path->slots[0] >= 315742e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 315842e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 315942e4b520STomohiro Misono if (ret < 0) { 316042e4b520STomohiro Misono goto out; 316142e4b520STomohiro Misono } else if (ret > 0) { 316242e4b520STomohiro Misono ret = -EUCLEAN; 316342e4b520STomohiro Misono goto out; 316442e4b520STomohiro Misono } 316542e4b520STomohiro Misono } 316642e4b520STomohiro Misono while (1) { 316742e4b520STomohiro Misono leaf = path->nodes[0]; 316842e4b520STomohiro Misono slot = path->slots[0]; 316942e4b520STomohiro Misono 317042e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 317142e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 317242e4b520STomohiro Misono ret = 0; 317342e4b520STomohiro Misono goto out; 317442e4b520STomohiro Misono } 317542e4b520STomohiro Misono 317642e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 317742e4b520STomohiro Misono ret = -EOVERFLOW; 317842e4b520STomohiro Misono goto out; 317942e4b520STomohiro Misono } 318042e4b520STomohiro Misono 318142e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 318242e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 318342e4b520STomohiro Misono rootrefs->rootref[found].dirid = 318442e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 318542e4b520STomohiro Misono found++; 318642e4b520STomohiro Misono 318742e4b520STomohiro Misono ret = btrfs_next_item(root, path); 318842e4b520STomohiro Misono if (ret < 0) { 318942e4b520STomohiro Misono goto out; 319042e4b520STomohiro Misono } else if (ret > 0) { 319142e4b520STomohiro Misono ret = -EUCLEAN; 319242e4b520STomohiro Misono goto out; 319342e4b520STomohiro Misono } 319442e4b520STomohiro Misono } 319542e4b520STomohiro Misono 319642e4b520STomohiro Misono out: 319742e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 319842e4b520STomohiro Misono rootrefs->num_items = found; 319942e4b520STomohiro Misono /* update min_treeid for next search */ 320042e4b520STomohiro Misono if (found) 320142e4b520STomohiro Misono rootrefs->min_treeid = 320242e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 320342e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 320442e4b520STomohiro Misono ret = -EFAULT; 320542e4b520STomohiro Misono } 320642e4b520STomohiro Misono 320742e4b520STomohiro Misono kfree(rootrefs); 320842e4b520STomohiro Misono btrfs_free_path(path); 320942e4b520STomohiro Misono 321042e4b520STomohiro Misono return ret; 321142e4b520STomohiro Misono } 321242e4b520STomohiro Misono 321376dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 3214949964c9SMarcos Paulo de Souza void __user *arg, 3215949964c9SMarcos Paulo de Souza bool destroy_v2) 321676dda93cSYan, Zheng { 321754563d41SAl Viro struct dentry *parent = file->f_path.dentry; 32180b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 321976dda93cSYan, Zheng struct dentry *dentry; 32202b0143b5SDavid Howells struct inode *dir = d_inode(parent); 322176dda93cSYan, Zheng struct inode *inode; 322276dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 322376dda93cSYan, Zheng struct btrfs_root *dest = NULL; 3224949964c9SMarcos Paulo de Souza struct btrfs_ioctl_vol_args *vol_args = NULL; 3225949964c9SMarcos Paulo de Souza struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL; 3226c4ed533bSChristian Brauner struct user_namespace *mnt_userns = file_mnt_user_ns(file); 3227949964c9SMarcos Paulo de Souza char *subvol_name, *subvol_name_ptr = NULL; 3228949964c9SMarcos Paulo de Souza int subvol_namelen; 322976dda93cSYan, Zheng int err = 0; 3230949964c9SMarcos Paulo de Souza bool destroy_parent = false; 323176dda93cSYan, Zheng 3232813febdbSJosef Bacik /* We don't support snapshots with extent tree v2 yet. */ 3233813febdbSJosef Bacik if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 3234813febdbSJosef Bacik btrfs_err(fs_info, 3235813febdbSJosef Bacik "extent tree v2 doesn't support snapshot deletion yet"); 3236813febdbSJosef Bacik return -EOPNOTSUPP; 3237813febdbSJosef Bacik } 3238813febdbSJosef Bacik 3239949964c9SMarcos Paulo de Souza if (destroy_v2) { 3240949964c9SMarcos Paulo de Souza vol_args2 = memdup_user(arg, sizeof(*vol_args2)); 3241949964c9SMarcos Paulo de Souza if (IS_ERR(vol_args2)) 3242949964c9SMarcos Paulo de Souza return PTR_ERR(vol_args2); 3243325c50e3SJeff Mahoney 3244949964c9SMarcos Paulo de Souza if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) { 3245949964c9SMarcos Paulo de Souza err = -EOPNOTSUPP; 3246949964c9SMarcos Paulo de Souza goto out; 3247949964c9SMarcos Paulo de Souza } 324876dda93cSYan, Zheng 3249949964c9SMarcos Paulo de Souza /* 3250949964c9SMarcos Paulo de Souza * If SPEC_BY_ID is not set, we are looking for the subvolume by 3251949964c9SMarcos Paulo de Souza * name, same as v1 currently does. 3252949964c9SMarcos Paulo de Souza */ 3253949964c9SMarcos Paulo de Souza if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { 3254949964c9SMarcos Paulo de Souza vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0; 3255949964c9SMarcos Paulo de Souza subvol_name = vol_args2->name; 3256949964c9SMarcos Paulo de Souza 3257949964c9SMarcos Paulo de Souza err = mnt_want_write_file(file); 3258949964c9SMarcos Paulo de Souza if (err) 3259949964c9SMarcos Paulo de Souza goto out; 3260949964c9SMarcos Paulo de Souza } else { 3261aabb34e7SChristian Brauner struct inode *old_dir; 3262c4ed533bSChristian Brauner 3263949964c9SMarcos Paulo de Souza if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) { 326476dda93cSYan, Zheng err = -EINVAL; 326576dda93cSYan, Zheng goto out; 326676dda93cSYan, Zheng } 326776dda93cSYan, Zheng 3268a561be71SAl Viro err = mnt_want_write_file(file); 326976dda93cSYan, Zheng if (err) 327076dda93cSYan, Zheng goto out; 327176dda93cSYan, Zheng 3272949964c9SMarcos Paulo de Souza dentry = btrfs_get_dentry(fs_info->sb, 3273949964c9SMarcos Paulo de Souza BTRFS_FIRST_FREE_OBJECTID, 3274949964c9SMarcos Paulo de Souza vol_args2->subvolid, 0, 0); 3275949964c9SMarcos Paulo de Souza if (IS_ERR(dentry)) { 3276949964c9SMarcos Paulo de Souza err = PTR_ERR(dentry); 3277949964c9SMarcos Paulo de Souza goto out_drop_write; 3278949964c9SMarcos Paulo de Souza } 3279949964c9SMarcos Paulo de Souza 3280949964c9SMarcos Paulo de Souza /* 3281949964c9SMarcos Paulo de Souza * Change the default parent since the subvolume being 3282949964c9SMarcos Paulo de Souza * deleted can be outside of the current mount point. 3283949964c9SMarcos Paulo de Souza */ 3284949964c9SMarcos Paulo de Souza parent = btrfs_get_parent(dentry); 3285949964c9SMarcos Paulo de Souza 3286949964c9SMarcos Paulo de Souza /* 3287949964c9SMarcos Paulo de Souza * At this point dentry->d_name can point to '/' if the 3288949964c9SMarcos Paulo de Souza * subvolume we want to destroy is outsite of the 3289949964c9SMarcos Paulo de Souza * current mount point, so we need to release the 3290949964c9SMarcos Paulo de Souza * current dentry and execute the lookup to return a new 3291949964c9SMarcos Paulo de Souza * one with ->d_name pointing to the 3292949964c9SMarcos Paulo de Souza * <mount point>/subvol_name. 3293949964c9SMarcos Paulo de Souza */ 3294949964c9SMarcos Paulo de Souza dput(dentry); 3295949964c9SMarcos Paulo de Souza if (IS_ERR(parent)) { 3296949964c9SMarcos Paulo de Souza err = PTR_ERR(parent); 3297949964c9SMarcos Paulo de Souza goto out_drop_write; 3298949964c9SMarcos Paulo de Souza } 3299aabb34e7SChristian Brauner old_dir = dir; 3300949964c9SMarcos Paulo de Souza dir = d_inode(parent); 3301949964c9SMarcos Paulo de Souza 3302949964c9SMarcos Paulo de Souza /* 3303949964c9SMarcos Paulo de Souza * If v2 was used with SPEC_BY_ID, a new parent was 3304949964c9SMarcos Paulo de Souza * allocated since the subvolume can be outside of the 3305949964c9SMarcos Paulo de Souza * current mount point. Later on we need to release this 3306949964c9SMarcos Paulo de Souza * new parent dentry. 3307949964c9SMarcos Paulo de Souza */ 3308949964c9SMarcos Paulo de Souza destroy_parent = true; 3309949964c9SMarcos Paulo de Souza 3310aabb34e7SChristian Brauner /* 3311aabb34e7SChristian Brauner * On idmapped mounts, deletion via subvolid is 3312aabb34e7SChristian Brauner * restricted to subvolumes that are immediate 3313aabb34e7SChristian Brauner * ancestors of the inode referenced by the file 3314aabb34e7SChristian Brauner * descriptor in the ioctl. Otherwise the idmapping 3315aabb34e7SChristian Brauner * could potentially be abused to delete subvolumes 3316aabb34e7SChristian Brauner * anywhere in the filesystem the user wouldn't be able 3317aabb34e7SChristian Brauner * to delete without an idmapped mount. 3318aabb34e7SChristian Brauner */ 3319aabb34e7SChristian Brauner if (old_dir != dir && mnt_userns != &init_user_ns) { 3320aabb34e7SChristian Brauner err = -EOPNOTSUPP; 3321aabb34e7SChristian Brauner goto free_parent; 3322aabb34e7SChristian Brauner } 3323aabb34e7SChristian Brauner 3324949964c9SMarcos Paulo de Souza subvol_name_ptr = btrfs_get_subvol_name_from_objectid( 3325949964c9SMarcos Paulo de Souza fs_info, vol_args2->subvolid); 3326949964c9SMarcos Paulo de Souza if (IS_ERR(subvol_name_ptr)) { 3327949964c9SMarcos Paulo de Souza err = PTR_ERR(subvol_name_ptr); 3328949964c9SMarcos Paulo de Souza goto free_parent; 3329949964c9SMarcos Paulo de Souza } 33301a9fd417SDavid Sterba /* subvol_name_ptr is already nul terminated */ 3331949964c9SMarcos Paulo de Souza subvol_name = (char *)kbasename(subvol_name_ptr); 3332949964c9SMarcos Paulo de Souza } 3333949964c9SMarcos Paulo de Souza } else { 3334949964c9SMarcos Paulo de Souza vol_args = memdup_user(arg, sizeof(*vol_args)); 3335949964c9SMarcos Paulo de Souza if (IS_ERR(vol_args)) 3336949964c9SMarcos Paulo de Souza return PTR_ERR(vol_args); 3337949964c9SMarcos Paulo de Souza 3338949964c9SMarcos Paulo de Souza vol_args->name[BTRFS_PATH_NAME_MAX] = 0; 3339949964c9SMarcos Paulo de Souza subvol_name = vol_args->name; 3340949964c9SMarcos Paulo de Souza 3341949964c9SMarcos Paulo de Souza err = mnt_want_write_file(file); 3342949964c9SMarcos Paulo de Souza if (err) 3343949964c9SMarcos Paulo de Souza goto out; 3344949964c9SMarcos Paulo de Souza } 3345949964c9SMarcos Paulo de Souza 3346949964c9SMarcos Paulo de Souza subvol_namelen = strlen(subvol_name); 3347949964c9SMarcos Paulo de Souza 3348949964c9SMarcos Paulo de Souza if (strchr(subvol_name, '/') || 3349949964c9SMarcos Paulo de Souza strncmp(subvol_name, "..", subvol_namelen) == 0) { 3350949964c9SMarcos Paulo de Souza err = -EINVAL; 3351949964c9SMarcos Paulo de Souza goto free_subvol_name; 3352949964c9SMarcos Paulo de Souza } 3353949964c9SMarcos Paulo de Souza 3354949964c9SMarcos Paulo de Souza if (!S_ISDIR(dir->i_mode)) { 3355949964c9SMarcos Paulo de Souza err = -ENOTDIR; 3356949964c9SMarcos Paulo de Souza goto free_subvol_name; 3357949964c9SMarcos Paulo de Souza } 3358521e0546SDavid Sterba 335900235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 336000235411SAl Viro if (err == -EINTR) 3361949964c9SMarcos Paulo de Souza goto free_subvol_name; 3362c4ed533bSChristian Brauner dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen); 336376dda93cSYan, Zheng if (IS_ERR(dentry)) { 336476dda93cSYan, Zheng err = PTR_ERR(dentry); 336576dda93cSYan, Zheng goto out_unlock_dir; 336676dda93cSYan, Zheng } 336776dda93cSYan, Zheng 33682b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 336976dda93cSYan, Zheng err = -ENOENT; 337076dda93cSYan, Zheng goto out_dput; 337176dda93cSYan, Zheng } 337276dda93cSYan, Zheng 33732b0143b5SDavid Howells inode = d_inode(dentry); 33744260f7c7SSage Weil dest = BTRFS_I(inode)->root; 33754260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 33764260f7c7SSage Weil /* 33774260f7c7SSage Weil * Regular user. Only allow this with a special mount 33784260f7c7SSage Weil * option, when the user has write+exec access to the 33794260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 33804260f7c7SSage Weil * allowed. 33814260f7c7SSage Weil * 33824260f7c7SSage Weil * Note that this is _not_ check that the subvol is 33834260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 33844260f7c7SSage Weil * otherwise be able to delete. 33854260f7c7SSage Weil * 33864260f7c7SSage Weil * Users who want to delete empty subvols should try 33874260f7c7SSage Weil * rmdir(2). 