1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 2f46b5a66SChristoph Hellwig /* 3f46b5a66SChristoph Hellwig * Copyright (C) 2007 Oracle. All rights reserved. 4f46b5a66SChristoph Hellwig */ 5f46b5a66SChristoph Hellwig 6f46b5a66SChristoph Hellwig #include <linux/kernel.h> 7f46b5a66SChristoph Hellwig #include <linux/bio.h> 8f46b5a66SChristoph Hellwig #include <linux/file.h> 9f46b5a66SChristoph Hellwig #include <linux/fs.h> 10cb8e7090SChristoph Hellwig #include <linux/fsnotify.h> 11f46b5a66SChristoph Hellwig #include <linux/pagemap.h> 12f46b5a66SChristoph Hellwig #include <linux/highmem.h> 13f46b5a66SChristoph Hellwig #include <linux/time.h> 14f46b5a66SChristoph Hellwig #include <linux/string.h> 15f46b5a66SChristoph Hellwig #include <linux/backing-dev.h> 16cb8e7090SChristoph Hellwig #include <linux/mount.h> 17cb8e7090SChristoph Hellwig #include <linux/namei.h> 18f46b5a66SChristoph Hellwig #include <linux/writeback.h> 19f46b5a66SChristoph Hellwig #include <linux/compat.h> 20cb8e7090SChristoph Hellwig #include <linux/security.h> 21f46b5a66SChristoph Hellwig #include <linux/xattr.h> 22f54de068SDavid Sterba #include <linux/mm.h> 235a0e3ad6STejun Heo #include <linux/slab.h> 24f7039b1dSLi Dongyang #include <linux/blkdev.h> 258ea05e3aSAlexander Block #include <linux/uuid.h> 2655e301fdSFilipe Brandenburger #include <linux/btrfs.h> 27416161dbSMark Fasheh #include <linux/uaccess.h> 28ae5e165dSJeff Layton #include <linux/iversion.h> 29f46b5a66SChristoph Hellwig #include "ctree.h" 30f46b5a66SChristoph Hellwig #include "disk-io.h" 31f46b5a66SChristoph Hellwig #include "transaction.h" 32f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 33f46b5a66SChristoph Hellwig #include "print-tree.h" 34f46b5a66SChristoph Hellwig #include "volumes.h" 35925baeddSChris Mason #include "locking.h" 36581bb050SLi Zefan #include "inode-map.h" 37d7728c96SJan Schmidt #include "backref.h" 38606686eeSJosef Bacik #include "rcu-string.h" 3931db9f7cSAlexander Block #include "send.h" 403f6bcfbdSStefan Behrens #include "dev-replace.h" 4163541927SFilipe David Borba Manana #include "props.h" 423b02a68aSJeff Mahoney #include "sysfs.h" 43fcebe456SJosef Bacik #include "qgroup.h" 441ec9a1aeSFilipe Manana #include "tree-log.h" 45ebb8765bSAnand Jain #include "compression.h" 46f46b5a66SChristoph Hellwig 47abccd00fSHugo Mills #ifdef CONFIG_64BIT 48abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 49abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 50abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 51abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 52abccd00fSHugo Mills */ 53abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 54abccd00fSHugo Mills __u64 sec; 55abccd00fSHugo Mills __u32 nsec; 56abccd00fSHugo Mills } __attribute__ ((__packed__)); 57abccd00fSHugo Mills 58abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 59abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 60abccd00fSHugo Mills __u64 stransid; /* in */ 61abccd00fSHugo Mills __u64 rtransid; /* out */ 62abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 63abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 64abccd00fSHugo Mills __u64 flags; /* in */ 65abccd00fSHugo Mills __u64 reserved[16]; /* in */ 66abccd00fSHugo Mills } __attribute__ ((__packed__)); 67abccd00fSHugo Mills 68abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 69abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 70abccd00fSHugo Mills #endif 71abccd00fSHugo Mills 722351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 732351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 742351f431SJosef Bacik __s64 send_fd; /* in */ 752351f431SJosef Bacik __u64 clone_sources_count; /* in */ 762351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 772351f431SJosef Bacik __u64 parent_root; /* in */ 782351f431SJosef Bacik __u64 flags; /* in */ 792351f431SJosef Bacik __u64 reserved[4]; /* in */ 802351f431SJosef Bacik } __attribute__ ((__packed__)); 812351f431SJosef Bacik 822351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 832351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 842351f431SJosef Bacik #endif 85abccd00fSHugo Mills 86416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 871c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 881c919a5eSMark Fasheh int no_time_update); 89416161dbSMark Fasheh 906cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 911905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 921905a0f7SDavid Sterba unsigned int flags) 936cbff00fSChristoph Hellwig { 941905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 956cbff00fSChristoph Hellwig return flags; 961905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 976cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 986cbff00fSChristoph Hellwig else 996cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1006cbff00fSChristoph Hellwig } 101f46b5a66SChristoph Hellwig 1026cbff00fSChristoph Hellwig /* 103a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 104a157d4fdSDavid Sterba * ioctl. 1056cbff00fSChristoph Hellwig */ 106a157d4fdSDavid Sterba static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags) 1076cbff00fSChristoph Hellwig { 1086cbff00fSChristoph Hellwig unsigned int iflags = 0; 1096cbff00fSChristoph Hellwig 1106cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1116cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1126cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1136cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1156cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1176cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1196cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1216cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 122d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 123d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 124d0092bddSLi Zefan 12513f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 126d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 12713f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 12813f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1296cbff00fSChristoph Hellwig 1306cbff00fSChristoph Hellwig return iflags; 1316cbff00fSChristoph Hellwig } 1326cbff00fSChristoph Hellwig 1336cbff00fSChristoph Hellwig /* 1346cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1356cbff00fSChristoph Hellwig */ 1367b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1376cbff00fSChristoph Hellwig { 1385c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1393cc79392SFilipe Manana unsigned int new_fl = 0; 1406cbff00fSChristoph Hellwig 1415c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1423cc79392SFilipe Manana new_fl |= S_SYNC; 1435c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1443cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1455c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1463cc79392SFilipe Manana new_fl |= S_APPEND; 1475c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1483cc79392SFilipe Manana new_fl |= S_NOATIME; 1495c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1503cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1513cc79392SFilipe Manana 1523cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1533cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1543cc79392SFilipe Manana new_fl); 1556cbff00fSChristoph Hellwig } 1566cbff00fSChristoph Hellwig 1576cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1586cbff00fSChristoph Hellwig { 1595c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 1605c57b8b6SDavid Sterba unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags); 1616cbff00fSChristoph Hellwig 1626cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1636cbff00fSChristoph Hellwig return -EFAULT; 1646cbff00fSChristoph Hellwig return 0; 1656cbff00fSChristoph Hellwig } 1666cbff00fSChristoph Hellwig 1675ba76abfSDavid Sterba /* Check if @flags are a supported and valid set of FS_*_FL flags */ 1685ba76abfSDavid Sterba static int check_fsflags(unsigned int flags) 16975e7cb7fSLiu Bo { 17075e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17175e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17275e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 173e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 174e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17575e7cb7fSLiu Bo return -EOPNOTSUPP; 17675e7cb7fSLiu Bo 17775e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 17875e7cb7fSLiu Bo return -EINVAL; 17975e7cb7fSLiu Bo 18075e7cb7fSLiu Bo return 0; 18175e7cb7fSLiu Bo } 18275e7cb7fSLiu Bo 1836cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1846cbff00fSChristoph Hellwig { 185496ad9aaSAl Viro struct inode *inode = file_inode(file); 1860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1875c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1885c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 1896cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1905aca2842SDarrick J. Wong unsigned int fsflags, old_fsflags; 1916cbff00fSChristoph Hellwig int ret; 192ff9fef55SAnand Jain const char *comp = NULL; 193d2b8fcfeSAnand Jain u32 binode_flags = binode->flags; 1946cbff00fSChristoph Hellwig 195bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 196bd60ea0fSDavid Sterba return -EPERM; 197bd60ea0fSDavid Sterba 198b83cc969SLi Zefan if (btrfs_root_readonly(root)) 199b83cc969SLi Zefan return -EROFS; 200b83cc969SLi Zefan 2015c57b8b6SDavid Sterba if (copy_from_user(&fsflags, arg, sizeof(fsflags))) 2026cbff00fSChristoph Hellwig return -EFAULT; 2036cbff00fSChristoph Hellwig 2045c57b8b6SDavid Sterba ret = check_fsflags(fsflags); 20575e7cb7fSLiu Bo if (ret) 20675e7cb7fSLiu Bo return ret; 2076cbff00fSChristoph Hellwig 208e7848683SJan Kara ret = mnt_want_write_file(file); 209e7848683SJan Kara if (ret) 210e7848683SJan Kara return ret; 211e7848683SJan Kara 2125955102cSAl Viro inode_lock(inode); 2136cbff00fSChristoph Hellwig 2145c57b8b6SDavid Sterba fsflags = btrfs_mask_fsflags_for_type(inode, fsflags); 2155aca2842SDarrick J. Wong old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags); 2165aca2842SDarrick J. Wong ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags); 2175aca2842SDarrick J. Wong if (ret) 2186cbff00fSChristoph Hellwig goto out_unlock; 2196cbff00fSChristoph Hellwig 2205c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 221d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_SYNC; 2226cbff00fSChristoph Hellwig else 223d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_SYNC; 2245c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 225d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_IMMUTABLE; 2266cbff00fSChristoph Hellwig else 227d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_IMMUTABLE; 2285c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 229d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_APPEND; 2306cbff00fSChristoph Hellwig else 231d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_APPEND; 2325c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 233d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODUMP; 2346cbff00fSChristoph Hellwig else 235d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODUMP; 2365c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 237d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOATIME; 2386cbff00fSChristoph Hellwig else 239d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOATIME; 2405c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 241d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_DIRSYNC; 2426cbff00fSChristoph Hellwig else 243d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_DIRSYNC; 2445c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 24544e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2467e97b8daSDavid Sterba /* 2477e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2487e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2497e97b8daSDavid Sterba * status of the file and will not set it. 2507e97b8daSDavid Sterba */ 2517e97b8daSDavid Sterba if (inode->i_size == 0) 252d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW | 253d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM; 2547e97b8daSDavid Sterba } else { 255d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NODATACOW; 2567e97b8daSDavid Sterba } 2577e97b8daSDavid Sterba } else { 2587e97b8daSDavid Sterba /* 25901327610SNicholas D Steeves * Revert back under same assumptions as above 2607e97b8daSDavid Sterba */ 26144e5194bSAnand Jain if (S_ISREG(inode->i_mode)) { 2627e97b8daSDavid Sterba if (inode->i_size == 0) 263d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_NODATACOW | 264d2b8fcfeSAnand Jain BTRFS_INODE_NODATASUM); 2657e97b8daSDavid Sterba } else { 266d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NODATACOW; 2677e97b8daSDavid Sterba } 2687e97b8daSDavid Sterba } 2696cbff00fSChristoph Hellwig 27075e7cb7fSLiu Bo /* 27175e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 27275e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 27375e7cb7fSLiu Bo * things smaller. 27475e7cb7fSLiu Bo */ 2755c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 276d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_COMPRESS; 277d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_NOCOMPRESS; 2785c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 27963541927SFilipe David Borba Manana 280eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 281eede2bf3SOmar Sandoval ret = -ETXTBSY; 282eede2bf3SOmar Sandoval goto out_unlock; 283eede2bf3SOmar Sandoval } 284eede2bf3SOmar Sandoval 285d2b8fcfeSAnand Jain binode_flags |= BTRFS_INODE_COMPRESS; 286d2b8fcfeSAnand Jain binode_flags &= ~BTRFS_INODE_NOCOMPRESS; 28763541927SFilipe David Borba Manana 28893370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 28993370509SDavid Sterba if (!comp || comp[0] == 0) 29093370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 291ebcb904dSLi Zefan } else { 292d2b8fcfeSAnand Jain binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 29375e7cb7fSLiu Bo } 2946cbff00fSChristoph Hellwig 295ff9fef55SAnand Jain /* 296ff9fef55SAnand Jain * 1 for inode item 297ff9fef55SAnand Jain * 2 for properties 298ff9fef55SAnand Jain */ 299ff9fef55SAnand Jain trans = btrfs_start_transaction(root, 3); 300f062abf0SLi Zefan if (IS_ERR(trans)) { 301f062abf0SLi Zefan ret = PTR_ERR(trans); 302d2b8fcfeSAnand Jain goto out_unlock; 303f062abf0SLi Zefan } 3046cbff00fSChristoph Hellwig 305ff9fef55SAnand Jain if (comp) { 306ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, 307ff9fef55SAnand Jain strlen(comp), 0); 308ff9fef55SAnand Jain if (ret) { 309ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 310ff9fef55SAnand Jain goto out_end_trans; 311ff9fef55SAnand Jain } 312ff9fef55SAnand Jain set_bit(BTRFS_INODE_COPY_EVERYTHING, 313ff9fef55SAnand Jain &BTRFS_I(inode)->runtime_flags); 314ff9fef55SAnand Jain } else { 315ff9fef55SAnand Jain ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 316ff9fef55SAnand Jain 0, 0); 317ff9fef55SAnand Jain if (ret && ret != -ENODATA) { 318ff9fef55SAnand Jain btrfs_abort_transaction(trans, ret); 319ff9fef55SAnand Jain goto out_end_trans; 320ff9fef55SAnand Jain } 321ff9fef55SAnand Jain } 322ff9fef55SAnand Jain 323d2b8fcfeSAnand Jain binode->flags = binode_flags; 3247b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3250c4d2d95SJosef Bacik inode_inc_iversion(inode); 326c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3276cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3286cbff00fSChristoph Hellwig 329ff9fef55SAnand Jain out_end_trans: 3303a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3316cbff00fSChristoph Hellwig out_unlock: 3325955102cSAl Viro inode_unlock(inode); 333e7848683SJan Kara mnt_drop_write_file(file); 3342d4e6f6aSliubo return ret; 3356cbff00fSChristoph Hellwig } 3366cbff00fSChristoph Hellwig 33719f93b3cSDavid Sterba /* 33819f93b3cSDavid Sterba * Translate btrfs internal inode flags to xflags as expected by the 33919f93b3cSDavid Sterba * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are 34019f93b3cSDavid Sterba * silently dropped. 34119f93b3cSDavid Sterba */ 34219f93b3cSDavid Sterba static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags) 34319f93b3cSDavid Sterba { 34419f93b3cSDavid Sterba unsigned int xflags = 0; 34519f93b3cSDavid Sterba 34619f93b3cSDavid Sterba if (flags & BTRFS_INODE_APPEND) 34719f93b3cSDavid Sterba xflags |= FS_XFLAG_APPEND; 34819f93b3cSDavid Sterba if (flags & BTRFS_INODE_IMMUTABLE) 34919f93b3cSDavid Sterba xflags |= FS_XFLAG_IMMUTABLE; 35019f93b3cSDavid Sterba if (flags & BTRFS_INODE_NOATIME) 35119f93b3cSDavid Sterba xflags |= FS_XFLAG_NOATIME; 35219f93b3cSDavid Sterba if (flags & BTRFS_INODE_NODUMP) 35319f93b3cSDavid Sterba xflags |= FS_XFLAG_NODUMP; 35419f93b3cSDavid Sterba if (flags & BTRFS_INODE_SYNC) 35519f93b3cSDavid Sterba xflags |= FS_XFLAG_SYNC; 35619f93b3cSDavid Sterba 35719f93b3cSDavid Sterba return xflags; 35819f93b3cSDavid Sterba } 35919f93b3cSDavid Sterba 36019f93b3cSDavid Sterba /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */ 36119f93b3cSDavid Sterba static int check_xflags(unsigned int flags) 36219f93b3cSDavid Sterba { 36319f93b3cSDavid Sterba if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME | 36419f93b3cSDavid Sterba FS_XFLAG_NODUMP | FS_XFLAG_SYNC)) 36519f93b3cSDavid Sterba return -EOPNOTSUPP; 36619f93b3cSDavid Sterba return 0; 36719f93b3cSDavid Sterba } 36819f93b3cSDavid Sterba 369e4202ac9SDavid Sterba /* 370e4202ac9SDavid Sterba * Set the xflags from the internal inode flags. The remaining items of fsxattr 371e4202ac9SDavid Sterba * are zeroed. 372e4202ac9SDavid Sterba */ 373e4202ac9SDavid Sterba static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg) 374e4202ac9SDavid Sterba { 375e4202ac9SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 376e4202ac9SDavid Sterba struct fsxattr fa; 377e4202ac9SDavid Sterba 378e4202ac9SDavid Sterba memset(&fa, 0, sizeof(fa)); 379e4202ac9SDavid Sterba fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags); 380e4202ac9SDavid Sterba 381e4202ac9SDavid Sterba if (copy_to_user(arg, &fa, sizeof(fa))) 382e4202ac9SDavid Sterba return -EFAULT; 383e4202ac9SDavid Sterba 384e4202ac9SDavid Sterba return 0; 385e4202ac9SDavid Sterba } 386e4202ac9SDavid Sterba 387025f2121SDavid Sterba static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg) 388025f2121SDavid Sterba { 389025f2121SDavid Sterba struct inode *inode = file_inode(file); 390025f2121SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 391025f2121SDavid Sterba struct btrfs_root *root = binode->root; 392025f2121SDavid Sterba struct btrfs_trans_handle *trans; 393025f2121SDavid Sterba struct fsxattr fa; 394025f2121SDavid Sterba unsigned old_flags; 395025f2121SDavid Sterba unsigned old_i_flags; 396025f2121SDavid Sterba int ret = 0; 397025f2121SDavid Sterba 398025f2121SDavid Sterba if (!inode_owner_or_capable(inode)) 399025f2121SDavid Sterba return -EPERM; 400025f2121SDavid Sterba 401025f2121SDavid Sterba if (btrfs_root_readonly(root)) 402025f2121SDavid Sterba return -EROFS; 403025f2121SDavid Sterba 404025f2121SDavid Sterba memset(&fa, 0, sizeof(fa)); 405025f2121SDavid Sterba if (copy_from_user(&fa, arg, sizeof(fa))) 406025f2121SDavid Sterba return -EFAULT; 407025f2121SDavid Sterba 408025f2121SDavid Sterba ret = check_xflags(fa.fsx_xflags); 409025f2121SDavid Sterba if (ret) 410025f2121SDavid Sterba return ret; 411025f2121SDavid Sterba 412025f2121SDavid Sterba if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0) 413025f2121SDavid Sterba return -EOPNOTSUPP; 414025f2121SDavid Sterba 415025f2121SDavid Sterba ret = mnt_want_write_file(file); 416025f2121SDavid Sterba if (ret) 417025f2121SDavid Sterba return ret; 418025f2121SDavid Sterba 419025f2121SDavid Sterba inode_lock(inode); 420025f2121SDavid Sterba 421025f2121SDavid Sterba old_flags = binode->flags; 422025f2121SDavid Sterba old_i_flags = inode->i_flags; 423025f2121SDavid Sterba 424025f2121SDavid Sterba /* We need the capabilities to change append-only or immutable inode */ 425025f2121SDavid Sterba if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) || 426025f2121SDavid Sterba (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) && 427025f2121SDavid Sterba !capable(CAP_LINUX_IMMUTABLE)) { 428025f2121SDavid Sterba ret = -EPERM; 429025f2121SDavid Sterba goto out_unlock; 430025f2121SDavid Sterba } 431025f2121SDavid Sterba 432025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_SYNC) 433025f2121SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 434025f2121SDavid Sterba else 435025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 436025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE) 437025f2121SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 438025f2121SDavid Sterba else 439025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 440025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_APPEND) 441025f2121SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 442025f2121SDavid Sterba else 443025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 444025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NODUMP) 445025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 446025f2121SDavid Sterba else 447025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 448025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NOATIME) 449025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 450025f2121SDavid Sterba else 451025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 452025f2121SDavid Sterba 453025f2121SDavid Sterba /* 1 item for the inode */ 454025f2121SDavid Sterba trans = btrfs_start_transaction(root, 1); 455025f2121SDavid Sterba if (IS_ERR(trans)) { 456025f2121SDavid Sterba ret = PTR_ERR(trans); 457025f2121SDavid Sterba goto out_unlock; 458025f2121SDavid Sterba } 459025f2121SDavid Sterba 460025f2121SDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 461025f2121SDavid Sterba inode_inc_iversion(inode); 462025f2121SDavid Sterba inode->i_ctime = current_time(inode); 463025f2121SDavid Sterba ret = btrfs_update_inode(trans, root, inode); 464025f2121SDavid Sterba 465025f2121SDavid Sterba btrfs_end_transaction(trans); 466025f2121SDavid Sterba 467025f2121SDavid Sterba out_unlock: 468025f2121SDavid Sterba if (ret) { 469025f2121SDavid Sterba binode->flags = old_flags; 470025f2121SDavid Sterba inode->i_flags = old_i_flags; 471025f2121SDavid Sterba } 472025f2121SDavid Sterba 473025f2121SDavid Sterba inode_unlock(inode); 474025f2121SDavid Sterba mnt_drop_write_file(file); 475025f2121SDavid Sterba 476025f2121SDavid Sterba return ret; 477025f2121SDavid Sterba } 478025f2121SDavid Sterba 4796cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4806cbff00fSChristoph Hellwig { 481496ad9aaSAl Viro struct inode *inode = file_inode(file); 4826cbff00fSChristoph Hellwig 4836cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4846cbff00fSChristoph Hellwig } 485f46b5a66SChristoph Hellwig 486f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 487f7039b1dSLi Dongyang { 4880b246afaSJeff Mahoney struct inode *inode = file_inode(file); 4890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 490f7039b1dSLi Dongyang struct btrfs_device *device; 491f7039b1dSLi Dongyang struct request_queue *q; 492f7039b1dSLi Dongyang struct fstrim_range range; 493f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 494f7039b1dSLi Dongyang u64 num_devices = 0; 495f7039b1dSLi Dongyang int ret; 496f7039b1dSLi Dongyang 497f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 498f7039b1dSLi Dongyang return -EPERM; 499f7039b1dSLi Dongyang 500f35f06c3SFilipe Manana /* 501f35f06c3SFilipe Manana * If the fs is mounted with nologreplay, which requires it to be 502f35f06c3SFilipe Manana * mounted in RO mode as well, we can not allow discard on free space 503f35f06c3SFilipe Manana * inside block groups, because log trees refer to extents that are not 504f35f06c3SFilipe Manana * pinned in a block group's free space cache (pinning the extents is 505f35f06c3SFilipe Manana * precisely the first phase of replaying a log tree). 506f35f06c3SFilipe Manana */ 507f35f06c3SFilipe Manana if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 508f35f06c3SFilipe Manana return -EROFS; 509f35f06c3SFilipe Manana 5101f78160cSXiao Guangrong rcu_read_lock(); 5111f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 5121f78160cSXiao Guangrong dev_list) { 513f7039b1dSLi Dongyang if (!device->bdev) 514f7039b1dSLi Dongyang continue; 515f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 516f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 517f7039b1dSLi Dongyang num_devices++; 51850d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 519f7039b1dSLi Dongyang minlen); 520f7039b1dSLi Dongyang } 521f7039b1dSLi Dongyang } 5221f78160cSXiao Guangrong rcu_read_unlock(); 523f4c697e6SLukas Czerner 524f7039b1dSLi Dongyang if (!num_devices) 525f7039b1dSLi Dongyang return -EOPNOTSUPP; 526f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 527f7039b1dSLi Dongyang return -EFAULT; 5286ba9fc8eSQu Wenruo 5296ba9fc8eSQu Wenruo /* 5306ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 5316ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 5326ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 5336ba9fc8eSQu Wenruo */ 5346ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 535f4c697e6SLukas Czerner return -EINVAL; 536f7039b1dSLi Dongyang 537f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5382ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 539f7039b1dSLi Dongyang if (ret < 0) 540f7039b1dSLi Dongyang return ret; 541f7039b1dSLi Dongyang 542f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 543f7039b1dSLi Dongyang return -EFAULT; 544f7039b1dSLi Dongyang 545f7039b1dSLi Dongyang return 0; 546f7039b1dSLi Dongyang } 547f7039b1dSLi Dongyang 548dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 549dd5f9615SStefan Behrens { 55046e0f66aSChris Mason int i; 55146e0f66aSChris Mason 55246e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 55346e0f66aSChris Mason if (uuid[i]) 55446e0f66aSChris Mason return 0; 55546e0f66aSChris Mason } 55646e0f66aSChris Mason return 1; 557dd5f9615SStefan Behrens } 558dd5f9615SStefan Behrens 559d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 560cb8e7090SChristoph Hellwig struct dentry *dentry, 56152f75f4fSDavid Sterba const char *name, int namelen, 5626f72c7e2SArne Jansen u64 *async_transid, 5638696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 564f46b5a66SChristoph Hellwig { 5650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 566f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 567f46b5a66SChristoph Hellwig struct btrfs_key key; 56849a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 569f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 570f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 571d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 57276dda93cSYan, Zheng struct btrfs_root *new_root; 573d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 57495582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5755662344bSTsutomu Itoh struct inode *inode; 576f46b5a66SChristoph Hellwig int ret; 577f46b5a66SChristoph Hellwig int err; 578f46b5a66SChristoph Hellwig u64 objectid; 579f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 5803de4586cSChris Mason u64 index = 0; 5818ea05e3aSAlexander Block uuid_le new_uuid; 582f46b5a66SChristoph Hellwig 58349a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 58449a3c4d9SDavid Sterba if (!root_item) 58549a3c4d9SDavid Sterba return -ENOMEM; 58649a3c4d9SDavid Sterba 5870b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5882fbe8c8aSAl Viro if (ret) 58949a3c4d9SDavid Sterba goto fail_free; 5906a912213SJosef Bacik 591e09fe2d2SQu Wenruo /* 592e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 59301327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 594e09fe2d2SQu Wenruo */ 59549a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 59649a3c4d9SDavid Sterba ret = -ENOSPC; 59749a3c4d9SDavid Sterba goto fail_free; 59849a3c4d9SDavid Sterba } 599e09fe2d2SQu Wenruo 600d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 6019ed74f2dSJosef Bacik /* 602d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 603d5c12070SMiao Xie * of create_snapshot(). 