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; 1905c57b8b6SDavid Sterba unsigned int fsflags, old_fsflags; 1916cbff00fSChristoph Hellwig int ret; 1925c57b8b6SDavid Sterba u64 old_flags; 1935c57b8b6SDavid Sterba unsigned int old_i_flags; 1947e97b8daSDavid Sterba umode_t mode; 1956cbff00fSChristoph Hellwig 196bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 197bd60ea0fSDavid Sterba return -EPERM; 198bd60ea0fSDavid Sterba 199b83cc969SLi Zefan if (btrfs_root_readonly(root)) 200b83cc969SLi Zefan return -EROFS; 201b83cc969SLi Zefan 2025c57b8b6SDavid Sterba if (copy_from_user(&fsflags, arg, sizeof(fsflags))) 2036cbff00fSChristoph Hellwig return -EFAULT; 2046cbff00fSChristoph Hellwig 2055c57b8b6SDavid Sterba ret = check_fsflags(fsflags); 20675e7cb7fSLiu Bo if (ret) 20775e7cb7fSLiu Bo return ret; 2086cbff00fSChristoph Hellwig 209e7848683SJan Kara ret = mnt_want_write_file(file); 210e7848683SJan Kara if (ret) 211e7848683SJan Kara return ret; 212e7848683SJan Kara 2135955102cSAl Viro inode_lock(inode); 2146cbff00fSChristoph Hellwig 2155c57b8b6SDavid Sterba old_flags = binode->flags; 2165c57b8b6SDavid Sterba old_i_flags = inode->i_flags; 2177e97b8daSDavid Sterba mode = inode->i_mode; 218f062abf0SLi Zefan 2195c57b8b6SDavid Sterba fsflags = btrfs_mask_fsflags_for_type(inode, fsflags); 2205c57b8b6SDavid Sterba old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags); 2215c57b8b6SDavid Sterba if ((fsflags ^ old_fsflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2226cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2236cbff00fSChristoph Hellwig ret = -EPERM; 2246cbff00fSChristoph Hellwig goto out_unlock; 2256cbff00fSChristoph Hellwig } 2266cbff00fSChristoph Hellwig } 2276cbff00fSChristoph Hellwig 2285c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 2295c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 2306cbff00fSChristoph Hellwig else 2315c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 2325c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 2335c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 2346cbff00fSChristoph Hellwig else 2355c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 2365c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 2375c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 2386cbff00fSChristoph Hellwig else 2395c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 2405c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 2415c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 2426cbff00fSChristoph Hellwig else 2435c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 2445c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 2455c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 2466cbff00fSChristoph Hellwig else 2475c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 2485c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 2495c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_DIRSYNC; 2506cbff00fSChristoph Hellwig else 2515c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_DIRSYNC; 2525c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 2537e97b8daSDavid Sterba if (S_ISREG(mode)) { 2547e97b8daSDavid Sterba /* 2557e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2567e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2577e97b8daSDavid Sterba * status of the file and will not set it. 2587e97b8daSDavid Sterba */ 2597e97b8daSDavid Sterba if (inode->i_size == 0) 2605c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NODATACOW 2617e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM; 2627e97b8daSDavid Sterba } else { 2635c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NODATACOW; 2647e97b8daSDavid Sterba } 2657e97b8daSDavid Sterba } else { 2667e97b8daSDavid Sterba /* 26701327610SNicholas D Steeves * Revert back under same assumptions as above 2687e97b8daSDavid Sterba */ 2697e97b8daSDavid Sterba if (S_ISREG(mode)) { 2707e97b8daSDavid Sterba if (inode->i_size == 0) 2715c57b8b6SDavid Sterba binode->flags &= ~(BTRFS_INODE_NODATACOW 2727e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM); 2737e97b8daSDavid Sterba } else { 2745c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_NODATACOW; 2757e97b8daSDavid Sterba } 2767e97b8daSDavid Sterba } 2776cbff00fSChristoph Hellwig 27875e7cb7fSLiu Bo /* 27975e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 28075e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 28175e7cb7fSLiu Bo * things smaller. 28275e7cb7fSLiu Bo */ 2835c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 2845c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_COMPRESS; 2855c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NOCOMPRESS; 28663541927SFilipe David Borba Manana 28763541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 28863541927SFilipe David Borba Manana if (ret && ret != -ENODATA) 28963541927SFilipe David Borba Manana goto out_drop; 2905c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 29163541927SFilipe David Borba Manana const char *comp; 29263541927SFilipe David Borba Manana 2935c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_COMPRESS; 2945c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_NOCOMPRESS; 29563541927SFilipe David Borba Manana 29693370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 29793370509SDavid Sterba if (!comp || comp[0] == 0) 29893370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 29993370509SDavid Sterba 30063541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", 30163541927SFilipe David Borba Manana comp, strlen(comp), 0); 30263541927SFilipe David Borba Manana if (ret) 30363541927SFilipe David Borba Manana goto out_drop; 30463541927SFilipe David Borba Manana 305ebcb904dSLi Zefan } else { 30678a017a2SFilipe Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 30778a017a2SFilipe Manana if (ret && ret != -ENODATA) 30878a017a2SFilipe Manana goto out_drop; 3095c57b8b6SDavid Sterba binode->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 31075e7cb7fSLiu Bo } 3116cbff00fSChristoph Hellwig 3124da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 313f062abf0SLi Zefan if (IS_ERR(trans)) { 314f062abf0SLi Zefan ret = PTR_ERR(trans); 315f062abf0SLi Zefan goto out_drop; 316f062abf0SLi Zefan } 3176cbff00fSChristoph Hellwig 3187b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3190c4d2d95SJosef Bacik inode_inc_iversion(inode); 320c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3216cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3226cbff00fSChristoph Hellwig 3233a45bb20SJeff Mahoney btrfs_end_transaction(trans); 324f062abf0SLi Zefan out_drop: 325f062abf0SLi Zefan if (ret) { 3265c57b8b6SDavid Sterba binode->flags = old_flags; 3275c57b8b6SDavid Sterba inode->i_flags = old_i_flags; 328f062abf0SLi Zefan } 3296cbff00fSChristoph Hellwig 3306cbff00fSChristoph Hellwig out_unlock: 3315955102cSAl Viro inode_unlock(inode); 332e7848683SJan Kara mnt_drop_write_file(file); 3332d4e6f6aSliubo return ret; 3346cbff00fSChristoph Hellwig } 3356cbff00fSChristoph Hellwig 33619f93b3cSDavid Sterba /* 33719f93b3cSDavid Sterba * Translate btrfs internal inode flags to xflags as expected by the 33819f93b3cSDavid Sterba * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are 33919f93b3cSDavid Sterba * silently dropped. 34019f93b3cSDavid Sterba */ 34119f93b3cSDavid Sterba static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags) 34219f93b3cSDavid Sterba { 34319f93b3cSDavid Sterba unsigned int xflags = 0; 34419f93b3cSDavid Sterba 34519f93b3cSDavid Sterba if (flags & BTRFS_INODE_APPEND) 34619f93b3cSDavid Sterba xflags |= FS_XFLAG_APPEND; 34719f93b3cSDavid Sterba if (flags & BTRFS_INODE_IMMUTABLE) 34819f93b3cSDavid Sterba xflags |= FS_XFLAG_IMMUTABLE; 34919f93b3cSDavid Sterba if (flags & BTRFS_INODE_NOATIME) 35019f93b3cSDavid Sterba xflags |= FS_XFLAG_NOATIME; 35119f93b3cSDavid Sterba if (flags & BTRFS_INODE_NODUMP) 35219f93b3cSDavid Sterba xflags |= FS_XFLAG_NODUMP; 35319f93b3cSDavid Sterba if (flags & BTRFS_INODE_SYNC) 35419f93b3cSDavid Sterba xflags |= FS_XFLAG_SYNC; 35519f93b3cSDavid Sterba 35619f93b3cSDavid Sterba return xflags; 35719f93b3cSDavid Sterba } 35819f93b3cSDavid Sterba 35919f93b3cSDavid Sterba /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */ 36019f93b3cSDavid Sterba static int check_xflags(unsigned int flags) 36119f93b3cSDavid Sterba { 36219f93b3cSDavid Sterba if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME | 36319f93b3cSDavid Sterba FS_XFLAG_NODUMP | FS_XFLAG_SYNC)) 36419f93b3cSDavid Sterba return -EOPNOTSUPP; 36519f93b3cSDavid Sterba return 0; 36619f93b3cSDavid Sterba } 36719f93b3cSDavid Sterba 368e4202ac9SDavid Sterba /* 369e4202ac9SDavid Sterba * Set the xflags from the internal inode flags. The remaining items of fsxattr 370e4202ac9SDavid Sterba * are zeroed. 371e4202ac9SDavid Sterba */ 372e4202ac9SDavid Sterba static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg) 373e4202ac9SDavid Sterba { 374e4202ac9SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 375e4202ac9SDavid Sterba struct fsxattr fa; 376e4202ac9SDavid Sterba 377e4202ac9SDavid Sterba memset(&fa, 0, sizeof(fa)); 378e4202ac9SDavid Sterba fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags); 379e4202ac9SDavid Sterba 380e4202ac9SDavid Sterba if (copy_to_user(arg, &fa, sizeof(fa))) 381e4202ac9SDavid Sterba return -EFAULT; 382e4202ac9SDavid Sterba 383e4202ac9SDavid Sterba return 0; 384e4202ac9SDavid Sterba } 385e4202ac9SDavid Sterba 386025f2121SDavid Sterba static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg) 387025f2121SDavid Sterba { 388025f2121SDavid Sterba struct inode *inode = file_inode(file); 389025f2121SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 390025f2121SDavid Sterba struct btrfs_root *root = binode->root; 391025f2121SDavid Sterba struct btrfs_trans_handle *trans; 392025f2121SDavid Sterba struct fsxattr fa; 393025f2121SDavid Sterba unsigned old_flags; 394025f2121SDavid Sterba unsigned old_i_flags; 395025f2121SDavid Sterba int ret = 0; 396025f2121SDavid Sterba 397025f2121SDavid Sterba if (!inode_owner_or_capable(inode)) 398025f2121SDavid Sterba return -EPERM; 399025f2121SDavid Sterba 400025f2121SDavid Sterba if (btrfs_root_readonly(root)) 401025f2121SDavid Sterba return -EROFS; 402025f2121SDavid Sterba 403025f2121SDavid Sterba memset(&fa, 0, sizeof(fa)); 404025f2121SDavid Sterba if (copy_from_user(&fa, arg, sizeof(fa))) 405025f2121SDavid Sterba return -EFAULT; 406025f2121SDavid Sterba 407025f2121SDavid Sterba ret = check_xflags(fa.fsx_xflags); 408025f2121SDavid Sterba if (ret) 409025f2121SDavid Sterba return ret; 410025f2121SDavid Sterba 411025f2121SDavid Sterba if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0) 412025f2121SDavid Sterba return -EOPNOTSUPP; 413025f2121SDavid Sterba 414025f2121SDavid Sterba ret = mnt_want_write_file(file); 415025f2121SDavid Sterba if (ret) 416025f2121SDavid Sterba return ret; 417025f2121SDavid Sterba 418025f2121SDavid Sterba inode_lock(inode); 419025f2121SDavid Sterba 420025f2121SDavid Sterba old_flags = binode->flags; 421025f2121SDavid Sterba old_i_flags = inode->i_flags; 422025f2121SDavid Sterba 423025f2121SDavid Sterba /* We need the capabilities to change append-only or immutable inode */ 424025f2121SDavid Sterba if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) || 425025f2121SDavid Sterba (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) && 426025f2121SDavid Sterba !capable(CAP_LINUX_IMMUTABLE)) { 427025f2121SDavid Sterba ret = -EPERM; 428025f2121SDavid Sterba goto out_unlock; 429025f2121SDavid Sterba } 430025f2121SDavid Sterba 431025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_SYNC) 432025f2121SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 433025f2121SDavid Sterba else 434025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 435025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE) 436025f2121SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 437025f2121SDavid Sterba else 438025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 439025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_APPEND) 440025f2121SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 441025f2121SDavid Sterba else 442025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 443025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NODUMP) 444025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 445025f2121SDavid Sterba else 446025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 447025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NOATIME) 448025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 449025f2121SDavid Sterba else 450025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 451025f2121SDavid Sterba 452025f2121SDavid Sterba /* 1 item for the inode */ 453025f2121SDavid Sterba trans = btrfs_start_transaction(root, 1); 454025f2121SDavid Sterba if (IS_ERR(trans)) { 455025f2121SDavid Sterba ret = PTR_ERR(trans); 456025f2121SDavid Sterba goto out_unlock; 457025f2121SDavid Sterba } 458025f2121SDavid Sterba 459025f2121SDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 460025f2121SDavid Sterba inode_inc_iversion(inode); 461025f2121SDavid Sterba inode->i_ctime = current_time(inode); 462025f2121SDavid Sterba ret = btrfs_update_inode(trans, root, inode); 463025f2121SDavid Sterba 464025f2121SDavid Sterba btrfs_end_transaction(trans); 465025f2121SDavid Sterba 466025f2121SDavid Sterba out_unlock: 467025f2121SDavid Sterba if (ret) { 468025f2121SDavid Sterba binode->flags = old_flags; 469025f2121SDavid Sterba inode->i_flags = old_i_flags; 470025f2121SDavid Sterba } 471025f2121SDavid Sterba 472025f2121SDavid Sterba inode_unlock(inode); 473025f2121SDavid Sterba mnt_drop_write_file(file); 474025f2121SDavid Sterba 475025f2121SDavid Sterba return ret; 476025f2121SDavid Sterba } 477025f2121SDavid Sterba 4786cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4796cbff00fSChristoph Hellwig { 480496ad9aaSAl Viro struct inode *inode = file_inode(file); 4816cbff00fSChristoph Hellwig 4826cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4836cbff00fSChristoph Hellwig } 484f46b5a66SChristoph Hellwig 485f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 486f7039b1dSLi Dongyang { 4870b246afaSJeff Mahoney struct inode *inode = file_inode(file); 4880b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 489f7039b1dSLi Dongyang struct btrfs_device *device; 490f7039b1dSLi Dongyang struct request_queue *q; 491f7039b1dSLi Dongyang struct fstrim_range range; 492f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 493f7039b1dSLi Dongyang u64 num_devices = 0; 494815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 495f7039b1dSLi Dongyang int ret; 496f7039b1dSLi Dongyang 497f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 498f7039b1dSLi Dongyang return -EPERM; 499f7039b1dSLi Dongyang 5001f78160cSXiao Guangrong rcu_read_lock(); 5011f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 5021f78160cSXiao Guangrong dev_list) { 503f7039b1dSLi Dongyang if (!device->bdev) 504f7039b1dSLi Dongyang continue; 505f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 506f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 507f7039b1dSLi Dongyang num_devices++; 50850d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 509f7039b1dSLi Dongyang minlen); 510f7039b1dSLi Dongyang } 511f7039b1dSLi Dongyang } 5121f78160cSXiao Guangrong rcu_read_unlock(); 513f4c697e6SLukas Czerner 514f7039b1dSLi Dongyang if (!num_devices) 515f7039b1dSLi Dongyang return -EOPNOTSUPP; 516f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 517f7039b1dSLi Dongyang return -EFAULT; 518e515c18bSLukas Czerner if (range.start > total_bytes || 519e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 520f4c697e6SLukas Czerner return -EINVAL; 521f7039b1dSLi Dongyang 522f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 523f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5242ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 525f7039b1dSLi Dongyang if (ret < 0) 526f7039b1dSLi Dongyang return ret; 527f7039b1dSLi Dongyang 528f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 529f7039b1dSLi Dongyang return -EFAULT; 530f7039b1dSLi Dongyang 531f7039b1dSLi Dongyang return 0; 532f7039b1dSLi Dongyang } 533f7039b1dSLi Dongyang 534dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 535dd5f9615SStefan Behrens { 53646e0f66aSChris Mason int i; 53746e0f66aSChris Mason 53846e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 53946e0f66aSChris Mason if (uuid[i]) 54046e0f66aSChris Mason return 0; 54146e0f66aSChris Mason } 54246e0f66aSChris Mason return 1; 543dd5f9615SStefan Behrens } 544dd5f9615SStefan Behrens 545d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 546cb8e7090SChristoph Hellwig struct dentry *dentry, 54752f75f4fSDavid Sterba const char *name, int namelen, 5486f72c7e2SArne Jansen u64 *async_transid, 5498696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 550f46b5a66SChristoph Hellwig { 5510b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 552f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 553f46b5a66SChristoph Hellwig struct btrfs_key key; 55449a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 555f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 556f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 557d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 55876dda93cSYan, Zheng struct btrfs_root *new_root; 559d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 56095582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5615662344bSTsutomu Itoh struct inode *inode; 562f46b5a66SChristoph Hellwig int ret; 563f46b5a66SChristoph Hellwig int err; 564f46b5a66SChristoph Hellwig u64 objectid; 565f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 5663de4586cSChris Mason u64 index = 0; 5678ea05e3aSAlexander Block uuid_le new_uuid; 568f46b5a66SChristoph Hellwig 56949a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 57049a3c4d9SDavid Sterba if (!root_item) 57149a3c4d9SDavid Sterba return -ENOMEM; 57249a3c4d9SDavid Sterba 5730b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5742fbe8c8aSAl Viro if (ret) 57549a3c4d9SDavid Sterba goto fail_free; 5766a912213SJosef Bacik 577e09fe2d2SQu Wenruo /* 578e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 57901327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 580e09fe2d2SQu Wenruo */ 58149a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 58249a3c4d9SDavid Sterba ret = -ENOSPC; 58349a3c4d9SDavid Sterba goto fail_free; 58449a3c4d9SDavid Sterba } 585e09fe2d2SQu Wenruo 586d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 5879ed74f2dSJosef Bacik /* 588d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 589d5c12070SMiao Xie * of create_snapshot(). 