33884260f7c7SSage Weil */ 33894260f7c7SSage Weil err = -EPERM; 33900b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 33914260f7c7SSage Weil goto out_dput; 33924260f7c7SSage Weil 33934260f7c7SSage Weil /* 33944260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 33954260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 33964260f7c7SSage Weil * must be called on the dentry referencing the root 33974260f7c7SSage Weil * of the subvol, not a random directory contained 33984260f7c7SSage Weil * within it. 33994260f7c7SSage Weil */ 34004260f7c7SSage Weil err = -EINVAL; 34014260f7c7SSage Weil if (root == dest) 34024260f7c7SSage Weil goto out_dput; 34034260f7c7SSage Weil 3404c4ed533bSChristian Brauner err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC); 34054260f7c7SSage Weil if (err) 34064260f7c7SSage Weil goto out_dput; 34075c39da5bSMiao Xie } 34084260f7c7SSage Weil 34095c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 3410c4ed533bSChristian Brauner err = btrfs_may_delete(mnt_userns, dir, dentry, 1); 34114260f7c7SSage Weil if (err) 34124260f7c7SSage Weil goto out_dput; 34134260f7c7SSage Weil 34144a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 341576dda93cSYan, Zheng err = -EINVAL; 341676dda93cSYan, Zheng goto out_dput; 341776dda93cSYan, Zheng } 341876dda93cSYan, Zheng 341964708539SJosef Bacik btrfs_inode_lock(inode, 0); 3420f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 342164708539SJosef Bacik btrfs_inode_unlock(inode, 0); 3422a37d9a17SAmir Goldstein if (!err) 3423a37d9a17SAmir Goldstein d_delete_notify(dir, dentry); 3424fa6ac876SLiu Bo 342576dda93cSYan, Zheng out_dput: 342676dda93cSYan, Zheng dput(dentry); 342776dda93cSYan, Zheng out_unlock_dir: 342864708539SJosef Bacik btrfs_inode_unlock(dir, 0); 3429949964c9SMarcos Paulo de Souza free_subvol_name: 3430949964c9SMarcos Paulo de Souza kfree(subvol_name_ptr); 3431949964c9SMarcos Paulo de Souza free_parent: 3432949964c9SMarcos Paulo de Souza if (destroy_parent) 3433949964c9SMarcos Paulo de Souza dput(parent); 343400235411SAl Viro out_drop_write: 34352a79f17eSAl Viro mnt_drop_write_file(file); 343676dda93cSYan, Zheng out: 3437949964c9SMarcos Paulo de Souza kfree(vol_args2); 343876dda93cSYan, Zheng kfree(vol_args); 343976dda93cSYan, Zheng return err; 344076dda93cSYan, Zheng } 344176dda93cSYan, Zheng 34421e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 3443f46b5a66SChristoph Hellwig { 3444496ad9aaSAl Viro struct inode *inode = file_inode(file); 3445f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3446c853a578SGoldwyn Rodrigues struct btrfs_ioctl_defrag_range_args range = {0}; 3447c146afadSYan Zheng int ret; 3448c146afadSYan Zheng 344925122d15SIlya Dryomov ret = mnt_want_write_file(file); 345025122d15SIlya Dryomov if (ret) 345125122d15SIlya Dryomov return ret; 3452b83cc969SLi Zefan 345325122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 345425122d15SIlya Dryomov ret = -EROFS; 345525122d15SIlya Dryomov goto out; 34565ac00addSStefan Behrens } 3457f46b5a66SChristoph Hellwig 3458f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 3459f46b5a66SChristoph Hellwig case S_IFDIR: 3460e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 3461e441d54dSChris Mason ret = -EPERM; 3462e441d54dSChris Mason goto out; 3463e441d54dSChris Mason } 3464de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 3465f46b5a66SChristoph Hellwig break; 3466f46b5a66SChristoph Hellwig case S_IFREG: 3467616d374eSAdam Borowski /* 3468616d374eSAdam Borowski * Note that this does not check the file descriptor for write 3469616d374eSAdam Borowski * access. This prevents defragmenting executables that are 3470616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 3471616d374eSAdam Borowski */ 3472616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 347347291baaSChristian Brauner inode_permission(&init_user_ns, inode, MAY_WRITE)) { 3474616d374eSAdam Borowski ret = -EPERM; 3475e441d54dSChris Mason goto out; 3476e441d54dSChris Mason } 34771e701a32SChris Mason 34781e701a32SChris Mason if (argp) { 3479c853a578SGoldwyn Rodrigues if (copy_from_user(&range, argp, sizeof(range))) { 34801e701a32SChris Mason ret = -EFAULT; 3481683be16eSDan Carpenter goto out; 34821e701a32SChris Mason } 34831e701a32SChris Mason /* compression requires us to start the IO */ 3484c853a578SGoldwyn Rodrigues if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 3485c853a578SGoldwyn Rodrigues range.flags |= BTRFS_DEFRAG_RANGE_START_IO; 3486c853a578SGoldwyn Rodrigues range.extent_thresh = (u32)-1; 34871e701a32SChris Mason } 34881e701a32SChris Mason } else { 34891e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 3490c853a578SGoldwyn Rodrigues range.len = (u64)-1; 34911e701a32SChris Mason } 34921ccc2e8aSQu Wenruo ret = btrfs_defrag_file(file_inode(file), &file->f_ra, 3493c853a578SGoldwyn Rodrigues &range, BTRFS_OLDEST_GENERATION, 0); 34944cb5300bSChris Mason if (ret > 0) 34954cb5300bSChris Mason ret = 0; 3496f46b5a66SChristoph Hellwig break; 34978929ecfaSYan, Zheng default: 34988929ecfaSYan, Zheng ret = -EINVAL; 3499f46b5a66SChristoph Hellwig } 3500e441d54dSChris Mason out: 350125122d15SIlya Dryomov mnt_drop_write_file(file); 3502e441d54dSChris Mason return ret; 3503f46b5a66SChristoph Hellwig } 3504f46b5a66SChristoph Hellwig 35052ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3506f46b5a66SChristoph Hellwig { 3507f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3508a174c0a2SNikolay Borisov bool restore_op = false; 3509f46b5a66SChristoph Hellwig int ret; 3510f46b5a66SChristoph Hellwig 3511e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3512e441d54dSChris Mason return -EPERM; 3513e441d54dSChris Mason 3514914a519bSJosef Bacik if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 3515914a519bSJosef Bacik btrfs_err(fs_info, "device add not supported on extent tree v2 yet"); 3516914a519bSJosef Bacik return -EINVAL; 3517914a519bSJosef Bacik } 3518914a519bSJosef Bacik 3519a174c0a2SNikolay Borisov if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) { 3520a174c0a2SNikolay Borisov if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD)) 3521e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3522c9e9f97bSIlya Dryomov 3523a174c0a2SNikolay Borisov /* 3524a174c0a2SNikolay Borisov * We can do the device add because we have a paused balanced, 3525a174c0a2SNikolay Borisov * change the exclusive op type and remember we should bring 3526a174c0a2SNikolay Borisov * back the paused balance 3527a174c0a2SNikolay Borisov */ 3528a174c0a2SNikolay Borisov fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD; 3529a174c0a2SNikolay Borisov btrfs_exclop_start_unlock(fs_info); 3530a174c0a2SNikolay Borisov restore_op = true; 3531a174c0a2SNikolay Borisov } 3532a174c0a2SNikolay Borisov 3533dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3534c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3535c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3536c9e9f97bSIlya Dryomov goto out; 3537c9e9f97bSIlya Dryomov } 3538f46b5a66SChristoph Hellwig 35395516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 35402ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3541f46b5a66SChristoph Hellwig 354243d20761SAnand Jain if (!ret) 35430b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 354443d20761SAnand Jain 3545f46b5a66SChristoph Hellwig kfree(vol_args); 3546c9e9f97bSIlya Dryomov out: 3547a174c0a2SNikolay Borisov if (restore_op) 3548a174c0a2SNikolay Borisov btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED); 3549a174c0a2SNikolay Borisov else 3550c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 3551f46b5a66SChristoph Hellwig return ret; 3552f46b5a66SChristoph Hellwig } 3553f46b5a66SChristoph Hellwig 35546b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3555f46b5a66SChristoph Hellwig { 35561a15eb72SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 35570b246afaSJeff Mahoney struct inode *inode = file_inode(file); 35580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 35596b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 35603fa421deSJosef Bacik struct block_device *bdev = NULL; 35613fa421deSJosef Bacik fmode_t mode; 3562f46b5a66SChristoph Hellwig int ret; 356367ae34b6SDavid Sterba bool cancel = false; 3564f46b5a66SChristoph Hellwig 3565e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3566e441d54dSChris Mason return -EPERM; 3567e441d54dSChris Mason 3568dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3569d815b3f2SDan Carpenter if (IS_ERR(vol_args)) 3570d815b3f2SDan Carpenter return PTR_ERR(vol_args); 3571f46b5a66SChristoph Hellwig 3572748449cdSDavid Sterba if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { 3573fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3574fd4e994bSOmar Sandoval goto out; 3575fd4e994bSOmar Sandoval } 35761a15eb72SJosef Bacik 35776b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 35781a15eb72SJosef Bacik if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 35791a15eb72SJosef Bacik args.devid = vol_args->devid; 35801a15eb72SJosef Bacik } else if (!strcmp("cancel", vol_args->name)) { 358167ae34b6SDavid Sterba cancel = true; 35821a15eb72SJosef Bacik } else { 35831a15eb72SJosef Bacik ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 35841a15eb72SJosef Bacik if (ret) 35851a15eb72SJosef Bacik goto out; 35861a15eb72SJosef Bacik } 35871a15eb72SJosef Bacik 35881a15eb72SJosef Bacik ret = mnt_want_write_file(file); 35891a15eb72SJosef Bacik if (ret) 35901a15eb72SJosef Bacik goto out; 359167ae34b6SDavid Sterba 359267ae34b6SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 359367ae34b6SDavid Sterba cancel); 359467ae34b6SDavid Sterba if (ret) 35951a15eb72SJosef Bacik goto err_drop; 359667ae34b6SDavid Sterba 35971a15eb72SJosef Bacik /* Exclusive operation is now claimed */ 35981a15eb72SJosef Bacik ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); 359967ae34b6SDavid Sterba 3600c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 3601183860f6SAnand Jain 36026b526ed7SAnand Jain if (!ret) { 3603735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 36040b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 36056b526ed7SAnand Jain vol_args->devid); 36066b526ed7SAnand Jain else 36070b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 36086b526ed7SAnand Jain vol_args->name); 36096b526ed7SAnand Jain } 3610c47ca32dSDan Carpenter err_drop: 36114ac20c70SIlya Dryomov mnt_drop_write_file(file); 36123fa421deSJosef Bacik if (bdev) 36133fa421deSJosef Bacik blkdev_put(bdev, mode); 36141a15eb72SJosef Bacik out: 36151a15eb72SJosef Bacik btrfs_put_dev_args_from_path(&args); 36161a15eb72SJosef Bacik kfree(vol_args); 3617f46b5a66SChristoph Hellwig return ret; 3618f46b5a66SChristoph Hellwig } 3619f46b5a66SChristoph Hellwig 3620f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3621f46b5a66SChristoph Hellwig { 36221a15eb72SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 36230b246afaSJeff Mahoney struct inode *inode = file_inode(file); 36240b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3625f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 36263fa421deSJosef Bacik struct block_device *bdev = NULL; 36273fa421deSJosef Bacik fmode_t mode; 3628f46b5a66SChristoph Hellwig int ret; 362937b45995STom Rix bool cancel = false; 3630f46b5a66SChristoph Hellwig 3631f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3632f46b5a66SChristoph Hellwig return -EPERM; 3633f46b5a66SChristoph Hellwig 36341a15eb72SJosef Bacik vol_args = memdup_user(arg, sizeof(*vol_args)); 36351a15eb72SJosef Bacik if (IS_ERR(vol_args)) 36361a15eb72SJosef Bacik return PTR_ERR(vol_args); 36371a15eb72SJosef Bacik 36381a15eb72SJosef Bacik vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 36391a15eb72SJosef Bacik if (!strcmp("cancel", vol_args->name)) { 36401a15eb72SJosef Bacik cancel = true; 36411a15eb72SJosef Bacik } else { 36421a15eb72SJosef Bacik ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 36431a15eb72SJosef Bacik if (ret) 36441a15eb72SJosef Bacik goto out; 36451a15eb72SJosef Bacik } 36461a15eb72SJosef Bacik 3647f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3648f46b5a66SChristoph Hellwig if (ret) 36491a15eb72SJosef Bacik goto out; 3650f46b5a66SChristoph Hellwig 365167ae34b6SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 365267ae34b6SDavid Sterba cancel); 365367ae34b6SDavid Sterba if (ret == 0) { 36541a15eb72SJosef Bacik ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); 3655f46b5a66SChristoph Hellwig if (!ret) 36560b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3657c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 365867ae34b6SDavid Sterba } 365967ae34b6SDavid Sterba 3660f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 36613fa421deSJosef Bacik if (bdev) 36623fa421deSJosef Bacik blkdev_put(bdev, mode); 36631a15eb72SJosef Bacik out: 36641a15eb72SJosef Bacik btrfs_put_dev_args_from_path(&args); 36651a15eb72SJosef Bacik kfree(vol_args); 3666f46b5a66SChristoph Hellwig return ret; 3667f46b5a66SChristoph Hellwig } 3668f46b5a66SChristoph Hellwig 36692ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 36702ff7e61eSJeff Mahoney void __user *arg) 3671475f6387SJan Schmidt { 3672027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3673475f6387SJan Schmidt struct btrfs_device *device; 36740b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3675137c5418SJohannes Thumshirn u64 flags_in; 3676027ed2f0SLi Zefan int ret = 0; 3677475f6387SJan Schmidt 3678137c5418SJohannes Thumshirn fi_args = memdup_user(arg, sizeof(*fi_args)); 3679137c5418SJohannes Thumshirn if (IS_ERR(fi_args)) 3680137c5418SJohannes Thumshirn return PTR_ERR(fi_args); 3681137c5418SJohannes Thumshirn 3682137c5418SJohannes Thumshirn flags_in = fi_args->flags; 3683137c5418SJohannes Thumshirn memset(fi_args, 0, sizeof(*fi_args)); 3684027ed2f0SLi Zefan 3685d03262c7SDavid Sterba rcu_read_lock(); 3686027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3687475f6387SJan Schmidt 3688d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3689027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3690027ed2f0SLi Zefan fi_args->max_id = device->devid; 3691475f6387SJan Schmidt } 3692d03262c7SDavid Sterba rcu_read_unlock(); 3693475f6387SJan Schmidt 3694de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3695bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3696bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3697bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 369880a773fbSDavid Sterba 3699137c5418SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) { 3700137c5418SJohannes Thumshirn fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy); 3701137c5418SJohannes Thumshirn fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy); 3702137c5418SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO; 3703137c5418SJohannes Thumshirn } 3704137c5418SJohannes Thumshirn 37050fb408a5SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) { 37060fb408a5SJohannes Thumshirn fi_args->generation = fs_info->generation; 37070fb408a5SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION; 37080fb408a5SJohannes Thumshirn } 37090fb408a5SJohannes Thumshirn 371049bac897SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) { 371149bac897SJohannes Thumshirn memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid, 371249bac897SJohannes Thumshirn sizeof(fi_args->metadata_uuid)); 371349bac897SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID; 371449bac897SJohannes Thumshirn } 371549bac897SJohannes Thumshirn 3716027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3717027ed2f0SLi Zefan ret = -EFAULT; 3718475f6387SJan Schmidt 3719027ed2f0SLi Zefan kfree(fi_args); 3720027ed2f0SLi Zefan return ret; 3721475f6387SJan Schmidt } 3722475f6387SJan Schmidt 37232ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 37242ff7e61eSJeff Mahoney void __user *arg) 3725475f6387SJan Schmidt { 3726562d7b15SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 3727475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3728475f6387SJan Schmidt struct btrfs_device *dev; 3729475f6387SJan Schmidt int ret = 0; 3730475f6387SJan Schmidt 3731475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3732475f6387SJan Schmidt if (IS_ERR(di_args)) 3733475f6387SJan Schmidt return PTR_ERR(di_args); 