6049ed74f2dSJosef Bacik */ 605c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 606d5c12070SMiao Xie if (ret) 60749a3c4d9SDavid Sterba goto fail_free; 608f46b5a66SChristoph Hellwig 609d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 610d5c12070SMiao Xie if (IS_ERR(trans)) { 611d5c12070SMiao Xie ret = PTR_ERR(trans); 6127775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 61349a3c4d9SDavid Sterba goto fail_free; 614d5c12070SMiao Xie } 615d5c12070SMiao Xie trans->block_rsv = &block_rsv; 616d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 617f46b5a66SChristoph Hellwig 618a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 6196f72c7e2SArne Jansen if (ret) 6206f72c7e2SArne Jansen goto fail; 6216f72c7e2SArne Jansen 6224d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6238e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6248e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6258e8a1e31SJosef Bacik goto fail; 6268e8a1e31SJosef Bacik } 627f46b5a66SChristoph Hellwig 628f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 629f46b5a66SChristoph Hellwig 63049a3c4d9SDavid Sterba inode_item = &root_item->inode; 6313cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6323cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6333cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 634da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6350b246afaSJeff Mahoney fs_info->nodesize); 6363cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 637f46b5a66SChristoph Hellwig 63849a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 63949a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6403cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 64108fe4db1SLi Zefan 64249a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 64349a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 64449a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 64549a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 64649a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 64749a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 648f46b5a66SChristoph Hellwig 64949a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 65049a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6518ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 65249a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 65349a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 65449a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 65549a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 65649a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 65749a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 658f46b5a66SChristoph Hellwig 659925baeddSChris Mason btrfs_tree_unlock(leaf); 660f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 661f46b5a66SChristoph Hellwig leaf = NULL; 662f46b5a66SChristoph Hellwig 66349a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 664f46b5a66SChristoph Hellwig 665f46b5a66SChristoph Hellwig key.objectid = objectid; 6665d4f98a2SYan Zheng key.offset = 0; 667962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6680b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 66949a3c4d9SDavid Sterba root_item); 670f46b5a66SChristoph Hellwig if (ret) 671f46b5a66SChristoph Hellwig goto fail; 672f46b5a66SChristoph Hellwig 67376dda93cSYan, Zheng key.offset = (u64)-1; 6740b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 67579787eaaSJeff Mahoney if (IS_ERR(new_root)) { 67679787eaaSJeff Mahoney ret = PTR_ERR(new_root); 67766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 67879787eaaSJeff Mahoney goto fail; 67979787eaaSJeff Mahoney } 68076dda93cSYan, Zheng 68176dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 68276dda93cSYan, Zheng 68363541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 684ce598979SMark Fasheh if (ret) { 685ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 68666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 687ce598979SMark Fasheh goto fail; 688ce598979SMark Fasheh } 689ce598979SMark Fasheh 690f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 691f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 692f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 693f32e48e9SChandan Rajendra 694f46b5a66SChristoph Hellwig /* 695f46b5a66SChristoph Hellwig * insert the directory item 696f46b5a66SChristoph Hellwig */ 697877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 69879787eaaSJeff Mahoney if (ret) { 69966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 70079787eaaSJeff Mahoney goto fail; 70179787eaaSJeff Mahoney } 7023de4586cSChris Mason 703684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 7043de4586cSChris Mason BTRFS_FT_DIR, index); 70579787eaaSJeff Mahoney if (ret) { 70666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 707f46b5a66SChristoph Hellwig goto fail; 70879787eaaSJeff Mahoney } 7090660b5afSChris Mason 7106ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 71152c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 71252c26179SYan Zheng BUG_ON(ret); 71352c26179SYan Zheng 7146025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7154a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7160660b5afSChris Mason BUG_ON(ret); 7170660b5afSChris Mason 718cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7196bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 720dd5f9615SStefan Behrens if (ret) 72166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 722dd5f9615SStefan Behrens 723f46b5a66SChristoph Hellwig fail: 72449a3c4d9SDavid Sterba kfree(root_item); 725d5c12070SMiao Xie trans->block_rsv = NULL; 726d5c12070SMiao Xie trans->bytes_reserved = 0; 7277775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 728de6e8200SLiu Bo 72972fd032eSSage Weil if (async_transid) { 73072fd032eSSage Weil *async_transid = trans->transid; 7313a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 73200d71c9cSMiao Xie if (err) 7333a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73472fd032eSSage Weil } else { 7353a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73672fd032eSSage Weil } 737f46b5a66SChristoph Hellwig if (err && !ret) 738f46b5a66SChristoph Hellwig ret = err; 7391a65e24bSChris Mason 7405662344bSTsutomu Itoh if (!ret) { 7415662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 742de6e8200SLiu Bo if (IS_ERR(inode)) 743de6e8200SLiu Bo return PTR_ERR(inode); 7445662344bSTsutomu Itoh d_instantiate(dentry, inode); 7455662344bSTsutomu Itoh } 746f46b5a66SChristoph Hellwig return ret; 74749a3c4d9SDavid Sterba 74849a3c4d9SDavid Sterba fail_free: 74949a3c4d9SDavid Sterba kfree(root_item); 75049a3c4d9SDavid Sterba return ret; 751f46b5a66SChristoph Hellwig } 752f46b5a66SChristoph Hellwig 753e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 75461d7e4cbSDavid Sterba struct dentry *dentry, 755e9662f70SMiao Xie u64 *async_transid, bool readonly, 756e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 757f46b5a66SChristoph Hellwig { 7580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7592e4bfab9SYan, Zheng struct inode *inode; 760f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 761f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7622e4bfab9SYan, Zheng int ret; 7638ecebf4dSRobbie Ko bool snapshot_force_cow = false; 764f46b5a66SChristoph Hellwig 76527cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 766f46b5a66SChristoph Hellwig return -EINVAL; 767f46b5a66SChristoph Hellwig 768eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 769eede2bf3SOmar Sandoval btrfs_warn(fs_info, 770eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 771eede2bf3SOmar Sandoval return -ETXTBSY; 772eede2bf3SOmar Sandoval } 773eede2bf3SOmar Sandoval 77423269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 775a1ee7362SDavid Sterba if (!pending_snapshot) 776a1ee7362SDavid Sterba return -ENOMEM; 777a1ee7362SDavid Sterba 778b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 77923269bf5SDavid Sterba GFP_KERNEL); 7808546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7818546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 782b0c0ea63SDavid Sterba ret = -ENOMEM; 783b0c0ea63SDavid Sterba goto free_pending; 784b0c0ea63SDavid Sterba } 785b0c0ea63SDavid Sterba 7868ecebf4dSRobbie Ko /* 7878ecebf4dSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 7888ecebf4dSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 7898ecebf4dSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 7908ecebf4dSRobbie Ko */ 791ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7924e857c58SPeter Zijlstra smp_mb__after_atomic(); 79345bac0f3SLiu Bo /* wait for no snapshot writes */ 79445bac0f3SLiu Bo wait_event(root->subv_writers->wait, 79545bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 7968257b2dcSMiao Xie 7973cd24c69SEthan Lien ret = btrfs_start_delalloc_snapshot(root); 7986a03843dSMiao Xie if (ret) 799a1ee7362SDavid Sterba goto dec_and_free; 8006a03843dSMiao Xie 8018ecebf4dSRobbie Ko /* 8028ecebf4dSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 8038ecebf4dSRobbie Ko * we force future writes to fallback to COW mode during snapshot 8048ecebf4dSRobbie Ko * creation. 8058ecebf4dSRobbie Ko */ 8068ecebf4dSRobbie Ko atomic_inc(&root->snapshot_force_cow); 8078ecebf4dSRobbie Ko snapshot_force_cow = true; 8088ecebf4dSRobbie Ko 8096374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 8106a03843dSMiao Xie 81166d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 81266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 813d5c12070SMiao Xie /* 814d5c12070SMiao Xie * 1 - parent dir inode 815d5c12070SMiao Xie * 2 - dir entries 816d5c12070SMiao Xie * 1 - root item 817d5c12070SMiao Xie * 2 - root ref/backref 818d5c12070SMiao Xie * 1 - root of snapshot 819dd5f9615SStefan Behrens * 1 - UUID item 820d5c12070SMiao Xie */ 821d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 822dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 823ee3441b4SJeff Mahoney false); 824d5c12070SMiao Xie if (ret) 825a1ee7362SDavid Sterba goto dec_and_free; 826d5c12070SMiao Xie 827a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 828a22285a6SYan, Zheng pending_snapshot->root = root; 829b83cc969SLi Zefan pending_snapshot->readonly = readonly; 830e9662f70SMiao Xie pending_snapshot->dir = dir; 8318696c533SMiao Xie pending_snapshot->inherit = inherit; 832a22285a6SYan, Zheng 833d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 834a22285a6SYan, Zheng if (IS_ERR(trans)) { 835a22285a6SYan, Zheng ret = PTR_ERR(trans); 836a22285a6SYan, Zheng goto fail; 837a22285a6SYan, Zheng } 838a22285a6SYan, Zheng 8390b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 840a22285a6SYan, Zheng list_add(&pending_snapshot->list, 841a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8420b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 84372fd032eSSage Weil if (async_transid) { 84472fd032eSSage Weil *async_transid = trans->transid; 8453a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 84600d71c9cSMiao Xie if (ret) 8473a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 84872fd032eSSage Weil } else { 8493a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 85072fd032eSSage Weil } 851aec8030aSMiao Xie if (ret) 852c37b2b62SJosef Bacik goto fail; 853a22285a6SYan, Zheng 854a22285a6SYan, Zheng ret = pending_snapshot->error; 855f46b5a66SChristoph Hellwig if (ret) 8562e4bfab9SYan, Zheng goto fail; 857f46b5a66SChristoph Hellwig 858d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 859d3797308SChris Mason if (ret) 860d3797308SChris Mason goto fail; 861d3797308SChris Mason 8622b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8632e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8642e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8652e4bfab9SYan, Zheng goto fail; 8662e4bfab9SYan, Zheng } 8675662344bSTsutomu Itoh 8682e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8692e4bfab9SYan, Zheng ret = 0; 8702e4bfab9SYan, Zheng fail: 8717775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 872a1ee7362SDavid Sterba dec_and_free: 8738ecebf4dSRobbie Ko if (snapshot_force_cow) 8748ecebf4dSRobbie Ko atomic_dec(&root->snapshot_force_cow); 875ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8764625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 877b0c0ea63SDavid Sterba free_pending: 878b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8798546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 880a1ee7362SDavid Sterba kfree(pending_snapshot); 881a1ee7362SDavid Sterba 882f46b5a66SChristoph Hellwig return ret; 883f46b5a66SChristoph Hellwig } 884f46b5a66SChristoph Hellwig 8854260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8864260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8874260f7c7SSage Weil * whether the type of victim is right. 8884260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8894260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8904260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8914260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8924260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8934260f7c7SSage Weil * a. be owner of dir, or 8944260f7c7SSage Weil * b. be owner of victim, or 8954260f7c7SSage Weil * c. have CAP_FOWNER capability 89601327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8974260f7c7SSage Weil * links pointing to it. 8984260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8994260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 9004260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 9014260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 9024260f7c7SSage Weil * nfs_async_unlink(). 9034260f7c7SSage Weil */ 9044260f7c7SSage Weil 9054260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 9064260f7c7SSage Weil { 9074260f7c7SSage Weil int error; 9084260f7c7SSage Weil 9092b0143b5SDavid Howells if (d_really_is_negative(victim)) 9104260f7c7SSage Weil return -ENOENT; 9114260f7c7SSage Weil 9122b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 9134fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 9144260f7c7SSage Weil 9154260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 9164260f7c7SSage Weil if (error) 9174260f7c7SSage Weil return error; 9184260f7c7SSage Weil if (IS_APPEND(dir)) 9194260f7c7SSage Weil return -EPERM; 9202b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9212b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9224260f7c7SSage Weil return -EPERM; 9234260f7c7SSage Weil if (isdir) { 924e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9254260f7c7SSage Weil return -ENOTDIR; 9264260f7c7SSage Weil if (IS_ROOT(victim)) 9274260f7c7SSage Weil return -EBUSY; 928e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9294260f7c7SSage Weil return -EISDIR; 9304260f7c7SSage Weil if (IS_DEADDIR(dir)) 9314260f7c7SSage Weil return -ENOENT; 9324260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9334260f7c7SSage Weil return -EBUSY; 9344260f7c7SSage Weil return 0; 9354260f7c7SSage Weil } 9364260f7c7SSage Weil 937cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 938cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 939cb8e7090SChristoph Hellwig { 9402b0143b5SDavid Howells if (d_really_is_positive(child)) 941cb8e7090SChristoph Hellwig return -EEXIST; 942cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 943cb8e7090SChristoph Hellwig return -ENOENT; 944cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 945cb8e7090SChristoph Hellwig } 946cb8e7090SChristoph Hellwig 947cb8e7090SChristoph Hellwig /* 948cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 949cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 950cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 951cb8e7090SChristoph Hellwig */ 95292872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 95352f75f4fSDavid Sterba const char *name, int namelen, 95472fd032eSSage Weil struct btrfs_root *snap_src, 9556f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9568696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 957cb8e7090SChristoph Hellwig { 9582b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 960cb8e7090SChristoph Hellwig struct dentry *dentry; 961cb8e7090SChristoph Hellwig int error; 962cb8e7090SChristoph Hellwig 96300235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 96400235411SAl Viro if (error == -EINTR) 96500235411SAl Viro return error; 966cb8e7090SChristoph Hellwig 967cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 968cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 969cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 970cb8e7090SChristoph Hellwig goto out_unlock; 971cb8e7090SChristoph Hellwig 97276dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 973cb8e7090SChristoph Hellwig if (error) 974a874a63eSLiu Bo goto out_dput; 975cb8e7090SChristoph Hellwig 9769c52057cSChris Mason /* 9779c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9789c52057cSChris Mason * check for them now when we can safely fail 9799c52057cSChris Mason */ 9809c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9819c52057cSChris Mason dir->i_ino, name, 9829c52057cSChris Mason namelen); 9839c52057cSChris Mason if (error) 9849c52057cSChris Mason goto out_dput; 9859c52057cSChris Mason 9860b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 98776dda93cSYan, Zheng 98876dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 98976dda93cSYan, Zheng goto out_up_read; 99076dda93cSYan, Zheng 9913de4586cSChris Mason if (snap_src) { 99261d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9936f72c7e2SArne Jansen async_transid, readonly, inherit); 9943de4586cSChris Mason } else { 995d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 996d5c12070SMiao Xie async_transid, inherit); 9973de4586cSChris Mason } 99876dda93cSYan, Zheng if (!error) 99976dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 100076dda93cSYan, Zheng out_up_read: 10010b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 1002cb8e7090SChristoph Hellwig out_dput: 1003cb8e7090SChristoph Hellwig dput(dentry); 1004cb8e7090SChristoph Hellwig out_unlock: 10055955102cSAl Viro inode_unlock(dir); 1006cb8e7090SChristoph Hellwig return error; 1007cb8e7090SChristoph Hellwig } 1008cb8e7090SChristoph Hellwig 10094cb5300bSChris Mason /* 10104cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 10114cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 10124cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 10134cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 10144cb5300bSChris Mason * part of the file 10154cb5300bSChris Mason */ 1016aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 10174cb5300bSChris Mason { 10184cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10194cb5300bSChris Mason struct extent_map *em = NULL; 10204cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10214cb5300bSChris Mason u64 end; 10224cb5300bSChris Mason 10234cb5300bSChris Mason read_lock(&em_tree->lock); 102409cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10254cb5300bSChris Mason read_unlock(&em_tree->lock); 10264cb5300bSChris Mason 10274cb5300bSChris Mason if (em) { 10284cb5300bSChris Mason end = extent_map_end(em); 10294cb5300bSChris Mason free_extent_map(em); 10304cb5300bSChris Mason if (end - offset > thresh) 10314cb5300bSChris Mason return 0; 10324cb5300bSChris Mason } 10334cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10344cb5300bSChris Mason thresh /= 2; 10354cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10364cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10374cb5300bSChris Mason if (end >= thresh) 10384cb5300bSChris Mason return 0; 10394cb5300bSChris Mason return 1; 10404cb5300bSChris Mason } 10414cb5300bSChris Mason 10424cb5300bSChris Mason /* 10434cb5300bSChris Mason * helper function to walk through a file and find extents 10444cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10454cb5300bSChris Mason * 10464cb5300bSChris Mason * This is used by the defragging code to find new and small 10474cb5300bSChris Mason * extents 10484cb5300bSChris Mason */ 10494cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10504cb5300bSChris Mason struct inode *inode, u64 newer_than, 1051aab110abSDavid Sterba u64 *off, u32 thresh) 10524cb5300bSChris Mason { 10534cb5300bSChris Mason struct btrfs_path *path; 10544cb5300bSChris Mason struct btrfs_key min_key; 10554cb5300bSChris Mason struct extent_buffer *leaf; 10564cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10574cb5300bSChris Mason int type; 10584cb5300bSChris Mason int ret; 10594a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10604cb5300bSChris Mason 10614cb5300bSChris Mason path = btrfs_alloc_path(); 10624cb5300bSChris Mason if (!path) 10634cb5300bSChris Mason return -ENOMEM; 10644cb5300bSChris Mason 1065a4689d2bSDavid Sterba min_key.objectid = ino; 10664cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10674cb5300bSChris Mason min_key.offset = *off; 10684cb5300bSChris Mason 10694cb5300bSChris Mason while (1) { 10706174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10714cb5300bSChris Mason if (ret != 0) 10724cb5300bSChris Mason goto none; 1073f094c9bdSFilipe Manana process_slot: 1074a4689d2bSDavid Sterba if (min_key.objectid != ino) 10754cb5300bSChris Mason goto none; 10764cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10774cb5300bSChris Mason goto none; 10784cb5300bSChris Mason 10794cb5300bSChris Mason leaf = path->nodes[0]; 10804cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10814cb5300bSChris Mason struct btrfs_file_extent_item); 10824cb5300bSChris Mason 10834cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10844cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10854cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10864cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10874cb5300bSChris Mason *off = min_key.offset; 10884cb5300bSChris Mason btrfs_free_path(path); 10894cb5300bSChris Mason return 0; 10904cb5300bSChris Mason } 10914cb5300bSChris Mason 1092f094c9bdSFilipe Manana path->slots[0]++; 1093f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1094f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1095f094c9bdSFilipe Manana goto process_slot; 1096f094c9bdSFilipe Manana } 1097f094c9bdSFilipe Manana 10984cb5300bSChris Mason if (min_key.offset == (u64)-1) 10994cb5300bSChris Mason goto none; 11004cb5300bSChris Mason 11014cb5300bSChris Mason min_key.offset++; 11024cb5300bSChris Mason btrfs_release_path(path); 11034cb5300bSChris Mason } 11044cb5300bSChris Mason none: 11054cb5300bSChris Mason btrfs_free_path(path); 11064cb5300bSChris Mason return -ENOENT; 11074cb5300bSChris Mason } 11084cb5300bSChris Mason 11096c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 111017ce6ef8SLiu Bo { 111117ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1112940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 11136c282eb4SLi Zefan struct extent_map *em; 111409cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1115940100a4SChris Mason 1116940100a4SChris Mason /* 1117940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1118940100a4SChris Mason * the full extent lock 1119940100a4SChris Mason */ 1120940100a4SChris Mason read_lock(&em_tree->lock); 1121940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1122940100a4SChris Mason read_unlock(&em_tree->lock); 1123940100a4SChris Mason 1124940100a4SChris Mason if (!em) { 1125308d9800SFilipe Manana struct extent_state *cached = NULL; 1126308d9800SFilipe Manana u64 end = start + len - 1; 1127308d9800SFilipe Manana 1128940100a4SChris Mason /* get the big lock and read metadata off disk */ 1129ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1130fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1131e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1132940100a4SChris Mason 11336cf8bfbfSDan Carpenter if (IS_ERR(em)) 11346c282eb4SLi Zefan return NULL; 1135940100a4SChris Mason } 1136940100a4SChris Mason 11376c282eb4SLi Zefan return em; 11386c282eb4SLi Zefan } 11396c282eb4SLi Zefan 11406c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11416c282eb4SLi Zefan { 11426c282eb4SLi Zefan struct extent_map *next; 11436c282eb4SLi Zefan bool ret = true; 11446c282eb4SLi Zefan 11456c282eb4SLi Zefan /* this is the last extent */ 11466c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11476c282eb4SLi Zefan return false; 11486c282eb4SLi Zefan 11496c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1150e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1151e9512d72SChris Mason ret = false; 1152e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1153ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11546c282eb4SLi Zefan ret = false; 11556c282eb4SLi Zefan 11566c282eb4SLi Zefan free_extent_map(next); 11576c282eb4SLi Zefan return ret; 11586c282eb4SLi Zefan } 11596c282eb4SLi Zefan 1160aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1161a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1162a43a2111SAndrew Mahone int compress) 11636c282eb4SLi Zefan { 11646c282eb4SLi Zefan struct extent_map *em; 11656c282eb4SLi Zefan int ret = 1; 11666c282eb4SLi Zefan bool next_mergeable = true; 11674a3560c4SLiu Bo bool prev_mergeable = true; 11686c282eb4SLi Zefan 11696c282eb4SLi Zefan /* 11706c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11716c282eb4SLi Zefan * defragging it 11726c282eb4SLi Zefan */ 11736c282eb4SLi Zefan if (start < *defrag_end) 11746c282eb4SLi Zefan return 1; 11756c282eb4SLi Zefan 11766c282eb4SLi Zefan *skip = 0; 11776c282eb4SLi Zefan 11786c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11796c282eb4SLi Zefan if (!em) 11806c282eb4SLi Zefan return 0; 11816c282eb4SLi Zefan 1182940100a4SChris Mason /* this will cover holes, and inline extents */ 118317ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1184940100a4SChris Mason ret = 0; 118517ce6ef8SLiu Bo goto out; 118617ce6ef8SLiu Bo } 118717ce6ef8SLiu Bo 11884a3560c4SLiu Bo if (!*defrag_end) 11894a3560c4SLiu Bo prev_mergeable = false; 11904a3560c4SLiu Bo 11916c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1192940100a4SChris Mason /* 11936c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11946c282eb4SLi Zefan * real extent, don't bother defragging it 1195940100a4SChris Mason */ 1196a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11974a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1198940100a4SChris Mason ret = 0; 119917ce6ef8SLiu Bo out: 1200940100a4SChris Mason /* 1201940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1202940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1203940100a4SChris Mason * so that the next tiny extent will force a defrag. 1204940100a4SChris Mason * 1205940100a4SChris Mason * The end result of this is that tiny extents before a single big 1206940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1207940100a4SChris Mason */ 1208940100a4SChris Mason if (ret) { 1209940100a4SChris Mason *defrag_end = extent_map_end(em); 1210940100a4SChris Mason } else { 1211940100a4SChris Mason *last_len = 0; 1212940100a4SChris Mason *skip = extent_map_end(em); 1213940100a4SChris Mason *defrag_end = 0; 1214940100a4SChris Mason } 1215940100a4SChris Mason 1216940100a4SChris Mason free_extent_map(em); 1217940100a4SChris Mason return ret; 1218940100a4SChris Mason } 1219940100a4SChris Mason 12204cb5300bSChris Mason /* 12214cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12224cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12234cb5300bSChris Mason * to COW and defrag. 12244cb5300bSChris Mason * 12254cb5300bSChris Mason * It also makes sure the delalloc code has enough 12264cb5300bSChris Mason * dirty data to avoid making new small extents as part 12274cb5300bSChris Mason * of the defrag 12284cb5300bSChris Mason * 12294cb5300bSChris Mason * It's a good idea to start RA on this range 12304cb5300bSChris Mason * before calling this. 12314cb5300bSChris Mason */ 12324cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12334cb5300bSChris Mason struct page **pages, 12344cb5300bSChris Mason unsigned long start_index, 1235c41570c9SJustin Maggard unsigned long num_pages) 1236f46b5a66SChristoph Hellwig { 12374cb5300bSChris Mason unsigned long file_end; 12384cb5300bSChris Mason u64 isize = i_size_read(inode); 1239f46b5a66SChristoph Hellwig u64 page_start; 1240f46b5a66SChristoph Hellwig u64 page_end; 12411f12bd06SLiu Bo u64 page_cnt; 12424cb5300bSChris Mason int ret; 12434cb5300bSChris Mason int i; 12444cb5300bSChris Mason int i_done; 12454cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12464cb5300bSChris Mason struct extent_state *cached_state = NULL; 1247600a45e1SMiao Xie struct extent_io_tree *tree; 1248364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12493b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12504cb5300bSChris Mason 125109cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12521f12bd06SLiu Bo if (!isize || start_index > file_end) 12531f12bd06SLiu Bo return 0; 12541f12bd06SLiu Bo 12551f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12564cb5300bSChris Mason 1257364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 125809cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 125909cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12604cb5300bSChris Mason if (ret) 12614cb5300bSChris Mason return ret; 12624cb5300bSChris Mason i_done = 0; 1263600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12644cb5300bSChris Mason 12654cb5300bSChris Mason /* step one, lock all the pages */ 12661f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12674cb5300bSChris Mason struct page *page; 1268600a45e1SMiao Xie again: 1269a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12703b16a4e3SJosef Bacik start_index + i, mask); 12714cb5300bSChris Mason if (!page) 12724cb5300bSChris Mason break; 12734cb5300bSChris Mason 1274600a45e1SMiao Xie page_start = page_offset(page); 127509cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1276600a45e1SMiao Xie while (1) { 1277308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1278ff13db41SDavid Sterba &cached_state); 1279600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1280600a45e1SMiao Xie page_start); 1281308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1282e43bbe5eSDavid Sterba &cached_state); 1283600a45e1SMiao Xie if (!ordered) 1284600a45e1SMiao Xie break; 1285600a45e1SMiao Xie 1286600a45e1SMiao Xie unlock_page(page); 1287600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1288600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1289600a45e1SMiao Xie lock_page(page); 12901f12bd06SLiu Bo /* 12911f12bd06SLiu Bo * we unlocked the page above, so we need check if 12921f12bd06SLiu Bo * it was released or not. 12931f12bd06SLiu Bo */ 12941f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12951f12bd06SLiu Bo unlock_page(page); 129609cbfeafSKirill A. Shutemov put_page(page); 12971f12bd06SLiu Bo goto again; 12981f12bd06SLiu Bo } 1299600a45e1SMiao Xie } 1300600a45e1SMiao Xie 13014cb5300bSChris Mason if (!PageUptodate(page)) { 13024cb5300bSChris Mason btrfs_readpage(NULL, page); 13034cb5300bSChris Mason lock_page(page); 13044cb5300bSChris Mason if (!PageUptodate(page)) { 13054cb5300bSChris Mason unlock_page(page); 130609cbfeafSKirill A. Shutemov put_page(page); 13074cb5300bSChris Mason ret = -EIO; 13084cb5300bSChris Mason break; 13094cb5300bSChris Mason } 13104cb5300bSChris Mason } 1311600a45e1SMiao Xie 1312600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1313600a45e1SMiao Xie unlock_page(page); 131409cbfeafSKirill A. Shutemov put_page(page); 1315600a45e1SMiao Xie goto again; 1316600a45e1SMiao Xie } 1317600a45e1SMiao Xie 13184cb5300bSChris Mason pages[i] = page; 13194cb5300bSChris Mason i_done++; 13204cb5300bSChris Mason } 13214cb5300bSChris Mason if (!i_done || ret) 13224cb5300bSChris Mason goto out; 13234cb5300bSChris Mason 13241751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13254cb5300bSChris Mason goto out; 13264cb5300bSChris Mason 13274cb5300bSChris Mason /* 13284cb5300bSChris Mason * so now we have a nice long stream of locked 13294cb5300bSChris Mason * and up to date pages, lets wait on them 13304cb5300bSChris Mason */ 13314cb5300bSChris Mason for (i = 0; i < i_done; i++) 13324cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13334cb5300bSChris Mason 13344cb5300bSChris Mason page_start = page_offset(pages[0]); 133509cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13364cb5300bSChris Mason 13374cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1338ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13394cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 13404cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 13419e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1342ae0f1625SDavid Sterba &cached_state); 13434cb5300bSChris Mason 13441f12bd06SLiu Bo if (i_done != page_cnt) { 13459e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 134628c4a3e2SSu Yue btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); 13479e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1348bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 134909cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 135043b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13514cb5300bSChris Mason } 13524cb5300bSChris Mason 13534cb5300bSChris Mason 13549e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1355018ed4f7SDavid Sterba &cached_state); 13564cb5300bSChris Mason 13574cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1358e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13594cb5300bSChris Mason 13604cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13614cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13624cb5300bSChris Mason ClearPageChecked(pages[i]); 13634cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13644cb5300bSChris Mason set_page_dirty(pages[i]); 13654cb5300bSChris Mason unlock_page(pages[i]); 136609cbfeafSKirill A. Shutemov put_page(pages[i]); 13674cb5300bSChris Mason } 136843b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 136943b18595SQu Wenruo false); 1370364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13714cb5300bSChris Mason return i_done; 13724cb5300bSChris Mason out: 13734cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13744cb5300bSChris Mason unlock_page(pages[i]); 137509cbfeafSKirill A. Shutemov put_page(pages[i]); 13764cb5300bSChris Mason } 1377bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 137809cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 137943b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 138043b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 138143b18595SQu Wenruo true); 1382364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13834cb5300bSChris Mason return ret; 13844cb5300bSChris Mason 13854cb5300bSChris Mason } 13864cb5300bSChris Mason 13874cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13884cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13894cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13904cb5300bSChris Mason { 13910b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13924cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13934cb5300bSChris Mason struct file_ra_state *ra = NULL; 13944cb5300bSChris Mason unsigned long last_index; 1395151a31b2SLi Zefan u64 isize = i_size_read(inode); 1396940100a4SChris Mason u64 last_len = 0; 1397940100a4SChris Mason u64 skip = 0; 1398940100a4SChris Mason u64 defrag_end = 0; 13994cb5300bSChris Mason u64 newer_off = range->start; 1400f46b5a66SChristoph Hellwig unsigned long i; 1401008873eaSLi Zefan unsigned long ra_index = 0; 1402f46b5a66SChristoph Hellwig int ret; 14034cb5300bSChris Mason int defrag_count = 0; 14041a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1405aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 140609cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1407c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1408ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 14094cb5300bSChris Mason struct page **pages = NULL; 14101e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14114cb5300bSChris Mason 14120abd5b17SLiu Bo if (isize == 0) 14130abd5b17SLiu Bo return 0; 14140abd5b17SLiu Bo 14150abd5b17SLiu Bo if (range->start >= isize) 14160abd5b17SLiu Bo return -EINVAL; 14171a419d85SLi Zefan 14181e2ef46dSDavid Sterba if (do_compress) { 14191a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 14201a419d85SLi Zefan return -EINVAL; 14211a419d85SLi Zefan if (range->compress_type) 14221a419d85SLi Zefan compress_type = range->compress_type; 14231a419d85SLi Zefan } 1424f46b5a66SChristoph Hellwig 14250abd5b17SLiu Bo if (extent_thresh == 0) 1426ee22184bSByongho Lee extent_thresh = SZ_256K; 1427f46b5a66SChristoph Hellwig 14284cb5300bSChris Mason /* 14290a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14300a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14310a52d108SDavid Sterba * we don't error out. 14324cb5300bSChris Mason */ 14334cb5300bSChris Mason if (!file) { 143463e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14350a52d108SDavid Sterba if (ra) 14364cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14374cb5300bSChris Mason } else { 14384cb5300bSChris Mason ra = &file->f_ra; 14394cb5300bSChris Mason } 14404cb5300bSChris Mason 144163e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14424cb5300bSChris Mason if (!pages) { 14434cb5300bSChris Mason ret = -ENOMEM; 14444cb5300bSChris Mason goto out_ra; 14454cb5300bSChris Mason } 14464cb5300bSChris Mason 14474cb5300bSChris Mason /* find the last page to defrag */ 14481e701a32SChris Mason if (range->start + range->len > range->start) { 1449151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 145009cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14511e701a32SChris Mason } else { 145209cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14531e701a32SChris Mason } 14541e701a32SChris Mason 14554cb5300bSChris Mason if (newer_than) { 14564cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1457ee22184bSByongho Lee &newer_off, SZ_64K); 14584cb5300bSChris Mason if (!ret) { 14594cb5300bSChris Mason range->start = newer_off; 14604cb5300bSChris Mason /* 14614cb5300bSChris Mason * we always align our defrag to help keep 14624cb5300bSChris Mason * the extents in the file evenly spaced 14634cb5300bSChris Mason */ 146409cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14654cb5300bSChris Mason } else 14664cb5300bSChris Mason goto out_ra; 14674cb5300bSChris Mason } else { 146809cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14694cb5300bSChris Mason } 14704cb5300bSChris Mason if (!max_to_defrag) 1471070034bdSchandan max_to_defrag = last_index - i + 1; 14724cb5300bSChris Mason 14732a0f7f57SLi Zefan /* 14742a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14752a0f7f57SLi Zefan * written sequentially. 