1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 2f46b5a66SChristoph Hellwig /* 3f46b5a66SChristoph Hellwig * Copyright (C) 2007 Oracle. All rights reserved. 4f46b5a66SChristoph Hellwig */ 5f46b5a66SChristoph Hellwig 6f46b5a66SChristoph Hellwig #include <linux/kernel.h> 7f46b5a66SChristoph Hellwig #include <linux/bio.h> 8f46b5a66SChristoph Hellwig #include <linux/file.h> 9f46b5a66SChristoph Hellwig #include <linux/fs.h> 10cb8e7090SChristoph Hellwig #include <linux/fsnotify.h> 11f46b5a66SChristoph Hellwig #include <linux/pagemap.h> 12f46b5a66SChristoph Hellwig #include <linux/highmem.h> 13f46b5a66SChristoph Hellwig #include <linux/time.h> 14f46b5a66SChristoph Hellwig #include <linux/string.h> 15f46b5a66SChristoph Hellwig #include <linux/backing-dev.h> 16cb8e7090SChristoph Hellwig #include <linux/mount.h> 17cb8e7090SChristoph Hellwig #include <linux/namei.h> 18f46b5a66SChristoph Hellwig #include <linux/writeback.h> 19f46b5a66SChristoph Hellwig #include <linux/compat.h> 20cb8e7090SChristoph Hellwig #include <linux/security.h> 21f46b5a66SChristoph Hellwig #include <linux/xattr.h> 22f54de068SDavid Sterba #include <linux/mm.h> 235a0e3ad6STejun Heo #include <linux/slab.h> 24f7039b1dSLi Dongyang #include <linux/blkdev.h> 258ea05e3aSAlexander Block #include <linux/uuid.h> 2655e301fdSFilipe Brandenburger #include <linux/btrfs.h> 27416161dbSMark Fasheh #include <linux/uaccess.h> 28ae5e165dSJeff Layton #include <linux/iversion.h> 2997fc2977SMiklos Szeredi #include <linux/fileattr.h> 3014605409SBoris Burkov #include <linux/fsverity.h> 31f46b5a66SChristoph Hellwig #include "ctree.h" 32f46b5a66SChristoph Hellwig #include "disk-io.h" 33949964c9SMarcos Paulo de Souza #include "export.h" 34f46b5a66SChristoph Hellwig #include "transaction.h" 35f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 36f46b5a66SChristoph Hellwig #include "print-tree.h" 37f46b5a66SChristoph Hellwig #include "volumes.h" 38925baeddSChris Mason #include "locking.h" 39d7728c96SJan Schmidt #include "backref.h" 40606686eeSJosef Bacik #include "rcu-string.h" 4131db9f7cSAlexander Block #include "send.h" 423f6bcfbdSStefan Behrens #include "dev-replace.h" 4363541927SFilipe David Borba Manana #include "props.h" 443b02a68aSJeff Mahoney #include "sysfs.h" 45fcebe456SJosef Bacik #include "qgroup.h" 461ec9a1aeSFilipe Manana #include "tree-log.h" 47ebb8765bSAnand Jain #include "compression.h" 488719aaaeSJosef Bacik #include "space-info.h" 4986736342SJosef Bacik #include "delalloc-space.h" 50aac0023cSJosef Bacik #include "block-group.h" 5122b398eeSQu Wenruo #include "subpage.h" 52f46b5a66SChristoph Hellwig 53abccd00fSHugo Mills #ifdef CONFIG_64BIT 54abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 55abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 56abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 57abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 58abccd00fSHugo Mills */ 59abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 60abccd00fSHugo Mills __u64 sec; 61abccd00fSHugo Mills __u32 nsec; 62abccd00fSHugo Mills } __attribute__ ((__packed__)); 63abccd00fSHugo Mills 64abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 65abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 66abccd00fSHugo Mills __u64 stransid; /* in */ 67abccd00fSHugo Mills __u64 rtransid; /* out */ 68abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 69abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 70abccd00fSHugo Mills __u64 flags; /* in */ 71abccd00fSHugo Mills __u64 reserved[16]; /* in */ 72abccd00fSHugo Mills } __attribute__ ((__packed__)); 73abccd00fSHugo Mills 74abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 75abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 76abccd00fSHugo Mills #endif 77abccd00fSHugo Mills 782351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 792351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 802351f431SJosef Bacik __s64 send_fd; /* in */ 812351f431SJosef Bacik __u64 clone_sources_count; /* in */ 822351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 832351f431SJosef Bacik __u64 parent_root; /* in */ 842351f431SJosef Bacik __u64 flags; /* in */ 852351f431SJosef Bacik __u64 reserved[4]; /* in */ 862351f431SJosef Bacik } __attribute__ ((__packed__)); 872351f431SJosef Bacik 882351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 892351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 902351f431SJosef Bacik #endif 91abccd00fSHugo Mills 926cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 931905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 941905a0f7SDavid Sterba unsigned int flags) 956cbff00fSChristoph Hellwig { 961905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 976cbff00fSChristoph Hellwig return flags; 981905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 996cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1006cbff00fSChristoph Hellwig else 1016cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1026cbff00fSChristoph Hellwig } 103f46b5a66SChristoph Hellwig 1046cbff00fSChristoph Hellwig /* 105a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 106a157d4fdSDavid Sterba * ioctl. 1076cbff00fSChristoph Hellwig */ 10877eea05eSBoris Burkov static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode) 1096cbff00fSChristoph Hellwig { 1106cbff00fSChristoph Hellwig unsigned int iflags = 0; 11177eea05eSBoris Burkov u32 flags = binode->flags; 11214605409SBoris Burkov u32 ro_flags = binode->ro_flags; 1136cbff00fSChristoph Hellwig 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1156cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1176cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1196cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1216cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1226cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1236cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1246cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1256cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 126d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 127d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 12814605409SBoris Burkov if (ro_flags & BTRFS_INODE_RO_VERITY) 12914605409SBoris Burkov iflags |= FS_VERITY_FL; 130d0092bddSLi Zefan 13113f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 132d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 13313f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13413f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1356cbff00fSChristoph Hellwig 1366cbff00fSChristoph Hellwig return iflags; 1376cbff00fSChristoph Hellwig } 1386cbff00fSChristoph Hellwig 1396cbff00fSChristoph Hellwig /* 1406cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1416cbff00fSChristoph Hellwig */ 1427b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1436cbff00fSChristoph Hellwig { 1445c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1453cc79392SFilipe Manana unsigned int new_fl = 0; 1466cbff00fSChristoph Hellwig 1475c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1483cc79392SFilipe Manana new_fl |= S_SYNC; 1495c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1503cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1515c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1523cc79392SFilipe Manana new_fl |= S_APPEND; 1535c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1543cc79392SFilipe Manana new_fl |= S_NOATIME; 1555c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1563cc79392SFilipe Manana new_fl |= S_DIRSYNC; 15714605409SBoris Burkov if (binode->ro_flags & BTRFS_INODE_RO_VERITY) 15814605409SBoris Burkov new_fl |= S_VERITY; 1593cc79392SFilipe Manana 1603cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 16114605409SBoris Burkov S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC | 16214605409SBoris Burkov S_VERITY, new_fl); 1636cbff00fSChristoph Hellwig } 1646cbff00fSChristoph Hellwig 165f37c563bSDavid Sterba /* 166f37c563bSDavid Sterba * Check if @flags are a supported and valid set of FS_*_FL flags and that 167f37c563bSDavid Sterba * the old and new flags are not conflicting 168f37c563bSDavid Sterba */ 169f37c563bSDavid Sterba static int check_fsflags(unsigned int old_flags, unsigned int flags) 17075e7cb7fSLiu Bo { 17175e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17275e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17375e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 174e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 175e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17675e7cb7fSLiu Bo return -EOPNOTSUPP; 17775e7cb7fSLiu Bo 178f37c563bSDavid Sterba /* COMPR and NOCOMP on new/old are valid */ 17975e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 18075e7cb7fSLiu Bo return -EINVAL; 18175e7cb7fSLiu Bo 182f37c563bSDavid Sterba if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL)) 183f37c563bSDavid Sterba return -EINVAL; 184f37c563bSDavid Sterba 185f37c563bSDavid Sterba /* NOCOW and compression options are mutually exclusive */ 186f37c563bSDavid Sterba if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL))) 187f37c563bSDavid Sterba return -EINVAL; 188f37c563bSDavid Sterba if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL))) 189f37c563bSDavid Sterba return -EINVAL; 190f37c563bSDavid Sterba 19175e7cb7fSLiu Bo return 0; 19275e7cb7fSLiu Bo } 19375e7cb7fSLiu Bo 194d206e9c9SNaohiro Aota static int check_fsflags_compatible(struct btrfs_fs_info *fs_info, 195d206e9c9SNaohiro Aota unsigned int flags) 196d206e9c9SNaohiro Aota { 197d206e9c9SNaohiro Aota if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL)) 198d206e9c9SNaohiro Aota return -EPERM; 199d206e9c9SNaohiro Aota 200d206e9c9SNaohiro Aota return 0; 201d206e9c9SNaohiro Aota } 202d206e9c9SNaohiro Aota 20397fc2977SMiklos Szeredi /* 20497fc2977SMiklos Szeredi * Set flags/xflags from the internal inode flags. The remaining items of 20597fc2977SMiklos Szeredi * fsxattr are zeroed. 20697fc2977SMiklos Szeredi */ 20797fc2977SMiklos Szeredi int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa) 2086cbff00fSChristoph Hellwig { 20997fc2977SMiklos Szeredi struct btrfs_inode *binode = BTRFS_I(d_inode(dentry)); 21097fc2977SMiklos Szeredi 21177eea05eSBoris Burkov fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode)); 21297fc2977SMiklos Szeredi return 0; 21397fc2977SMiklos Szeredi } 21497fc2977SMiklos Szeredi 21597fc2977SMiklos Szeredi int btrfs_fileattr_set(struct user_namespace *mnt_userns, 21697fc2977SMiklos Szeredi struct dentry *dentry, struct fileattr *fa) 21797fc2977SMiklos Szeredi { 21897fc2977SMiklos Szeredi struct inode *inode = d_inode(dentry); 2190b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2205c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 2215c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 2226cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 2235aca2842SDarrick J. Wong unsigned int fsflags, old_fsflags; 2246cbff00fSChristoph Hellwig int ret; 225ff9fef55SAnand Jain const char *comp = NULL; 226f37c563bSDavid Sterba u32 binode_flags; 2276cbff00fSChristoph Hellwig 228b83cc969SLi Zefan if (btrfs_root_readonly(root)) 229b83cc969SLi Zefan return -EROFS; 230b83cc969SLi Zefan 23197fc2977SMiklos Szeredi if (fileattr_has_fsx(fa)) 23297fc2977SMiklos Szeredi return -EOPNOTSUPP; 2336cbff00fSChristoph Hellwig 23497fc2977SMiklos Szeredi fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags); 23577eea05eSBoris Burkov old_fsflags = btrfs_inode_flags_to_fsflags(binode); 23697fc2977SMiklos Szeredi ret = check_fsflags(old_fsflags, fsflags); 237e7848683SJan Kara if (ret) 238e7848683SJan Kara return ret; 239e7848683SJan Kara 240d206e9c9SNaohiro Aota ret = check_fsflags_compatible(fs_info, fsflags); 241d206e9c9SNaohiro Aota if (ret) 24297fc2977SMiklos Szeredi return ret; 243d206e9c9SNaohiro Aota 244f37c563bSDavid Sterba binode_flags = binode->flags; 2455c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 246d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_SYNC; 2476cbff00fSChristoph Hellwig else 248d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_SYNC; 2495c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 250d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_IMMUTABLE; 2516cbff00fSChristoph Hellwig else 252d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_IMMUTABLE; 2535c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 254d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_APPEND; 2556cbff00fSChristoph Hellwig else 256d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_APPEND; 2575c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 258d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODUMP; 2596cbff00fSChristoph Hellwig else 260d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODUMP; 2615c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 262d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOATIME; 2636cbff00fSChristoph Hellwig else 264d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOATIME; 26597fc2977SMiklos Szeredi 26697fc2977SMiklos Szeredi /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */ 26797fc2977SMiklos Szeredi if (!fa->flags_valid) { 26897fc2977SMiklos Szeredi /* 1 item for the inode */ 26997fc2977SMiklos Szeredi trans = btrfs_start_transaction(root, 1); 2709b8a233bSRitesh Harjani if (IS_ERR(trans)) 2719b8a233bSRitesh Harjani return PTR_ERR(trans); 27297fc2977SMiklos Szeredi goto update_flags; 27397fc2977SMiklos Szeredi } 27497fc2977SMiklos Szeredi 2755c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 276d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_DIRSYNC; 2776cbff00fSChristoph Hellwig else 278d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_DIRSYNC; 2795c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 28044e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2817e97b8daSDavid Sterba /* 2827e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2837e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2847e97b8daSDavid Sterba * status of the file and will not set it. 2857e97b8daSDavid Sterba */ 2867e97b8daSDavid Sterba if (inode->i_size == 0) 287d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW | 288d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM; 2897e97b8daSDavid Sterba } else { 290d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW; 2917e97b8daSDavid Sterba } 2927e97b8daSDavid Sterba } else { 2937e97b8daSDavid Sterba /* 29401327610SNicholas D Steeves * Revert back under same assumptions as above 2957e97b8daSDavid Sterba */ 29644e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2977e97b8daSDavid Sterba if (inode->i_size == 0) 298d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_NODATACOW | 299d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM); 3007e97b8daSDavid Sterba } else { 301d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODATACOW; 3027e97b8daSDavid Sterba } 3037e97b8daSDavid Sterba } 3046cbff00fSChristoph Hellwig 30575e7cb7fSLiu Bo /* 30675e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 30775e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 30875e7cb7fSLiu Bo * things smaller. 30975e7cb7fSLiu Bo */ 3105c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 311d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_COMPRESS; 312d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOCOMPRESS; 3135c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 31463541927SFilipe David Borba Manana 31597fc2977SMiklos Szeredi if (IS_SWAPFILE(inode)) 31697fc2977SMiklos Szeredi return -ETXTBSY; 317eede2bf3SOmar Sandoval 318d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_COMPRESS; 319d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOCOMPRESS; 32063541927SFilipe David Borba Manana 32193370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 32293370509SDavid Sterba if (!comp || comp[0] == 0) 32393370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 324ebcb904dSLi Zefan } else { 325d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 32675e7cb7fSLiu Bo } 3276cbff00fSChristoph Hellwig 328ff9fef55SAnand Jain /* 329ff9fef55SAnand Jain * 1 for inode item 330ff9fef55SAnand Jain * 2 for properties 331ff9fef55SAnand Jain */ 332ff9fef55SAnand Jain trans = btrfs_start_transaction(root, 3); 33397fc2977SMiklos Szeredi if (IS_ERR(trans)) 33497fc2977SMiklos Szeredi return PTR_ERR(trans); 3356cbff00fSChristoph Hellwig 336ff9fef55SAnand Jain if (comp) { 337ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, 338ff9fef55SAnand Jain strlen(comp), 0); 339ff9fef55SAnand Jain if (ret) { 340ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 341ff9fef55SAnand Jain goto out_end_trans; 342ff9fef55SAnand Jain } 343ff9fef55SAnand Jain } else { 344ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 345ff9fef55SAnand Jain 0, 0); 346ff9fef55SAnand Jain if (ret && ret != -ENODATA) { 347ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 348ff9fef55SAnand Jain goto out_end_trans; 349ff9fef55SAnand Jain } 350ff9fef55SAnand Jain } 351ff9fef55SAnand Jain 35297fc2977SMiklos Szeredi update_flags: 353d2b8fcfeSAnand Jain binode->flags = binode_flags; 3547b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3550c4d2d95SJosef Bacik inode_inc_iversion(inode); 356c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3579a56fcd1SNikolay Borisov ret = btrfs_update_inode(trans, root, BTRFS_I(inode)); 3586cbff00fSChristoph Hellwig 359ff9fef55SAnand Jain out_end_trans: 3603a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3612d4e6f6aSliubo return ret; 3626cbff00fSChristoph Hellwig } 3636cbff00fSChristoph Hellwig 3640d7ed32cSDavid Sterba /* 3650d7ed32cSDavid Sterba * Start exclusive operation @type, return true on success 3660d7ed32cSDavid Sterba */ 367c3e1f96cSGoldwyn Rodrigues bool btrfs_exclop_start(struct btrfs_fs_info *fs_info, 368c3e1f96cSGoldwyn Rodrigues enum btrfs_exclusive_operation type) 369c3e1f96cSGoldwyn Rodrigues { 3700d7ed32cSDavid Sterba bool ret = false; 3710d7ed32cSDavid Sterba 3720d7ed32cSDavid Sterba spin_lock(&fs_info->super_lock); 3730d7ed32cSDavid Sterba if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) { 3740d7ed32cSDavid Sterba fs_info->exclusive_operation = type; 3750d7ed32cSDavid Sterba ret = true; 3760d7ed32cSDavid Sterba } 3770d7ed32cSDavid Sterba spin_unlock(&fs_info->super_lock); 3780d7ed32cSDavid Sterba 3790d7ed32cSDavid Sterba return ret; 380c3e1f96cSGoldwyn Rodrigues } 381c3e1f96cSGoldwyn Rodrigues 382578bda9eSDavid Sterba /* 383578bda9eSDavid Sterba * Conditionally allow to enter the exclusive operation in case it's compatible 384578bda9eSDavid Sterba * with the running one. This must be paired with btrfs_exclop_start_unlock and 385578bda9eSDavid Sterba * btrfs_exclop_finish. 386578bda9eSDavid Sterba * 387578bda9eSDavid Sterba * Compatibility: 388578bda9eSDavid Sterba * - the same type is already running 389578bda9eSDavid Sterba * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller 390578bda9eSDavid Sterba * must check the condition first that would allow none -> @type 391578bda9eSDavid Sterba */ 392578bda9eSDavid Sterba bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info, 393578bda9eSDavid Sterba enum btrfs_exclusive_operation type) 394578bda9eSDavid Sterba { 395578bda9eSDavid Sterba spin_lock(&fs_info->super_lock); 396578bda9eSDavid Sterba if (fs_info->exclusive_operation == type) 397578bda9eSDavid Sterba return true; 398578bda9eSDavid Sterba 399578bda9eSDavid Sterba spin_unlock(&fs_info->super_lock); 400578bda9eSDavid Sterba return false; 401578bda9eSDavid Sterba } 402578bda9eSDavid Sterba 403578bda9eSDavid Sterba void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info) 404578bda9eSDavid Sterba { 405578bda9eSDavid Sterba spin_unlock(&fs_info->super_lock); 406578bda9eSDavid Sterba } 407578bda9eSDavid Sterba 408c3e1f96cSGoldwyn Rodrigues void btrfs_exclop_finish(struct btrfs_fs_info *fs_info) 409c3e1f96cSGoldwyn Rodrigues { 4100d7ed32cSDavid Sterba spin_lock(&fs_info->super_lock); 411c3e1f96cSGoldwyn Rodrigues WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE); 4120d7ed32cSDavid Sterba spin_unlock(&fs_info->super_lock); 41366a2823cSGoldwyn Rodrigues sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation"); 414c3e1f96cSGoldwyn Rodrigues } 415c3e1f96cSGoldwyn Rodrigues 4166cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4176cbff00fSChristoph Hellwig { 418496ad9aaSAl Viro struct inode *inode = file_inode(file); 4196cbff00fSChristoph Hellwig 4206cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4216cbff00fSChristoph Hellwig } 422f46b5a66SChristoph Hellwig 423b929c1d8SMarcos Paulo de Souza static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, 424b929c1d8SMarcos Paulo de Souza void __user *arg) 425f7039b1dSLi Dongyang { 426f7039b1dSLi Dongyang struct btrfs_device *device; 427f7039b1dSLi Dongyang struct request_queue *q; 428f7039b1dSLi Dongyang struct fstrim_range range; 429f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 430f7039b1dSLi Dongyang u64 num_devices = 0; 431f7039b1dSLi Dongyang int ret; 432f7039b1dSLi Dongyang 433f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 434f7039b1dSLi Dongyang return -EPERM; 435f7039b1dSLi Dongyang 436f35f06c3SFilipe Manana /* 4371cb3dc3fSNaohiro Aota * btrfs_trim_block_group() depends on space cache, which is not 4381cb3dc3fSNaohiro Aota * available in zoned filesystem. So, disallow fitrim on a zoned 4391cb3dc3fSNaohiro Aota * filesystem for now. 4401cb3dc3fSNaohiro Aota */ 4411cb3dc3fSNaohiro Aota if (btrfs_is_zoned(fs_info)) 4421cb3dc3fSNaohiro Aota return -EOPNOTSUPP; 4431cb3dc3fSNaohiro Aota 4441cb3dc3fSNaohiro Aota /* 445f35f06c3SFilipe Manana * If the fs is mounted with nologreplay, which requires it to be 446f35f06c3SFilipe Manana * mounted in RO mode as well, we can not allow discard on free space 447f35f06c3SFilipe Manana * inside block groups, because log trees refer to extents that are not 448f35f06c3SFilipe Manana * pinned in a block group's free space cache (pinning the extents is 449f35f06c3SFilipe Manana * precisely the first phase of replaying a log tree). 450f35f06c3SFilipe Manana */ 451f35f06c3SFilipe Manana if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 452f35f06c3SFilipe Manana return -EROFS; 453f35f06c3SFilipe Manana 4541f78160cSXiao Guangrong rcu_read_lock(); 4551f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 4561f78160cSXiao Guangrong dev_list) { 457f7039b1dSLi Dongyang if (!device->bdev) 458f7039b1dSLi Dongyang continue; 459f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 460f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 461f7039b1dSLi Dongyang num_devices++; 46250d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 463f7039b1dSLi Dongyang minlen); 464f7039b1dSLi Dongyang } 465f7039b1dSLi Dongyang } 4661f78160cSXiao Guangrong rcu_read_unlock(); 467f4c697e6SLukas Czerner 468f7039b1dSLi Dongyang if (!num_devices) 469f7039b1dSLi Dongyang return -EOPNOTSUPP; 470f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 471f7039b1dSLi Dongyang return -EFAULT; 4726ba9fc8eSQu Wenruo 4736ba9fc8eSQu Wenruo /* 4746ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 4756ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 4766ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 4776ba9fc8eSQu Wenruo */ 4786ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 479f4c697e6SLukas Czerner return -EINVAL; 480f7039b1dSLi Dongyang 481f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 4822ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 483f7039b1dSLi Dongyang if (ret < 0) 484f7039b1dSLi Dongyang return ret; 485f7039b1dSLi Dongyang 486f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 487f7039b1dSLi Dongyang return -EFAULT; 488f7039b1dSLi Dongyang 489f7039b1dSLi Dongyang return 0; 490f7039b1dSLi Dongyang } 491f7039b1dSLi Dongyang 492e1f60a65SDavid Sterba int __pure btrfs_is_empty_uuid(u8 *uuid) 493dd5f9615SStefan Behrens { 49446e0f66aSChris Mason int i; 49546e0f66aSChris Mason 49646e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 49746e0f66aSChris Mason if (uuid[i]) 49846e0f66aSChris Mason return 0; 49946e0f66aSChris Mason } 50046e0f66aSChris Mason return 1; 501dd5f9615SStefan Behrens } 502dd5f9615SStefan Behrens 5034d4340c9SChristian Brauner static noinline int create_subvol(struct user_namespace *mnt_userns, 5044d4340c9SChristian Brauner struct inode *dir, struct dentry *dentry, 50552f75f4fSDavid Sterba const char *name, int namelen, 5068696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 507f46b5a66SChristoph Hellwig { 5080b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 509f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 510f46b5a66SChristoph Hellwig struct btrfs_key key; 51149a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 512f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 513f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 514d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 51576dda93cSYan, Zheng struct btrfs_root *new_root; 516d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 51795582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5185662344bSTsutomu Itoh struct inode *inode; 519f46b5a66SChristoph Hellwig int ret; 520f46b5a66SChristoph Hellwig int err; 5212dfb1e43SQu Wenruo dev_t anon_dev = 0; 522f46b5a66SChristoph Hellwig u64 objectid; 5233de4586cSChris Mason u64 index = 0; 524f46b5a66SChristoph Hellwig 52549a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 52649a3c4d9SDavid Sterba if (!root_item) 52749a3c4d9SDavid Sterba return -ENOMEM; 52849a3c4d9SDavid Sterba 529543068a2SNikolay Borisov ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid); 5302fbe8c8aSAl Viro if (ret) 53149a3c4d9SDavid Sterba goto fail_free; 5326a912213SJosef Bacik 5332dfb1e43SQu Wenruo ret = get_anon_bdev(&anon_dev); 5342dfb1e43SQu Wenruo if (ret < 0) 5352dfb1e43SQu Wenruo goto fail_free; 5362dfb1e43SQu Wenruo 537e09fe2d2SQu Wenruo /* 538e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 53901327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 540e09fe2d2SQu Wenruo */ 54149a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 54249a3c4d9SDavid Sterba ret = -ENOSPC; 54349a3c4d9SDavid Sterba goto fail_free; 54449a3c4d9SDavid Sterba } 545e09fe2d2SQu Wenruo 546d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 5479ed74f2dSJosef Bacik /* 548d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 549d5c12070SMiao Xie * of create_snapshot(). 