3734475f6387SJan Schmidt 3735562d7b15SJosef Bacik args.devid = di_args->devid; 3736dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3737562d7b15SJosef Bacik args.uuid = di_args->uuid; 3738475f6387SJan Schmidt 3739c5593ca3SDavid Sterba rcu_read_lock(); 3740562d7b15SJosef Bacik dev = btrfs_find_device(fs_info->fs_devices, &args); 3741475f6387SJan Schmidt if (!dev) { 3742475f6387SJan Schmidt ret = -ENODEV; 3743475f6387SJan Schmidt goto out; 3744475f6387SJan Schmidt } 3745475f6387SJan Schmidt 3746475f6387SJan Schmidt di_args->devid = dev->devid; 37477cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 37487cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3749475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3750a27202fbSJim Meyering if (dev->name) { 3751672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3752672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3753a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3754a27202fbSJim Meyering } else { 375599ba55adSStefan Behrens di_args->path[0] = '\0'; 3756a27202fbSJim Meyering } 3757475f6387SJan Schmidt 3758475f6387SJan Schmidt out: 3759c5593ca3SDavid Sterba rcu_read_unlock(); 3760475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3761475f6387SJan Schmidt ret = -EFAULT; 3762475f6387SJan Schmidt 3763475f6387SJan Schmidt kfree(di_args); 3764475f6387SJan Schmidt return ret; 3765475f6387SJan Schmidt } 3766475f6387SJan Schmidt 37676ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 37686ef5ed0dSJosef Bacik { 3769496ad9aaSAl Viro struct inode *inode = file_inode(file); 37700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 37716ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 37726ef5ed0dSJosef Bacik struct btrfs_root *new_root; 37736ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 37746ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 37752a2b5d62SJosef Bacik struct btrfs_path *path = NULL; 37766ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 37776ef5ed0dSJosef Bacik u64 objectid = 0; 37786ef5ed0dSJosef Bacik u64 dir_id; 37793c04ce01SMiao Xie int ret; 37806ef5ed0dSJosef Bacik 37816ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 37826ef5ed0dSJosef Bacik return -EPERM; 37836ef5ed0dSJosef Bacik 37843c04ce01SMiao Xie ret = mnt_want_write_file(file); 37853c04ce01SMiao Xie if (ret) 37863c04ce01SMiao Xie return ret; 37873c04ce01SMiao Xie 37883c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 37893c04ce01SMiao Xie ret = -EFAULT; 37903c04ce01SMiao Xie goto out; 37913c04ce01SMiao Xie } 37926ef5ed0dSJosef Bacik 37936ef5ed0dSJosef Bacik if (!objectid) 37941cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 37956ef5ed0dSJosef Bacik 379656e9357aSDavid Sterba new_root = btrfs_get_fs_root(fs_info, objectid, true); 37973c04ce01SMiao Xie if (IS_ERR(new_root)) { 37983c04ce01SMiao Xie ret = PTR_ERR(new_root); 37993c04ce01SMiao Xie goto out; 38003c04ce01SMiao Xie } 38012a2b5d62SJosef Bacik if (!is_fstree(new_root->root_key.objectid)) { 38022a2b5d62SJosef Bacik ret = -ENOENT; 38032a2b5d62SJosef Bacik goto out_free; 38042a2b5d62SJosef Bacik } 38056ef5ed0dSJosef Bacik 38066ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 38073c04ce01SMiao Xie if (!path) { 38083c04ce01SMiao Xie ret = -ENOMEM; 38092a2b5d62SJosef Bacik goto out_free; 38103c04ce01SMiao Xie } 38116ef5ed0dSJosef Bacik 38126ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 381398d5dc13STsutomu Itoh if (IS_ERR(trans)) { 38143c04ce01SMiao Xie ret = PTR_ERR(trans); 38152a2b5d62SJosef Bacik goto out_free; 38166ef5ed0dSJosef Bacik } 38176ef5ed0dSJosef Bacik 38180b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 38190b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 38206ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 3821cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 38222a2b5d62SJosef Bacik btrfs_release_path(path); 38233a45bb20SJeff Mahoney btrfs_end_transaction(trans); 38240b246afaSJeff Mahoney btrfs_err(fs_info, 38255d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 38263c04ce01SMiao Xie ret = -ENOENT; 38272a2b5d62SJosef Bacik goto out_free; 38286ef5ed0dSJosef Bacik } 38296ef5ed0dSJosef Bacik 38306ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 38316ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 38326ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 38332a2b5d62SJosef Bacik btrfs_release_path(path); 38346ef5ed0dSJosef Bacik 38350b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 38363a45bb20SJeff Mahoney btrfs_end_transaction(trans); 38372a2b5d62SJosef Bacik out_free: 383800246528SJosef Bacik btrfs_put_root(new_root); 38392a2b5d62SJosef Bacik btrfs_free_path(path); 38403c04ce01SMiao Xie out: 38413c04ce01SMiao Xie mnt_drop_write_file(file); 38423c04ce01SMiao Xie return ret; 38436ef5ed0dSJosef Bacik } 38446ef5ed0dSJosef Bacik 3845c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 3846bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 3847bf5fc093SJosef Bacik { 384832da5386SDavid Sterba struct btrfs_block_group *block_group; 3849bf5fc093SJosef Bacik 3850bf5fc093SJosef Bacik space->total_bytes = 0; 3851bf5fc093SJosef Bacik space->used_bytes = 0; 3852bf5fc093SJosef Bacik space->flags = 0; 3853bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 3854bf5fc093SJosef Bacik space->flags = block_group->flags; 3855b3470b5dSDavid Sterba space->total_bytes += block_group->length; 3856bf38be65SDavid Sterba space->used_bytes += block_group->used; 3857bf5fc093SJosef Bacik } 3858bf5fc093SJosef Bacik } 3859bf5fc093SJosef Bacik 38602ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 38612ff7e61eSJeff Mahoney void __user *arg) 38621406e432SJosef Bacik { 38631406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 38641406e432SJosef Bacik struct btrfs_ioctl_space_info space; 38651406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 38667fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 386713f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 38681406e432SJosef Bacik struct btrfs_space_info *info; 3869315d8e98SColin Ian King static const u64 types[] = { 3870315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 3871bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 3872bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 3873315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 3874315d8e98SColin Ian King }; 3875bf5fc093SJosef Bacik int num_types = 4; 38767fde62bfSChris Mason int alloc_size; 38771406e432SJosef Bacik int ret = 0; 387851788b1bSDan Rosenberg u64 slot_count = 0; 3879bf5fc093SJosef Bacik int i, c; 38801406e432SJosef Bacik 38811406e432SJosef Bacik if (copy_from_user(&space_args, 38821406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 38831406e432SJosef Bacik sizeof(space_args))) 38841406e432SJosef Bacik return -EFAULT; 38851406e432SJosef Bacik 3886bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3887bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3888bf5fc093SJosef Bacik 3889bf5fc093SJosef Bacik info = NULL; 389072804905SJosef Bacik list_for_each_entry(tmp, &fs_info->space_info, list) { 3891bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3892bf5fc093SJosef Bacik info = tmp; 3893bf5fc093SJosef Bacik break; 3894bf5fc093SJosef Bacik } 3895bf5fc093SJosef Bacik } 38961406e432SJosef Bacik 3897bf5fc093SJosef Bacik if (!info) 3898bf5fc093SJosef Bacik continue; 3899bf5fc093SJosef Bacik 3900bf5fc093SJosef Bacik down_read(&info->groups_sem); 3901bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3902bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 3903bf5fc093SJosef Bacik slot_count++; 3904bf5fc093SJosef Bacik } 3905bf5fc093SJosef Bacik up_read(&info->groups_sem); 3906bf5fc093SJosef Bacik } 3907bf5fc093SJosef Bacik 390836523e95SDavid Sterba /* 390936523e95SDavid Sterba * Global block reserve, exported as a space_info 391036523e95SDavid Sterba */ 391136523e95SDavid Sterba slot_count++; 391236523e95SDavid Sterba 39137fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 39147fde62bfSChris Mason if (space_args.space_slots == 0) { 39157fde62bfSChris Mason space_args.total_spaces = slot_count; 39167fde62bfSChris Mason goto out; 39177fde62bfSChris Mason } 3918bf5fc093SJosef Bacik 391951788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 3920bf5fc093SJosef Bacik 39217fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 3922bf5fc093SJosef Bacik 39237fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 39247fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 39257fde62bfSChris Mason */ 392609cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 39277fde62bfSChris Mason return -ENOMEM; 39287fde62bfSChris Mason 39297fde62bfSChris Mason space_args.total_spaces = 0; 39308d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 39317fde62bfSChris Mason if (!dest) 39327fde62bfSChris Mason return -ENOMEM; 39337fde62bfSChris Mason dest_orig = dest; 39347fde62bfSChris Mason 39357fde62bfSChris Mason /* now we have a buffer to copy into */ 3936bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3937bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3938bf5fc093SJosef Bacik 393951788b1bSDan Rosenberg if (!slot_count) 394051788b1bSDan Rosenberg break; 394151788b1bSDan Rosenberg 3942bf5fc093SJosef Bacik info = NULL; 394372804905SJosef Bacik list_for_each_entry(tmp, &fs_info->space_info, list) { 3944bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3945bf5fc093SJosef Bacik info = tmp; 39467fde62bfSChris Mason break; 3947bf5fc093SJosef Bacik } 3948bf5fc093SJosef Bacik } 39497fde62bfSChris Mason 3950bf5fc093SJosef Bacik if (!info) 3951bf5fc093SJosef Bacik continue; 3952bf5fc093SJosef Bacik down_read(&info->groups_sem); 3953bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3954bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 3955c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 3956c065f5b1SSu Yue &space); 39577fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 39581406e432SJosef Bacik dest++; 39591406e432SJosef Bacik space_args.total_spaces++; 396051788b1bSDan Rosenberg slot_count--; 39611406e432SJosef Bacik } 396251788b1bSDan Rosenberg if (!slot_count) 396351788b1bSDan Rosenberg break; 3964bf5fc093SJosef Bacik } 3965bf5fc093SJosef Bacik up_read(&info->groups_sem); 3966bf5fc093SJosef Bacik } 39671406e432SJosef Bacik 396836523e95SDavid Sterba /* 396936523e95SDavid Sterba * Add global block reserve 397036523e95SDavid Sterba */ 397136523e95SDavid Sterba if (slot_count) { 39720b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 397336523e95SDavid Sterba 397436523e95SDavid Sterba spin_lock(&block_rsv->lock); 397536523e95SDavid Sterba space.total_bytes = block_rsv->size; 397636523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 397736523e95SDavid Sterba spin_unlock(&block_rsv->lock); 397836523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 397936523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 398036523e95SDavid Sterba space_args.total_spaces++; 398136523e95SDavid Sterba } 398236523e95SDavid Sterba 39832eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 39847fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 39857fde62bfSChris Mason 39867fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 39877fde62bfSChris Mason ret = -EFAULT; 39887fde62bfSChris Mason 39897fde62bfSChris Mason kfree(dest_orig); 39907fde62bfSChris Mason out: 39917fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 39921406e432SJosef Bacik ret = -EFAULT; 39931406e432SJosef Bacik 39941406e432SJosef Bacik return ret; 39951406e432SJosef Bacik } 39961406e432SJosef Bacik 39979a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 39989a8c28beSMiao Xie void __user *argp) 399946204592SSage Weil { 400046204592SSage Weil struct btrfs_trans_handle *trans; 400146204592SSage Weil u64 transid; 400246204592SSage Weil 4003d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4004ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4005ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 400698d5dc13STsutomu Itoh return PTR_ERR(trans); 4007ff7c1d33SMiao Xie 4008ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4009ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4010ff7c1d33SMiao Xie goto out; 4011ff7c1d33SMiao Xie } 401246204592SSage Weil transid = trans->transid; 4013fdfbf020SJosef Bacik btrfs_commit_transaction_async(trans); 4014ff7c1d33SMiao Xie out: 401546204592SSage Weil if (argp) 401646204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 401746204592SSage Weil return -EFAULT; 401846204592SSage Weil return 0; 401946204592SSage Weil } 402046204592SSage Weil 40212ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 40229a8c28beSMiao Xie void __user *argp) 402346204592SSage Weil { 402446204592SSage Weil u64 transid; 402546204592SSage Weil 402646204592SSage Weil if (argp) { 402746204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 402846204592SSage Weil return -EFAULT; 402946204592SSage Weil } else { 403046204592SSage Weil transid = 0; /* current trans */ 403146204592SSage Weil } 40322ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 403346204592SSage Weil } 403446204592SSage Weil 4035b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4036475f6387SJan Schmidt { 40370b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4038475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4039b8e95489SMiao Xie int ret; 4040475f6387SJan Schmidt 4041475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4042475f6387SJan Schmidt return -EPERM; 4043475f6387SJan Schmidt 4044da32c6d5SJosef Bacik if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 4045da32c6d5SJosef Bacik btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet"); 4046da32c6d5SJosef Bacik return -EINVAL; 4047da32c6d5SJosef Bacik } 4048da32c6d5SJosef Bacik 4049475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4050475f6387SJan Schmidt if (IS_ERR(sa)) 4051475f6387SJan Schmidt return PTR_ERR(sa); 4052475f6387SJan Schmidt 4053b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4054b8e95489SMiao Xie ret = mnt_want_write_file(file); 4055b8e95489SMiao Xie if (ret) 4056b8e95489SMiao Xie goto out; 4057b8e95489SMiao Xie } 4058b8e95489SMiao Xie 40590b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 406063a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 406163a212abSStefan Behrens 0); 4062475f6387SJan Schmidt 40635afe6ce7SFilipe Manana /* 40645afe6ce7SFilipe Manana * Copy scrub args to user space even if btrfs_scrub_dev() returned an 40655afe6ce7SFilipe Manana * error. This is important as it allows user space to know how much 40665afe6ce7SFilipe Manana * progress scrub has done. For example, if scrub is canceled we get 40675afe6ce7SFilipe Manana * -ECANCELED from btrfs_scrub_dev() and return that error back to user 40685afe6ce7SFilipe Manana * space. Later user space can inspect the progress from the structure 40695afe6ce7SFilipe Manana * btrfs_ioctl_scrub_args and resume scrub from where it left off 40705afe6ce7SFilipe Manana * previously (btrfs-progs does this). 