5909ed74f2dSJosef Bacik */ 591c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 592d5c12070SMiao Xie if (ret) 59349a3c4d9SDavid Sterba goto fail_free; 594f46b5a66SChristoph Hellwig 595d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 596d5c12070SMiao Xie if (IS_ERR(trans)) { 597d5c12070SMiao Xie ret = PTR_ERR(trans); 5987775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 59949a3c4d9SDavid Sterba goto fail_free; 600d5c12070SMiao Xie } 601d5c12070SMiao Xie trans->block_rsv = &block_rsv; 602d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 603f46b5a66SChristoph Hellwig 6040b246afaSJeff Mahoney ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); 6056f72c7e2SArne Jansen if (ret) 6066f72c7e2SArne Jansen goto fail; 6076f72c7e2SArne Jansen 6084d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6098e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6108e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6118e8a1e31SJosef Bacik goto fail; 6128e8a1e31SJosef Bacik } 613f46b5a66SChristoph Hellwig 614f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 615f46b5a66SChristoph Hellwig 61649a3c4d9SDavid Sterba inode_item = &root_item->inode; 6173cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6183cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6193cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 620da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6210b246afaSJeff Mahoney fs_info->nodesize); 6223cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 623f46b5a66SChristoph Hellwig 62449a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 62549a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6263cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 62708fe4db1SLi Zefan 62849a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 62949a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 63049a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 63149a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 63249a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 63349a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 634f46b5a66SChristoph Hellwig 63549a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 63649a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6378ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 63849a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 63949a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 64049a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 64149a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 64249a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 64349a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 644f46b5a66SChristoph Hellwig 645925baeddSChris Mason btrfs_tree_unlock(leaf); 646f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 647f46b5a66SChristoph Hellwig leaf = NULL; 648f46b5a66SChristoph Hellwig 64949a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 650f46b5a66SChristoph Hellwig 651f46b5a66SChristoph Hellwig key.objectid = objectid; 6525d4f98a2SYan Zheng key.offset = 0; 653962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6540b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 65549a3c4d9SDavid Sterba root_item); 656f46b5a66SChristoph Hellwig if (ret) 657f46b5a66SChristoph Hellwig goto fail; 658f46b5a66SChristoph Hellwig 65976dda93cSYan, Zheng key.offset = (u64)-1; 6600b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 66179787eaaSJeff Mahoney if (IS_ERR(new_root)) { 66279787eaaSJeff Mahoney ret = PTR_ERR(new_root); 66366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 66479787eaaSJeff Mahoney goto fail; 66579787eaaSJeff Mahoney } 66676dda93cSYan, Zheng 66776dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 66876dda93cSYan, Zheng 66963541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 670ce598979SMark Fasheh if (ret) { 671ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 67266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 673ce598979SMark Fasheh goto fail; 674ce598979SMark Fasheh } 675ce598979SMark Fasheh 676f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 677f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 678f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 679f32e48e9SChandan Rajendra 680f46b5a66SChristoph Hellwig /* 681f46b5a66SChristoph Hellwig * insert the directory item 682f46b5a66SChristoph Hellwig */ 683877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 68479787eaaSJeff Mahoney if (ret) { 68566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 68679787eaaSJeff Mahoney goto fail; 68779787eaaSJeff Mahoney } 6883de4586cSChris Mason 6893de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 6908e7611cfSNikolay Borisov name, namelen, BTRFS_I(dir), &key, 6913de4586cSChris Mason BTRFS_FT_DIR, index); 69279787eaaSJeff Mahoney if (ret) { 69366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 694f46b5a66SChristoph Hellwig goto fail; 69579787eaaSJeff Mahoney } 6960660b5afSChris Mason 6976ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 69852c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 69952c26179SYan Zheng BUG_ON(ret); 70052c26179SYan Zheng 7010b246afaSJeff Mahoney ret = btrfs_add_root_ref(trans, fs_info, 7024df27c4dSYan, Zheng objectid, root->root_key.objectid, 7034a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7040660b5afSChris Mason BUG_ON(ret); 7050660b5afSChris Mason 706cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7076bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 708dd5f9615SStefan Behrens if (ret) 70966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 710dd5f9615SStefan Behrens 711f46b5a66SChristoph Hellwig fail: 71249a3c4d9SDavid Sterba kfree(root_item); 713d5c12070SMiao Xie trans->block_rsv = NULL; 714d5c12070SMiao Xie trans->bytes_reserved = 0; 7157775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 716de6e8200SLiu Bo 71772fd032eSSage Weil if (async_transid) { 71872fd032eSSage Weil *async_transid = trans->transid; 7193a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 72000d71c9cSMiao Xie if (err) 7213a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 72272fd032eSSage Weil } else { 7233a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 72472fd032eSSage Weil } 725f46b5a66SChristoph Hellwig if (err && !ret) 726f46b5a66SChristoph Hellwig ret = err; 7271a65e24bSChris Mason 7285662344bSTsutomu Itoh if (!ret) { 7295662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 730de6e8200SLiu Bo if (IS_ERR(inode)) 731de6e8200SLiu Bo return PTR_ERR(inode); 7325662344bSTsutomu Itoh d_instantiate(dentry, inode); 7335662344bSTsutomu Itoh } 734f46b5a66SChristoph Hellwig return ret; 73549a3c4d9SDavid Sterba 73649a3c4d9SDavid Sterba fail_free: 73749a3c4d9SDavid Sterba kfree(root_item); 73849a3c4d9SDavid Sterba return ret; 739f46b5a66SChristoph Hellwig } 740f46b5a66SChristoph Hellwig 741e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 74261d7e4cbSDavid Sterba struct dentry *dentry, 743e9662f70SMiao Xie u64 *async_transid, bool readonly, 744e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 745f46b5a66SChristoph Hellwig { 7460b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7472e4bfab9SYan, Zheng struct inode *inode; 748f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 749f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7502e4bfab9SYan, Zheng int ret; 751f46b5a66SChristoph Hellwig 75227cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 753f46b5a66SChristoph Hellwig return -EINVAL; 754f46b5a66SChristoph Hellwig 75523269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 756a1ee7362SDavid Sterba if (!pending_snapshot) 757a1ee7362SDavid Sterba return -ENOMEM; 758a1ee7362SDavid Sterba 759b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 76023269bf5SDavid Sterba GFP_KERNEL); 7618546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7628546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 763b0c0ea63SDavid Sterba ret = -ENOMEM; 764b0c0ea63SDavid Sterba goto free_pending; 765b0c0ea63SDavid Sterba } 766b0c0ea63SDavid Sterba 767ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7684e857c58SPeter Zijlstra smp_mb__after_atomic(); 76945bac0f3SLiu Bo /* wait for no snapshot writes */ 77045bac0f3SLiu Bo wait_event(root->subv_writers->wait, 77145bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 7728257b2dcSMiao Xie 77376f32e24SNikolay Borisov ret = btrfs_start_delalloc_inodes(root); 7746a03843dSMiao Xie if (ret) 775a1ee7362SDavid Sterba goto dec_and_free; 7766a03843dSMiao Xie 7776374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 7786a03843dSMiao Xie 77966d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 78066d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 781d5c12070SMiao Xie /* 782d5c12070SMiao Xie * 1 - parent dir inode 783d5c12070SMiao Xie * 2 - dir entries 784d5c12070SMiao Xie * 1 - root item 785d5c12070SMiao Xie * 2 - root ref/backref 786d5c12070SMiao Xie * 1 - root of snapshot 787dd5f9615SStefan Behrens * 1 - UUID item 788d5c12070SMiao Xie */ 789d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 790dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 791ee3441b4SJeff Mahoney false); 792d5c12070SMiao Xie if (ret) 793a1ee7362SDavid Sterba goto dec_and_free; 794d5c12070SMiao Xie 795a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 796a22285a6SYan, Zheng pending_snapshot->root = root; 797b83cc969SLi Zefan pending_snapshot->readonly = readonly; 798e9662f70SMiao Xie pending_snapshot->dir = dir; 7998696c533SMiao Xie pending_snapshot->inherit = inherit; 800a22285a6SYan, Zheng 801d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 802a22285a6SYan, Zheng if (IS_ERR(trans)) { 803a22285a6SYan, Zheng ret = PTR_ERR(trans); 804a22285a6SYan, Zheng goto fail; 805a22285a6SYan, Zheng } 806a22285a6SYan, Zheng 8070b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 808a22285a6SYan, Zheng list_add(&pending_snapshot->list, 809a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8100b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 81172fd032eSSage Weil if (async_transid) { 81272fd032eSSage Weil *async_transid = trans->transid; 8133a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 81400d71c9cSMiao Xie if (ret) 8153a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 81672fd032eSSage Weil } else { 8173a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 81872fd032eSSage Weil } 819aec8030aSMiao Xie if (ret) 820c37b2b62SJosef Bacik goto fail; 821a22285a6SYan, Zheng 822a22285a6SYan, Zheng ret = pending_snapshot->error; 823f46b5a66SChristoph Hellwig if (ret) 8242e4bfab9SYan, Zheng goto fail; 825f46b5a66SChristoph Hellwig 826d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 827d3797308SChris Mason if (ret) 828d3797308SChris Mason goto fail; 829d3797308SChris Mason 8302b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8312e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8322e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8332e4bfab9SYan, Zheng goto fail; 8342e4bfab9SYan, Zheng } 8355662344bSTsutomu Itoh 8362e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8372e4bfab9SYan, Zheng ret = 0; 8382e4bfab9SYan, Zheng fail: 8397775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 840a1ee7362SDavid Sterba dec_and_free: 841ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8424625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 843b0c0ea63SDavid Sterba free_pending: 844b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8458546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 846a1ee7362SDavid Sterba kfree(pending_snapshot); 847a1ee7362SDavid Sterba 848f46b5a66SChristoph Hellwig return ret; 849f46b5a66SChristoph Hellwig } 850f46b5a66SChristoph Hellwig 8514260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8524260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8534260f7c7SSage Weil * whether the type of victim is right. 8544260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8554260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8564260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8574260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8584260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8594260f7c7SSage Weil * a. be owner of dir, or 8604260f7c7SSage Weil * b. be owner of victim, or 8614260f7c7SSage Weil * c. have CAP_FOWNER capability 86201327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8634260f7c7SSage Weil * links pointing to it. 8644260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8654260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8664260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8674260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8684260f7c7SSage Weil * nfs_async_unlink(). 8694260f7c7SSage Weil */ 8704260f7c7SSage Weil 8714260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 8724260f7c7SSage Weil { 8734260f7c7SSage Weil int error; 8744260f7c7SSage Weil 8752b0143b5SDavid Howells if (d_really_is_negative(victim)) 8764260f7c7SSage Weil return -ENOENT; 8774260f7c7SSage Weil 8782b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8794fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8804260f7c7SSage Weil 8814260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 8824260f7c7SSage Weil if (error) 8834260f7c7SSage Weil return error; 8844260f7c7SSage Weil if (IS_APPEND(dir)) 8854260f7c7SSage Weil return -EPERM; 8862b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 8872b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 8884260f7c7SSage Weil return -EPERM; 8894260f7c7SSage Weil if (isdir) { 890e36cb0b8SDavid Howells if (!d_is_dir(victim)) 8914260f7c7SSage Weil return -ENOTDIR; 8924260f7c7SSage Weil if (IS_ROOT(victim)) 8934260f7c7SSage Weil return -EBUSY; 894e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 8954260f7c7SSage Weil return -EISDIR; 8964260f7c7SSage Weil if (IS_DEADDIR(dir)) 8974260f7c7SSage Weil return -ENOENT; 8984260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 8994260f7c7SSage Weil return -EBUSY; 9004260f7c7SSage Weil return 0; 9014260f7c7SSage Weil } 9024260f7c7SSage Weil 903cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 904cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 905cb8e7090SChristoph Hellwig { 9062b0143b5SDavid Howells if (d_really_is_positive(child)) 907cb8e7090SChristoph Hellwig return -EEXIST; 908cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 909cb8e7090SChristoph Hellwig return -ENOENT; 910cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 911cb8e7090SChristoph Hellwig } 912cb8e7090SChristoph Hellwig 913cb8e7090SChristoph Hellwig /* 914cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 915cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 916cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 917cb8e7090SChristoph Hellwig */ 91892872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 91952f75f4fSDavid Sterba const char *name, int namelen, 92072fd032eSSage Weil struct btrfs_root *snap_src, 9216f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9228696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 923cb8e7090SChristoph Hellwig { 9242b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9250b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 926cb8e7090SChristoph Hellwig struct dentry *dentry; 927cb8e7090SChristoph Hellwig int error; 928cb8e7090SChristoph Hellwig 92900235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 93000235411SAl Viro if (error == -EINTR) 93100235411SAl Viro return error; 932cb8e7090SChristoph Hellwig 933cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 934cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 935cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 936cb8e7090SChristoph Hellwig goto out_unlock; 937cb8e7090SChristoph Hellwig 93876dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 939cb8e7090SChristoph Hellwig if (error) 940a874a63eSLiu Bo goto out_dput; 941cb8e7090SChristoph Hellwig 9429c52057cSChris Mason /* 9439c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9449c52057cSChris Mason * check for them now when we can safely fail 9459c52057cSChris Mason */ 9469c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9479c52057cSChris Mason dir->i_ino, name, 9489c52057cSChris Mason namelen); 9499c52057cSChris Mason if (error) 9509c52057cSChris Mason goto out_dput; 9519c52057cSChris Mason 9520b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 95376dda93cSYan, Zheng 95476dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 95576dda93cSYan, Zheng goto out_up_read; 95676dda93cSYan, Zheng 9573de4586cSChris Mason if (snap_src) { 95861d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9596f72c7e2SArne Jansen async_transid, readonly, inherit); 9603de4586cSChris Mason } else { 961d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 962d5c12070SMiao Xie async_transid, inherit); 9633de4586cSChris Mason } 96476dda93cSYan, Zheng if (!error) 96576dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 96676dda93cSYan, Zheng out_up_read: 9670b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 968cb8e7090SChristoph Hellwig out_dput: 969cb8e7090SChristoph Hellwig dput(dentry); 970cb8e7090SChristoph Hellwig out_unlock: 9715955102cSAl Viro inode_unlock(dir); 972cb8e7090SChristoph Hellwig return error; 973cb8e7090SChristoph Hellwig } 974cb8e7090SChristoph Hellwig 9754cb5300bSChris Mason /* 9764cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 9774cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 9784cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 9794cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 9804cb5300bSChris Mason * part of the file 9814cb5300bSChris Mason */ 982aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 9834cb5300bSChris Mason { 9844cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9854cb5300bSChris Mason struct extent_map *em = NULL; 9864cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 9874cb5300bSChris Mason u64 end; 9884cb5300bSChris Mason 9894cb5300bSChris Mason read_lock(&em_tree->lock); 99009cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 9914cb5300bSChris Mason read_unlock(&em_tree->lock); 9924cb5300bSChris Mason 9934cb5300bSChris Mason if (em) { 9944cb5300bSChris Mason end = extent_map_end(em); 9954cb5300bSChris Mason free_extent_map(em); 9964cb5300bSChris Mason if (end - offset > thresh) 9974cb5300bSChris Mason return 0; 9984cb5300bSChris Mason } 9994cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10004cb5300bSChris Mason thresh /= 2; 10014cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10024cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10034cb5300bSChris Mason if (end >= thresh) 10044cb5300bSChris