5509ed74f2dSJosef Bacik */ 551c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 552d5c12070SMiao Xie if (ret) 55349a3c4d9SDavid Sterba goto fail_free; 554f46b5a66SChristoph Hellwig 555d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 556d5c12070SMiao Xie if (IS_ERR(trans)) { 557d5c12070SMiao Xie ret = PTR_ERR(trans); 558e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &block_rsv); 55949a3c4d9SDavid Sterba goto fail_free; 560d5c12070SMiao Xie } 561d5c12070SMiao Xie trans->block_rsv = &block_rsv; 562d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 563f46b5a66SChristoph Hellwig 564a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 5656f72c7e2SArne Jansen if (ret) 5666f72c7e2SArne Jansen goto fail; 5676f72c7e2SArne Jansen 5689631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 5699631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 5708e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 5718e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 5728e8a1e31SJosef Bacik goto fail; 5738e8a1e31SJosef Bacik } 574f46b5a66SChristoph Hellwig 575f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 576f46b5a66SChristoph Hellwig 57749a3c4d9SDavid Sterba inode_item = &root_item->inode; 5783cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 5793cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 5803cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 581da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 5820b246afaSJeff Mahoney fs_info->nodesize); 5833cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 584f46b5a66SChristoph Hellwig 58549a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 58649a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 5873cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 58808fe4db1SLi Zefan 58949a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 59049a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 59149a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 59249a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 59349a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 59449a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 595f46b5a66SChristoph Hellwig 59649a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 59749a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 598807fc790SAndy Shevchenko generate_random_guid(root_item->uuid); 59949a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 60049a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 60149a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 60249a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 60349a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 604f46b5a66SChristoph Hellwig 605925baeddSChris Mason btrfs_tree_unlock(leaf); 606f46b5a66SChristoph Hellwig 60769948022SNikolay Borisov btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID); 608f46b5a66SChristoph Hellwig 609f46b5a66SChristoph Hellwig key.objectid = objectid; 6105d4f98a2SYan Zheng key.offset = 0; 611962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6120b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 61349a3c4d9SDavid Sterba root_item); 61467addf29SFilipe Manana if (ret) { 61567addf29SFilipe Manana /* 61667addf29SFilipe Manana * Since we don't abort the transaction in this case, free the 61767addf29SFilipe Manana * tree block so that we don't leak space and leave the 61867addf29SFilipe Manana * filesystem in an inconsistent state (an extent item in the 61967addf29SFilipe Manana * extent tree without backreferences). Also no need to have 62067addf29SFilipe Manana * the tree block locked since it is not in any tree at this 62167addf29SFilipe Manana * point, so no other task can find it and use it. 62267addf29SFilipe Manana */ 62367addf29SFilipe Manana btrfs_free_tree_block(trans, root, leaf, 0, 1); 62467addf29SFilipe Manana free_extent_buffer(leaf); 625f46b5a66SChristoph Hellwig goto fail; 62667addf29SFilipe Manana } 62767addf29SFilipe Manana 62867addf29SFilipe Manana free_extent_buffer(leaf); 62967addf29SFilipe Manana leaf = NULL; 630f46b5a66SChristoph Hellwig 63176dda93cSYan, Zheng key.offset = (u64)-1; 6322dfb1e43SQu Wenruo new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev); 63379787eaaSJeff Mahoney if (IS_ERR(new_root)) { 6342dfb1e43SQu Wenruo free_anon_bdev(anon_dev); 63579787eaaSJeff Mahoney ret = PTR_ERR(new_root); 63666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 63779787eaaSJeff Mahoney goto fail; 63879787eaaSJeff Mahoney } 6392dfb1e43SQu Wenruo /* Freeing will be done in btrfs_put_root() of new_root */ 6402dfb1e43SQu Wenruo anon_dev = 0; 64176dda93cSYan, Zheng 642221581e4SJosef Bacik ret = btrfs_record_root_in_trans(trans, new_root); 643221581e4SJosef Bacik if (ret) { 644221581e4SJosef Bacik btrfs_put_root(new_root); 645221581e4SJosef Bacik btrfs_abort_transaction(trans, ret); 646221581e4SJosef Bacik goto fail; 647221581e4SJosef Bacik } 64876dda93cSYan, Zheng 6494d4340c9SChristian Brauner ret = btrfs_create_subvol_root(trans, new_root, root, mnt_userns); 65000246528SJosef Bacik btrfs_put_root(new_root); 651ce598979SMark Fasheh if (ret) { 652ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 65366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 654ce598979SMark Fasheh goto fail; 655ce598979SMark Fasheh } 656ce598979SMark Fasheh 657f46b5a66SChristoph Hellwig /* 658f46b5a66SChristoph Hellwig * insert the directory item 659f46b5a66SChristoph Hellwig */ 660877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 66179787eaaSJeff Mahoney if (ret) { 66266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 66379787eaaSJeff Mahoney goto fail; 66479787eaaSJeff Mahoney } 6653de4586cSChris Mason 666684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 6673de4586cSChris Mason BTRFS_FT_DIR, index); 66879787eaaSJeff Mahoney if (ret) { 66966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 670f46b5a66SChristoph Hellwig goto fail; 67179787eaaSJeff Mahoney } 6720660b5afSChris Mason 6736ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 6749a56fcd1SNikolay Borisov ret = btrfs_update_inode(trans, root, BTRFS_I(dir)); 675c7e54b51SJosef Bacik if (ret) { 676c7e54b51SJosef Bacik btrfs_abort_transaction(trans, ret); 677c7e54b51SJosef Bacik goto fail; 678c7e54b51SJosef Bacik } 67952c26179SYan Zheng 6806025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 6814a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 682c7e54b51SJosef Bacik if (ret) { 683c7e54b51SJosef Bacik btrfs_abort_transaction(trans, ret); 684c7e54b51SJosef Bacik goto fail; 685c7e54b51SJosef Bacik } 6860660b5afSChris Mason 687cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 6886bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 689dd5f9615SStefan Behrens if (ret) 69066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 691dd5f9615SStefan Behrens 692f46b5a66SChristoph Hellwig fail: 69349a3c4d9SDavid Sterba kfree(root_item); 694d5c12070SMiao Xie trans->block_rsv = NULL; 695d5c12070SMiao Xie trans->bytes_reserved = 0; 696e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &block_rsv); 697de6e8200SLiu Bo 6983a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 699f46b5a66SChristoph Hellwig if (err && !ret) 700f46b5a66SChristoph Hellwig ret = err; 7011a65e24bSChris Mason 7025662344bSTsutomu Itoh if (!ret) { 7035662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 704de6e8200SLiu Bo if (IS_ERR(inode)) 705de6e8200SLiu Bo return PTR_ERR(inode); 7065662344bSTsutomu Itoh d_instantiate(dentry, inode); 7075662344bSTsutomu Itoh } 708f46b5a66SChristoph Hellwig return ret; 70949a3c4d9SDavid Sterba 71049a3c4d9SDavid Sterba fail_free: 7112dfb1e43SQu Wenruo if (anon_dev) 7122dfb1e43SQu Wenruo free_anon_bdev(anon_dev); 71349a3c4d9SDavid Sterba kfree(root_item); 71449a3c4d9SDavid Sterba return ret; 715f46b5a66SChristoph Hellwig } 716f46b5a66SChristoph Hellwig 717e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 7189babda9fSNikolay Borisov struct dentry *dentry, bool readonly, 719e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 720f46b5a66SChristoph Hellwig { 7210b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7222e4bfab9SYan, Zheng struct inode *inode; 723f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 724f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7252e4bfab9SYan, Zheng int ret; 726f46b5a66SChristoph Hellwig 72792a7cc42SQu Wenruo if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) 728f46b5a66SChristoph Hellwig return -EINVAL; 729f46b5a66SChristoph Hellwig 730eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 731eede2bf3SOmar Sandoval btrfs_warn(fs_info, 732eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 733eede2bf3SOmar Sandoval return -ETXTBSY; 734eede2bf3SOmar Sandoval } 735eede2bf3SOmar Sandoval 73623269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 737a1ee7362SDavid Sterba if (!pending_snapshot) 738a1ee7362SDavid Sterba return -ENOMEM; 739a1ee7362SDavid Sterba 7402dfb1e43SQu Wenruo ret = get_anon_bdev(&pending_snapshot->anon_dev); 7412dfb1e43SQu Wenruo if (ret < 0) 7422dfb1e43SQu Wenruo goto free_pending; 743b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 74423269bf5SDavid Sterba GFP_KERNEL); 7458546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7468546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 747b0c0ea63SDavid Sterba ret = -ENOMEM; 748b0c0ea63SDavid Sterba goto free_pending; 749b0c0ea63SDavid Sterba } 750b0c0ea63SDavid Sterba 75166d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 75266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 753d5c12070SMiao Xie /* 754d5c12070SMiao Xie * 1 - parent dir inode 755d5c12070SMiao Xie * 2 - dir entries 756d5c12070SMiao Xie * 1 - root item 757d5c12070SMiao Xie * 2 - root ref/backref 758d5c12070SMiao Xie * 1 - root of snapshot 759dd5f9615SStefan Behrens * 1 - UUID item 760d5c12070SMiao Xie */ 761d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 762dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 763ee3441b4SJeff Mahoney false); 764d5c12070SMiao Xie if (ret) 765c11fbb6eSRobbie Ko goto free_pending; 766d5c12070SMiao Xie 767a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 768a22285a6SYan, Zheng pending_snapshot->root = root; 769b83cc969SLi Zefan pending_snapshot->readonly = readonly; 770e9662f70SMiao Xie pending_snapshot->dir = dir; 7718696c533SMiao Xie pending_snapshot->inherit = inherit; 772a22285a6SYan, Zheng 773d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 774a22285a6SYan, Zheng if (IS_ERR(trans)) { 775a22285a6SYan, Zheng ret = PTR_ERR(trans); 776a22285a6SYan, Zheng goto fail; 777a22285a6SYan, Zheng } 778a22285a6SYan, Zheng 7790b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 780a22285a6SYan, Zheng list_add(&pending_snapshot->list, 781a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 7820b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 7839babda9fSNikolay Borisov 7843a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 785aec8030aSMiao Xie if (ret) 786c37b2b62SJosef Bacik goto fail; 787a22285a6SYan, Zheng 788a22285a6SYan, Zheng ret = pending_snapshot->error; 789f46b5a66SChristoph Hellwig if (ret) 7902e4bfab9SYan, Zheng goto fail; 791f46b5a66SChristoph Hellwig 792d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 793d3797308SChris Mason if (ret) 794d3797308SChris Mason goto fail; 795d3797308SChris Mason 7962b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7972e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7982e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7992e4bfab9SYan, Zheng goto fail; 8002e4bfab9SYan, Zheng } 8015662344bSTsutomu Itoh 8022e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8032e4bfab9SYan, Zheng ret = 0; 8042dfb1e43SQu Wenruo pending_snapshot->anon_dev = 0; 8052e4bfab9SYan, Zheng fail: 8062dfb1e43SQu Wenruo /* Prevent double freeing of anon_dev */ 8072dfb1e43SQu Wenruo if (ret && pending_snapshot->snap) 8082dfb1e43SQu Wenruo pending_snapshot->snap->anon_dev = 0; 80900246528SJosef Bacik btrfs_put_root(pending_snapshot->snap); 810e85fde51SQu Wenruo btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv); 811b0c0ea63SDavid Sterba free_pending: 8122dfb1e43SQu Wenruo if (pending_snapshot->anon_dev) 8132dfb1e43SQu Wenruo free_anon_bdev(pending_snapshot->anon_dev); 814b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8158546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 816a1ee7362SDavid Sterba kfree(pending_snapshot); 817a1ee7362SDavid Sterba 818f46b5a66SChristoph Hellwig return ret; 819f46b5a66SChristoph Hellwig } 820f46b5a66SChristoph Hellwig 8214260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8224260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8234260f7c7SSage Weil * whether the type of victim is right. 8244260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8254260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8264260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8274260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8284260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8294260f7c7SSage Weil * a. be owner of dir, or 8304260f7c7SSage Weil * b. be owner of victim, or 8314260f7c7SSage Weil * c. have CAP_FOWNER capability 83201327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8334260f7c7SSage Weil * links pointing to it. 8344260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8354260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8364260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8374260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8384260f7c7SSage Weil * nfs_async_unlink(). 8394260f7c7SSage Weil */ 8404260f7c7SSage Weil 841c4ed533bSChristian Brauner static int btrfs_may_delete(struct user_namespace *mnt_userns, 842c4ed533bSChristian Brauner struct inode *dir, struct dentry *victim, int isdir) 8434260f7c7SSage Weil { 8444260f7c7SSage Weil int error; 8454260f7c7SSage Weil 8462b0143b5SDavid Howells if (d_really_is_negative(victim)) 8474260f7c7SSage Weil return -ENOENT; 8484260f7c7SSage Weil 8492b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8504fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8514260f7c7SSage Weil 852c4ed533bSChristian Brauner error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); 8534260f7c7SSage Weil if (error) 8544260f7c7SSage Weil return error; 8554260f7c7SSage Weil if (IS_APPEND(dir)) 8564260f7c7SSage Weil return -EPERM; 857c4ed533bSChristian Brauner if (check_sticky(mnt_userns, dir, d_inode(victim)) || 858ba73d987SChristian Brauner IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) || 859ba73d987SChristian Brauner IS_SWAPFILE(d_inode(victim))) 8604260f7c7SSage Weil return -EPERM; 8614260f7c7SSage Weil if (isdir) { 862e36cb0b8SDavid Howells if (!d_is_dir(victim)) 8634260f7c7SSage Weil return -ENOTDIR; 8644260f7c7SSage Weil if (IS_ROOT(victim)) 8654260f7c7SSage Weil return -EBUSY; 866e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 8674260f7c7SSage Weil return -EISDIR; 8684260f7c7SSage Weil if (IS_DEADDIR(dir)) 8694260f7c7SSage Weil return -ENOENT; 8704260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 8714260f7c7SSage Weil return -EBUSY; 8724260f7c7SSage Weil return 0; 8734260f7c7SSage Weil } 8744260f7c7SSage Weil 875cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 8764d4340c9SChristian Brauner static inline int btrfs_may_create(struct user_namespace *mnt_userns, 8774d4340c9SChristian Brauner struct inode *dir, struct dentry *child) 878cb8e7090SChristoph Hellwig { 8792b0143b5SDavid Howells if (d_really_is_positive(child)) 880cb8e7090SChristoph Hellwig return -EEXIST; 881cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 882cb8e7090SChristoph Hellwig return -ENOENT; 8834d4340c9SChristian Brauner if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns)) 8845474bf40SChristian Brauner return -EOVERFLOW; 8854d4340c9SChristian Brauner return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); 886cb8e7090SChristoph Hellwig } 887cb8e7090SChristoph Hellwig 888cb8e7090SChristoph Hellwig /* 889cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 890cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 891cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 892cb8e7090SChristoph Hellwig */ 89392872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 8944d4340c9SChristian Brauner struct user_namespace *mnt_userns, 89552f75f4fSDavid Sterba const char *name, int namelen, 89672fd032eSSage Weil struct btrfs_root *snap_src, 8979babda9fSNikolay Borisov bool readonly, 8988696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 899cb8e7090SChristoph Hellwig { 9002b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9010b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 902cb8e7090SChristoph Hellwig struct dentry *dentry; 903cb8e7090SChristoph Hellwig int error; 904cb8e7090SChristoph Hellwig 90500235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 90600235411SAl Viro if (error == -EINTR) 90700235411SAl Viro return error; 908cb8e7090SChristoph Hellwig 9094d4340c9SChristian Brauner dentry = lookup_one(mnt_userns, name, parent->dentry, namelen); 910cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 911cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 912cb8e7090SChristoph Hellwig goto out_unlock; 913cb8e7090SChristoph Hellwig 9144d4340c9SChristian Brauner error = btrfs_may_create(mnt_userns, dir, dentry); 915cb8e7090SChristoph Hellwig if (error) 916a874a63eSLiu Bo goto out_dput; 917cb8e7090SChristoph Hellwig 9189c52057cSChris Mason /* 9199c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9209c52057cSChris Mason * check for them now when we can safely fail 9219c52057cSChris Mason */ 9229c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9239c52057cSChris Mason dir->i_ino, name, 9249c52057cSChris Mason namelen); 9259c52057cSChris Mason if (error) 9269c52057cSChris Mason goto out_dput; 9279c52057cSChris Mason 9280b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 92976dda93cSYan, Zheng 93076dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 93176dda93cSYan, Zheng goto out_up_read; 93276dda93cSYan, Zheng 9339babda9fSNikolay Borisov if (snap_src) 9349babda9fSNikolay Borisov error = create_snapshot(snap_src, dir, dentry, readonly, inherit); 9359babda9fSNikolay Borisov else 9364d4340c9SChristian Brauner error = create_subvol(mnt_userns, dir, dentry, name, namelen, inherit); 9379babda9fSNikolay Borisov 93876dda93cSYan, Zheng if (!error) 93976dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 94076dda93cSYan, Zheng out_up_read: 9410b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 942cb8e7090SChristoph Hellwig out_dput: 943cb8e7090SChristoph Hellwig dput(dentry); 944cb8e7090SChristoph Hellwig out_unlock: 94564708539SJosef Bacik btrfs_inode_unlock(dir, 0); 946cb8e7090SChristoph Hellwig return error; 947cb8e7090SChristoph Hellwig } 948cb8e7090SChristoph Hellwig 949c11fbb6eSRobbie Ko static noinline int btrfs_mksnapshot(const struct path *parent, 9504d4340c9SChristian Brauner struct user_namespace *mnt_userns, 951c11fbb6eSRobbie Ko const char *name, int namelen, 952c11fbb6eSRobbie Ko struct btrfs_root *root, 953c11fbb6eSRobbie Ko bool readonly, 954c11fbb6eSRobbie Ko struct btrfs_qgroup_inherit *inherit) 955c11fbb6eSRobbie Ko { 956c11fbb6eSRobbie Ko int ret; 957c11fbb6eSRobbie Ko bool snapshot_force_cow = false; 958c11fbb6eSRobbie Ko 959c11fbb6eSRobbie Ko /* 960c11fbb6eSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 961c11fbb6eSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 962c11fbb6eSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 963c11fbb6eSRobbie Ko */ 964c11fbb6eSRobbie Ko btrfs_drew_read_lock(&root->snapshot_lock); 965c11fbb6eSRobbie Ko 966f9baa501SFilipe Manana ret = btrfs_start_delalloc_snapshot(root, false); 967c11fbb6eSRobbie Ko if (ret) 968c11fbb6eSRobbie Ko goto out; 969c11fbb6eSRobbie Ko 970c11fbb6eSRobbie Ko /* 971c11fbb6eSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 972c11fbb6eSRobbie Ko * we force future writes to fallback to COW mode during snapshot 973c11fbb6eSRobbie Ko * creation. 974c11fbb6eSRobbie Ko */ 975c11fbb6eSRobbie Ko atomic_inc(&root->snapshot_force_cow); 976c11fbb6eSRobbie Ko snapshot_force_cow = true; 977c11fbb6eSRobbie Ko 978c11fbb6eSRobbie Ko btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 979c11fbb6eSRobbie Ko 9804d4340c9SChristian Brauner ret = btrfs_mksubvol(parent, mnt_userns, name, namelen, 981c11fbb6eSRobbie Ko root, readonly, inherit); 982c11fbb6eSRobbie Ko out: 983c11fbb6eSRobbie Ko if (snapshot_force_cow) 984c11fbb6eSRobbie Ko atomic_dec(&root->snapshot_force_cow); 985c11fbb6eSRobbie Ko btrfs_drew_read_unlock(&root->snapshot_lock); 986c11fbb6eSRobbie Ko return ret; 987c11fbb6eSRobbie Ko } 988c11fbb6eSRobbie Ko 989e9eec721SQu Wenruo static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start, 990e9eec721SQu Wenruo bool locked) 99117ce6ef8SLiu Bo { 99217ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 993940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9946c282eb4SLi Zefan struct extent_map *em; 99576068caeSQu Wenruo const u32 sectorsize = BTRFS_I(inode)->root->fs_info->sectorsize; 996940100a4SChris Mason 997940100a4SChris Mason /* 998940100a4SChris Mason * hopefully we have this extent in the tree already, try without 999940100a4SChris Mason * the full extent lock 1000940100a4SChris Mason */ 1001940100a4SChris Mason read_lock(&em_tree->lock); 100276068caeSQu Wenruo em = lookup_extent_mapping(em_tree, start, sectorsize); 1003940100a4SChris Mason read_unlock(&em_tree->lock); 1004940100a4SChris Mason 1005940100a4SChris Mason if (!em) { 1006308d9800SFilipe Manana struct extent_state *cached = NULL; 100776068caeSQu Wenruo u64 end = start + sectorsize - 1; 1008308d9800SFilipe Manana 1009940100a4SChris Mason /* get the big lock and read metadata off disk */ 1010e9eec721SQu Wenruo if (!locked) 1011ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 101276068caeSQu Wenruo em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, sectorsize); 1013e9eec721SQu Wenruo if (!locked) 1014e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1015940100a4SChris Mason 10166cf8bfbfSDan Carpenter if (IS_ERR(em)) 10176c282eb4SLi Zefan return NULL; 1018940100a4SChris Mason } 1019940100a4SChris Mason 10206c282eb4SLi Zefan return em; 10216c282eb4SLi Zefan } 10226c282eb4SLi Zefan 1023e9eec721SQu Wenruo static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em, 1024e9eec721SQu Wenruo bool locked) 10256c282eb4SLi Zefan { 10266c282eb4SLi Zefan struct extent_map *next; 10276c282eb4SLi Zefan bool ret = true; 10286c282eb4SLi Zefan 10296c282eb4SLi Zefan /* this is the last extent */ 10306c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 10316c282eb4SLi Zefan return false; 10326c282eb4SLi Zefan 1033e9eec721SQu Wenruo next = defrag_lookup_extent(inode, em->start + em->len, locked); 1034e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1035e9512d72SChris Mason ret = false; 1036e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1037ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 10386c282eb4SLi Zefan ret = false; 10396c282eb4SLi Zefan 10406c282eb4SLi Zefan free_extent_map(next); 10416c282eb4SLi Zefan return ret; 10426c282eb4SLi Zefan } 10436c282eb4SLi Zefan 10444cb5300bSChris Mason /* 10455767b50cSQu Wenruo * Prepare one page to be defragged. 10465767b50cSQu Wenruo * 10475767b50cSQu Wenruo * This will ensure: 10485767b50cSQu Wenruo * 10495767b50cSQu Wenruo * - Returned page is locked and has been set up properly. 10505767b50cSQu Wenruo * - No ordered extent exists in the page. 10515767b50cSQu Wenruo * - The page is uptodate. 10525767b50cSQu Wenruo * 10535767b50cSQu Wenruo * NOTE: Caller should also wait for page writeback after the cluster is 10545767b50cSQu Wenruo * prepared, here we don't do writeback wait for each page. 10555767b50cSQu Wenruo */ 10565767b50cSQu Wenruo static struct page *defrag_prepare_one_page(struct btrfs_inode *inode, 10575767b50cSQu Wenruo pgoff_t index) 10585767b50cSQu Wenruo { 10595767b50cSQu Wenruo struct address_space *mapping = inode->vfs_inode.i_mapping; 10605767b50cSQu Wenruo gfp_t mask = btrfs_alloc_write_mask(mapping); 10615767b50cSQu Wenruo u64 page_start = (u64)index << PAGE_SHIFT; 10625767b50cSQu Wenruo u64 page_end = page_start + PAGE_SIZE - 1; 10635767b50cSQu Wenruo struct extent_state *cached_state = NULL; 10645767b50cSQu Wenruo struct page *page; 10655767b50cSQu Wenruo int ret; 10665767b50cSQu Wenruo 10675767b50cSQu Wenruo again: 10685767b50cSQu Wenruo page = find_or_create_page(mapping, index, mask); 10695767b50cSQu Wenruo if (!page) 10705767b50cSQu Wenruo return ERR_PTR(-ENOMEM); 10715767b50cSQu Wenruo 10725767b50cSQu Wenruo ret = set_page_extent_mapped(page); 10735767b50cSQu Wenruo if (ret < 0) { 10745767b50cSQu Wenruo unlock_page(page); 10755767b50cSQu Wenruo put_page(page); 10765767b50cSQu Wenruo return ERR_PTR(ret); 10775767b50cSQu Wenruo } 10785767b50cSQu Wenruo 10795767b50cSQu Wenruo /* Wait for any existing ordered extent in the range */ 10805767b50cSQu Wenruo while (1) { 10815767b50cSQu Wenruo struct btrfs_ordered_extent *ordered; 10825767b50cSQu Wenruo 10835767b50cSQu Wenruo lock_extent_bits(&inode->io_tree, page_start, page_end, &cached_state); 10845767b50cSQu Wenruo ordered = btrfs_lookup_ordered_range(inode, page_start, PAGE_SIZE); 10855767b50cSQu Wenruo unlock_extent_cached(&inode->io_tree, page_start, page_end, 10865767b50cSQu Wenruo &cached_state); 10875767b50cSQu Wenruo if (!ordered) 10885767b50cSQu Wenruo break; 10895767b50cSQu Wenruo 10905767b50cSQu Wenruo unlock_page(page); 10915767b50cSQu Wenruo btrfs_start_ordered_extent(ordered, 1); 10925767b50cSQu Wenruo btrfs_put_ordered_extent(ordered); 10935767b50cSQu Wenruo lock_page(page); 10945767b50cSQu Wenruo /* 10955767b50cSQu Wenruo * We unlocked the page above, so we need check if it was 10965767b50cSQu Wenruo * released or not. 10975767b50cSQu Wenruo */ 10985767b50cSQu Wenruo if (page->mapping != mapping || !PagePrivate(page)) { 10995767b50cSQu Wenruo unlock_page(page); 11005767b50cSQu Wenruo put_page(page); 11015767b50cSQu Wenruo goto again; 11025767b50cSQu Wenruo } 11035767b50cSQu Wenruo } 11045767b50cSQu Wenruo 11055767b50cSQu Wenruo /* 11065767b50cSQu Wenruo * Now the page range has no ordered extent any more. Read the page to 11075767b50cSQu Wenruo * make it uptodate. 11085767b50cSQu Wenruo */ 11095767b50cSQu Wenruo if (!PageUptodate(page)) { 11105767b50cSQu Wenruo btrfs_readpage(NULL, page); 11115767b50cSQu Wenruo lock_page(page); 11125767b50cSQu Wenruo if (page->mapping != mapping || !PagePrivate(page)) { 11135767b50cSQu Wenruo unlock_page(page); 11145767b50cSQu Wenruo put_page(page); 11155767b50cSQu Wenruo goto again; 11165767b50cSQu Wenruo } 11175767b50cSQu Wenruo if (!PageUptodate(page)) { 11185767b50cSQu Wenruo unlock_page(page); 11195767b50cSQu Wenruo put_page(page); 11205767b50cSQu Wenruo return ERR_PTR(-EIO); 11215767b50cSQu Wenruo } 11225767b50cSQu Wenruo } 11235767b50cSQu Wenruo return page; 11245767b50cSQu Wenruo } 11255767b50cSQu Wenruo 1126eb793cf8SQu Wenruo struct defrag_target_range { 1127eb793cf8SQu Wenruo struct list_head list; 1128eb793cf8SQu Wenruo u64 start; 1129eb793cf8SQu Wenruo u64 len; 1130eb793cf8SQu Wenruo }; 1131eb793cf8SQu Wenruo 1132eb793cf8SQu Wenruo /* 1133eb793cf8SQu Wenruo * Collect all valid target extents. 1134eb793cf8SQu Wenruo * 1135eb793cf8SQu Wenruo * @start: file offset to lookup 1136eb793cf8SQu Wenruo * @len: length to lookup 1137eb793cf8SQu Wenruo * @extent_thresh: file extent size threshold, any extent size >= this value 1138eb793cf8SQu Wenruo * will be ignored 1139eb793cf8SQu Wenruo * @newer_than: only defrag extents newer than this value 1140eb793cf8SQu Wenruo * @do_compress: whether the defrag is doing compression 1141eb793cf8SQu Wenruo * if true, @extent_thresh will be ignored and all regular 1142eb793cf8SQu Wenruo * file extents meeting @newer_than will be targets. 1143e9eec721SQu Wenruo * @locked: if the range has already held extent lock 1144eb793cf8SQu Wenruo * @target_list: list of targets file extents 1145eb793cf8SQu Wenruo */ 1146eb793cf8SQu Wenruo static int defrag_collect_targets(struct btrfs_inode *inode, 1147eb793cf8SQu Wenruo u64 start, u64 len, u32 extent_thresh, 1148eb793cf8SQu Wenruo u64 newer_than, bool do_compress, 1149e9eec721SQu Wenruo bool locked, struct list_head *target_list) 1150eb793cf8SQu Wenruo { 1151eb793cf8SQu Wenruo u64 cur = start; 1152eb793cf8SQu Wenruo int ret = 0; 1153eb793cf8SQu Wenruo 1154eb793cf8SQu Wenruo while (cur < start + len) { 1155eb793cf8SQu Wenruo struct extent_map *em; 1156eb793cf8SQu Wenruo struct defrag_target_range *new; 1157eb793cf8SQu Wenruo bool next_mergeable = true; 1158eb793cf8SQu Wenruo u64 range_len; 1159eb793cf8SQu Wenruo 1160e9eec721SQu Wenruo em = defrag_lookup_extent(&inode->vfs_inode, cur, locked); 1161eb793cf8SQu Wenruo if (!em) 1162eb793cf8SQu Wenruo break; 1163eb793cf8SQu Wenruo 1164eb793cf8SQu Wenruo /* Skip hole/inline/preallocated extents */ 1165eb793cf8SQu Wenruo if (em->block_start >= EXTENT_MAP_LAST_BYTE || 1166eb793cf8SQu Wenruo test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) 1167eb793cf8SQu Wenruo goto next; 1168eb793cf8SQu Wenruo 1169eb793cf8SQu Wenruo /* Skip older extent */ 1170eb793cf8SQu Wenruo if (em->generation < newer_than) 1171eb793cf8SQu Wenruo goto next; 1172eb793cf8SQu Wenruo 1173eb793cf8SQu Wenruo /* 1174eb793cf8SQu Wenruo * For do_compress case, we want to compress all valid file 1175eb793cf8SQu Wenruo * extents, thus no @extent_thresh or mergeable check. 1176eb793cf8SQu Wenruo */ 1177eb793cf8SQu Wenruo if (do_compress) 1178eb793cf8SQu Wenruo goto add; 1179eb793cf8SQu Wenruo 1180eb793cf8SQu Wenruo /* Skip too large extent */ 1181eb793cf8SQu Wenruo if (em->len >= extent_thresh) 1182eb793cf8SQu Wenruo goto next; 1183eb793cf8SQu Wenruo 1184e9eec721SQu Wenruo next_mergeable = defrag_check_next_extent(&inode->vfs_inode, em, 1185e9eec721SQu Wenruo locked); 1186eb793cf8SQu Wenruo if (!next_mergeable) { 1187eb793cf8SQu Wenruo struct defrag_target_range *last; 1188eb793cf8SQu Wenruo 1189eb793cf8SQu Wenruo /* Empty target list, no way to merge with last entry */ 1190eb793cf8SQu Wenruo if (list_empty(target_list)) 1191eb793cf8SQu Wenruo goto next; 1192eb793cf8SQu Wenruo last = list_entry(target_list->prev, 1193eb793cf8SQu Wenruo struct defrag_target_range, list); 1194eb793cf8SQu Wenruo /* Not mergeable with last entry */ 1195eb793cf8SQu Wenruo if (last->start + last->len != cur) 1196eb793cf8SQu Wenruo goto next; 1197eb793cf8SQu Wenruo 1198eb793cf8SQu Wenruo /* Mergeable, fall through to add it to @target_list. */ 1199eb793cf8SQu Wenruo } 1200eb793cf8SQu Wenruo 1201eb793cf8SQu Wenruo add: 1202eb793cf8SQu Wenruo range_len = min(extent_map_end(em), start + len) - cur; 1203eb793cf8SQu Wenruo /* 1204eb793cf8SQu Wenruo * This one is a good target, check if it can be merged into 1205eb793cf8SQu Wenruo * last range of the target list. 1206eb793cf8SQu Wenruo */ 1207eb793cf8SQu Wenruo if (!list_empty(target_list)) { 1208eb793cf8SQu Wenruo struct defrag_target_range *last; 1209eb793cf8SQu Wenruo 1210eb793cf8SQu Wenruo last = list_entry(target_list->prev, 1211eb793cf8SQu Wenruo struct defrag_target_range, list); 1212eb793cf8SQu Wenruo ASSERT(last->start + last->len <= cur); 1213eb793cf8SQu Wenruo if (last->start + last->len == cur) { 1214eb793cf8SQu Wenruo /* Mergeable, enlarge the last entry */ 1215eb793cf8SQu Wenruo last->len += range_len; 1216eb793cf8SQu Wenruo goto next; 1217eb793cf8SQu Wenruo } 1218eb793cf8SQu Wenruo /* Fall through to allocate a new entry */ 1219eb793cf8SQu Wenruo } 1220eb793cf8SQu Wenruo 1221eb793cf8SQu Wenruo /* Allocate new defrag_target_range */ 1222eb793cf8SQu Wenruo new = kmalloc(sizeof(*new), GFP_NOFS); 1223eb793cf8SQu Wenruo if (!new) { 1224eb793cf8SQu Wenruo free_extent_map(em); 1225eb793cf8SQu Wenruo ret = -ENOMEM; 1226eb793cf8SQu Wenruo break; 1227eb793cf8SQu Wenruo } 1228eb793cf8SQu Wenruo new->start = cur; 1229eb793cf8SQu Wenruo new->len = range_len; 1230eb793cf8SQu Wenruo list_add_tail(&new->list, target_list); 1231eb793cf8SQu Wenruo 1232eb793cf8SQu Wenruo next: 1233eb793cf8SQu Wenruo cur = extent_map_end(em); 1234eb793cf8SQu Wenruo free_extent_map(em); 1235eb793cf8SQu Wenruo } 1236eb793cf8SQu Wenruo if (ret < 0) { 1237eb793cf8SQu Wenruo struct defrag_target_range *entry; 1238eb793cf8SQu Wenruo struct defrag_target_range *tmp; 1239eb793cf8SQu Wenruo 1240eb793cf8SQu Wenruo list_for_each_entry_safe(entry, tmp, target_list, list) { 1241eb793cf8SQu Wenruo list_del_init(&entry->list); 1242eb793cf8SQu Wenruo kfree(entry); 1243eb793cf8SQu Wenruo } 1244eb793cf8SQu Wenruo } 1245eb793cf8SQu Wenruo return ret; 1246eb793cf8SQu Wenruo } 1247eb793cf8SQu Wenruo 124822b398eeSQu Wenruo #define CLUSTER_SIZE (SZ_256K) 124922b398eeSQu Wenruo 125022b398eeSQu Wenruo /* 125122b398eeSQu Wenruo * Defrag one contiguous target range. 