40715afe6ce7SFilipe Manana * If we fail to copy the btrfs_ioctl_scrub_args structure to user space 40725afe6ce7SFilipe Manana * then return -EFAULT to signal the structure was not copied or it may 40735afe6ce7SFilipe Manana * be corrupt and unreliable due to a partial copy. 40745afe6ce7SFilipe Manana */ 40755afe6ce7SFilipe Manana if (copy_to_user(arg, sa, sizeof(*sa))) 4076475f6387SJan Schmidt ret = -EFAULT; 4077475f6387SJan Schmidt 4078b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4079b8e95489SMiao Xie mnt_drop_write_file(file); 4080b8e95489SMiao Xie out: 4081475f6387SJan Schmidt kfree(sa); 4082475f6387SJan Schmidt return ret; 4083475f6387SJan Schmidt } 4084475f6387SJan Schmidt 40852ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4086475f6387SJan Schmidt { 4087475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4088475f6387SJan Schmidt return -EPERM; 4089475f6387SJan Schmidt 40902ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4091475f6387SJan Schmidt } 4092475f6387SJan Schmidt 40932ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4094475f6387SJan Schmidt void __user *arg) 4095475f6387SJan Schmidt { 4096475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4097475f6387SJan Schmidt int ret; 4098475f6387SJan Schmidt 4099475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4100475f6387SJan Schmidt return -EPERM; 4101475f6387SJan Schmidt 4102475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4103475f6387SJan Schmidt if (IS_ERR(sa)) 4104475f6387SJan Schmidt return PTR_ERR(sa); 4105475f6387SJan Schmidt 41062ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4107475f6387SJan Schmidt 41084fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4109475f6387SJan Schmidt ret = -EFAULT; 4110475f6387SJan Schmidt 4111475f6387SJan Schmidt kfree(sa); 4112475f6387SJan Schmidt return ret; 4113475f6387SJan Schmidt } 4114475f6387SJan Schmidt 41152ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4116b27f7c0cSDavid Sterba void __user *arg) 4117c11d2c23SStefan Behrens { 4118c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4119c11d2c23SStefan Behrens int ret; 4120c11d2c23SStefan Behrens 4121c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4122c11d2c23SStefan Behrens if (IS_ERR(sa)) 4123c11d2c23SStefan Behrens return PTR_ERR(sa); 4124c11d2c23SStefan Behrens 4125b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4126b27f7c0cSDavid Sterba kfree(sa); 4127b27f7c0cSDavid Sterba return -EPERM; 4128b27f7c0cSDavid Sterba } 4129b27f7c0cSDavid Sterba 41302ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4131c11d2c23SStefan Behrens 4132eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4133c11d2c23SStefan Behrens ret = -EFAULT; 4134c11d2c23SStefan Behrens 4135c11d2c23SStefan Behrens kfree(sa); 4136c11d2c23SStefan Behrens return ret; 4137c11d2c23SStefan Behrens } 4138c11d2c23SStefan Behrens 41392ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 41402ff7e61eSJeff Mahoney void __user *arg) 41413f6bcfbdSStefan Behrens { 41423f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 41433f6bcfbdSStefan Behrens int ret; 41443f6bcfbdSStefan Behrens 41453f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 41463f6bcfbdSStefan Behrens return -EPERM; 41473f6bcfbdSStefan Behrens 4148914a519bSJosef Bacik if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 4149914a519bSJosef Bacik btrfs_err(fs_info, "device replace not supported on extent tree v2 yet"); 4150914a519bSJosef Bacik return -EINVAL; 4151914a519bSJosef Bacik } 4152914a519bSJosef Bacik 41533f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 41543f6bcfbdSStefan Behrens if (IS_ERR(p)) 41553f6bcfbdSStefan Behrens return PTR_ERR(p); 41563f6bcfbdSStefan Behrens 41573f6bcfbdSStefan Behrens switch (p->cmd) { 41583f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4159bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4160adfa97cbSIlya Dryomov ret = -EROFS; 4161adfa97cbSIlya Dryomov goto out; 4162adfa97cbSIlya Dryomov } 4163c3e1f96cSGoldwyn Rodrigues if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) { 4164e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 41653f6bcfbdSStefan Behrens } else { 41662ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4167c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 41683f6bcfbdSStefan Behrens } 41693f6bcfbdSStefan Behrens break; 41703f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 41710b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 41723f6bcfbdSStefan Behrens ret = 0; 41733f6bcfbdSStefan Behrens break; 41743f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 417517d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 417697282031SAnand Jain ret = 0; 41773f6bcfbdSStefan Behrens break; 41783f6bcfbdSStefan Behrens default: 41793f6bcfbdSStefan Behrens ret = -EINVAL; 41803f6bcfbdSStefan Behrens break; 41813f6bcfbdSStefan Behrens } 41823f6bcfbdSStefan Behrens 4183d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 41843f6bcfbdSStefan Behrens ret = -EFAULT; 4185adfa97cbSIlya Dryomov out: 41863f6bcfbdSStefan Behrens kfree(p); 41873f6bcfbdSStefan Behrens return ret; 41883f6bcfbdSStefan Behrens } 41893f6bcfbdSStefan Behrens 4190d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4191d7728c96SJan Schmidt { 4192d7728c96SJan Schmidt int ret = 0; 4193d7728c96SJan Schmidt int i; 4194740c3d22SChris Mason u64 rel_ptr; 4195d7728c96SJan Schmidt int size; 4196806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4197d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4198d7728c96SJan Schmidt struct btrfs_path *path; 4199d7728c96SJan Schmidt 420082b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4201d7728c96SJan Schmidt return -EPERM; 4202d7728c96SJan Schmidt 4203d7728c96SJan Schmidt path = btrfs_alloc_path(); 4204d7728c96SJan Schmidt if (!path) { 4205d7728c96SJan Schmidt ret = -ENOMEM; 4206d7728c96SJan Schmidt goto out; 4207d7728c96SJan Schmidt } 4208d7728c96SJan Schmidt 4209d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4210d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4211d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4212d7728c96SJan Schmidt ipa = NULL; 4213d7728c96SJan Schmidt goto out; 4214d7728c96SJan Schmidt } 4215d7728c96SJan Schmidt 4216d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4217d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4218d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4219d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4220d7728c96SJan Schmidt ipath = NULL; 4221d7728c96SJan Schmidt goto out; 4222d7728c96SJan Schmidt } 4223d7728c96SJan Schmidt 4224d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4225d7728c96SJan Schmidt if (ret < 0) 4226d7728c96SJan Schmidt goto out; 4227d7728c96SJan Schmidt 4228d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4229745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4230745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4231740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4232d7728c96SJan Schmidt } 4233d7728c96SJan Schmidt 4234718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4235718dc5faSOmar Sandoval ipath->fspath, size); 4236d7728c96SJan Schmidt if (ret) { 4237d7728c96SJan Schmidt ret = -EFAULT; 4238d7728c96SJan Schmidt goto out; 4239d7728c96SJan Schmidt } 4240d7728c96SJan Schmidt 4241d7728c96SJan Schmidt out: 4242d7728c96SJan Schmidt btrfs_free_path(path); 4243d7728c96SJan Schmidt free_ipath(ipath); 4244d7728c96SJan Schmidt kfree(ipa); 4245d7728c96SJan Schmidt 4246d7728c96SJan Schmidt return ret; 4247d7728c96SJan Schmidt } 4248d7728c96SJan Schmidt 42492ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4250d24a67b2SZygo Blaxell void __user *arg, int version) 4251d7728c96SJan Schmidt { 4252d7728c96SJan Schmidt int ret = 0; 4253d7728c96SJan Schmidt int size; 4254d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4255d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4256d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4257d24a67b2SZygo Blaxell bool ignore_offset; 4258d7728c96SJan Schmidt 4259d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4260d7728c96SJan Schmidt return -EPERM; 4261d7728c96SJan Schmidt 4262d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 42637b9ea627SShailendra Verma if (IS_ERR(loi)) 42647b9ea627SShailendra Verma return PTR_ERR(loi); 4265d7728c96SJan Schmidt 4266d24a67b2SZygo Blaxell if (version == 1) { 4267d24a67b2SZygo Blaxell ignore_offset = false; 4268b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4269d24a67b2SZygo Blaxell } else { 4270d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4271d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4272d24a67b2SZygo Blaxell ret = -EINVAL; 4273d24a67b2SZygo Blaxell goto out_loi; 4274d24a67b2SZygo Blaxell } 4275d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4276d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4277d24a67b2SZygo Blaxell ret = -EINVAL; 4278d24a67b2SZygo Blaxell goto out_loi; 4279d24a67b2SZygo Blaxell } 4280d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4281b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4282d24a67b2SZygo Blaxell } 4283d24a67b2SZygo Blaxell 4284d7728c96SJan Schmidt path = btrfs_alloc_path(); 4285d7728c96SJan Schmidt if (!path) { 4286d7728c96SJan Schmidt ret = -ENOMEM; 4287d7728c96SJan Schmidt goto out; 4288d7728c96SJan Schmidt } 4289d7728c96SJan Schmidt 4290d7728c96SJan Schmidt inodes = init_data_container(size); 4291d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4292d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4293d7728c96SJan Schmidt inodes = NULL; 4294d7728c96SJan Schmidt goto out; 4295d7728c96SJan Schmidt } 4296d7728c96SJan Schmidt 42972ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4298e3059ec0SDavid Sterba inodes, ignore_offset); 4299df031f07SLiu Bo if (ret == -EINVAL) 4300d7728c96SJan Schmidt ret = -ENOENT; 4301d7728c96SJan Schmidt if (ret < 0) 4302d7728c96SJan Schmidt goto out; 4303d7728c96SJan Schmidt 4304718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4305718dc5faSOmar Sandoval size); 4306d7728c96SJan Schmidt if (ret) 4307d7728c96SJan Schmidt ret = -EFAULT; 4308d7728c96SJan Schmidt 4309d7728c96SJan Schmidt out: 4310d7728c96SJan Schmidt btrfs_free_path(path); 4311f54de068SDavid Sterba kvfree(inodes); 4312d24a67b2SZygo Blaxell out_loi: 4313d7728c96SJan Schmidt kfree(loi); 4314d7728c96SJan Schmidt 4315d7728c96SJan Schmidt return ret; 4316d7728c96SJan Schmidt } 4317d7728c96SJan Schmidt 4318008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4319c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4320c9e9f97bSIlya Dryomov { 4321c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4322c9e9f97bSIlya Dryomov 4323c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4324c9e9f97bSIlya Dryomov 43253009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4326837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4327837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4328837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4329a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4330a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4331837d5b6eSIlya Dryomov 4332c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4333c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4334c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 433519a39dceSIlya Dryomov 433619a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 433719a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 433819a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4339c9e9f97bSIlya Dryomov } 4340c9e9f97bSIlya Dryomov 43417fb10ed8SNikolay Borisov /** 43427fb10ed8SNikolay Borisov * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as 43437fb10ed8SNikolay Borisov * required. 43447fb10ed8SNikolay Borisov * 43457fb10ed8SNikolay Borisov * @fs_info: the filesystem 43467fb10ed8SNikolay Borisov * @excl_acquired: ptr to boolean value which is set to false in case balance 43477fb10ed8SNikolay Borisov * is being resumed 43487fb10ed8SNikolay Borisov * 43497fb10ed8SNikolay Borisov * Return 0 on success in which case both fs_info::balance is acquired as well 43507fb10ed8SNikolay Borisov * as exclusive ops are blocked. In case of failure return an error code. 43517fb10ed8SNikolay Borisov */ 43527fb10ed8SNikolay Borisov static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired) 43537fb10ed8SNikolay Borisov { 43547fb10ed8SNikolay Borisov int ret; 43557fb10ed8SNikolay Borisov 43567fb10ed8SNikolay Borisov /* 43577fb10ed8SNikolay Borisov * Exclusive operation is locked. Three possibilities: 43587fb10ed8SNikolay Borisov * (1) some other op is running 43597fb10ed8SNikolay Borisov * (2) balance is running 43607fb10ed8SNikolay Borisov * (3) balance is paused -- special case (think resume) 43617fb10ed8SNikolay Borisov */ 43627fb10ed8SNikolay Borisov while (1) { 43637fb10ed8SNikolay Borisov if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { 43647fb10ed8SNikolay Borisov *excl_acquired = true; 43657fb10ed8SNikolay Borisov mutex_lock(&fs_info->balance_mutex); 43667fb10ed8SNikolay Borisov return 0; 43677fb10ed8SNikolay Borisov } 43687fb10ed8SNikolay Borisov 43697fb10ed8SNikolay Borisov mutex_lock(&fs_info->balance_mutex); 43707fb10ed8SNikolay Borisov if (fs_info->balance_ctl) { 43717fb10ed8SNikolay Borisov /* This is either (2) or (3) */ 43727fb10ed8SNikolay Borisov if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 43737fb10ed8SNikolay Borisov /* This is (2) */ 43747fb10ed8SNikolay Borisov ret = -EINPROGRESS; 43757fb10ed8SNikolay Borisov goto out_failure; 43767fb10ed8SNikolay Borisov 43777fb10ed8SNikolay Borisov } else { 43787fb10ed8SNikolay Borisov mutex_unlock(&fs_info->balance_mutex); 43797fb10ed8SNikolay Borisov /* 43807fb10ed8SNikolay Borisov * Lock released to allow other waiters to 43817fb10ed8SNikolay Borisov * continue, we'll reexamine the status again. 