14762a0f7f57SLi Zefan */ 14772a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14782a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14792a0f7f57SLi Zefan 1480f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 148109cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14824cb5300bSChris Mason /* 14834cb5300bSChris Mason * make sure we stop running if someone unmounts 14844cb5300bSChris Mason * the FS 14854cb5300bSChris Mason */ 14861751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14874cb5300bSChris Mason break; 14884cb5300bSChris Mason 14890b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14900b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1491210549ebSDavid Sterba ret = -EAGAIN; 1492210549ebSDavid Sterba break; 1493210549ebSDavid Sterba } 1494210549ebSDavid Sterba 149509cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14966c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14971e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1498940100a4SChris Mason unsigned long next; 1499940100a4SChris Mason /* 1500940100a4SChris Mason * the should_defrag function tells us how much to skip 1501940100a4SChris Mason * bump our counter by the suggested amount 1502940100a4SChris Mason */ 150309cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1504940100a4SChris Mason i = max(i + 1, next); 1505940100a4SChris Mason continue; 1506940100a4SChris Mason } 1507008873eaSLi Zefan 1508008873eaSLi Zefan if (!newer_than) { 150909cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 151009cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1511008873eaSLi Zefan cluster = min(cluster, max_cluster); 1512008873eaSLi Zefan } else { 1513008873eaSLi Zefan cluster = max_cluster; 1514008873eaSLi Zefan } 1515008873eaSLi Zefan 1516008873eaSLi Zefan if (i + cluster > ra_index) { 1517008873eaSLi Zefan ra_index = max(i, ra_index); 15180a52d108SDavid Sterba if (ra) 1519d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1520d3c0bab5SDavid Sterba file, ra_index, cluster); 1521e4826a5bSchandan ra_index += cluster; 1522008873eaSLi Zefan } 15234cb5300bSChris Mason 15245955102cSAl Viro inode_lock(inode); 1525eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1526eede2bf3SOmar Sandoval ret = -ETXTBSY; 1527eede2bf3SOmar Sandoval } else { 15281e2ef46dSDavid Sterba if (do_compress) 1529eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1530008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1531eede2bf3SOmar Sandoval } 1532ecb8bea8SLiu Bo if (ret < 0) { 15335955102cSAl Viro inode_unlock(inode); 15344cb5300bSChris Mason goto out_ra; 1535ecb8bea8SLiu Bo } 15364cb5300bSChris Mason 15374cb5300bSChris Mason defrag_count += ret; 1538d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15395955102cSAl Viro inode_unlock(inode); 15404cb5300bSChris Mason 15414cb5300bSChris Mason if (newer_than) { 15424cb5300bSChris Mason if (newer_off == (u64)-1) 15434cb5300bSChris Mason break; 15444cb5300bSChris Mason 1545e1f041e1SLiu Bo if (ret > 0) 1546e1f041e1SLiu Bo i += ret; 1547e1f041e1SLiu Bo 15484cb5300bSChris Mason newer_off = max(newer_off + 1, 154909cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15504cb5300bSChris Mason 1551ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1552ee22184bSByongho Lee &newer_off, SZ_64K); 15534cb5300bSChris Mason if (!ret) { 15544cb5300bSChris Mason range->start = newer_off; 155509cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15564cb5300bSChris Mason } else { 15574cb5300bSChris Mason break; 1558940100a4SChris Mason } 15594cb5300bSChris Mason } else { 1560008873eaSLi Zefan if (ret > 0) { 1561cbcc8326SLi Zefan i += ret; 156209cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1563008873eaSLi Zefan } else { 1564940100a4SChris Mason i++; 1565008873eaSLi Zefan last_len = 0; 1566008873eaSLi Zefan } 1567f46b5a66SChristoph Hellwig } 15684cb5300bSChris Mason } 1569f46b5a66SChristoph Hellwig 1570dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15711e701a32SChris Mason filemap_flush(inode->i_mapping); 1572dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1573dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1574dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1575dec8ef90SFilipe Manana } 15761e701a32SChris Mason 15771a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15780b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15795c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15805c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15811a419d85SLi Zefan } 15821a419d85SLi Zefan 158360ccf82fSDiego Calleja ret = defrag_count; 1584940100a4SChris Mason 15854cb5300bSChris Mason out_ra: 15861e2ef46dSDavid Sterba if (do_compress) { 15875955102cSAl Viro inode_lock(inode); 1588eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15895955102cSAl Viro inode_unlock(inode); 1590633085c7SFilipe David Borba Manana } 15914cb5300bSChris Mason if (!file) 15924cb5300bSChris Mason kfree(ra); 15934cb5300bSChris Mason kfree(pages); 1594940100a4SChris Mason return ret; 1595f46b5a66SChristoph Hellwig } 1596f46b5a66SChristoph Hellwig 1597198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 159876dda93cSYan, Zheng void __user *arg) 1599f46b5a66SChristoph Hellwig { 16000b246afaSJeff Mahoney struct inode *inode = file_inode(file); 16010b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1602f46b5a66SChristoph Hellwig u64 new_size; 1603f46b5a66SChristoph Hellwig u64 old_size; 1604f46b5a66SChristoph Hellwig u64 devid = 1; 16050b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1606f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1607f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1608f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1609f46b5a66SChristoph Hellwig char *sizestr; 16109a40f122SGui Hecheng char *retptr; 1611f46b5a66SChristoph Hellwig char *devstr = NULL; 1612f46b5a66SChristoph Hellwig int ret = 0; 1613f46b5a66SChristoph Hellwig int mod = 0; 1614f46b5a66SChristoph Hellwig 1615e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1616e441d54dSChris Mason return -EPERM; 1617e441d54dSChris Mason 1618198605a8SMiao Xie ret = mnt_want_write_file(file); 1619198605a8SMiao Xie if (ret) 1620198605a8SMiao Xie return ret; 1621198605a8SMiao Xie 1622171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 162397547676SMiao Xie mnt_drop_write_file(file); 1624e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1625c9e9f97bSIlya Dryomov } 1626c9e9f97bSIlya Dryomov 1627dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1628c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1629c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1630c9e9f97bSIlya Dryomov goto out; 1631c9e9f97bSIlya Dryomov } 16325516e595SMark Fasheh 16335516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1634f46b5a66SChristoph Hellwig 1635f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1636f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1637f46b5a66SChristoph Hellwig if (devstr) { 1638f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1639f46b5a66SChristoph Hellwig *devstr = '\0'; 1640f46b5a66SChristoph Hellwig devstr = vol_args->name; 164158dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 164258dfae63SZhangZhen if (ret) 164358dfae63SZhangZhen goto out_free; 1644dfd79829SMiao Xie if (!devid) { 1645dfd79829SMiao Xie ret = -EINVAL; 1646dfd79829SMiao Xie goto out_free; 1647dfd79829SMiao Xie } 16480b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1649f46b5a66SChristoph Hellwig } 1650dba60f3fSMiao Xie 165109ba3bc9SAnand Jain device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true); 1652f46b5a66SChristoph Hellwig if (!device) { 16530b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1654c1c9ff7cSGeert Uytterhoeven devid); 1655dfd79829SMiao Xie ret = -ENODEV; 1656c9e9f97bSIlya Dryomov goto out_free; 1657f46b5a66SChristoph Hellwig } 1658dba60f3fSMiao Xie 1659ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16600b246afaSJeff Mahoney btrfs_info(fs_info, 1661efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1662c1c9ff7cSGeert Uytterhoeven devid); 1663dfd79829SMiao Xie ret = -EPERM; 16644e42ae1bSLiu Bo goto out_free; 16654e42ae1bSLiu Bo } 16664e42ae1bSLiu Bo 1667f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1668f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1669f46b5a66SChristoph Hellwig else { 1670f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1671f46b5a66SChristoph Hellwig mod = -1; 1672f46b5a66SChristoph Hellwig sizestr++; 1673f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1674f46b5a66SChristoph Hellwig mod = 1; 1675f46b5a66SChristoph Hellwig sizestr++; 1676f46b5a66SChristoph Hellwig } 16779a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16789a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1679f46b5a66SChristoph Hellwig ret = -EINVAL; 1680c9e9f97bSIlya Dryomov goto out_free; 1681f46b5a66SChristoph Hellwig } 1682f46b5a66SChristoph Hellwig } 1683f46b5a66SChristoph Hellwig 1684401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1685dfd79829SMiao Xie ret = -EPERM; 168663a212abSStefan Behrens goto out_free; 168763a212abSStefan Behrens } 168863a212abSStefan Behrens 16897cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1690f46b5a66SChristoph Hellwig 1691f46b5a66SChristoph Hellwig if (mod < 0) { 1692f46b5a66SChristoph Hellwig if (new_size > old_size) { 1693f46b5a66SChristoph Hellwig ret = -EINVAL; 1694c9e9f97bSIlya Dryomov goto out_free; 1695f46b5a66SChristoph Hellwig } 1696f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1697f46b5a66SChristoph Hellwig } else if (mod > 0) { 1698eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1699902c68a4SGui Hecheng ret = -ERANGE; 1700eb8052e0SWenliang Fan goto out_free; 1701eb8052e0SWenliang Fan } 1702f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1703f46b5a66SChristoph Hellwig } 1704f46b5a66SChristoph Hellwig 1705ee22184bSByongho Lee if (new_size < SZ_256M) { 1706f46b5a66SChristoph Hellwig ret = -EINVAL; 1707c9e9f97bSIlya Dryomov goto out_free; 1708f46b5a66SChristoph Hellwig } 1709f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1710f46b5a66SChristoph Hellwig ret = -EFBIG; 1711c9e9f97bSIlya Dryomov goto out_free; 1712f46b5a66SChristoph Hellwig } 1713f46b5a66SChristoph Hellwig 171447f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1715f46b5a66SChristoph Hellwig 17160b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1717c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1718f46b5a66SChristoph Hellwig 1719f46b5a66SChristoph Hellwig if (new_size > old_size) { 1720a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 172198d5dc13STsutomu Itoh if (IS_ERR(trans)) { 172298d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1723c9e9f97bSIlya Dryomov goto out_free; 172498d5dc13STsutomu Itoh } 1725f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17263a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1727ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1728f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17290253f40eSjeff.liu } /* equal, nothing need to do */ 1730f46b5a66SChristoph Hellwig 1731c9e9f97bSIlya Dryomov out_free: 1732f46b5a66SChristoph Hellwig kfree(vol_args); 1733c9e9f97bSIlya Dryomov out: 1734171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 173518f39c41SIlya Dryomov mnt_drop_write_file(file); 1736f46b5a66SChristoph Hellwig return ret; 1737f46b5a66SChristoph Hellwig } 1738f46b5a66SChristoph Hellwig 173972fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 174052f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17416f72c7e2SArne Jansen u64 *transid, bool readonly, 17428696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1743f46b5a66SChristoph Hellwig { 1744f46b5a66SChristoph Hellwig int namelen; 17453de4586cSChris Mason int ret = 0; 1746f46b5a66SChristoph Hellwig 1747325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1748325c50e3SJeff Mahoney return -ENOTDIR; 1749325c50e3SJeff Mahoney 1750a874a63eSLiu Bo ret = mnt_want_write_file(file); 1751a874a63eSLiu Bo if (ret) 1752a874a63eSLiu Bo goto out; 1753a874a63eSLiu Bo 175472fd032eSSage Weil namelen = strlen(name); 175572fd032eSSage Weil if (strchr(name, '/')) { 1756f46b5a66SChristoph Hellwig ret = -EINVAL; 1757a874a63eSLiu Bo goto out_drop_write; 1758f46b5a66SChristoph Hellwig } 1759f46b5a66SChristoph Hellwig 176016780cabSChris Mason if (name[0] == '.' && 176116780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 176216780cabSChris Mason ret = -EEXIST; 1763a874a63eSLiu Bo goto out_drop_write; 176416780cabSChris Mason } 176516780cabSChris Mason 17663de4586cSChris Mason if (subvol) { 176772fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17686f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1769cb8e7090SChristoph Hellwig } else { 17702903ff01SAl Viro struct fd src = fdget(fd); 17713de4586cSChris Mason struct inode *src_inode; 17722903ff01SAl Viro if (!src.file) { 17733de4586cSChris Mason ret = -EINVAL; 1774a874a63eSLiu Bo goto out_drop_write; 17753de4586cSChris Mason } 17763de4586cSChris Mason 1777496ad9aaSAl Viro src_inode = file_inode(src.file); 1778496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1779c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1780efe120a0SFrank Holton "Snapshot src from another FS"); 178123ad5b17SKusanagi Kouichi ret = -EXDEV; 1782d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1783d0242061SDavid Sterba /* 1784d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1785d0242061SDavid Sterba * are limited to own subvolumes only 1786d0242061SDavid Sterba */ 1787d0242061SDavid Sterba ret = -EPERM; 1788ecd18815SAl Viro } else { 178972fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 179072fd032eSSage Weil BTRFS_I(src_inode)->root, 17916f72c7e2SArne Jansen transid, readonly, inherit); 1792ecd18815SAl Viro } 17932903ff01SAl Viro fdput(src); 1794cb8e7090SChristoph Hellwig } 1795a874a63eSLiu Bo out_drop_write: 1796a874a63eSLiu Bo mnt_drop_write_file(file); 1797f46b5a66SChristoph Hellwig out: 179872fd032eSSage Weil return ret; 179972fd032eSSage Weil } 180072fd032eSSage Weil 180172fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1802fa0d2b9bSLi Zefan void __user *arg, int subvol) 180372fd032eSSage Weil { 1804fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 180572fd032eSSage Weil int ret; 180672fd032eSSage Weil 1807325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1808325c50e3SJeff Mahoney return -ENOTDIR; 1809325c50e3SJeff Mahoney 1810fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1811fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1812fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1813fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1814fa0d2b9bSLi Zefan 1815fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1816b83cc969SLi Zefan vol_args->fd, subvol, 18176f72c7e2SArne Jansen NULL, false, NULL); 1818fa0d2b9bSLi Zefan 1819fa0d2b9bSLi Zefan kfree(vol_args); 1820fa0d2b9bSLi Zefan return ret; 1821fa0d2b9bSLi Zefan } 1822fa0d2b9bSLi Zefan 1823fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1824fa0d2b9bSLi Zefan void __user *arg, int subvol) 1825fa0d2b9bSLi Zefan { 1826fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1827fa0d2b9bSLi Zefan int ret; 1828fdfb1e4fSLi Zefan u64 transid = 0; 1829fdfb1e4fSLi Zefan u64 *ptr = NULL; 1830b83cc969SLi Zefan bool readonly = false; 18316f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 183272fd032eSSage Weil 1833325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1834325c50e3SJeff Mahoney return -ENOTDIR; 1835325c50e3SJeff Mahoney 1836fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1837fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1838fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1839fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1840fdfb1e4fSLi Zefan 1841b83cc969SLi Zefan if (vol_args->flags & 18426f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18436f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1844b83cc969SLi Zefan ret = -EOPNOTSUPP; 1845c47ca32dSDan Carpenter goto free_args; 1846fdfb1e4fSLi Zefan } 1847fdfb1e4fSLi Zefan 1848fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1849fdfb1e4fSLi Zefan ptr = &transid; 1850b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1851b83cc969SLi Zefan readonly = true; 18526f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 185309cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18546f72c7e2SArne Jansen ret = -EINVAL; 1855c47ca32dSDan Carpenter goto free_args; 18566f72c7e2SArne Jansen } 18576f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18586f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18596f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1860c47ca32dSDan Carpenter goto free_args; 18616f72c7e2SArne Jansen } 18626f72c7e2SArne Jansen } 186375eaa0e2SSage Weil 1864fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18656f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18668696c533SMiao Xie readonly, inherit); 1867c47ca32dSDan Carpenter if (ret) 1868c47ca32dSDan Carpenter goto free_inherit; 186975eaa0e2SSage Weil 1870c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1871fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1872c47ca32dSDan Carpenter transid), 1873c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 187475eaa0e2SSage Weil ret = -EFAULT; 1875c47ca32dSDan Carpenter 1876c47ca32dSDan Carpenter free_inherit: 18776f72c7e2SArne Jansen kfree(inherit); 1878c47ca32dSDan Carpenter free_args: 1879c47ca32dSDan Carpenter kfree(vol_args); 1880f46b5a66SChristoph Hellwig return ret; 1881f46b5a66SChristoph Hellwig } 1882f46b5a66SChristoph Hellwig 18830caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18840caa102dSLi Zefan void __user *arg) 18850caa102dSLi Zefan { 1886496ad9aaSAl Viro struct inode *inode = file_inode(file); 18870b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18880caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18890caa102dSLi Zefan int ret = 0; 18900caa102dSLi Zefan u64 flags = 0; 18910caa102dSLi Zefan 18924a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18930caa102dSLi Zefan return -EINVAL; 18940caa102dSLi Zefan 18950b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 18960caa102dSLi Zefan if (btrfs_root_readonly(root)) 18970caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 18980b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 18990caa102dSLi Zefan 19000caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 19010caa102dSLi Zefan ret = -EFAULT; 19020caa102dSLi Zefan 19030caa102dSLi Zefan return ret; 19040caa102dSLi Zefan } 19050caa102dSLi Zefan 19060caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19070caa102dSLi Zefan void __user *arg) 19080caa102dSLi Zefan { 1909496ad9aaSAl Viro struct inode *inode = file_inode(file); 19100b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19110caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19120caa102dSLi Zefan struct btrfs_trans_handle *trans; 19130caa102dSLi Zefan u64 root_flags; 19140caa102dSLi Zefan u64 flags; 19150caa102dSLi Zefan int ret = 0; 19160caa102dSLi Zefan 1917bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1918bd60ea0fSDavid Sterba return -EPERM; 1919bd60ea0fSDavid Sterba 1920b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1921b9ca0664SLiu Bo if (ret) 1922b9ca0664SLiu Bo goto out; 19230caa102dSLi Zefan 19244a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1925b9ca0664SLiu Bo ret = -EINVAL; 1926b9ca0664SLiu Bo goto out_drop_write; 1927b9ca0664SLiu Bo } 19280caa102dSLi Zefan 1929b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1930b9ca0664SLiu Bo ret = -EFAULT; 1931b9ca0664SLiu Bo goto out_drop_write; 1932b9ca0664SLiu Bo } 19330caa102dSLi Zefan 1934b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1935b9ca0664SLiu Bo ret = -EINVAL; 1936b9ca0664SLiu Bo goto out_drop_write; 1937b9ca0664SLiu Bo } 19380caa102dSLi Zefan 1939b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1940b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1941b9ca0664SLiu Bo goto out_drop_write; 1942b9ca0664SLiu Bo } 19430caa102dSLi Zefan 19440b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19450caa102dSLi Zefan 19460caa102dSLi Zefan /* nothing to do */ 19470caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1948b9ca0664SLiu Bo goto out_drop_sem; 19490caa102dSLi Zefan 19500caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19512c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19520caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19530caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19542c686537SDavid Sterba } else { 19552c686537SDavid Sterba /* 19562c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19572c686537SDavid Sterba * send 19582c686537SDavid Sterba */ 19592c686537SDavid Sterba spin_lock(&root->root_item_lock); 19602c686537SDavid Sterba if (root->send_in_progress == 0) { 19610caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19620caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19632c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19642c686537SDavid Sterba } else { 19652c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19660b246afaSJeff Mahoney btrfs_warn(fs_info, 19672c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19682c686537SDavid Sterba root->root_key.objectid); 19692c686537SDavid Sterba ret = -EPERM; 19702c686537SDavid Sterba goto out_drop_sem; 19712c686537SDavid Sterba } 19722c686537SDavid Sterba } 19730caa102dSLi Zefan 19740caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19750caa102dSLi Zefan if (IS_ERR(trans)) { 19760caa102dSLi Zefan ret = PTR_ERR(trans); 19770caa102dSLi Zefan goto out_reset; 19780caa102dSLi Zefan } 19790caa102dSLi Zefan 19800b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19810caa102dSLi Zefan &root->root_key, &root->root_item); 19829417ebc8SNikolay Borisov if (ret < 0) { 19839417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19849417ebc8SNikolay Borisov goto out_reset; 19859417ebc8SNikolay Borisov } 19860caa102dSLi Zefan 19879417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19889417ebc8SNikolay Borisov 19890caa102dSLi Zefan out_reset: 19900caa102dSLi Zefan if (ret) 19910caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1992b9ca0664SLiu Bo out_drop_sem: 19930b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1994b9ca0664SLiu Bo out_drop_write: 1995b9ca0664SLiu Bo mnt_drop_write_file(file); 1996b9ca0664SLiu Bo out: 19970caa102dSLi Zefan return ret; 19980caa102dSLi Zefan } 19990caa102dSLi Zefan 2000ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2001ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2002ac8e9819SChris Mason { 2003abc6e134SChris Mason struct btrfs_key test; 2004abc6e134SChris Mason int ret; 2005abc6e134SChris Mason 2006abc6e134SChris Mason test.objectid = sk->min_objectid; 2007abc6e134SChris Mason test.type = sk->min_type; 2008abc6e134SChris Mason test.offset = sk->min_offset; 2009abc6e134SChris Mason 2010abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2011abc6e134SChris Mason if (ret < 0) 2012ac8e9819SChris Mason return 0; 2013abc6e134SChris Mason 2014abc6e134SChris Mason test.objectid = sk->max_objectid; 2015abc6e134SChris Mason test.type = sk->max_type; 2016abc6e134SChris Mason test.offset = sk->max_offset; 2017abc6e134SChris Mason 2018abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2019abc6e134SChris Mason if (ret > 0) 2020ac8e9819SChris Mason return 0; 2021ac8e9819SChris Mason return 1; 2022ac8e9819SChris Mason } 2023ac8e9819SChris Mason 2024df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2025ac8e9819SChris Mason struct btrfs_key *key, 2026ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20279b6e817dSGerhard Heift size_t *buf_size, 2028ba346b35SGerhard Heift char __user *ubuf, 2029ac8e9819SChris Mason unsigned long *sk_offset, 2030ac8e9819SChris Mason int *num_found) 2031ac8e9819SChris Mason { 2032ac8e9819SChris Mason u64 found_transid; 2033ac8e9819SChris Mason struct extent_buffer *leaf; 2034ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2035dd81d459SNaohiro Aota struct btrfs_key test; 2036ac8e9819SChris Mason unsigned long item_off; 2037ac8e9819SChris Mason unsigned long item_len; 2038ac8e9819SChris Mason int nritems; 2039ac8e9819SChris Mason int i; 2040ac8e9819SChris Mason int slot; 2041ac8e9819SChris Mason int ret = 0; 2042ac8e9819SChris Mason 2043ac8e9819SChris Mason leaf = path->nodes[0]; 2044ac8e9819SChris Mason slot = path->slots[0]; 2045ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2046ac8e9819SChris Mason 2047ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2048ac8e9819SChris Mason i = nritems; 2049ac8e9819SChris Mason goto advance_key; 2050ac8e9819SChris Mason } 2051ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2052ac8e9819SChris Mason 2053ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2054ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2055ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2056ac8e9819SChris Mason 205703b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 205803b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 205903b71c6cSGabriel de Perthuis continue; 206003b71c6cSGabriel de Perthuis 20619b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20628f5f6178SGerhard Heift if (*num_found) { 2063ac8e9819SChris Mason ret = 1; 20648f5f6178SGerhard Heift goto out; 20658f5f6178SGerhard Heift } 20668f5f6178SGerhard Heift 20678f5f6178SGerhard Heift /* 20688f5f6178SGerhard Heift * return one empty item back for v1, which does not 20698f5f6178SGerhard Heift * handle -EOVERFLOW 20708f5f6178SGerhard Heift */ 20718f5f6178SGerhard Heift 20729b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2073ac8e9819SChris Mason item_len = 0; 20748f5f6178SGerhard Heift ret = -EOVERFLOW; 20758f5f6178SGerhard Heift } 2076ac8e9819SChris Mason 20779b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2078ac8e9819SChris Mason ret = 1; 207925c9bc2eSGerhard Heift goto out; 2080ac8e9819SChris Mason } 2081ac8e9819SChris Mason 2082ac8e9819SChris Mason sh.objectid = key->objectid; 2083ac8e9819SChris Mason sh.offset = key->offset; 2084ac8e9819SChris Mason sh.type = key->type; 2085ac8e9819SChris Mason sh.len = item_len; 2086ac8e9819SChris Mason sh.transid = found_transid; 2087ac8e9819SChris Mason 2088ac8e9819SChris Mason /* copy search result header */ 2089ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2090ba346b35SGerhard Heift ret = -EFAULT; 2091ba346b35SGerhard Heift goto out; 2092ba346b35SGerhard Heift } 2093ba346b35SGerhard Heift 2094ac8e9819SChris Mason *sk_offset += sizeof(sh); 2095ac8e9819SChris Mason 2096ac8e9819SChris Mason if (item_len) { 2097ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2098ac8e9819SChris Mason /* copy the item */ 2099ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2100ba346b35SGerhard Heift item_off, item_len)) { 2101ba346b35SGerhard Heift ret = -EFAULT; 2102ba346b35SGerhard Heift goto out; 2103ba346b35SGerhard Heift } 2104ba346b35SGerhard Heift 2105ac8e9819SChris Mason *sk_offset += item_len; 2106ac8e9819SChris Mason } 2107e2156867SHugo Mills (*num_found)++; 2108ac8e9819SChris Mason 21098f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21108f5f6178SGerhard Heift goto out; 21118f5f6178SGerhard Heift 211225c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 211325c9bc2eSGerhard Heift ret = 1; 211425c9bc2eSGerhard Heift goto out; 211525c9bc2eSGerhard Heift } 2116ac8e9819SChris Mason } 2117ac8e9819SChris Mason advance_key: 2118ac8e9819SChris Mason ret = 0; 2119dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2120dd81d459SNaohiro Aota test.type = sk->max_type; 2121dd81d459SNaohiro Aota test.offset = sk->max_offset; 2122dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2123dd81d459SNaohiro Aota ret = 1; 2124dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2125abc6e134SChris Mason key->offset++; 2126dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2127abc6e134SChris Mason key->offset = 0; 2128abc6e134SChris Mason key->type++; 2129dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2130abc6e134SChris Mason key->offset = 0; 2131abc6e134SChris Mason key->type = 0; 2132abc6e134SChris Mason key->objectid++; 2133abc6e134SChris Mason } else 2134abc6e134SChris Mason ret = 1; 213525c9bc2eSGerhard Heift out: 2136ba346b35SGerhard Heift /* 2137ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2138ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2139ba346b35SGerhard Heift * * all items were found 2140ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2141ba346b35SGerhard Heift * leaf 2142ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2143ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2144ba346b35SGerhard Heift */ 2145ac8e9819SChris Mason return ret; 2146ac8e9819SChris Mason } 2147ac8e9819SChris Mason 2148ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 214912544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21509b6e817dSGerhard Heift size_t *buf_size, 2151ba346b35SGerhard Heift char __user *ubuf) 2152ac8e9819SChris Mason { 21530b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2154ac8e9819SChris Mason struct btrfs_root *root; 2155ac8e9819SChris Mason struct btrfs_key key; 2156ac8e9819SChris Mason struct btrfs_path *path; 2157ac8e9819SChris Mason int ret; 2158ac8e9819SChris Mason int num_found = 0; 2159ac8e9819SChris Mason unsigned long sk_offset = 0; 2160ac8e9819SChris Mason 21619b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21629b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 216312544442SGerhard Heift return -EOVERFLOW; 21649b6e817dSGerhard Heift } 216512544442SGerhard Heift 2166ac8e9819SChris Mason path = btrfs_alloc_path(); 2167ac8e9819SChris Mason if (!path) 2168ac8e9819SChris Mason return -ENOMEM; 2169ac8e9819SChris Mason 2170ac8e9819SChris Mason if (sk->tree_id == 0) { 2171ac8e9819SChris Mason /* search the root of the inode that was passed */ 2172ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2173ac8e9819SChris Mason } else { 2174ac8e9819SChris Mason key.objectid = sk->tree_id; 2175ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2176ac8e9819SChris Mason key.offset = (u64)-1; 2177ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2178ac8e9819SChris Mason if (IS_ERR(root)) { 2179ac8e9819SChris Mason btrfs_free_path(path); 2180ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2181ac8e9819SChris Mason } 2182ac8e9819SChris Mason } 2183ac8e9819SChris Mason 2184ac8e9819SChris Mason key.objectid = sk->min_objectid; 2185ac8e9819SChris Mason key.type = sk->min_type; 2186ac8e9819SChris Mason key.offset = sk->min_offset; 2187ac8e9819SChris Mason 2188ac8e9819SChris Mason while (1) { 21896174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2190ac8e9819SChris Mason if (ret != 0) { 2191ac8e9819SChris Mason if (ret > 0) 2192ac8e9819SChris Mason ret = 0; 2193ac8e9819SChris Mason goto err; 2194ac8e9819SChris Mason } 2195df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2196ac8e9819SChris Mason &sk_offset, &num_found); 2197b3b4aa74SDavid Sterba btrfs_release_path(path); 219825c9bc2eSGerhard Heift if (ret) 2199ac8e9819SChris Mason break; 2200ac8e9819SChris Mason 2201ac8e9819SChris Mason } 22028f5f6178SGerhard Heift if (ret > 0) 2203ac8e9819SChris Mason ret = 0; 2204ac8e9819SChris Mason err: 2205ac8e9819SChris Mason sk->nr_items = num_found; 2206ac8e9819SChris Mason btrfs_free_path(path); 2207ac8e9819SChris Mason return ret; 2208ac8e9819SChris Mason } 2209ac8e9819SChris Mason 2210ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2211ac8e9819SChris Mason void __user *argp) 2212ac8e9819SChris Mason { 2213ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2214ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2215ac8e9819SChris Mason struct inode *inode; 2216ac8e9819SChris Mason int ret; 22179b6e817dSGerhard Heift size_t buf_size; 2218ac8e9819SChris Mason 2219ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2220ac8e9819SChris Mason return -EPERM; 2221ac8e9819SChris Mason 2222ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2223ac8e9819SChris Mason 2224ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2225ba346b35SGerhard Heift return -EFAULT; 2226ba346b35SGerhard Heift 2227ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2228ac8e9819SChris Mason 2229496ad9aaSAl Viro inode = file_inode(file); 2230ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22318f5f6178SGerhard Heift 22328f5f6178SGerhard Heift /* 22338f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22348f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22358f5f6178SGerhard Heift */ 22368f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22378f5f6178SGerhard Heift ret = 0; 22388f5f6178SGerhard Heift 2239ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2240ac8e9819SChris Mason ret = -EFAULT; 2241ac8e9819SChris Mason return ret; 2242ac8e9819SChris Mason } 2243ac8e9819SChris Mason 2244cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2245cc68a8a5SGerhard Heift void __user *argp) 2246cc68a8a5SGerhard Heift { 2247cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2248cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2249cc68a8a5SGerhard Heift struct inode *inode; 2250cc68a8a5SGerhard Heift int ret; 2251cc68a8a5SGerhard Heift size_t buf_size; 2252ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2253cc68a8a5SGerhard Heift 2254cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2255cc68a8a5SGerhard Heift return -EPERM; 2256cc68a8a5SGerhard Heift 2257cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2258cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2259cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2260cc68a8a5SGerhard Heift return -EFAULT; 2261cc68a8a5SGerhard Heift 2262cc68a8a5SGerhard Heift buf_size = args.buf_size; 2263cc68a8a5SGerhard Heift 2264cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2265cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2266cc68a8a5SGerhard Heift buf_size = buf_limit; 2267cc68a8a5SGerhard Heift 2268cc68a8a5SGerhard Heift inode = file_inode(file); 2269cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2270718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2271cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2272cc68a8a5SGerhard Heift ret = -EFAULT; 2273cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2274cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2275cc68a8a5SGerhard Heift ret = -EFAULT; 2276cc68a8a5SGerhard Heift 227776dda93cSYan, Zheng return ret; 227876dda93cSYan, Zheng } 227976dda93cSYan, Zheng 228098d377a0STARUISI Hiroaki /* 2281ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2282ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 228398d377a0STARUISI Hiroaki */ 228498d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 228598d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 228698d377a0STARUISI Hiroaki { 228798d377a0STARUISI Hiroaki struct btrfs_root *root; 228898d377a0STARUISI Hiroaki struct btrfs_key key; 2289ac8e9819SChris Mason char *ptr; 229098d377a0STARUISI Hiroaki int ret = -1; 229198d377a0STARUISI Hiroaki int slot; 229298d377a0STARUISI Hiroaki int len; 229398d377a0STARUISI Hiroaki int total_len = 0; 229498d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 229598d377a0STARUISI Hiroaki struct extent_buffer *l; 229698d377a0STARUISI Hiroaki struct btrfs_path *path; 229798d377a0STARUISI Hiroaki 229898d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 229998d377a0STARUISI Hiroaki name[0]='\0'; 230098d377a0STARUISI Hiroaki return 0; 230198d377a0STARUISI Hiroaki } 230298d377a0STARUISI Hiroaki 230398d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 230498d377a0STARUISI Hiroaki if (!path) 230598d377a0STARUISI Hiroaki return -ENOMEM; 230698d377a0STARUISI Hiroaki 2307c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 230898d377a0STARUISI Hiroaki 230998d377a0STARUISI Hiroaki key.objectid = tree_id; 231098d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 231198d377a0STARUISI Hiroaki key.offset = (u64)-1; 231298d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 231398d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2314ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 23158ad6fcabSChris Mason goto out; 231698d377a0STARUISI Hiroaki } 231798d377a0STARUISI Hiroaki 231898d377a0STARUISI Hiroaki key.objectid = dirid; 231998d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23208ad6fcabSChris Mason key.offset = (u64)-1; 232198d377a0STARUISI Hiroaki 232298d377a0STARUISI Hiroaki while (1) { 232398d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 232498d377a0STARUISI Hiroaki if (ret < 0) 232598d377a0STARUISI Hiroaki goto out; 232618674c6cSFilipe David Borba Manana else if (ret > 0) { 232718674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 232818674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 232918674c6cSFilipe David Borba Manana if (ret < 0) 233018674c6cSFilipe David Borba Manana goto out; 233118674c6cSFilipe David Borba Manana else if (ret > 0) { 2332ac8e9819SChris Mason ret = -ENOENT; 233398d377a0STARUISI Hiroaki goto out; 2334ac8e9819SChris Mason } 233518674c6cSFilipe David Borba Manana } 233618674c6cSFilipe David Borba Manana 233718674c6cSFilipe David Borba Manana l = path->nodes[0]; 233818674c6cSFilipe David Borba Manana slot = path->slots[0]; 233918674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 234098d377a0STARUISI Hiroaki 234198d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 234298d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 234398d377a0STARUISI Hiroaki ptr -= len + 1; 234498d377a0STARUISI Hiroaki total_len += len + 1; 2345a696cf35SFilipe David Borba Manana if (ptr < name) { 2346a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 234798d377a0STARUISI Hiroaki goto out; 2348a696cf35SFilipe David Borba Manana } 234998d377a0STARUISI Hiroaki 235098d377a0STARUISI Hiroaki *(ptr + len) = '/'; 235198d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 235298d377a0STARUISI Hiroaki 235398d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 235498d377a0STARUISI Hiroaki break; 235598d377a0STARUISI Hiroaki 2356b3b4aa74SDavid Sterba btrfs_release_path(path); 235798d377a0STARUISI Hiroaki key.objectid = key.offset; 23588ad6fcabSChris Mason key.offset = (u64)-1; 235998d377a0STARUISI Hiroaki dirid = key.objectid; 236098d377a0STARUISI Hiroaki } 236177906a50SLi Zefan memmove(name, ptr, total_len); 236298d377a0STARUISI Hiroaki name[total_len] = '\0'; 236398d377a0STARUISI Hiroaki ret = 0; 236498d377a0STARUISI Hiroaki out: 236598d377a0STARUISI Hiroaki btrfs_free_path(path); 2366ac8e9819SChris Mason return ret; 2367ac8e9819SChris Mason } 2368ac8e9819SChris Mason 236923d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 237023d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 237123d0b79dSTomohiro Misono { 237223d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 237323d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 237423d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 237523d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 237623d0b79dSTomohiro Misono u64 dirid = args->dirid; 237723d0b79dSTomohiro Misono unsigned long item_off; 237823d0b79dSTomohiro Misono unsigned long item_len; 237923d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 238023d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 238123d0b79dSTomohiro Misono struct btrfs_root *root; 238223d0b79dSTomohiro Misono struct btrfs_path *path; 238323d0b79dSTomohiro Misono struct btrfs_key key, key2; 238423d0b79dSTomohiro Misono struct extent_buffer *leaf; 238523d0b79dSTomohiro Misono struct inode *temp_inode; 238623d0b79dSTomohiro Misono char *ptr; 238723d0b79dSTomohiro Misono int slot; 238823d0b79dSTomohiro Misono int len; 238923d0b79dSTomohiro Misono int total_len = 0; 239023d0b79dSTomohiro Misono int ret; 239123d0b79dSTomohiro Misono 239223d0b79dSTomohiro Misono path = btrfs_alloc_path(); 239323d0b79dSTomohiro Misono if (!path) 239423d0b79dSTomohiro Misono return -ENOMEM; 239523d0b79dSTomohiro Misono 239623d0b79dSTomohiro Misono /* 239723d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 239823d0b79dSTomohiro Misono * construct the path in from the bottom up. 239923d0b79dSTomohiro Misono */ 240023d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 240123d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 240223d0b79dSTomohiro Misono 240323d0b79dSTomohiro Misono key.objectid = treeid; 240423d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 240523d0b79dSTomohiro Misono key.offset = (u64)-1; 240623d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 240723d0b79dSTomohiro Misono if (IS_ERR(root)) { 240823d0b79dSTomohiro Misono ret = PTR_ERR(root); 240923d0b79dSTomohiro Misono goto out; 241023d0b79dSTomohiro Misono } 241123d0b79dSTomohiro Misono 241223d0b79dSTomohiro Misono key.objectid = dirid; 241323d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 241423d0b79dSTomohiro Misono key.offset = (u64)-1; 241523d0b79dSTomohiro Misono while (1) { 241623d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 241723d0b79dSTomohiro Misono if (ret < 0) { 241823d0b79dSTomohiro Misono goto out; 241923d0b79dSTomohiro Misono } else if (ret > 0) { 242023d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 242123d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 242223d0b79dSTomohiro Misono if (ret < 0) { 242323d0b79dSTomohiro Misono goto out; 242423d0b79dSTomohiro Misono } else if (ret > 0) { 242523d0b79dSTomohiro Misono ret = -ENOENT; 242623d0b79dSTomohiro Misono goto out; 242723d0b79dSTomohiro Misono } 242823d0b79dSTomohiro Misono } 242923d0b79dSTomohiro Misono 243023d0b79dSTomohiro Misono leaf = path->nodes[0]; 243123d0b79dSTomohiro Misono slot = path->slots[0]; 243223d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 243323d0b79dSTomohiro Misono 243423d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 243523d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 243623d0b79dSTomohiro Misono ptr -= len + 1; 243723d0b79dSTomohiro Misono total_len += len + 1; 243823d0b79dSTomohiro Misono if (ptr < args->path) { 243923d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 244023d0b79dSTomohiro Misono goto out; 244123d0b79dSTomohiro Misono } 244223d0b79dSTomohiro Misono 244323d0b79dSTomohiro Misono *(ptr + len) = '/'; 244423d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 244523d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 244623d0b79dSTomohiro Misono 244723d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 244823d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 244923d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 245023d0b79dSTomohiro Misono if (ret < 0) { 245123d0b79dSTomohiro Misono goto out; 245223d0b79dSTomohiro Misono } else if (ret > 0) { 245323d0b79dSTomohiro Misono ret = -ENOENT; 245423d0b79dSTomohiro Misono goto out; 245523d0b79dSTomohiro Misono } 245623d0b79dSTomohiro Misono 245723d0b79dSTomohiro Misono leaf = path->nodes[0]; 245823d0b79dSTomohiro Misono slot = path->slots[0]; 245923d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 246023d0b79dSTomohiro Misono if (key2.objectid != dirid) { 246123d0b79dSTomohiro Misono ret = -ENOENT; 246223d0b79dSTomohiro Misono goto out; 246323d0b79dSTomohiro Misono } 246423d0b79dSTomohiro Misono 246523d0b79dSTomohiro Misono temp_inode = btrfs_iget(sb, &key2, root, NULL); 24663ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24673ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24683ca57bd6SMisono Tomohiro goto out; 24693ca57bd6SMisono Tomohiro } 247023d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 247123d0b79dSTomohiro Misono iput(temp_inode); 247223d0b79dSTomohiro Misono if (ret) { 247323d0b79dSTomohiro Misono ret = -EACCES; 247423d0b79dSTomohiro Misono goto out; 247523d0b79dSTomohiro Misono } 247623d0b79dSTomohiro Misono 247723d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 247823d0b79dSTomohiro Misono break; 247923d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 248023d0b79dSTomohiro Misono ret = -EACCES; 248123d0b79dSTomohiro Misono goto out; 248223d0b79dSTomohiro Misono } 248323d0b79dSTomohiro Misono 248423d0b79dSTomohiro Misono btrfs_release_path(path); 248523d0b79dSTomohiro Misono key.objectid = key.offset; 248623d0b79dSTomohiro Misono key.offset = (u64)-1; 248723d0b79dSTomohiro Misono dirid = key.objectid; 248823d0b79dSTomohiro Misono } 248923d0b79dSTomohiro Misono 249023d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 249123d0b79dSTomohiro Misono args->path[total_len] = '\0'; 249223d0b79dSTomohiro Misono btrfs_release_path(path); 249323d0b79dSTomohiro Misono } 249423d0b79dSTomohiro Misono 249523d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 249623d0b79dSTomohiro Misono root = fs_info->tree_root; 249723d0b79dSTomohiro Misono key.objectid = treeid; 249823d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 249923d0b79dSTomohiro Misono key.offset = args->treeid; 250023d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 250123d0b79dSTomohiro Misono if (ret < 0) { 250223d0b79dSTomohiro Misono goto out; 250323d0b79dSTomohiro Misono } else if (ret > 0) { 250423d0b79dSTomohiro Misono ret = -ENOENT; 250523d0b79dSTomohiro Misono goto out; 250623d0b79dSTomohiro Misono } 250723d0b79dSTomohiro Misono 250823d0b79dSTomohiro Misono leaf = path->nodes[0]; 250923d0b79dSTomohiro Misono slot = path->slots[0]; 251023d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 251123d0b79dSTomohiro Misono 251223d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 251323d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 251423d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 251523d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 251623d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 251723d0b79dSTomohiro Misono ret = -EINVAL; 251823d0b79dSTomohiro Misono goto out; 251923d0b79dSTomohiro Misono } 252023d0b79dSTomohiro Misono 252123d0b79dSTomohiro Misono /* Copy subvolume's name */ 252223d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 252323d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 252423d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 252523d0b79dSTomohiro Misono args->name[item_len] = 0; 252623d0b79dSTomohiro Misono 252723d0b79dSTomohiro Misono out: 252823d0b79dSTomohiro Misono btrfs_free_path(path); 252923d0b79dSTomohiro Misono return ret; 253023d0b79dSTomohiro Misono } 253123d0b79dSTomohiro Misono 2532ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2533ac8e9819SChris Mason void __user *argp) 2534ac8e9819SChris Mason { 2535ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2536ac8e9819SChris Mason struct inode *inode; 253701b810b8SDavid Sterba int ret = 0; 2538ac8e9819SChris Mason 25392354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25402354d08fSJulia Lawall if (IS_ERR(args)) 25412354d08fSJulia Lawall return PTR_ERR(args); 2542c2b96929SDan Carpenter 2543496ad9aaSAl Viro inode = file_inode(file); 2544ac8e9819SChris Mason 254501b810b8SDavid Sterba /* 254601b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 254701b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 254801b810b8SDavid Sterba */ 25491b53ac4dSChris Mason if (args->treeid == 0) 25501b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25511b53ac4dSChris Mason 255201b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 255301b810b8SDavid Sterba args->name[0] = 0; 255401b810b8SDavid Sterba goto out; 255501b810b8SDavid Sterba } 255601b810b8SDavid Sterba 255701b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 255801b810b8SDavid Sterba ret = -EPERM; 255901b810b8SDavid Sterba goto out; 256001b810b8SDavid Sterba } 256101b810b8SDavid Sterba 2562ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2563ac8e9819SChris Mason args->treeid, args->objectid, 2564ac8e9819SChris Mason args->name); 2565ac8e9819SChris Mason 256601b810b8SDavid Sterba out: 2567ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2568ac8e9819SChris Mason ret = -EFAULT; 2569ac8e9819SChris Mason 2570ac8e9819SChris Mason kfree(args); 257198d377a0STARUISI Hiroaki return ret; 257298d377a0STARUISI Hiroaki } 257398d377a0STARUISI Hiroaki 257423d0b79dSTomohiro Misono /* 257523d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 257623d0b79dSTomohiro Misono * 257723d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 257823d0b79dSTomohiro Misono * 257923d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 258023d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 258123d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 258223d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 258323d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 258423d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 258523d0b79dSTomohiro Misono */ 258623d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 258723d0b79dSTomohiro Misono { 258823d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 258923d0b79dSTomohiro Misono struct inode *inode; 259023d0b79dSTomohiro Misono int ret; 259123d0b79dSTomohiro Misono 259223d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 259323d0b79dSTomohiro Misono if (IS_ERR(args)) 259423d0b79dSTomohiro Misono return PTR_ERR(args); 259523d0b79dSTomohiro Misono 259623d0b79dSTomohiro Misono inode = file_inode(file); 259723d0b79dSTomohiro Misono 259823d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 259923d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 260023d0b79dSTomohiro Misono /* 260123d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 260223d0b79dSTomohiro Misono * called 260323d0b79dSTomohiro Misono */ 260423d0b79dSTomohiro Misono kfree(args); 260523d0b79dSTomohiro Misono return -EACCES; 260623d0b79dSTomohiro Misono } 260723d0b79dSTomohiro Misono 260823d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 260923d0b79dSTomohiro Misono 261023d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 261123d0b79dSTomohiro Misono ret = -EFAULT; 261223d0b79dSTomohiro Misono 261323d0b79dSTomohiro Misono kfree(args); 261423d0b79dSTomohiro Misono return ret; 261523d0b79dSTomohiro Misono } 261623d0b79dSTomohiro Misono 2617b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2618b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2619b64ec075STomohiro Misono { 2620b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2621b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2622b64ec075STomohiro Misono struct btrfs_root *root; 2623b64ec075STomohiro Misono struct btrfs_path *path; 2624b64ec075STomohiro Misono struct btrfs_key key; 2625b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2626b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2627b64ec075STomohiro Misono struct extent_buffer *leaf; 2628b64ec075STomohiro Misono unsigned long item_off; 2629b64ec075STomohiro Misono unsigned long item_len; 2630b64ec075STomohiro Misono struct inode *inode; 2631b64ec075STomohiro Misono int slot; 2632b64ec075STomohiro Misono int ret = 0; 2633b64ec075STomohiro Misono 2634b64ec075STomohiro Misono path = btrfs_alloc_path(); 2635b64ec075STomohiro Misono if (!path) 2636b64ec075STomohiro Misono return -ENOMEM; 2637b64ec075STomohiro Misono 2638b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2639b64ec075STomohiro Misono if (!subvol_info) { 2640b64ec075STomohiro Misono btrfs_free_path(path); 2641b64ec075STomohiro Misono return -ENOMEM; 2642b64ec075STomohiro Misono } 2643b64ec075STomohiro Misono 2644b64ec075STomohiro Misono inode = file_inode(file); 2645b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2646b64ec075STomohiro Misono 2647b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2648b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2649b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2650b64ec075STomohiro Misono key.offset = (u64)-1; 2651b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2652b64ec075STomohiro Misono if (IS_ERR(root)) { 2653b64ec075STomohiro Misono ret = PTR_ERR(root); 2654b64ec075STomohiro Misono goto out; 2655b64ec075STomohiro Misono } 2656b64ec075STomohiro Misono root_item = &root->root_item; 2657b64ec075STomohiro Misono 2658b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2659b64ec075STomohiro Misono 2660b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2661b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2662b64ec075STomohiro Misono 2663b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2664b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2665b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2666b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2667b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2668b64ec075STomohiro Misono 2669b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2670b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2671b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2672b64ec075STomohiro Misono 2673b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2674b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2675b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2676b64ec075STomohiro Misono 2677b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2678b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2679b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2680b64ec075STomohiro Misono 2681b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2682b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2683b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2684b64ec075STomohiro Misono 2685b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2686b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2687b64ec075STomohiro Misono root = fs_info->tree_root; 2688b64ec075STomohiro Misono 2689b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2690b64ec075STomohiro Misono key.offset = 0; 2691b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2692b64ec075STomohiro Misono if (ret < 0) { 2693b64ec075STomohiro Misono goto out; 2694b64ec075STomohiro Misono } else if (path->slots[0] >= 2695b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2696b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2697b64ec075STomohiro Misono if (ret < 0) { 2698b64ec075STomohiro Misono goto out; 2699b64ec075STomohiro Misono } else if (ret > 0) { 2700b64ec075STomohiro Misono ret = -EUCLEAN; 2701b64ec075STomohiro Misono goto out; 2702b64ec075STomohiro Misono } 2703b64ec075STomohiro Misono } 2704b64ec075STomohiro Misono 2705b64ec075STomohiro Misono leaf = path->nodes[0]; 2706b64ec075STomohiro Misono slot = path->slots[0]; 2707b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2708b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2709b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2710b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2711b64ec075STomohiro Misono 2712b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2713b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2714b64ec075STomohiro Misono 2715b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2716b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2717b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2718b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2719b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2720b64ec075STomohiro Misono item_off, item_len); 2721b64ec075STomohiro Misono } else { 2722b64ec075STomohiro Misono ret = -ENOENT; 2723b64ec075STomohiro Misono goto out; 2724b64ec075STomohiro Misono } 2725b64ec075STomohiro Misono } 2726b64ec075STomohiro Misono 2727b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2728b64ec075STomohiro Misono ret = -EFAULT; 2729b64ec075STomohiro Misono 2730b64ec075STomohiro Misono out: 2731b64ec075STomohiro Misono btrfs_free_path(path); 2732b64ec075STomohiro Misono kzfree(subvol_info); 2733b64ec075STomohiro Misono return ret; 2734b64ec075STomohiro Misono } 2735b64ec075STomohiro Misono 273642e4b520STomohiro Misono /* 273742e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 273842e4b520STomohiro Misono * except the subvolume name. 273942e4b520STomohiro Misono */ 274042e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 274142e4b520STomohiro Misono { 274242e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 274342e4b520STomohiro Misono struct btrfs_root_ref *rref; 274442e4b520STomohiro Misono struct btrfs_root *root; 274542e4b520STomohiro Misono struct btrfs_path *path; 274642e4b520STomohiro Misono struct btrfs_key key; 274742e4b520STomohiro Misono struct extent_buffer *leaf; 274842e4b520STomohiro Misono struct inode *inode; 274942e4b520STomohiro Misono u64 objectid; 275042e4b520STomohiro Misono int slot; 275142e4b520STomohiro Misono int ret; 275242e4b520STomohiro Misono u8 found; 275342e4b520STomohiro Misono 275442e4b520STomohiro Misono path = btrfs_alloc_path(); 275542e4b520STomohiro Misono if (!path) 275642e4b520STomohiro Misono return -ENOMEM; 275742e4b520STomohiro Misono 275842e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 275942e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 276042e4b520STomohiro Misono btrfs_free_path(path); 276142e4b520STomohiro Misono return PTR_ERR(rootrefs); 276242e4b520STomohiro Misono } 276342e4b520STomohiro Misono 276442e4b520STomohiro Misono inode = file_inode(file); 276542e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 276642e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 276742e4b520STomohiro Misono 276842e4b520STomohiro Misono key.objectid = objectid; 276942e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 277042e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 277142e4b520STomohiro Misono found = 0; 277242e4b520STomohiro Misono 277342e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 277442e4b520STomohiro Misono if (ret < 0) { 277542e4b520STomohiro Misono goto out; 277642e4b520STomohiro Misono } else if (path->slots[0] >= 277742e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 277842e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 277942e4b520STomohiro Misono if (ret < 0) { 278042e4b520STomohiro Misono goto out; 278142e4b520STomohiro Misono } else if (ret > 0) { 278242e4b520STomohiro Misono ret = -EUCLEAN; 278342e4b520STomohiro Misono goto out; 278442e4b520STomohiro Misono } 278542e4b520STomohiro Misono } 278642e4b520STomohiro Misono while (1) { 278742e4b520STomohiro Misono leaf = path->nodes[0]; 278842e4b520STomohiro Misono slot = path->slots[0]; 278942e4b520STomohiro Misono 279042e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 279142e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 279242e4b520STomohiro Misono ret = 0; 279342e4b520STomohiro Misono goto out; 279442e4b520STomohiro Misono } 279542e4b520STomohiro Misono 279642e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 279742e4b520STomohiro Misono ret = -EOVERFLOW; 279842e4b520STomohiro Misono goto out; 279942e4b520STomohiro Misono } 280042e4b520STomohiro Misono 280142e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 280242e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 280342e4b520STomohiro Misono rootrefs->rootref[found].dirid = 280442e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 280542e4b520STomohiro Misono found++; 280642e4b520STomohiro Misono 280742e4b520STomohiro Misono ret = btrfs_next_item(root, path); 280842e4b520STomohiro Misono if (ret < 0) { 280942e4b520STomohiro Misono goto out; 281042e4b520STomohiro Misono } else if (ret > 0) { 281142e4b520STomohiro Misono ret = -EUCLEAN; 281242e4b520STomohiro Misono goto out; 281342e4b520STomohiro Misono } 281442e4b520STomohiro Misono } 281542e4b520STomohiro Misono 281642e4b520STomohiro Misono out: 281742e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 281842e4b520STomohiro Misono rootrefs->num_items = found; 281942e4b520STomohiro Misono /* update min_treeid for next search */ 282042e4b520STomohiro Misono if (found) 282142e4b520STomohiro Misono rootrefs->min_treeid = 282242e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 282342e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 282442e4b520STomohiro Misono ret = -EFAULT; 282542e4b520STomohiro Misono } 282642e4b520STomohiro Misono 282742e4b520STomohiro Misono kfree(rootrefs); 282842e4b520STomohiro Misono btrfs_free_path(path); 282942e4b520STomohiro Misono 283042e4b520STomohiro Misono return ret; 283142e4b520STomohiro Misono } 283242e4b520STomohiro Misono 283376dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 283476dda93cSYan, Zheng void __user *arg) 283576dda93cSYan, Zheng { 283654563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28370b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 283876dda93cSYan, Zheng struct dentry *dentry; 28392b0143b5SDavid Howells struct inode *dir = d_inode(parent); 284076dda93cSYan, Zheng struct inode *inode; 284176dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 284276dda93cSYan, Zheng struct btrfs_root *dest = NULL; 284376dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 284476dda93cSYan, Zheng int namelen; 284576dda93cSYan, Zheng int err = 0; 284676dda93cSYan, Zheng 2847325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2848325c50e3SJeff Mahoney return -ENOTDIR; 2849325c50e3SJeff Mahoney 285076dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 285176dda93cSYan, Zheng if (IS_ERR(vol_args)) 285276dda93cSYan, Zheng return PTR_ERR(vol_args); 285376dda93cSYan, Zheng 285476dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 285576dda93cSYan, Zheng namelen = strlen(vol_args->name); 285676dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 285776dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 285876dda93cSYan, Zheng err = -EINVAL; 285976dda93cSYan, Zheng goto out; 286076dda93cSYan, Zheng } 286176dda93cSYan, Zheng 2862a561be71SAl Viro err = mnt_want_write_file(file); 286376dda93cSYan, Zheng if (err) 286476dda93cSYan, Zheng goto out; 286576dda93cSYan, Zheng 2866521e0546SDavid Sterba 286700235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 286800235411SAl Viro if (err == -EINTR) 286900235411SAl Viro goto out_drop_write; 287076dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 287176dda93cSYan, Zheng if (IS_ERR(dentry)) { 287276dda93cSYan, Zheng err = PTR_ERR(dentry); 287376dda93cSYan, Zheng goto out_unlock_dir; 287476dda93cSYan, Zheng } 287576dda93cSYan, Zheng 28762b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 287776dda93cSYan, Zheng err = -ENOENT; 287876dda93cSYan, Zheng goto out_dput; 287976dda93cSYan, Zheng } 288076dda93cSYan, Zheng 28812b0143b5SDavid Howells inode = d_inode(dentry); 28824260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28834260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28844260f7c7SSage Weil /* 28854260f7c7SSage Weil * Regular user. Only allow this with a special mount 28864260f7c7SSage Weil * option, when the user has write+exec access to the 28874260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28884260f7c7SSage Weil * allowed. 