125222b398eeSQu Wenruo * 125322b398eeSQu Wenruo * @inode: target inode 125422b398eeSQu Wenruo * @target: target range to defrag 125522b398eeSQu Wenruo * @pages: locked pages covering the defrag range 125622b398eeSQu Wenruo * @nr_pages: number of locked pages 125722b398eeSQu Wenruo * 125822b398eeSQu Wenruo * Caller should ensure: 125922b398eeSQu Wenruo * 126022b398eeSQu Wenruo * - Pages are prepared 126122b398eeSQu Wenruo * Pages should be locked, no ordered extent in the pages range, 126222b398eeSQu Wenruo * no writeback. 126322b398eeSQu Wenruo * 126422b398eeSQu Wenruo * - Extent bits are locked 126522b398eeSQu Wenruo */ 126622b398eeSQu Wenruo static int defrag_one_locked_target(struct btrfs_inode *inode, 126722b398eeSQu Wenruo struct defrag_target_range *target, 126822b398eeSQu Wenruo struct page **pages, int nr_pages, 126922b398eeSQu Wenruo struct extent_state **cached_state) 127022b398eeSQu Wenruo { 127122b398eeSQu Wenruo struct btrfs_fs_info *fs_info = inode->root->fs_info; 127222b398eeSQu Wenruo struct extent_changeset *data_reserved = NULL; 127322b398eeSQu Wenruo const u64 start = target->start; 127422b398eeSQu Wenruo const u64 len = target->len; 127522b398eeSQu Wenruo unsigned long last_index = (start + len - 1) >> PAGE_SHIFT; 127622b398eeSQu Wenruo unsigned long start_index = start >> PAGE_SHIFT; 127722b398eeSQu Wenruo unsigned long first_index = page_index(pages[0]); 127822b398eeSQu Wenruo int ret = 0; 127922b398eeSQu Wenruo int i; 128022b398eeSQu Wenruo 128122b398eeSQu Wenruo ASSERT(last_index - first_index + 1 <= nr_pages); 128222b398eeSQu Wenruo 128322b398eeSQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, start, len); 128422b398eeSQu Wenruo if (ret < 0) 128522b398eeSQu Wenruo return ret; 128622b398eeSQu Wenruo clear_extent_bit(&inode->io_tree, start, start + len - 1, 128722b398eeSQu Wenruo EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | 128822b398eeSQu Wenruo EXTENT_DEFRAG, 0, 0, cached_state); 128922b398eeSQu Wenruo set_extent_defrag(&inode->io_tree, start, start + len - 1, cached_state); 129022b398eeSQu Wenruo 129122b398eeSQu Wenruo /* Update the page status */ 129222b398eeSQu Wenruo for (i = start_index - first_index; i <= last_index - first_index; i++) { 129322b398eeSQu Wenruo ClearPageChecked(pages[i]); 129422b398eeSQu Wenruo btrfs_page_clamp_set_dirty(fs_info, pages[i], start, len); 129522b398eeSQu Wenruo } 129622b398eeSQu Wenruo btrfs_delalloc_release_extents(inode, len); 129722b398eeSQu Wenruo extent_changeset_free(data_reserved); 129822b398eeSQu Wenruo 129922b398eeSQu Wenruo return ret; 130022b398eeSQu Wenruo } 130122b398eeSQu Wenruo 1302e9eec721SQu Wenruo static int defrag_one_range(struct btrfs_inode *inode, u64 start, u32 len, 1303e9eec721SQu Wenruo u32 extent_thresh, u64 newer_than, bool do_compress) 1304e9eec721SQu Wenruo { 1305e9eec721SQu Wenruo struct extent_state *cached_state = NULL; 1306e9eec721SQu Wenruo struct defrag_target_range *entry; 1307e9eec721SQu Wenruo struct defrag_target_range *tmp; 1308e9eec721SQu Wenruo LIST_HEAD(target_list); 1309e9eec721SQu Wenruo struct page **pages; 1310e9eec721SQu Wenruo const u32 sectorsize = inode->root->fs_info->sectorsize; 1311e9eec721SQu Wenruo u64 last_index = (start + len - 1) >> PAGE_SHIFT; 1312e9eec721SQu Wenruo u64 start_index = start >> PAGE_SHIFT; 1313e9eec721SQu Wenruo unsigned int nr_pages = last_index - start_index + 1; 1314e9eec721SQu Wenruo int ret = 0; 1315e9eec721SQu Wenruo int i; 1316e9eec721SQu Wenruo 1317e9eec721SQu Wenruo ASSERT(nr_pages <= CLUSTER_SIZE / PAGE_SIZE); 1318e9eec721SQu Wenruo ASSERT(IS_ALIGNED(start, sectorsize) && IS_ALIGNED(len, sectorsize)); 1319e9eec721SQu Wenruo 1320e9eec721SQu Wenruo pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS); 1321e9eec721SQu Wenruo if (!pages) 1322e9eec721SQu Wenruo return -ENOMEM; 1323e9eec721SQu Wenruo 1324e9eec721SQu Wenruo /* Prepare all pages */ 1325e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) { 1326e9eec721SQu Wenruo pages[i] = defrag_prepare_one_page(inode, start_index + i); 1327e9eec721SQu Wenruo if (IS_ERR(pages[i])) { 1328e9eec721SQu Wenruo ret = PTR_ERR(pages[i]); 1329e9eec721SQu Wenruo pages[i] = NULL; 1330e9eec721SQu Wenruo goto free_pages; 1331e9eec721SQu Wenruo } 1332e9eec721SQu Wenruo } 1333e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) 1334e9eec721SQu Wenruo wait_on_page_writeback(pages[i]); 1335e9eec721SQu Wenruo 1336e9eec721SQu Wenruo /* Lock the pages range */ 1337e9eec721SQu Wenruo lock_extent_bits(&inode->io_tree, start_index << PAGE_SHIFT, 1338e9eec721SQu Wenruo (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, 1339e9eec721SQu Wenruo &cached_state); 1340e9eec721SQu Wenruo /* 1341e9eec721SQu Wenruo * Now we have a consistent view about the extent map, re-check 1342e9eec721SQu Wenruo * which range really needs to be defragged. 1343e9eec721SQu Wenruo * 1344e9eec721SQu Wenruo * And this time we have extent locked already, pass @locked = true 1345e9eec721SQu Wenruo * so that we won't relock the extent range and cause deadlock. 1346e9eec721SQu Wenruo */ 1347e9eec721SQu Wenruo ret = defrag_collect_targets(inode, start, len, extent_thresh, 1348e9eec721SQu Wenruo newer_than, do_compress, true, 1349e9eec721SQu Wenruo &target_list); 1350e9eec721SQu Wenruo if (ret < 0) 1351e9eec721SQu Wenruo goto unlock_extent; 1352e9eec721SQu Wenruo 1353e9eec721SQu Wenruo list_for_each_entry(entry, &target_list, list) { 1354e9eec721SQu Wenruo ret = defrag_one_locked_target(inode, entry, pages, nr_pages, 1355e9eec721SQu Wenruo &cached_state); 1356e9eec721SQu Wenruo if (ret < 0) 1357e9eec721SQu Wenruo break; 1358e9eec721SQu Wenruo } 1359e9eec721SQu Wenruo 1360e9eec721SQu Wenruo list_for_each_entry_safe(entry, tmp, &target_list, list) { 1361e9eec721SQu Wenruo list_del_init(&entry->list); 1362e9eec721SQu Wenruo kfree(entry); 1363e9eec721SQu Wenruo } 1364e9eec721SQu Wenruo unlock_extent: 1365e9eec721SQu Wenruo unlock_extent_cached(&inode->io_tree, start_index << PAGE_SHIFT, 1366e9eec721SQu Wenruo (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, 1367e9eec721SQu Wenruo &cached_state); 1368e9eec721SQu Wenruo free_pages: 1369e9eec721SQu Wenruo for (i = 0; i < nr_pages; i++) { 1370e9eec721SQu Wenruo if (pages[i]) { 1371e9eec721SQu Wenruo unlock_page(pages[i]); 1372e9eec721SQu Wenruo put_page(pages[i]); 1373e9eec721SQu Wenruo } 1374e9eec721SQu Wenruo } 1375e9eec721SQu Wenruo kfree(pages); 1376e9eec721SQu Wenruo return ret; 1377e9eec721SQu Wenruo } 1378e9eec721SQu Wenruo 1379b18c3ab2SQu Wenruo static int defrag_one_cluster(struct btrfs_inode *inode, 1380b18c3ab2SQu Wenruo struct file_ra_state *ra, 1381b18c3ab2SQu Wenruo u64 start, u32 len, u32 extent_thresh, 1382b18c3ab2SQu Wenruo u64 newer_than, bool do_compress, 1383b18c3ab2SQu Wenruo unsigned long *sectors_defragged, 1384b18c3ab2SQu Wenruo unsigned long max_sectors) 1385b18c3ab2SQu Wenruo { 1386b18c3ab2SQu Wenruo const u32 sectorsize = inode->root->fs_info->sectorsize; 1387b18c3ab2SQu Wenruo struct defrag_target_range *entry; 1388b18c3ab2SQu Wenruo struct defrag_target_range *tmp; 1389b18c3ab2SQu Wenruo LIST_HEAD(target_list); 1390b18c3ab2SQu Wenruo int ret; 1391b18c3ab2SQu Wenruo 1392b18c3ab2SQu Wenruo BUILD_BUG_ON(!IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE)); 1393b18c3ab2SQu Wenruo ret = defrag_collect_targets(inode, start, len, extent_thresh, 1394b18c3ab2SQu Wenruo newer_than, do_compress, false, 1395b18c3ab2SQu Wenruo &target_list); 1396b18c3ab2SQu Wenruo if (ret < 0) 1397b18c3ab2SQu Wenruo goto out; 1398b18c3ab2SQu Wenruo 1399b18c3ab2SQu Wenruo list_for_each_entry(entry, &target_list, list) { 1400b18c3ab2SQu Wenruo u32 range_len = entry->len; 1401b18c3ab2SQu Wenruo 1402b18c3ab2SQu Wenruo /* Reached the limit */ 1403b18c3ab2SQu Wenruo if (max_sectors && max_sectors == *sectors_defragged) 1404b18c3ab2SQu Wenruo break; 1405b18c3ab2SQu Wenruo 1406b18c3ab2SQu Wenruo if (max_sectors) 1407b18c3ab2SQu Wenruo range_len = min_t(u32, range_len, 1408b18c3ab2SQu Wenruo (max_sectors - *sectors_defragged) * sectorsize); 1409b18c3ab2SQu Wenruo 1410b18c3ab2SQu Wenruo if (ra) 1411b18c3ab2SQu Wenruo page_cache_sync_readahead(inode->vfs_inode.i_mapping, 1412b18c3ab2SQu Wenruo ra, NULL, entry->start >> PAGE_SHIFT, 1413b18c3ab2SQu Wenruo ((entry->start + range_len - 1) >> PAGE_SHIFT) - 1414b18c3ab2SQu Wenruo (entry->start >> PAGE_SHIFT) + 1); 1415b18c3ab2SQu Wenruo /* 1416b18c3ab2SQu Wenruo * Here we may not defrag any range if holes are punched before 1417b18c3ab2SQu Wenruo * we locked the pages. 1418b18c3ab2SQu Wenruo * But that's fine, it only affects the @sectors_defragged 1419b18c3ab2SQu Wenruo * accounting. 1420b18c3ab2SQu Wenruo */ 1421b18c3ab2SQu Wenruo ret = defrag_one_range(inode, entry->start, range_len, 1422b18c3ab2SQu Wenruo extent_thresh, newer_than, do_compress); 1423b18c3ab2SQu Wenruo if (ret < 0) 1424b18c3ab2SQu Wenruo break; 1425b18c3ab2SQu Wenruo *sectors_defragged += range_len; 1426b18c3ab2SQu Wenruo } 1427b18c3ab2SQu Wenruo out: 1428b18c3ab2SQu Wenruo list_for_each_entry_safe(entry, tmp, &target_list, list) { 1429b18c3ab2SQu Wenruo list_del_init(&entry->list); 1430b18c3ab2SQu Wenruo kfree(entry); 1431b18c3ab2SQu Wenruo } 1432b18c3ab2SQu Wenruo return ret; 1433b18c3ab2SQu Wenruo } 1434b18c3ab2SQu Wenruo 14351ccc2e8aSQu Wenruo /* 14361ccc2e8aSQu Wenruo * Entry point to file defragmentation. 14371ccc2e8aSQu Wenruo * 14381ccc2e8aSQu Wenruo * @inode: inode to be defragged 14391ccc2e8aSQu Wenruo * @ra: readahead state (can be NUL) 14401ccc2e8aSQu Wenruo * @range: defrag options including range and flags 14411ccc2e8aSQu Wenruo * @newer_than: minimum transid to defrag 14421ccc2e8aSQu Wenruo * @max_to_defrag: max number of sectors to be defragged, if 0, the whole inode 14431ccc2e8aSQu Wenruo * will be defragged. 14441ccc2e8aSQu Wenruo */ 14451ccc2e8aSQu Wenruo int btrfs_defrag_file(struct inode *inode, struct file_ra_state *ra, 14464cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 14474cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 14484cb5300bSChris Mason { 14490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 14507b508037SQu Wenruo unsigned long sectors_defragged = 0; 1451151a31b2SLi Zefan u64 isize = i_size_read(inode); 14527b508037SQu Wenruo u64 cur; 14537b508037SQu Wenruo u64 last_byte; 14541e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14551ccc2e8aSQu Wenruo bool ra_allocated = false; 14567b508037SQu Wenruo int compress_type = BTRFS_COMPRESS_ZLIB; 14577b508037SQu Wenruo int ret = 0; 14587b508037SQu Wenruo u32 extent_thresh = range->extent_thresh; 14594cb5300bSChris Mason 14600abd5b17SLiu Bo if (isize == 0) 14610abd5b17SLiu Bo return 0; 14620abd5b17SLiu Bo 14630abd5b17SLiu Bo if (range->start >= isize) 14640abd5b17SLiu Bo return -EINVAL; 14651a419d85SLi Zefan 14661e2ef46dSDavid Sterba if (do_compress) { 1467ce96b7ffSChengguang Xu if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES) 14681a419d85SLi Zefan return -EINVAL; 14691a419d85SLi Zefan if (range->compress_type) 14701a419d85SLi Zefan compress_type = range->compress_type; 14711a419d85SLi Zefan } 1472f46b5a66SChristoph Hellwig 14730abd5b17SLiu Bo if (extent_thresh == 0) 1474ee22184bSByongho Lee extent_thresh = SZ_256K; 1475f46b5a66SChristoph Hellwig 14767b508037SQu Wenruo if (range->start + range->len > range->start) { 14777b508037SQu Wenruo /* Got a specific range */ 14787b508037SQu Wenruo last_byte = min(isize, range->start + range->len) - 1; 14797b508037SQu Wenruo } else { 14807b508037SQu Wenruo /* Defrag until file end */ 14817b508037SQu Wenruo last_byte = isize - 1; 14827b508037SQu Wenruo } 14837b508037SQu Wenruo 14844cb5300bSChris Mason /* 14851ccc2e8aSQu Wenruo * If we were not given a ra, allocate a readahead context. As 14860a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14870a52d108SDavid Sterba * we don't error out. 14884cb5300bSChris Mason */ 14891ccc2e8aSQu Wenruo if (!ra) { 14901ccc2e8aSQu Wenruo ra_allocated = true; 149163e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14920a52d108SDavid Sterba if (ra) 14934cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14944cb5300bSChris Mason } 14954cb5300bSChris Mason 14967b508037SQu Wenruo /* Align the range */ 14977b508037SQu Wenruo cur = round_down(range->start, fs_info->sectorsize); 14987b508037SQu Wenruo last_byte = round_up(last_byte, fs_info->sectorsize) - 1; 14994cb5300bSChris Mason 15007b508037SQu Wenruo while (cur < last_byte) { 15017b508037SQu Wenruo u64 cluster_end; 15021e701a32SChris Mason 15037b508037SQu Wenruo /* The cluster size 256K should always be page aligned */ 15047b508037SQu Wenruo BUILD_BUG_ON(!IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE)); 15054cb5300bSChris Mason 15067b508037SQu Wenruo /* We want the cluster end at page boundary when possible */ 15077b508037SQu Wenruo cluster_end = (((cur >> PAGE_SHIFT) + 15087b508037SQu Wenruo (SZ_256K >> PAGE_SHIFT)) << PAGE_SHIFT) - 1; 15097b508037SQu Wenruo cluster_end = min(cluster_end, last_byte); 15104cb5300bSChris Mason 151164708539SJosef Bacik btrfs_inode_lock(inode, 0); 1512eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1513eede2bf3SOmar Sandoval ret = -ETXTBSY; 15147b508037SQu Wenruo btrfs_inode_unlock(inode, 0); 15157b508037SQu Wenruo break; 15167b508037SQu Wenruo } 15177b508037SQu Wenruo if (!(inode->i_sb->s_flags & SB_ACTIVE)) { 15187b508037SQu Wenruo btrfs_inode_unlock(inode, 0); 15197b508037SQu Wenruo break; 15207b508037SQu Wenruo } 15211e2ef46dSDavid Sterba if (do_compress) 1522eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 15237b508037SQu Wenruo ret = defrag_one_cluster(BTRFS_I(inode), ra, cur, 15247b508037SQu Wenruo cluster_end + 1 - cur, extent_thresh, 15257b508037SQu Wenruo newer_than, do_compress, 15267b508037SQu Wenruo §ors_defragged, max_to_defrag); 152764708539SJosef Bacik btrfs_inode_unlock(inode, 0); 15287b508037SQu Wenruo if (ret < 0) 15294cb5300bSChris Mason break; 15307b508037SQu Wenruo cur = cluster_end + 1; 15314cb5300bSChris Mason } 1532f46b5a66SChristoph Hellwig 15337b508037SQu Wenruo if (ra_allocated) 15347b508037SQu Wenruo kfree(ra); 15357b508037SQu Wenruo if (sectors_defragged) { 15367b508037SQu Wenruo /* 15377b508037SQu Wenruo * We have defragged some sectors, for compression case they 15387b508037SQu Wenruo * need to be written back immediately. 15397b508037SQu Wenruo */ 15407b508037SQu Wenruo if (range->flags & BTRFS_DEFRAG_RANGE_START_IO) { 15411e701a32SChris Mason filemap_flush(inode->i_mapping); 1542dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1543dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1544dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1545dec8ef90SFilipe Manana } 15467b508037SQu Wenruo if (range->compress_type == BTRFS_COMPRESS_LZO) 15470b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15487b508037SQu Wenruo else if (range->compress_type == BTRFS_COMPRESS_ZSTD) 15495c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15507b508037SQu Wenruo ret = sectors_defragged; 15511a419d85SLi Zefan } 15521e2ef46dSDavid Sterba if (do_compress) { 155364708539SJosef Bacik btrfs_inode_lock(inode, 0); 1554eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 155564708539SJosef Bacik btrfs_inode_unlock(inode, 0); 1556633085c7SFilipe David Borba Manana } 1557940100a4SChris Mason return ret; 1558f46b5a66SChristoph Hellwig } 1559f46b5a66SChristoph Hellwig 156017aaa434SDavid Sterba /* 156117aaa434SDavid Sterba * Try to start exclusive operation @type or cancel it if it's running. 156217aaa434SDavid Sterba * 156317aaa434SDavid Sterba * Return: 156417aaa434SDavid Sterba * 0 - normal mode, newly claimed op started 156517aaa434SDavid Sterba * >0 - normal mode, something else is running, 156617aaa434SDavid Sterba * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space 156717aaa434SDavid Sterba * ECANCELED - cancel mode, successful cancel 156817aaa434SDavid Sterba * ENOTCONN - cancel mode, operation not running anymore 156917aaa434SDavid Sterba */ 157017aaa434SDavid Sterba static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info, 157117aaa434SDavid Sterba enum btrfs_exclusive_operation type, bool cancel) 157217aaa434SDavid Sterba { 157317aaa434SDavid Sterba if (!cancel) { 157417aaa434SDavid Sterba /* Start normal op */ 157517aaa434SDavid Sterba if (!btrfs_exclop_start(fs_info, type)) 157617aaa434SDavid Sterba return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 157717aaa434SDavid Sterba /* Exclusive operation is now claimed */ 157817aaa434SDavid Sterba return 0; 157917aaa434SDavid Sterba } 158017aaa434SDavid Sterba 158117aaa434SDavid Sterba /* Cancel running op */ 158217aaa434SDavid Sterba if (btrfs_exclop_start_try_lock(fs_info, type)) { 158317aaa434SDavid Sterba /* 158417aaa434SDavid Sterba * This blocks any exclop finish from setting it to NONE, so we 158517aaa434SDavid Sterba * request cancellation. Either it runs and we will wait for it, 158617aaa434SDavid Sterba * or it has finished and no waiting will happen. 158717aaa434SDavid Sterba */ 158817aaa434SDavid Sterba atomic_inc(&fs_info->reloc_cancel_req); 158917aaa434SDavid Sterba btrfs_exclop_start_unlock(fs_info); 159017aaa434SDavid Sterba 159117aaa434SDavid Sterba if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) 159217aaa434SDavid Sterba wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING, 159317aaa434SDavid Sterba TASK_INTERRUPTIBLE); 159417aaa434SDavid Sterba 159517aaa434SDavid Sterba return -ECANCELED; 159617aaa434SDavid Sterba } 159717aaa434SDavid Sterba 159817aaa434SDavid Sterba /* Something else is running or none */ 159917aaa434SDavid Sterba return -ENOTCONN; 160017aaa434SDavid Sterba } 160117aaa434SDavid Sterba 1602198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 160376dda93cSYan, Zheng void __user *arg) 1604f46b5a66SChristoph Hellwig { 1605562d7b15SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 16060b246afaSJeff Mahoney struct inode *inode = file_inode(file); 16070b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1608f46b5a66SChristoph Hellwig u64 new_size; 1609f46b5a66SChristoph Hellwig u64 old_size; 1610f46b5a66SChristoph Hellwig u64 devid = 1; 16110b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1612f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1613f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1614f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1615f46b5a66SChristoph Hellwig char *sizestr; 16169a40f122SGui Hecheng char *retptr; 1617f46b5a66SChristoph Hellwig char *devstr = NULL; 1618f46b5a66SChristoph Hellwig int ret = 0; 1619f46b5a66SChristoph Hellwig int mod = 0; 1620bb059a37SDavid Sterba bool cancel; 1621f46b5a66SChristoph Hellwig 1622e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1623e441d54dSChris Mason return -EPERM; 1624e441d54dSChris Mason 1625198605a8SMiao Xie ret = mnt_want_write_file(file); 1626198605a8SMiao Xie if (ret) 1627198605a8SMiao Xie return ret; 1628198605a8SMiao Xie 1629bb059a37SDavid Sterba /* 1630bb059a37SDavid Sterba * Read the arguments before checking exclusivity to be able to 1631bb059a37SDavid Sterba * distinguish regular resize and cancel 1632bb059a37SDavid Sterba */ 1633dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1634c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1635c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1636bb059a37SDavid Sterba goto out_drop; 1637c9e9f97bSIlya Dryomov } 16385516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1639f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1640bb059a37SDavid Sterba cancel = (strcmp("cancel", sizestr) == 0); 1641bb059a37SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel); 1642bb059a37SDavid Sterba if (ret) 1643bb059a37SDavid Sterba goto out_free; 1644bb059a37SDavid Sterba /* Exclusive operation is now claimed */ 1645bb059a37SDavid Sterba 1646f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1647f46b5a66SChristoph Hellwig if (devstr) { 1648f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1649f46b5a66SChristoph Hellwig *devstr = '\0'; 1650f46b5a66SChristoph Hellwig devstr = vol_args->name; 165158dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 165258dfae63SZhangZhen if (ret) 1653bb059a37SDavid Sterba goto out_finish; 1654dfd79829SMiao Xie if (!devid) { 1655dfd79829SMiao Xie ret = -EINVAL; 1656bb059a37SDavid Sterba goto out_finish; 1657dfd79829SMiao Xie } 16580b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1659f46b5a66SChristoph Hellwig } 1660dba60f3fSMiao Xie 1661562d7b15SJosef Bacik args.devid = devid; 1662562d7b15SJosef Bacik device = btrfs_find_device(fs_info->fs_devices, &args); 1663f46b5a66SChristoph Hellwig if (!device) { 16640b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1665c1c9ff7cSGeert Uytterhoeven devid); 1666dfd79829SMiao Xie ret = -ENODEV; 1667bb059a37SDavid Sterba goto out_finish; 1668f46b5a66SChristoph Hellwig } 1669dba60f3fSMiao Xie 1670ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16710b246afaSJeff Mahoney btrfs_info(fs_info, 1672efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1673c1c9ff7cSGeert Uytterhoeven devid); 1674dfd79829SMiao Xie ret = -EPERM; 1675bb059a37SDavid Sterba goto out_finish; 16764e42ae1bSLiu Bo } 16774e42ae1bSLiu Bo 1678f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1679f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1680f46b5a66SChristoph Hellwig else { 1681f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1682f46b5a66SChristoph Hellwig mod = -1; 1683f46b5a66SChristoph Hellwig sizestr++; 1684f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1685f46b5a66SChristoph Hellwig mod = 1; 1686f46b5a66SChristoph Hellwig sizestr++; 1687f46b5a66SChristoph Hellwig } 16889a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16899a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1690f46b5a66SChristoph Hellwig ret = -EINVAL; 1691bb059a37SDavid Sterba goto out_finish; 1692f46b5a66SChristoph Hellwig } 1693f46b5a66SChristoph Hellwig } 1694f46b5a66SChristoph Hellwig 1695401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1696dfd79829SMiao Xie ret = -EPERM; 1697bb059a37SDavid Sterba goto out_finish; 169863a212abSStefan Behrens } 169963a212abSStefan Behrens 17007cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1701f46b5a66SChristoph Hellwig 1702f46b5a66SChristoph Hellwig if (mod < 0) { 1703f46b5a66SChristoph Hellwig if (new_size > old_size) { 1704f46b5a66SChristoph Hellwig ret = -EINVAL; 1705bb059a37SDavid Sterba goto out_finish; 1706f46b5a66SChristoph Hellwig } 1707f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1708f46b5a66SChristoph Hellwig } else if (mod > 0) { 1709eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1710902c68a4SGui Hecheng ret = -ERANGE; 1711bb059a37SDavid Sterba goto out_finish; 1712eb8052e0SWenliang Fan } 1713f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1714f46b5a66SChristoph Hellwig } 1715f46b5a66SChristoph Hellwig 1716ee22184bSByongho Lee if (new_size < SZ_256M) { 1717f46b5a66SChristoph Hellwig ret = -EINVAL; 1718bb059a37SDavid Sterba goto out_finish; 1719f46b5a66SChristoph Hellwig } 1720f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1721f46b5a66SChristoph Hellwig ret = -EFBIG; 1722bb059a37SDavid Sterba goto out_finish; 1723f46b5a66SChristoph Hellwig } 1724f46b5a66SChristoph Hellwig 172547f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1726f46b5a66SChristoph Hellwig 1727f46b5a66SChristoph Hellwig if (new_size > old_size) { 1728a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 172998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 173098d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1731bb059a37SDavid Sterba goto out_finish; 173298d5dc13STsutomu Itoh } 1733f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17343a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1735ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1736f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17370253f40eSjeff.liu } /* equal, nothing need to do */ 1738f46b5a66SChristoph Hellwig 1739faf8f7b9SMarcos Paulo de Souza if (ret == 0 && new_size != old_size) 1740faf8f7b9SMarcos Paulo de Souza btrfs_info_in_rcu(fs_info, 1741faf8f7b9SMarcos Paulo de Souza "resize device %s (devid %llu) from %llu to %llu", 1742faf8f7b9SMarcos Paulo de Souza rcu_str_deref(device->name), device->devid, 1743faf8f7b9SMarcos Paulo de Souza old_size, new_size); 1744bb059a37SDavid Sterba out_finish: 1745bb059a37SDavid Sterba btrfs_exclop_finish(fs_info); 1746c9e9f97bSIlya Dryomov out_free: 1747f46b5a66SChristoph Hellwig kfree(vol_args); 1748bb059a37SDavid Sterba out_drop: 174918f39c41SIlya Dryomov mnt_drop_write_file(file); 1750f46b5a66SChristoph Hellwig return ret; 1751f46b5a66SChristoph Hellwig } 1752f46b5a66SChristoph Hellwig 17535d54c67eSNikolay Borisov static noinline int __btrfs_ioctl_snap_create(struct file *file, 17544d4340c9SChristian Brauner struct user_namespace *mnt_userns, 175552f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17565d54c67eSNikolay Borisov bool readonly, 17578696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1758f46b5a66SChristoph Hellwig { 1759f46b5a66SChristoph Hellwig int namelen; 17603de4586cSChris Mason int ret = 0; 1761f46b5a66SChristoph Hellwig 1762325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1763325c50e3SJeff Mahoney return -ENOTDIR; 1764325c50e3SJeff Mahoney 1765a874a63eSLiu Bo ret = mnt_want_write_file(file); 1766a874a63eSLiu Bo if (ret) 1767a874a63eSLiu Bo goto out; 1768a874a63eSLiu Bo 176972fd032eSSage Weil namelen = strlen(name); 177072fd032eSSage Weil if (strchr(name, '/')) { 1771f46b5a66SChristoph Hellwig ret = -EINVAL; 1772a874a63eSLiu Bo goto out_drop_write; 1773f46b5a66SChristoph Hellwig } 1774f46b5a66SChristoph Hellwig 177516780cabSChris Mason if (name[0] == '.' && 177616780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 177716780cabSChris Mason ret = -EEXIST; 1778a874a63eSLiu Bo goto out_drop_write; 177916780cabSChris Mason } 178016780cabSChris Mason 17813de4586cSChris Mason if (subvol) { 17824d4340c9SChristian Brauner ret = btrfs_mksubvol(&file->f_path, mnt_userns, name, 17834d4340c9SChristian Brauner namelen, NULL, readonly, inherit); 1784cb8e7090SChristoph Hellwig } else { 17852903ff01SAl Viro struct fd src = fdget(fd); 17863de4586cSChris Mason struct inode *src_inode; 17872903ff01SAl Viro if (!src.file) { 17883de4586cSChris Mason ret = -EINVAL; 1789a874a63eSLiu Bo goto out_drop_write; 17903de4586cSChris Mason } 17913de4586cSChris Mason 1792496ad9aaSAl Viro src_inode = file_inode(src.file); 1793496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1794c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1795efe120a0SFrank Holton "Snapshot src from another FS"); 179623ad5b17SKusanagi Kouichi ret = -EXDEV; 17974d4340c9SChristian Brauner } else if (!inode_owner_or_capable(mnt_userns, src_inode)) { 1798d0242061SDavid Sterba /* 1799d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1800d0242061SDavid Sterba * are limited to own subvolumes only 1801d0242061SDavid Sterba */ 1802d0242061SDavid Sterba ret = -EPERM; 1803ecd18815SAl Viro } else { 18044d4340c9SChristian Brauner ret = btrfs_mksnapshot(&file->f_path, mnt_userns, 18054d4340c9SChristian Brauner name, namelen, 180672fd032eSSage Weil BTRFS_I(src_inode)->root, 18075d54c67eSNikolay Borisov readonly, inherit); 1808ecd18815SAl Viro } 18092903ff01SAl Viro fdput(src); 1810cb8e7090SChristoph Hellwig } 1811a874a63eSLiu Bo out_drop_write: 1812a874a63eSLiu Bo mnt_drop_write_file(file); 1813f46b5a66SChristoph Hellwig out: 181472fd032eSSage Weil return ret; 181572fd032eSSage Weil } 181672fd032eSSage Weil 181772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1818fa0d2b9bSLi Zefan void __user *arg, int subvol) 181972fd032eSSage Weil { 1820fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 182172fd032eSSage Weil int ret; 182272fd032eSSage Weil 1823325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1824325c50e3SJeff Mahoney return -ENOTDIR; 1825325c50e3SJeff Mahoney 1826fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1827fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1828fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1829fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1830fa0d2b9bSLi Zefan 18314d4340c9SChristian Brauner ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), 18324d4340c9SChristian Brauner vol_args->name, vol_args->fd, subvol, 