43827fb10ed8SNikolay Borisov */ 43837fb10ed8SNikolay Borisov mutex_lock(&fs_info->balance_mutex); 43847fb10ed8SNikolay Borisov 43857fb10ed8SNikolay Borisov if (fs_info->balance_ctl && 43867fb10ed8SNikolay Borisov !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 43877fb10ed8SNikolay Borisov /* This is (3) */ 43887fb10ed8SNikolay Borisov *excl_acquired = false; 43897fb10ed8SNikolay Borisov return 0; 43907fb10ed8SNikolay Borisov } 43917fb10ed8SNikolay Borisov } 43927fb10ed8SNikolay Borisov } else { 43937fb10ed8SNikolay Borisov /* This is (1) */ 43947fb10ed8SNikolay Borisov ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 43957fb10ed8SNikolay Borisov goto out_failure; 43967fb10ed8SNikolay Borisov } 43977fb10ed8SNikolay Borisov 43987fb10ed8SNikolay Borisov mutex_unlock(&fs_info->balance_mutex); 43997fb10ed8SNikolay Borisov } 44007fb10ed8SNikolay Borisov 44017fb10ed8SNikolay Borisov out_failure: 44027fb10ed8SNikolay Borisov mutex_unlock(&fs_info->balance_mutex); 44037fb10ed8SNikolay Borisov *excl_acquired = false; 44047fb10ed8SNikolay Borisov return ret; 44057fb10ed8SNikolay Borisov } 44067fb10ed8SNikolay Borisov 44079ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4408c9e9f97bSIlya Dryomov { 4409496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4410c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4411c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4412c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4413099aa972SNikolay Borisov bool need_unlock = true; 4414c9e9f97bSIlya Dryomov int ret; 4415c9e9f97bSIlya Dryomov 4416c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4417c9e9f97bSIlya Dryomov return -EPERM; 4418c9e9f97bSIlya Dryomov 4419e54bfa31SLiu Bo ret = mnt_want_write_file(file); 44209ba1f6e4SLiu Bo if (ret) 44219ba1f6e4SLiu Bo return ret; 44229ba1f6e4SLiu Bo 4423d8645462SNikolay Borisov bargs = memdup_user(arg, sizeof(*bargs)); 4424d8645462SNikolay Borisov if (IS_ERR(bargs)) { 4425d8645462SNikolay Borisov ret = PTR_ERR(bargs); 4426d8645462SNikolay Borisov bargs = NULL; 4427d8645462SNikolay Borisov goto out; 4428d8645462SNikolay Borisov } 4429d8645462SNikolay Borisov 4430099aa972SNikolay Borisov ret = btrfs_try_lock_balance(fs_info, &need_unlock); 4431099aa972SNikolay Borisov if (ret) 4432ed0fb78fSIlya Dryomov goto out; 4433ed0fb78fSIlya Dryomov 4434099aa972SNikolay Borisov lockdep_assert_held(&fs_info->balance_mutex); 4435099aa972SNikolay Borisov 4436de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4437de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4438de322263SIlya Dryomov ret = -ENOTCONN; 4439d8645462SNikolay Borisov goto out_unlock; 4440de322263SIlya Dryomov } 4441de322263SIlya Dryomov 4442de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4443de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4444de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4445de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4446efc0e69cSNikolay Borisov btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE); 4447de322263SIlya Dryomov 4448de322263SIlya Dryomov goto do_balance; 4449de322263SIlya Dryomov } 4450c9e9f97bSIlya Dryomov 4451d8645462SNikolay Borisov if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 4452d8645462SNikolay Borisov ret = -EINVAL; 4453d8645462SNikolay Borisov goto out_unlock; 4454c9e9f97bSIlya Dryomov } 4455c9e9f97bSIlya Dryomov 4456ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4457837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4458d8645462SNikolay Borisov goto out_unlock; 4459837d5b6eSIlya Dryomov } 4460837d5b6eSIlya Dryomov 44618d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4462c9e9f97bSIlya Dryomov if (!bctl) { 4463c9e9f97bSIlya Dryomov ret = -ENOMEM; 4464d8645462SNikolay Borisov goto out_unlock; 4465c9e9f97bSIlya Dryomov } 4466c9e9f97bSIlya Dryomov 4467c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4468c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4469c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4470c9e9f97bSIlya Dryomov 4471c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4472de322263SIlya Dryomov do_balance: 4473c9e9f97bSIlya Dryomov /* 4474c3e1f96cSGoldwyn Rodrigues * Ownership of bctl and exclusive operation goes to btrfs_balance. 4475c3e1f96cSGoldwyn Rodrigues * bctl is freed in reset_balance_state, or, if restriper was paused 4476c3e1f96cSGoldwyn Rodrigues * all the way until unmount, in free_fs_info. The flag should be 4477c3e1f96cSGoldwyn Rodrigues * cleared after reset_balance_state. 4478c9e9f97bSIlya Dryomov */ 4479ed0fb78fSIlya Dryomov need_unlock = false; 4480ed0fb78fSIlya Dryomov 44816fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 44820f89abf5SChristian Engelmayer bctl = NULL; 4483ed0fb78fSIlya Dryomov 448439864601SNikolay Borisov if (ret == 0 || ret == -ECANCELED) { 4485c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4486c9e9f97bSIlya Dryomov ret = -EFAULT; 4487c9e9f97bSIlya Dryomov } 4488c9e9f97bSIlya Dryomov 44890f89abf5SChristian Engelmayer kfree(bctl); 4490ed0fb78fSIlya Dryomov out_unlock: 4491c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4492ed0fb78fSIlya Dryomov if (need_unlock) 4493c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 4494ed0fb78fSIlya Dryomov out: 4495e54bfa31SLiu Bo mnt_drop_write_file(file); 4496d8645462SNikolay Borisov kfree(bargs); 4497c9e9f97bSIlya Dryomov return ret; 4498c9e9f97bSIlya Dryomov } 4499c9e9f97bSIlya Dryomov 45002ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4501837d5b6eSIlya Dryomov { 4502837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4503837d5b6eSIlya Dryomov return -EPERM; 4504837d5b6eSIlya Dryomov 4505837d5b6eSIlya Dryomov switch (cmd) { 4506837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 45070b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4508a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 45090b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4510837d5b6eSIlya Dryomov } 4511837d5b6eSIlya Dryomov 4512837d5b6eSIlya Dryomov return -EINVAL; 4513837d5b6eSIlya Dryomov } 4514837d5b6eSIlya Dryomov 45152ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 451619a39dceSIlya Dryomov void __user *arg) 451719a39dceSIlya Dryomov { 451819a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 451919a39dceSIlya Dryomov int ret = 0; 452019a39dceSIlya Dryomov 452119a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 452219a39dceSIlya Dryomov return -EPERM; 452319a39dceSIlya Dryomov 452419a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 452519a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 452619a39dceSIlya Dryomov ret = -ENOTCONN; 452719a39dceSIlya Dryomov goto out; 452819a39dceSIlya Dryomov } 452919a39dceSIlya Dryomov 45308d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 453119a39dceSIlya Dryomov if (!bargs) { 453219a39dceSIlya Dryomov ret = -ENOMEM; 453319a39dceSIlya Dryomov goto out; 453419a39dceSIlya Dryomov } 453519a39dceSIlya Dryomov 4536008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 453719a39dceSIlya Dryomov 453819a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 453919a39dceSIlya Dryomov ret = -EFAULT; 454019a39dceSIlya Dryomov 454119a39dceSIlya Dryomov kfree(bargs); 454219a39dceSIlya Dryomov out: 454319a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 454419a39dceSIlya Dryomov return ret; 454519a39dceSIlya Dryomov } 454619a39dceSIlya Dryomov 4547905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 45485d13a37bSArne Jansen { 45490b246afaSJeff Mahoney struct inode *inode = file_inode(file); 45500b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 45515d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 45525d13a37bSArne Jansen int ret; 45535d13a37bSArne Jansen 45545d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 45555d13a37bSArne Jansen return -EPERM; 45565d13a37bSArne Jansen 4557905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4558905b0ddaSMiao Xie if (ret) 4559905b0ddaSMiao Xie return ret; 45605d13a37bSArne Jansen 45615d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4562905b0ddaSMiao Xie if (IS_ERR(sa)) { 4563905b0ddaSMiao Xie ret = PTR_ERR(sa); 4564905b0ddaSMiao Xie goto drop_write; 4565905b0ddaSMiao Xie } 45665d13a37bSArne Jansen 45670b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 45685d13a37bSArne Jansen 45695d13a37bSArne Jansen switch (sa->cmd) { 45705d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4571340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 45725d13a37bSArne Jansen break; 45735d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4574340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 45755d13a37bSArne Jansen break; 45765d13a37bSArne Jansen default: 45775d13a37bSArne Jansen ret = -EINVAL; 45785d13a37bSArne Jansen break; 45795d13a37bSArne Jansen } 45805d13a37bSArne Jansen 45815d13a37bSArne Jansen kfree(sa); 45820b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4583905b0ddaSMiao Xie drop_write: 4584905b0ddaSMiao Xie mnt_drop_write_file(file); 45855d13a37bSArne Jansen return ret; 45865d13a37bSArne Jansen } 45875d13a37bSArne Jansen 4588905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 45895d13a37bSArne Jansen { 45900b246afaSJeff Mahoney struct inode *inode = file_inode(file); 45910b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 45920b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 45935d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 45945d13a37bSArne Jansen struct btrfs_trans_handle *trans; 45955d13a37bSArne Jansen int ret; 45965d13a37bSArne Jansen int err; 45975d13a37bSArne Jansen 45985d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 45995d13a37bSArne Jansen return -EPERM; 46005d13a37bSArne Jansen 4601905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4602905b0ddaSMiao Xie if (ret) 4603905b0ddaSMiao Xie return ret; 46045d13a37bSArne Jansen 46055d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4606905b0ddaSMiao Xie if (IS_ERR(sa)) { 4607905b0ddaSMiao Xie ret = PTR_ERR(sa); 4608905b0ddaSMiao Xie goto drop_write; 4609905b0ddaSMiao Xie } 46105d13a37bSArne Jansen 46115d13a37bSArne Jansen trans = btrfs_join_transaction(root); 46125d13a37bSArne Jansen if (IS_ERR(trans)) { 46135d13a37bSArne Jansen ret = PTR_ERR(trans); 46145d13a37bSArne Jansen goto out; 46155d13a37bSArne Jansen } 46165d13a37bSArne Jansen 46175d13a37bSArne Jansen if (sa->assign) { 46189f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 46195d13a37bSArne Jansen } else { 462039616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 46215d13a37bSArne Jansen } 46225d13a37bSArne Jansen 4623e082f563SQu Wenruo /* update qgroup status and info */ 4624280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4625e082f563SQu Wenruo if (err < 0) 46260b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4627ad8403dfSAnand Jain "failed to update qgroup status and info"); 46283a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 46295d13a37bSArne Jansen if (err && !ret) 46305d13a37bSArne Jansen ret = err; 46315d13a37bSArne Jansen 46325d13a37bSArne Jansen out: 46335d13a37bSArne Jansen kfree(sa); 4634905b0ddaSMiao Xie drop_write: 4635905b0ddaSMiao Xie mnt_drop_write_file(file); 46365d13a37bSArne Jansen return ret; 46375d13a37bSArne Jansen } 46385d13a37bSArne Jansen 4639905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 46405d13a37bSArne Jansen { 46410b246afaSJeff Mahoney struct inode *inode = file_inode(file); 46420b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 46435d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 46445d13a37bSArne Jansen struct btrfs_trans_handle *trans; 46455d13a37bSArne Jansen int ret; 46465d13a37bSArne Jansen int err; 46475d13a37bSArne Jansen 46485d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 46495d13a37bSArne Jansen return -EPERM; 46505d13a37bSArne Jansen 4651905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4652905b0ddaSMiao Xie if (ret) 4653905b0ddaSMiao Xie return ret; 46545d13a37bSArne Jansen 46555d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4656905b0ddaSMiao Xie if (IS_ERR(sa)) { 4657905b0ddaSMiao Xie ret = PTR_ERR(sa); 4658905b0ddaSMiao Xie goto drop_write; 4659905b0ddaSMiao Xie } 46605d13a37bSArne Jansen 4661d86e56cfSMiao Xie if (!sa->qgroupid) { 4662d86e56cfSMiao Xie ret = -EINVAL; 4663d86e56cfSMiao Xie goto out; 4664d86e56cfSMiao Xie } 4665d86e56cfSMiao Xie 46665d13a37bSArne Jansen trans = btrfs_join_transaction(root); 46675d13a37bSArne Jansen if (IS_ERR(trans)) { 46685d13a37bSArne Jansen ret = PTR_ERR(trans); 46695d13a37bSArne Jansen goto out; 46705d13a37bSArne Jansen } 46715d13a37bSArne Jansen 46725d13a37bSArne Jansen if (sa->create) { 467349a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 46745d13a37bSArne Jansen } else { 46753efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 46765d13a37bSArne Jansen } 46775d13a37bSArne Jansen 46783a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 46795d13a37bSArne Jansen if (err && !ret) 46805d13a37bSArne Jansen ret = err; 46815d13a37bSArne Jansen 46825d13a37bSArne Jansen out: 46835d13a37bSArne Jansen kfree(sa); 4684905b0ddaSMiao Xie drop_write: 4685905b0ddaSMiao Xie mnt_drop_write_file(file); 46865d13a37bSArne Jansen return ret; 46875d13a37bSArne Jansen } 46885d13a37bSArne Jansen 4689905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 46905d13a37bSArne Jansen { 46910b246afaSJeff Mahoney struct inode *inode = file_inode(file); 46920b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 46935d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 46945d13a37bSArne Jansen struct btrfs_trans_handle *trans; 46955d13a37bSArne Jansen int ret; 46965d13a37bSArne Jansen int err; 46975d13a37bSArne Jansen u64 qgroupid; 46985d13a37bSArne Jansen 46995d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47005d13a37bSArne Jansen return -EPERM; 47015d13a37bSArne Jansen 4702905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4703905b0ddaSMiao