28894260f7c7SSage Weil * 28904260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28914260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28924260f7c7SSage Weil * otherwise be able to delete. 28934260f7c7SSage Weil * 28944260f7c7SSage Weil * Users who want to delete empty subvols should try 28954260f7c7SSage Weil * rmdir(2). 28964260f7c7SSage Weil */ 28974260f7c7SSage Weil err = -EPERM; 28980b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 28994260f7c7SSage Weil goto out_dput; 29004260f7c7SSage Weil 29014260f7c7SSage Weil /* 29024260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 29034260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 29044260f7c7SSage Weil * must be called on the dentry referencing the root 29054260f7c7SSage Weil * of the subvol, not a random directory contained 29064260f7c7SSage Weil * within it. 29074260f7c7SSage Weil */ 29084260f7c7SSage Weil err = -EINVAL; 29094260f7c7SSage Weil if (root == dest) 29104260f7c7SSage Weil goto out_dput; 29114260f7c7SSage Weil 29124260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 29134260f7c7SSage Weil if (err) 29144260f7c7SSage Weil goto out_dput; 29155c39da5bSMiao Xie } 29164260f7c7SSage Weil 29175c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 29184260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 29194260f7c7SSage Weil if (err) 29204260f7c7SSage Weil goto out_dput; 29214260f7c7SSage Weil 29224a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 292376dda93cSYan, Zheng err = -EINVAL; 292476dda93cSYan, Zheng goto out_dput; 292576dda93cSYan, Zheng } 292676dda93cSYan, Zheng 29275955102cSAl Viro inode_lock(inode); 2928f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 29295955102cSAl Viro inode_unlock(inode); 2930f60a2364SMisono Tomohiro if (!err) 293176dda93cSYan, Zheng d_delete(dentry); 2932fa6ac876SLiu Bo 293376dda93cSYan, Zheng out_dput: 293476dda93cSYan, Zheng dput(dentry); 293576dda93cSYan, Zheng out_unlock_dir: 29365955102cSAl Viro inode_unlock(dir); 293700235411SAl Viro out_drop_write: 29382a79f17eSAl Viro mnt_drop_write_file(file); 293976dda93cSYan, Zheng out: 294076dda93cSYan, Zheng kfree(vol_args); 294176dda93cSYan, Zheng return err; 294276dda93cSYan, Zheng } 294376dda93cSYan, Zheng 29441e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2945f46b5a66SChristoph Hellwig { 2946496ad9aaSAl Viro struct inode *inode = file_inode(file); 2947f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29481e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2949c146afadSYan Zheng int ret; 2950c146afadSYan Zheng 295125122d15SIlya Dryomov ret = mnt_want_write_file(file); 295225122d15SIlya Dryomov if (ret) 295325122d15SIlya Dryomov return ret; 2954b83cc969SLi Zefan 295525122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 295625122d15SIlya Dryomov ret = -EROFS; 295725122d15SIlya Dryomov goto out; 29585ac00addSStefan Behrens } 2959f46b5a66SChristoph Hellwig 2960f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2961f46b5a66SChristoph Hellwig case S_IFDIR: 2962e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2963e441d54dSChris Mason ret = -EPERM; 2964e441d54dSChris Mason goto out; 2965e441d54dSChris Mason } 2966de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2967f46b5a66SChristoph Hellwig break; 2968f46b5a66SChristoph Hellwig case S_IFREG: 2969616d374eSAdam Borowski /* 2970616d374eSAdam Borowski * Note that this does not check the file descriptor for write 2971616d374eSAdam Borowski * access. This prevents defragmenting executables that are 2972616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 2973616d374eSAdam Borowski */ 2974616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 2975616d374eSAdam Borowski inode_permission(inode, MAY_WRITE)) { 2976616d374eSAdam Borowski ret = -EPERM; 2977e441d54dSChris Mason goto out; 2978e441d54dSChris Mason } 29791e701a32SChris Mason 29801e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29811e701a32SChris Mason if (!range) { 29821e701a32SChris Mason ret = -ENOMEM; 29831e701a32SChris Mason goto out; 29841e701a32SChris Mason } 29851e701a32SChris Mason 29861e701a32SChris Mason if (argp) { 29871e701a32SChris Mason if (copy_from_user(range, argp, 29881e701a32SChris Mason sizeof(*range))) { 29891e701a32SChris Mason ret = -EFAULT; 29901e701a32SChris Mason kfree(range); 2991683be16eSDan Carpenter goto out; 29921e701a32SChris Mason } 29931e701a32SChris Mason /* compression requires us to start the IO */ 29941e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 29951e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 29961e701a32SChris Mason range->extent_thresh = (u32)-1; 29971e701a32SChris Mason } 29981e701a32SChris Mason } else { 29991e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 30001e701a32SChris Mason range->len = (u64)-1; 30011e701a32SChris Mason } 3002496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 30037c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 30044cb5300bSChris Mason if (ret > 0) 30054cb5300bSChris Mason ret = 0; 30061e701a32SChris Mason kfree(range); 3007f46b5a66SChristoph Hellwig break; 30088929ecfaSYan, Zheng default: 30098929ecfaSYan, Zheng ret = -EINVAL; 3010f46b5a66SChristoph Hellwig } 3011e441d54dSChris Mason out: 301225122d15SIlya Dryomov mnt_drop_write_file(file); 3013e441d54dSChris Mason return ret; 3014f46b5a66SChristoph Hellwig } 3015f46b5a66SChristoph Hellwig 30162ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3017f46b5a66SChristoph Hellwig { 3018f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3019f46b5a66SChristoph Hellwig int ret; 3020f46b5a66SChristoph Hellwig 3021e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3022e441d54dSChris Mason return -EPERM; 3023e441d54dSChris Mason 3024171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 3025e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3026c9e9f97bSIlya Dryomov 3027dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3028c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3029c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3030c9e9f97bSIlya Dryomov goto out; 3031c9e9f97bSIlya Dryomov } 3032f46b5a66SChristoph Hellwig 30335516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30342ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3035f46b5a66SChristoph Hellwig 303643d20761SAnand Jain if (!ret) 30370b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 303843d20761SAnand Jain 3039f46b5a66SChristoph Hellwig kfree(vol_args); 3040c9e9f97bSIlya Dryomov out: 3041171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3042f46b5a66SChristoph Hellwig return ret; 3043f46b5a66SChristoph Hellwig } 3044f46b5a66SChristoph Hellwig 30456b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3046f46b5a66SChristoph Hellwig { 30470b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30480b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30496b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3050f46b5a66SChristoph Hellwig int ret; 3051f46b5a66SChristoph Hellwig 3052e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3053e441d54dSChris Mason return -EPERM; 3054e441d54dSChris Mason 3055da24927bSMiao Xie ret = mnt_want_write_file(file); 3056da24927bSMiao Xie if (ret) 3057da24927bSMiao Xie return ret; 3058c146afadSYan Zheng 3059dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3060c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3061c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3062c47ca32dSDan Carpenter goto err_drop; 3063c9e9f97bSIlya Dryomov } 3064f46b5a66SChristoph Hellwig 30656b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3066fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3067fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3068fd4e994bSOmar Sandoval goto out; 3069fd4e994bSOmar Sandoval } 3070f46b5a66SChristoph Hellwig 3071171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3072183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3073183860f6SAnand Jain goto out; 3074183860f6SAnand Jain } 3075183860f6SAnand Jain 3076735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30772ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30786b526ed7SAnand Jain } else { 30796b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30802ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30816b526ed7SAnand Jain } 3082171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3083183860f6SAnand Jain 30846b526ed7SAnand Jain if (!ret) { 3085735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30860b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30876b526ed7SAnand Jain vol_args->devid); 30886b526ed7SAnand Jain else 30890b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30906b526ed7SAnand Jain vol_args->name); 30916b526ed7SAnand Jain } 3092183860f6SAnand Jain out: 3093183860f6SAnand Jain kfree(vol_args); 3094c47ca32dSDan Carpenter err_drop: 30954ac20c70SIlya Dryomov mnt_drop_write_file(file); 3096f46b5a66SChristoph Hellwig return ret; 3097f46b5a66SChristoph Hellwig } 3098f46b5a66SChristoph Hellwig 3099f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3100f46b5a66SChristoph Hellwig { 31010b246afaSJeff Mahoney struct inode *inode = file_inode(file); 31020b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3103f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3104f46b5a66SChristoph Hellwig int ret; 3105f46b5a66SChristoph Hellwig 3106f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3107f46b5a66SChristoph Hellwig return -EPERM; 3108f46b5a66SChristoph Hellwig 3109f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3110f46b5a66SChristoph Hellwig if (ret) 3111f46b5a66SChristoph Hellwig return ret; 3112f46b5a66SChristoph Hellwig 3113171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3114f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 311558d7bbf8SDavid Sterba goto out_drop_write; 311658d7bbf8SDavid Sterba } 311758d7bbf8SDavid Sterba 311858d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 311958d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 312058d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3121f46b5a66SChristoph Hellwig goto out; 3122f46b5a66SChristoph Hellwig } 3123f46b5a66SChristoph Hellwig 312458d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31252ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3126f46b5a66SChristoph Hellwig 3127f46b5a66SChristoph Hellwig if (!ret) 31280b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3129f46b5a66SChristoph Hellwig kfree(vol_args); 313058d7bbf8SDavid Sterba out: 3131171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 313258d7bbf8SDavid Sterba out_drop_write: 3133f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 313458d7bbf8SDavid Sterba 3135f46b5a66SChristoph Hellwig return ret; 3136f46b5a66SChristoph Hellwig } 3137f46b5a66SChristoph Hellwig 31382ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 31392ff7e61eSJeff Mahoney void __user *arg) 3140475f6387SJan Schmidt { 3141027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3142475f6387SJan Schmidt struct btrfs_device *device; 31430b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3144027ed2f0SLi Zefan int ret = 0; 3145475f6387SJan Schmidt 3146027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3147027ed2f0SLi Zefan if (!fi_args) 3148027ed2f0SLi Zefan return -ENOMEM; 3149027ed2f0SLi Zefan 3150d03262c7SDavid Sterba rcu_read_lock(); 3151027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3152475f6387SJan Schmidt 3153d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3154027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3155027ed2f0SLi Zefan fi_args->max_id = device->devid; 3156475f6387SJan Schmidt } 3157d03262c7SDavid Sterba rcu_read_unlock(); 3158475f6387SJan Schmidt 3159de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3160bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3161bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3162bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 316380a773fbSDavid Sterba 3164027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3165027ed2f0SLi Zefan ret = -EFAULT; 3166475f6387SJan Schmidt 3167027ed2f0SLi Zefan kfree(fi_args); 3168027ed2f0SLi Zefan return ret; 3169475f6387SJan Schmidt } 3170475f6387SJan Schmidt 31712ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31722ff7e61eSJeff Mahoney void __user *arg) 3173475f6387SJan Schmidt { 3174475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3175475f6387SJan Schmidt struct btrfs_device *dev; 3176475f6387SJan Schmidt int ret = 0; 3177475f6387SJan Schmidt char *s_uuid = NULL; 3178475f6387SJan Schmidt 3179475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3180475f6387SJan Schmidt if (IS_ERR(di_args)) 3181475f6387SJan Schmidt return PTR_ERR(di_args); 3182475f6387SJan Schmidt 3183dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3184475f6387SJan Schmidt s_uuid = di_args->uuid; 3185475f6387SJan Schmidt 3186c5593ca3SDavid Sterba rcu_read_lock(); 3187e4319cd9SAnand Jain dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid, 318809ba3bc9SAnand Jain NULL, true); 3189475f6387SJan Schmidt 3190475f6387SJan Schmidt if (!dev) { 3191475f6387SJan Schmidt ret = -ENODEV; 3192475f6387SJan Schmidt goto out; 3193475f6387SJan Schmidt } 3194475f6387SJan Schmidt 3195475f6387SJan Schmidt di_args->devid = dev->devid; 31967cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 31977cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3198475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3199a27202fbSJim Meyering if (dev->name) { 3200672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3201672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3202a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3203a27202fbSJim Meyering } else { 320499ba55adSStefan Behrens di_args->path[0] = '\0'; 3205a27202fbSJim Meyering } 3206475f6387SJan Schmidt 3207475f6387SJan Schmidt out: 3208c5593ca3SDavid Sterba rcu_read_unlock(); 3209475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3210475f6387SJan Schmidt ret = -EFAULT; 3211475f6387SJan Schmidt 3212475f6387SJan Schmidt kfree(di_args); 3213475f6387SJan Schmidt return ret; 3214475f6387SJan Schmidt } 3215475f6387SJan Schmidt 3216d8b55242SFilipe Manana static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3217d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3218d8b55242SFilipe Manana { 3219d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3220d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3221d8b55242SFilipe Manana } 3222d8b55242SFilipe Manana 3223d8b55242SFilipe Manana static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3224d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3225d8b55242SFilipe Manana { 3226d8b55242SFilipe Manana if (inode1 < inode2) { 3227d8b55242SFilipe Manana swap(inode1, inode2); 3228d8b55242SFilipe Manana swap(loff1, loff2); 3229d8b55242SFilipe Manana } else if (inode1 == inode2 && loff2 < loff1) { 3230d8b55242SFilipe Manana swap(loff1, loff2); 3231d8b55242SFilipe Manana } 3232d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3233d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3234d8b55242SFilipe Manana } 3235d8b55242SFilipe Manana 323657a50e25SFilipe Manana static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len, 323734a28e3dSFilipe Manana struct inode *dst, u64 dst_loff) 3238416161dbSMark Fasheh { 3239416161dbSMark Fasheh int ret; 32400efa9f48SMark Fasheh 3241de02b9f6SFilipe Manana /* 3242d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3243d8b55242SFilipe Manana * source range to serialize with relocation. 3244de02b9f6SFilipe Manana */ 3245d8b55242SFilipe Manana btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 324657a50e25SFilipe Manana ret = btrfs_clone(src, dst, loff, len, len, dst_loff, 1); 3247d8b55242SFilipe Manana btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 32483973909dSTimofey Titovets 32493973909dSTimofey Titovets return ret; 32503973909dSTimofey Titovets } 32513973909dSTimofey Titovets 3252b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3253b6728768STimofey Titovets 32543973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 32553973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 32563973909dSTimofey Titovets { 32573973909dSTimofey Titovets int ret; 3258b6728768STimofey Titovets u64 i, tail_len, chunk_count; 325962d54f3aSFilipe Manana struct btrfs_root *root_dst = BTRFS_I(dst)->root; 326062d54f3aSFilipe Manana 326162d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 326262d54f3aSFilipe Manana if (root_dst->send_in_progress) { 326362d54f3aSFilipe Manana btrfs_warn_rl(root_dst->fs_info, 326462d54f3aSFilipe Manana "cannot deduplicate to root %llu while send operations are using it (%d in progress)", 326562d54f3aSFilipe Manana root_dst->root_key.objectid, 326662d54f3aSFilipe Manana root_dst->send_in_progress); 326762d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 326862d54f3aSFilipe Manana return -EAGAIN; 326962d54f3aSFilipe Manana } 327062d54f3aSFilipe Manana root_dst->dedupe_in_progress++; 327162d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 32723973909dSTimofey Titovets 3273b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3274b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 327567b07bd4STimofey Titovets 3276b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3277b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 327834a28e3dSFilipe Manana dst, dst_loff); 3279b6728768STimofey Titovets if (ret) 328062d54f3aSFilipe Manana goto out; 3281b6728768STimofey Titovets 3282b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3283b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3284b6728768STimofey Titovets } 3285b6728768STimofey Titovets 3286b6728768STimofey Titovets if (tail_len > 0) 328767b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 328834a28e3dSFilipe Manana dst_loff); 328962d54f3aSFilipe Manana out: 329062d54f3aSFilipe Manana spin_lock(&root_dst->root_item_lock); 329162d54f3aSFilipe Manana root_dst->dedupe_in_progress--; 329262d54f3aSFilipe Manana spin_unlock(&root_dst->root_item_lock); 3293416161dbSMark Fasheh 3294416161dbSMark Fasheh return ret; 3295416161dbSMark Fasheh } 3296416161dbSMark Fasheh 3297f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3298f82a9901SFilipe Manana struct inode *inode, 3299f82a9901SFilipe Manana u64 endoff, 3300f82a9901SFilipe Manana const u64 destoff, 33011c919a5eSMark Fasheh const u64 olen, 33021c919a5eSMark Fasheh int no_time_update) 3303f82a9901SFilipe Manana { 3304f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3305f82a9901SFilipe Manana int ret; 3306f82a9901SFilipe Manana 3307f82a9901SFilipe Manana inode_inc_iversion(inode); 33081c919a5eSMark Fasheh if (!no_time_update) 3309c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3310f82a9901SFilipe Manana /* 3311f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3312f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3313f82a9901SFilipe Manana */ 3314f82a9901SFilipe Manana if (endoff > destoff + olen) 3315f82a9901SFilipe Manana endoff = destoff + olen; 3316f82a9901SFilipe Manana if (endoff > inode->i_size) 33176ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3318f82a9901SFilipe Manana 3319f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3320f82a9901SFilipe Manana if (ret) { 332166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 33223a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3323f82a9901SFilipe Manana goto out; 3324f82a9901SFilipe Manana } 33253a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3326f82a9901SFilipe Manana out: 3327f82a9901SFilipe Manana return ret; 3328f82a9901SFilipe Manana } 3329f82a9901SFilipe Manana 3330a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 33317ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 33327ffbb598SFilipe Manana const struct btrfs_path *path, 33337ffbb598SFilipe Manana const u64 hole_offset, 33347ffbb598SFilipe Manana const u64 hole_len) 33357ffbb598SFilipe Manana { 3336a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 33377ffbb598SFilipe Manana struct extent_map *em; 33387ffbb598SFilipe Manana int ret; 33397ffbb598SFilipe Manana 33407ffbb598SFilipe Manana em = alloc_extent_map(); 33417ffbb598SFilipe Manana if (!em) { 3342a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33437ffbb598SFilipe Manana return; 33447ffbb598SFilipe Manana } 33457ffbb598SFilipe Manana 334614f59796SFilipe Manana if (path) { 334714f59796SFilipe Manana struct btrfs_file_extent_item *fi; 334814f59796SFilipe Manana 334914f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 335014f59796SFilipe Manana struct btrfs_file_extent_item); 3351a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 33527ffbb598SFilipe Manana em->generation = -1; 33537ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 33547ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33557ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3356a2f392e4SNikolay Borisov &inode->runtime_flags); 33577ffbb598SFilipe Manana } else { 33587ffbb598SFilipe Manana em->start = hole_offset; 33597ffbb598SFilipe Manana em->len = hole_len; 33607ffbb598SFilipe Manana em->ram_bytes = em->len; 33617ffbb598SFilipe Manana em->orig_start = hole_offset; 33627ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33637ffbb598SFilipe Manana em->block_len = 0; 33647ffbb598SFilipe Manana em->orig_block_len = 0; 33657ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33667ffbb598SFilipe Manana em->generation = trans->transid; 33677ffbb598SFilipe Manana } 33687ffbb598SFilipe Manana 33697ffbb598SFilipe Manana while (1) { 33707ffbb598SFilipe Manana write_lock(&em_tree->lock); 33717ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33727ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33737ffbb598SFilipe Manana if (ret != -EEXIST) { 33747ffbb598SFilipe Manana free_extent_map(em); 33757ffbb598SFilipe Manana break; 33767ffbb598SFilipe Manana } 3377a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 33787ffbb598SFilipe Manana em->start + em->len - 1, 0); 33797ffbb598SFilipe Manana } 33807ffbb598SFilipe Manana 3381ee39b432SDavid Sterba if (ret) 3382a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33837ffbb598SFilipe Manana } 33847ffbb598SFilipe Manana 33858039d87dSFilipe Manana /* 33868039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33878039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33888039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33898039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33908039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33918039d87dSFilipe Manana * 33928039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33938039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33948039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33958039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33968039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33978039d87dSFilipe Manana * i_mutex, which we are holding here). 33988039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33998039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 34008039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 34018039d87dSFilipe Manana * 34028039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 34038039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 34048039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 34058039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 34068039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 34078039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 34088039d87dSFilipe Manana * the former. 34098039d87dSFilipe Manana */ 34104a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 34118039d87dSFilipe Manana struct btrfs_trans_handle *trans, 34128039d87dSFilipe Manana struct btrfs_path *path, 34138039d87dSFilipe Manana struct btrfs_key *new_key, 34148039d87dSFilipe Manana const u64 drop_start, 34158039d87dSFilipe Manana const u64 datal, 34168039d87dSFilipe Manana const u64 skip, 34178039d87dSFilipe Manana const u64 size, 34188039d87dSFilipe Manana char *inline_data) 34198039d87dSFilipe Manana { 34200b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 34218039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 34228039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 34230b246afaSJeff Mahoney fs_info->sectorsize); 34248039d87dSFilipe Manana int ret; 34258039d87dSFilipe Manana struct btrfs_key key; 34268039d87dSFilipe Manana 34278039d87dSFilipe Manana if (new_key->offset > 0) 34288039d87dSFilipe Manana return -EOPNOTSUPP; 34298039d87dSFilipe Manana 34304a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 34318039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 34328039d87dSFilipe Manana key.offset = 0; 34338039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 34348039d87dSFilipe Manana if (ret < 0) { 34358039d87dSFilipe Manana return ret; 34368039d87dSFilipe Manana } else if (ret > 0) { 34378039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 34388039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 34398039d87dSFilipe Manana if (ret < 0) 34408039d87dSFilipe Manana return ret; 34418039d87dSFilipe Manana else if (ret > 0) 34428039d87dSFilipe Manana goto copy_inline_extent; 34438039d87dSFilipe Manana } 34448039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 34454a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34468039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 34478039d87dSFilipe Manana ASSERT(key.offset > 0); 34488039d87dSFilipe Manana return -EOPNOTSUPP; 34498039d87dSFilipe Manana } 34508039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34518039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34528039d87dSFilipe Manana u64 ext_len; 34538039d87dSFilipe Manana 34548039d87dSFilipe Manana /* 34558039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34568039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34578039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34588039d87dSFilipe Manana */ 34598039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34608039d87dSFilipe Manana struct btrfs_file_extent_item); 34618039d87dSFilipe Manana /* 34628039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34638039d87dSFilipe Manana * following it. 34648039d87dSFilipe Manana */ 34658039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34668039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34678039d87dSFilipe Manana goto copy_inline_extent; 34688039d87dSFilipe Manana 34698039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34708039d87dSFilipe Manana if (ext_len > aligned_end) 34718039d87dSFilipe Manana return -EOPNOTSUPP; 34728039d87dSFilipe Manana 34738039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34748039d87dSFilipe Manana if (ret < 0) { 34758039d87dSFilipe Manana return ret; 34768039d87dSFilipe Manana } else if (ret == 0) { 34778039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34788039d87dSFilipe Manana path->slots[0]); 34794a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34808039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34818039d87dSFilipe Manana return -EOPNOTSUPP; 34828039d87dSFilipe Manana } 34838039d87dSFilipe Manana } 34848039d87dSFilipe Manana 34858039d87dSFilipe Manana copy_inline_extent: 34868039d87dSFilipe Manana /* 34878039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34888039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34898039d87dSFilipe Manana */ 34908039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34918039d87dSFilipe Manana /* 34928039d87dSFilipe Manana * If the destination inode has an inline extent... 34938039d87dSFilipe Manana * This would require copying the data from the source inline 34948039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34958039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34968039d87dSFilipe Manana * or just one of them, which would require decompressing and 34978039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34988039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34998039d87dSFilipe Manana * inline extent). 