18334d4340c9SChristian Brauner false, NULL); 1834fa0d2b9bSLi Zefan 1835fa0d2b9bSLi Zefan kfree(vol_args); 1836fa0d2b9bSLi Zefan return ret; 1837fa0d2b9bSLi Zefan } 1838fa0d2b9bSLi Zefan 1839fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1840fa0d2b9bSLi Zefan void __user *arg, int subvol) 1841fa0d2b9bSLi Zefan { 1842fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1843fa0d2b9bSLi Zefan int ret; 1844b83cc969SLi Zefan bool readonly = false; 18456f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 184672fd032eSSage Weil 1847325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1848325c50e3SJeff Mahoney return -ENOTDIR; 1849325c50e3SJeff Mahoney 1850fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1851fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1852fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1853fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1854fdfb1e4fSLi Zefan 1855673990dbSDavid Sterba if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) { 1856b83cc969SLi Zefan ret = -EOPNOTSUPP; 1857c47ca32dSDan Carpenter goto free_args; 1858fdfb1e4fSLi Zefan } 1859fdfb1e4fSLi Zefan 1860b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1861b83cc969SLi Zefan readonly = true; 18626f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 18635011c5a6SDan Carpenter u64 nums; 18645011c5a6SDan Carpenter 18655011c5a6SDan Carpenter if (vol_args->size < sizeof(*inherit) || 18665011c5a6SDan Carpenter vol_args->size > PAGE_SIZE) { 18676f72c7e2SArne Jansen ret = -EINVAL; 1868c47ca32dSDan Carpenter goto free_args; 18696f72c7e2SArne Jansen } 18706f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18716f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18726f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1873c47ca32dSDan Carpenter goto free_args; 18746f72c7e2SArne Jansen } 18755011c5a6SDan Carpenter 18765011c5a6SDan Carpenter if (inherit->num_qgroups > PAGE_SIZE || 18775011c5a6SDan Carpenter inherit->num_ref_copies > PAGE_SIZE || 18785011c5a6SDan Carpenter inherit->num_excl_copies > PAGE_SIZE) { 18795011c5a6SDan Carpenter ret = -EINVAL; 18805011c5a6SDan Carpenter goto free_inherit; 18815011c5a6SDan Carpenter } 18825011c5a6SDan Carpenter 18835011c5a6SDan Carpenter nums = inherit->num_qgroups + 2 * inherit->num_ref_copies + 18845011c5a6SDan Carpenter 2 * inherit->num_excl_copies; 18855011c5a6SDan Carpenter if (vol_args->size != struct_size(inherit, qgroups, nums)) { 18865011c5a6SDan Carpenter ret = -EINVAL; 18875011c5a6SDan Carpenter goto free_inherit; 18885011c5a6SDan Carpenter } 18896f72c7e2SArne Jansen } 189075eaa0e2SSage Weil 18914d4340c9SChristian Brauner ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), 18924d4340c9SChristian Brauner vol_args->name, vol_args->fd, subvol, 18934d4340c9SChristian Brauner readonly, inherit); 1894c47ca32dSDan Carpenter if (ret) 1895c47ca32dSDan Carpenter goto free_inherit; 1896c47ca32dSDan Carpenter free_inherit: 18976f72c7e2SArne Jansen kfree(inherit); 1898c47ca32dSDan Carpenter free_args: 1899c47ca32dSDan Carpenter kfree(vol_args); 1900f46b5a66SChristoph Hellwig return ret; 1901f46b5a66SChristoph Hellwig } 1902f46b5a66SChristoph Hellwig 19030caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 19040caa102dSLi Zefan void __user *arg) 19050caa102dSLi Zefan { 1906496ad9aaSAl Viro struct inode *inode = file_inode(file); 19070b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19080caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19090caa102dSLi Zefan int ret = 0; 19100caa102dSLi Zefan u64 flags = 0; 19110caa102dSLi Zefan 19124a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 19130caa102dSLi Zefan return -EINVAL; 19140caa102dSLi Zefan 19150b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 19160caa102dSLi Zefan if (btrfs_root_readonly(root)) 19170caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 19180b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 19190caa102dSLi Zefan 19200caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 19210caa102dSLi Zefan ret = -EFAULT; 19220caa102dSLi Zefan 19230caa102dSLi Zefan return ret; 19240caa102dSLi Zefan } 19250caa102dSLi Zefan 19260caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19270caa102dSLi Zefan void __user *arg) 19280caa102dSLi Zefan { 1929496ad9aaSAl Viro struct inode *inode = file_inode(file); 19300b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19310caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19320caa102dSLi Zefan struct btrfs_trans_handle *trans; 19330caa102dSLi Zefan u64 root_flags; 19340caa102dSLi Zefan u64 flags; 19350caa102dSLi Zefan int ret = 0; 19360caa102dSLi Zefan 193739e1674fSChristian Brauner if (!inode_owner_or_capable(file_mnt_user_ns(file), inode)) 1938bd60ea0fSDavid Sterba return -EPERM; 1939bd60ea0fSDavid Sterba 1940b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1941b9ca0664SLiu Bo if (ret) 1942b9ca0664SLiu Bo goto out; 19430caa102dSLi Zefan 19444a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1945b9ca0664SLiu Bo ret = -EINVAL; 1946b9ca0664SLiu Bo goto out_drop_write; 1947b9ca0664SLiu Bo } 19480caa102dSLi Zefan 1949b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1950b9ca0664SLiu Bo ret = -EFAULT; 1951b9ca0664SLiu Bo goto out_drop_write; 1952b9ca0664SLiu Bo } 19530caa102dSLi Zefan 1954b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1955b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1956b9ca0664SLiu Bo goto out_drop_write; 1957b9ca0664SLiu Bo } 19580caa102dSLi Zefan 19590b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19600caa102dSLi Zefan 19610caa102dSLi Zefan /* nothing to do */ 19620caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1963b9ca0664SLiu Bo goto out_drop_sem; 19640caa102dSLi Zefan 19650caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19662c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19670caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19680caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19692c686537SDavid Sterba } else { 19702c686537SDavid Sterba /* 19712c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19722c686537SDavid Sterba * send 19732c686537SDavid Sterba */ 19742c686537SDavid Sterba spin_lock(&root->root_item_lock); 19752c686537SDavid Sterba if (root->send_in_progress == 0) { 19760caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19770caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19782c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19792c686537SDavid Sterba } else { 19802c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19810b246afaSJeff Mahoney btrfs_warn(fs_info, 19822c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19832c686537SDavid Sterba root->root_key.objectid); 19842c686537SDavid Sterba ret = -EPERM; 19852c686537SDavid Sterba goto out_drop_sem; 19862c686537SDavid Sterba } 19872c686537SDavid Sterba } 19880caa102dSLi Zefan 19890caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19900caa102dSLi Zefan if (IS_ERR(trans)) { 19910caa102dSLi Zefan ret = PTR_ERR(trans); 19920caa102dSLi Zefan goto out_reset; 19930caa102dSLi Zefan } 19940caa102dSLi Zefan 19950b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19960caa102dSLi Zefan &root->root_key, &root->root_item); 19979417ebc8SNikolay Borisov if (ret < 0) { 19989417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19999417ebc8SNikolay Borisov goto out_reset; 20009417ebc8SNikolay Borisov } 20010caa102dSLi Zefan 20029417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 20039417ebc8SNikolay Borisov 20040caa102dSLi Zefan out_reset: 20050caa102dSLi Zefan if (ret) 20060caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 2007b9ca0664SLiu Bo out_drop_sem: 20080b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 2009b9ca0664SLiu Bo out_drop_write: 2010b9ca0664SLiu Bo mnt_drop_write_file(file); 2011b9ca0664SLiu Bo out: 20120caa102dSLi Zefan return ret; 20130caa102dSLi Zefan } 20140caa102dSLi Zefan 2015ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2016ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2017ac8e9819SChris Mason { 2018abc6e134SChris Mason struct btrfs_key test; 2019abc6e134SChris Mason int ret; 2020abc6e134SChris Mason 2021abc6e134SChris Mason test.objectid = sk->min_objectid; 2022abc6e134SChris Mason test.type = sk->min_type; 2023abc6e134SChris Mason test.offset = sk->min_offset; 2024abc6e134SChris Mason 2025abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2026abc6e134SChris Mason if (ret < 0) 2027ac8e9819SChris Mason return 0; 2028abc6e134SChris Mason 2029abc6e134SChris Mason test.objectid = sk->max_objectid; 2030abc6e134SChris Mason test.type = sk->max_type; 2031abc6e134SChris Mason test.offset = sk->max_offset; 2032abc6e134SChris Mason 2033abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2034abc6e134SChris Mason if (ret > 0) 2035ac8e9819SChris Mason return 0; 2036ac8e9819SChris Mason return 1; 2037ac8e9819SChris Mason } 2038ac8e9819SChris Mason 2039df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2040ac8e9819SChris Mason struct btrfs_key *key, 2041ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20429b6e817dSGerhard Heift size_t *buf_size, 2043ba346b35SGerhard Heift char __user *ubuf, 2044ac8e9819SChris Mason unsigned long *sk_offset, 2045ac8e9819SChris Mason int *num_found) 2046ac8e9819SChris Mason { 2047ac8e9819SChris Mason u64 found_transid; 2048ac8e9819SChris Mason struct extent_buffer *leaf; 2049ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2050dd81d459SNaohiro Aota struct btrfs_key test; 2051ac8e9819SChris Mason unsigned long item_off; 2052ac8e9819SChris Mason unsigned long item_len; 2053ac8e9819SChris Mason int nritems; 2054ac8e9819SChris Mason int i; 2055ac8e9819SChris Mason int slot; 2056ac8e9819SChris Mason int ret = 0; 2057ac8e9819SChris Mason 2058ac8e9819SChris Mason leaf = path->nodes[0]; 2059ac8e9819SChris Mason slot = path->slots[0]; 2060ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2061ac8e9819SChris Mason 2062ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2063ac8e9819SChris Mason i = nritems; 2064ac8e9819SChris Mason goto advance_key; 2065ac8e9819SChris Mason } 2066ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2067ac8e9819SChris Mason 2068ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2069ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2070ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2071ac8e9819SChris Mason 207203b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 207303b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 207403b71c6cSGabriel de Perthuis continue; 207503b71c6cSGabriel de Perthuis 20769b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20778f5f6178SGerhard Heift if (*num_found) { 2078ac8e9819SChris Mason ret = 1; 20798f5f6178SGerhard Heift goto out; 20808f5f6178SGerhard Heift } 20818f5f6178SGerhard Heift 20828f5f6178SGerhard Heift /* 20838f5f6178SGerhard Heift * return one empty item back for v1, which does not 20848f5f6178SGerhard Heift * handle -EOVERFLOW 20858f5f6178SGerhard Heift */ 20868f5f6178SGerhard Heift 20879b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2088ac8e9819SChris Mason item_len = 0; 20898f5f6178SGerhard Heift ret = -EOVERFLOW; 20908f5f6178SGerhard Heift } 2091ac8e9819SChris Mason 20929b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2093ac8e9819SChris Mason ret = 1; 209425c9bc2eSGerhard Heift goto out; 2095ac8e9819SChris Mason } 2096ac8e9819SChris Mason 2097ac8e9819SChris Mason sh.objectid = key->objectid; 2098ac8e9819SChris Mason sh.offset = key->offset; 2099ac8e9819SChris Mason sh.type = key->type; 2100ac8e9819SChris Mason sh.len = item_len; 2101ac8e9819SChris Mason sh.transid = found_transid; 2102ac8e9819SChris Mason 2103a48b73ecSJosef Bacik /* 2104a48b73ecSJosef Bacik * Copy search result header. If we fault then loop again so we 2105a48b73ecSJosef Bacik * can fault in the pages and -EFAULT there if there's a 2106a48b73ecSJosef Bacik * problem. Otherwise we'll fault and then copy the buffer in 2107a48b73ecSJosef Bacik * properly this next time through 2108a48b73ecSJosef Bacik */ 2109a48b73ecSJosef Bacik if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) { 2110a48b73ecSJosef Bacik ret = 0; 2111ba346b35SGerhard Heift goto out; 2112ba346b35SGerhard Heift } 2113ba346b35SGerhard Heift 2114ac8e9819SChris Mason *sk_offset += sizeof(sh); 2115ac8e9819SChris Mason 2116ac8e9819SChris Mason if (item_len) { 2117ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2118a48b73ecSJosef Bacik /* 2119a48b73ecSJosef Bacik * Copy the item, same behavior as above, but reset the 2120a48b73ecSJosef Bacik * * sk_offset so we copy the full thing again. 2121a48b73ecSJosef Bacik */ 2122a48b73ecSJosef Bacik if (read_extent_buffer_to_user_nofault(leaf, up, 2123ba346b35SGerhard Heift item_off, item_len)) { 2124a48b73ecSJosef Bacik ret = 0; 2125a48b73ecSJosef Bacik *sk_offset -= sizeof(sh); 2126ba346b35SGerhard Heift goto out; 2127ba346b35SGerhard Heift } 2128ba346b35SGerhard Heift 2129ac8e9819SChris Mason *sk_offset += item_len; 2130ac8e9819SChris Mason } 2131e2156867SHugo Mills (*num_found)++; 2132ac8e9819SChris Mason 21338f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21348f5f6178SGerhard Heift goto out; 21358f5f6178SGerhard Heift 213625c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 213725c9bc2eSGerhard Heift ret = 1; 213825c9bc2eSGerhard Heift goto out; 213925c9bc2eSGerhard Heift } 2140ac8e9819SChris Mason } 2141ac8e9819SChris Mason advance_key: 2142ac8e9819SChris Mason ret = 0; 2143dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2144dd81d459SNaohiro Aota test.type = sk->max_type; 2145dd81d459SNaohiro Aota test.offset = sk->max_offset; 2146dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2147dd81d459SNaohiro Aota ret = 1; 2148dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2149abc6e134SChris Mason key->offset++; 2150dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2151abc6e134SChris Mason key->offset = 0; 2152abc6e134SChris Mason key->type++; 2153dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2154abc6e134SChris Mason key->offset = 0; 2155abc6e134SChris Mason key->type = 0; 2156abc6e134SChris Mason key->objectid++; 2157abc6e134SChris Mason } else 2158abc6e134SChris Mason ret = 1; 215925c9bc2eSGerhard Heift out: 2160ba346b35SGerhard Heift /* 2161ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2162ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2163ba346b35SGerhard Heift * * all items were found 2164ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2165ba346b35SGerhard Heift * leaf 2166ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2167ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2168ba346b35SGerhard Heift */ 2169ac8e9819SChris Mason return ret; 2170ac8e9819SChris Mason } 2171ac8e9819SChris Mason 2172ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 217312544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21749b6e817dSGerhard Heift size_t *buf_size, 2175ba346b35SGerhard Heift char __user *ubuf) 2176ac8e9819SChris Mason { 21770b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2178ac8e9819SChris Mason struct btrfs_root *root; 2179ac8e9819SChris Mason struct btrfs_key key; 2180ac8e9819SChris Mason struct btrfs_path *path; 2181ac8e9819SChris Mason int ret; 2182ac8e9819SChris Mason int num_found = 0; 2183ac8e9819SChris Mason unsigned long sk_offset = 0; 2184ac8e9819SChris Mason 21859b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21869b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 218712544442SGerhard Heift return -EOVERFLOW; 21889b6e817dSGerhard Heift } 218912544442SGerhard Heift 2190ac8e9819SChris Mason path = btrfs_alloc_path(); 2191ac8e9819SChris Mason if (!path) 2192ac8e9819SChris Mason return -ENOMEM; 2193ac8e9819SChris Mason 2194ac8e9819SChris Mason if (sk->tree_id == 0) { 2195ac8e9819SChris Mason /* search the root of the inode that was passed */ 219600246528SJosef Bacik root = btrfs_grab_root(BTRFS_I(inode)->root); 2197ac8e9819SChris Mason } else { 219856e9357aSDavid Sterba root = btrfs_get_fs_root(info, sk->tree_id, true); 2199ac8e9819SChris Mason if (IS_ERR(root)) { 2200ac8e9819SChris Mason btrfs_free_path(path); 2201ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2202ac8e9819SChris Mason } 2203ac8e9819SChris Mason } 2204ac8e9819SChris Mason 2205ac8e9819SChris Mason key.objectid = sk->min_objectid; 2206ac8e9819SChris Mason key.type = sk->min_type; 2207ac8e9819SChris Mason key.offset = sk->min_offset; 2208ac8e9819SChris Mason 2209ac8e9819SChris Mason while (1) { 22101c78544eSFilipe Manana ret = fault_in_pages_writeable(ubuf + sk_offset, 22111c78544eSFilipe Manana *buf_size - sk_offset); 2212a48b73ecSJosef Bacik if (ret) 2213a48b73ecSJosef Bacik break; 2214a48b73ecSJosef Bacik 22156174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2216ac8e9819SChris Mason if (ret != 0) { 2217ac8e9819SChris Mason if (ret > 0) 2218ac8e9819SChris Mason ret = 0; 2219ac8e9819SChris Mason goto err; 2220ac8e9819SChris Mason } 2221df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2222ac8e9819SChris Mason &sk_offset, &num_found); 2223b3b4aa74SDavid Sterba btrfs_release_path(path); 222425c9bc2eSGerhard Heift if (ret) 2225ac8e9819SChris Mason break; 2226ac8e9819SChris Mason 2227ac8e9819SChris Mason } 22288f5f6178SGerhard Heift if (ret > 0) 2229ac8e9819SChris Mason ret = 0; 2230ac8e9819SChris Mason err: 2231ac8e9819SChris Mason sk->nr_items = num_found; 223200246528SJosef Bacik btrfs_put_root(root); 2233ac8e9819SChris Mason btrfs_free_path(path); 2234ac8e9819SChris Mason return ret; 2235ac8e9819SChris Mason } 2236ac8e9819SChris Mason 2237ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2238ac8e9819SChris Mason void __user *argp) 2239ac8e9819SChris Mason { 2240ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2241ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2242ac8e9819SChris Mason struct inode *inode; 2243ac8e9819SChris Mason int ret; 22449b6e817dSGerhard Heift size_t buf_size; 2245ac8e9819SChris Mason 2246ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2247ac8e9819SChris Mason return -EPERM; 2248ac8e9819SChris Mason 2249ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2250ac8e9819SChris Mason 2251ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2252ba346b35SGerhard Heift return -EFAULT; 2253ba346b35SGerhard Heift 2254ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2255ac8e9819SChris Mason 2256496ad9aaSAl Viro inode = file_inode(file); 2257ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22588f5f6178SGerhard Heift 22598f5f6178SGerhard Heift /* 22608f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22618f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22628f5f6178SGerhard Heift */ 22638f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22648f5f6178SGerhard Heift ret = 0; 22658f5f6178SGerhard Heift 2266ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2267ac8e9819SChris Mason ret = -EFAULT; 2268ac8e9819SChris Mason return ret; 2269ac8e9819SChris Mason } 2270ac8e9819SChris Mason 2271cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2272cc68a8a5SGerhard Heift void __user *argp) 2273cc68a8a5SGerhard Heift { 2274cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2275cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2276cc68a8a5SGerhard Heift struct inode *inode; 2277cc68a8a5SGerhard Heift int ret; 2278cc68a8a5SGerhard Heift size_t buf_size; 2279ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2280cc68a8a5SGerhard Heift 2281cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2282cc68a8a5SGerhard Heift return -EPERM; 2283cc68a8a5SGerhard Heift 2284cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2285cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2286cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2287cc68a8a5SGerhard Heift return -EFAULT; 2288cc68a8a5SGerhard Heift 2289cc68a8a5SGerhard Heift buf_size = args.buf_size; 2290cc68a8a5SGerhard Heift 2291cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2292cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2293cc68a8a5SGerhard Heift buf_size = buf_limit; 2294cc68a8a5SGerhard Heift 2295cc68a8a5SGerhard Heift inode = file_inode(file); 2296cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2297718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2298cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2299cc68a8a5SGerhard Heift ret = -EFAULT; 2300cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2301cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2302cc68a8a5SGerhard Heift ret = -EFAULT; 2303cc68a8a5SGerhard Heift 230476dda93cSYan, Zheng return ret; 230576dda93cSYan, Zheng } 230676dda93cSYan, Zheng 230798d377a0STARUISI Hiroaki /* 2308ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2309ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 231098d377a0STARUISI Hiroaki */ 231198d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 231298d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 231398d377a0STARUISI Hiroaki { 231498d377a0STARUISI Hiroaki struct btrfs_root *root; 231598d377a0STARUISI Hiroaki struct btrfs_key key; 2316ac8e9819SChris Mason char *ptr; 231798d377a0STARUISI Hiroaki int ret = -1; 231898d377a0STARUISI Hiroaki int slot; 231998d377a0STARUISI Hiroaki int len; 232098d377a0STARUISI Hiroaki int total_len = 0; 232198d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 232298d377a0STARUISI Hiroaki struct extent_buffer *l; 232398d377a0STARUISI Hiroaki struct btrfs_path *path; 232498d377a0STARUISI Hiroaki 232598d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 232698d377a0STARUISI Hiroaki name[0]='\0'; 232798d377a0STARUISI Hiroaki return 0; 232898d377a0STARUISI Hiroaki } 232998d377a0STARUISI Hiroaki 233098d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 233198d377a0STARUISI Hiroaki if (!path) 233298d377a0STARUISI Hiroaki return -ENOMEM; 233398d377a0STARUISI Hiroaki 2334c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 233598d377a0STARUISI Hiroaki 233656e9357aSDavid Sterba root = btrfs_get_fs_root(info, tree_id, true); 233798d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2338ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 233988234012SJosef Bacik root = NULL; 234088234012SJosef Bacik goto out; 234188234012SJosef Bacik } 234298d377a0STARUISI Hiroaki 234398d377a0STARUISI Hiroaki key.objectid = dirid; 234498d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23458ad6fcabSChris Mason key.offset = (u64)-1; 234698d377a0STARUISI Hiroaki 234798d377a0STARUISI Hiroaki while (1) { 23480ff40a91SMarcos Paulo de Souza ret = btrfs_search_backwards(root, &key, path); 234918674c6cSFilipe David Borba Manana if (ret < 0) 235018674c6cSFilipe David Borba Manana goto out; 235118674c6cSFilipe David Borba Manana else if (ret > 0) { 2352ac8e9819SChris Mason ret = -ENOENT; 235398d377a0STARUISI Hiroaki goto out; 2354ac8e9819SChris Mason } 235518674c6cSFilipe David Borba Manana 235618674c6cSFilipe David Borba Manana l = path->nodes[0]; 235718674c6cSFilipe David Borba Manana slot = path->slots[0]; 235898d377a0STARUISI Hiroaki 235998d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 236098d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 236198d377a0STARUISI Hiroaki ptr -= len + 1; 236298d377a0STARUISI Hiroaki total_len += len + 1; 2363a696cf35SFilipe David Borba Manana if (ptr < name) { 2364a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 236598d377a0STARUISI Hiroaki goto out; 2366a696cf35SFilipe David Borba Manana } 236798d377a0STARUISI Hiroaki 236898d377a0STARUISI Hiroaki *(ptr + len) = '/'; 236998d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 237098d377a0STARUISI Hiroaki 237198d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 237298d377a0STARUISI Hiroaki break; 237398d377a0STARUISI Hiroaki 2374b3b4aa74SDavid Sterba btrfs_release_path(path); 237598d377a0STARUISI Hiroaki key.objectid = key.offset; 23768ad6fcabSChris Mason key.offset = (u64)-1; 237798d377a0STARUISI Hiroaki dirid = key.objectid; 237898d377a0STARUISI Hiroaki } 237977906a50SLi Zefan memmove(name, ptr, total_len); 238098d377a0STARUISI Hiroaki name[total_len] = '\0'; 238198d377a0STARUISI Hiroaki ret = 0; 238298d377a0STARUISI Hiroaki out: 238300246528SJosef Bacik btrfs_put_root(root); 238498d377a0STARUISI Hiroaki btrfs_free_path(path); 2385ac8e9819SChris Mason return ret; 2386ac8e9819SChris Mason } 2387ac8e9819SChris Mason 23886623d9a0SChristian Brauner static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns, 23896623d9a0SChristian Brauner struct inode *inode, 239023d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 239123d0b79dSTomohiro Misono { 239223d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 239323d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 239423d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 239523d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 239623d0b79dSTomohiro Misono u64 dirid = args->dirid; 239723d0b79dSTomohiro Misono unsigned long item_off; 239823d0b79dSTomohiro Misono unsigned long item_len; 239923d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 240023d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 2401b8a49ae1SJosef Bacik struct btrfs_root *root = NULL; 240223d0b79dSTomohiro Misono struct btrfs_path *path; 240323d0b79dSTomohiro Misono struct btrfs_key key, key2; 240423d0b79dSTomohiro Misono struct extent_buffer *leaf; 240523d0b79dSTomohiro Misono struct inode *temp_inode; 240623d0b79dSTomohiro Misono char *ptr; 240723d0b79dSTomohiro Misono int slot; 240823d0b79dSTomohiro Misono int len; 240923d0b79dSTomohiro Misono int total_len = 0; 241023d0b79dSTomohiro Misono int ret; 241123d0b79dSTomohiro Misono 241223d0b79dSTomohiro Misono path = btrfs_alloc_path(); 241323d0b79dSTomohiro Misono if (!path) 241423d0b79dSTomohiro Misono return -ENOMEM; 241523d0b79dSTomohiro Misono 241623d0b79dSTomohiro Misono /* 241723d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 241823d0b79dSTomohiro Misono * construct the path in from the bottom up. 241923d0b79dSTomohiro Misono */ 242023d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 242123d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 242223d0b79dSTomohiro Misono 242356e9357aSDavid Sterba root = btrfs_get_fs_root(fs_info, treeid, true); 242423d0b79dSTomohiro Misono if (IS_ERR(root)) { 242523d0b79dSTomohiro Misono ret = PTR_ERR(root); 242623d0b79dSTomohiro Misono goto out; 242723d0b79dSTomohiro Misono } 242823d0b79dSTomohiro Misono 242923d0b79dSTomohiro Misono key.objectid = dirid; 243023d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 243123d0b79dSTomohiro Misono key.offset = (u64)-1; 243223d0b79dSTomohiro Misono while (1) { 24330ff40a91SMarcos Paulo de Souza ret = btrfs_search_backwards(root, &key, path); 24340ff40a91SMarcos Paulo de Souza if (ret < 0) 2435b8a49ae1SJosef Bacik goto out_put; 24360ff40a91SMarcos Paulo de Souza else if (ret > 0) { 243723d0b79dSTomohiro Misono ret = -ENOENT; 2438b8a49ae1SJosef Bacik goto out_put; 243923d0b79dSTomohiro Misono } 244023d0b79dSTomohiro Misono 244123d0b79dSTomohiro Misono leaf = path->nodes[0]; 244223d0b79dSTomohiro Misono slot = path->slots[0]; 244323d0b79dSTomohiro Misono 244423d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 244523d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 244623d0b79dSTomohiro Misono ptr -= len + 1; 244723d0b79dSTomohiro Misono total_len += len + 1; 244823d0b79dSTomohiro Misono if (ptr < args->path) { 244923d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 2450b8a49ae1SJosef Bacik goto out_put; 245123d0b79dSTomohiro Misono } 245223d0b79dSTomohiro Misono 245323d0b79dSTomohiro Misono *(ptr + len) = '/'; 245423d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 245523d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 245623d0b79dSTomohiro Misono 245723d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 245823d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 245923d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 246023d0b79dSTomohiro Misono if (ret < 0) { 2461b8a49ae1SJosef Bacik goto out_put; 246223d0b79dSTomohiro Misono } else if (ret > 0) { 246323d0b79dSTomohiro Misono ret = -ENOENT; 2464b8a49ae1SJosef Bacik goto out_put; 246523d0b79dSTomohiro Misono } 246623d0b79dSTomohiro Misono 246723d0b79dSTomohiro Misono leaf = path->nodes[0]; 246823d0b79dSTomohiro Misono slot = path->slots[0]; 246923d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 247023d0b79dSTomohiro Misono if (key2.objectid != dirid) { 247123d0b79dSTomohiro Misono ret = -ENOENT; 2472b8a49ae1SJosef Bacik goto out_put; 247323d0b79dSTomohiro Misono } 247423d0b79dSTomohiro Misono 24750202e83fSDavid Sterba temp_inode = btrfs_iget(sb, key2.objectid, root); 24763ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24773ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 2478b8a49ae1SJosef Bacik goto out_put; 24793ca57bd6SMisono Tomohiro } 24806623d9a0SChristian Brauner ret = inode_permission(mnt_userns, temp_inode, 248147291baaSChristian Brauner MAY_READ | MAY_EXEC); 248223d0b79dSTomohiro Misono iput(temp_inode); 248323d0b79dSTomohiro Misono if (ret) { 248423d0b79dSTomohiro Misono ret = -EACCES; 2485b8a49ae1SJosef Bacik goto out_put; 248623d0b79dSTomohiro