Xie if (ret) 4704905b0ddaSMiao Xie return ret; 47055d13a37bSArne Jansen 47065d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4707905b0ddaSMiao Xie if (IS_ERR(sa)) { 4708905b0ddaSMiao Xie ret = PTR_ERR(sa); 4709905b0ddaSMiao Xie goto drop_write; 4710905b0ddaSMiao Xie } 47115d13a37bSArne Jansen 47125d13a37bSArne Jansen trans = btrfs_join_transaction(root); 47135d13a37bSArne Jansen if (IS_ERR(trans)) { 47145d13a37bSArne Jansen ret = PTR_ERR(trans); 47155d13a37bSArne Jansen goto out; 47165d13a37bSArne Jansen } 47175d13a37bSArne Jansen 47185d13a37bSArne Jansen qgroupid = sa->qgroupid; 47195d13a37bSArne Jansen if (!qgroupid) { 47205d13a37bSArne Jansen /* take the current subvol as qgroup */ 47215d13a37bSArne Jansen qgroupid = root->root_key.objectid; 47225d13a37bSArne Jansen } 47235d13a37bSArne Jansen 4724f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 47255d13a37bSArne Jansen 47263a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 47275d13a37bSArne Jansen if (err && !ret) 47285d13a37bSArne Jansen ret = err; 47295d13a37bSArne Jansen 47305d13a37bSArne Jansen out: 47315d13a37bSArne Jansen kfree(sa); 4732905b0ddaSMiao Xie drop_write: 4733905b0ddaSMiao Xie mnt_drop_write_file(file); 47345d13a37bSArne Jansen return ret; 47355d13a37bSArne Jansen } 47365d13a37bSArne Jansen 47372f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 47382f232036SJan Schmidt { 47390b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47400b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47412f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 47422f232036SJan Schmidt int ret; 47432f232036SJan Schmidt 47442f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 47452f232036SJan Schmidt return -EPERM; 47462f232036SJan Schmidt 47472f232036SJan Schmidt ret = mnt_want_write_file(file); 47482f232036SJan Schmidt if (ret) 47492f232036SJan Schmidt return ret; 47502f232036SJan Schmidt 47512f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 47522f232036SJan Schmidt if (IS_ERR(qsa)) { 47532f232036SJan Schmidt ret = PTR_ERR(qsa); 47542f232036SJan Schmidt goto drop_write; 47552f232036SJan Schmidt } 47562f232036SJan Schmidt 47572f232036SJan Schmidt if (qsa->flags) { 47582f232036SJan Schmidt ret = -EINVAL; 47592f232036SJan Schmidt goto out; 47602f232036SJan Schmidt } 47612f232036SJan Schmidt 47620b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 47632f232036SJan Schmidt 47642f232036SJan Schmidt out: 47652f232036SJan Schmidt kfree(qsa); 47662f232036SJan Schmidt drop_write: 47672f232036SJan Schmidt mnt_drop_write_file(file); 47682f232036SJan Schmidt return ret; 47692f232036SJan Schmidt } 47702f232036SJan Schmidt 4771b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, 4772b929c1d8SMarcos Paulo de Souza void __user *arg) 47732f232036SJan Schmidt { 47740afb603aSGoldwyn Rodrigues struct btrfs_ioctl_quota_rescan_args qsa = {0}; 47752f232036SJan Schmidt 47762f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 47772f232036SJan Schmidt return -EPERM; 47782f232036SJan Schmidt 47790b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 47800afb603aSGoldwyn Rodrigues qsa.flags = 1; 47810afb603aSGoldwyn Rodrigues qsa.progress = fs_info->qgroup_rescan_progress.objectid; 47822f232036SJan Schmidt } 47832f232036SJan Schmidt 47840afb603aSGoldwyn Rodrigues if (copy_to_user(arg, &qsa, sizeof(qsa))) 4785991a3daeSAnand Jain return -EFAULT; 47862f232036SJan Schmidt 4787991a3daeSAnand Jain return 0; 47882f232036SJan Schmidt } 47892f232036SJan Schmidt 4790b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info, 4791b929c1d8SMarcos Paulo de Souza void __user *arg) 479257254b6eSJan Schmidt { 479357254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 479457254b6eSJan Schmidt return -EPERM; 479557254b6eSJan Schmidt 47960b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 479757254b6eSJan Schmidt } 479857254b6eSJan Schmidt 4799abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 4800e4fed17aSChristian Brauner struct user_namespace *mnt_userns, 4801abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 48028ea05e3aSAlexander Block { 4803496ad9aaSAl Viro struct inode *inode = file_inode(file); 48040b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48058ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 48068ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 48078ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 480895582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 48098ea05e3aSAlexander Block int ret = 0; 4810dd5f9615SStefan Behrens int received_uuid_changed; 48118ea05e3aSAlexander Block 4812e4fed17aSChristian Brauner if (!inode_owner_or_capable(mnt_userns, inode)) 4813bd60ea0fSDavid Sterba return -EPERM; 4814bd60ea0fSDavid Sterba 48158ea05e3aSAlexander Block ret = mnt_want_write_file(file); 48168ea05e3aSAlexander Block if (ret < 0) 48178ea05e3aSAlexander Block return ret; 48188ea05e3aSAlexander Block 48190b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48208ea05e3aSAlexander Block 48214a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 48228ea05e3aSAlexander Block ret = -EINVAL; 48238ea05e3aSAlexander Block goto out; 48248ea05e3aSAlexander Block } 48258ea05e3aSAlexander Block 48268ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 48278ea05e3aSAlexander Block ret = -EROFS; 48288ea05e3aSAlexander Block goto out; 48298ea05e3aSAlexander Block } 48308ea05e3aSAlexander Block 4831dd5f9615SStefan Behrens /* 4832dd5f9615SStefan Behrens * 1 - root item 4833dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 4834dd5f9615SStefan Behrens */ 4835dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 48368ea05e3aSAlexander Block if (IS_ERR(trans)) { 48378ea05e3aSAlexander Block ret = PTR_ERR(trans); 48388ea05e3aSAlexander Block trans = NULL; 48398ea05e3aSAlexander Block goto out; 48408ea05e3aSAlexander Block } 48418ea05e3aSAlexander Block 48428ea05e3aSAlexander Block sa->rtransid = trans->transid; 48438ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 48448ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 48458ea05e3aSAlexander Block 4846dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 4847dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 4848dd5f9615SStefan Behrens if (received_uuid_changed && 4849d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 4850d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 4851dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4852dd5f9615SStefan Behrens root->root_key.objectid); 4853d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 4854d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 4855d87ff758SNikolay Borisov btrfs_end_transaction(trans); 4856d87ff758SNikolay Borisov goto out; 4857d87ff758SNikolay Borisov } 4858d87ff758SNikolay Borisov } 48598ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 48608ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 48618ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 48623cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 48633cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 48643cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 48653cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 48668ea05e3aSAlexander Block 48670b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 48688ea05e3aSAlexander Block &root->root_key, &root->root_item); 48698ea05e3aSAlexander Block if (ret < 0) { 48703a45bb20SJeff Mahoney btrfs_end_transaction(trans); 48718ea05e3aSAlexander Block goto out; 4872dd5f9615SStefan Behrens } 4873dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 4874cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 4875dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4876dd5f9615SStefan Behrens root->root_key.objectid); 4877dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 487866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 4879efd38150SNikolay Borisov btrfs_end_transaction(trans); 4880dd5f9615SStefan Behrens goto out; 4881dd5f9615SStefan Behrens } 4882dd5f9615SStefan Behrens } 48833a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 4884abccd00fSHugo Mills out: 48850b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4886abccd00fSHugo Mills mnt_drop_write_file(file); 4887abccd00fSHugo Mills return ret; 4888abccd00fSHugo Mills } 4889abccd00fSHugo Mills 4890abccd00fSHugo Mills #ifdef CONFIG_64BIT 4891abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 4892abccd00fSHugo Mills void __user *arg) 4893abccd00fSHugo Mills { 4894abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 4895abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 4896abccd00fSHugo Mills int ret = 0; 4897abccd00fSHugo Mills 4898abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 48997b9ea627SShailendra Verma if (IS_ERR(args32)) 49007b9ea627SShailendra Verma return PTR_ERR(args32); 4901abccd00fSHugo Mills 49028d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 490384dbeb87SDan Carpenter if (!args64) { 490484dbeb87SDan Carpenter ret = -ENOMEM; 4905abccd00fSHugo Mills goto out; 4906abccd00fSHugo Mills } 4907abccd00fSHugo Mills 4908abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 4909abccd00fSHugo Mills args64->stransid = args32->stransid; 4910abccd00fSHugo Mills args64->rtransid = args32->rtransid; 4911abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 4912abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 4913abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 4914abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 4915abccd00fSHugo Mills args64->flags = args32->flags; 4916abccd00fSHugo Mills 4917e4fed17aSChristian Brauner ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64); 4918abccd00fSHugo Mills if (ret) 4919abccd00fSHugo Mills goto out; 4920abccd00fSHugo Mills 4921abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 4922abccd00fSHugo Mills args32->stransid = args64->stransid; 4923abccd00fSHugo Mills args32->rtransid = args64->rtransid; 4924abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 4925abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 4926abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 4927abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 4928abccd00fSHugo Mills args32->flags = args64->flags; 4929abccd00fSHugo Mills 4930abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 4931abccd00fSHugo Mills if (ret) 4932abccd00fSHugo Mills ret = -EFAULT; 4933abccd00fSHugo Mills 4934abccd00fSHugo Mills out: 4935abccd00fSHugo Mills kfree(args32); 4936abccd00fSHugo Mills kfree(args64); 4937abccd00fSHugo Mills return ret; 4938abccd00fSHugo Mills } 4939abccd00fSHugo Mills #endif 4940abccd00fSHugo Mills 4941abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 4942abccd00fSHugo Mills void __user *arg) 4943abccd00fSHugo Mills { 4944abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 4945abccd00fSHugo Mills int ret = 0; 4946abccd00fSHugo Mills 4947abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 49487b9ea627SShailendra Verma if (IS_ERR(sa)) 49497b9ea627SShailendra Verma return PTR_ERR(sa); 4950abccd00fSHugo Mills 4951e4fed17aSChristian Brauner ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa); 4952abccd00fSHugo Mills 4953abccd00fSHugo Mills if (ret) 4954abccd00fSHugo Mills goto out; 4955abccd00fSHugo Mills 49568ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 49578ea05e3aSAlexander Block if (ret) 49588ea05e3aSAlexander Block ret = -EFAULT; 49598ea05e3aSAlexander Block 49608ea05e3aSAlexander Block out: 49618ea05e3aSAlexander Block kfree(sa); 49628ea05e3aSAlexander Block return ret; 49638ea05e3aSAlexander Block } 49648ea05e3aSAlexander Block 4965b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, 4966b929c1d8SMarcos Paulo de Souza void __user *arg) 4967867ab667Sjeff.liu { 4968a1b83ac5SAnand Jain size_t len; 4969867ab667Sjeff.liu int ret; 4970a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 4971a1b83ac5SAnand Jain 49720b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 49730b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 49740b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 4975a1b83ac5SAnand Jain 4976a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 4977867ab667Sjeff.liu 4978867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 49790b246afaSJeff Mahoney btrfs_warn(fs_info, 49800b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 49810b246afaSJeff Mahoney --len); 4982867ab667Sjeff.liu } 4983867ab667Sjeff.liu 4984867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 4985867ab667Sjeff.liu 4986867ab667Sjeff.liu return ret ? -EFAULT : 0; 4987867ab667Sjeff.liu } 4988867ab667Sjeff.liu 4989a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 4990a8bfd4abSjeff.liu { 49910b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49920b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49930b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49940b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 4995a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 4996a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 4997a8bfd4abSjeff.liu int ret; 4998a8bfd4abSjeff.liu 4999a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5000a8bfd4abSjeff.liu return -EPERM; 5001a8bfd4abSjeff.liu 5002a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5003a8bfd4abSjeff.liu return -EFAULT; 5004a8bfd4abSjeff.liu 5005a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 50060b246afaSJeff Mahoney btrfs_err(fs_info, 50075d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5008a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5009a8bfd4abSjeff.liu return -EINVAL; 5010a8bfd4abSjeff.liu } 5011a8bfd4abSjeff.liu 5012a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5013a8bfd4abSjeff.liu if (ret) 5014a8bfd4abSjeff.liu return ret; 5015a8bfd4abSjeff.liu 5016a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5017a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5018a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5019a8bfd4abSjeff.liu goto out_unlock; 5020a8bfd4abSjeff.liu } 5021a8bfd4abSjeff.liu 50220b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5023a8bfd4abSjeff.liu strcpy(super_block->label, label); 50240b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 