Mason return 0; 10054cb5300bSChris Mason return 1; 10064cb5300bSChris Mason } 10074cb5300bSChris Mason 10084cb5300bSChris Mason /* 10094cb5300bSChris Mason * helper function to walk through a file and find extents 10104cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10114cb5300bSChris Mason * 10124cb5300bSChris Mason * This is used by the defragging code to find new and small 10134cb5300bSChris Mason * extents 10144cb5300bSChris Mason */ 10154cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10164cb5300bSChris Mason struct inode *inode, u64 newer_than, 1017aab110abSDavid Sterba u64 *off, u32 thresh) 10184cb5300bSChris Mason { 10194cb5300bSChris Mason struct btrfs_path *path; 10204cb5300bSChris Mason struct btrfs_key min_key; 10214cb5300bSChris Mason struct extent_buffer *leaf; 10224cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10234cb5300bSChris Mason int type; 10244cb5300bSChris Mason int ret; 10254a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10264cb5300bSChris Mason 10274cb5300bSChris Mason path = btrfs_alloc_path(); 10284cb5300bSChris Mason if (!path) 10294cb5300bSChris Mason return -ENOMEM; 10304cb5300bSChris Mason 1031a4689d2bSDavid Sterba min_key.objectid = ino; 10324cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10334cb5300bSChris Mason min_key.offset = *off; 10344cb5300bSChris Mason 10354cb5300bSChris Mason while (1) { 10366174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10374cb5300bSChris Mason if (ret != 0) 10384cb5300bSChris Mason goto none; 1039f094c9bdSFilipe Manana process_slot: 1040a4689d2bSDavid Sterba if (min_key.objectid != ino) 10414cb5300bSChris Mason goto none; 10424cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10434cb5300bSChris Mason goto none; 10444cb5300bSChris Mason 10454cb5300bSChris Mason leaf = path->nodes[0]; 10464cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10474cb5300bSChris Mason struct btrfs_file_extent_item); 10484cb5300bSChris Mason 10494cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10504cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10514cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10524cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10534cb5300bSChris Mason *off = min_key.offset; 10544cb5300bSChris Mason btrfs_free_path(path); 10554cb5300bSChris Mason return 0; 10564cb5300bSChris Mason } 10574cb5300bSChris Mason 1058f094c9bdSFilipe Manana path->slots[0]++; 1059f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1060f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1061f094c9bdSFilipe Manana goto process_slot; 1062f094c9bdSFilipe Manana } 1063f094c9bdSFilipe Manana 10644cb5300bSChris Mason if (min_key.offset == (u64)-1) 10654cb5300bSChris Mason goto none; 10664cb5300bSChris Mason 10674cb5300bSChris Mason min_key.offset++; 10684cb5300bSChris Mason btrfs_release_path(path); 10694cb5300bSChris Mason } 10704cb5300bSChris Mason none: 10714cb5300bSChris Mason btrfs_free_path(path); 10724cb5300bSChris Mason return -ENOENT; 10734cb5300bSChris Mason } 10744cb5300bSChris Mason 10756c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 107617ce6ef8SLiu Bo { 107717ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1078940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10796c282eb4SLi Zefan struct extent_map *em; 108009cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1081940100a4SChris Mason 1082940100a4SChris Mason /* 1083940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1084940100a4SChris Mason * the full extent lock 1085940100a4SChris Mason */ 1086940100a4SChris Mason read_lock(&em_tree->lock); 1087940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1088940100a4SChris Mason read_unlock(&em_tree->lock); 1089940100a4SChris Mason 1090940100a4SChris Mason if (!em) { 1091308d9800SFilipe Manana struct extent_state *cached = NULL; 1092308d9800SFilipe Manana u64 end = start + len - 1; 1093308d9800SFilipe Manana 1094940100a4SChris Mason /* get the big lock and read metadata off disk */ 1095ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1096fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1097e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1098940100a4SChris Mason 10996cf8bfbfSDan Carpenter if (IS_ERR(em)) 11006c282eb4SLi Zefan return NULL; 1101940100a4SChris Mason } 1102940100a4SChris Mason 11036c282eb4SLi Zefan return em; 11046c282eb4SLi Zefan } 11056c282eb4SLi Zefan 11066c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11076c282eb4SLi Zefan { 11086c282eb4SLi Zefan struct extent_map *next; 11096c282eb4SLi Zefan bool ret = true; 11106c282eb4SLi Zefan 11116c282eb4SLi Zefan /* this is the last extent */ 11126c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11136c282eb4SLi Zefan return false; 11146c282eb4SLi Zefan 11156c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1116e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1117e9512d72SChris Mason ret = false; 1118e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1119ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11206c282eb4SLi Zefan ret = false; 11216c282eb4SLi Zefan 11226c282eb4SLi Zefan free_extent_map(next); 11236c282eb4SLi Zefan return ret; 11246c282eb4SLi Zefan } 11256c282eb4SLi Zefan 1126aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1127a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1128a43a2111SAndrew Mahone int compress) 11296c282eb4SLi Zefan { 11306c282eb4SLi Zefan struct extent_map *em; 11316c282eb4SLi Zefan int ret = 1; 11326c282eb4SLi Zefan bool next_mergeable = true; 11334a3560c4SLiu Bo bool prev_mergeable = true; 11346c282eb4SLi Zefan 11356c282eb4SLi Zefan /* 11366c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11376c282eb4SLi Zefan * defragging it 11386c282eb4SLi Zefan */ 11396c282eb4SLi Zefan if (start < *defrag_end) 11406c282eb4SLi Zefan return 1; 11416c282eb4SLi Zefan 11426c282eb4SLi Zefan *skip = 0; 11436c282eb4SLi Zefan 11446c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11456c282eb4SLi Zefan if (!em) 11466c282eb4SLi Zefan return 0; 11476c282eb4SLi Zefan 1148940100a4SChris Mason /* this will cover holes, and inline extents */ 114917ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1150940100a4SChris Mason ret = 0; 115117ce6ef8SLiu Bo goto out; 115217ce6ef8SLiu Bo } 115317ce6ef8SLiu Bo 11544a3560c4SLiu Bo if (!*defrag_end) 11554a3560c4SLiu Bo prev_mergeable = false; 11564a3560c4SLiu Bo 11576c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1158940100a4SChris Mason /* 11596c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11606c282eb4SLi Zefan * real extent, don't bother defragging it 1161940100a4SChris Mason */ 1162a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11634a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1164940100a4SChris Mason ret = 0; 116517ce6ef8SLiu Bo out: 1166940100a4SChris Mason /* 1167940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1168940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1169940100a4SChris Mason * so that the next tiny extent will force a defrag. 1170940100a4SChris Mason * 1171940100a4SChris Mason * The end result of this is that tiny extents before a single big 1172940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1173940100a4SChris Mason */ 1174940100a4SChris Mason if (ret) { 1175940100a4SChris Mason *defrag_end = extent_map_end(em); 1176940100a4SChris Mason } else { 1177940100a4SChris Mason *last_len = 0; 1178940100a4SChris Mason *skip = extent_map_end(em); 1179940100a4SChris Mason *defrag_end = 0; 1180940100a4SChris Mason } 1181940100a4SChris Mason 1182940100a4SChris Mason free_extent_map(em); 1183940100a4SChris Mason return ret; 1184940100a4SChris Mason } 1185940100a4SChris Mason 11864cb5300bSChris Mason /* 11874cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 11884cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 11894cb5300bSChris Mason * to COW and defrag. 11904cb5300bSChris Mason * 11914cb5300bSChris Mason * It also makes sure the delalloc code has enough 11924cb5300bSChris Mason * dirty data to avoid making new small extents as part 11934cb5300bSChris Mason * of the defrag 11944cb5300bSChris Mason * 11954cb5300bSChris Mason * It's a good idea to start RA on this range 11964cb5300bSChris Mason * before calling this. 11974cb5300bSChris Mason */ 11984cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 11994cb5300bSChris Mason struct page **pages, 12004cb5300bSChris Mason unsigned long start_index, 1201c41570c9SJustin Maggard unsigned long num_pages) 1202f46b5a66SChristoph Hellwig { 12034cb5300bSChris Mason unsigned long file_end; 12044cb5300bSChris Mason u64 isize = i_size_read(inode); 1205f46b5a66SChristoph Hellwig u64 page_start; 1206f46b5a66SChristoph Hellwig u64 page_end; 12071f12bd06SLiu Bo u64 page_cnt; 12084cb5300bSChris Mason int ret; 12094cb5300bSChris Mason int i; 12104cb5300bSChris Mason int i_done; 12114cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12124cb5300bSChris Mason struct extent_state *cached_state = NULL; 1213600a45e1SMiao Xie struct extent_io_tree *tree; 1214364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12153b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12164cb5300bSChris Mason 121709cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12181f12bd06SLiu Bo if (!isize || start_index > file_end) 12191f12bd06SLiu Bo return 0; 12201f12bd06SLiu Bo 12211f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12224cb5300bSChris Mason 1223364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 122409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 122509cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12264cb5300bSChris Mason if (ret) 12274cb5300bSChris Mason return ret; 12284cb5300bSChris Mason i_done = 0; 1229600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12304cb5300bSChris Mason 12314cb5300bSChris Mason /* step one, lock all the pages */ 12321f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12334cb5300bSChris Mason struct page *page; 1234600a45e1SMiao Xie again: 1235a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12363b16a4e3SJosef Bacik start_index + i, mask); 12374cb5300bSChris Mason if (!page) 12384cb5300bSChris Mason break; 12394cb5300bSChris Mason 1240600a45e1SMiao Xie page_start = page_offset(page); 124109cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1242600a45e1SMiao Xie while (1) { 1243308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1244ff13db41SDavid Sterba &cached_state); 1245600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1246600a45e1SMiao Xie page_start); 1247308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1248e43bbe5eSDavid Sterba &cached_state); 1249600a45e1SMiao Xie if (!ordered) 1250600a45e1SMiao Xie break; 1251600a45e1SMiao Xie 1252600a45e1SMiao Xie unlock_page(page); 1253600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1254600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1255600a45e1SMiao Xie lock_page(page); 12561f12bd06SLiu Bo /* 12571f12bd06SLiu Bo * we unlocked the page above, so we need check if 12581f12bd06SLiu Bo * it was released or not. 12591f12bd06SLiu Bo */ 12601f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12611f12bd06SLiu Bo unlock_page(page); 126209cbfeafSKirill A. Shutemov put_page(page); 12631f12bd06SLiu Bo goto again; 12641f12bd06SLiu Bo } 1265600a45e1SMiao Xie } 1266600a45e1SMiao Xie 12674cb5300bSChris Mason if (!PageUptodate(page)) { 12684cb5300bSChris Mason btrfs_readpage(NULL, page); 12694cb5300bSChris Mason lock_page(page); 12704cb5300bSChris Mason if (!PageUptodate(page)) { 12714cb5300bSChris Mason unlock_page(page); 127209cbfeafSKirill A. Shutemov put_page(page); 12734cb5300bSChris Mason ret = -EIO; 12744cb5300bSChris Mason break; 12754cb5300bSChris Mason } 12764cb5300bSChris Mason } 1277600a45e1SMiao Xie 1278600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1279600a45e1SMiao Xie unlock_page(page); 128009cbfeafSKirill A. Shutemov put_page(page); 1281600a45e1SMiao Xie goto again; 1282600a45e1SMiao Xie } 1283600a45e1SMiao Xie 12844cb5300bSChris Mason pages[i] = page; 12854cb5300bSChris Mason i_done++; 12864cb5300bSChris Mason } 12874cb5300bSChris Mason if (!i_done || ret) 12884cb5300bSChris Mason goto out; 12894cb5300bSChris Mason 12901751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 12914cb5300bSChris Mason goto out; 12924cb5300bSChris Mason 12934cb5300bSChris Mason /* 12944cb5300bSChris Mason * so now we have a nice long stream of locked 12954cb5300bSChris Mason * and up to date pages, lets wait on them 12964cb5300bSChris Mason */ 12974cb5300bSChris Mason for (i = 0; i < i_done; i++) 12984cb5300bSChris Mason wait_on_page_writeback(pages[i]); 12994cb5300bSChris Mason 13004cb5300bSChris Mason page_start = page_offset(pages[0]); 130109cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13024cb5300bSChris Mason 13034cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1304ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13054cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 13064cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 13079e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1308ae0f1625SDavid Sterba &cached_state); 13094cb5300bSChris Mason 13101f12bd06SLiu Bo if (i_done != page_cnt) { 13119e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 13129e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 13139e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1314bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 131509cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 131643b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13174cb5300bSChris Mason } 13184cb5300bSChris Mason 13194cb5300bSChris Mason 13209e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1321018ed4f7SDavid Sterba &cached_state); 13224cb5300bSChris Mason 13234cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1324e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13254cb5300bSChris Mason 13264cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13274cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13284cb5300bSChris Mason ClearPageChecked(pages[i]); 13294cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13304cb5300bSChris Mason set_page_dirty(pages[i]); 13314cb5300bSChris Mason unlock_page(pages[i]); 133209cbfeafSKirill A. Shutemov put_page(pages[i]); 13334cb5300bSChris Mason } 133443b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 133543b18595SQu Wenruo false); 1336364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13374cb5300bSChris Mason return i_done; 13384cb5300bSChris Mason out: 13394cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13404cb5300bSChris Mason unlock_page(pages[i]); 134109cbfeafSKirill A. Shutemov put_page(pages[i]); 13424cb5300bSChris Mason } 1343bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 134409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 134543b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 134643b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 134743b18595SQu Wenruo true); 1348364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13494cb5300bSChris Mason return ret; 13504cb5300bSChris Mason 13514cb5300bSChris Mason } 13524cb5300bSChris Mason 13534cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13544cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13554cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13564cb5300bSChris Mason { 13570b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13584cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13594cb5300bSChris Mason struct file_ra_state *ra = NULL; 13604cb5300bSChris Mason unsigned long last_index; 1361151a31b2SLi Zefan u64 isize = i_size_read(inode); 1362940100a4SChris Mason u64 last_len = 0; 1363940100a4SChris Mason u64 skip = 0; 1364940100a4SChris Mason u64 defrag_end = 0; 13654cb5300bSChris Mason u64 newer_off = range->start; 1366f46b5a66SChristoph Hellwig unsigned long i; 1367008873eaSLi Zefan unsigned long ra_index = 0; 1368f46b5a66SChristoph Hellwig int ret; 13694cb5300bSChris Mason int defrag_count = 0; 13701a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1371aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 137209cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1373c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1374ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 13754cb5300bSChris Mason struct page **pages = NULL; 13761e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 13774cb5300bSChris Mason 13780abd5b17SLiu Bo if (isize == 0) 13790abd5b17SLiu Bo return 0; 13800abd5b17SLiu Bo 13810abd5b17SLiu Bo if (range->start >= isize) 13820abd5b17SLiu Bo return -EINVAL; 13831a419d85SLi Zefan 13841e2ef46dSDavid Sterba if (do_compress) { 13851a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 13861a419d85SLi Zefan return -EINVAL; 13871a419d85SLi Zefan if (range->compress_type) 13881a419d85SLi Zefan compress_type = range->compress_type; 13891a419d85SLi Zefan } 1390f46b5a66SChristoph Hellwig 13910abd5b17SLiu Bo if (extent_thresh == 0) 1392ee22184bSByongho Lee extent_thresh = SZ_256K; 1393f46b5a66SChristoph Hellwig 13944cb5300bSChris Mason /* 13950a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 13960a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 13970a52d108SDavid Sterba * we don't error out. 13984cb5300bSChris Mason */ 13994cb5300bSChris Mason if (!file) { 140063e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14010a52d108SDavid Sterba if (ra) 14024cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14034cb5300bSChris Mason } else { 14044cb5300bSChris Mason ra = &file->f_ra; 14054cb5300bSChris Mason } 14064cb5300bSChris Mason 140763e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14084cb5300bSChris Mason if (!pages) { 14094cb5300bSChris Mason ret = -ENOMEM; 14104cb5300bSChris Mason goto out_ra; 14114cb5300bSChris Mason } 14124cb5300bSChris Mason 14134cb5300bSChris Mason /* find the last page to defrag */ 14141e701a32SChris Mason if (range->start + range->len > range->start) { 1415151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 141609cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14171e701a32SChris Mason } else { 141809cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14191e701a32SChris Mason } 14201e701a32SChris Mason 14214cb5300bSChris Mason if (newer_than) { 14224cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1423ee22184bSByongho Lee &newer_off, SZ_64K); 14244cb5300bSChris Mason if (!ret) { 14254cb5300bSChris Mason range->start = newer_off; 14264cb5300bSChris Mason /* 14274cb5300bSChris Mason * we always align our defrag to help keep 14284cb5300bSChris Mason * the extents in the file evenly spaced 14294cb5300bSChris Mason */ 143009cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14314cb5300bSChris Mason } else 14324cb5300bSChris Mason goto out_ra; 14334cb5300bSChris Mason } else { 143409cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14354cb5300bSChris Mason } 14364cb5300bSChris Mason if (!max_to_defrag) 1437070034bdSchandan max_to_defrag = last_index - i + 1; 14384cb5300bSChris Mason 14392a0f7f57SLi Zefan /* 14402a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14412a0f7f57SLi Zefan * written sequentially. 