Misono } 248723d0b79dSTomohiro Misono 248823d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 248923d0b79dSTomohiro Misono break; 249023d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 249123d0b79dSTomohiro Misono ret = -EACCES; 2492b8a49ae1SJosef Bacik goto out_put; 249323d0b79dSTomohiro Misono } 249423d0b79dSTomohiro Misono 249523d0b79dSTomohiro Misono btrfs_release_path(path); 249623d0b79dSTomohiro Misono key.objectid = key.offset; 249723d0b79dSTomohiro Misono key.offset = (u64)-1; 249823d0b79dSTomohiro Misono dirid = key.objectid; 249923d0b79dSTomohiro Misono } 250023d0b79dSTomohiro Misono 250123d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 250223d0b79dSTomohiro Misono args->path[total_len] = '\0'; 250300246528SJosef Bacik btrfs_put_root(root); 2504b8a49ae1SJosef Bacik root = NULL; 250523d0b79dSTomohiro Misono btrfs_release_path(path); 250623d0b79dSTomohiro Misono } 250723d0b79dSTomohiro Misono 250823d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 250923d0b79dSTomohiro Misono key.objectid = treeid; 251023d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 251123d0b79dSTomohiro Misono key.offset = args->treeid; 2512b8a49ae1SJosef Bacik ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 251323d0b79dSTomohiro Misono if (ret < 0) { 251423d0b79dSTomohiro Misono goto out; 251523d0b79dSTomohiro Misono } else if (ret > 0) { 251623d0b79dSTomohiro Misono ret = -ENOENT; 251723d0b79dSTomohiro Misono goto out; 251823d0b79dSTomohiro Misono } 251923d0b79dSTomohiro Misono 252023d0b79dSTomohiro Misono leaf = path->nodes[0]; 252123d0b79dSTomohiro Misono slot = path->slots[0]; 252223d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 252323d0b79dSTomohiro Misono 252423d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 252523d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 252623d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 252723d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 252823d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 252923d0b79dSTomohiro Misono ret = -EINVAL; 253023d0b79dSTomohiro Misono goto out; 253123d0b79dSTomohiro Misono } 253223d0b79dSTomohiro Misono 253323d0b79dSTomohiro Misono /* Copy subvolume's name */ 253423d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 253523d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 253623d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 253723d0b79dSTomohiro Misono args->name[item_len] = 0; 253823d0b79dSTomohiro Misono 2539b8a49ae1SJosef Bacik out_put: 254000246528SJosef Bacik btrfs_put_root(root); 254123d0b79dSTomohiro Misono out: 254223d0b79dSTomohiro Misono btrfs_free_path(path); 254323d0b79dSTomohiro Misono return ret; 254423d0b79dSTomohiro Misono } 254523d0b79dSTomohiro Misono 2546ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2547ac8e9819SChris Mason void __user *argp) 2548ac8e9819SChris Mason { 2549ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2550ac8e9819SChris Mason struct inode *inode; 255101b810b8SDavid Sterba int ret = 0; 2552ac8e9819SChris Mason 25532354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25542354d08fSJulia Lawall if (IS_ERR(args)) 25552354d08fSJulia Lawall return PTR_ERR(args); 2556c2b96929SDan Carpenter 2557496ad9aaSAl Viro inode = file_inode(file); 2558ac8e9819SChris Mason 255901b810b8SDavid Sterba /* 256001b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 256101b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 256201b810b8SDavid Sterba */ 25631b53ac4dSChris Mason if (args->treeid == 0) 25641b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25651b53ac4dSChris Mason 256601b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 256701b810b8SDavid Sterba args->name[0] = 0; 256801b810b8SDavid Sterba goto out; 256901b810b8SDavid Sterba } 257001b810b8SDavid Sterba 257101b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 257201b810b8SDavid Sterba ret = -EPERM; 257301b810b8SDavid Sterba goto out; 257401b810b8SDavid Sterba } 257501b810b8SDavid Sterba 2576ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2577ac8e9819SChris Mason args->treeid, args->objectid, 2578ac8e9819SChris Mason args->name); 2579ac8e9819SChris Mason 258001b810b8SDavid Sterba out: 2581ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2582ac8e9819SChris Mason ret = -EFAULT; 2583ac8e9819SChris Mason 2584ac8e9819SChris Mason kfree(args); 258598d377a0STARUISI Hiroaki return ret; 258698d377a0STARUISI Hiroaki } 258798d377a0STARUISI Hiroaki 258823d0b79dSTomohiro Misono /* 258923d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 259023d0b79dSTomohiro Misono * 259123d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 259223d0b79dSTomohiro Misono * 259323d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 259423d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 259523d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 259623d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 259723d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 259823d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 259923d0b79dSTomohiro Misono */ 260023d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 260123d0b79dSTomohiro Misono { 260223d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 260323d0b79dSTomohiro Misono struct inode *inode; 260423d0b79dSTomohiro Misono int ret; 260523d0b79dSTomohiro Misono 260623d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 260723d0b79dSTomohiro Misono if (IS_ERR(args)) 260823d0b79dSTomohiro Misono return PTR_ERR(args); 260923d0b79dSTomohiro Misono 261023d0b79dSTomohiro Misono inode = file_inode(file); 261123d0b79dSTomohiro Misono 261223d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 261323d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 261423d0b79dSTomohiro Misono /* 261523d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 261623d0b79dSTomohiro Misono * called 261723d0b79dSTomohiro Misono */ 261823d0b79dSTomohiro Misono kfree(args); 261923d0b79dSTomohiro Misono return -EACCES; 262023d0b79dSTomohiro Misono } 262123d0b79dSTomohiro Misono 26226623d9a0SChristian Brauner ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args); 262323d0b79dSTomohiro Misono 262423d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 262523d0b79dSTomohiro Misono ret = -EFAULT; 262623d0b79dSTomohiro Misono 262723d0b79dSTomohiro Misono kfree(args); 262823d0b79dSTomohiro Misono return ret; 262923d0b79dSTomohiro Misono } 263023d0b79dSTomohiro Misono 2631b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2632b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2633b64ec075STomohiro Misono { 2634b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2635b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2636b64ec075STomohiro Misono struct btrfs_root *root; 2637b64ec075STomohiro Misono struct btrfs_path *path; 2638b64ec075STomohiro Misono struct btrfs_key key; 2639b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2640b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2641b64ec075STomohiro Misono struct extent_buffer *leaf; 2642b64ec075STomohiro Misono unsigned long item_off; 2643b64ec075STomohiro Misono unsigned long item_len; 2644b64ec075STomohiro Misono struct inode *inode; 2645b64ec075STomohiro Misono int slot; 2646b64ec075STomohiro Misono int ret = 0; 2647b64ec075STomohiro Misono 2648b64ec075STomohiro Misono path = btrfs_alloc_path(); 2649b64ec075STomohiro Misono if (!path) 2650b64ec075STomohiro Misono return -ENOMEM; 2651b64ec075STomohiro Misono 2652b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2653b64ec075STomohiro Misono if (!subvol_info) { 2654b64ec075STomohiro Misono btrfs_free_path(path); 2655b64ec075STomohiro Misono return -ENOMEM; 2656b64ec075STomohiro Misono } 2657b64ec075STomohiro Misono 2658b64ec075STomohiro Misono inode = file_inode(file); 2659b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2660b64ec075STomohiro Misono 2661b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2662b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 266356e9357aSDavid Sterba root = btrfs_get_fs_root(fs_info, key.objectid, true); 2664b64ec075STomohiro Misono if (IS_ERR(root)) { 2665b64ec075STomohiro Misono ret = PTR_ERR(root); 266604734e84SJosef Bacik goto out_free; 266704734e84SJosef Bacik } 2668b64ec075STomohiro Misono root_item = &root->root_item; 2669b64ec075STomohiro Misono 2670b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2671b64ec075STomohiro Misono 2672b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2673b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2674b64ec075STomohiro Misono 2675b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2676b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2677b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2678b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2679b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2680b64ec075STomohiro Misono 2681b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2682b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2683b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2684b64ec075STomohiro Misono 2685b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2686b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2687b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2688b64ec075STomohiro Misono 2689b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2690b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2691b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2692b64ec075STomohiro Misono 2693b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2694b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2695b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2696b64ec075STomohiro Misono 2697b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2698b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2699b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2700b64ec075STomohiro Misono key.offset = 0; 270104734e84SJosef Bacik ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 2702b64ec075STomohiro Misono if (ret < 0) { 2703b64ec075STomohiro Misono goto out; 2704b64ec075STomohiro Misono } else if (path->slots[0] >= 2705b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 270604734e84SJosef Bacik ret = btrfs_next_leaf(fs_info->tree_root, path); 2707b64ec075STomohiro Misono if (ret < 0) { 2708b64ec075STomohiro Misono goto out; 2709b64ec075STomohiro Misono } else if (ret > 0) { 2710b64ec075STomohiro Misono ret = -EUCLEAN; 2711b64ec075STomohiro Misono goto out; 2712b64ec075STomohiro Misono } 2713b64ec075STomohiro Misono } 2714b64ec075STomohiro Misono 2715b64ec075STomohiro Misono leaf = path->nodes[0]; 2716b64ec075STomohiro Misono slot = path->slots[0]; 2717b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2718b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2719b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2720b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2721b64ec075STomohiro Misono 2722b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2723b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2724b64ec075STomohiro Misono 2725b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2726b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2727b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2728b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2729b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2730b64ec075STomohiro Misono item_off, item_len); 2731b64ec075STomohiro Misono } else { 2732b64ec075STomohiro Misono ret = -ENOENT; 2733b64ec075STomohiro Misono goto out; 2734b64ec075STomohiro Misono } 2735b64ec075STomohiro Misono } 2736b64ec075STomohiro Misono 2737b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2738b64ec075STomohiro Misono ret = -EFAULT; 2739b64ec075STomohiro Misono 2740b64ec075STomohiro Misono out: 274100246528SJosef Bacik btrfs_put_root(root); 274204734e84SJosef Bacik out_free: 2743b64ec075STomohiro Misono btrfs_free_path(path); 2744b091f7feSWaiman Long kfree(subvol_info); 2745b64ec075STomohiro Misono return ret; 2746b64ec075STomohiro Misono } 2747b64ec075STomohiro Misono 274842e4b520STomohiro Misono /* 274942e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 275042e4b520STomohiro Misono * except the subvolume name. 275142e4b520STomohiro Misono */ 275242e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 275342e4b520STomohiro Misono { 275442e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 275542e4b520STomohiro Misono struct btrfs_root_ref *rref; 275642e4b520STomohiro Misono struct btrfs_root *root; 275742e4b520STomohiro Misono struct btrfs_path *path; 275842e4b520STomohiro Misono struct btrfs_key key; 275942e4b520STomohiro Misono struct extent_buffer *leaf; 276042e4b520STomohiro Misono struct inode *inode; 276142e4b520STomohiro Misono u64 objectid; 276242e4b520STomohiro Misono int slot; 276342e4b520STomohiro Misono int ret; 276442e4b520STomohiro Misono u8 found; 276542e4b520STomohiro Misono 276642e4b520STomohiro Misono path = btrfs_alloc_path(); 276742e4b520STomohiro Misono if (!path) 276842e4b520STomohiro Misono return -ENOMEM; 276942e4b520STomohiro Misono 277042e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 277142e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 277242e4b520STomohiro Misono btrfs_free_path(path); 277342e4b520STomohiro Misono return PTR_ERR(rootrefs); 277442e4b520STomohiro Misono } 277542e4b520STomohiro Misono 277642e4b520STomohiro Misono inode = file_inode(file); 277742e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 277842e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 277942e4b520STomohiro Misono 278042e4b520STomohiro Misono key.objectid = objectid; 278142e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 278242e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 278342e4b520STomohiro Misono found = 0; 278442e4b520STomohiro Misono 278542e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 278642e4b520STomohiro Misono if (ret < 0) { 278742e4b520STomohiro Misono goto out; 278842e4b520STomohiro Misono } else if (path->slots[0] >= 278942e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 279042e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 279142e4b520STomohiro Misono if (ret < 0) { 279242e4b520STomohiro Misono goto out; 279342e4b520STomohiro Misono } else if (ret > 0) { 279442e4b520STomohiro Misono ret = -EUCLEAN; 279542e4b520STomohiro Misono goto out; 279642e4b520STomohiro Misono } 279742e4b520STomohiro Misono } 279842e4b520STomohiro Misono while (1) { 279942e4b520STomohiro Misono leaf = path->nodes[0]; 280042e4b520STomohiro Misono slot = path->slots[0]; 280142e4b520STomohiro Misono 280242e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 280342e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 280442e4b520STomohiro Misono ret = 0; 280542e4b520STomohiro Misono goto out; 280642e4b520STomohiro Misono } 280742e4b520STomohiro Misono 280842e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 280942e4b520STomohiro Misono ret = -EOVERFLOW; 281042e4b520STomohiro Misono goto out; 281142e4b520STomohiro Misono } 281242e4b520STomohiro Misono 281342e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 281442e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 281542e4b520STomohiro Misono rootrefs->rootref[found].dirid = 281642e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 281742e4b520STomohiro Misono found++; 281842e4b520STomohiro Misono 281942e4b520STomohiro Misono ret = btrfs_next_item(root, path); 282042e4b520STomohiro Misono if (ret < 0) { 282142e4b520STomohiro Misono goto out; 282242e4b520STomohiro Misono } else if (ret > 0) { 282342e4b520STomohiro Misono ret = -EUCLEAN; 282442e4b520STomohiro Misono goto out; 282542e4b520STomohiro Misono } 282642e4b520STomohiro Misono } 282742e4b520STomohiro Misono 282842e4b520STomohiro Misono out: 282942e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 283042e4b520STomohiro Misono rootrefs->num_items = found; 283142e4b520STomohiro Misono /* update min_treeid for next search */ 283242e4b520STomohiro Misono if (found) 283342e4b520STomohiro Misono rootrefs->min_treeid = 283442e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 283542e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 283642e4b520STomohiro Misono ret = -EFAULT; 283742e4b520STomohiro Misono } 283842e4b520STomohiro Misono 283942e4b520STomohiro Misono kfree(rootrefs); 284042e4b520STomohiro Misono btrfs_free_path(path); 284142e4b520STomohiro Misono 284242e4b520STomohiro Misono return ret; 284342e4b520STomohiro Misono } 284442e4b520STomohiro Misono 284576dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 2846949964c9SMarcos Paulo de Souza void __user *arg, 2847949964c9SMarcos Paulo de Souza bool destroy_v2) 284876dda93cSYan, Zheng { 284954563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28500b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 285176dda93cSYan, Zheng struct dentry *dentry; 28522b0143b5SDavid Howells struct inode *dir = d_inode(parent); 285376dda93cSYan, Zheng struct inode *inode; 285476dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 285576dda93cSYan, Zheng struct btrfs_root *dest = NULL; 2856949964c9SMarcos Paulo de Souza struct btrfs_ioctl_vol_args *vol_args = NULL; 2857949964c9SMarcos Paulo de Souza struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL; 2858c4ed533bSChristian Brauner struct user_namespace *mnt_userns = file_mnt_user_ns(file); 2859949964c9SMarcos Paulo de Souza char *subvol_name, *subvol_name_ptr = NULL; 2860949964c9SMarcos Paulo de Souza int subvol_namelen; 286176dda93cSYan, Zheng int err = 0; 2862949964c9SMarcos Paulo de Souza bool destroy_parent = false; 286376dda93cSYan, Zheng 2864949964c9SMarcos Paulo de Souza if (destroy_v2) { 2865949964c9SMarcos Paulo de Souza vol_args2 = memdup_user(arg, sizeof(*vol_args2)); 2866949964c9SMarcos Paulo de Souza if (IS_ERR(vol_args2)) 2867949964c9SMarcos Paulo de Souza return PTR_ERR(vol_args2); 2868325c50e3SJeff Mahoney 2869949964c9SMarcos Paulo de Souza if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) { 2870949964c9SMarcos Paulo de Souza err = -EOPNOTSUPP; 2871949964c9SMarcos Paulo de Souza goto out; 2872949964c9SMarcos Paulo de Souza } 287376dda93cSYan, Zheng 2874949964c9SMarcos Paulo de Souza /* 2875949964c9SMarcos Paulo de Souza * If SPEC_BY_ID is not set, we are looking for the subvolume by 2876949964c9SMarcos Paulo de Souza * name, same as v1 currently does. 2877949964c9SMarcos Paulo de Souza */ 2878949964c9SMarcos Paulo de Souza if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { 2879949964c9SMarcos Paulo de Souza vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0; 2880949964c9SMarcos Paulo de Souza subvol_name = vol_args2->name; 2881949964c9SMarcos Paulo de Souza 2882949964c9SMarcos Paulo de Souza err = mnt_want_write_file(file); 2883949964c9SMarcos Paulo de Souza if (err) 2884949964c9SMarcos Paulo de Souza goto out; 2885949964c9SMarcos Paulo de Souza } else { 2886aabb34e7SChristian Brauner struct inode *old_dir; 2887c4ed533bSChristian Brauner 2888949964c9SMarcos Paulo de Souza if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) { 288976dda93cSYan, Zheng err = -EINVAL; 289076dda93cSYan, Zheng goto out; 289176dda93cSYan, Zheng } 289276dda93cSYan, Zheng 2893a561be71SAl Viro err = mnt_want_write_file(file); 289476dda93cSYan, Zheng if (err) 289576dda93cSYan, Zheng goto out; 289676dda93cSYan, Zheng 2897949964c9SMarcos Paulo de Souza dentry = btrfs_get_dentry(fs_info->sb, 2898949964c9SMarcos Paulo de Souza BTRFS_FIRST_FREE_OBJECTID, 2899949964c9SMarcos Paulo de Souza vol_args2->subvolid, 0, 0); 2900949964c9SMarcos Paulo de Souza if (IS_ERR(dentry)) { 2901949964c9SMarcos Paulo de Souza err = PTR_ERR(dentry); 2902949964c9SMarcos Paulo de Souza goto out_drop_write; 2903949964c9SMarcos Paulo de Souza } 2904949964c9SMarcos Paulo de Souza 2905949964c9SMarcos Paulo de Souza /* 2906949964c9SMarcos Paulo de Souza * Change the default parent since the subvolume being 2907949964c9SMarcos Paulo de Souza * deleted can be outside of the current mount point. 2908949964c9SMarcos Paulo de Souza */ 2909949964c9SMarcos Paulo de Souza parent = btrfs_get_parent(dentry); 2910949964c9SMarcos Paulo de Souza 2911949964c9SMarcos Paulo de Souza /* 2912949964c9SMarcos Paulo de Souza * At this point dentry->d_name can point to '/' if the 2913949964c9SMarcos Paulo de Souza * subvolume we want to destroy is outsite of the 2914949964c9SMarcos Paulo de Souza * current mount point, so we need to release the 2915949964c9SMarcos Paulo de Souza * current dentry and execute the lookup to return a new 2916949964c9SMarcos Paulo de Souza * one with ->d_name pointing to the 2917949964c9SMarcos Paulo de Souza * <mount point>/subvol_name. 2918949964c9SMarcos Paulo de Souza */ 2919949964c9SMarcos Paulo de Souza dput(dentry); 2920949964c9SMarcos Paulo de Souza if (IS_ERR(parent)) { 2921949964c9SMarcos Paulo de Souza err = PTR_ERR(parent); 2922949964c9SMarcos Paulo de Souza goto out_drop_write; 2923949964c9SMarcos Paulo de Souza } 2924aabb34e7SChristian Brauner old_dir = dir; 2925949964c9SMarcos Paulo de Souza dir = d_inode(parent); 2926949964c9SMarcos Paulo de Souza 2927949964c9SMarcos Paulo de Souza /* 2928949964c9SMarcos Paulo de Souza * If v2 was used with SPEC_BY_ID, a new parent was 2929949964c9SMarcos Paulo de Souza * allocated since the subvolume can be outside of the 2930949964c9SMarcos Paulo de Souza * current mount point. Later on we need to release this 2931949964c9SMarcos Paulo de Souza * new parent dentry. 2932949964c9SMarcos Paulo de Souza */ 2933949964c9SMarcos Paulo de Souza destroy_parent = true; 2934949964c9SMarcos Paulo de Souza 2935aabb34e7SChristian Brauner /* 2936aabb34e7SChristian Brauner * On idmapped mounts, deletion via subvolid is 2937aabb34e7SChristian Brauner * restricted to subvolumes that are immediate 2938aabb34e7SChristian Brauner * ancestors of the inode referenced by the file 2939aabb34e7SChristian Brauner * descriptor in the ioctl. Otherwise the idmapping 2940aabb34e7SChristian Brauner * could potentially be abused to delete subvolumes 2941aabb34e7SChristian Brauner * anywhere in the filesystem the user wouldn't be able 2942aabb34e7SChristian Brauner * to delete without an idmapped mount. 2943aabb34e7SChristian Brauner */ 2944aabb34e7SChristian Brauner if (old_dir != dir && mnt_userns != &init_user_ns) { 2945aabb34e7SChristian Brauner err = -EOPNOTSUPP; 2946aabb34e7SChristian Brauner goto free_parent; 2947aabb34e7SChristian Brauner } 2948aabb34e7SChristian Brauner 2949949964c9SMarcos Paulo de Souza subvol_name_ptr = btrfs_get_subvol_name_from_objectid( 2950949964c9SMarcos Paulo de Souza fs_info, vol_args2->subvolid); 2951949964c9SMarcos Paulo de Souza if (IS_ERR(subvol_name_ptr)) { 2952949964c9SMarcos Paulo de Souza err = PTR_ERR(subvol_name_ptr); 2953949964c9SMarcos Paulo de Souza goto free_parent; 2954949964c9SMarcos Paulo de Souza } 29551a9fd417SDavid Sterba /* subvol_name_ptr is already nul terminated */ 2956949964c9SMarcos Paulo de Souza subvol_name = (char *)kbasename(subvol_name_ptr); 2957949964c9SMarcos Paulo de Souza } 2958949964c9SMarcos Paulo de Souza } else { 2959949964c9SMarcos Paulo de Souza vol_args = memdup_user(arg, sizeof(*vol_args)); 2960949964c9SMarcos Paulo de Souza if (IS_ERR(vol_args)) 2961949964c9SMarcos Paulo de Souza return PTR_ERR(vol_args); 2962949964c9SMarcos Paulo de Souza 2963949964c9SMarcos Paulo de Souza vol_args->name[BTRFS_PATH_NAME_MAX] = 0; 2964949964c9SMarcos Paulo de Souza subvol_name = vol_args->name; 2965949964c9SMarcos Paulo de Souza 2966949964c9SMarcos Paulo de Souza err = mnt_want_write_file(file); 2967949964c9SMarcos Paulo de Souza if (err) 2968949964c9SMarcos Paulo de Souza goto out; 2969949964c9SMarcos Paulo de Souza } 2970949964c9SMarcos Paulo de Souza 2971949964c9SMarcos Paulo de Souza subvol_namelen = strlen(subvol_name); 2972949964c9SMarcos Paulo de Souza 2973949964c9SMarcos Paulo de Souza if (strchr(subvol_name, '/') || 2974949964c9SMarcos Paulo de Souza strncmp(subvol_name, "..", subvol_namelen) == 0) { 2975949964c9SMarcos Paulo de Souza err = -EINVAL; 2976949964c9SMarcos Paulo de Souza goto free_subvol_name; 2977949964c9SMarcos Paulo de Souza } 2978949964c9SMarcos Paulo de Souza 2979949964c9SMarcos Paulo de Souza if (!S_ISDIR(dir->i_mode)) { 2980949964c9SMarcos Paulo de Souza err = -ENOTDIR; 2981949964c9SMarcos Paulo de Souza goto free_subvol_name; 2982949964c9SMarcos Paulo de Souza } 2983521e0546SDavid Sterba 298400235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 298500235411SAl Viro if (err == -EINTR) 2986949964c9SMarcos Paulo de Souza goto free_subvol_name; 2987c4ed533bSChristian Brauner dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen); 298876dda93cSYan, Zheng if (IS_ERR(dentry)) { 298976dda93cSYan, Zheng err = PTR_ERR(dentry); 299076dda93cSYan, Zheng goto out_unlock_dir; 299176dda93cSYan, Zheng } 299276dda93cSYan, Zheng 29932b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 299476dda93cSYan, Zheng err = -ENOENT; 299576dda93cSYan, Zheng goto out_dput; 299676dda93cSYan, Zheng } 299776dda93cSYan, Zheng 29982b0143b5SDavid Howells inode = d_inode(dentry); 29994260f7c7SSage Weil dest = BTRFS_I(inode)->root; 30004260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 30014260f7c7SSage Weil /* 30024260f7c7SSage Weil * Regular user. Only allow this with a special mount 30034260f7c7SSage Weil * option, when the user has write+exec access to the 30044260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 30054260f7c7SSage Weil * allowed. 30064260f7c7SSage Weil * 30074260f7c7SSage Weil * Note that this is _not_ check that the subvol is 30084260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 30094260f7c7SSage Weil * otherwise be able to delete. 30104260f7c7SSage Weil * 30114260f7c7SSage Weil * Users who want to delete empty subvols should try 30124260f7c7SSage Weil * rmdir(2). 