50253a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5026a8bfd4abSjeff.liu 5027a8bfd4abSjeff.liu out_unlock: 5028a8bfd4abSjeff.liu mnt_drop_write_file(file); 5029a8bfd4abSjeff.liu return ret; 5030a8bfd4abSjeff.liu } 5031a8bfd4abSjeff.liu 50322eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 50332eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 50342eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 50352eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 50362eaa055fSJeff Mahoney 5037d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 50382eaa055fSJeff Mahoney { 50394d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 50402eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 50412eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 50422eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 50432eaa055fSJeff Mahoney }; 50442eaa055fSJeff Mahoney 50452eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 50462eaa055fSJeff Mahoney return -EFAULT; 50472eaa055fSJeff Mahoney 50482eaa055fSJeff Mahoney return 0; 50492eaa055fSJeff Mahoney } 50502eaa055fSJeff Mahoney 5051b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, 5052b929c1d8SMarcos Paulo de Souza void __user *arg) 50532eaa055fSJeff Mahoney { 50540b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 50552eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 50562eaa055fSJeff Mahoney 50572eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 50582eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 50592eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 50602eaa055fSJeff Mahoney 50612eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 50622eaa055fSJeff Mahoney return -EFAULT; 50632eaa055fSJeff Mahoney 50642eaa055fSJeff Mahoney return 0; 50652eaa055fSJeff Mahoney } 50662eaa055fSJeff Mahoney 50672ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 50683b02a68aSJeff Mahoney enum btrfs_feature_set set, 50692eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 50702eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 50712eaa055fSJeff Mahoney { 5072f10152bcSDavid Sterba const char *type = btrfs_feature_set_name(set); 50733b02a68aSJeff Mahoney char *names; 50742eaa055fSJeff Mahoney u64 disallowed, unsupported; 50752eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 50762eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 50772eaa055fSJeff Mahoney 50782eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 50792eaa055fSJeff Mahoney if (unsupported) { 50803b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 50813b02a68aSJeff Mahoney if (names) { 50820b246afaSJeff Mahoney btrfs_warn(fs_info, 50833b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 50843b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 50853b02a68aSJeff Mahoney kfree(names); 50863b02a68aSJeff Mahoney } else 50870b246afaSJeff Mahoney btrfs_warn(fs_info, 50882eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 50892eaa055fSJeff Mahoney type, unsupported); 50902eaa055fSJeff Mahoney return -EOPNOTSUPP; 50912eaa055fSJeff Mahoney } 50922eaa055fSJeff Mahoney 50932eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 50942eaa055fSJeff Mahoney if (disallowed) { 50953b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 50963b02a68aSJeff Mahoney if (names) { 50970b246afaSJeff Mahoney btrfs_warn(fs_info, 50983b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 50993b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 51003b02a68aSJeff Mahoney kfree(names); 51013b02a68aSJeff Mahoney } else 51020b246afaSJeff Mahoney btrfs_warn(fs_info, 51032eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 51042eaa055fSJeff Mahoney type, disallowed); 51052eaa055fSJeff Mahoney return -EPERM; 51062eaa055fSJeff Mahoney } 51072eaa055fSJeff Mahoney 51082eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 51092eaa055fSJeff Mahoney if (disallowed) { 51103b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 51113b02a68aSJeff Mahoney if (names) { 51120b246afaSJeff Mahoney btrfs_warn(fs_info, 51133b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 51143b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 51153b02a68aSJeff Mahoney kfree(names); 51163b02a68aSJeff Mahoney } else 51170b246afaSJeff Mahoney btrfs_warn(fs_info, 51182eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 51192eaa055fSJeff Mahoney type, disallowed); 51202eaa055fSJeff Mahoney return -EPERM; 51212eaa055fSJeff Mahoney } 51222eaa055fSJeff Mahoney 51232eaa055fSJeff Mahoney return 0; 51242eaa055fSJeff Mahoney } 51252eaa055fSJeff Mahoney 51262ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 51272ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 51282eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 51292eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 51302eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 51312eaa055fSJeff Mahoney 51322eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 51332eaa055fSJeff Mahoney { 51340b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51350b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51360b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51370b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 51382eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 51392eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 51402eaa055fSJeff Mahoney u64 newflags; 51412eaa055fSJeff Mahoney int ret; 51422eaa055fSJeff Mahoney 51432eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 51442eaa055fSJeff Mahoney return -EPERM; 51452eaa055fSJeff Mahoney 51462eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 51472eaa055fSJeff Mahoney return -EFAULT; 51482eaa055fSJeff Mahoney 51492eaa055fSJeff Mahoney /* Nothing to do */ 51502eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 51512eaa055fSJeff Mahoney !flags[0].incompat_flags) 51522eaa055fSJeff Mahoney return 0; 51532eaa055fSJeff Mahoney 51542ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 51552eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 51562eaa055fSJeff Mahoney if (ret) 51572eaa055fSJeff Mahoney return ret; 51582eaa055fSJeff Mahoney 51592ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 51602eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 51612eaa055fSJeff Mahoney if (ret) 51622eaa055fSJeff Mahoney return ret; 51632eaa055fSJeff Mahoney 51642ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 51652eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 51662eaa055fSJeff Mahoney if (ret) 51672eaa055fSJeff Mahoney return ret; 51682eaa055fSJeff Mahoney 51697ab19625SDavid Sterba ret = mnt_want_write_file(file); 51707ab19625SDavid Sterba if (ret) 51717ab19625SDavid Sterba return ret; 51727ab19625SDavid Sterba 51738051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 51747ab19625SDavid Sterba if (IS_ERR(trans)) { 51757ab19625SDavid Sterba ret = PTR_ERR(trans); 51767ab19625SDavid Sterba goto out_drop_write; 51777ab19625SDavid Sterba } 51782eaa055fSJeff Mahoney 51790b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 51802eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 51812eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 51822eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 51832eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 51842eaa055fSJeff Mahoney 51852eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 51862eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 51872eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 51882eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 51892eaa055fSJeff Mahoney 51902eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 51912eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 51922eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 51932eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 51940b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 51952eaa055fSJeff Mahoney 51963a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 51977ab19625SDavid Sterba out_drop_write: 51987ab19625SDavid Sterba mnt_drop_write_file(file); 51997ab19625SDavid Sterba 52007ab19625SDavid Sterba return ret; 52012eaa055fSJeff Mahoney } 52022eaa055fSJeff Mahoney 52039ad12305SSahil Kang static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat) 52042351f431SJosef Bacik { 52052351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 52062351f431SJosef Bacik int ret; 52072351f431SJosef Bacik 52082351f431SJosef Bacik if (compat) { 52092351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 52102351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 52112351f431SJosef Bacik 52122351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 52132351f431SJosef Bacik if (ret) 52142351f431SJosef Bacik return -EFAULT; 52152351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 52162351f431SJosef Bacik if (!arg) 52172351f431SJosef Bacik return -ENOMEM; 52182351f431SJosef Bacik arg->send_fd = args32.send_fd; 52192351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 52202351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 52212351f431SJosef Bacik arg->parent_root = args32.parent_root; 52222351f431SJosef Bacik arg->flags = args32.flags; 52232351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 52242351f431SJosef Bacik sizeof(args32.reserved)); 52252351f431SJosef Bacik #else 52262351f431SJosef Bacik return -ENOTTY; 52272351f431SJosef Bacik #endif 52282351f431SJosef Bacik } else { 52292351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 52302351f431SJosef Bacik if (IS_ERR(arg)) 52312351f431SJosef Bacik return PTR_ERR(arg); 52322351f431SJosef Bacik } 52339ad12305SSahil Kang ret = btrfs_ioctl_send(inode, arg); 52342351f431SJosef Bacik kfree(arg); 52352351f431SJosef Bacik return ret; 52362351f431SJosef Bacik } 52372351f431SJosef Bacik 52381881fba8SOmar Sandoval static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp, 52391881fba8SOmar Sandoval bool compat) 52401881fba8SOmar Sandoval { 52411881fba8SOmar Sandoval struct btrfs_ioctl_encoded_io_args args = { 0 }; 52421881fba8SOmar Sandoval size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args, 52431881fba8SOmar Sandoval flags); 52441881fba8SOmar Sandoval size_t copy_end; 52451881fba8SOmar Sandoval struct iovec iovstack[UIO_FASTIOV]; 52461881fba8SOmar Sandoval struct iovec *iov = iovstack; 52471881fba8SOmar Sandoval struct iov_iter iter; 52481881fba8SOmar Sandoval loff_t pos; 52491881fba8SOmar Sandoval struct kiocb kiocb; 52501881fba8SOmar Sandoval ssize_t ret; 52511881fba8SOmar Sandoval 52521881fba8SOmar Sandoval if (!capable(CAP_SYS_ADMIN)) { 52531881fba8SOmar Sandoval ret = -EPERM; 52541881fba8SOmar Sandoval goto out_acct; 52551881fba8SOmar Sandoval } 52561881fba8SOmar Sandoval 52571881fba8SOmar Sandoval if (compat) { 52581881fba8SOmar Sandoval #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 52591881fba8SOmar Sandoval struct btrfs_ioctl_encoded_io_args_32 args32; 52601881fba8SOmar Sandoval 52611881fba8SOmar Sandoval copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32, 52621881fba8SOmar Sandoval flags); 52631881fba8SOmar Sandoval if (copy_from_user(&args32, argp, copy_end)) { 52641881fba8SOmar Sandoval ret = -EFAULT; 52651881fba8SOmar Sandoval goto out_acct; 52661881fba8SOmar Sandoval } 52671881fba8SOmar Sandoval args.iov = compat_ptr(args32.iov); 52681881fba8SOmar Sandoval args.iovcnt = args32.iovcnt; 52691881fba8SOmar Sandoval args.offset = args32.offset; 52701881fba8SOmar Sandoval args.flags = args32.flags; 52711881fba8SOmar Sandoval #else 52721881fba8SOmar Sandoval return -ENOTTY; 52731881fba8SOmar Sandoval #endif 52741881fba8SOmar Sandoval } else { 52751881fba8SOmar Sandoval copy_end = copy_end_kernel; 52761881fba8SOmar Sandoval if (copy_from_user(&args, argp, copy_end)) { 52771881fba8SOmar Sandoval ret = -EFAULT; 52781881fba8SOmar Sandoval goto out_acct; 52791881fba8SOmar Sandoval } 52801881fba8SOmar Sandoval } 52811881fba8SOmar Sandoval if (args.flags != 0) { 52821881fba8SOmar Sandoval ret = -EINVAL; 52831881fba8SOmar Sandoval goto out_acct; 52841881fba8SOmar Sandoval } 52851881fba8SOmar Sandoval 5286*de4eda9dSAl Viro ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack), 52871881fba8SOmar Sandoval &iov, &iter); 52881881fba8SOmar Sandoval if (ret < 0) 52891881fba8SOmar Sandoval goto out_acct; 52901881fba8SOmar Sandoval 52911881fba8SOmar Sandoval if (iov_iter_count(&iter) == 0) { 52921881fba8SOmar Sandoval ret = 0; 52931881fba8SOmar Sandoval goto out_iov; 52941881fba8SOmar Sandoval } 52951881fba8SOmar Sandoval pos = args.offset; 52961881fba8SOmar Sandoval ret = rw_verify_area(READ, file, &pos, args.len); 52971881fba8SOmar Sandoval if (ret < 0) 52981881fba8SOmar Sandoval goto out_iov; 52991881fba8SOmar Sandoval 53001881fba8SOmar Sandoval init_sync_kiocb(&kiocb, file); 53011881fba8SOmar Sandoval kiocb.ki_pos = pos; 53021881fba8SOmar Sandoval 53031881fba8SOmar Sandoval ret = btrfs_encoded_read(&kiocb, &iter, &args); 53041881fba8SOmar Sandoval if (ret >= 0) { 53051881fba8SOmar Sandoval fsnotify_access(file); 53061881fba8SOmar Sandoval if (copy_to_user(argp + copy_end, 53071881fba8SOmar Sandoval (char *)&args + copy_end_kernel, 53081881fba8SOmar Sandoval sizeof(args) - copy_end_kernel)) 53091881fba8SOmar Sandoval ret = -EFAULT; 53101881fba8SOmar Sandoval } 53111881fba8SOmar Sandoval 53121881fba8SOmar Sandoval out_iov: 53131881fba8SOmar Sandoval kfree(iov); 53141881fba8SOmar Sandoval out_acct: 53151881fba8SOmar Sandoval if (ret > 0) 53161881fba8SOmar Sandoval add_rchar(current, ret); 53171881fba8SOmar Sandoval inc_syscr(current); 53181881fba8SOmar Sandoval return ret; 53191881fba8SOmar Sandoval } 53201881fba8SOmar Sandoval 53217c0c7269SOmar Sandoval static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat) 53227c0c7269SOmar Sandoval { 53237c0c7269SOmar Sandoval struct btrfs_ioctl_encoded_io_args args; 53247c0c7269SOmar Sandoval struct iovec iovstack[UIO_FASTIOV]; 53257c0c7269SOmar Sandoval struct iovec *iov = iovstack; 53267c0c7269SOmar Sandoval struct iov_iter iter; 53277c0c7269SOmar Sandoval loff_t pos; 53287c0c7269SOmar Sandoval struct kiocb kiocb; 53297c0c7269SOmar Sandoval ssize_t ret; 53307c0c7269SOmar Sandoval 53317c0c7269SOmar Sandoval if (!capable(CAP_SYS_ADMIN)) { 53327c0c7269SOmar Sandoval ret = -EPERM; 53337c0c7269SOmar Sandoval goto out_acct; 53347c0c7269SOmar Sandoval } 53357c0c7269SOmar Sandoval 53367c0c7269SOmar Sandoval if (!