14422a0f7f57SLi Zefan */ 14432a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14442a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14452a0f7f57SLi Zefan 1446f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 144709cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14484cb5300bSChris Mason /* 14494cb5300bSChris Mason * make sure we stop running if someone unmounts 14504cb5300bSChris Mason * the FS 14514cb5300bSChris Mason */ 14521751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14534cb5300bSChris Mason break; 14544cb5300bSChris Mason 14550b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14560b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1457210549ebSDavid Sterba ret = -EAGAIN; 1458210549ebSDavid Sterba break; 1459210549ebSDavid Sterba } 1460210549ebSDavid Sterba 146109cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14626c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14631e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1464940100a4SChris Mason unsigned long next; 1465940100a4SChris Mason /* 1466940100a4SChris Mason * the should_defrag function tells us how much to skip 1467940100a4SChris Mason * bump our counter by the suggested amount 1468940100a4SChris Mason */ 146909cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1470940100a4SChris Mason i = max(i + 1, next); 1471940100a4SChris Mason continue; 1472940100a4SChris Mason } 1473008873eaSLi Zefan 1474008873eaSLi Zefan if (!newer_than) { 147509cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 147609cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1477008873eaSLi Zefan cluster = min(cluster, max_cluster); 1478008873eaSLi Zefan } else { 1479008873eaSLi Zefan cluster = max_cluster; 1480008873eaSLi Zefan } 1481008873eaSLi Zefan 1482008873eaSLi Zefan if (i + cluster > ra_index) { 1483008873eaSLi Zefan ra_index = max(i, ra_index); 14840a52d108SDavid Sterba if (ra) 1485d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1486d3c0bab5SDavid Sterba file, ra_index, cluster); 1487e4826a5bSchandan ra_index += cluster; 1488008873eaSLi Zefan } 14894cb5300bSChris Mason 14905955102cSAl Viro inode_lock(inode); 14911e2ef46dSDavid Sterba if (do_compress) 1492eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1493008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1494ecb8bea8SLiu Bo if (ret < 0) { 14955955102cSAl Viro inode_unlock(inode); 14964cb5300bSChris Mason goto out_ra; 1497ecb8bea8SLiu Bo } 14984cb5300bSChris Mason 14994cb5300bSChris Mason defrag_count += ret; 1500d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15015955102cSAl Viro inode_unlock(inode); 15024cb5300bSChris Mason 15034cb5300bSChris Mason if (newer_than) { 15044cb5300bSChris Mason if (newer_off == (u64)-1) 15054cb5300bSChris Mason break; 15064cb5300bSChris Mason 1507e1f041e1SLiu Bo if (ret > 0) 1508e1f041e1SLiu Bo i += ret; 1509e1f041e1SLiu Bo 15104cb5300bSChris Mason newer_off = max(newer_off + 1, 151109cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15124cb5300bSChris Mason 1513ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1514ee22184bSByongho Lee &newer_off, SZ_64K); 15154cb5300bSChris Mason if (!ret) { 15164cb5300bSChris Mason range->start = newer_off; 151709cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15184cb5300bSChris Mason } else { 15194cb5300bSChris Mason break; 1520940100a4SChris Mason } 15214cb5300bSChris Mason } else { 1522008873eaSLi Zefan if (ret > 0) { 1523cbcc8326SLi Zefan i += ret; 152409cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1525008873eaSLi Zefan } else { 1526940100a4SChris Mason i++; 1527008873eaSLi Zefan last_len = 0; 1528008873eaSLi Zefan } 1529f46b5a66SChristoph Hellwig } 15304cb5300bSChris Mason } 1531f46b5a66SChristoph Hellwig 1532dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15331e701a32SChris Mason filemap_flush(inode->i_mapping); 1534dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1535dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1536dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1537dec8ef90SFilipe Manana } 15381e701a32SChris Mason 15391a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15400b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15415c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15425c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15431a419d85SLi Zefan } 15441a419d85SLi Zefan 154560ccf82fSDiego Calleja ret = defrag_count; 1546940100a4SChris Mason 15474cb5300bSChris Mason out_ra: 15481e2ef46dSDavid Sterba if (do_compress) { 15495955102cSAl Viro inode_lock(inode); 1550eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15515955102cSAl Viro inode_unlock(inode); 1552633085c7SFilipe David Borba Manana } 15534cb5300bSChris Mason if (!file) 15544cb5300bSChris Mason kfree(ra); 15554cb5300bSChris Mason kfree(pages); 1556940100a4SChris Mason return ret; 1557f46b5a66SChristoph Hellwig } 1558f46b5a66SChristoph Hellwig 1559198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 156076dda93cSYan, Zheng void __user *arg) 1561f46b5a66SChristoph Hellwig { 15620b246afaSJeff Mahoney struct inode *inode = file_inode(file); 15630b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1564f46b5a66SChristoph Hellwig u64 new_size; 1565f46b5a66SChristoph Hellwig u64 old_size; 1566f46b5a66SChristoph Hellwig u64 devid = 1; 15670b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1568f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1569f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1570f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1571f46b5a66SChristoph Hellwig char *sizestr; 15729a40f122SGui Hecheng char *retptr; 1573f46b5a66SChristoph Hellwig char *devstr = NULL; 1574f46b5a66SChristoph Hellwig int ret = 0; 1575f46b5a66SChristoph Hellwig int mod = 0; 1576f46b5a66SChristoph Hellwig 1577e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1578e441d54dSChris Mason return -EPERM; 1579e441d54dSChris Mason 1580198605a8SMiao Xie ret = mnt_want_write_file(file); 1581198605a8SMiao Xie if (ret) 1582198605a8SMiao Xie return ret; 1583198605a8SMiao Xie 1584171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 158597547676SMiao Xie mnt_drop_write_file(file); 1586e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1587c9e9f97bSIlya Dryomov } 1588c9e9f97bSIlya Dryomov 1589dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1590c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1591c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1592c9e9f97bSIlya Dryomov goto out; 1593c9e9f97bSIlya Dryomov } 15945516e595SMark Fasheh 15955516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1596f46b5a66SChristoph Hellwig 1597f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1598f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1599f46b5a66SChristoph Hellwig if (devstr) { 1600f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1601f46b5a66SChristoph Hellwig *devstr = '\0'; 1602f46b5a66SChristoph Hellwig devstr = vol_args->name; 160358dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 160458dfae63SZhangZhen if (ret) 160558dfae63SZhangZhen goto out_free; 1606dfd79829SMiao Xie if (!devid) { 1607dfd79829SMiao Xie ret = -EINVAL; 1608dfd79829SMiao Xie goto out_free; 1609dfd79829SMiao Xie } 16100b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1611f46b5a66SChristoph Hellwig } 1612dba60f3fSMiao Xie 16130b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1614f46b5a66SChristoph Hellwig if (!device) { 16150b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1616c1c9ff7cSGeert Uytterhoeven devid); 1617dfd79829SMiao Xie ret = -ENODEV; 1618c9e9f97bSIlya Dryomov goto out_free; 1619f46b5a66SChristoph Hellwig } 1620dba60f3fSMiao Xie 1621ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16220b246afaSJeff Mahoney btrfs_info(fs_info, 1623efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1624c1c9ff7cSGeert Uytterhoeven devid); 1625dfd79829SMiao Xie ret = -EPERM; 16264e42ae1bSLiu Bo goto out_free; 16274e42ae1bSLiu Bo } 16284e42ae1bSLiu Bo 1629f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1630f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1631f46b5a66SChristoph Hellwig else { 1632f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1633f46b5a66SChristoph Hellwig mod = -1; 1634f46b5a66SChristoph Hellwig sizestr++; 1635f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1636f46b5a66SChristoph Hellwig mod = 1; 1637f46b5a66SChristoph Hellwig sizestr++; 1638f46b5a66SChristoph Hellwig } 16399a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16409a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1641f46b5a66SChristoph Hellwig ret = -EINVAL; 1642c9e9f97bSIlya Dryomov goto out_free; 1643f46b5a66SChristoph Hellwig } 1644f46b5a66SChristoph Hellwig } 1645f46b5a66SChristoph Hellwig 1646401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1647dfd79829SMiao Xie ret = -EPERM; 164863a212abSStefan Behrens goto out_free; 164963a212abSStefan Behrens } 165063a212abSStefan Behrens 16517cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1652f46b5a66SChristoph Hellwig 1653f46b5a66SChristoph Hellwig if (mod < 0) { 1654f46b5a66SChristoph Hellwig if (new_size > old_size) { 1655f46b5a66SChristoph Hellwig ret = -EINVAL; 1656c9e9f97bSIlya Dryomov goto out_free; 1657f46b5a66SChristoph Hellwig } 1658f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1659f46b5a66SChristoph Hellwig } else if (mod > 0) { 1660eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1661902c68a4SGui Hecheng ret = -ERANGE; 1662eb8052e0SWenliang Fan goto out_free; 1663eb8052e0SWenliang Fan } 1664f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1665f46b5a66SChristoph Hellwig } 1666f46b5a66SChristoph Hellwig 1667ee22184bSByongho Lee if (new_size < SZ_256M) { 1668f46b5a66SChristoph Hellwig ret = -EINVAL; 1669c9e9f97bSIlya Dryomov goto out_free; 1670f46b5a66SChristoph Hellwig } 1671f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1672f46b5a66SChristoph Hellwig ret = -EFBIG; 1673c9e9f97bSIlya Dryomov goto out_free; 1674f46b5a66SChristoph Hellwig } 1675f46b5a66SChristoph Hellwig 167647f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1677f46b5a66SChristoph Hellwig 16780b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1679c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1680f46b5a66SChristoph Hellwig 1681f46b5a66SChristoph Hellwig if (new_size > old_size) { 1682a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 168398d5dc13STsutomu Itoh if (IS_ERR(trans)) { 168498d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1685c9e9f97bSIlya Dryomov goto out_free; 168698d5dc13STsutomu Itoh } 1687f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 16883a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1689ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1690f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 16910253f40eSjeff.liu } /* equal, nothing need to do */ 1692f46b5a66SChristoph Hellwig 1693c9e9f97bSIlya Dryomov out_free: 1694f46b5a66SChristoph Hellwig kfree(vol_args); 1695c9e9f97bSIlya Dryomov out: 1696171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 169718f39c41SIlya Dryomov mnt_drop_write_file(file); 1698f46b5a66SChristoph Hellwig return ret; 1699f46b5a66SChristoph Hellwig } 1700f46b5a66SChristoph Hellwig 170172fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 170252f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17036f72c7e2SArne Jansen u64 *transid, bool readonly, 17048696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1705f46b5a66SChristoph Hellwig { 1706f46b5a66SChristoph Hellwig int namelen; 17073de4586cSChris Mason int ret = 0; 1708f46b5a66SChristoph Hellwig 1709325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1710325c50e3SJeff Mahoney return -ENOTDIR; 1711325c50e3SJeff Mahoney 1712a874a63eSLiu Bo ret = mnt_want_write_file(file); 1713a874a63eSLiu Bo if (ret) 1714a874a63eSLiu Bo goto out; 1715a874a63eSLiu Bo 171672fd032eSSage Weil namelen = strlen(name); 171772fd032eSSage Weil if (strchr(name, '/')) { 1718f46b5a66SChristoph Hellwig ret = -EINVAL; 1719a874a63eSLiu Bo goto out_drop_write; 1720f46b5a66SChristoph Hellwig } 1721f46b5a66SChristoph Hellwig 172216780cabSChris Mason if (name[0] == '.' && 172316780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 172416780cabSChris Mason ret = -EEXIST; 1725a874a63eSLiu Bo goto out_drop_write; 172616780cabSChris Mason } 172716780cabSChris Mason 17283de4586cSChris Mason if (subvol) { 172972fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17306f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1731cb8e7090SChristoph Hellwig } else { 17322903ff01SAl Viro struct fd src = fdget(fd); 17333de4586cSChris Mason struct inode *src_inode; 17342903ff01SAl Viro if (!src.file) { 17353de4586cSChris Mason ret = -EINVAL; 1736a874a63eSLiu Bo goto out_drop_write; 17373de4586cSChris Mason } 17383de4586cSChris Mason 1739496ad9aaSAl Viro src_inode = file_inode(src.file); 1740496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1741c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1742efe120a0SFrank Holton "Snapshot src from another FS"); 174323ad5b17SKusanagi Kouichi ret = -EXDEV; 1744d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1745d0242061SDavid Sterba /* 1746d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1747d0242061SDavid Sterba * are limited to own subvolumes only 1748d0242061SDavid Sterba */ 1749d0242061SDavid Sterba ret = -EPERM; 1750ecd18815SAl Viro } else { 175172fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 175272fd032eSSage Weil BTRFS_I(src_inode)->root, 17536f72c7e2SArne Jansen transid, readonly, inherit); 1754ecd18815SAl Viro } 17552903ff01SAl Viro fdput(src); 1756cb8e7090SChristoph Hellwig } 1757a874a63eSLiu Bo out_drop_write: 1758a874a63eSLiu Bo mnt_drop_write_file(file); 1759f46b5a66SChristoph Hellwig out: 176072fd032eSSage Weil return ret; 176172fd032eSSage Weil } 176272fd032eSSage Weil 176372fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1764fa0d2b9bSLi Zefan void __user *arg, int subvol) 176572fd032eSSage Weil { 1766fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 176772fd032eSSage Weil int ret; 176872fd032eSSage Weil 1769325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1770325c50e3SJeff Mahoney return -ENOTDIR; 1771325c50e3SJeff Mahoney 1772fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1773fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1774fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1775fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1776fa0d2b9bSLi Zefan 1777fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1778b83cc969SLi Zefan vol_args->fd, subvol, 17796f72c7e2SArne Jansen NULL, false, NULL); 1780fa0d2b9bSLi Zefan 1781fa0d2b9bSLi Zefan kfree(vol_args); 1782fa0d2b9bSLi Zefan return ret; 1783fa0d2b9bSLi Zefan } 1784fa0d2b9bSLi Zefan 1785fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1786fa0d2b9bSLi Zefan void __user *arg, int subvol) 1787fa0d2b9bSLi Zefan { 1788fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1789fa0d2b9bSLi Zefan int ret; 1790fdfb1e4fSLi Zefan u64 transid = 0; 1791fdfb1e4fSLi Zefan u64 *ptr = NULL; 1792b83cc969SLi Zefan bool readonly = false; 17936f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 179472fd032eSSage Weil 1795325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1796325c50e3SJeff Mahoney return -ENOTDIR; 1797325c50e3SJeff Mahoney 1798fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1799fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1800fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1801fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1802fdfb1e4fSLi Zefan 1803b83cc969SLi Zefan if (vol_args->flags & 18046f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18056f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1806b83cc969SLi Zefan ret = -EOPNOTSUPP; 1807c47ca32dSDan Carpenter goto free_args; 1808fdfb1e4fSLi Zefan } 1809fdfb1e4fSLi Zefan 1810fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1811fdfb1e4fSLi Zefan ptr = &transid; 1812b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1813b83cc969SLi Zefan readonly = true; 18146f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 181509cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18166f72c7e2SArne Jansen ret = -EINVAL; 1817c47ca32dSDan Carpenter goto free_args; 18186f72c7e2SArne Jansen } 18196f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18206f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18216f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1822c47ca32dSDan Carpenter goto free_args; 18236f72c7e2SArne Jansen } 18246f72c7e2SArne Jansen } 182575eaa0e2SSage Weil 1826fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18276f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18288696c533SMiao Xie readonly, inherit); 1829c47ca32dSDan Carpenter if (ret) 1830c47ca32dSDan Carpenter goto free_inherit; 183175eaa0e2SSage Weil 1832c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1833fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1834c47ca32dSDan Carpenter transid), 1835c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 183675eaa0e2SSage Weil ret = -EFAULT; 1837c47ca32dSDan Carpenter 1838c47ca32dSDan Carpenter free_inherit: 18396f72c7e2SArne Jansen kfree(inherit); 1840c47ca32dSDan Carpenter free_args: 1841c47ca32dSDan Carpenter kfree(vol_args); 1842f46b5a66SChristoph Hellwig return ret; 1843f46b5a66SChristoph Hellwig } 1844f46b5a66SChristoph Hellwig 18450caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18460caa102dSLi Zefan void __user *arg) 18470caa102dSLi Zefan { 1848496ad9aaSAl Viro struct inode *inode = file_inode(file); 18490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18500caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18510caa102dSLi Zefan int ret = 0; 18520caa102dSLi Zefan u64 flags = 0; 18530caa102dSLi Zefan 18544a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18550caa102dSLi Zefan return -EINVAL; 18560caa102dSLi Zefan 18570b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 18580caa102dSLi Zefan if (btrfs_root_readonly(root)) 18590caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 18600b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 18610caa102dSLi Zefan 