30134260f7c7SSage Weil */ 30144260f7c7SSage Weil err = -EPERM; 30150b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 30164260f7c7SSage Weil goto out_dput; 30174260f7c7SSage Weil 30184260f7c7SSage Weil /* 30194260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 30204260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 30214260f7c7SSage Weil * must be called on the dentry referencing the root 30224260f7c7SSage Weil * of the subvol, not a random directory contained 30234260f7c7SSage Weil * within it. 30244260f7c7SSage Weil */ 30254260f7c7SSage Weil err = -EINVAL; 30264260f7c7SSage Weil if (root == dest) 30274260f7c7SSage Weil goto out_dput; 30284260f7c7SSage Weil 3029c4ed533bSChristian Brauner err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC); 30304260f7c7SSage Weil if (err) 30314260f7c7SSage Weil goto out_dput; 30325c39da5bSMiao Xie } 30334260f7c7SSage Weil 30345c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 3035c4ed533bSChristian Brauner err = btrfs_may_delete(mnt_userns, dir, dentry, 1); 30364260f7c7SSage Weil if (err) 30374260f7c7SSage Weil goto out_dput; 30384260f7c7SSage Weil 30394a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 304076dda93cSYan, Zheng err = -EINVAL; 304176dda93cSYan, Zheng goto out_dput; 304276dda93cSYan, Zheng } 304376dda93cSYan, Zheng 304464708539SJosef Bacik btrfs_inode_lock(inode, 0); 3045f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 304664708539SJosef Bacik btrfs_inode_unlock(inode, 0); 304746008d9dSAmir Goldstein if (!err) { 304846008d9dSAmir Goldstein fsnotify_rmdir(dir, dentry); 304976dda93cSYan, Zheng d_delete(dentry); 305046008d9dSAmir Goldstein } 3051fa6ac876SLiu Bo 305276dda93cSYan, Zheng out_dput: 305376dda93cSYan, Zheng dput(dentry); 305476dda93cSYan, Zheng out_unlock_dir: 305564708539SJosef Bacik btrfs_inode_unlock(dir, 0); 3056949964c9SMarcos Paulo de Souza free_subvol_name: 3057949964c9SMarcos Paulo de Souza kfree(subvol_name_ptr); 3058949964c9SMarcos Paulo de Souza free_parent: 3059949964c9SMarcos Paulo de Souza if (destroy_parent) 3060949964c9SMarcos Paulo de Souza dput(parent); 306100235411SAl Viro out_drop_write: 30622a79f17eSAl Viro mnt_drop_write_file(file); 306376dda93cSYan, Zheng out: 3064949964c9SMarcos Paulo de Souza kfree(vol_args2); 306576dda93cSYan, Zheng kfree(vol_args); 306676dda93cSYan, Zheng return err; 306776dda93cSYan, Zheng } 306876dda93cSYan, Zheng 30691e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 3070f46b5a66SChristoph Hellwig { 3071496ad9aaSAl Viro struct inode *inode = file_inode(file); 3072f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3073c853a578SGoldwyn Rodrigues struct btrfs_ioctl_defrag_range_args range = {0}; 3074c146afadSYan Zheng int ret; 3075c146afadSYan Zheng 307625122d15SIlya Dryomov ret = mnt_want_write_file(file); 307725122d15SIlya Dryomov if (ret) 307825122d15SIlya Dryomov return ret; 3079b83cc969SLi Zefan 308025122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 308125122d15SIlya Dryomov ret = -EROFS; 308225122d15SIlya Dryomov goto out; 30835ac00addSStefan Behrens } 3084f46b5a66SChristoph Hellwig 3085f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 3086f46b5a66SChristoph Hellwig case S_IFDIR: 3087e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 3088e441d54dSChris Mason ret = -EPERM; 3089e441d54dSChris Mason goto out; 3090e441d54dSChris Mason } 3091de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 3092f46b5a66SChristoph Hellwig break; 3093f46b5a66SChristoph Hellwig case S_IFREG: 3094616d374eSAdam Borowski /* 3095616d374eSAdam Borowski * Note that this does not check the file descriptor for write 3096616d374eSAdam Borowski * access. This prevents defragmenting executables that are 3097616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 3098616d374eSAdam Borowski */ 3099616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 310047291baaSChristian Brauner inode_permission(&init_user_ns, inode, MAY_WRITE)) { 3101616d374eSAdam Borowski ret = -EPERM; 3102e441d54dSChris Mason goto out; 3103e441d54dSChris Mason } 31041e701a32SChris Mason 31051e701a32SChris Mason if (argp) { 3106c853a578SGoldwyn Rodrigues if (copy_from_user(&range, argp, sizeof(range))) { 31071e701a32SChris Mason ret = -EFAULT; 3108683be16eSDan Carpenter goto out; 31091e701a32SChris Mason } 31101e701a32SChris Mason /* compression requires us to start the IO */ 3111c853a578SGoldwyn Rodrigues if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 3112c853a578SGoldwyn Rodrigues range.flags |= BTRFS_DEFRAG_RANGE_START_IO; 3113c853a578SGoldwyn Rodrigues range.extent_thresh = (u32)-1; 31141e701a32SChris Mason } 31151e701a32SChris Mason } else { 31161e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 3117c853a578SGoldwyn Rodrigues range.len = (u64)-1; 31181e701a32SChris Mason } 31191ccc2e8aSQu Wenruo ret = btrfs_defrag_file(file_inode(file), &file->f_ra, 3120c853a578SGoldwyn Rodrigues &range, BTRFS_OLDEST_GENERATION, 0); 31214cb5300bSChris Mason if (ret > 0) 31224cb5300bSChris Mason ret = 0; 3123f46b5a66SChristoph Hellwig break; 31248929ecfaSYan, Zheng default: 31258929ecfaSYan, Zheng ret = -EINVAL; 3126f46b5a66SChristoph Hellwig } 3127e441d54dSChris Mason out: 312825122d15SIlya Dryomov mnt_drop_write_file(file); 3129e441d54dSChris Mason return ret; 3130f46b5a66SChristoph Hellwig } 3131f46b5a66SChristoph Hellwig 31322ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3133f46b5a66SChristoph Hellwig { 3134f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3135f46b5a66SChristoph Hellwig int ret; 3136f46b5a66SChristoph Hellwig 3137e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3138e441d54dSChris Mason return -EPERM; 3139e441d54dSChris Mason 3140c3e1f96cSGoldwyn Rodrigues if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) 3141e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3142c9e9f97bSIlya Dryomov 3143dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3144c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3145c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3146c9e9f97bSIlya Dryomov goto out; 3147c9e9f97bSIlya Dryomov } 3148f46b5a66SChristoph Hellwig 31495516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31502ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3151f46b5a66SChristoph Hellwig 315243d20761SAnand Jain if (!ret) 31530b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 315443d20761SAnand Jain 3155f46b5a66SChristoph Hellwig kfree(vol_args); 3156c9e9f97bSIlya Dryomov out: 3157c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 3158f46b5a66SChristoph Hellwig return ret; 3159f46b5a66SChristoph Hellwig } 3160f46b5a66SChristoph Hellwig 31616b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3162f46b5a66SChristoph Hellwig { 3163*1a15eb72SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 31640b246afaSJeff Mahoney struct inode *inode = file_inode(file); 31650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 31666b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 31673fa421deSJosef Bacik struct block_device *bdev = NULL; 31683fa421deSJosef Bacik fmode_t mode; 3169f46b5a66SChristoph Hellwig int ret; 317067ae34b6SDavid Sterba bool cancel = false; 3171f46b5a66SChristoph Hellwig 3172e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3173e441d54dSChris Mason return -EPERM; 3174e441d54dSChris Mason 3175dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3176c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3177c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3178*1a15eb72SJosef Bacik goto out; 3179c9e9f97bSIlya Dryomov } 3180f46b5a66SChristoph Hellwig 3181748449cdSDavid Sterba if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { 3182fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3183fd4e994bSOmar Sandoval goto out; 3184fd4e994bSOmar Sandoval } 3185*1a15eb72SJosef Bacik 31866b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 3187*1a15eb72SJosef Bacik if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 3188*1a15eb72SJosef Bacik args.devid = vol_args->devid; 3189*1a15eb72SJosef Bacik } else if (!strcmp("cancel", vol_args->name)) { 319067ae34b6SDavid Sterba cancel = true; 3191*1a15eb72SJosef Bacik } else { 3192*1a15eb72SJosef Bacik ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 3193*1a15eb72SJosef Bacik if (ret) 3194*1a15eb72SJosef Bacik goto out; 3195*1a15eb72SJosef Bacik } 3196*1a15eb72SJosef Bacik 3197*1a15eb72SJosef Bacik ret = mnt_want_write_file(file); 3198*1a15eb72SJosef Bacik if (ret) 3199*1a15eb72SJosef Bacik goto out; 320067ae34b6SDavid Sterba 320167ae34b6SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 320267ae34b6SDavid Sterba cancel); 320367ae34b6SDavid Sterba if (ret) 3204*1a15eb72SJosef Bacik goto err_drop; 320567ae34b6SDavid Sterba 3206*1a15eb72SJosef Bacik /* Exclusive operation is now claimed */ 3207*1a15eb72SJosef Bacik ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); 320867ae34b6SDavid Sterba 3209c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 3210183860f6SAnand Jain 32116b526ed7SAnand Jain if (!ret) { 3212735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 32130b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 32146b526ed7SAnand Jain vol_args->devid); 32156b526ed7SAnand Jain else 32160b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 32176b526ed7SAnand Jain vol_args->name); 32186b526ed7SAnand Jain } 3219c47ca32dSDan Carpenter err_drop: 32204ac20c70SIlya Dryomov mnt_drop_write_file(file); 32213fa421deSJosef Bacik if (bdev) 32223fa421deSJosef Bacik blkdev_put(bdev, mode); 3223*1a15eb72SJosef Bacik out: 3224*1a15eb72SJosef Bacik btrfs_put_dev_args_from_path(&args); 3225*1a15eb72SJosef Bacik kfree(vol_args); 3226f46b5a66SChristoph Hellwig return ret; 3227f46b5a66SChristoph Hellwig } 3228f46b5a66SChristoph Hellwig 3229f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3230f46b5a66SChristoph Hellwig { 3231*1a15eb72SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 32320b246afaSJeff Mahoney struct inode *inode = file_inode(file); 32330b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3234f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 32353fa421deSJosef Bacik struct block_device *bdev = NULL; 32363fa421deSJosef Bacik fmode_t mode; 3237f46b5a66SChristoph Hellwig int ret; 323867ae34b6SDavid Sterba bool cancel; 3239f46b5a66SChristoph Hellwig 3240f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3241f46b5a66SChristoph Hellwig return -EPERM; 3242f46b5a66SChristoph Hellwig 3243*1a15eb72SJosef Bacik vol_args = memdup_user(arg, sizeof(*vol_args)); 3244*1a15eb72SJosef Bacik if (IS_ERR(vol_args)) 3245*1a15eb72SJosef Bacik return PTR_ERR(vol_args); 3246*1a15eb72SJosef Bacik 3247*1a15eb72SJosef Bacik vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 3248*1a15eb72SJosef Bacik if (!strcmp("cancel", vol_args->name)) { 3249*1a15eb72SJosef Bacik cancel = true; 3250*1a15eb72SJosef Bacik } else { 3251*1a15eb72SJosef Bacik ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 3252*1a15eb72SJosef Bacik if (ret) 3253*1a15eb72SJosef Bacik goto out; 3254*1a15eb72SJosef Bacik } 3255*1a15eb72SJosef Bacik 3256f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3257f46b5a66SChristoph Hellwig if (ret) 3258*1a15eb72SJosef Bacik goto out; 3259f46b5a66SChristoph Hellwig 326067ae34b6SDavid Sterba ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 326167ae34b6SDavid Sterba cancel); 326267ae34b6SDavid Sterba if (ret == 0) { 3263*1a15eb72SJosef Bacik ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); 3264f46b5a66SChristoph Hellwig if (!ret) 32650b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3266c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 326767ae34b6SDavid Sterba } 326867ae34b6SDavid Sterba 3269f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 32703fa421deSJosef Bacik if (bdev) 32713fa421deSJosef Bacik blkdev_put(bdev, mode); 3272*1a15eb72SJosef Bacik out: 3273*1a15eb72SJosef Bacik btrfs_put_dev_args_from_path(&args); 3274*1a15eb72SJosef Bacik kfree(vol_args); 3275f46b5a66SChristoph Hellwig return ret; 3276f46b5a66SChristoph Hellwig } 3277f46b5a66SChristoph Hellwig 32782ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 32792ff7e61eSJeff Mahoney void __user *arg) 3280475f6387SJan Schmidt { 3281027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3282475f6387SJan Schmidt struct btrfs_device *device; 32830b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3284137c5418SJohannes Thumshirn u64 flags_in; 3285027ed2f0SLi Zefan int ret = 0; 3286475f6387SJan Schmidt 3287137c5418SJohannes Thumshirn fi_args = memdup_user(arg, sizeof(*fi_args)); 3288137c5418SJohannes Thumshirn if (IS_ERR(fi_args)) 3289137c5418SJohannes Thumshirn return PTR_ERR(fi_args); 3290137c5418SJohannes Thumshirn 3291137c5418SJohannes Thumshirn flags_in = fi_args->flags; 3292137c5418SJohannes Thumshirn memset(fi_args, 0, sizeof(*fi_args)); 3293027ed2f0SLi Zefan 3294d03262c7SDavid Sterba rcu_read_lock(); 3295027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3296475f6387SJan Schmidt 3297d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3298027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3299027ed2f0SLi Zefan fi_args->max_id = device->devid; 3300475f6387SJan Schmidt } 3301d03262c7SDavid Sterba rcu_read_unlock(); 3302475f6387SJan Schmidt 3303de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3304bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3305bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3306bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 330780a773fbSDavid Sterba 3308137c5418SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) { 3309137c5418SJohannes Thumshirn fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy); 3310137c5418SJohannes Thumshirn fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy); 3311137c5418SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO; 3312137c5418SJohannes Thumshirn } 3313137c5418SJohannes Thumshirn 33140fb408a5SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) { 33150fb408a5SJohannes Thumshirn fi_args->generation = fs_info->generation; 33160fb408a5SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION; 33170fb408a5SJohannes Thumshirn } 33180fb408a5SJohannes Thumshirn 331949bac897SJohannes Thumshirn if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) { 332049bac897SJohannes Thumshirn memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid, 332149bac897SJohannes Thumshirn sizeof(fi_args->metadata_uuid)); 332249bac897SJohannes Thumshirn fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID; 332349bac897SJohannes Thumshirn } 332449bac897SJohannes Thumshirn 3325027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3326027ed2f0SLi Zefan ret = -EFAULT; 3327475f6387SJan Schmidt 3328027ed2f0SLi Zefan kfree(fi_args); 3329027ed2f0SLi Zefan return ret; 3330475f6387SJan Schmidt } 3331475f6387SJan Schmidt 33322ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 33332ff7e61eSJeff Mahoney void __user *arg) 3334475f6387SJan Schmidt { 3335562d7b15SJosef Bacik BTRFS_DEV_LOOKUP_ARGS(args); 3336475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3337475f6387SJan Schmidt struct btrfs_device *dev; 3338475f6387SJan Schmidt int ret = 0; 3339475f6387SJan Schmidt 3340475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3341475f6387SJan Schmidt if (IS_ERR(di_args)) 3342475f6387SJan Schmidt return PTR_ERR(di_args); 3343475f6387SJan Schmidt 3344562d7b15SJosef Bacik args.devid = di_args->devid; 3345dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3346562d7b15SJosef Bacik args.uuid = di_args->uuid; 3347475f6387SJan Schmidt 3348c5593ca3SDavid Sterba rcu_read_lock(); 3349562d7b15SJosef Bacik dev = btrfs_find_device(fs_info->fs_devices, &args); 3350475f6387SJan Schmidt if (!dev) { 3351475f6387SJan Schmidt ret = -ENODEV; 3352475f6387SJan Schmidt goto out; 3353475f6387SJan Schmidt } 3354475f6387SJan Schmidt 3355475f6387SJan Schmidt di_args->devid = dev->devid; 33567cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 33577cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3358475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3359a27202fbSJim Meyering if (dev->name) { 3360672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3361672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3362a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3363a27202fbSJim Meyering } else { 336499ba55adSStefan Behrens di_args->path[0] = '\0'; 3365a27202fbSJim Meyering } 3366475f6387SJan Schmidt 3367475f6387SJan Schmidt out: 3368c5593ca3SDavid Sterba rcu_read_unlock(); 3369475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3370475f6387SJan Schmidt ret = -EFAULT; 3371475f6387SJan Schmidt 3372475f6387SJan Schmidt kfree(di_args); 3373475f6387SJan Schmidt return ret; 3374475f6387SJan Schmidt } 3375475f6387SJan Schmidt 33766ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 33776ef5ed0dSJosef Bacik { 3378496ad9aaSAl Viro struct inode *inode = file_inode(file); 33790b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 33806ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 33816ef5ed0dSJosef Bacik struct btrfs_root *new_root; 33826ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 33836ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 33842a2b5d62SJosef Bacik struct btrfs_path *path = NULL; 33856ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 33866ef5ed0dSJosef Bacik u64 objectid = 0; 33876ef5ed0dSJosef Bacik u64 dir_id; 33883c04ce01SMiao Xie int ret; 33896ef5ed0dSJosef Bacik 33906ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 33916ef5ed0dSJosef Bacik return -EPERM; 33926ef5ed0dSJosef Bacik 33933c04ce01SMiao Xie ret = mnt_want_write_file(file); 33943c04ce01SMiao Xie if (ret) 33953c04ce01SMiao Xie return ret; 33963c04ce01SMiao Xie 33973c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 33983c04ce01SMiao Xie ret = -EFAULT; 33993c04ce01SMiao Xie goto out; 34003c04ce01SMiao Xie } 34016ef5ed0dSJosef Bacik 34026ef5ed0dSJosef Bacik if (!objectid) 34031cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 34046ef5ed0dSJosef Bacik 340556e9357aSDavid Sterba new_root = btrfs_get_fs_root(fs_info, objectid, true); 34063c04ce01SMiao Xie if (IS_ERR(new_root)) { 34073c04ce01SMiao Xie ret = PTR_ERR(new_root); 34083c04ce01SMiao Xie goto out; 34093c04ce01SMiao Xie } 34102a2b5d62SJosef Bacik if (!is_fstree(new_root->root_key.objectid)) { 34112a2b5d62SJosef Bacik ret = -ENOENT; 34122a2b5d62SJosef Bacik goto out_free; 34132a2b5d62SJosef Bacik } 34146ef5ed0dSJosef Bacik 34156ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 34163c04ce01SMiao Xie if (!path) { 34173c04ce01SMiao Xie ret = -ENOMEM; 34182a2b5d62SJosef Bacik goto out_free; 34193c04ce01SMiao Xie } 34206ef5ed0dSJosef Bacik 34216ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 342298d5dc13STsutomu Itoh if (IS_ERR(trans)) { 34233c04ce01SMiao Xie ret = PTR_ERR(trans); 34242a2b5d62SJosef Bacik goto out_free; 34256ef5ed0dSJosef Bacik } 34266ef5ed0dSJosef Bacik 34270b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 34280b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 34296ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 3430cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 34312a2b5d62SJosef Bacik btrfs_release_path(path); 34323a45bb20SJeff Mahoney btrfs_end_transaction(trans); 34330b246afaSJeff Mahoney btrfs_err(fs_info, 34345d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 34353c04ce01SMiao Xie ret = -ENOENT; 34362a2b5d62SJosef Bacik goto out_free; 34376ef5ed0dSJosef Bacik } 34386ef5ed0dSJosef Bacik 34396ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 34406ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 34416ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 34422a2b5d62SJosef Bacik btrfs_release_path(path); 34436ef5ed0dSJosef Bacik 34440b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 34453a45bb20SJeff Mahoney btrfs_end_transaction(trans); 34462a2b5d62SJosef Bacik out_free: 344700246528SJosef Bacik btrfs_put_root(new_root); 34482a2b5d62SJosef Bacik btrfs_free_path(path); 34493c04ce01SMiao Xie out: 34503c04ce01SMiao Xie mnt_drop_write_file(file); 34513c04ce01SMiao Xie return ret; 34526ef5ed0dSJosef Bacik } 34536ef5ed0dSJosef Bacik 3454c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 3455bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 3456bf5fc093SJosef Bacik { 345732da5386SDavid Sterba struct btrfs_block_group *block_group; 3458bf5fc093SJosef Bacik 3459bf5fc093SJosef Bacik space->total_bytes = 0; 3460bf5fc093SJosef Bacik space->used_bytes = 0; 3461bf5fc093SJosef Bacik space->flags = 0; 3462bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 3463bf5fc093SJosef Bacik space->flags = block_group->flags; 3464b3470b5dSDavid Sterba space->total_bytes += block_group->length; 3465bf38be65SDavid Sterba space->used_bytes += block_group->used; 3466bf5fc093SJosef Bacik } 3467bf5fc093SJosef Bacik } 3468bf5fc093SJosef Bacik 34692ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 34702ff7e61eSJeff Mahoney void __user *arg) 34711406e432SJosef Bacik { 34721406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 34731406e432SJosef Bacik struct btrfs_ioctl_space_info space; 34741406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 34757fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 347613f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 34771406e432SJosef Bacik struct btrfs_space_info *info; 3478315d8e98SColin Ian King static const u64 types[] = { 3479315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 3480bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 3481bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 3482315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 3483315d8e98SColin Ian King }; 3484bf5fc093SJosef Bacik int num_types = 4; 34857fde62bfSChris Mason int alloc_size; 34861406e432SJosef Bacik int ret = 0; 348751788b1bSDan Rosenberg u64 slot_count = 0; 3488bf5fc093SJosef Bacik int i, c; 34891406e432SJosef Bacik 34901406e432SJosef Bacik if (copy_from_user(&space_args, 34911406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 34921406e432SJosef Bacik sizeof(space_args))) 34931406e432SJosef Bacik return -EFAULT; 34941406e432SJosef Bacik 3495bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3496bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3497bf5fc093SJosef Bacik 3498bf5fc093SJosef Bacik info = NULL; 349972804905SJosef Bacik list_for_each_entry(tmp, &fs_info->space_info, list) { 3500bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3501bf5fc093SJosef Bacik info = tmp; 3502bf5fc093SJosef Bacik break; 3503bf5fc093SJosef Bacik } 3504bf5fc093SJosef Bacik } 35051406e432SJosef Bacik 3506bf5fc093SJosef Bacik if (!info) 3507bf5fc093SJosef Bacik continue; 3508bf5fc093SJosef Bacik 3509bf5fc093SJosef Bacik down_read(&info->groups_sem); 3510bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3511bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 3512bf5fc093SJosef Bacik slot_count++; 3513bf5fc093SJosef Bacik } 3514bf5fc093SJosef Bacik up_read(&info->groups_sem); 3515bf5fc093SJosef Bacik } 3516bf5fc093SJosef Bacik 351736523e95SDavid Sterba /* 351836523e95SDavid Sterba * Global block reserve, exported as a space_info 351936523e95SDavid Sterba */ 352036523e95SDavid Sterba slot_count++; 352136523e95SDavid Sterba 35227fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 35237fde62bfSChris Mason if (space_args.space_slots == 0) { 35247fde62bfSChris Mason space_args.total_spaces = slot_count; 35257fde62bfSChris Mason goto out; 35267fde62bfSChris Mason } 3527bf5fc093SJosef Bacik 352851788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 3529bf5fc093SJosef Bacik 35307fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 3531bf5fc093SJosef Bacik 35327fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 35337fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 35347fde62bfSChris Mason */ 353509cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 35367fde62bfSChris Mason return -ENOMEM; 35377fde62bfSChris Mason 35387fde62bfSChris Mason space_args.total_spaces = 0; 35398d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 35407fde62bfSChris Mason if (!dest) 35417fde62bfSChris Mason return -ENOMEM; 35427fde62bfSChris Mason dest_orig = dest; 35437fde62bfSChris Mason 35447fde62bfSChris Mason /* now we have a buffer to copy into */ 3545bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3546bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3547bf5fc093SJosef Bacik 354851788b1bSDan Rosenberg if (!slot_count) 354951788b1bSDan Rosenberg break; 355051788b1bSDan Rosenberg 3551bf5fc093SJosef Bacik info = NULL; 355272804905SJosef Bacik list_for_each_entry(tmp, &fs_info->space_info, list) { 3553bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3554bf5fc093SJosef Bacik info = tmp; 35557fde62bfSChris Mason break; 3556bf5fc093SJosef Bacik } 3557bf5fc093SJosef Bacik } 35587fde62bfSChris Mason 3559bf5fc093SJosef Bacik if (!info) 3560bf5fc093SJosef Bacik continue; 3561bf5fc093SJosef Bacik down_read(&info->groups_sem); 3562bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3563bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 3564c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 3565c065f5b1SSu Yue &space); 35667fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 35671406e432SJosef Bacik dest++; 35681406e432SJosef Bacik space_args.total_spaces++; 356951788b1bSDan Rosenberg slot_count--; 35701406e432SJosef Bacik } 357151788b1bSDan Rosenberg if (!slot_count) 357251788b1bSDan Rosenberg break; 3573bf5fc093SJosef Bacik } 3574bf5fc093SJosef Bacik up_read(&info->groups_sem); 3575bf5fc093SJosef Bacik } 35761406e432SJosef Bacik 357736523e95SDavid Sterba /* 357836523e95SDavid Sterba * Add global block reserve 357936523e95SDavid Sterba */ 358036523e95SDavid Sterba if (slot_count) { 35810b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 358236523e95SDavid Sterba 358336523e95SDavid Sterba spin_lock(&block_rsv->lock); 358436523e95SDavid Sterba space.total_bytes = block_rsv->size; 358536523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 358636523e95SDavid Sterba spin_unlock(&block_rsv->lock); 358736523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 358836523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 358936523e95SDavid Sterba space_args.total_spaces++; 359036523e95SDavid Sterba } 359136523e95SDavid Sterba 35922eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 35937fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 35947fde62bfSChris Mason 35957fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 35967fde62bfSChris Mason ret = -EFAULT; 35977fde62bfSChris Mason 35987fde62bfSChris Mason kfree(dest_orig); 35997fde62bfSChris Mason out: 36007fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 36011406e432SJosef Bacik ret = -EFAULT; 36021406e432SJosef Bacik 36031406e432SJosef Bacik return ret; 36041406e432SJosef Bacik } 36051406e432SJosef Bacik 36069a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 36079a8c28beSMiao Xie void __user *argp) 360846204592SSage Weil { 360946204592SSage Weil struct btrfs_trans_handle *trans; 361046204592SSage Weil u64 transid; 3611db5b493aSTsutomu Itoh int ret; 361246204592SSage Weil 3613d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 3614ff7c1d33SMiao Xie if (IS_ERR(trans)) { 3615ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 361698d5dc13STsutomu Itoh return PTR_ERR(trans); 3617ff7c1d33SMiao Xie 3618ff7c1d33SMiao Xie /* No running transaction, don't bother */ 3619ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 3620ff7c1d33SMiao Xie goto out; 3621ff7c1d33SMiao Xie } 362246204592SSage Weil transid = trans->transid; 362332cc4f87SDavid Sterba ret = btrfs_commit_transaction_async(trans); 36248b2b2d3cSTsutomu Itoh if (ret) { 36253a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3626db5b493aSTsutomu Itoh return ret; 36278b2b2d3cSTsutomu Itoh } 3628ff7c1d33SMiao Xie out: 362946204592SSage Weil if (argp) 363046204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 363146204592SSage Weil return -EFAULT; 363246204592SSage Weil return 0; 363346204592SSage Weil } 363446204592SSage Weil 36352ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 36369a8c28beSMiao Xie void __user *argp) 363746204592SSage Weil { 363846204592SSage Weil u64 transid; 363946204592SSage Weil 364046204592SSage Weil if (argp) { 364146204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 364246204592SSage Weil return -EFAULT; 364346204592SSage Weil } else { 364446204592SSage Weil transid = 0; /* current trans */ 364546204592SSage Weil } 36462ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 364746204592SSage Weil } 364846204592SSage Weil 3649b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 3650475f6387SJan Schmidt { 36510b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 3652475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 3653b8e95489SMiao Xie int ret; 3654475f6387SJan Schmidt 3655475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3656475f6387SJan Schmidt return -EPERM; 3657475f6387SJan Schmidt 3658475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 3659475f6387SJan Schmidt if (IS_ERR(sa)) 3660475f6387SJan Schmidt return PTR_ERR(sa); 3661475f6387SJan Schmidt 3662b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 3663b8e95489SMiao Xie ret = mnt_want_write_file(file); 3664b8e95489SMiao Xie if (ret) 3665b8e95489SMiao Xie goto out; 3666b8e95489SMiao Xie } 3667b8e95489SMiao Xie 36680b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 366963a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 367063a212abSStefan Behrens 0); 3671475f6387SJan Schmidt 36725afe6ce7SFilipe Manana /* 36735afe6ce7SFilipe Manana * Copy scrub args to user space even if btrfs_scrub_dev() returned an 36745afe6ce7SFilipe Manana * error. This is important as it allows user space to know how much 36755afe6ce7SFilipe Manana * progress scrub has done. For example, if scrub is canceled we get 36765afe6ce7SFilipe Manana * -ECANCELED from btrfs_scrub_dev() and return that error back to user 36775afe6ce7SFilipe Manana * space. Later user space can inspect the progress from the structure 36785afe6ce7SFilipe Manana * btrfs_ioctl_scrub_args and resume scrub from where it left off 36795afe6ce7SFilipe Manana * previously (btrfs-progs does this). 