(file->f_mode & FMODE_WRITE)) { 53377c0c7269SOmar Sandoval ret = -EBADF; 53387c0c7269SOmar Sandoval goto out_acct; 53397c0c7269SOmar Sandoval } 53407c0c7269SOmar Sandoval 53417c0c7269SOmar Sandoval if (compat) { 53427c0c7269SOmar Sandoval #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 53437c0c7269SOmar Sandoval struct btrfs_ioctl_encoded_io_args_32 args32; 53447c0c7269SOmar Sandoval 53457c0c7269SOmar Sandoval if (copy_from_user(&args32, argp, sizeof(args32))) { 53467c0c7269SOmar Sandoval ret = -EFAULT; 53477c0c7269SOmar Sandoval goto out_acct; 53487c0c7269SOmar Sandoval } 53497c0c7269SOmar Sandoval args.iov = compat_ptr(args32.iov); 53507c0c7269SOmar Sandoval args.iovcnt = args32.iovcnt; 53517c0c7269SOmar Sandoval args.offset = args32.offset; 53527c0c7269SOmar Sandoval args.flags = args32.flags; 53537c0c7269SOmar Sandoval args.len = args32.len; 53547c0c7269SOmar Sandoval args.unencoded_len = args32.unencoded_len; 53557c0c7269SOmar Sandoval args.unencoded_offset = args32.unencoded_offset; 53567c0c7269SOmar Sandoval args.compression = args32.compression; 53577c0c7269SOmar Sandoval args.encryption = args32.encryption; 53587c0c7269SOmar Sandoval memcpy(args.reserved, args32.reserved, sizeof(args.reserved)); 53597c0c7269SOmar Sandoval #else 53607c0c7269SOmar Sandoval return -ENOTTY; 53617c0c7269SOmar Sandoval #endif 53627c0c7269SOmar Sandoval } else { 53637c0c7269SOmar Sandoval if (copy_from_user(&args, argp, sizeof(args))) { 53647c0c7269SOmar Sandoval ret = -EFAULT; 53657c0c7269SOmar Sandoval goto out_acct; 53667c0c7269SOmar Sandoval } 53677c0c7269SOmar Sandoval } 53687c0c7269SOmar Sandoval 53697c0c7269SOmar Sandoval ret = -EINVAL; 53707c0c7269SOmar Sandoval if (args.flags != 0) 53717c0c7269SOmar Sandoval goto out_acct; 53727c0c7269SOmar Sandoval if (memchr_inv(args.reserved, 0, sizeof(args.reserved))) 53737c0c7269SOmar Sandoval goto out_acct; 53747c0c7269SOmar Sandoval if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE && 53757c0c7269SOmar Sandoval args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE) 53767c0c7269SOmar Sandoval goto out_acct; 53777c0c7269SOmar Sandoval if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES || 53787c0c7269SOmar Sandoval args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES) 53797c0c7269SOmar Sandoval goto out_acct; 53807c0c7269SOmar Sandoval if (args.unencoded_offset > args.unencoded_len) 53817c0c7269SOmar Sandoval goto out_acct; 53827c0c7269SOmar Sandoval if (args.len > args.unencoded_len - args.unencoded_offset) 53837c0c7269SOmar Sandoval goto out_acct; 53847c0c7269SOmar Sandoval 5385*de4eda9dSAl Viro ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack), 53867c0c7269SOmar Sandoval &iov, &iter); 53877c0c7269SOmar Sandoval if (ret < 0) 53887c0c7269SOmar Sandoval goto out_acct; 53897c0c7269SOmar Sandoval 53907c0c7269SOmar Sandoval file_start_write(file); 53917c0c7269SOmar Sandoval 53927c0c7269SOmar Sandoval if (iov_iter_count(&iter) == 0) { 53937c0c7269SOmar Sandoval ret = 0; 53947c0c7269SOmar Sandoval goto out_end_write; 53957c0c7269SOmar Sandoval } 53967c0c7269SOmar Sandoval pos = args.offset; 53977c0c7269SOmar Sandoval ret = rw_verify_area(WRITE, file, &pos, args.len); 53987c0c7269SOmar Sandoval if (ret < 0) 53997c0c7269SOmar Sandoval goto out_end_write; 54007c0c7269SOmar Sandoval 54017c0c7269SOmar Sandoval init_sync_kiocb(&kiocb, file); 54027c0c7269SOmar Sandoval ret = kiocb_set_rw_flags(&kiocb, 0); 54037c0c7269SOmar Sandoval if (ret) 54047c0c7269SOmar Sandoval goto out_end_write; 54057c0c7269SOmar Sandoval kiocb.ki_pos = pos; 54067c0c7269SOmar Sandoval 54077c0c7269SOmar Sandoval ret = btrfs_do_write_iter(&kiocb, &iter, &args); 54087c0c7269SOmar Sandoval if (ret > 0) 54097c0c7269SOmar Sandoval fsnotify_modify(file); 54107c0c7269SOmar Sandoval 54117c0c7269SOmar Sandoval out_end_write: 54127c0c7269SOmar Sandoval file_end_write(file); 54137c0c7269SOmar Sandoval kfree(iov); 54147c0c7269SOmar Sandoval out_acct: 54157c0c7269SOmar Sandoval if (ret > 0) 54167c0c7269SOmar Sandoval add_wchar(current, ret); 54177c0c7269SOmar Sandoval inc_syscw(current); 54187c0c7269SOmar Sandoval return ret; 54197c0c7269SOmar Sandoval } 54207c0c7269SOmar Sandoval 5421f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5422f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5423f46b5a66SChristoph Hellwig { 54240b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54250b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54260b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54274bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5428f46b5a66SChristoph Hellwig 5429f46b5a66SChristoph Hellwig switch (cmd) { 54306cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 5431dc408ccdSSahil Kang return btrfs_ioctl_getversion(inode, argp); 543240cf931fSEric Sandeen case FS_IOC_GETFSLABEL: 5433b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_fslabel(fs_info, argp); 543440cf931fSEric Sandeen case FS_IOC_SETFSLABEL: 543540cf931fSEric Sandeen return btrfs_ioctl_set_fslabel(file, argp); 5436f7039b1dSLi Dongyang case FITRIM: 5437b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_fitrim(fs_info, argp); 5438f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5439fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5440fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5441fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 54423de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5443fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 54446f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 54456f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 544676dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 5447949964c9SMarcos Paulo de Souza return btrfs_ioctl_snap_destroy(file, argp, false); 5448949964c9SMarcos Paulo de Souza case BTRFS_IOC_SNAP_DESTROY_V2: 5449949964c9SMarcos Paulo de Souza return btrfs_ioctl_snap_destroy(file, argp, true); 54500caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 54519ad12305SSahil Kang return btrfs_ioctl_subvol_getflags(inode, argp); 54520caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 54530caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 54546ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 54556ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5456f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 54571e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 54581e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 54591e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5460f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5461198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5462f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 54632ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5464f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5465da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 54666b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 54676b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5468475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 54692ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5470475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 54712ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5472ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 54739ad12305SSahil Kang return btrfs_ioctl_tree_search(inode, argp); 5474cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 54759ad12305SSahil Kang return btrfs_ioctl_tree_search_v2(inode, argp); 5476ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5477dc408ccdSSahil Kang return btrfs_ioctl_ino_lookup(root, argp); 5478d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5479d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5480d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5481d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5482d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5483d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 54841406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 54852ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 54869b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 54879b199859SFilipe David Borba Manana int ret; 54889b199859SFilipe David Borba Manana 54899db4dc24SNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false); 54909b199859SFilipe David Borba Manana if (ret) 54919b199859SFilipe David Borba Manana return ret; 54920b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 54932fad4e83SDavid Sterba /* 54942fad4e83SDavid Sterba * The transaction thread may want to do more work, 549501327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 54962fad4e83SDavid Sterba * processing uncleaned subvols. 54972fad4e83SDavid Sterba */ 54980b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 54999b199859SFilipe David Borba Manana return ret; 55009b199859SFilipe David Borba Manana } 550146204592SSage Weil case BTRFS_IOC_START_SYNC: 55029a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 550346204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55042ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5505475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5506b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5507475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 55082ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5509475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 55102ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5511c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55129ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5513837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 55142ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 551519a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 55162ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 55178ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55188ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5519abccd00fSHugo Mills #ifdef CONFIG_64BIT 5520abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5521abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5522abccd00fSHugo Mills #endif 552331db9f7cSAlexander Block case BTRFS_IOC_SEND: 55249ad12305SSahil Kang return _btrfs_ioctl_send(inode, argp, false); 55252351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55262351f431SJosef Bacik case BTRFS_IOC_SEND_32: 55279ad12305SSahil Kang return _btrfs_ioctl_send(inode, argp, true); 55282351f431SJosef Bacik #endif 5529c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 55302ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 55315d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5532905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55335d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5534905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55355d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5536905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55375d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5538905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 55392f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 55402f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 55412f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 5542b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_status(fs_info, argp); 554357254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 5544b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_wait(fs_info, argp); 55453f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 55462ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 55472eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5548d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 55492eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 5550b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_features(fs_info, argp); 55512eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 55522eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5553b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 55549ad12305SSahil Kang return btrfs_ioctl_get_subvol_info(inode, argp); 555542e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 5556dc408ccdSSahil Kang return btrfs_ioctl_get_subvol_rootref(root, argp); 555723d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 555823d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 555914605409SBoris Burkov case FS_IOC_ENABLE_VERITY: 556014605409SBoris Burkov return fsverity_ioctl_enable(file, (const void __user *)argp); 556114605409SBoris Burkov case FS_IOC_MEASURE_VERITY: 556214605409SBoris Burkov return fsverity_ioctl_measure(file, argp); 55631881fba8SOmar Sandoval case BTRFS_IOC_ENCODED_READ: 55641881fba8SOmar Sandoval return btrfs_ioctl_encoded_read(file, argp, false); 55657c0c7269SOmar Sandoval case BTRFS_IOC_ENCODED_WRITE: 55667c0c7269SOmar Sandoval return btrfs_ioctl_encoded_write(file, argp, false); 55671881fba8SOmar Sandoval #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55681881fba8SOmar Sandoval case BTRFS_IOC_ENCODED_READ_32: 55691881fba8SOmar Sandoval return btrfs_ioctl_encoded_read(file, argp, true); 55707c0c7269SOmar Sandoval case BTRFS_IOC_ENCODED_WRITE_32: 55717c0c7269SOmar Sandoval return btrfs_ioctl_encoded_write(file, argp, true); 55721881fba8SOmar Sandoval #endif 5573f46b5a66SChristoph Hellwig } 5574f46b5a66SChristoph Hellwig 5575f46b5a66SChristoph Hellwig return -ENOTTY; 5576f46b5a66SChristoph Hellwig } 55774c63c245SLuke Dashjr 55784c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 55794c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 55804c63c245SLuke Dashjr { 55812a362249SJeff Mahoney /* 55822a362249SJeff Mahoney * These all access 32-bit values anyway so no further 55832a362249SJeff Mahoney * handling is necessary. 55842a362249SJeff Mahoney */ 55854c63c245SLuke Dashjr switch (cmd) { 55864c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 55874c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 55884c63c245SLuke Dashjr break; 55894c63c245SLuke Dashjr } 55904c63c245SLuke Dashjr 55914c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 55924c63c245SLuke Dashjr } 55934c63c245SLuke Dashjr #endif 5594