18620caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 18630caa102dSLi Zefan ret = -EFAULT; 18640caa102dSLi Zefan 18650caa102dSLi Zefan return ret; 18660caa102dSLi Zefan } 18670caa102dSLi Zefan 18680caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 18690caa102dSLi Zefan void __user *arg) 18700caa102dSLi Zefan { 1871496ad9aaSAl Viro struct inode *inode = file_inode(file); 18720b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18730caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18740caa102dSLi Zefan struct btrfs_trans_handle *trans; 18750caa102dSLi Zefan u64 root_flags; 18760caa102dSLi Zefan u64 flags; 18770caa102dSLi Zefan int ret = 0; 18780caa102dSLi Zefan 1879bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1880bd60ea0fSDavid Sterba return -EPERM; 1881bd60ea0fSDavid Sterba 1882b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1883b9ca0664SLiu Bo if (ret) 1884b9ca0664SLiu Bo goto out; 18850caa102dSLi Zefan 18864a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1887b9ca0664SLiu Bo ret = -EINVAL; 1888b9ca0664SLiu Bo goto out_drop_write; 1889b9ca0664SLiu Bo } 18900caa102dSLi Zefan 1891b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1892b9ca0664SLiu Bo ret = -EFAULT; 1893b9ca0664SLiu Bo goto out_drop_write; 1894b9ca0664SLiu Bo } 18950caa102dSLi Zefan 1896b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1897b9ca0664SLiu Bo ret = -EINVAL; 1898b9ca0664SLiu Bo goto out_drop_write; 1899b9ca0664SLiu Bo } 19000caa102dSLi Zefan 1901b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1902b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1903b9ca0664SLiu Bo goto out_drop_write; 1904b9ca0664SLiu Bo } 19050caa102dSLi Zefan 19060b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19070caa102dSLi Zefan 19080caa102dSLi Zefan /* nothing to do */ 19090caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1910b9ca0664SLiu Bo goto out_drop_sem; 19110caa102dSLi Zefan 19120caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19132c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19140caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19150caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19162c686537SDavid Sterba } else { 19172c686537SDavid Sterba /* 19182c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19192c686537SDavid Sterba * send 19202c686537SDavid Sterba */ 19212c686537SDavid Sterba spin_lock(&root->root_item_lock); 19222c686537SDavid Sterba if (root->send_in_progress == 0) { 19230caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19240caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19252c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19262c686537SDavid Sterba } else { 19272c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19280b246afaSJeff Mahoney btrfs_warn(fs_info, 19292c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19302c686537SDavid Sterba root->root_key.objectid); 19312c686537SDavid Sterba ret = -EPERM; 19322c686537SDavid Sterba goto out_drop_sem; 19332c686537SDavid Sterba } 19342c686537SDavid Sterba } 19350caa102dSLi Zefan 19360caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19370caa102dSLi Zefan if (IS_ERR(trans)) { 19380caa102dSLi Zefan ret = PTR_ERR(trans); 19390caa102dSLi Zefan goto out_reset; 19400caa102dSLi Zefan } 19410caa102dSLi Zefan 19420b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19430caa102dSLi Zefan &root->root_key, &root->root_item); 19449417ebc8SNikolay Borisov if (ret < 0) { 19459417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19469417ebc8SNikolay Borisov goto out_reset; 19479417ebc8SNikolay Borisov } 19480caa102dSLi Zefan 19499417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19509417ebc8SNikolay Borisov 19510caa102dSLi Zefan out_reset: 19520caa102dSLi Zefan if (ret) 19530caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1954b9ca0664SLiu Bo out_drop_sem: 19550b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1956b9ca0664SLiu Bo out_drop_write: 1957b9ca0664SLiu Bo mnt_drop_write_file(file); 1958b9ca0664SLiu Bo out: 19590caa102dSLi Zefan return ret; 19600caa102dSLi Zefan } 19610caa102dSLi Zefan 1962ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1963ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1964ac8e9819SChris Mason { 1965abc6e134SChris Mason struct btrfs_key test; 1966abc6e134SChris Mason int ret; 1967abc6e134SChris Mason 1968abc6e134SChris Mason test.objectid = sk->min_objectid; 1969abc6e134SChris Mason test.type = sk->min_type; 1970abc6e134SChris Mason test.offset = sk->min_offset; 1971abc6e134SChris Mason 1972abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1973abc6e134SChris Mason if (ret < 0) 1974ac8e9819SChris Mason return 0; 1975abc6e134SChris Mason 1976abc6e134SChris Mason test.objectid = sk->max_objectid; 1977abc6e134SChris Mason test.type = sk->max_type; 1978abc6e134SChris Mason test.offset = sk->max_offset; 1979abc6e134SChris Mason 1980abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1981abc6e134SChris Mason if (ret > 0) 1982ac8e9819SChris Mason return 0; 1983ac8e9819SChris Mason return 1; 1984ac8e9819SChris Mason } 1985ac8e9819SChris Mason 1986df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 1987ac8e9819SChris Mason struct btrfs_key *key, 1988ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19899b6e817dSGerhard Heift size_t *buf_size, 1990ba346b35SGerhard Heift char __user *ubuf, 1991ac8e9819SChris Mason unsigned long *sk_offset, 1992ac8e9819SChris Mason int *num_found) 1993ac8e9819SChris Mason { 1994ac8e9819SChris Mason u64 found_transid; 1995ac8e9819SChris Mason struct extent_buffer *leaf; 1996ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1997dd81d459SNaohiro Aota struct btrfs_key test; 1998ac8e9819SChris Mason unsigned long item_off; 1999ac8e9819SChris Mason unsigned long item_len; 2000ac8e9819SChris Mason int nritems; 2001ac8e9819SChris Mason int i; 2002ac8e9819SChris Mason int slot; 2003ac8e9819SChris Mason int ret = 0; 2004ac8e9819SChris Mason 2005ac8e9819SChris Mason leaf = path->nodes[0]; 2006ac8e9819SChris Mason slot = path->slots[0]; 2007ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2008ac8e9819SChris Mason 2009ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2010ac8e9819SChris Mason i = nritems; 2011ac8e9819SChris Mason goto advance_key; 2012ac8e9819SChris Mason } 2013ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2014ac8e9819SChris Mason 2015ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2016ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2017ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2018ac8e9819SChris Mason 201903b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 202003b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 202103b71c6cSGabriel de Perthuis continue; 202203b71c6cSGabriel de Perthuis 20239b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20248f5f6178SGerhard Heift if (*num_found) { 2025ac8e9819SChris Mason ret = 1; 20268f5f6178SGerhard Heift goto out; 20278f5f6178SGerhard Heift } 20288f5f6178SGerhard Heift 20298f5f6178SGerhard Heift /* 20308f5f6178SGerhard Heift * return one empty item back for v1, which does not 20318f5f6178SGerhard Heift * handle -EOVERFLOW 20328f5f6178SGerhard Heift */ 20338f5f6178SGerhard Heift 20349b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2035ac8e9819SChris Mason item_len = 0; 20368f5f6178SGerhard Heift ret = -EOVERFLOW; 20378f5f6178SGerhard Heift } 2038ac8e9819SChris Mason 20399b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2040ac8e9819SChris Mason ret = 1; 204125c9bc2eSGerhard Heift goto out; 2042ac8e9819SChris Mason } 2043ac8e9819SChris Mason 2044ac8e9819SChris Mason sh.objectid = key->objectid; 2045ac8e9819SChris Mason sh.offset = key->offset; 2046ac8e9819SChris Mason sh.type = key->type; 2047ac8e9819SChris Mason sh.len = item_len; 2048ac8e9819SChris Mason sh.transid = found_transid; 2049ac8e9819SChris Mason 2050ac8e9819SChris Mason /* copy search result header */ 2051ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2052ba346b35SGerhard Heift ret = -EFAULT; 2053ba346b35SGerhard Heift goto out; 2054ba346b35SGerhard Heift } 2055ba346b35SGerhard Heift 2056ac8e9819SChris Mason *sk_offset += sizeof(sh); 2057ac8e9819SChris Mason 2058ac8e9819SChris Mason if (item_len) { 2059ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2060ac8e9819SChris Mason /* copy the item */ 2061ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2062ba346b35SGerhard Heift item_off, item_len)) { 2063ba346b35SGerhard Heift ret = -EFAULT; 2064ba346b35SGerhard Heift goto out; 2065ba346b35SGerhard Heift } 2066ba346b35SGerhard Heift 2067ac8e9819SChris Mason *sk_offset += item_len; 2068ac8e9819SChris Mason } 2069e2156867SHugo Mills (*num_found)++; 2070ac8e9819SChris Mason 20718f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20728f5f6178SGerhard Heift goto out; 20738f5f6178SGerhard Heift 207425c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 207525c9bc2eSGerhard Heift ret = 1; 207625c9bc2eSGerhard Heift goto out; 207725c9bc2eSGerhard Heift } 2078ac8e9819SChris Mason } 2079ac8e9819SChris Mason advance_key: 2080ac8e9819SChris Mason ret = 0; 2081dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2082dd81d459SNaohiro Aota test.type = sk->max_type; 2083dd81d459SNaohiro Aota test.offset = sk->max_offset; 2084dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2085dd81d459SNaohiro Aota ret = 1; 2086dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2087abc6e134SChris Mason key->offset++; 2088dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2089abc6e134SChris Mason key->offset = 0; 2090abc6e134SChris Mason key->type++; 2091dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2092abc6e134SChris Mason key->offset = 0; 2093abc6e134SChris Mason key->type = 0; 2094abc6e134SChris Mason key->objectid++; 2095abc6e134SChris Mason } else 2096abc6e134SChris Mason ret = 1; 209725c9bc2eSGerhard Heift out: 2098ba346b35SGerhard Heift /* 2099ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2100ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2101ba346b35SGerhard Heift * * all items were found 2102ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2103ba346b35SGerhard Heift * leaf 2104ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2105ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2106ba346b35SGerhard Heift */ 2107ac8e9819SChris Mason return ret; 2108ac8e9819SChris Mason } 2109ac8e9819SChris Mason 2110ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 211112544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21129b6e817dSGerhard Heift size_t *buf_size, 2113ba346b35SGerhard Heift char __user *ubuf) 2114ac8e9819SChris Mason { 21150b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2116ac8e9819SChris Mason struct btrfs_root *root; 2117ac8e9819SChris Mason struct btrfs_key key; 2118ac8e9819SChris Mason struct btrfs_path *path; 2119ac8e9819SChris Mason int ret; 2120ac8e9819SChris Mason int num_found = 0; 2121ac8e9819SChris Mason unsigned long sk_offset = 0; 2122ac8e9819SChris Mason 21239b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21249b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 212512544442SGerhard Heift return -EOVERFLOW; 21269b6e817dSGerhard Heift } 212712544442SGerhard Heift 2128ac8e9819SChris Mason path = btrfs_alloc_path(); 2129ac8e9819SChris Mason if (!path) 2130ac8e9819SChris Mason return -ENOMEM; 2131ac8e9819SChris Mason 2132ac8e9819SChris Mason if (sk->tree_id == 0) { 2133ac8e9819SChris Mason /* search the root of the inode that was passed */ 2134ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2135ac8e9819SChris Mason } else { 2136ac8e9819SChris Mason key.objectid = sk->tree_id; 2137ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2138ac8e9819SChris Mason key.offset = (u64)-1; 2139ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2140ac8e9819SChris Mason if (IS_ERR(root)) { 2141ac8e9819SChris Mason btrfs_free_path(path); 2142ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2143ac8e9819SChris Mason } 2144ac8e9819SChris Mason } 2145ac8e9819SChris Mason 2146ac8e9819SChris Mason key.objectid = sk->min_objectid; 2147ac8e9819SChris Mason key.type = sk->min_type; 2148ac8e9819SChris Mason key.offset = sk->min_offset; 2149ac8e9819SChris Mason 2150ac8e9819SChris Mason while (1) { 21516174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2152ac8e9819SChris Mason if (ret != 0) { 2153ac8e9819SChris Mason if (ret > 0) 2154ac8e9819SChris Mason ret = 0; 2155ac8e9819SChris Mason goto err; 2156ac8e9819SChris Mason } 2157df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2158ac8e9819SChris Mason &sk_offset, &num_found); 2159b3b4aa74SDavid Sterba btrfs_release_path(path); 216025c9bc2eSGerhard Heift if (ret) 2161ac8e9819SChris Mason break; 2162ac8e9819SChris Mason 2163ac8e9819SChris Mason } 21648f5f6178SGerhard Heift if (ret > 0) 2165ac8e9819SChris Mason ret = 0; 2166ac8e9819SChris Mason err: 2167ac8e9819SChris Mason sk->nr_items = num_found; 2168ac8e9819SChris Mason btrfs_free_path(path); 2169ac8e9819SChris Mason return ret; 2170ac8e9819SChris Mason } 2171ac8e9819SChris Mason 2172ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2173ac8e9819SChris Mason void __user *argp) 2174ac8e9819SChris Mason { 2175ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2176ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2177ac8e9819SChris Mason struct inode *inode; 2178ac8e9819SChris Mason int ret; 21799b6e817dSGerhard Heift size_t buf_size; 2180ac8e9819SChris Mason 2181ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2182ac8e9819SChris Mason return -EPERM; 2183ac8e9819SChris Mason 2184ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2185ac8e9819SChris Mason 2186ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2187ba346b35SGerhard Heift return -EFAULT; 2188ba346b35SGerhard Heift 2189ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2190ac8e9819SChris Mason 2191496ad9aaSAl Viro inode = file_inode(file); 2192ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21938f5f6178SGerhard Heift 21948f5f6178SGerhard Heift /* 21958f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21968f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21978f5f6178SGerhard Heift */ 21988f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21998f5f6178SGerhard Heift ret = 0; 22008f5f6178SGerhard Heift 2201ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2202ac8e9819SChris Mason ret = -EFAULT; 2203ac8e9819SChris Mason return ret; 2204ac8e9819SChris Mason } 2205ac8e9819SChris Mason 2206cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2207cc68a8a5SGerhard Heift void __user *argp) 2208cc68a8a5SGerhard Heift { 2209cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2210cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2211cc68a8a5SGerhard Heift struct inode *inode; 2212cc68a8a5SGerhard Heift int ret; 2213cc68a8a5SGerhard Heift size_t buf_size; 2214ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2215cc68a8a5SGerhard Heift 2216cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2217cc68a8a5SGerhard Heift return -EPERM; 2218cc68a8a5SGerhard Heift 2219cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2220cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2221cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2222cc68a8a5SGerhard Heift return -EFAULT; 2223cc68a8a5SGerhard Heift 2224cc68a8a5SGerhard Heift buf_size = args.buf_size; 2225cc68a8a5SGerhard Heift 2226cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2227cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2228cc68a8a5SGerhard Heift buf_size = buf_limit; 2229cc68a8a5SGerhard Heift 2230cc68a8a5SGerhard Heift inode = file_inode(file); 2231cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2232718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2233cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2234cc68a8a5SGerhard Heift ret = -EFAULT; 2235cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2236cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2237cc68a8a5SGerhard Heift ret = -EFAULT; 2238cc68a8a5SGerhard Heift 223976dda93cSYan, Zheng return ret; 224076dda93cSYan, Zheng } 224176dda93cSYan, Zheng 224298d377a0STARUISI Hiroaki /* 2243ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2244ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 224598d377a0STARUISI Hiroaki */ 224698d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 224798d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 224898d377a0STARUISI Hiroaki { 224998d377a0STARUISI Hiroaki struct btrfs_root *root; 225098d377a0STARUISI Hiroaki struct btrfs_key key; 2251ac8e9819SChris Mason char *ptr; 225298d377a0STARUISI Hiroaki int ret = -1; 225398d377a0STARUISI Hiroaki int slot; 225498d377a0STARUISI Hiroaki int len; 225598d377a0STARUISI Hiroaki int total_len = 0; 225698d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 225798d377a0STARUISI Hiroaki struct extent_buffer *l; 225898d377a0STARUISI Hiroaki struct btrfs_path *path; 225998d377a0STARUISI Hiroaki 226098d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 226198d377a0STARUISI Hiroaki name[0]='\0'; 226298d377a0STARUISI Hiroaki return 0; 226398d377a0STARUISI Hiroaki } 226498d377a0STARUISI Hiroaki 226598d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 226698d377a0STARUISI Hiroaki if (!path) 226798d377a0STARUISI Hiroaki return -ENOMEM; 226898d377a0STARUISI Hiroaki 2269c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 227098d377a0STARUISI Hiroaki 227198d377a0STARUISI Hiroaki key.objectid = tree_id; 227298d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 227398d377a0STARUISI Hiroaki key.offset = (u64)-1; 227498d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 227598d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2276ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 22778ad6fcabSChris Mason goto out; 227898d377a0STARUISI Hiroaki } 227998d377a0STARUISI Hiroaki 228098d377a0STARUISI Hiroaki key.objectid = dirid; 228198d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22828ad6fcabSChris Mason key.offset = (u64)-1; 228398d377a0STARUISI Hiroaki 228498d377a0STARUISI Hiroaki while (1) { 228598d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 228698d377a0STARUISI Hiroaki if (ret < 0) 228798d377a0STARUISI Hiroaki goto out; 228818674c6cSFilipe David Borba Manana else if (ret > 0) { 228918674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 229018674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 229118674c6cSFilipe David Borba Manana if (ret < 0) 229218674c6cSFilipe David Borba Manana goto out; 229318674c6cSFilipe David Borba Manana else if (ret > 0) { 2294ac8e9819SChris Mason ret = -ENOENT; 229598d377a0STARUISI Hiroaki goto out; 2296ac8e9819SChris