36805afe6ce7SFilipe Manana * If we fail to copy the btrfs_ioctl_scrub_args structure to user space 36815afe6ce7SFilipe Manana * then return -EFAULT to signal the structure was not copied or it may 36825afe6ce7SFilipe Manana * be corrupt and unreliable due to a partial copy. 36835afe6ce7SFilipe Manana */ 36845afe6ce7SFilipe Manana if (copy_to_user(arg, sa, sizeof(*sa))) 3685475f6387SJan Schmidt ret = -EFAULT; 3686475f6387SJan Schmidt 3687b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 3688b8e95489SMiao Xie mnt_drop_write_file(file); 3689b8e95489SMiao Xie out: 3690475f6387SJan Schmidt kfree(sa); 3691475f6387SJan Schmidt return ret; 3692475f6387SJan Schmidt } 3693475f6387SJan Schmidt 36942ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 3695475f6387SJan Schmidt { 3696475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3697475f6387SJan Schmidt return -EPERM; 3698475f6387SJan Schmidt 36992ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 3700475f6387SJan Schmidt } 3701475f6387SJan Schmidt 37022ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 3703475f6387SJan Schmidt void __user *arg) 3704475f6387SJan Schmidt { 3705475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 3706475f6387SJan Schmidt int ret; 3707475f6387SJan Schmidt 3708475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3709475f6387SJan Schmidt return -EPERM; 3710475f6387SJan Schmidt 3711475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 3712475f6387SJan Schmidt if (IS_ERR(sa)) 3713475f6387SJan Schmidt return PTR_ERR(sa); 3714475f6387SJan Schmidt 37152ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 3716475f6387SJan Schmidt 37174fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 3718475f6387SJan Schmidt ret = -EFAULT; 3719475f6387SJan Schmidt 3720475f6387SJan Schmidt kfree(sa); 3721475f6387SJan Schmidt return ret; 3722475f6387SJan Schmidt } 3723475f6387SJan Schmidt 37242ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 3725b27f7c0cSDavid Sterba void __user *arg) 3726c11d2c23SStefan Behrens { 3727c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 3728c11d2c23SStefan Behrens int ret; 3729c11d2c23SStefan Behrens 3730c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 3731c11d2c23SStefan Behrens if (IS_ERR(sa)) 3732c11d2c23SStefan Behrens return PTR_ERR(sa); 3733c11d2c23SStefan Behrens 3734b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 3735b27f7c0cSDavid Sterba kfree(sa); 3736b27f7c0cSDavid Sterba return -EPERM; 3737b27f7c0cSDavid Sterba } 3738b27f7c0cSDavid Sterba 37392ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 3740c11d2c23SStefan Behrens 3741eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 3742c11d2c23SStefan Behrens ret = -EFAULT; 3743c11d2c23SStefan Behrens 3744c11d2c23SStefan Behrens kfree(sa); 3745c11d2c23SStefan Behrens return ret; 3746c11d2c23SStefan Behrens } 3747c11d2c23SStefan Behrens 37482ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 37492ff7e61eSJeff Mahoney void __user *arg) 37503f6bcfbdSStefan Behrens { 37513f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 37523f6bcfbdSStefan Behrens int ret; 37533f6bcfbdSStefan Behrens 37543f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 37553f6bcfbdSStefan Behrens return -EPERM; 37563f6bcfbdSStefan Behrens 37573f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 37583f6bcfbdSStefan Behrens if (IS_ERR(p)) 37593f6bcfbdSStefan Behrens return PTR_ERR(p); 37603f6bcfbdSStefan Behrens 37613f6bcfbdSStefan Behrens switch (p->cmd) { 37623f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 3763bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 3764adfa97cbSIlya Dryomov ret = -EROFS; 3765adfa97cbSIlya Dryomov goto out; 3766adfa97cbSIlya Dryomov } 3767c3e1f96cSGoldwyn Rodrigues if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) { 3768e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 37693f6bcfbdSStefan Behrens } else { 37702ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 3771c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 37723f6bcfbdSStefan Behrens } 37733f6bcfbdSStefan Behrens break; 37743f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 37750b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 37763f6bcfbdSStefan Behrens ret = 0; 37773f6bcfbdSStefan Behrens break; 37783f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 377917d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 378097282031SAnand Jain ret = 0; 37813f6bcfbdSStefan Behrens break; 37823f6bcfbdSStefan Behrens default: 37833f6bcfbdSStefan Behrens ret = -EINVAL; 37843f6bcfbdSStefan Behrens break; 37853f6bcfbdSStefan Behrens } 37863f6bcfbdSStefan Behrens 3787d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 37883f6bcfbdSStefan Behrens ret = -EFAULT; 3789adfa97cbSIlya Dryomov out: 37903f6bcfbdSStefan Behrens kfree(p); 37913f6bcfbdSStefan Behrens return ret; 37923f6bcfbdSStefan Behrens } 37933f6bcfbdSStefan Behrens 3794d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 3795d7728c96SJan Schmidt { 3796d7728c96SJan Schmidt int ret = 0; 3797d7728c96SJan Schmidt int i; 3798740c3d22SChris Mason u64 rel_ptr; 3799d7728c96SJan Schmidt int size; 3800806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 3801d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 3802d7728c96SJan Schmidt struct btrfs_path *path; 3803d7728c96SJan Schmidt 380482b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 3805d7728c96SJan Schmidt return -EPERM; 3806d7728c96SJan Schmidt 3807d7728c96SJan Schmidt path = btrfs_alloc_path(); 3808d7728c96SJan Schmidt if (!path) { 3809d7728c96SJan Schmidt ret = -ENOMEM; 3810d7728c96SJan Schmidt goto out; 3811d7728c96SJan Schmidt } 3812d7728c96SJan Schmidt 3813d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 3814d7728c96SJan Schmidt if (IS_ERR(ipa)) { 3815d7728c96SJan Schmidt ret = PTR_ERR(ipa); 3816d7728c96SJan Schmidt ipa = NULL; 3817d7728c96SJan Schmidt goto out; 3818d7728c96SJan Schmidt } 3819d7728c96SJan Schmidt 3820d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 3821d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 3822d7728c96SJan Schmidt if (IS_ERR(ipath)) { 3823d7728c96SJan Schmidt ret = PTR_ERR(ipath); 3824d7728c96SJan Schmidt ipath = NULL; 3825d7728c96SJan Schmidt goto out; 3826d7728c96SJan Schmidt } 3827d7728c96SJan Schmidt 3828d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 3829d7728c96SJan Schmidt if (ret < 0) 3830d7728c96SJan Schmidt goto out; 3831d7728c96SJan Schmidt 3832d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 3833745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 3834745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 3835740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 3836d7728c96SJan Schmidt } 3837d7728c96SJan Schmidt 3838718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 3839718dc5faSOmar Sandoval ipath->fspath, size); 3840d7728c96SJan Schmidt if (ret) { 3841d7728c96SJan Schmidt ret = -EFAULT; 3842d7728c96SJan Schmidt goto out; 3843d7728c96SJan Schmidt } 3844d7728c96SJan Schmidt 3845d7728c96SJan Schmidt out: 3846d7728c96SJan Schmidt btrfs_free_path(path); 3847d7728c96SJan Schmidt free_ipath(ipath); 3848d7728c96SJan Schmidt kfree(ipa); 3849d7728c96SJan Schmidt 3850d7728c96SJan Schmidt return ret; 3851d7728c96SJan Schmidt } 3852d7728c96SJan Schmidt 3853d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 3854d7728c96SJan Schmidt { 3855d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 3856d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 3857d7728c96SJan Schmidt 3858d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 3859d7728c96SJan Schmidt inodes->bytes_left -= c; 3860d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 3861d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 3862d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 3863d7728c96SJan Schmidt inodes->elem_cnt += 3; 3864d7728c96SJan Schmidt } else { 3865d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 3866d7728c96SJan Schmidt inodes->bytes_left = 0; 3867d7728c96SJan Schmidt inodes->elem_missed += 3; 3868d7728c96SJan Schmidt } 3869d7728c96SJan Schmidt 3870d7728c96SJan Schmidt return 0; 3871d7728c96SJan Schmidt } 3872d7728c96SJan Schmidt 38732ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 3874d24a67b2SZygo Blaxell void __user *arg, int version) 3875d7728c96SJan Schmidt { 3876d7728c96SJan Schmidt int ret = 0; 3877d7728c96SJan Schmidt int size; 3878d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 3879d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 3880d7728c96SJan Schmidt struct btrfs_path *path = NULL; 3881d24a67b2SZygo Blaxell bool ignore_offset; 3882d7728c96SJan Schmidt 3883d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3884d7728c96SJan Schmidt return -EPERM; 3885d7728c96SJan Schmidt 3886d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 38877b9ea627SShailendra Verma if (IS_ERR(loi)) 38887b9ea627SShailendra Verma return PTR_ERR(loi); 3889d7728c96SJan Schmidt 3890d24a67b2SZygo Blaxell if (version == 1) { 3891d24a67b2SZygo Blaxell ignore_offset = false; 3892b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 3893d24a67b2SZygo Blaxell } else { 3894d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 3895d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 3896d24a67b2SZygo Blaxell ret = -EINVAL; 3897d24a67b2SZygo Blaxell goto out_loi; 3898d24a67b2SZygo Blaxell } 3899d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 3900d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 3901d24a67b2SZygo Blaxell ret = -EINVAL; 3902d24a67b2SZygo Blaxell goto out_loi; 3903d24a67b2SZygo Blaxell } 3904d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 3905b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 3906d24a67b2SZygo Blaxell } 3907d24a67b2SZygo Blaxell 3908d7728c96SJan Schmidt path = btrfs_alloc_path(); 3909d7728c96SJan Schmidt if (!path) { 3910d7728c96SJan Schmidt ret = -ENOMEM; 3911d7728c96SJan Schmidt goto out; 3912d7728c96SJan Schmidt } 3913d7728c96SJan Schmidt 3914d7728c96SJan Schmidt inodes = init_data_container(size); 3915d7728c96SJan Schmidt if (IS_ERR(inodes)) { 3916d7728c96SJan Schmidt ret = PTR_ERR(inodes); 3917d7728c96SJan Schmidt inodes = NULL; 3918d7728c96SJan Schmidt goto out; 3919d7728c96SJan Schmidt } 3920d7728c96SJan Schmidt 39212ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 3922d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 3923df031f07SLiu Bo if (ret == -EINVAL) 3924d7728c96SJan Schmidt ret = -ENOENT; 3925d7728c96SJan Schmidt if (ret < 0) 3926d7728c96SJan Schmidt goto out; 3927d7728c96SJan Schmidt 3928718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 3929718dc5faSOmar Sandoval size); 3930d7728c96SJan Schmidt if (ret) 3931d7728c96SJan Schmidt ret = -EFAULT; 3932d7728c96SJan Schmidt 3933d7728c96SJan Schmidt out: 3934d7728c96SJan Schmidt btrfs_free_path(path); 3935f54de068SDavid Sterba kvfree(inodes); 3936d24a67b2SZygo Blaxell out_loi: 3937d7728c96SJan Schmidt kfree(loi); 3938d7728c96SJan Schmidt 3939d7728c96SJan Schmidt return ret; 3940d7728c96SJan Schmidt } 3941d7728c96SJan Schmidt 3942008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 3943c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 3944c9e9f97bSIlya Dryomov { 3945c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 3946c9e9f97bSIlya Dryomov 3947c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 3948c9e9f97bSIlya Dryomov 39493009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 3950837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 3951837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 3952837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 3953a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 3954a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 3955837d5b6eSIlya Dryomov 3956c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 3957c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 3958c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 395919a39dceSIlya Dryomov 396019a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 396119a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 396219a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 3963c9e9f97bSIlya Dryomov } 3964c9e9f97bSIlya Dryomov 39659ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 3966c9e9f97bSIlya Dryomov { 3967496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 3968c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 3969c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 3970c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 3971ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 3972c9e9f97bSIlya Dryomov int ret; 3973c9e9f97bSIlya Dryomov 3974c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 3975c9e9f97bSIlya Dryomov return -EPERM; 3976c9e9f97bSIlya Dryomov 3977e54bfa31SLiu Bo ret = mnt_want_write_file(file); 39789ba1f6e4SLiu Bo if (ret) 39799ba1f6e4SLiu Bo return ret; 39809ba1f6e4SLiu Bo 3981ed0fb78fSIlya Dryomov again: 3982c3e1f96cSGoldwyn Rodrigues if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { 3983c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 3984ed0fb78fSIlya Dryomov need_unlock = true; 3985ed0fb78fSIlya Dryomov goto locked; 3986ed0fb78fSIlya Dryomov } 3987ed0fb78fSIlya Dryomov 3988ed0fb78fSIlya Dryomov /* 398901327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 3990ed0fb78fSIlya Dryomov * (1) some other op is running 3991ed0fb78fSIlya Dryomov * (2) balance is running 3992ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 3993ed0fb78fSIlya Dryomov */ 3994ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 3995ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 3996ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 39973009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 3998ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 3999dccdb07bSDavid Sterba /* 4000dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4001dccdb07bSDavid Sterba * we'll reexamine the status again. 4002dccdb07bSDavid Sterba */ 4003ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4004ed0fb78fSIlya Dryomov 4005ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 40063009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4007ed0fb78fSIlya Dryomov /* this is (3) */ 4008ed0fb78fSIlya Dryomov need_unlock = false; 4009ed0fb78fSIlya Dryomov goto locked; 4010ed0fb78fSIlya Dryomov } 4011ed0fb78fSIlya Dryomov 4012ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4013ed0fb78fSIlya Dryomov goto again; 4014ed0fb78fSIlya Dryomov } else { 4015ed0fb78fSIlya Dryomov /* this is (2) */ 4016ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4017ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4018ed0fb78fSIlya Dryomov goto out; 4019ed0fb78fSIlya Dryomov } 4020ed0fb78fSIlya Dryomov } else { 4021ed0fb78fSIlya Dryomov /* this is (1) */ 4022ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4023e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4024ed0fb78fSIlya Dryomov goto out; 4025ed0fb78fSIlya Dryomov } 4026ed0fb78fSIlya Dryomov 4027ed0fb78fSIlya Dryomov locked: 4028c9e9f97bSIlya Dryomov 4029c9e9f97bSIlya Dryomov if (arg) { 4030c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4031c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4032c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4033ed0fb78fSIlya Dryomov goto out_unlock; 4034c9e9f97bSIlya Dryomov } 4035de322263SIlya Dryomov 4036de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4037de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4038de322263SIlya Dryomov ret = -ENOTCONN; 4039de322263SIlya Dryomov goto out_bargs; 4040de322263SIlya Dryomov } 4041de322263SIlya Dryomov 4042de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4043de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4044de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4045de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4046de322263SIlya Dryomov 4047de322263SIlya Dryomov goto do_balance; 4048de322263SIlya Dryomov } 4049c9e9f97bSIlya Dryomov } else { 4050c9e9f97bSIlya Dryomov bargs = NULL; 4051c9e9f97bSIlya Dryomov } 4052c9e9f97bSIlya Dryomov 4053ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4054837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4055837d5b6eSIlya Dryomov goto out_bargs; 4056837d5b6eSIlya Dryomov } 4057837d5b6eSIlya Dryomov 40588d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4059c9e9f97bSIlya Dryomov if (!bctl) { 4060c9e9f97bSIlya Dryomov ret = -ENOMEM; 4061c9e9f97bSIlya Dryomov goto out_bargs; 4062c9e9f97bSIlya Dryomov } 4063c9e9f97bSIlya Dryomov 4064c9e9f97bSIlya Dryomov if (arg) { 4065c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4066c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4067c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4068c9e9f97bSIlya Dryomov 4069c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4070f43ffb60SIlya Dryomov } else { 4071f43ffb60SIlya Dryomov /* balance everything - no filters */ 4072f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4073c9e9f97bSIlya Dryomov } 4074c9e9f97bSIlya Dryomov 40758eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 40768eb93459SDavid Sterba ret = -EINVAL; 40770f89abf5SChristian Engelmayer goto out_bctl; 40788eb93459SDavid Sterba } 40798eb93459SDavid Sterba 4080de322263SIlya Dryomov do_balance: 4081c9e9f97bSIlya Dryomov /* 4082c3e1f96cSGoldwyn Rodrigues * Ownership of bctl and exclusive operation goes to btrfs_balance. 4083c3e1f96cSGoldwyn Rodrigues * bctl is freed in reset_balance_state, or, if restriper was paused 4084c3e1f96cSGoldwyn Rodrigues * all the way until unmount, in free_fs_info. The flag should be 4085c3e1f96cSGoldwyn Rodrigues * cleared after reset_balance_state. 4086c9e9f97bSIlya Dryomov */ 4087ed0fb78fSIlya Dryomov need_unlock = false; 4088ed0fb78fSIlya Dryomov 40896fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 40900f89abf5SChristian Engelmayer bctl = NULL; 4091ed0fb78fSIlya Dryomov 4092d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4093c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4094c9e9f97bSIlya Dryomov ret = -EFAULT; 4095c9e9f97bSIlya Dryomov } 4096c9e9f97bSIlya Dryomov 40970f89abf5SChristian Engelmayer out_bctl: 40980f89abf5SChristian Engelmayer kfree(bctl); 4099c9e9f97bSIlya Dryomov out_bargs: 4100c9e9f97bSIlya Dryomov kfree(bargs); 4101ed0fb78fSIlya Dryomov out_unlock: 4102c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4103ed0fb78fSIlya Dryomov if (need_unlock) 4104c3e1f96cSGoldwyn Rodrigues btrfs_exclop_finish(fs_info); 4105ed0fb78fSIlya Dryomov out: 4106e54bfa31SLiu Bo mnt_drop_write_file(file); 4107c9e9f97bSIlya Dryomov return ret; 4108c9e9f97bSIlya Dryomov } 4109c9e9f97bSIlya Dryomov 41102ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4111837d5b6eSIlya Dryomov { 4112837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4113837d5b6eSIlya Dryomov return -EPERM; 4114837d5b6eSIlya Dryomov 4115837d5b6eSIlya Dryomov switch (cmd) { 4116837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 41170b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4118a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 41190b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4120837d5b6eSIlya Dryomov } 4121837d5b6eSIlya Dryomov 4122837d5b6eSIlya Dryomov return -EINVAL; 4123837d5b6eSIlya Dryomov } 4124837d5b6eSIlya Dryomov 41252ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 412619a39dceSIlya Dryomov void __user *arg) 412719a39dceSIlya Dryomov { 412819a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 412919a39dceSIlya Dryomov int ret = 0; 413019a39dceSIlya Dryomov 413119a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 413219a39dceSIlya Dryomov return -EPERM; 413319a39dceSIlya Dryomov 413419a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 413519a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 413619a39dceSIlya Dryomov ret = -ENOTCONN; 413719a39dceSIlya Dryomov goto out; 413819a39dceSIlya Dryomov } 413919a39dceSIlya Dryomov 41408d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 414119a39dceSIlya Dryomov if (!bargs) { 414219a39dceSIlya Dryomov ret = -ENOMEM; 414319a39dceSIlya Dryomov goto out; 414419a39dceSIlya Dryomov } 414519a39dceSIlya Dryomov 4146008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 414719a39dceSIlya Dryomov 414819a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 414919a39dceSIlya Dryomov ret = -EFAULT; 415019a39dceSIlya Dryomov 415119a39dceSIlya Dryomov kfree(bargs); 415219a39dceSIlya Dryomov out: 415319a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 415419a39dceSIlya Dryomov return ret; 415519a39dceSIlya Dryomov } 415619a39dceSIlya Dryomov 4157905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 41585d13a37bSArne Jansen { 41590b246afaSJeff Mahoney struct inode *inode = file_inode(file); 41600b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 41615d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 41625d13a37bSArne Jansen int ret; 41635d13a37bSArne Jansen 41645d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 41655d13a37bSArne Jansen return -EPERM; 41665d13a37bSArne Jansen 4167905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4168905b0ddaSMiao Xie if (ret) 4169905b0ddaSMiao Xie return ret; 41705d13a37bSArne Jansen 41715d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4172905b0ddaSMiao Xie if (IS_ERR(sa)) { 4173905b0ddaSMiao Xie ret = PTR_ERR(sa); 4174905b0ddaSMiao Xie goto drop_write; 4175905b0ddaSMiao Xie } 41765d13a37bSArne Jansen 41770b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 41785d13a37bSArne Jansen 41795d13a37bSArne Jansen switch (sa->cmd) { 41805d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4181340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 41825d13a37bSArne Jansen break; 41835d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4184340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 41855d13a37bSArne Jansen break; 41865d13a37bSArne Jansen default: 41875d13a37bSArne Jansen ret = -EINVAL; 41885d13a37bSArne Jansen break; 41895d13a37bSArne Jansen } 41905d13a37bSArne Jansen 41915d13a37bSArne Jansen kfree(sa); 41920b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4193905b0ddaSMiao Xie drop_write: 4194905b0ddaSMiao Xie mnt_drop_write_file(file); 41955d13a37bSArne Jansen return ret; 41965d13a37bSArne Jansen } 41975d13a37bSArne Jansen 4198905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 41995d13a37bSArne Jansen { 42000b246afaSJeff Mahoney struct inode *inode = file_inode(file); 42010b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 42020b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 42035d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 42045d13a37bSArne Jansen struct btrfs_trans_handle *trans; 42055d13a37bSArne Jansen int ret; 42065d13a37bSArne Jansen int err; 42075d13a37bSArne Jansen 42085d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 42095d13a37bSArne Jansen return -EPERM; 42105d13a37bSArne Jansen 4211905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4212905b0ddaSMiao Xie if (ret) 4213905b0ddaSMiao Xie return ret; 42145d13a37bSArne Jansen 42155d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4216905b0ddaSMiao Xie if (IS_ERR(sa)) { 4217905b0ddaSMiao Xie ret = PTR_ERR(sa); 4218905b0ddaSMiao Xie goto drop_write; 4219905b0ddaSMiao Xie } 42205d13a37bSArne Jansen 42215d13a37bSArne Jansen trans = btrfs_join_transaction(root); 42225d13a37bSArne Jansen if (IS_ERR(trans)) { 42235d13a37bSArne Jansen ret = PTR_ERR(trans); 42245d13a37bSArne Jansen goto out; 42255d13a37bSArne Jansen } 42265d13a37bSArne Jansen 42275d13a37bSArne Jansen if (sa->assign) { 42289f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 42295d13a37bSArne Jansen } else { 423039616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 42315d13a37bSArne Jansen } 42325d13a37bSArne Jansen 4233e082f563SQu Wenruo /* update qgroup status and info */ 4234280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4235e082f563SQu Wenruo if (err < 0) 42360b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4237ad8403dfSAnand Jain "failed to update qgroup status and info"); 42383a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 42395d13a37bSArne Jansen if (err && !ret) 42405d13a37bSArne Jansen ret = err; 42415d13a37bSArne Jansen 42425d13a37bSArne Jansen out: 42435d13a37bSArne Jansen kfree(sa); 4244905b0ddaSMiao Xie drop_write: 4245905b0ddaSMiao Xie mnt_drop_write_file(file); 42465d13a37bSArne Jansen return ret; 42475d13a37bSArne Jansen } 42485d13a37bSArne Jansen 4249905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 42505d13a37bSArne Jansen { 42510b246afaSJeff Mahoney struct inode *inode = file_inode(file); 42520b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 42535d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 42545d13a37bSArne Jansen struct btrfs_trans_handle *trans; 42555d13a37bSArne Jansen int ret; 42565d13a37bSArne Jansen int err; 42575d13a37bSArne Jansen 42585d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 42595d13a37bSArne Jansen return -EPERM; 42605d13a37bSArne Jansen 4261905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4262905b0ddaSMiao Xie if (ret) 4263905b0ddaSMiao Xie return ret; 42645d13a37bSArne Jansen 42655d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4266905b0ddaSMiao Xie if (IS_ERR(sa)) { 4267905b0ddaSMiao Xie ret = PTR_ERR(sa); 4268905b0ddaSMiao Xie goto drop_write; 4269905b0ddaSMiao Xie } 42705d13a37bSArne Jansen 4271d86e56cfSMiao Xie if (!sa->qgroupid) { 4272d86e56cfSMiao Xie ret = -EINVAL; 4273d86e56cfSMiao Xie goto out; 4274d86e56cfSMiao Xie } 4275d86e56cfSMiao Xie 42765d13a37bSArne Jansen trans = btrfs_join_transaction(root); 42775d13a37bSArne Jansen if (IS_ERR(trans)) { 42785d13a37bSArne Jansen ret = PTR_ERR(trans); 42795d13a37bSArne Jansen goto out; 42805d13a37bSArne Jansen } 42815d13a37bSArne Jansen 42825d13a37bSArne Jansen if (sa->create) { 428349a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 42845d13a37bSArne Jansen } else { 42853efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 42865d13a37bSArne Jansen } 42875d13a37bSArne Jansen 42883a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 42895d13a37bSArne Jansen if (err && !ret) 42905d13a37bSArne Jansen ret = err; 42915d13a37bSArne Jansen 42925d13a37bSArne Jansen out: 42935d13a37bSArne Jansen kfree(sa); 4294905b0ddaSMiao Xie drop_write: 4295905b0ddaSMiao Xie mnt_drop_write_file(file); 42965d13a37bSArne Jansen return ret; 42975d13a37bSArne Jansen } 42985d13a37bSArne Jansen 4299905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 43005d13a37bSArne Jansen { 43010b246afaSJeff Mahoney struct inode *inode = file_inode(file); 43020b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 43035d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 43045d13a37bSArne Jansen struct btrfs_trans_handle *trans; 43055d13a37bSArne Jansen int ret; 43065d13a37bSArne Jansen int err; 43075d13a37bSArne Jansen u64 qgroupid; 43085d13a37bSArne Jansen 43095d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 43105d13a37bSArne Jansen return -EPERM; 43115d13a37bSArne Jansen 4312905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4313905b0ddaSMiao Xie if (ret) 4314905b0ddaSMiao Xie return ret; 43155d13a37bSArne Jansen 43165d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4317905b0ddaSMiao Xie if (IS_ERR(sa)) { 4318905b0ddaSMiao Xie ret = PTR_ERR(sa); 4319905b0ddaSMiao Xie goto drop_write; 4320905b0ddaSMiao Xie } 43215d13a37bSArne Jansen 43225d13a37bSArne Jansen trans = btrfs_join_transaction(root); 43235d13a37bSArne Jansen if (IS_ERR(trans)) { 43245d13a37bSArne Jansen ret = PTR_ERR(trans); 43255d13a37bSArne Jansen goto out; 43265d13a37bSArne Jansen } 43275d13a37bSArne Jansen 43285d13a37bSArne Jansen qgroupid = sa->qgroupid; 43295d13a37bSArne Jansen if (!qgroupid) { 43305d13a37bSArne Jansen /* take the current subvol as qgroup */ 43315d13a37bSArne Jansen qgroupid = root->root_key.objectid; 43325d13a37bSArne Jansen } 43335d13a37bSArne Jansen 4334f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 43355d13a37bSArne Jansen 43363a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 43375d13a37bSArne Jansen if (err && !ret) 43385d13a37bSArne Jansen ret = err; 43395d13a37bSArne Jansen 43405d13a37bSArne Jansen out: 43415d13a37bSArne Jansen kfree(sa); 4342905b0ddaSMiao Xie drop_write: 4343905b0ddaSMiao Xie mnt_drop_write_file(file); 43445d13a37bSArne Jansen return ret; 43455d13a37bSArne Jansen } 43465d13a37bSArne Jansen 43472f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 43482f232036SJan Schmidt { 43490b246afaSJeff Mahoney struct inode *inode = file_inode(file); 43500b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 43512f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 43522f232036SJan Schmidt int ret; 43532f232036SJan Schmidt 43542f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 43552f232036SJan Schmidt return -EPERM; 43562f232036SJan Schmidt 43572f232036SJan Schmidt ret = mnt_want_write_file(file); 43582f232036SJan Schmidt if (ret) 43592f232036SJan Schmidt return ret; 43602f232036SJan Schmidt 43612f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 43622f232036SJan Schmidt if (IS_ERR(qsa)) { 43632f232036SJan Schmidt ret = PTR_ERR(qsa); 43642f232036SJan Schmidt goto drop_write; 43652f232036SJan Schmidt } 43662f232036SJan Schmidt 43672f232036SJan Schmidt if (qsa->flags) { 43682f232036SJan Schmidt ret = -EINVAL; 43692f232036SJan Schmidt goto out; 43702f232036SJan Schmidt } 43712f232036SJan Schmidt 43720b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 43732f232036SJan Schmidt 43742f232036SJan Schmidt out: 43752f232036SJan Schmidt kfree(qsa); 43762f232036SJan Schmidt drop_write: 43772f232036SJan Schmidt mnt_drop_write_file(file); 43782f232036SJan Schmidt return ret; 43792f232036SJan Schmidt } 43802f232036SJan Schmidt 4381b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, 4382b929c1d8SMarcos Paulo de Souza void __user *arg) 43832f232036SJan Schmidt { 43840afb603aSGoldwyn Rodrigues struct btrfs_ioctl_quota_rescan_args qsa = {0}; 43852f232036SJan Schmidt 43862f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 43872f232036SJan Schmidt return -EPERM; 43882f232036SJan Schmidt 43890b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 43900afb603aSGoldwyn Rodrigues qsa.flags = 1; 43910afb603aSGoldwyn Rodrigues qsa.progress = fs_info->qgroup_rescan_progress.objectid; 43922f232036SJan Schmidt } 43932f232036SJan Schmidt 43940afb603aSGoldwyn Rodrigues if (copy_to_user(arg, &qsa, sizeof(qsa))) 4395991a3daeSAnand Jain return -EFAULT; 43962f232036SJan Schmidt 4397991a3daeSAnand Jain return 0; 43982f232036SJan Schmidt } 43992f232036SJan Schmidt 4400b929c1d8SMarcos Paulo de Souza static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info, 4401b929c1d8SMarcos Paulo de Souza void __user *arg) 440257254b6eSJan Schmidt { 440357254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 440457254b6eSJan Schmidt return -EPERM; 440557254b6eSJan Schmidt 44060b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 440757254b6eSJan Schmidt } 440857254b6eSJan Schmidt 4409abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 4410e4fed17aSChristian Brauner struct user_namespace *mnt_userns, 4411abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 44128ea05e3aSAlexander Block { 4413496ad9aaSAl Viro struct inode *inode = file_inode(file); 44140b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 