35008039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 35018039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 35028039d87dSFilipe Manana */ 35038039d87dSFilipe Manana return -EOPNOTSUPP; 35048039d87dSFilipe Manana } 35058039d87dSFilipe Manana 35068039d87dSFilipe Manana btrfs_release_path(path); 35078039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 35088039d87dSFilipe Manana if (ret) 35098039d87dSFilipe Manana return ret; 35108039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 35118039d87dSFilipe Manana if (ret) 35128039d87dSFilipe Manana return ret; 35138039d87dSFilipe Manana 35148039d87dSFilipe Manana if (skip) { 35158039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 35168039d87dSFilipe Manana 35178039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 35188039d87dSFilipe Manana } 35198039d87dSFilipe Manana 35208039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 35218039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 35228039d87dSFilipe Manana path->slots[0]), 35238039d87dSFilipe Manana size); 35248039d87dSFilipe Manana inode_add_bytes(dst, datal); 35258039d87dSFilipe Manana 35268039d87dSFilipe Manana return 0; 35278039d87dSFilipe Manana } 35288039d87dSFilipe Manana 352932b7c687SMark Fasheh /** 353032b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 353132b7c687SMark Fasheh * 353232b7c687SMark Fasheh * @src: Inode to clone from 353332b7c687SMark Fasheh * @inode: Inode to clone to 353432b7c687SMark Fasheh * @off: Offset within source to start clone from 353532b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 35361c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 353732b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 35381c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 353932b7c687SMark Fasheh */ 354032b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3541f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 35421c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3543f46b5a66SChristoph Hellwig { 35440b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3545f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 354632b7c687SMark Fasheh struct btrfs_path *path = NULL; 3547f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 354832b7c687SMark Fasheh struct btrfs_trans_handle *trans; 354932b7c687SMark Fasheh char *buf = NULL; 3550ae01a0abSYan Zheng struct btrfs_key key; 3551f46b5a66SChristoph Hellwig u32 nritems; 3552f46b5a66SChristoph Hellwig int slot; 3553ae01a0abSYan Zheng int ret; 3554f82a9901SFilipe Manana const u64 len = olen_aligned; 3555f82a9901SFilipe Manana u64 last_dest_end = destoff; 3556ae01a0abSYan Zheng 3557ae01a0abSYan Zheng ret = -ENOMEM; 3558752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3559ae01a0abSYan Zheng if (!buf) 356032b7c687SMark Fasheh return ret; 3561ae01a0abSYan Zheng 3562ae01a0abSYan Zheng path = btrfs_alloc_path(); 3563ae01a0abSYan Zheng if (!path) { 356415351955SDavid Sterba kvfree(buf); 356532b7c687SMark Fasheh return ret; 3566ae01a0abSYan Zheng } 356732b7c687SMark Fasheh 3568e4058b54SDavid Sterba path->reada = READA_FORWARD; 3569c5c9cd4dSSage Weil /* clone data */ 35704a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3571ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35722c463823SFilipe Manana key.offset = off; 3573f46b5a66SChristoph Hellwig 3574f46b5a66SChristoph Hellwig while (1) { 3575de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3576df858e76SFilipe Manana 3577f46b5a66SChristoph Hellwig /* 3578f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3579f46b5a66SChristoph Hellwig * tree. 3580f46b5a66SChristoph Hellwig */ 3581e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3582362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3583362a20c5SDavid Sterba 0, 0); 3584f46b5a66SChristoph Hellwig if (ret < 0) 3585f46b5a66SChristoph Hellwig goto out; 35862c463823SFilipe Manana /* 35872c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35882c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35892c463823SFilipe Manana * it might overlap our target range, therefore process it. 35902c463823SFilipe Manana */ 35912c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35922c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35932c463823SFilipe Manana path->slots[0] - 1); 35942c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35952c463823SFilipe Manana path->slots[0]--; 35962c463823SFilipe Manana } 3597f46b5a66SChristoph Hellwig 3598ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3599e4355f34SFilipe David Borba Manana process_slot: 3600ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3601362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3602f46b5a66SChristoph Hellwig if (ret < 0) 3603f46b5a66SChristoph Hellwig goto out; 3604f46b5a66SChristoph Hellwig if (ret > 0) 3605f46b5a66SChristoph Hellwig break; 3606ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3607f46b5a66SChristoph Hellwig } 3608f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3609f46b5a66SChristoph Hellwig slot = path->slots[0]; 3610f46b5a66SChristoph Hellwig 3611ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3612962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 36134a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3614f46b5a66SChristoph Hellwig break; 3615f46b5a66SChristoph Hellwig 3616962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3617c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3618c5c9cd4dSSage Weil int type; 361931840ae1SZheng Yan u32 size; 362031840ae1SZheng Yan struct btrfs_key new_key; 3621c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3622c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3623c5c9cd4dSSage Weil u8 comp; 3624f82a9901SFilipe Manana u64 drop_start; 362531840ae1SZheng Yan 3626c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3627c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3628c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3629c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3630c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3631c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3632d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3633d397712bSChris Mason extent); 3634d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3635d397712bSChris Mason extent); 3636c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3637d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3638d397712bSChris Mason extent); 3639c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3640c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3641c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3642c5c9cd4dSSage Weil extent); 3643c5c9cd4dSSage Weil } 364431840ae1SZheng Yan 36452c463823SFilipe Manana /* 36462c463823SFilipe Manana * The first search might have left us at an extent 36472c463823SFilipe Manana * item that ends before our target range's start, can 36482c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 36492c463823SFilipe Manana */ 36502c463823SFilipe Manana if (key.offset + datal <= off) { 3651e4355f34SFilipe David Borba Manana path->slots[0]++; 3652e4355f34SFilipe David Borba Manana goto process_slot; 36532c463823SFilipe Manana } else if (key.offset >= off + len) { 36542c463823SFilipe Manana break; 3655e4355f34SFilipe David Borba Manana } 3656df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3657e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3658e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3659e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3660e4355f34SFilipe David Borba Manana size); 3661e4355f34SFilipe David Borba Manana 3662e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3663e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3664c5c9cd4dSSage Weil 366531840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36664a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36674d728ec7SLi Zefan if (off <= key.offset) 3668c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36694d728ec7SLi Zefan else 36704d728ec7SLi Zefan new_key.offset = destoff; 3671c5c9cd4dSSage Weil 3672b6f3409bSSage Weil /* 3673f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3674f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3675f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3676f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3677f82a9901SFilipe Manana * partially overlaps it). 3678f82a9901SFilipe Manana */ 3679f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3680f82a9901SFilipe Manana drop_start = last_dest_end; 3681f82a9901SFilipe Manana else 3682f82a9901SFilipe Manana drop_start = new_key.offset; 3683f82a9901SFilipe Manana 3684f82a9901SFilipe Manana /* 3685b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3686b6f3409bSSage Weil * 1 - add new extent 3687b6f3409bSSage Weil * 1 - inode update 3688b6f3409bSSage Weil */ 3689b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3690a22285a6SYan, Zheng if (IS_ERR(trans)) { 3691a22285a6SYan, Zheng ret = PTR_ERR(trans); 3692a22285a6SYan, Zheng goto out; 3693a22285a6SYan, Zheng } 3694a22285a6SYan, Zheng 3695c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3696c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3697d72c0842SLi Zefan /* 3698d72c0842SLi Zefan * a | --- range to clone ---| b 3699d72c0842SLi Zefan * | ------------- extent ------------- | 3700d72c0842SLi Zefan */ 3701d72c0842SLi Zefan 370293915584SAntonio Ospite /* subtract range b */ 3703d72c0842SLi Zefan if (key.offset + datal > off + len) 3704d72c0842SLi Zefan datal = off + len - key.offset; 3705d72c0842SLi Zefan 370693915584SAntonio Ospite /* subtract range a */ 3707a22285a6SYan, Zheng if (off > key.offset) { 3708a22285a6SYan, Zheng datao += off - key.offset; 3709a22285a6SYan, Zheng datal -= off - key.offset; 3710a22285a6SYan, Zheng } 3711a22285a6SYan, Zheng 37125dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3713f82a9901SFilipe Manana drop_start, 3714a22285a6SYan, Zheng new_key.offset + datal, 37152671485dSJosef Bacik 1); 371679787eaaSJeff Mahoney if (ret) { 37173f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 371800fdf13aSLiu Bo btrfs_abort_transaction(trans, 371966642832SJeff Mahoney ret); 37203a45bb20SJeff Mahoney btrfs_end_transaction(trans); 372179787eaaSJeff Mahoney goto out; 372279787eaaSJeff Mahoney } 3723a22285a6SYan, Zheng 372431840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 372531840ae1SZheng Yan &new_key, size); 372679787eaaSJeff Mahoney if (ret) { 372766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 37283a45bb20SJeff Mahoney btrfs_end_transaction(trans); 372979787eaaSJeff Mahoney goto out; 373079787eaaSJeff Mahoney } 373131840ae1SZheng Yan 373231840ae1SZheng Yan leaf = path->nodes[0]; 373331840ae1SZheng Yan slot = path->slots[0]; 373431840ae1SZheng Yan write_extent_buffer(leaf, buf, 373531840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 373631840ae1SZheng Yan size); 3737ae01a0abSYan Zheng 3738f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3739f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3740c5c9cd4dSSage Weil 3741c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3742c5c9cd4dSSage Weil if (!disko) 3743c5c9cd4dSSage Weil datao = 0; 3744c5c9cd4dSSage Weil 3745c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3746c5c9cd4dSSage Weil datao); 3747c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3748c5c9cd4dSSage Weil datal); 3749fcebe456SJosef Bacik 3750c5c9cd4dSSage Weil if (disko) { 375182fa113fSQu Wenruo struct btrfs_ref ref = { 0 }; 3752c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 375382fa113fSQu Wenruo btrfs_init_generic_ref(&ref, 375482fa113fSQu Wenruo BTRFS_ADD_DELAYED_REF, disko, 375582fa113fSQu Wenruo diskl, 0); 375682fa113fSQu Wenruo btrfs_init_data_ref(&ref, 3757f46b5a66SChristoph Hellwig root->root_key.objectid, 37584a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 3759b06c4bf5SFilipe Manana new_key.offset - datao); 376082fa113fSQu Wenruo ret = btrfs_inc_extent_ref(trans, &ref); 376179787eaaSJeff Mahoney if (ret) { 376279787eaaSJeff Mahoney btrfs_abort_transaction(trans, 376379787eaaSJeff Mahoney ret); 37643a45bb20SJeff Mahoney btrfs_end_transaction(trans); 376579787eaaSJeff Mahoney goto out; 376679787eaaSJeff Mahoney 376779787eaaSJeff Mahoney } 3768f46b5a66SChristoph Hellwig } 3769c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3770c5c9cd4dSSage Weil u64 skip = 0; 3771c5c9cd4dSSage Weil u64 trim = 0; 3772ed958762SFilipe Manana 3773c5c9cd4dSSage Weil if (off > key.offset) { 3774c5c9cd4dSSage Weil skip = off - key.offset; 3775c5c9cd4dSSage Weil new_key.offset += skip; 377631840ae1SZheng Yan } 3777d397712bSChris Mason 3778c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3779c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3780d397712bSChris Mason 3781c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3782c5c9cd4dSSage Weil ret = -EINVAL; 37833a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3784c5c9cd4dSSage Weil goto out; 378531840ae1SZheng Yan } 3786c5c9cd4dSSage Weil size -= skip + trim; 3787c5c9cd4dSSage Weil datal -= skip + trim; 3788a22285a6SYan, Zheng 37894a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 37908039d87dSFilipe Manana trans, path, 37918039d87dSFilipe Manana &new_key, 3792f82a9901SFilipe Manana drop_start, 37938039d87dSFilipe Manana datal, 37948039d87dSFilipe Manana skip, size, buf); 379579787eaaSJeff Mahoney if (ret) { 37963f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37973a29bc09SChris Mason btrfs_abort_transaction(trans, 379879787eaaSJeff Mahoney ret); 37993a45bb20SJeff Mahoney btrfs_end_transaction(trans); 380079787eaaSJeff Mahoney goto out; 380179787eaaSJeff Mahoney } 3802c5c9cd4dSSage Weil leaf = path->nodes[0]; 3803c5c9cd4dSSage Weil slot = path->slots[0]; 3804c5c9cd4dSSage Weil } 3805c5c9cd4dSSage Weil 38067ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 38077ffbb598SFilipe Manana if (drop_start < new_key.offset) 3808a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 380914f59796SFilipe Manana NULL, drop_start, 38107ffbb598SFilipe Manana new_key.offset - drop_start); 38117ffbb598SFilipe Manana 3812a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 3813a2f392e4SNikolay Borisov path, 0, 0); 38147ffbb598SFilipe Manana 3815c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3816b3b4aa74SDavid Sterba btrfs_release_path(path); 3817c5c9cd4dSSage Weil 381862e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 38190b246afaSJeff Mahoney fs_info->sectorsize); 3820f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3821f82a9901SFilipe Manana last_dest_end, 38221c919a5eSMark Fasheh destoff, olen, 38231c919a5eSMark Fasheh no_time_update); 3824f82a9901SFilipe Manana if (ret) 382579787eaaSJeff Mahoney goto out; 38262c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 38272c463823SFilipe Manana break; 3828a22285a6SYan, Zheng } 3829b3b4aa74SDavid Sterba btrfs_release_path(path); 3830df858e76SFilipe Manana key.offset = next_key_min_offset; 383169ae5e44SWang Xiaoguang 383269ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 383369ae5e44SWang Xiaoguang ret = -EINTR; 383469ae5e44SWang Xiaoguang goto out; 383569ae5e44SWang Xiaoguang } 3836ae01a0abSYan Zheng } 3837f46b5a66SChristoph Hellwig ret = 0; 383832b7c687SMark Fasheh 3839f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3840f82a9901SFilipe Manana /* 3841f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3842f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3843f82a9901SFilipe Manana */ 3844b3b4aa74SDavid Sterba btrfs_release_path(path); 3845f82a9901SFilipe Manana 3846f82a9901SFilipe Manana /* 3847f82a9901SFilipe Manana * 1 - remove extent(s) 3848f82a9901SFilipe Manana * 1 - inode update 3849f82a9901SFilipe Manana */ 3850f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3851f82a9901SFilipe Manana if (IS_ERR(trans)) { 3852f82a9901SFilipe Manana ret = PTR_ERR(trans); 3853f82a9901SFilipe Manana goto out; 3854f82a9901SFilipe Manana } 3855f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3856f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3857f82a9901SFilipe Manana if (ret) { 3858f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 385966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 38603a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3861f82a9901SFilipe Manana goto out; 3862f82a9901SFilipe Manana } 3863a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 3864a2f392e4SNikolay Borisov last_dest_end, 386514f59796SFilipe Manana destoff + len - last_dest_end); 3866f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38671c919a5eSMark Fasheh destoff, olen, no_time_update); 3868f82a9901SFilipe Manana } 3869f82a9901SFilipe Manana 3870f46b5a66SChristoph Hellwig out: 387132b7c687SMark Fasheh btrfs_free_path(path); 387215351955SDavid Sterba kvfree(buf); 387332b7c687SMark Fasheh return ret; 387432b7c687SMark Fasheh } 387532b7c687SMark Fasheh 38763db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 387732b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 387832b7c687SMark Fasheh { 387954563d41SAl Viro struct inode *inode = file_inode(file); 38803db11b2eSZach Brown struct inode *src = file_inode(file_src); 38810b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 388232b7c687SMark Fasheh int ret; 388332b7c687SMark Fasheh u64 len = olen; 38840b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 388532b7c687SMark Fasheh 388632b7c687SMark Fasheh /* 388732b7c687SMark Fasheh * TODO: 388832b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 388932b7c687SMark Fasheh * decompress into destination's address_space (the file offset 389032b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 389132b7c687SMark Fasheh * otherwise reinsert) a subrange. 389200fdf13aSLiu Bo * 389300fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 389400fdf13aSLiu Bo * be either compressed or non-compressed. 389532b7c687SMark Fasheh */ 389632b7c687SMark Fasheh 3897ac765f83SFilipe Manana /* 389834a28e3dSFilipe Manana * VFS's generic_remap_file_range_prep() protects us from cloning the 389934a28e3dSFilipe Manana * eof block into the middle of a file, which would result in corruption 390034a28e3dSFilipe Manana * if the file size is not blocksize aligned. So we don't need to check 390134a28e3dSFilipe Manana * for that case here. 3902ac765f83SFilipe Manana */ 390334a28e3dSFilipe Manana if (off + len == src->i_size) 390432b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 390532b7c687SMark Fasheh 390632b7c687SMark Fasheh if (destoff > inode->i_size) { 3907f7fa1107SFilipe Manana const u64 wb_start = ALIGN_DOWN(inode->i_size, bs); 3908f7fa1107SFilipe Manana 390932b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 391032b7c687SMark Fasheh if (ret) 391134a28e3dSFilipe Manana return ret; 3912f7fa1107SFilipe Manana /* 3913f7fa1107SFilipe Manana * We may have truncated the last block if the inode's size is 3914f7fa1107SFilipe Manana * not sector size aligned, so we need to wait for writeback to 3915f7fa1107SFilipe Manana * complete before proceeding further, otherwise we can race 3916f7fa1107SFilipe Manana * with cloning and attempt to increment a reference to an 3917f7fa1107SFilipe Manana * extent that no longer exists (writeback completed right after 3918f7fa1107SFilipe Manana * we found the previous extent covering eof and before we 3919f7fa1107SFilipe Manana * attempted to increment its reference count). 3920f7fa1107SFilipe Manana */ 3921f7fa1107SFilipe Manana ret = btrfs_wait_ordered_range(inode, wb_start, 3922f7fa1107SFilipe Manana destoff - wb_start); 3923f7fa1107SFilipe Manana if (ret) 3924f7fa1107SFilipe Manana return ret; 392532b7c687SMark Fasheh } 392632b7c687SMark Fasheh 3927c125b8bfSFilipe Manana /* 3928d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3929d8b55242SFilipe Manana * source range to serialize with relocation. 3930c125b8bfSFilipe Manana */ 3931d8b55242SFilipe Manana btrfs_double_extent_lock(src, off, inode, destoff, len); 39321c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 3933d8b55242SFilipe Manana btrfs_double_extent_unlock(src, off, inode, destoff, len); 3934c125b8bfSFilipe Manana /* 3935c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3936c125b8bfSFilipe Manana * data immediately and not the previous data. 3937c125b8bfSFilipe Manana */ 393865bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 393909cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 394009cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 394134a28e3dSFilipe Manana 394234a28e3dSFilipe Manana return ret; 394334a28e3dSFilipe Manana } 394434a28e3dSFilipe Manana 394534a28e3dSFilipe Manana static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in, 394634a28e3dSFilipe Manana struct file *file_out, loff_t pos_out, 394734a28e3dSFilipe Manana loff_t *len, unsigned int remap_flags) 394834a28e3dSFilipe Manana { 394934a28e3dSFilipe Manana struct inode *inode_in = file_inode(file_in); 395034a28e3dSFilipe Manana struct inode *inode_out = file_inode(file_out); 395134a28e3dSFilipe Manana u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize; 395234a28e3dSFilipe Manana bool same_inode = inode_out == inode_in; 395334a28e3dSFilipe Manana u64 wb_len; 395434a28e3dSFilipe Manana int ret; 395534a28e3dSFilipe Manana 395634a28e3dSFilipe Manana if (!(remap_flags & REMAP_FILE_DEDUP)) { 395734a28e3dSFilipe Manana struct btrfs_root *root_out = BTRFS_I(inode_out)->root; 395834a28e3dSFilipe Manana 395934a28e3dSFilipe Manana if (btrfs_root_readonly(root_out)) 396034a28e3dSFilipe Manana return -EROFS; 396134a28e3dSFilipe Manana 396234a28e3dSFilipe Manana if (file_in->f_path.mnt != file_out->f_path.mnt || 396334a28e3dSFilipe Manana inode_in->i_sb != inode_out->i_sb) 396434a28e3dSFilipe Manana return -EXDEV; 396534a28e3dSFilipe Manana } 396634a28e3dSFilipe Manana 3967500710d3SFilipe Manana /* don't make the dst file partly checksummed */ 3968500710d3SFilipe Manana if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) != 3969500710d3SFilipe Manana (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) { 39707984ae52SGoldwyn Rodrigues return -EINVAL; 3971500710d3SFilipe Manana } 3972500710d3SFilipe Manana 397334a28e3dSFilipe Manana /* 397434a28e3dSFilipe Manana * Now that the inodes are locked, we need to start writeback ourselves 397534a28e3dSFilipe Manana * and can not rely on the writeback from the VFS's generic helper 397634a28e3dSFilipe Manana * generic_remap_file_range_prep() because: 397734a28e3dSFilipe Manana * 397834a28e3dSFilipe Manana * 1) For compression we must call filemap_fdatawrite_range() range 397934a28e3dSFilipe Manana * twice (btrfs_fdatawrite_range() does it for us), and the generic 398034a28e3dSFilipe Manana * helper only calls it once; 398134a28e3dSFilipe Manana * 398234a28e3dSFilipe Manana * 2) filemap_fdatawrite_range(), called by the generic helper only 398334a28e3dSFilipe Manana * waits for the writeback to complete, i.e. for IO to be done, and 398434a28e3dSFilipe Manana * not for the ordered extents to complete. We need to wait for them 398534a28e3dSFilipe Manana * to complete so that new file extent items are in the fs tree. 398634a28e3dSFilipe Manana */ 398734a28e3dSFilipe Manana if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP)) 398834a28e3dSFilipe Manana wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs); 398934a28e3dSFilipe Manana else 399034a28e3dSFilipe Manana wb_len = ALIGN(*len, bs); 399134a28e3dSFilipe Manana 399234a28e3dSFilipe Manana /* 399334a28e3dSFilipe Manana * Since we don't lock ranges, wait for ongoing lockless dio writes (as 399434a28e3dSFilipe Manana * any in progress could create its ordered extents after we wait for 399534a28e3dSFilipe Manana * existing ordered extents below). 399634a28e3dSFilipe Manana */ 399734a28e3dSFilipe Manana inode_dio_wait(inode_in); 399834a28e3dSFilipe Manana if (!same_inode) 399934a28e3dSFilipe Manana inode_dio_wait(inode_out); 400034a28e3dSFilipe Manana 400134a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs), 400234a28e3dSFilipe Manana wb_len); 400334a28e3dSFilipe Manana if (ret < 0) 40047984ae52SGoldwyn Rodrigues return ret; 400534a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs), 400634a28e3dSFilipe Manana wb_len); 400734a28e3dSFilipe Manana if (ret < 0) 40083db11b2eSZach Brown return ret; 40097984ae52SGoldwyn Rodrigues 40107984ae52SGoldwyn Rodrigues return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out, 40117984ae52SGoldwyn Rodrigues len, remap_flags); 40123db11b2eSZach Brown } 40133db11b2eSZach Brown 401442ec3d4cSDarrick J. Wong loff_t btrfs_remap_file_range(struct file *src_file, loff_t off, 401542ec3d4cSDarrick J. Wong struct file *dst_file, loff_t destoff, loff_t len, 40162e5dfc99SDarrick J. Wong unsigned int remap_flags) 40173db11b2eSZach Brown { 401834a28e3dSFilipe Manana struct inode *src_inode = file_inode(src_file); 401934a28e3dSFilipe Manana struct inode *dst_inode = file_inode(dst_file); 402034a28e3dSFilipe Manana bool same_inode = dst_inode == src_inode; 402142ec3d4cSDarrick J. Wong int ret; 402242ec3d4cSDarrick J. Wong 40232e5dfc99SDarrick J. Wong if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY)) 40242e5dfc99SDarrick J. Wong return -EINVAL; 40252e5dfc99SDarrick J. Wong 40267984ae52SGoldwyn Rodrigues if (same_inode) 40277984ae52SGoldwyn Rodrigues inode_lock(src_inode); 40287984ae52SGoldwyn Rodrigues else 40297984ae52SGoldwyn Rodrigues lock_two_nondirectories(src_inode, dst_inode); 40307984ae52SGoldwyn Rodrigues 403134a28e3dSFilipe Manana ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff, 403234a28e3dSFilipe Manana &len, remap_flags); 403334a28e3dSFilipe Manana if (ret < 0 || len == 0) 40347984ae52SGoldwyn Rodrigues goto out_unlock; 40352e5dfc99SDarrick J. Wong 403634a28e3dSFilipe Manana if (remap_flags & REMAP_FILE_DEDUP) 403734a28e3dSFilipe Manana ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff); 403834a28e3dSFilipe Manana else 403942ec3d4cSDarrick J. Wong ret = btrfs_clone_files(dst_file, src_file, off, len, destoff); 404034a28e3dSFilipe Manana 40417984ae52SGoldwyn Rodrigues out_unlock: 404234a28e3dSFilipe Manana if (same_inode) 404334a28e3dSFilipe Manana inode_unlock(src_inode); 404434a28e3dSFilipe Manana else 40454ea748e1SFilipe Manana unlock_two_nondirectories(src_inode, dst_inode); 404634a28e3dSFilipe Manana 404742ec3d4cSDarrick J. Wong return ret < 0 ? ret : len; 4048c5c9cd4dSSage Weil } 4049c5c9cd4dSSage Weil 40506ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40516ef5ed0dSJosef Bacik { 4052496ad9aaSAl Viro struct inode *inode = file_inode(file); 40530b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 40546ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40556ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40566ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40576ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40586ef5ed0dSJosef Bacik struct btrfs_path *path; 40596ef5ed0dSJosef Bacik struct btrfs_key location; 40606ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40616ef5ed0dSJosef Bacik u64 objectid = 0; 40626ef5ed0dSJosef Bacik u64 dir_id; 40633c04ce01SMiao Xie int ret; 40646ef5ed0dSJosef Bacik 40656ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40666ef5ed0dSJosef Bacik return -EPERM; 40676ef5ed0dSJosef Bacik 40683c04ce01SMiao Xie ret = mnt_want_write_file(file); 40693c04ce01SMiao Xie if (ret) 40703c04ce01SMiao Xie return ret; 40713c04ce01SMiao Xie 40723c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40733c04ce01SMiao Xie ret = -EFAULT; 40743c04ce01SMiao Xie goto out; 40753c04ce01SMiao Xie } 40766ef5ed0dSJosef Bacik 40776ef5ed0dSJosef Bacik if (!objectid) 40781cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40796ef5ed0dSJosef Bacik 40806ef5ed0dSJosef Bacik location.objectid = objectid; 40816ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40826ef5ed0dSJosef Bacik location.offset = (u64)-1; 40836ef5ed0dSJosef Bacik 40840b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 40853c04ce01SMiao Xie if (IS_ERR(new_root)) { 40863c04ce01SMiao Xie ret = PTR_ERR(new_root); 40873c04ce01SMiao Xie goto out; 40883c04ce01SMiao Xie } 40894fd786e6SMisono Tomohiro if (!is_fstree(new_root->root_key.objectid)) { 40906d6d2829Ssatoru takeuchi ret = -ENOENT; 40916d6d2829Ssatoru takeuchi goto out; 40926d6d2829Ssatoru takeuchi } 40936ef5ed0dSJosef Bacik 40946ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40953c04ce01SMiao Xie if (!path) { 40963c04ce01SMiao Xie ret = -ENOMEM; 40973c04ce01SMiao Xie goto out; 40983c04ce01SMiao Xie } 40996ef5ed0dSJosef Bacik path->leave_spinning = 1; 41006ef5ed0dSJosef Bacik 41016ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 410298d5dc13STsutomu Itoh if (IS_ERR(trans)) { 41036ef5ed0dSJosef Bacik btrfs_free_path(path); 41043c04ce01SMiao Xie ret = PTR_ERR(trans); 41053c04ce01SMiao Xie goto out; 41066ef5ed0dSJosef Bacik } 41076ef5ed0dSJosef Bacik 41080b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 41090b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 41106ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4111cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 41126ef5ed0dSJosef Bacik btrfs_free_path(path); 41133a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41140b246afaSJeff Mahoney btrfs_err(fs_info, 41155d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 41163c04ce01SMiao Xie ret = -ENOENT; 41173c04ce01SMiao Xie goto out; 41186ef5ed0dSJosef Bacik } 41196ef5ed0dSJosef Bacik 41206ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 41216ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 41226ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 41236ef5ed0dSJosef Bacik btrfs_free_path(path); 41246ef5ed0dSJosef Bacik 41250b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 41263a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41273c04ce01SMiao Xie out: 41283c04ce01SMiao Xie mnt_drop_write_file(file); 41293c04ce01SMiao Xie return ret; 41306ef5ed0dSJosef Bacik } 41316ef5ed0dSJosef Bacik 4132c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4133bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4134bf5fc093SJosef Bacik { 4135bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4136bf5fc093SJosef Bacik 4137bf5fc093SJosef Bacik space->total_bytes = 0; 4138bf5fc093SJosef Bacik space->used_bytes = 0; 4139bf5fc093SJosef Bacik space->flags = 0; 4140bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4141bf5fc093SJosef Bacik space->flags = block_group->flags; 4142bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4143bf5fc093SJosef Bacik space->used_bytes += 4144bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4145bf5fc093SJosef Bacik } 4146bf5fc093SJosef Bacik } 4147bf5fc093SJosef Bacik 41482ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 41492ff7e61eSJeff Mahoney void __user *arg) 41501406e432SJosef Bacik { 41511406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41521406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41531406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41547fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 415513f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41561406e432SJosef Bacik struct btrfs_space_info *info; 4157315d8e98SColin Ian King static const u64 types[] = { 4158315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4159bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4160bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4161315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4162315d8e98SColin Ian King }; 4163bf5fc093SJosef Bacik int num_types = 4; 41647fde62bfSChris Mason int alloc_size; 41651406e432SJosef Bacik int ret = 0; 416651788b1bSDan Rosenberg u64 slot_count = 0; 4167bf5fc093SJosef Bacik int i, c; 41681406e432SJosef Bacik 41691406e432SJosef Bacik if (copy_from_user(&space_args, 41701406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41711406e432SJosef Bacik sizeof(space_args))) 41721406e432SJosef Bacik return -EFAULT; 41731406e432SJosef Bacik 4174bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4175bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4176bf5fc093SJosef Bacik 4177bf5fc093SJosef Bacik info = NULL; 41787fde62bfSChris Mason rcu_read_lock(); 41790b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4180bf5fc093SJosef Bacik list) { 4181bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4182bf5fc093SJosef Bacik info = tmp; 4183bf5fc093SJosef Bacik break; 4184bf5fc093SJosef Bacik } 4185bf5fc093SJosef Bacik } 41867fde62bfSChris Mason rcu_read_unlock(); 41871406e432SJosef Bacik 4188bf5fc093SJosef Bacik if (!info) 4189bf5fc093SJosef Bacik continue; 4190bf5fc093SJosef Bacik 4191bf5fc093SJosef Bacik down_read(&info->groups_sem); 4192bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4193bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4194bf5fc093SJosef Bacik slot_count++; 4195bf5fc093SJosef Bacik } 4196bf5fc093SJosef Bacik up_read(&info->groups_sem); 4197bf5fc093SJosef Bacik } 4198bf5fc093SJosef Bacik 419936523e95SDavid Sterba /* 420036523e95SDavid Sterba * Global block reserve, exported as a space_info 420136523e95SDavid Sterba */ 420236523e95SDavid Sterba slot_count++; 420336523e95SDavid Sterba 42047fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 42057fde62bfSChris Mason if (space_args.space_slots == 0) { 42067fde62bfSChris Mason space_args.total_spaces = slot_count; 42077fde62bfSChris Mason goto out; 42087fde62bfSChris Mason } 4209bf5fc093SJosef Bacik 421051788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4211bf5fc093SJosef Bacik 42127fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4213bf5fc093SJosef Bacik 42147fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 42157fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 42167fde62bfSChris Mason */ 421709cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 42187fde62bfSChris Mason return -ENOMEM; 42197fde62bfSChris Mason 42207fde62bfSChris Mason space_args.total_spaces = 0; 42218d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 42227fde62bfSChris Mason if (!dest) 42237fde62bfSChris Mason return -ENOMEM; 42247fde62bfSChris Mason dest_orig = dest; 42257fde62bfSChris Mason 42267fde62bfSChris Mason /* now we have a buffer to copy into */ 4227bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4228bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4229bf5fc093SJosef Bacik 423051788b1bSDan Rosenberg if (!slot_count) 423151788b1bSDan Rosenberg break; 423251788b1bSDan Rosenberg 4233bf5fc093SJosef Bacik info = NULL; 42341406e432SJosef Bacik rcu_read_lock(); 42350b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4236bf5fc093SJosef Bacik list) { 4237bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4238bf5fc093SJosef Bacik info = tmp; 42397fde62bfSChris Mason break; 4240bf5fc093SJosef Bacik } 4241bf5fc093SJosef Bacik } 4242bf5fc093SJosef Bacik rcu_read_unlock(); 42437fde62bfSChris Mason 4244bf5fc093SJosef Bacik if (!info) 4245bf5fc093SJosef Bacik continue; 4246bf5fc093SJosef Bacik down_read(&info->groups_sem); 4247bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4248bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4249c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4250c065f5b1SSu Yue &space); 42517fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42521406e432SJosef Bacik dest++; 42531406e432SJosef Bacik space_args.total_spaces++; 