Mason } 229718674c6cSFilipe David Borba Manana } 229818674c6cSFilipe David Borba Manana 229918674c6cSFilipe David Borba Manana l = path->nodes[0]; 230018674c6cSFilipe David Borba Manana slot = path->slots[0]; 230118674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 230298d377a0STARUISI Hiroaki 230398d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 230498d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 230598d377a0STARUISI Hiroaki ptr -= len + 1; 230698d377a0STARUISI Hiroaki total_len += len + 1; 2307a696cf35SFilipe David Borba Manana if (ptr < name) { 2308a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 230998d377a0STARUISI Hiroaki goto out; 2310a696cf35SFilipe David Borba Manana } 231198d377a0STARUISI Hiroaki 231298d377a0STARUISI Hiroaki *(ptr + len) = '/'; 231398d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 231498d377a0STARUISI Hiroaki 231598d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 231698d377a0STARUISI Hiroaki break; 231798d377a0STARUISI Hiroaki 2318b3b4aa74SDavid Sterba btrfs_release_path(path); 231998d377a0STARUISI Hiroaki key.objectid = key.offset; 23208ad6fcabSChris Mason key.offset = (u64)-1; 232198d377a0STARUISI Hiroaki dirid = key.objectid; 232298d377a0STARUISI Hiroaki } 232377906a50SLi Zefan memmove(name, ptr, total_len); 232498d377a0STARUISI Hiroaki name[total_len] = '\0'; 232598d377a0STARUISI Hiroaki ret = 0; 232698d377a0STARUISI Hiroaki out: 232798d377a0STARUISI Hiroaki btrfs_free_path(path); 2328ac8e9819SChris Mason return ret; 2329ac8e9819SChris Mason } 2330ac8e9819SChris Mason 233123d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 233223d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 233323d0b79dSTomohiro Misono { 233423d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 233523d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 233623d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 233723d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 233823d0b79dSTomohiro Misono u64 dirid = args->dirid; 233923d0b79dSTomohiro Misono unsigned long item_off; 234023d0b79dSTomohiro Misono unsigned long item_len; 234123d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 234223d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 234323d0b79dSTomohiro Misono struct btrfs_root *root; 234423d0b79dSTomohiro Misono struct btrfs_path *path; 234523d0b79dSTomohiro Misono struct btrfs_key key, key2; 234623d0b79dSTomohiro Misono struct extent_buffer *leaf; 234723d0b79dSTomohiro Misono struct inode *temp_inode; 234823d0b79dSTomohiro Misono char *ptr; 234923d0b79dSTomohiro Misono int slot; 235023d0b79dSTomohiro Misono int len; 235123d0b79dSTomohiro Misono int total_len = 0; 235223d0b79dSTomohiro Misono int ret; 235323d0b79dSTomohiro Misono 235423d0b79dSTomohiro Misono path = btrfs_alloc_path(); 235523d0b79dSTomohiro Misono if (!path) 235623d0b79dSTomohiro Misono return -ENOMEM; 235723d0b79dSTomohiro Misono 235823d0b79dSTomohiro Misono /* 235923d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 236023d0b79dSTomohiro Misono * construct the path in from the bottom up. 236123d0b79dSTomohiro Misono */ 236223d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 236323d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 236423d0b79dSTomohiro Misono 236523d0b79dSTomohiro Misono key.objectid = treeid; 236623d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 236723d0b79dSTomohiro Misono key.offset = (u64)-1; 236823d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 236923d0b79dSTomohiro Misono if (IS_ERR(root)) { 237023d0b79dSTomohiro Misono ret = PTR_ERR(root); 237123d0b79dSTomohiro Misono goto out; 237223d0b79dSTomohiro Misono } 237323d0b79dSTomohiro Misono 237423d0b79dSTomohiro Misono key.objectid = dirid; 237523d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 237623d0b79dSTomohiro Misono key.offset = (u64)-1; 237723d0b79dSTomohiro Misono while (1) { 237823d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 237923d0b79dSTomohiro Misono if (ret < 0) { 238023d0b79dSTomohiro Misono goto out; 238123d0b79dSTomohiro Misono } else if (ret > 0) { 238223d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 238323d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 238423d0b79dSTomohiro Misono if (ret < 0) { 238523d0b79dSTomohiro Misono goto out; 238623d0b79dSTomohiro Misono } else if (ret > 0) { 238723d0b79dSTomohiro Misono ret = -ENOENT; 238823d0b79dSTomohiro Misono goto out; 238923d0b79dSTomohiro Misono } 239023d0b79dSTomohiro Misono } 239123d0b79dSTomohiro Misono 239223d0b79dSTomohiro Misono leaf = path->nodes[0]; 239323d0b79dSTomohiro Misono slot = path->slots[0]; 239423d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 239523d0b79dSTomohiro Misono 239623d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 239723d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 239823d0b79dSTomohiro Misono ptr -= len + 1; 239923d0b79dSTomohiro Misono total_len += len + 1; 240023d0b79dSTomohiro Misono if (ptr < args->path) { 240123d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 240223d0b79dSTomohiro Misono goto out; 240323d0b79dSTomohiro Misono } 240423d0b79dSTomohiro Misono 240523d0b79dSTomohiro Misono *(ptr + len) = '/'; 240623d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 240723d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 240823d0b79dSTomohiro Misono 240923d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 241023d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 241123d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 241223d0b79dSTomohiro Misono if (ret < 0) { 241323d0b79dSTomohiro Misono goto out; 241423d0b79dSTomohiro Misono } else if (ret > 0) { 241523d0b79dSTomohiro Misono ret = -ENOENT; 241623d0b79dSTomohiro Misono goto out; 241723d0b79dSTomohiro Misono } 241823d0b79dSTomohiro Misono 241923d0b79dSTomohiro Misono leaf = path->nodes[0]; 242023d0b79dSTomohiro Misono slot = path->slots[0]; 242123d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 242223d0b79dSTomohiro Misono if (key2.objectid != dirid) { 242323d0b79dSTomohiro Misono ret = -ENOENT; 242423d0b79dSTomohiro Misono goto out; 242523d0b79dSTomohiro Misono } 242623d0b79dSTomohiro Misono 242723d0b79dSTomohiro Misono temp_inode = btrfs_iget(sb, &key2, root, NULL); 24283ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24293ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24303ca57bd6SMisono Tomohiro goto out; 24313ca57bd6SMisono Tomohiro } 243223d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 243323d0b79dSTomohiro Misono iput(temp_inode); 243423d0b79dSTomohiro Misono if (ret) { 243523d0b79dSTomohiro Misono ret = -EACCES; 243623d0b79dSTomohiro Misono goto out; 243723d0b79dSTomohiro Misono } 243823d0b79dSTomohiro Misono 243923d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 244023d0b79dSTomohiro Misono break; 244123d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 244223d0b79dSTomohiro Misono ret = -EACCES; 244323d0b79dSTomohiro Misono goto out; 244423d0b79dSTomohiro Misono } 244523d0b79dSTomohiro Misono 244623d0b79dSTomohiro Misono btrfs_release_path(path); 244723d0b79dSTomohiro Misono key.objectid = key.offset; 244823d0b79dSTomohiro Misono key.offset = (u64)-1; 244923d0b79dSTomohiro Misono dirid = key.objectid; 245023d0b79dSTomohiro Misono } 245123d0b79dSTomohiro Misono 245223d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 245323d0b79dSTomohiro Misono args->path[total_len] = '\0'; 245423d0b79dSTomohiro Misono btrfs_release_path(path); 245523d0b79dSTomohiro Misono } 245623d0b79dSTomohiro Misono 245723d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 245823d0b79dSTomohiro Misono root = fs_info->tree_root; 245923d0b79dSTomohiro Misono key.objectid = treeid; 246023d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 246123d0b79dSTomohiro Misono key.offset = args->treeid; 246223d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 246323d0b79dSTomohiro Misono if (ret < 0) { 246423d0b79dSTomohiro Misono goto out; 246523d0b79dSTomohiro Misono } else if (ret > 0) { 246623d0b79dSTomohiro Misono ret = -ENOENT; 246723d0b79dSTomohiro Misono goto out; 246823d0b79dSTomohiro Misono } 246923d0b79dSTomohiro Misono 247023d0b79dSTomohiro Misono leaf = path->nodes[0]; 247123d0b79dSTomohiro Misono slot = path->slots[0]; 247223d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 247323d0b79dSTomohiro Misono 247423d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 247523d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 247623d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 247723d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 247823d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 247923d0b79dSTomohiro Misono ret = -EINVAL; 248023d0b79dSTomohiro Misono goto out; 248123d0b79dSTomohiro Misono } 248223d0b79dSTomohiro Misono 248323d0b79dSTomohiro Misono /* Copy subvolume's name */ 248423d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 248523d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 248623d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 248723d0b79dSTomohiro Misono args->name[item_len] = 0; 248823d0b79dSTomohiro Misono 248923d0b79dSTomohiro Misono out: 249023d0b79dSTomohiro Misono btrfs_free_path(path); 249123d0b79dSTomohiro Misono return ret; 249223d0b79dSTomohiro Misono } 249323d0b79dSTomohiro Misono 2494ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2495ac8e9819SChris Mason void __user *argp) 2496ac8e9819SChris Mason { 2497ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2498ac8e9819SChris Mason struct inode *inode; 249901b810b8SDavid Sterba int ret = 0; 2500ac8e9819SChris Mason 25012354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25022354d08fSJulia Lawall if (IS_ERR(args)) 25032354d08fSJulia Lawall return PTR_ERR(args); 2504c2b96929SDan Carpenter 2505496ad9aaSAl Viro inode = file_inode(file); 2506ac8e9819SChris Mason 250701b810b8SDavid Sterba /* 250801b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 250901b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 251001b810b8SDavid Sterba */ 25111b53ac4dSChris Mason if (args->treeid == 0) 25121b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25131b53ac4dSChris Mason 251401b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 251501b810b8SDavid Sterba args->name[0] = 0; 251601b810b8SDavid Sterba goto out; 251701b810b8SDavid Sterba } 251801b810b8SDavid Sterba 251901b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 252001b810b8SDavid Sterba ret = -EPERM; 252101b810b8SDavid Sterba goto out; 252201b810b8SDavid Sterba } 252301b810b8SDavid Sterba 2524ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2525ac8e9819SChris Mason args->treeid, args->objectid, 2526ac8e9819SChris Mason args->name); 2527ac8e9819SChris Mason 252801b810b8SDavid Sterba out: 2529ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2530ac8e9819SChris Mason ret = -EFAULT; 2531ac8e9819SChris Mason 2532ac8e9819SChris Mason kfree(args); 253398d377a0STARUISI Hiroaki return ret; 253498d377a0STARUISI Hiroaki } 253598d377a0STARUISI Hiroaki 253623d0b79dSTomohiro Misono /* 253723d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 253823d0b79dSTomohiro Misono * 253923d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 254023d0b79dSTomohiro Misono * 254123d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 254223d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 254323d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 254423d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 254523d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 254623d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 254723d0b79dSTomohiro Misono */ 254823d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 254923d0b79dSTomohiro Misono { 255023d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 255123d0b79dSTomohiro Misono struct inode *inode; 255223d0b79dSTomohiro Misono int ret; 255323d0b79dSTomohiro Misono 255423d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 255523d0b79dSTomohiro Misono if (IS_ERR(args)) 255623d0b79dSTomohiro Misono return PTR_ERR(args); 255723d0b79dSTomohiro Misono 255823d0b79dSTomohiro Misono inode = file_inode(file); 255923d0b79dSTomohiro Misono 256023d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 256123d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 256223d0b79dSTomohiro Misono /* 256323d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 256423d0b79dSTomohiro Misono * called 256523d0b79dSTomohiro Misono */ 256623d0b79dSTomohiro Misono kfree(args); 256723d0b79dSTomohiro Misono return -EACCES; 256823d0b79dSTomohiro Misono } 256923d0b79dSTomohiro Misono 257023d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 257123d0b79dSTomohiro Misono 257223d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 257323d0b79dSTomohiro Misono ret = -EFAULT; 257423d0b79dSTomohiro Misono 257523d0b79dSTomohiro Misono kfree(args); 257623d0b79dSTomohiro Misono return ret; 257723d0b79dSTomohiro Misono } 257823d0b79dSTomohiro Misono 2579b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2580b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2581b64ec075STomohiro Misono { 2582b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2583b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2584b64ec075STomohiro Misono struct btrfs_root *root; 2585b64ec075STomohiro Misono struct btrfs_path *path; 2586b64ec075STomohiro Misono struct btrfs_key key; 2587b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2588b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2589b64ec075STomohiro Misono struct extent_buffer *leaf; 2590b64ec075STomohiro Misono unsigned long item_off; 2591b64ec075STomohiro Misono unsigned long item_len; 2592b64ec075STomohiro Misono struct inode *inode; 2593b64ec075STomohiro Misono int slot; 2594b64ec075STomohiro Misono int ret = 0; 2595b64ec075STomohiro Misono 2596b64ec075STomohiro Misono path = btrfs_alloc_path(); 2597b64ec075STomohiro Misono if (!path) 2598b64ec075STomohiro Misono return -ENOMEM; 2599b64ec075STomohiro Misono 2600b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2601b64ec075STomohiro Misono if (!subvol_info) { 2602b64ec075STomohiro Misono btrfs_free_path(path); 2603b64ec075STomohiro Misono return -ENOMEM; 2604b64ec075STomohiro Misono } 2605b64ec075STomohiro Misono 2606b64ec075STomohiro Misono inode = file_inode(file); 2607b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2608b64ec075STomohiro Misono 2609b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2610b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2611b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2612b64ec075STomohiro Misono key.offset = (u64)-1; 2613b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2614b64ec075STomohiro Misono if (IS_ERR(root)) { 2615b64ec075STomohiro Misono ret = PTR_ERR(root); 2616b64ec075STomohiro Misono goto out; 2617b64ec075STomohiro Misono } 2618b64ec075STomohiro Misono root_item = &root->root_item; 2619b64ec075STomohiro Misono 2620b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2621b64ec075STomohiro Misono 2622b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2623b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2624b64ec075STomohiro Misono 2625b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2626b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2627b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2628b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2629b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2630b64ec075STomohiro Misono 2631b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2632b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2633b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2634b64ec075STomohiro Misono 2635b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2636b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2637b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2638b64ec075STomohiro Misono 2639b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2640b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2641b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2642b64ec075STomohiro Misono 2643b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2644b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2645b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2646b64ec075STomohiro Misono 2647b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2648b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2649b64ec075STomohiro Misono root = fs_info->tree_root; 2650b64ec075STomohiro Misono 2651b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2652b64ec075STomohiro Misono key.offset = 0; 2653b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2654b64ec075STomohiro Misono if (ret < 0) { 2655b64ec075STomohiro Misono goto out; 2656b64ec075STomohiro Misono } else if (path->slots[0] >= 2657b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2658b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2659b64ec075STomohiro Misono if (ret < 0) { 2660b64ec075STomohiro Misono goto out; 2661b64ec075STomohiro Misono } else if (ret > 0) { 2662b64ec075STomohiro Misono ret = -EUCLEAN; 2663b64ec075STomohiro Misono goto out; 2664b64ec075STomohiro Misono } 2665b64ec075STomohiro Misono } 2666b64ec075STomohiro Misono 2667b64ec075STomohiro Misono leaf = path->nodes[0]; 2668b64ec075STomohiro Misono slot = path->slots[0]; 2669b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2670b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2671b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2672b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2673b64ec075STomohiro Misono 2674b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2675b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2676b64ec075STomohiro Misono 2677b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2678b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2679b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2680b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2681b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2682b64ec075STomohiro Misono item_off, item_len); 2683b64ec075STomohiro Misono } else { 2684b64ec075STomohiro Misono ret = -ENOENT; 2685b64ec075STomohiro Misono goto out; 2686b64ec075STomohiro Misono } 2687b64ec075STomohiro Misono } 2688b64ec075STomohiro Misono 2689b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2690b64ec075STomohiro Misono ret = -EFAULT; 2691b64ec075STomohiro Misono 2692b64ec075STomohiro Misono out: 2693b64ec075STomohiro Misono btrfs_free_path(path); 2694b64ec075STomohiro Misono kzfree(subvol_info); 2695b64ec075STomohiro Misono return ret; 2696b64ec075STomohiro Misono } 2697b64ec075STomohiro Misono 269842e4b520STomohiro