44158ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 44168ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 44178ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 441895582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 44198ea05e3aSAlexander Block int ret = 0; 4420dd5f9615SStefan Behrens int received_uuid_changed; 44218ea05e3aSAlexander Block 4422e4fed17aSChristian Brauner if (!inode_owner_or_capable(mnt_userns, inode)) 4423bd60ea0fSDavid Sterba return -EPERM; 4424bd60ea0fSDavid Sterba 44258ea05e3aSAlexander Block ret = mnt_want_write_file(file); 44268ea05e3aSAlexander Block if (ret < 0) 44278ea05e3aSAlexander Block return ret; 44288ea05e3aSAlexander Block 44290b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 44308ea05e3aSAlexander Block 44314a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 44328ea05e3aSAlexander Block ret = -EINVAL; 44338ea05e3aSAlexander Block goto out; 44348ea05e3aSAlexander Block } 44358ea05e3aSAlexander Block 44368ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 44378ea05e3aSAlexander Block ret = -EROFS; 44388ea05e3aSAlexander Block goto out; 44398ea05e3aSAlexander Block } 44408ea05e3aSAlexander Block 4441dd5f9615SStefan Behrens /* 4442dd5f9615SStefan Behrens * 1 - root item 4443dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 4444dd5f9615SStefan Behrens */ 4445dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 44468ea05e3aSAlexander Block if (IS_ERR(trans)) { 44478ea05e3aSAlexander Block ret = PTR_ERR(trans); 44488ea05e3aSAlexander Block trans = NULL; 44498ea05e3aSAlexander Block goto out; 44508ea05e3aSAlexander Block } 44518ea05e3aSAlexander Block 44528ea05e3aSAlexander Block sa->rtransid = trans->transid; 44538ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 44548ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 44558ea05e3aSAlexander Block 4456dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 4457dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 4458dd5f9615SStefan Behrens if (received_uuid_changed && 4459d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 4460d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 4461dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4462dd5f9615SStefan Behrens root->root_key.objectid); 4463d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 4464d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 4465d87ff758SNikolay Borisov btrfs_end_transaction(trans); 4466d87ff758SNikolay Borisov goto out; 4467d87ff758SNikolay Borisov } 4468d87ff758SNikolay Borisov } 44698ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 44708ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 44718ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 44723cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 44733cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 44743cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 44753cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 44768ea05e3aSAlexander Block 44770b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 44788ea05e3aSAlexander Block &root->root_key, &root->root_item); 44798ea05e3aSAlexander Block if (ret < 0) { 44803a45bb20SJeff Mahoney btrfs_end_transaction(trans); 44818ea05e3aSAlexander Block goto out; 4482dd5f9615SStefan Behrens } 4483dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 4484cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 4485dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4486dd5f9615SStefan Behrens root->root_key.objectid); 4487dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 448866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 4489efd38150SNikolay Borisov btrfs_end_transaction(trans); 4490dd5f9615SStefan Behrens goto out; 4491dd5f9615SStefan Behrens } 4492dd5f9615SStefan Behrens } 44933a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 4494abccd00fSHugo Mills out: 44950b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4496abccd00fSHugo Mills mnt_drop_write_file(file); 4497abccd00fSHugo Mills return ret; 4498abccd00fSHugo Mills } 4499abccd00fSHugo Mills 4500abccd00fSHugo Mills #ifdef CONFIG_64BIT 4501abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 4502abccd00fSHugo Mills void __user *arg) 4503abccd00fSHugo Mills { 4504abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 4505abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 4506abccd00fSHugo Mills int ret = 0; 4507abccd00fSHugo Mills 4508abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 45097b9ea627SShailendra Verma if (IS_ERR(args32)) 45107b9ea627SShailendra Verma return PTR_ERR(args32); 4511abccd00fSHugo Mills 45128d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 451384dbeb87SDan Carpenter if (!args64) { 451484dbeb87SDan Carpenter ret = -ENOMEM; 4515abccd00fSHugo Mills goto out; 4516abccd00fSHugo Mills } 4517abccd00fSHugo Mills 4518abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 4519abccd00fSHugo Mills args64->stransid = args32->stransid; 4520abccd00fSHugo Mills args64->rtransid = args32->rtransid; 4521abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 4522abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 4523abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 4524abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 4525abccd00fSHugo Mills args64->flags = args32->flags; 4526abccd00fSHugo Mills 4527e4fed17aSChristian Brauner ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64); 4528abccd00fSHugo Mills if (ret) 4529abccd00fSHugo Mills goto out; 4530abccd00fSHugo Mills 4531abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 4532abccd00fSHugo Mills args32->stransid = args64->stransid; 4533abccd00fSHugo Mills args32->rtransid = args64->rtransid; 4534abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 4535abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 4536abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 4537abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 4538abccd00fSHugo Mills args32->flags = args64->flags; 4539abccd00fSHugo Mills 4540abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 4541abccd00fSHugo Mills if (ret) 4542abccd00fSHugo Mills ret = -EFAULT; 4543abccd00fSHugo Mills 4544abccd00fSHugo Mills out: 4545abccd00fSHugo Mills kfree(args32); 4546abccd00fSHugo Mills kfree(args64); 4547abccd00fSHugo Mills return ret; 4548abccd00fSHugo Mills } 4549abccd00fSHugo Mills #endif 4550abccd00fSHugo Mills 4551abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 4552abccd00fSHugo Mills void __user *arg) 4553abccd00fSHugo Mills { 4554abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 4555abccd00fSHugo Mills int ret = 0; 4556abccd00fSHugo Mills 4557abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 45587b9ea627SShailendra Verma if (IS_ERR(sa)) 45597b9ea627SShailendra Verma return PTR_ERR(sa); 4560abccd00fSHugo Mills 4561e4fed17aSChristian Brauner ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa); 4562abccd00fSHugo Mills 4563abccd00fSHugo Mills if (ret) 4564abccd00fSHugo Mills goto out; 4565abccd00fSHugo Mills 45668ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 45678ea05e3aSAlexander Block if (ret) 45688ea05e3aSAlexander Block ret = -EFAULT; 45698ea05e3aSAlexander Block 45708ea05e3aSAlexander Block out: 45718ea05e3aSAlexander Block kfree(sa); 45728ea05e3aSAlexander Block return ret; 45738ea05e3aSAlexander Block } 45748ea05e3aSAlexander Block 4575b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, 4576b929c1d8SMarcos Paulo de Souza void __user *arg) 4577867ab667Sjeff.liu { 4578a1b83ac5SAnand Jain size_t len; 4579867ab667Sjeff.liu int ret; 4580a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 4581a1b83ac5SAnand Jain 45820b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 45830b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 45840b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 4585a1b83ac5SAnand Jain 4586a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 4587867ab667Sjeff.liu 4588867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 45890b246afaSJeff Mahoney btrfs_warn(fs_info, 45900b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 45910b246afaSJeff Mahoney --len); 4592867ab667Sjeff.liu } 4593867ab667Sjeff.liu 4594867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 4595867ab667Sjeff.liu 4596867ab667Sjeff.liu return ret ? -EFAULT : 0; 4597867ab667Sjeff.liu } 4598867ab667Sjeff.liu 4599a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 4600a8bfd4abSjeff.liu { 46010b246afaSJeff Mahoney struct inode *inode = file_inode(file); 46020b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 46030b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 46040b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 4605a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 4606a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 4607a8bfd4abSjeff.liu int ret; 4608a8bfd4abSjeff.liu 4609a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 4610a8bfd4abSjeff.liu return -EPERM; 4611a8bfd4abSjeff.liu 4612a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 4613a8bfd4abSjeff.liu return -EFAULT; 4614a8bfd4abSjeff.liu 4615a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 46160b246afaSJeff Mahoney btrfs_err(fs_info, 46175d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 4618a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 4619a8bfd4abSjeff.liu return -EINVAL; 4620a8bfd4abSjeff.liu } 4621a8bfd4abSjeff.liu 4622a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 4623a8bfd4abSjeff.liu if (ret) 4624a8bfd4abSjeff.liu return ret; 4625a8bfd4abSjeff.liu 4626a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 4627a8bfd4abSjeff.liu if (IS_ERR(trans)) { 4628a8bfd4abSjeff.liu ret = PTR_ERR(trans); 4629a8bfd4abSjeff.liu goto out_unlock; 4630a8bfd4abSjeff.liu } 4631a8bfd4abSjeff.liu 46320b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 4633a8bfd4abSjeff.liu strcpy(super_block->label, label); 46340b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 46353a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 4636a8bfd4abSjeff.liu 4637a8bfd4abSjeff.liu out_unlock: 4638a8bfd4abSjeff.liu mnt_drop_write_file(file); 4639a8bfd4abSjeff.liu return ret; 4640a8bfd4abSjeff.liu } 4641a8bfd4abSjeff.liu 46422eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 46432eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 46442eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 46452eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 46462eaa055fSJeff Mahoney 4647d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 46482eaa055fSJeff Mahoney { 46494d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 46502eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 46512eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 46522eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 46532eaa055fSJeff Mahoney }; 46542eaa055fSJeff Mahoney 46552eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 46562eaa055fSJeff Mahoney return -EFAULT; 46572eaa055fSJeff Mahoney 46582eaa055fSJeff Mahoney return 0; 46592eaa055fSJeff Mahoney } 46602eaa055fSJeff Mahoney 4661b929c1d8SMarcos Paulo de Souza static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, 4662b929c1d8SMarcos Paulo de Souza void __user *arg) 46632eaa055fSJeff Mahoney { 46640b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 46652eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 46662eaa055fSJeff Mahoney 46672eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 46682eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 46692eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 46702eaa055fSJeff Mahoney 46712eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 46722eaa055fSJeff Mahoney return -EFAULT; 46732eaa055fSJeff Mahoney 46742eaa055fSJeff Mahoney return 0; 46752eaa055fSJeff Mahoney } 46762eaa055fSJeff Mahoney 46772ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 46783b02a68aSJeff Mahoney enum btrfs_feature_set set, 46792eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 46802eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 46812eaa055fSJeff Mahoney { 4682f10152bcSDavid Sterba const char *type = btrfs_feature_set_name(set); 46833b02a68aSJeff Mahoney char *names; 46842eaa055fSJeff Mahoney u64 disallowed, unsupported; 46852eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 46862eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 46872eaa055fSJeff Mahoney 46882eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 46892eaa055fSJeff Mahoney if (unsupported) { 46903b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 46913b02a68aSJeff Mahoney if (names) { 46920b246afaSJeff Mahoney btrfs_warn(fs_info, 46933b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 46943b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 46953b02a68aSJeff Mahoney kfree(names); 46963b02a68aSJeff Mahoney } else 46970b246afaSJeff Mahoney btrfs_warn(fs_info, 46982eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 46992eaa055fSJeff Mahoney type, unsupported); 47002eaa055fSJeff Mahoney return -EOPNOTSUPP; 47012eaa055fSJeff Mahoney } 47022eaa055fSJeff Mahoney 47032eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 47042eaa055fSJeff Mahoney if (disallowed) { 47053b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 47063b02a68aSJeff Mahoney if (names) { 47070b246afaSJeff Mahoney btrfs_warn(fs_info, 47083b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 47093b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 47103b02a68aSJeff Mahoney kfree(names); 47113b02a68aSJeff Mahoney } else 47120b246afaSJeff Mahoney btrfs_warn(fs_info, 47132eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 47142eaa055fSJeff Mahoney type, disallowed); 47152eaa055fSJeff Mahoney return -EPERM; 47162eaa055fSJeff Mahoney } 47172eaa055fSJeff Mahoney 47182eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 47192eaa055fSJeff Mahoney if (disallowed) { 47203b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 47213b02a68aSJeff Mahoney if (names) { 47220b246afaSJeff Mahoney btrfs_warn(fs_info, 47233b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 47243b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 47253b02a68aSJeff Mahoney kfree(names); 47263b02a68aSJeff Mahoney } else 47270b246afaSJeff Mahoney btrfs_warn(fs_info, 47282eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 47292eaa055fSJeff Mahoney type, disallowed); 47302eaa055fSJeff Mahoney return -EPERM; 47312eaa055fSJeff Mahoney } 47322eaa055fSJeff Mahoney 47332eaa055fSJeff Mahoney return 0; 47342eaa055fSJeff Mahoney } 47352eaa055fSJeff Mahoney 47362ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 47372ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 47382eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 47392eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 47402eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 47412eaa055fSJeff Mahoney 47422eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 47432eaa055fSJeff Mahoney { 47440b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47450b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47460b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 47470b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 47482eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 47492eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 47502eaa055fSJeff Mahoney u64 newflags; 47512eaa055fSJeff Mahoney int ret; 47522eaa055fSJeff Mahoney 47532eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 47542eaa055fSJeff Mahoney return -EPERM; 47552eaa055fSJeff Mahoney 47562eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 47572eaa055fSJeff Mahoney return -EFAULT; 47582eaa055fSJeff Mahoney 47592eaa055fSJeff Mahoney /* Nothing to do */ 47602eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 47612eaa055fSJeff Mahoney !flags[0].incompat_flags) 47622eaa055fSJeff Mahoney return 0; 47632eaa055fSJeff Mahoney 47642ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 47652eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 47662eaa055fSJeff Mahoney if (ret) 47672eaa055fSJeff Mahoney return ret; 47682eaa055fSJeff Mahoney 47692ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 47702eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 47712eaa055fSJeff Mahoney if (ret) 47722eaa055fSJeff Mahoney return ret; 47732eaa055fSJeff Mahoney 47742ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 47752eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 47762eaa055fSJeff Mahoney if (ret) 47772eaa055fSJeff Mahoney return ret; 47782eaa055fSJeff Mahoney 47797ab19625SDavid Sterba ret = mnt_want_write_file(file); 47807ab19625SDavid Sterba if (ret) 47817ab19625SDavid Sterba return ret; 47827ab19625SDavid Sterba 47838051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 47847ab19625SDavid Sterba if (IS_ERR(trans)) { 47857ab19625SDavid Sterba ret = PTR_ERR(trans); 47867ab19625SDavid Sterba goto out_drop_write; 47877ab19625SDavid Sterba } 47882eaa055fSJeff Mahoney 47890b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 47902eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 47912eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 47922eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 47932eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 47942eaa055fSJeff Mahoney 47952eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 47962eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 47972eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 47982eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 47992eaa055fSJeff Mahoney 48002eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 48012eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 48022eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 48032eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 48040b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 48052eaa055fSJeff Mahoney 48063a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 48077ab19625SDavid Sterba out_drop_write: 48087ab19625SDavid Sterba mnt_drop_write_file(file); 48097ab19625SDavid Sterba 48107ab19625SDavid Sterba return ret; 48112eaa055fSJeff Mahoney } 48122eaa055fSJeff Mahoney 48132351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 48142351f431SJosef Bacik { 48152351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 48162351f431SJosef Bacik int ret; 48172351f431SJosef Bacik 48182351f431SJosef Bacik if (compat) { 48192351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 48202351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 48212351f431SJosef Bacik 48222351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 48232351f431SJosef Bacik if (ret) 48242351f431SJosef Bacik return -EFAULT; 48252351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 48262351f431SJosef Bacik if (!arg) 48272351f431SJosef Bacik return -ENOMEM; 48282351f431SJosef Bacik arg->send_fd = args32.send_fd; 48292351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 48302351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 48312351f431SJosef Bacik arg->parent_root = args32.parent_root; 48322351f431SJosef Bacik arg->flags = args32.flags; 48332351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 48342351f431SJosef Bacik sizeof(args32.reserved)); 48352351f431SJosef Bacik #else 48362351f431SJosef Bacik return -ENOTTY; 48372351f431SJosef Bacik #endif 48382351f431SJosef Bacik } else { 48392351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 48402351f431SJosef Bacik if (IS_ERR(arg)) 48412351f431SJosef Bacik return PTR_ERR(arg); 48422351f431SJosef Bacik } 48432351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 48442351f431SJosef Bacik kfree(arg); 48452351f431SJosef Bacik return ret; 48462351f431SJosef Bacik } 48472351f431SJosef Bacik 4848f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 4849f46b5a66SChristoph Hellwig cmd, unsigned long arg) 4850f46b5a66SChristoph Hellwig { 48510b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48520b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48530b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48544bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 4855f46b5a66SChristoph Hellwig 4856f46b5a66SChristoph Hellwig switch (cmd) { 48576cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 48586cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 485940cf931fSEric Sandeen case FS_IOC_GETFSLABEL: 4860b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_fslabel(fs_info, argp); 486140cf931fSEric Sandeen case FS_IOC_SETFSLABEL: 486240cf931fSEric Sandeen return btrfs_ioctl_set_fslabel(file, argp); 4863f7039b1dSLi Dongyang case FITRIM: 4864b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_fitrim(fs_info, argp); 4865f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 4866fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 4867fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 4868fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 48693de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 4870fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 48716f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 48726f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 487376dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 4874949964c9SMarcos Paulo de Souza return btrfs_ioctl_snap_destroy(file, argp, false); 4875949964c9SMarcos Paulo de Souza case BTRFS_IOC_SNAP_DESTROY_V2: 4876949964c9SMarcos Paulo de Souza return btrfs_ioctl_snap_destroy(file, argp, true); 48770caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 48780caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 48790caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 48800caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 48816ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 48826ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 4883f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 48841e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 48851e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 48861e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 4887f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 4888198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 4889f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 48902ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 4891f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 4892da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 48936b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 48946b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 4895475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 48962ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 4897475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 48982ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 4899f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 49009ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 4901ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 4902ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 4903cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 4904cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 4905ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 4906ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 4907d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 4908d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 4909d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 4910d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 4911d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 4912d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 49131406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 49142ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 49159b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 49169b199859SFilipe David Borba Manana int ret; 49179b199859SFilipe David Borba Manana 49189db4dc24SNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false); 49199b199859SFilipe David Borba Manana if (ret) 49209b199859SFilipe David Borba Manana return ret; 49210b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 49222fad4e83SDavid Sterba /* 49232fad4e83SDavid Sterba * The transaction thread may want to do more work, 492401327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 49252fad4e83SDavid Sterba * processing uncleaned subvols. 49262fad4e83SDavid Sterba */ 49270b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 49289b199859SFilipe David Borba Manana return ret; 49299b199859SFilipe David Borba Manana } 493046204592SSage Weil case BTRFS_IOC_START_SYNC: 49319a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 493246204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 49332ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 4934475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 4935b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 4936475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 49372ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 4938475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 49392ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 4940c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 49419ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 4942837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 49432ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 494419a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 49452ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 49468ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 49478ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 4948abccd00fSHugo Mills #ifdef CONFIG_64BIT 4949abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 4950abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 4951abccd00fSHugo Mills #endif 495231db9f7cSAlexander Block case BTRFS_IOC_SEND: 49532351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 49542351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 49552351f431SJosef Bacik case BTRFS_IOC_SEND_32: 49562351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 49572351f431SJosef Bacik #endif 4958c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 49592ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 49605d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 4961905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 49625d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 4963905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 49645d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 4965905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 49665d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 4967905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 49682f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 49692f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 49702f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 4971b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_status(fs_info, argp); 497257254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 4973b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_quota_rescan_wait(fs_info, argp); 49743f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 49752ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 49762eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 4977d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 49782eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 4979b929c1d8SMarcos Paulo de Souza return btrfs_ioctl_get_features(fs_info, argp); 49802eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 49812eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 4982b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 4983b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 498442e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 498542e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 498623d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 498723d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 498814605409SBoris Burkov case FS_IOC_ENABLE_VERITY: 498914605409SBoris Burkov return fsverity_ioctl_enable(file, (const void __user *)argp); 499014605409SBoris Burkov case FS_IOC_MEASURE_VERITY: 499114605409SBoris Burkov return fsverity_ioctl_measure(file, argp); 4992f46b5a66SChristoph Hellwig } 4993f46b5a66SChristoph Hellwig 4994f46b5a66SChristoph Hellwig return -ENOTTY; 4995f46b5a66SChristoph Hellwig } 49964c63c245SLuke Dashjr 49974c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 49984c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 49994c63c245SLuke Dashjr { 50002a362249SJeff Mahoney /* 50012a362249SJeff Mahoney * These all access 32-bit values anyway so no further 50022a362249SJeff Mahoney * handling is necessary. 50032a362249SJeff Mahoney */ 50044c63c245SLuke Dashjr switch (cmd) { 50054c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 50064c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 50074c63c245SLuke Dashjr break; 50084c63c245SLuke Dashjr } 50094c63c245SLuke Dashjr 50104c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 50114c63c245SLuke Dashjr } 50124c63c245SLuke Dashjr #endif 5013