425451788b1bSDan Rosenberg slot_count--; 42551406e432SJosef Bacik } 425651788b1bSDan Rosenberg if (!slot_count) 425751788b1bSDan Rosenberg break; 4258bf5fc093SJosef Bacik } 4259bf5fc093SJosef Bacik up_read(&info->groups_sem); 4260bf5fc093SJosef Bacik } 42611406e432SJosef Bacik 426236523e95SDavid Sterba /* 426336523e95SDavid Sterba * Add global block reserve 426436523e95SDavid Sterba */ 426536523e95SDavid Sterba if (slot_count) { 42660b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 426736523e95SDavid Sterba 426836523e95SDavid Sterba spin_lock(&block_rsv->lock); 426936523e95SDavid Sterba space.total_bytes = block_rsv->size; 427036523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 427136523e95SDavid Sterba spin_unlock(&block_rsv->lock); 427236523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 427336523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 427436523e95SDavid Sterba space_args.total_spaces++; 427536523e95SDavid Sterba } 427636523e95SDavid Sterba 42772eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42787fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42797fde62bfSChris Mason 42807fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42817fde62bfSChris Mason ret = -EFAULT; 42827fde62bfSChris Mason 42837fde62bfSChris Mason kfree(dest_orig); 42847fde62bfSChris Mason out: 42857fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42861406e432SJosef Bacik ret = -EFAULT; 42871406e432SJosef Bacik 42881406e432SJosef Bacik return ret; 42891406e432SJosef Bacik } 42901406e432SJosef Bacik 42919a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42929a8c28beSMiao Xie void __user *argp) 429346204592SSage Weil { 429446204592SSage Weil struct btrfs_trans_handle *trans; 429546204592SSage Weil u64 transid; 4296db5b493aSTsutomu Itoh int ret; 429746204592SSage Weil 4298d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4299ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4300ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 430198d5dc13STsutomu Itoh return PTR_ERR(trans); 4302ff7c1d33SMiao Xie 4303ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4304ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4305ff7c1d33SMiao Xie goto out; 4306ff7c1d33SMiao Xie } 430746204592SSage Weil transid = trans->transid; 43083a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 43098b2b2d3cSTsutomu Itoh if (ret) { 43103a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4311db5b493aSTsutomu Itoh return ret; 43128b2b2d3cSTsutomu Itoh } 4313ff7c1d33SMiao Xie out: 431446204592SSage Weil if (argp) 431546204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 431646204592SSage Weil return -EFAULT; 431746204592SSage Weil return 0; 431846204592SSage Weil } 431946204592SSage Weil 43202ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 43219a8c28beSMiao Xie void __user *argp) 432246204592SSage Weil { 432346204592SSage Weil u64 transid; 432446204592SSage Weil 432546204592SSage Weil if (argp) { 432646204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 432746204592SSage Weil return -EFAULT; 432846204592SSage Weil } else { 432946204592SSage Weil transid = 0; /* current trans */ 433046204592SSage Weil } 43312ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 433246204592SSage Weil } 433346204592SSage Weil 4334b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4335475f6387SJan Schmidt { 43360b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4337475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4338b8e95489SMiao Xie int ret; 4339475f6387SJan Schmidt 4340475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4341475f6387SJan Schmidt return -EPERM; 4342475f6387SJan Schmidt 4343475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4344475f6387SJan Schmidt if (IS_ERR(sa)) 4345475f6387SJan Schmidt return PTR_ERR(sa); 4346475f6387SJan Schmidt 4347b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4348b8e95489SMiao Xie ret = mnt_want_write_file(file); 4349b8e95489SMiao Xie if (ret) 4350b8e95489SMiao Xie goto out; 4351b8e95489SMiao Xie } 4352b8e95489SMiao Xie 43530b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 435463a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 435563a212abSStefan Behrens 0); 4356475f6387SJan Schmidt 435706fe39abSFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4358475f6387SJan Schmidt ret = -EFAULT; 4359475f6387SJan Schmidt 4360b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4361b8e95489SMiao Xie mnt_drop_write_file(file); 4362b8e95489SMiao Xie out: 4363475f6387SJan Schmidt kfree(sa); 4364475f6387SJan Schmidt return ret; 4365475f6387SJan Schmidt } 4366475f6387SJan Schmidt 43672ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4368475f6387SJan Schmidt { 4369475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4370475f6387SJan Schmidt return -EPERM; 4371475f6387SJan Schmidt 43722ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4373475f6387SJan Schmidt } 4374475f6387SJan Schmidt 43752ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4376475f6387SJan Schmidt void __user *arg) 4377475f6387SJan Schmidt { 4378475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4379475f6387SJan Schmidt int ret; 4380475f6387SJan Schmidt 4381475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4382475f6387SJan Schmidt return -EPERM; 4383475f6387SJan Schmidt 4384475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4385475f6387SJan Schmidt if (IS_ERR(sa)) 4386475f6387SJan Schmidt return PTR_ERR(sa); 4387475f6387SJan Schmidt 43882ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4389475f6387SJan Schmidt 43904fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4391475f6387SJan Schmidt ret = -EFAULT; 4392475f6387SJan Schmidt 4393475f6387SJan Schmidt kfree(sa); 4394475f6387SJan Schmidt return ret; 4395475f6387SJan Schmidt } 4396475f6387SJan Schmidt 43972ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4398b27f7c0cSDavid Sterba void __user *arg) 4399c11d2c23SStefan Behrens { 4400c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4401c11d2c23SStefan Behrens int ret; 4402c11d2c23SStefan Behrens 4403c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4404c11d2c23SStefan Behrens if (IS_ERR(sa)) 4405c11d2c23SStefan Behrens return PTR_ERR(sa); 4406c11d2c23SStefan Behrens 4407b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4408b27f7c0cSDavid Sterba kfree(sa); 4409b27f7c0cSDavid Sterba return -EPERM; 4410b27f7c0cSDavid Sterba } 4411b27f7c0cSDavid Sterba 44122ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4413c11d2c23SStefan Behrens 4414eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4415c11d2c23SStefan Behrens ret = -EFAULT; 4416c11d2c23SStefan Behrens 4417c11d2c23SStefan Behrens kfree(sa); 4418c11d2c23SStefan Behrens return ret; 4419c11d2c23SStefan Behrens } 4420c11d2c23SStefan Behrens 44212ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 44222ff7e61eSJeff Mahoney void __user *arg) 44233f6bcfbdSStefan Behrens { 44243f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 44253f6bcfbdSStefan Behrens int ret; 44263f6bcfbdSStefan Behrens 44273f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44283f6bcfbdSStefan Behrens return -EPERM; 44293f6bcfbdSStefan Behrens 44303f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44313f6bcfbdSStefan Behrens if (IS_ERR(p)) 44323f6bcfbdSStefan Behrens return PTR_ERR(p); 44333f6bcfbdSStefan Behrens 44343f6bcfbdSStefan Behrens switch (p->cmd) { 44353f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4436bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4437adfa97cbSIlya Dryomov ret = -EROFS; 4438adfa97cbSIlya Dryomov goto out; 4439adfa97cbSIlya Dryomov } 4440171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4441e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44423f6bcfbdSStefan Behrens } else { 44432ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4444171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 44453f6bcfbdSStefan Behrens } 44463f6bcfbdSStefan Behrens break; 44473f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44480b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 44493f6bcfbdSStefan Behrens ret = 0; 44503f6bcfbdSStefan Behrens break; 44513f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 445217d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 445397282031SAnand Jain ret = 0; 44543f6bcfbdSStefan Behrens break; 44553f6bcfbdSStefan Behrens default: 44563f6bcfbdSStefan Behrens ret = -EINVAL; 44573f6bcfbdSStefan Behrens break; 44583f6bcfbdSStefan Behrens } 44593f6bcfbdSStefan Behrens 4460d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 44613f6bcfbdSStefan Behrens ret = -EFAULT; 4462adfa97cbSIlya Dryomov out: 44633f6bcfbdSStefan Behrens kfree(p); 44643f6bcfbdSStefan Behrens return ret; 44653f6bcfbdSStefan Behrens } 44663f6bcfbdSStefan Behrens 4467d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4468d7728c96SJan Schmidt { 4469d7728c96SJan Schmidt int ret = 0; 4470d7728c96SJan Schmidt int i; 4471740c3d22SChris Mason u64 rel_ptr; 4472d7728c96SJan Schmidt int size; 4473806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4474d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4475d7728c96SJan Schmidt struct btrfs_path *path; 4476d7728c96SJan Schmidt 447782b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4478d7728c96SJan Schmidt return -EPERM; 4479d7728c96SJan Schmidt 4480d7728c96SJan Schmidt path = btrfs_alloc_path(); 4481d7728c96SJan Schmidt if (!path) { 4482d7728c96SJan Schmidt ret = -ENOMEM; 4483d7728c96SJan Schmidt goto out; 4484d7728c96SJan Schmidt } 4485d7728c96SJan Schmidt 4486d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4487d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4488d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4489d7728c96SJan Schmidt ipa = NULL; 4490d7728c96SJan Schmidt goto out; 4491d7728c96SJan Schmidt } 4492d7728c96SJan Schmidt 4493d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4494d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4495d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4496d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4497d7728c96SJan Schmidt ipath = NULL; 4498d7728c96SJan Schmidt goto out; 4499d7728c96SJan Schmidt } 4500d7728c96SJan Schmidt 4501d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4502d7728c96SJan Schmidt if (ret < 0) 4503d7728c96SJan Schmidt goto out; 4504d7728c96SJan Schmidt 4505d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4506745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4507745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4508740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4509d7728c96SJan Schmidt } 4510d7728c96SJan Schmidt 4511718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4512718dc5faSOmar Sandoval ipath->fspath, size); 4513d7728c96SJan Schmidt if (ret) { 4514d7728c96SJan Schmidt ret = -EFAULT; 4515d7728c96SJan Schmidt goto out; 4516d7728c96SJan Schmidt } 4517d7728c96SJan Schmidt 4518d7728c96SJan Schmidt out: 4519d7728c96SJan Schmidt btrfs_free_path(path); 4520d7728c96SJan Schmidt free_ipath(ipath); 4521d7728c96SJan Schmidt kfree(ipa); 4522d7728c96SJan Schmidt 4523d7728c96SJan Schmidt return ret; 4524d7728c96SJan Schmidt } 4525d7728c96SJan Schmidt 4526d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4527d7728c96SJan Schmidt { 4528d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4529d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4530d7728c96SJan Schmidt 4531d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4532d7728c96SJan Schmidt inodes->bytes_left -= c; 4533d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4534d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4535d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4536d7728c96SJan Schmidt inodes->elem_cnt += 3; 4537d7728c96SJan Schmidt } else { 4538d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4539d7728c96SJan Schmidt inodes->bytes_left = 0; 4540d7728c96SJan Schmidt inodes->elem_missed += 3; 4541d7728c96SJan Schmidt } 4542d7728c96SJan Schmidt 4543d7728c96SJan Schmidt return 0; 4544d7728c96SJan Schmidt } 4545d7728c96SJan Schmidt 45462ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4547d24a67b2SZygo Blaxell void __user *arg, int version) 4548d7728c96SJan Schmidt { 4549d7728c96SJan Schmidt int ret = 0; 4550d7728c96SJan Schmidt int size; 4551d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4552d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4553d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4554d24a67b2SZygo Blaxell bool ignore_offset; 4555d7728c96SJan Schmidt 4556d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4557d7728c96SJan Schmidt return -EPERM; 4558d7728c96SJan Schmidt 4559d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 45607b9ea627SShailendra Verma if (IS_ERR(loi)) 45617b9ea627SShailendra Verma return PTR_ERR(loi); 4562d7728c96SJan Schmidt 4563d24a67b2SZygo Blaxell if (version == 1) { 4564d24a67b2SZygo Blaxell ignore_offset = false; 4565b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4566d24a67b2SZygo Blaxell } else { 4567d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4568d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4569d24a67b2SZygo Blaxell ret = -EINVAL; 4570d24a67b2SZygo Blaxell goto out_loi; 4571d24a67b2SZygo Blaxell } 4572d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4573d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4574d24a67b2SZygo Blaxell ret = -EINVAL; 4575d24a67b2SZygo Blaxell goto out_loi; 4576d24a67b2SZygo Blaxell } 4577d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4578b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4579d24a67b2SZygo Blaxell } 4580d24a67b2SZygo Blaxell 4581d7728c96SJan Schmidt path = btrfs_alloc_path(); 4582d7728c96SJan Schmidt if (!path) { 4583d7728c96SJan Schmidt ret = -ENOMEM; 4584d7728c96SJan Schmidt goto out; 4585d7728c96SJan Schmidt } 4586d7728c96SJan Schmidt 4587d7728c96SJan Schmidt inodes = init_data_container(size); 4588d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4589d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4590d7728c96SJan Schmidt inodes = NULL; 4591d7728c96SJan Schmidt goto out; 4592d7728c96SJan Schmidt } 4593d7728c96SJan Schmidt 45942ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4595d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4596df031f07SLiu Bo if (ret == -EINVAL) 4597d7728c96SJan Schmidt ret = -ENOENT; 4598d7728c96SJan Schmidt if (ret < 0) 4599d7728c96SJan Schmidt goto out; 4600d7728c96SJan Schmidt 4601718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4602718dc5faSOmar Sandoval size); 4603d7728c96SJan Schmidt if (ret) 4604d7728c96SJan Schmidt ret = -EFAULT; 4605d7728c96SJan Schmidt 4606d7728c96SJan Schmidt out: 4607d7728c96SJan Schmidt btrfs_free_path(path); 4608f54de068SDavid Sterba kvfree(inodes); 4609d24a67b2SZygo Blaxell out_loi: 4610d7728c96SJan Schmidt kfree(loi); 4611d7728c96SJan Schmidt 4612d7728c96SJan Schmidt return ret; 4613d7728c96SJan Schmidt } 4614d7728c96SJan Schmidt 4615008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4616c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4617c9e9f97bSIlya Dryomov { 4618c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4619c9e9f97bSIlya Dryomov 4620c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4621c9e9f97bSIlya Dryomov 46223009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4623837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4624837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4625837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4626a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4627a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4628837d5b6eSIlya Dryomov 4629c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4630c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4631c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 463219a39dceSIlya Dryomov 463319a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 463419a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 463519a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4636c9e9f97bSIlya Dryomov } 4637c9e9f97bSIlya Dryomov 46389ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4639c9e9f97bSIlya Dryomov { 4640496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4641c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4642c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4643c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4644ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4645c9e9f97bSIlya Dryomov int ret; 4646c9e9f97bSIlya Dryomov 4647c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4648c9e9f97bSIlya Dryomov return -EPERM; 4649c9e9f97bSIlya Dryomov 4650e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46519ba1f6e4SLiu Bo if (ret) 46529ba1f6e4SLiu Bo return ret; 46539ba1f6e4SLiu Bo 4654ed0fb78fSIlya Dryomov again: 4655171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4656c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4657ed0fb78fSIlya Dryomov need_unlock = true; 4658ed0fb78fSIlya Dryomov goto locked; 4659ed0fb78fSIlya Dryomov } 4660ed0fb78fSIlya Dryomov 4661ed0fb78fSIlya Dryomov /* 466201327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4663ed0fb78fSIlya Dryomov * (1) some other op is running 4664ed0fb78fSIlya Dryomov * (2) balance is running 4665ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4666ed0fb78fSIlya Dryomov */ 4667ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4668ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4669ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 46703009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4671ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4672dccdb07bSDavid Sterba /* 4673dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4674dccdb07bSDavid Sterba * we'll reexamine the status again. 4675dccdb07bSDavid Sterba */ 4676ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4677ed0fb78fSIlya Dryomov 4678ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 46793009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4680ed0fb78fSIlya Dryomov /* this is (3) */ 4681ed0fb78fSIlya Dryomov need_unlock = false; 4682ed0fb78fSIlya Dryomov goto locked; 4683ed0fb78fSIlya Dryomov } 4684ed0fb78fSIlya Dryomov 4685ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4686ed0fb78fSIlya Dryomov goto again; 4687ed0fb78fSIlya Dryomov } else { 4688ed0fb78fSIlya Dryomov /* this is (2) */ 4689ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4690ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4691ed0fb78fSIlya Dryomov goto out; 4692ed0fb78fSIlya Dryomov } 4693ed0fb78fSIlya Dryomov } else { 4694ed0fb78fSIlya Dryomov /* this is (1) */ 4695ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4696e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4697ed0fb78fSIlya Dryomov goto out; 4698ed0fb78fSIlya Dryomov } 4699ed0fb78fSIlya Dryomov 4700ed0fb78fSIlya Dryomov locked: 4701171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4702c9e9f97bSIlya Dryomov 4703c9e9f97bSIlya Dryomov if (arg) { 4704c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4705c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4706c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4707ed0fb78fSIlya Dryomov goto out_unlock; 4708c9e9f97bSIlya Dryomov } 4709de322263SIlya Dryomov 4710de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4711de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4712de322263SIlya Dryomov ret = -ENOTCONN; 4713de322263SIlya Dryomov goto out_bargs; 4714de322263SIlya Dryomov } 4715de322263SIlya Dryomov 4716de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4717de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4718de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4719de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4720de322263SIlya Dryomov 4721de322263SIlya Dryomov goto do_balance; 4722de322263SIlya Dryomov } 4723c9e9f97bSIlya Dryomov } else { 4724c9e9f97bSIlya Dryomov bargs = NULL; 4725c9e9f97bSIlya Dryomov } 4726c9e9f97bSIlya Dryomov 4727ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4728837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4729837d5b6eSIlya Dryomov goto out_bargs; 4730837d5b6eSIlya Dryomov } 4731837d5b6eSIlya Dryomov 47328d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4733c9e9f97bSIlya Dryomov if (!bctl) { 4734c9e9f97bSIlya Dryomov ret = -ENOMEM; 4735c9e9f97bSIlya Dryomov goto out_bargs; 4736c9e9f97bSIlya Dryomov } 4737c9e9f97bSIlya Dryomov 4738c9e9f97bSIlya Dryomov if (arg) { 4739c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4740c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4741c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4742c9e9f97bSIlya Dryomov 4743c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4744f43ffb60SIlya Dryomov } else { 4745f43ffb60SIlya Dryomov /* balance everything - no filters */ 4746f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4747c9e9f97bSIlya Dryomov } 4748c9e9f97bSIlya Dryomov 47498eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47508eb93459SDavid Sterba ret = -EINVAL; 47510f89abf5SChristian Engelmayer goto out_bctl; 47528eb93459SDavid Sterba } 47538eb93459SDavid Sterba 4754de322263SIlya Dryomov do_balance: 4755c9e9f97bSIlya Dryomov /* 4756149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4757149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4758149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4759149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4760c9e9f97bSIlya Dryomov */ 4761ed0fb78fSIlya Dryomov need_unlock = false; 4762ed0fb78fSIlya Dryomov 47636fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 47640f89abf5SChristian Engelmayer bctl = NULL; 4765ed0fb78fSIlya Dryomov 4766d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4767c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4768c9e9f97bSIlya Dryomov ret = -EFAULT; 4769c9e9f97bSIlya Dryomov } 4770c9e9f97bSIlya Dryomov 47710f89abf5SChristian Engelmayer out_bctl: 47720f89abf5SChristian Engelmayer kfree(bctl); 4773c9e9f97bSIlya Dryomov out_bargs: 4774c9e9f97bSIlya Dryomov kfree(bargs); 4775ed0fb78fSIlya Dryomov out_unlock: 4776c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4777ed0fb78fSIlya Dryomov if (need_unlock) 4778171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4779ed0fb78fSIlya Dryomov out: 4780e54bfa31SLiu Bo mnt_drop_write_file(file); 4781c9e9f97bSIlya Dryomov return ret; 4782c9e9f97bSIlya Dryomov } 4783c9e9f97bSIlya Dryomov 47842ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4785837d5b6eSIlya Dryomov { 4786837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4787837d5b6eSIlya Dryomov return -EPERM; 4788837d5b6eSIlya Dryomov 4789837d5b6eSIlya Dryomov switch (cmd) { 4790837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 47910b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4792a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47930b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4794837d5b6eSIlya Dryomov } 4795837d5b6eSIlya Dryomov 4796837d5b6eSIlya Dryomov return -EINVAL; 4797837d5b6eSIlya Dryomov } 4798837d5b6eSIlya Dryomov 47992ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 480019a39dceSIlya Dryomov void __user *arg) 480119a39dceSIlya Dryomov { 480219a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 480319a39dceSIlya Dryomov int ret = 0; 480419a39dceSIlya Dryomov 480519a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 480619a39dceSIlya Dryomov return -EPERM; 480719a39dceSIlya Dryomov 480819a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 480919a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 481019a39dceSIlya Dryomov ret = -ENOTCONN; 481119a39dceSIlya Dryomov goto out; 481219a39dceSIlya Dryomov } 481319a39dceSIlya Dryomov 48148d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 481519a39dceSIlya Dryomov if (!bargs) { 481619a39dceSIlya Dryomov ret = -ENOMEM; 481719a39dceSIlya Dryomov goto out; 481819a39dceSIlya Dryomov } 481919a39dceSIlya Dryomov 4820008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 482119a39dceSIlya Dryomov 482219a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 482319a39dceSIlya Dryomov ret = -EFAULT; 482419a39dceSIlya Dryomov 482519a39dceSIlya Dryomov kfree(bargs); 482619a39dceSIlya Dryomov out: 482719a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 482819a39dceSIlya Dryomov return ret; 482919a39dceSIlya Dryomov } 483019a39dceSIlya Dryomov 4831905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 48325d13a37bSArne Jansen { 48330b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48340b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48355d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 48365d13a37bSArne Jansen int ret; 48375d13a37bSArne Jansen 48385d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48395d13a37bSArne Jansen return -EPERM; 48405d13a37bSArne Jansen 4841905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4842905b0ddaSMiao Xie if (ret) 4843905b0ddaSMiao Xie return ret; 48445d13a37bSArne Jansen 48455d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4846905b0ddaSMiao Xie if (IS_ERR(sa)) { 4847905b0ddaSMiao Xie ret = PTR_ERR(sa); 4848905b0ddaSMiao Xie goto drop_write; 4849905b0ddaSMiao Xie } 48505d13a37bSArne Jansen 48510b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48525d13a37bSArne Jansen 48535d13a37bSArne Jansen switch (sa->cmd) { 48545d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4855340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 48565d13a37bSArne Jansen break; 48575d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4858340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 48595d13a37bSArne Jansen break; 48605d13a37bSArne Jansen default: 48615d13a37bSArne Jansen ret = -EINVAL; 48625d13a37bSArne Jansen break; 48635d13a37bSArne Jansen } 48645d13a37bSArne Jansen 48655d13a37bSArne Jansen kfree(sa); 48660b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4867905b0ddaSMiao Xie drop_write: 4868905b0ddaSMiao Xie mnt_drop_write_file(file); 48695d13a37bSArne Jansen return ret; 48705d13a37bSArne Jansen } 48715d13a37bSArne Jansen 4872905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48735d13a37bSArne Jansen { 48740b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48750b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48760b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48775d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48785d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48795d13a37bSArne Jansen int ret; 48805d13a37bSArne Jansen int err; 48815d13a37bSArne Jansen 48825d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48835d13a37bSArne Jansen return -EPERM; 48845d13a37bSArne Jansen 4885905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4886905b0ddaSMiao Xie if (ret) 4887905b0ddaSMiao Xie return ret; 48885d13a37bSArne Jansen 48895d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4890905b0ddaSMiao Xie if (IS_ERR(sa)) { 4891905b0ddaSMiao Xie ret = PTR_ERR(sa); 4892905b0ddaSMiao Xie goto drop_write; 4893905b0ddaSMiao Xie } 48945d13a37bSArne Jansen 48955d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48965d13a37bSArne Jansen if (IS_ERR(trans)) { 48975d13a37bSArne Jansen ret = PTR_ERR(trans); 48985d13a37bSArne Jansen goto out; 48995d13a37bSArne Jansen } 49005d13a37bSArne Jansen 49015d13a37bSArne Jansen if (sa->assign) { 49029f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 49035d13a37bSArne Jansen } else { 490439616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 49055d13a37bSArne Jansen } 49065d13a37bSArne Jansen 4907e082f563SQu Wenruo /* update qgroup status and info */ 4908280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4909e082f563SQu Wenruo if (err < 0) 49100b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4911ad8403dfSAnand Jain "failed to update qgroup status and info"); 49123a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49135d13a37bSArne Jansen if (err && !ret) 49145d13a37bSArne Jansen ret = err; 49155d13a37bSArne Jansen 49165d13a37bSArne Jansen out: 49175d13a37bSArne Jansen kfree(sa); 4918905b0ddaSMiao Xie drop_write: 4919905b0ddaSMiao Xie mnt_drop_write_file(file); 49205d13a37bSArne Jansen return ret; 49215d13a37bSArne Jansen } 49225d13a37bSArne Jansen 4923905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 49245d13a37bSArne Jansen { 49250b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49260b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49275d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49285d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49295d13a37bSArne Jansen int ret; 49305d13a37bSArne Jansen int err; 49315d13a37bSArne Jansen 49325d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49335d13a37bSArne Jansen return -EPERM; 49345d13a37bSArne Jansen 4935905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4936905b0ddaSMiao Xie if (ret) 4937905b0ddaSMiao Xie return ret; 49385d13a37bSArne Jansen 49395d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4940905b0ddaSMiao Xie if (IS_ERR(sa)) { 4941905b0ddaSMiao Xie ret = PTR_ERR(sa); 4942905b0ddaSMiao Xie goto drop_write; 4943905b0ddaSMiao Xie } 49445d13a37bSArne Jansen 4945d86e56cfSMiao Xie if (!sa->qgroupid) { 4946d86e56cfSMiao Xie ret = -EINVAL; 4947d86e56cfSMiao Xie goto out; 4948d86e56cfSMiao Xie } 4949d86e56cfSMiao Xie 49505d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49515d13a37bSArne Jansen if (IS_ERR(trans)) { 49525d13a37bSArne Jansen ret = PTR_ERR(trans); 49535d13a37bSArne Jansen goto out; 49545d13a37bSArne Jansen } 49555d13a37bSArne Jansen 49565d13a37bSArne Jansen if (sa->create) { 495749a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 49585d13a37bSArne Jansen } else { 49593efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 49605d13a37bSArne Jansen } 49615d13a37bSArne Jansen 49623a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49635d13a37bSArne Jansen if (err && !ret) 49645d13a37bSArne Jansen ret = err; 49655d13a37bSArne Jansen 49665d13a37bSArne Jansen out: 49675d13a37bSArne Jansen kfree(sa); 4968905b0ddaSMiao Xie drop_write: 4969905b0ddaSMiao Xie mnt_drop_write_file(file); 49705d13a37bSArne Jansen return ret; 49715d13a37bSArne Jansen } 49725d13a37bSArne Jansen 4973905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49745d13a37bSArne Jansen { 49750b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49760b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49775d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49785d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49795d13a37bSArne Jansen int ret; 49805d13a37bSArne Jansen int err; 49815d13a37bSArne Jansen u64 qgroupid; 49825d13a37bSArne Jansen 49835d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49845d13a37bSArne Jansen return -EPERM; 49855d13a37bSArne Jansen 4986905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4987905b0ddaSMiao Xie if (ret) 4988905b0ddaSMiao Xie return ret; 49895d13a37bSArne Jansen 49905d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4991905b0ddaSMiao Xie if (IS_ERR(sa)) { 4992905b0ddaSMiao Xie ret = PTR_ERR(sa); 4993905b0ddaSMiao Xie goto drop_write; 4994905b0ddaSMiao Xie } 49955d13a37bSArne Jansen 49965d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49975d13a37bSArne Jansen if (IS_ERR(trans)) { 49985d13a37bSArne Jansen ret = PTR_ERR(trans); 49995d13a37bSArne Jansen goto out; 50005d13a37bSArne Jansen } 50015d13a37bSArne Jansen 50025d13a37bSArne Jansen qgroupid = sa->qgroupid; 50035d13a37bSArne Jansen if (!qgroupid) { 50045d13a37bSArne Jansen /* take the current subvol as qgroup */ 50055d13a37bSArne Jansen qgroupid = root->root_key.objectid; 50065d13a37bSArne Jansen } 50075d13a37bSArne Jansen 5008f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 50095d13a37bSArne Jansen 50103a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 50115d13a37bSArne Jansen if (err && !ret) 50125d13a37bSArne Jansen ret = err; 50135d13a37bSArne Jansen 50145d13a37bSArne Jansen out: 50155d13a37bSArne Jansen kfree(sa); 5016905b0ddaSMiao Xie drop_write: 5017905b0ddaSMiao Xie mnt_drop_write_file(file); 50185d13a37bSArne Jansen return ret; 50195d13a37bSArne Jansen } 50205d13a37bSArne Jansen 