Misono /* 269942e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 270042e4b520STomohiro Misono * except the subvolume name. 270142e4b520STomohiro Misono */ 270242e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 270342e4b520STomohiro Misono { 270442e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 270542e4b520STomohiro Misono struct btrfs_root_ref *rref; 270642e4b520STomohiro Misono struct btrfs_root *root; 270742e4b520STomohiro Misono struct btrfs_path *path; 270842e4b520STomohiro Misono struct btrfs_key key; 270942e4b520STomohiro Misono struct extent_buffer *leaf; 271042e4b520STomohiro Misono struct inode *inode; 271142e4b520STomohiro Misono u64 objectid; 271242e4b520STomohiro Misono int slot; 271342e4b520STomohiro Misono int ret; 271442e4b520STomohiro Misono u8 found; 271542e4b520STomohiro Misono 271642e4b520STomohiro Misono path = btrfs_alloc_path(); 271742e4b520STomohiro Misono if (!path) 271842e4b520STomohiro Misono return -ENOMEM; 271942e4b520STomohiro Misono 272042e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 272142e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 272242e4b520STomohiro Misono btrfs_free_path(path); 272342e4b520STomohiro Misono return PTR_ERR(rootrefs); 272442e4b520STomohiro Misono } 272542e4b520STomohiro Misono 272642e4b520STomohiro Misono inode = file_inode(file); 272742e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 272842e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 272942e4b520STomohiro Misono 273042e4b520STomohiro Misono key.objectid = objectid; 273142e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 273242e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 273342e4b520STomohiro Misono found = 0; 273442e4b520STomohiro Misono 273542e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 273642e4b520STomohiro Misono if (ret < 0) { 273742e4b520STomohiro Misono goto out; 273842e4b520STomohiro Misono } else if (path->slots[0] >= 273942e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 274042e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 274142e4b520STomohiro Misono if (ret < 0) { 274242e4b520STomohiro Misono goto out; 274342e4b520STomohiro Misono } else if (ret > 0) { 274442e4b520STomohiro Misono ret = -EUCLEAN; 274542e4b520STomohiro Misono goto out; 274642e4b520STomohiro Misono } 274742e4b520STomohiro Misono } 274842e4b520STomohiro Misono while (1) { 274942e4b520STomohiro Misono leaf = path->nodes[0]; 275042e4b520STomohiro Misono slot = path->slots[0]; 275142e4b520STomohiro Misono 275242e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 275342e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 275442e4b520STomohiro Misono ret = 0; 275542e4b520STomohiro Misono goto out; 275642e4b520STomohiro Misono } 275742e4b520STomohiro Misono 275842e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 275942e4b520STomohiro Misono ret = -EOVERFLOW; 276042e4b520STomohiro Misono goto out; 276142e4b520STomohiro Misono } 276242e4b520STomohiro Misono 276342e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 276442e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 276542e4b520STomohiro Misono rootrefs->rootref[found].dirid = 276642e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 276742e4b520STomohiro Misono found++; 276842e4b520STomohiro Misono 276942e4b520STomohiro Misono ret = btrfs_next_item(root, path); 277042e4b520STomohiro Misono if (ret < 0) { 277142e4b520STomohiro Misono goto out; 277242e4b520STomohiro Misono } else if (ret > 0) { 277342e4b520STomohiro Misono ret = -EUCLEAN; 277442e4b520STomohiro Misono goto out; 277542e4b520STomohiro Misono } 277642e4b520STomohiro Misono } 277742e4b520STomohiro Misono 277842e4b520STomohiro Misono out: 277942e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 278042e4b520STomohiro Misono rootrefs->num_items = found; 278142e4b520STomohiro Misono /* update min_treeid for next search */ 278242e4b520STomohiro Misono if (found) 278342e4b520STomohiro Misono rootrefs->min_treeid = 278442e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 278542e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 278642e4b520STomohiro Misono ret = -EFAULT; 278742e4b520STomohiro Misono } 278842e4b520STomohiro Misono 278942e4b520STomohiro Misono kfree(rootrefs); 279042e4b520STomohiro Misono btrfs_free_path(path); 279142e4b520STomohiro Misono 279242e4b520STomohiro Misono return ret; 279342e4b520STomohiro Misono } 279442e4b520STomohiro Misono 279576dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 279676dda93cSYan, Zheng void __user *arg) 279776dda93cSYan, Zheng { 279854563d41SAl Viro struct dentry *parent = file->f_path.dentry; 27990b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 280076dda93cSYan, Zheng struct dentry *dentry; 28012b0143b5SDavid Howells struct inode *dir = d_inode(parent); 280276dda93cSYan, Zheng struct inode *inode; 280376dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 280476dda93cSYan, Zheng struct btrfs_root *dest = NULL; 280576dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 280676dda93cSYan, Zheng int namelen; 280776dda93cSYan, Zheng int err = 0; 280876dda93cSYan, Zheng 2809325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2810325c50e3SJeff Mahoney return -ENOTDIR; 2811325c50e3SJeff Mahoney 281276dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 281376dda93cSYan, Zheng if (IS_ERR(vol_args)) 281476dda93cSYan, Zheng return PTR_ERR(vol_args); 281576dda93cSYan, Zheng 281676dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 281776dda93cSYan, Zheng namelen = strlen(vol_args->name); 281876dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 281976dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 282076dda93cSYan, Zheng err = -EINVAL; 282176dda93cSYan, Zheng goto out; 282276dda93cSYan, Zheng } 282376dda93cSYan, Zheng 2824a561be71SAl Viro err = mnt_want_write_file(file); 282576dda93cSYan, Zheng if (err) 282676dda93cSYan, Zheng goto out; 282776dda93cSYan, Zheng 2828521e0546SDavid Sterba 282900235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 283000235411SAl Viro if (err == -EINTR) 283100235411SAl Viro goto out_drop_write; 283276dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 283376dda93cSYan, Zheng if (IS_ERR(dentry)) { 283476dda93cSYan, Zheng err = PTR_ERR(dentry); 283576dda93cSYan, Zheng goto out_unlock_dir; 283676dda93cSYan, Zheng } 283776dda93cSYan, Zheng 28382b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 283976dda93cSYan, Zheng err = -ENOENT; 284076dda93cSYan, Zheng goto out_dput; 284176dda93cSYan, Zheng } 284276dda93cSYan, Zheng 28432b0143b5SDavid Howells inode = d_inode(dentry); 28444260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28454260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28464260f7c7SSage Weil /* 28474260f7c7SSage Weil * Regular user. Only allow this with a special mount 28484260f7c7SSage Weil * option, when the user has write+exec access to the 28494260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28504260f7c7SSage Weil * allowed. 28514260f7c7SSage Weil * 28524260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28534260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28544260f7c7SSage Weil * otherwise be able to delete. 28554260f7c7SSage Weil * 28564260f7c7SSage Weil * Users who want to delete empty subvols should try 28574260f7c7SSage Weil * rmdir(2). 28584260f7c7SSage Weil */ 28594260f7c7SSage Weil err = -EPERM; 28600b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 28614260f7c7SSage Weil goto out_dput; 28624260f7c7SSage Weil 28634260f7c7SSage Weil /* 28644260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 28654260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 28664260f7c7SSage Weil * must be called on the dentry referencing the root 28674260f7c7SSage Weil * of the subvol, not a random directory contained 28684260f7c7SSage Weil * within it. 28694260f7c7SSage Weil */ 28704260f7c7SSage Weil err = -EINVAL; 28714260f7c7SSage Weil if (root == dest) 28724260f7c7SSage Weil goto out_dput; 28734260f7c7SSage Weil 28744260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 28754260f7c7SSage Weil if (err) 28764260f7c7SSage Weil goto out_dput; 28775c39da5bSMiao Xie } 28784260f7c7SSage Weil 28795c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 28804260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 28814260f7c7SSage Weil if (err) 28824260f7c7SSage Weil goto out_dput; 28834260f7c7SSage Weil 28844a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 288576dda93cSYan, Zheng err = -EINVAL; 288676dda93cSYan, Zheng goto out_dput; 288776dda93cSYan, Zheng } 288876dda93cSYan, Zheng 28895955102cSAl Viro inode_lock(inode); 2890f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 28915955102cSAl Viro inode_unlock(inode); 2892f60a2364SMisono Tomohiro if (!err) 289376dda93cSYan, Zheng d_delete(dentry); 2894fa6ac876SLiu Bo 289576dda93cSYan, Zheng out_dput: 289676dda93cSYan, Zheng dput(dentry); 289776dda93cSYan, Zheng out_unlock_dir: 28985955102cSAl Viro inode_unlock(dir); 289900235411SAl Viro out_drop_write: 29002a79f17eSAl Viro mnt_drop_write_file(file); 290176dda93cSYan, Zheng out: 290276dda93cSYan, Zheng kfree(vol_args); 290376dda93cSYan, Zheng return err; 290476dda93cSYan, Zheng } 290576dda93cSYan, Zheng 29061e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2907f46b5a66SChristoph Hellwig { 2908496ad9aaSAl Viro struct inode *inode = file_inode(file); 2909f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29101e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2911c146afadSYan Zheng int ret; 2912c146afadSYan Zheng 291325122d15SIlya Dryomov ret = mnt_want_write_file(file); 291425122d15SIlya Dryomov if (ret) 291525122d15SIlya Dryomov return ret; 2916b83cc969SLi Zefan 291725122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 291825122d15SIlya Dryomov ret = -EROFS; 291925122d15SIlya Dryomov goto out; 29205ac00addSStefan Behrens } 2921f46b5a66SChristoph Hellwig 2922f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2923f46b5a66SChristoph Hellwig case S_IFDIR: 2924e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2925e441d54dSChris Mason ret = -EPERM; 2926e441d54dSChris Mason goto out; 2927e441d54dSChris Mason } 2928de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2929f46b5a66SChristoph Hellwig break; 2930f46b5a66SChristoph Hellwig case S_IFREG: 2931e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2932e441d54dSChris Mason ret = -EINVAL; 2933e441d54dSChris Mason goto out; 2934e441d54dSChris Mason } 29351e701a32SChris Mason 29361e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29371e701a32SChris Mason if (!range) { 29381e701a32SChris Mason ret = -ENOMEM; 29391e701a32SChris Mason goto out; 29401e701a32SChris Mason } 29411e701a32SChris Mason 29421e701a32SChris Mason if (argp) { 29431e701a32SChris Mason if (copy_from_user(range, argp, 29441e701a32SChris Mason sizeof(*range))) { 29451e701a32SChris Mason ret = -EFAULT; 29461e701a32SChris Mason kfree(range); 2947683be16eSDan Carpenter goto out; 29481e701a32SChris Mason } 29491e701a32SChris Mason /* compression requires us to start the IO */ 29501e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 29511e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 29521e701a32SChris Mason range->extent_thresh = (u32)-1; 29531e701a32SChris Mason } 29541e701a32SChris Mason } else { 29551e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 29561e701a32SChris Mason range->len = (u64)-1; 29571e701a32SChris Mason } 2958496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 29597c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 29604cb5300bSChris Mason if (ret > 0) 29614cb5300bSChris Mason ret = 0; 29621e701a32SChris Mason kfree(range); 2963f46b5a66SChristoph Hellwig break; 29648929ecfaSYan, Zheng default: 29658929ecfaSYan, Zheng ret = -EINVAL; 2966f46b5a66SChristoph Hellwig } 2967e441d54dSChris Mason out: 296825122d15SIlya Dryomov mnt_drop_write_file(file); 2969e441d54dSChris Mason return ret; 2970f46b5a66SChristoph Hellwig } 2971f46b5a66SChristoph Hellwig 29722ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2973f46b5a66SChristoph Hellwig { 2974f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2975f46b5a66SChristoph Hellwig int ret; 2976f46b5a66SChristoph Hellwig 2977e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2978e441d54dSChris Mason return -EPERM; 2979e441d54dSChris Mason 2980171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 2981e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2982c9e9f97bSIlya Dryomov 2983dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2984c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2985c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2986c9e9f97bSIlya Dryomov goto out; 2987c9e9f97bSIlya Dryomov } 2988f46b5a66SChristoph Hellwig 29895516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 29902ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 2991f46b5a66SChristoph Hellwig 299243d20761SAnand Jain if (!ret) 29930b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 299443d20761SAnand Jain 2995f46b5a66SChristoph Hellwig kfree(vol_args); 2996c9e9f97bSIlya Dryomov out: 2997171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2998f46b5a66SChristoph Hellwig return ret; 2999f46b5a66SChristoph Hellwig } 3000f46b5a66SChristoph Hellwig 30016b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3002f46b5a66SChristoph Hellwig { 30030b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30040b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30056b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3006f46b5a66SChristoph Hellwig int ret; 3007f46b5a66SChristoph Hellwig 3008e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3009e441d54dSChris Mason return -EPERM; 3010e441d54dSChris Mason 3011da24927bSMiao Xie ret = mnt_want_write_file(file); 3012da24927bSMiao Xie if (ret) 3013da24927bSMiao Xie return ret; 3014c146afadSYan Zheng 3015dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3016c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3017c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3018c47ca32dSDan Carpenter goto err_drop; 3019c9e9f97bSIlya Dryomov } 3020f46b5a66SChristoph Hellwig 30216b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3022fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3023fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3024fd4e994bSOmar Sandoval goto out; 3025fd4e994bSOmar Sandoval } 3026f46b5a66SChristoph Hellwig 3027171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3028183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3029183860f6SAnand Jain goto out; 3030183860f6SAnand Jain } 3031183860f6SAnand Jain 3032735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30332ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30346b526ed7SAnand Jain } else { 30356b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30362ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30376b526ed7SAnand Jain } 3038171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3039183860f6SAnand Jain 30406b526ed7SAnand Jain if (!ret) { 3041735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30420b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30436b526ed7SAnand Jain vol_args->devid); 30446b526ed7SAnand Jain else 30450b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30466b526ed7SAnand Jain vol_args->name); 30476b526ed7SAnand Jain } 3048183860f6SAnand Jain out: 3049183860f6SAnand Jain kfree(vol_args); 3050c47ca32dSDan Carpenter err_drop: 30514ac20c70SIlya Dryomov mnt_drop_write_file(file); 3052f46b5a66SChristoph Hellwig return ret; 3053f46b5a66SChristoph Hellwig } 3054f46b5a66SChristoph Hellwig 3055f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3056f46b5a66SChristoph Hellwig { 30570b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3059f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3060f46b5a66SChristoph Hellwig int ret; 3061f46b5a66SChristoph Hellwig 3062f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3063f46b5a66SChristoph Hellwig return -EPERM; 3064f46b5a66SChristoph Hellwig 3065f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3066f46b5a66SChristoph Hellwig if (ret) 3067f46b5a66SChristoph Hellwig return ret; 3068f46b5a66SChristoph Hellwig 3069171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3070f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 307158d7bbf8SDavid Sterba goto out_drop_write; 307258d7bbf8SDavid Sterba } 307358d7bbf8SDavid Sterba 307458d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 307558d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 307658d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3077f46b5a66SChristoph Hellwig goto out; 3078f46b5a66SChristoph Hellwig } 3079f46b5a66SChristoph Hellwig 308058d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30812ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3082f46b5a66SChristoph Hellwig 3083f46b5a66SChristoph Hellwig if (!ret) 30840b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3085f46b5a66SChristoph Hellwig kfree(vol_args); 308658d7bbf8SDavid Sterba out: 3087171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 308858d7bbf8SDavid Sterba out_drop_write: 3089f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 309058d7bbf8SDavid Sterba 3091f46b5a66SChristoph Hellwig return ret; 3092f46b5a66SChristoph Hellwig } 3093f46b5a66SChristoph Hellwig 30942ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 30952ff7e61eSJeff Mahoney void __user *arg) 3096475f6387SJan Schmidt { 3097027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3098475f6387SJan Schmidt struct btrfs_device *device; 30990b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3100027ed2f0SLi Zefan int ret = 0; 3101475f6387SJan Schmidt 3102027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3103027ed2f0SLi Zefan if (!fi_args) 3104027ed2f0SLi Zefan return -ENOMEM; 3105027ed2f0SLi Zefan 3106d03262c7SDavid Sterba rcu_read_lock(); 3107027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3108475f6387SJan Schmidt 3109d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3110027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3111027ed2f0SLi Zefan fi_args->max_id = device->devid; 3112475f6387SJan Schmidt } 3113d03262c7SDavid Sterba rcu_read_unlock(); 3114475f6387SJan Schmidt 3115d03262c7SDavid Sterba memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 3116bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3117bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3118bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 311980a773fbSDavid Sterba 3120027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3121027ed2f0SLi Zefan ret = -EFAULT; 3122475f6387SJan Schmidt 3123027ed2f0SLi