50212f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 50222f232036SJan Schmidt { 50230b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50240b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50252f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50262f232036SJan Schmidt int ret; 50272f232036SJan Schmidt 50282f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50292f232036SJan Schmidt return -EPERM; 50302f232036SJan Schmidt 50312f232036SJan Schmidt ret = mnt_want_write_file(file); 50322f232036SJan Schmidt if (ret) 50332f232036SJan Schmidt return ret; 50342f232036SJan Schmidt 50352f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50362f232036SJan Schmidt if (IS_ERR(qsa)) { 50372f232036SJan Schmidt ret = PTR_ERR(qsa); 50382f232036SJan Schmidt goto drop_write; 50392f232036SJan Schmidt } 50402f232036SJan Schmidt 50412f232036SJan Schmidt if (qsa->flags) { 50422f232036SJan Schmidt ret = -EINVAL; 50432f232036SJan Schmidt goto out; 50442f232036SJan Schmidt } 50452f232036SJan Schmidt 50460b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 50472f232036SJan Schmidt 50482f232036SJan Schmidt out: 50492f232036SJan Schmidt kfree(qsa); 50502f232036SJan Schmidt drop_write: 50512f232036SJan Schmidt mnt_drop_write_file(file); 50522f232036SJan Schmidt return ret; 50532f232036SJan Schmidt } 50542f232036SJan Schmidt 50552f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50562f232036SJan Schmidt { 50570b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50592f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50602f232036SJan Schmidt int ret = 0; 50612f232036SJan Schmidt 50622f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50632f232036SJan Schmidt return -EPERM; 50642f232036SJan Schmidt 50658d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 50662f232036SJan Schmidt if (!qsa) 50672f232036SJan Schmidt return -ENOMEM; 50682f232036SJan Schmidt 50690b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50702f232036SJan Schmidt qsa->flags = 1; 50710b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 50722f232036SJan Schmidt } 50732f232036SJan Schmidt 50742f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50752f232036SJan Schmidt ret = -EFAULT; 50762f232036SJan Schmidt 50772f232036SJan Schmidt kfree(qsa); 50782f232036SJan Schmidt return ret; 50792f232036SJan Schmidt } 50802f232036SJan Schmidt 508157254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 508257254b6eSJan Schmidt { 50830b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50840b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 508557254b6eSJan Schmidt 508657254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 508757254b6eSJan Schmidt return -EPERM; 508857254b6eSJan Schmidt 50890b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 509057254b6eSJan Schmidt } 509157254b6eSJan Schmidt 5092abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5093abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50948ea05e3aSAlexander Block { 5095496ad9aaSAl Viro struct inode *inode = file_inode(file); 50960b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50978ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50988ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50998ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 510095582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 51018ea05e3aSAlexander Block int ret = 0; 5102dd5f9615SStefan Behrens int received_uuid_changed; 51038ea05e3aSAlexander Block 5104bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5105bd60ea0fSDavid Sterba return -EPERM; 5106bd60ea0fSDavid Sterba 51078ea05e3aSAlexander Block ret = mnt_want_write_file(file); 51088ea05e3aSAlexander Block if (ret < 0) 51098ea05e3aSAlexander Block return ret; 51108ea05e3aSAlexander Block 51110b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51128ea05e3aSAlexander Block 51134a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 51148ea05e3aSAlexander Block ret = -EINVAL; 51158ea05e3aSAlexander Block goto out; 51168ea05e3aSAlexander Block } 51178ea05e3aSAlexander Block 51188ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 51198ea05e3aSAlexander Block ret = -EROFS; 51208ea05e3aSAlexander Block goto out; 51218ea05e3aSAlexander Block } 51228ea05e3aSAlexander Block 5123dd5f9615SStefan Behrens /* 5124dd5f9615SStefan Behrens * 1 - root item 5125dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5126dd5f9615SStefan Behrens */ 5127dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51288ea05e3aSAlexander Block if (IS_ERR(trans)) { 51298ea05e3aSAlexander Block ret = PTR_ERR(trans); 51308ea05e3aSAlexander Block trans = NULL; 51318ea05e3aSAlexander Block goto out; 51328ea05e3aSAlexander Block } 51338ea05e3aSAlexander Block 51348ea05e3aSAlexander Block sa->rtransid = trans->transid; 51358ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51368ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51378ea05e3aSAlexander Block 5138dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5139dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5140dd5f9615SStefan Behrens if (received_uuid_changed && 5141d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5142d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5143dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5144dd5f9615SStefan Behrens root->root_key.objectid); 5145d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5146d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5147d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5148d87ff758SNikolay Borisov goto out; 5149d87ff758SNikolay Borisov } 5150d87ff758SNikolay Borisov } 51518ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51528ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51538ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51543cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51553cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51563cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51573cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51588ea05e3aSAlexander Block 51590b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 51608ea05e3aSAlexander Block &root->root_key, &root->root_item); 51618ea05e3aSAlexander Block if (ret < 0) { 51623a45bb20SJeff Mahoney btrfs_end_transaction(trans); 51638ea05e3aSAlexander Block goto out; 5164dd5f9615SStefan Behrens } 5165dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5166cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5167dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5168dd5f9615SStefan Behrens root->root_key.objectid); 5169dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 517066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5171efd38150SNikolay Borisov btrfs_end_transaction(trans); 5172dd5f9615SStefan Behrens goto out; 5173dd5f9615SStefan Behrens } 5174dd5f9615SStefan Behrens } 51753a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5176abccd00fSHugo Mills out: 51770b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5178abccd00fSHugo Mills mnt_drop_write_file(file); 5179abccd00fSHugo Mills return ret; 5180abccd00fSHugo Mills } 5181abccd00fSHugo Mills 5182abccd00fSHugo Mills #ifdef CONFIG_64BIT 5183abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5184abccd00fSHugo Mills void __user *arg) 5185abccd00fSHugo Mills { 5186abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5187abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5188abccd00fSHugo Mills int ret = 0; 5189abccd00fSHugo Mills 5190abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 51917b9ea627SShailendra Verma if (IS_ERR(args32)) 51927b9ea627SShailendra Verma return PTR_ERR(args32); 5193abccd00fSHugo Mills 51948d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 519584dbeb87SDan Carpenter if (!args64) { 519684dbeb87SDan Carpenter ret = -ENOMEM; 5197abccd00fSHugo Mills goto out; 5198abccd00fSHugo Mills } 5199abccd00fSHugo Mills 5200abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5201abccd00fSHugo Mills args64->stransid = args32->stransid; 5202abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5203abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5204abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5205abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5206abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5207abccd00fSHugo Mills args64->flags = args32->flags; 5208abccd00fSHugo Mills 5209abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5210abccd00fSHugo Mills if (ret) 5211abccd00fSHugo Mills goto out; 5212abccd00fSHugo Mills 5213abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5214abccd00fSHugo Mills args32->stransid = args64->stransid; 5215abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5216abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5217abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5218abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5219abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5220abccd00fSHugo Mills args32->flags = args64->flags; 5221abccd00fSHugo Mills 5222abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5223abccd00fSHugo Mills if (ret) 5224abccd00fSHugo Mills ret = -EFAULT; 5225abccd00fSHugo Mills 5226abccd00fSHugo Mills out: 5227abccd00fSHugo Mills kfree(args32); 5228abccd00fSHugo Mills kfree(args64); 5229abccd00fSHugo Mills return ret; 5230abccd00fSHugo Mills } 5231abccd00fSHugo Mills #endif 5232abccd00fSHugo Mills 5233abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5234abccd00fSHugo Mills void __user *arg) 5235abccd00fSHugo Mills { 5236abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5237abccd00fSHugo Mills int ret = 0; 5238abccd00fSHugo Mills 5239abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 52407b9ea627SShailendra Verma if (IS_ERR(sa)) 52417b9ea627SShailendra Verma return PTR_ERR(sa); 5242abccd00fSHugo Mills 5243abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5244abccd00fSHugo Mills 5245abccd00fSHugo Mills if (ret) 5246abccd00fSHugo Mills goto out; 5247abccd00fSHugo Mills 52488ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52498ea05e3aSAlexander Block if (ret) 52508ea05e3aSAlexander Block ret = -EFAULT; 52518ea05e3aSAlexander Block 52528ea05e3aSAlexander Block out: 52538ea05e3aSAlexander Block kfree(sa); 52548ea05e3aSAlexander Block return ret; 52558ea05e3aSAlexander Block } 52568ea05e3aSAlexander Block 5257867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5258867ab667Sjeff.liu { 52590b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52600b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5261a1b83ac5SAnand Jain size_t len; 5262867ab667Sjeff.liu int ret; 5263a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5264a1b83ac5SAnand Jain 52650b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 52660b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 52670b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5268a1b83ac5SAnand Jain 5269a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5270867ab667Sjeff.liu 5271867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 52720b246afaSJeff Mahoney btrfs_warn(fs_info, 52730b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 52740b246afaSJeff Mahoney --len); 5275867ab667Sjeff.liu } 5276867ab667Sjeff.liu 5277867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5278867ab667Sjeff.liu 5279867ab667Sjeff.liu return ret ? -EFAULT : 0; 5280867ab667Sjeff.liu } 5281867ab667Sjeff.liu 5282a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5283a8bfd4abSjeff.liu { 52840b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52850b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52860b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52870b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5288a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5289a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5290a8bfd4abSjeff.liu int ret; 5291a8bfd4abSjeff.liu 5292a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5293a8bfd4abSjeff.liu return -EPERM; 5294a8bfd4abSjeff.liu 5295a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5296a8bfd4abSjeff.liu return -EFAULT; 5297a8bfd4abSjeff.liu 5298a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 52990b246afaSJeff Mahoney btrfs_err(fs_info, 53005d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5301a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5302a8bfd4abSjeff.liu return -EINVAL; 5303a8bfd4abSjeff.liu } 5304a8bfd4abSjeff.liu 5305a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5306a8bfd4abSjeff.liu if (ret) 5307a8bfd4abSjeff.liu return ret; 5308a8bfd4abSjeff.liu 5309a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5310a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5311a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5312a8bfd4abSjeff.liu goto out_unlock; 5313a8bfd4abSjeff.liu } 5314a8bfd4abSjeff.liu 53150b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5316a8bfd4abSjeff.liu strcpy(super_block->label, label); 53170b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53183a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5319a8bfd4abSjeff.liu 5320a8bfd4abSjeff.liu out_unlock: 5321a8bfd4abSjeff.liu mnt_drop_write_file(file); 5322a8bfd4abSjeff.liu return ret; 5323a8bfd4abSjeff.liu } 5324a8bfd4abSjeff.liu 53252eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 53262eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 53272eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53282eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53292eaa055fSJeff Mahoney 5330d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53312eaa055fSJeff Mahoney { 53324d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53332eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53342eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53352eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53362eaa055fSJeff Mahoney }; 53372eaa055fSJeff Mahoney 53382eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53392eaa055fSJeff Mahoney return -EFAULT; 53402eaa055fSJeff Mahoney 53412eaa055fSJeff Mahoney return 0; 53422eaa055fSJeff Mahoney } 53432eaa055fSJeff Mahoney 53442eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53452eaa055fSJeff Mahoney { 53460b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53480b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53492eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53502eaa055fSJeff Mahoney 53512eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53522eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53532eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53542eaa055fSJeff Mahoney 53552eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53562eaa055fSJeff Mahoney return -EFAULT; 53572eaa055fSJeff Mahoney 53582eaa055fSJeff Mahoney return 0; 53592eaa055fSJeff Mahoney } 53602eaa055fSJeff Mahoney 53612ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 53623b02a68aSJeff Mahoney enum btrfs_feature_set set, 53632eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53642eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53652eaa055fSJeff Mahoney { 53663b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 53673b02a68aSJeff Mahoney char *names; 53682eaa055fSJeff Mahoney u64 disallowed, unsupported; 53692eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53702eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53712eaa055fSJeff Mahoney 53722eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53732eaa055fSJeff Mahoney if (unsupported) { 53743b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53753b02a68aSJeff Mahoney if (names) { 53760b246afaSJeff Mahoney btrfs_warn(fs_info, 53773b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53783b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53793b02a68aSJeff Mahoney kfree(names); 53803b02a68aSJeff Mahoney } else 53810b246afaSJeff Mahoney btrfs_warn(fs_info, 53822eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53832eaa055fSJeff Mahoney type, unsupported); 53842eaa055fSJeff Mahoney return -EOPNOTSUPP; 53852eaa055fSJeff Mahoney } 53862eaa055fSJeff Mahoney 53872eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53882eaa055fSJeff Mahoney if (disallowed) { 53893b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53903b02a68aSJeff Mahoney if (names) { 53910b246afaSJeff Mahoney btrfs_warn(fs_info, 53923b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53933b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53943b02a68aSJeff Mahoney kfree(names); 53953b02a68aSJeff Mahoney } else 53960b246afaSJeff Mahoney btrfs_warn(fs_info, 53972eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53982eaa055fSJeff Mahoney type, disallowed); 53992eaa055fSJeff Mahoney return -EPERM; 54002eaa055fSJeff Mahoney } 54012eaa055fSJeff Mahoney 54022eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 54032eaa055fSJeff Mahoney if (disallowed) { 54043b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54053b02a68aSJeff Mahoney if (names) { 54060b246afaSJeff Mahoney btrfs_warn(fs_info, 54073b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 54083b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54093b02a68aSJeff Mahoney kfree(names); 54103b02a68aSJeff Mahoney } else 54110b246afaSJeff Mahoney btrfs_warn(fs_info, 54122eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 54132eaa055fSJeff Mahoney type, disallowed); 54142eaa055fSJeff Mahoney return -EPERM; 54152eaa055fSJeff Mahoney } 54162eaa055fSJeff Mahoney 54172eaa055fSJeff Mahoney return 0; 54182eaa055fSJeff Mahoney } 54192eaa055fSJeff Mahoney 54202ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 54212ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 54222eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 54232eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 54242eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 54252eaa055fSJeff Mahoney 54262eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 54272eaa055fSJeff Mahoney { 54280b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54290b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54300b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54310b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 54322eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54332eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54342eaa055fSJeff Mahoney u64 newflags; 54352eaa055fSJeff Mahoney int ret; 54362eaa055fSJeff Mahoney 54372eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54382eaa055fSJeff Mahoney return -EPERM; 54392eaa055fSJeff Mahoney 54402eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54412eaa055fSJeff Mahoney return -EFAULT; 54422eaa055fSJeff Mahoney 54432eaa055fSJeff Mahoney /* Nothing to do */ 54442eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54452eaa055fSJeff Mahoney !flags[0].incompat_flags) 54462eaa055fSJeff Mahoney return 0; 54472eaa055fSJeff Mahoney 54482ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 54492eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54502eaa055fSJeff Mahoney if (ret) 54512eaa055fSJeff Mahoney return ret; 54522eaa055fSJeff Mahoney 54532ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 54542eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54552eaa055fSJeff Mahoney if (ret) 54562eaa055fSJeff Mahoney return ret; 54572eaa055fSJeff Mahoney 54582ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 54592eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54602eaa055fSJeff Mahoney if (ret) 54612eaa055fSJeff Mahoney return ret; 54622eaa055fSJeff Mahoney 54637ab19625SDavid Sterba ret = mnt_want_write_file(file); 54647ab19625SDavid Sterba if (ret) 54657ab19625SDavid Sterba return ret; 54667ab19625SDavid Sterba 54678051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54687ab19625SDavid Sterba if (IS_ERR(trans)) { 54697ab19625SDavid Sterba ret = PTR_ERR(trans); 54707ab19625SDavid Sterba goto out_drop_write; 54717ab19625SDavid Sterba } 54722eaa055fSJeff Mahoney 54730b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 54742eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54752eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54762eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54772eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54782eaa055fSJeff Mahoney 54792eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54802eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54812eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54822eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54832eaa055fSJeff Mahoney 54842eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54852eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54862eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54872eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54880b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 54892eaa055fSJeff Mahoney 54903a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 54917ab19625SDavid Sterba out_drop_write: 54927ab19625SDavid Sterba mnt_drop_write_file(file); 54937ab19625SDavid Sterba 54947ab19625SDavid Sterba return ret; 54952eaa055fSJeff Mahoney } 54962eaa055fSJeff Mahoney 54972351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 54982351f431SJosef Bacik { 54992351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 55002351f431SJosef Bacik int ret; 55012351f431SJosef Bacik 55022351f431SJosef Bacik if (compat) { 55032351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55042351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 55052351f431SJosef Bacik 55062351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 55072351f431SJosef Bacik if (ret) 55082351f431SJosef Bacik return -EFAULT; 55092351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 55102351f431SJosef Bacik if (!arg) 55112351f431SJosef Bacik return -ENOMEM; 55122351f431SJosef Bacik arg->send_fd = args32.send_fd; 55132351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 55142351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 55152351f431SJosef Bacik arg->parent_root = args32.parent_root; 55162351f431SJosef Bacik arg->flags = args32.flags; 55172351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 55182351f431SJosef Bacik sizeof(args32.reserved)); 55192351f431SJosef Bacik #else 55202351f431SJosef Bacik return -ENOTTY; 55212351f431SJosef Bacik #endif 55222351f431SJosef Bacik } else { 55232351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 55242351f431SJosef Bacik if (IS_ERR(arg)) 55252351f431SJosef Bacik return PTR_ERR(arg); 55262351f431SJosef Bacik } 55272351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 55282351f431SJosef Bacik kfree(arg); 55292351f431SJosef Bacik return ret; 55302351f431SJosef Bacik } 55312351f431SJosef Bacik 5532f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5533f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5534f46b5a66SChristoph Hellwig { 55350b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55360b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55370b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55384bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5539f46b5a66SChristoph Hellwig 5540f46b5a66SChristoph Hellwig switch (cmd) { 55416cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 55426cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 55436cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 55446cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 55456cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 55466cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5547f7039b1dSLi Dongyang case FITRIM: 5548f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5549f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5550fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5551fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5552fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 55533de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5554fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 55556f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 55566f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 555776dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 555876dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 55590caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 55600caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 55610caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 55620caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55636ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55646ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5565f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55661e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55671e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55681e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5569f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5570198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5571f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55722ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5573f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5574da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55756b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55766b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5577475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 55782ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5579475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 55802ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5581f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55829ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5583ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5584ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5585cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5586cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5587ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5588ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5589d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5590d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5591d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5592d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5593d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5594d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 55951406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55962ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 55979b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55989b199859SFilipe David Borba Manana int ret; 55999b199859SFilipe David Borba Manana 560082b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 56019b199859SFilipe David Borba Manana if (ret) 56029b199859SFilipe David Borba Manana return ret; 56030b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 56042fad4e83SDavid Sterba /* 56052fad4e83SDavid Sterba * The transaction thread may want to do more work, 560601327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 56072fad4e83SDavid Sterba * processing uncleaned subvols. 56082fad4e83SDavid Sterba */ 56090b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 56109b199859SFilipe David Borba Manana return ret; 56119b199859SFilipe David Borba Manana } 561246204592SSage Weil case BTRFS_IOC_START_SYNC: 56139a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 561446204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 56152ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5616475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5617b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5618475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 56192ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5620475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 56212ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5622c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 56239ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5624837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 56252ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 562619a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 56272ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 56288ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 56298ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5630abccd00fSHugo Mills #ifdef CONFIG_64BIT 5631abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5632abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5633abccd00fSHugo Mills #endif 563431db9f7cSAlexander Block case BTRFS_IOC_SEND: 56352351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 56362351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 56372351f431SJosef Bacik case BTRFS_IOC_SEND_32: 56382351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 56392351f431SJosef Bacik #endif 5640c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 56412ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 56425d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5643905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 56445d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5645905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 56465d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5647905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 56485d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5649905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 56502f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 56512f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 56522f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 56532f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 565457254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 565557254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 56563f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 56572ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5658867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5659867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5660a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5661a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 56622eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5663d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 56642eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 56652eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 56662eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56672eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5668e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5669e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5670025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5671025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5672b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5673b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 567442e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 567542e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 567623d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 567723d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5678f46b5a66SChristoph Hellwig } 5679f46b5a66SChristoph Hellwig 5680f46b5a66SChristoph Hellwig return -ENOTTY; 5681f46b5a66SChristoph Hellwig } 56824c63c245SLuke Dashjr 56834c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56844c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56854c63c245SLuke Dashjr { 56862a362249SJeff Mahoney /* 56872a362249SJeff Mahoney * These all access 32-bit values anyway so no further 56882a362249SJeff Mahoney * handling is necessary. 56892a362249SJeff Mahoney */ 56904c63c245SLuke Dashjr switch (cmd) { 56914c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56924c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56934c63c245SLuke Dashjr break; 56944c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56954c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56964c63c245SLuke Dashjr break; 56974c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56984c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56994c63c245SLuke Dashjr break; 57004c63c245SLuke Dashjr } 57014c63c245SLuke Dashjr 57024c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 57034c63c245SLuke Dashjr } 57044c63c245SLuke Dashjr #endif 5705