Zefan kfree(fi_args); 3124027ed2f0SLi Zefan return ret; 3125475f6387SJan Schmidt } 3126475f6387SJan Schmidt 31272ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31282ff7e61eSJeff Mahoney void __user *arg) 3129475f6387SJan Schmidt { 3130475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3131475f6387SJan Schmidt struct btrfs_device *dev; 3132475f6387SJan Schmidt int ret = 0; 3133475f6387SJan Schmidt char *s_uuid = NULL; 3134475f6387SJan Schmidt 3135475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3136475f6387SJan Schmidt if (IS_ERR(di_args)) 3137475f6387SJan Schmidt return PTR_ERR(di_args); 3138475f6387SJan Schmidt 3139dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3140475f6387SJan Schmidt s_uuid = di_args->uuid; 3141475f6387SJan Schmidt 3142c5593ca3SDavid Sterba rcu_read_lock(); 31430b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 3144475f6387SJan Schmidt 3145475f6387SJan Schmidt if (!dev) { 3146475f6387SJan Schmidt ret = -ENODEV; 3147475f6387SJan Schmidt goto out; 3148475f6387SJan Schmidt } 3149475f6387SJan Schmidt 3150475f6387SJan Schmidt di_args->devid = dev->devid; 31517cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 31527cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3153475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3154a27202fbSJim Meyering if (dev->name) { 3155606686eeSJosef Bacik struct rcu_string *name; 3156606686eeSJosef Bacik 3157606686eeSJosef Bacik name = rcu_dereference(dev->name); 31586670d4c2SXiongfeng Wang strncpy(di_args->path, name->str, sizeof(di_args->path) - 1); 3159a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3160a27202fbSJim Meyering } else { 316199ba55adSStefan Behrens di_args->path[0] = '\0'; 3162a27202fbSJim Meyering } 3163475f6387SJan Schmidt 3164475f6387SJan Schmidt out: 3165c5593ca3SDavid Sterba rcu_read_unlock(); 3166475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3167475f6387SJan Schmidt ret = -EFAULT; 3168475f6387SJan Schmidt 3169475f6387SJan Schmidt kfree(di_args); 3170475f6387SJan Schmidt return ret; 3171475f6387SJan Schmidt } 3172475f6387SJan Schmidt 3173f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 3174416161dbSMark Fasheh { 3175416161dbSMark Fasheh struct page *page; 3176416161dbSMark Fasheh 3177416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 3178416161dbSMark Fasheh if (!page) 317931314002SFilipe Manana return ERR_PTR(-ENOMEM); 3180416161dbSMark Fasheh 3181416161dbSMark Fasheh if (!PageUptodate(page)) { 318231314002SFilipe Manana int ret; 318331314002SFilipe Manana 318431314002SFilipe Manana ret = btrfs_readpage(NULL, page); 318531314002SFilipe Manana if (ret) 318631314002SFilipe Manana return ERR_PTR(ret); 3187416161dbSMark Fasheh lock_page(page); 3188416161dbSMark Fasheh if (!PageUptodate(page)) { 3189416161dbSMark Fasheh unlock_page(page); 319009cbfeafSKirill A. Shutemov put_page(page); 319131314002SFilipe Manana return ERR_PTR(-EIO); 3192416161dbSMark Fasheh } 319331314002SFilipe Manana if (page->mapping != inode->i_mapping) { 3194416161dbSMark Fasheh unlock_page(page); 319509cbfeafSKirill A. Shutemov put_page(page); 319631314002SFilipe Manana return ERR_PTR(-EAGAIN); 3197416161dbSMark Fasheh } 3198416161dbSMark Fasheh } 3199416161dbSMark Fasheh 3200416161dbSMark Fasheh return page; 3201416161dbSMark Fasheh } 3202416161dbSMark Fasheh 3203f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 3204f4414602SMark Fasheh int num_pages, u64 off) 3205f4414602SMark Fasheh { 3206f4414602SMark Fasheh int i; 320709cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 3208f4414602SMark Fasheh 3209f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 321031314002SFilipe Manana again: 3211f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 321231314002SFilipe Manana if (IS_ERR(pages[i])) { 321331314002SFilipe Manana int err = PTR_ERR(pages[i]); 321431314002SFilipe Manana 321531314002SFilipe Manana if (err == -EAGAIN) 321631314002SFilipe Manana goto again; 321731314002SFilipe Manana pages[i] = NULL; 321831314002SFilipe Manana return err; 321931314002SFilipe Manana } 3220f4414602SMark Fasheh } 3221f4414602SMark Fasheh return 0; 3222f4414602SMark Fasheh } 3223f4414602SMark Fasheh 3224e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 3225e0bd70c6SFilipe Manana bool retry_range_locking) 322677fe20dcSMark Fasheh { 3227e0bd70c6SFilipe Manana /* 3228e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 3229e0bd70c6SFilipe Manana * another, and lock file content. 3230e0bd70c6SFilipe Manana * The locking order is: 3231e0bd70c6SFilipe Manana * 3232e0bd70c6SFilipe Manana * 1) pages 3233e0bd70c6SFilipe Manana * 2) range in the inode's io tree 3234e0bd70c6SFilipe Manana */ 323577fe20dcSMark Fasheh while (1) { 323677fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 323777fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 323877fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 323977fe20dcSMark Fasheh off + len - 1); 3240ff5df9b8SFilipe Manana if ((!ordered || 3241ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 3242ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 324377fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 3244ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 3245ff5df9b8SFilipe Manana if (ordered) 3246ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 324777fe20dcSMark Fasheh break; 3248ff5df9b8SFilipe Manana } 324977fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 325077fe20dcSMark Fasheh if (ordered) 325177fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 3252e0bd70c6SFilipe Manana if (!retry_range_locking) 3253e0bd70c6SFilipe Manana return -EAGAIN; 325477fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 325577fe20dcSMark Fasheh } 3256e0bd70c6SFilipe Manana return 0; 325777fe20dcSMark Fasheh } 325877fe20dcSMark Fasheh 3259f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 3260416161dbSMark Fasheh { 32615955102cSAl Viro inode_unlock(inode1); 32625955102cSAl Viro inode_unlock(inode2); 3263416161dbSMark Fasheh } 3264416161dbSMark Fasheh 3265f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 3266f4414602SMark Fasheh { 3267f4414602SMark Fasheh if (inode1 < inode2) 3268f4414602SMark Fasheh swap(inode1, inode2); 3269f4414602SMark Fasheh 32705955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 32715955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 3272f4414602SMark Fasheh } 3273f4414602SMark Fasheh 3274f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3275f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 3276f4414602SMark Fasheh { 3277f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3278f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3279f4414602SMark Fasheh } 3280f4414602SMark Fasheh 3281e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3282e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 3283e0bd70c6SFilipe Manana bool retry_range_locking) 3284416161dbSMark Fasheh { 3285e0bd70c6SFilipe Manana int ret; 3286e0bd70c6SFilipe Manana 3287416161dbSMark Fasheh if (inode1 < inode2) { 3288416161dbSMark Fasheh swap(inode1, inode2); 3289416161dbSMark Fasheh swap(loff1, loff2); 3290416161dbSMark Fasheh } 3291e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 3292e0bd70c6SFilipe Manana if (ret) 3293e0bd70c6SFilipe Manana return ret; 3294e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 3295e0bd70c6SFilipe Manana if (ret) 3296e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 3297e0bd70c6SFilipe Manana loff1 + len - 1); 3298e0bd70c6SFilipe Manana return ret; 3299416161dbSMark Fasheh } 3300f4414602SMark Fasheh 3301f4414602SMark Fasheh struct cmp_pages { 3302f4414602SMark Fasheh int num_pages; 3303f4414602SMark Fasheh struct page **src_pages; 3304f4414602SMark Fasheh struct page **dst_pages; 3305f4414602SMark Fasheh }; 3306f4414602SMark Fasheh 3307f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 3308f4414602SMark Fasheh { 3309f4414602SMark Fasheh int i; 3310f4414602SMark Fasheh struct page *pg; 3311f4414602SMark Fasheh 3312f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 3313f4414602SMark Fasheh pg = cmp->src_pages[i]; 3314e0bd70c6SFilipe Manana if (pg) { 3315e0bd70c6SFilipe Manana unlock_page(pg); 331609cbfeafSKirill A. Shutemov put_page(pg); 331797b19170SNaohiro Aota cmp->src_pages[i] = NULL; 3318e0bd70c6SFilipe Manana } 3319f4414602SMark Fasheh pg = cmp->dst_pages[i]; 3320e0bd70c6SFilipe Manana if (pg) { 3321e0bd70c6SFilipe Manana unlock_page(pg); 332209cbfeafSKirill A. Shutemov put_page(pg); 332397b19170SNaohiro Aota cmp->dst_pages[i] = NULL; 3324f4414602SMark Fasheh } 3325e0bd70c6SFilipe Manana } 3326f4414602SMark Fasheh } 3327f4414602SMark Fasheh 3328f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 3329f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 3330f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 3331f4414602SMark Fasheh { 3332f4414602SMark Fasheh int ret; 333309cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 3334f4414602SMark Fasheh 3335f4414602SMark Fasheh cmp->num_pages = num_pages; 3336f4414602SMark Fasheh 333767b07bd4STimofey Titovets ret = gather_extent_pages(src, cmp->src_pages, num_pages, loff); 3338f4414602SMark Fasheh if (ret) 3339f4414602SMark Fasheh goto out; 3340f4414602SMark Fasheh 334167b07bd4STimofey Titovets ret = gather_extent_pages(dst, cmp->dst_pages, num_pages, dst_loff); 3342f4414602SMark Fasheh 3343f4414602SMark Fasheh out: 3344f4414602SMark Fasheh if (ret) 3345f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 334678ad4ce0SNaohiro Aota return ret; 3347416161dbSMark Fasheh } 3348416161dbSMark Fasheh 33491a287cfeSDavid Sterba static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp) 3350416161dbSMark Fasheh { 3351416161dbSMark Fasheh int ret = 0; 3352f4414602SMark Fasheh int i; 3353416161dbSMark Fasheh struct page *src_page, *dst_page; 335409cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 3355416161dbSMark Fasheh void *addr, *dst_addr; 3356416161dbSMark Fasheh 3357f4414602SMark Fasheh i = 0; 3358416161dbSMark Fasheh while (len) { 335909cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 3360416161dbSMark Fasheh cmp_len = len; 3361416161dbSMark Fasheh 3362f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 3363f4414602SMark Fasheh 3364f4414602SMark Fasheh src_page = cmp->src_pages[i]; 3365f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 3366e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 3367e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 3368f4414602SMark Fasheh 3369416161dbSMark Fasheh addr = kmap_atomic(src_page); 3370416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 3371416161dbSMark Fasheh 3372416161dbSMark Fasheh flush_dcache_page(src_page); 3373416161dbSMark Fasheh flush_dcache_page(dst_page); 3374416161dbSMark Fasheh 3375416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 33762b3909f8SDarrick J. Wong ret = -EBADE; 3377416161dbSMark Fasheh 3378416161dbSMark Fasheh kunmap_atomic(addr); 3379416161dbSMark Fasheh kunmap_atomic(dst_addr); 3380416161dbSMark Fasheh 3381416161dbSMark Fasheh if (ret) 3382416161dbSMark Fasheh break; 3383416161dbSMark Fasheh 3384416161dbSMark Fasheh len -= cmp_len; 3385f4414602SMark Fasheh i++; 3386416161dbSMark Fasheh } 3387416161dbSMark Fasheh 3388416161dbSMark Fasheh return ret; 3389416161dbSMark Fasheh } 3390416161dbSMark Fasheh 3391e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3392e1d227a4SMark Fasheh u64 olen) 3393416161dbSMark Fasheh { 3394e1d227a4SMark Fasheh u64 len = *plen; 3395416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3396416161dbSMark Fasheh 3397e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 3398416161dbSMark Fasheh return -EINVAL; 3399e1d227a4SMark Fasheh 3400e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 3401e1d227a4SMark Fasheh if (off + len == inode->i_size) 3402e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 3403e1d227a4SMark Fasheh 3404416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 3405416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3406416161dbSMark Fasheh return -EINVAL; 3407416161dbSMark Fasheh 3408416161dbSMark Fasheh return 0; 3409416161dbSMark Fasheh } 3410416161dbSMark Fasheh 34113973909dSTimofey Titovets static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 olen, 341267b07bd4STimofey Titovets struct inode *dst, u64 dst_loff, 341367b07bd4STimofey Titovets struct cmp_pages *cmp) 3414416161dbSMark Fasheh { 3415416161dbSMark Fasheh int ret; 3416e1d227a4SMark Fasheh u64 len = olen; 3417fc4badd9SOmar Sandoval bool same_inode = (src == dst); 34180efa9f48SMark Fasheh u64 same_lock_start = 0; 34190efa9f48SMark Fasheh u64 same_lock_len = 0; 3420416161dbSMark Fasheh 34210efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 34220efa9f48SMark Fasheh if (ret) 34233973909dSTimofey Titovets return ret; 3424fc4badd9SOmar Sandoval 3425fc4badd9SOmar Sandoval ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3426f4dfe687SFilipe Manana if (ret) 34273973909dSTimofey Titovets return ret; 34280efa9f48SMark Fasheh 3429fc4badd9SOmar Sandoval if (same_inode) { 34300efa9f48SMark Fasheh /* 34310efa9f48SMark Fasheh * Single inode case wants the same checks, except we 34320efa9f48SMark Fasheh * don't want our length pushed out past i_size as 34330efa9f48SMark Fasheh * comparing that data range makes no sense. 34340efa9f48SMark Fasheh * 34350efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 34360efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 34370efa9f48SMark Fasheh * reject the request. 34380efa9f48SMark Fasheh * 34390efa9f48SMark Fasheh * This effectively means we require aligned extents 34400efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 34410efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 34420efa9f48SMark Fasheh * i_size. 34430efa9f48SMark Fasheh */ 34443973909dSTimofey Titovets if (len != olen) 34453973909dSTimofey Titovets return -EINVAL; 34460efa9f48SMark Fasheh 34470efa9f48SMark Fasheh /* Check for overlapping ranges */ 34483973909dSTimofey Titovets if (dst_loff + len > loff && dst_loff < loff + len) 34493973909dSTimofey Titovets return -EINVAL; 34500efa9f48SMark Fasheh 34510efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 34520efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 34530efa9f48SMark Fasheh } 3454416161dbSMark Fasheh 3455e0bd70c6SFilipe Manana again: 345667b07bd4STimofey Titovets ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, cmp); 3457f4414602SMark Fasheh if (ret) 34583973909dSTimofey Titovets return ret; 3459f4414602SMark Fasheh 34600efa9f48SMark Fasheh if (same_inode) 3461e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3462e0bd70c6SFilipe Manana false); 34630efa9f48SMark Fasheh else 3464e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3465e0bd70c6SFilipe Manana false); 3466e0bd70c6SFilipe Manana /* 3467e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3468e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3469e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3470e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3471e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3472e0bd70c6SFilipe Manana * range locking). 3473e0bd70c6SFilipe Manana */ 3474e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3475e0bd70c6SFilipe Manana /* 3476e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3477e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3478e0bd70c6SFilipe Manana */ 347967b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 3480e0bd70c6SFilipe Manana if (same_inode) { 3481e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3482e0bd70c6SFilipe Manana same_lock_len); 3483e0bd70c6SFilipe Manana } else { 3484e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3485e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3486e0bd70c6SFilipe Manana } 3487e0bd70c6SFilipe Manana goto again; 3488e0bd70c6SFilipe Manana } 3489e0bd70c6SFilipe Manana ASSERT(ret == 0); 3490e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3491e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 349267b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 3493e0bd70c6SFilipe Manana return ret; 3494e0bd70c6SFilipe Manana } 3495f4414602SMark Fasheh 3496207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 349767b07bd4STimofey Titovets ret = btrfs_cmp_data(olen, cmp); 3498416161dbSMark Fasheh if (ret == 0) 34991c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3500416161dbSMark Fasheh 35010efa9f48SMark Fasheh if (same_inode) 35020efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 35030efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 35040efa9f48SMark Fasheh else 3505f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3506f4414602SMark Fasheh 350767b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 35083973909dSTimofey Titovets 35093973909dSTimofey Titovets return ret; 35103973909dSTimofey Titovets } 35113973909dSTimofey Titovets 3512b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3513b6728768STimofey Titovets 35143973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 35153973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 35163973909dSTimofey Titovets { 35173973909dSTimofey Titovets int ret; 351867b07bd4STimofey Titovets struct cmp_pages cmp; 351967b07bd4STimofey Titovets int num_pages = PAGE_ALIGN(BTRFS_MAX_DEDUPE_LEN) >> PAGE_SHIFT; 35203973909dSTimofey Titovets bool same_inode = (src == dst); 3521b6728768STimofey Titovets u64 i, tail_len, chunk_count; 35223973909dSTimofey Titovets 35233973909dSTimofey Titovets if (olen == 0) 35243973909dSTimofey Titovets return 0; 35253973909dSTimofey Titovets 35263973909dSTimofey Titovets if (same_inode) 35273973909dSTimofey Titovets inode_lock(src); 35283973909dSTimofey Titovets else 35293973909dSTimofey Titovets btrfs_double_inode_lock(src, dst); 35303973909dSTimofey Titovets 35313973909dSTimofey Titovets /* don't make the dst file partly checksummed */ 35323973909dSTimofey Titovets if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 35333973909dSTimofey Titovets (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 35343973909dSTimofey Titovets ret = -EINVAL; 35353973909dSTimofey Titovets goto out_unlock; 35363973909dSTimofey Titovets } 35373973909dSTimofey Titovets 3538b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3539b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 354067b07bd4STimofey Titovets if (chunk_count == 0) 354167b07bd4STimofey Titovets num_pages = PAGE_ALIGN(tail_len) >> PAGE_SHIFT; 354267b07bd4STimofey Titovets 354367b07bd4STimofey Titovets /* 354467b07bd4STimofey Titovets * If deduping ranges in the same inode, locking rules make it 354567b07bd4STimofey Titovets * mandatory to always lock pages in ascending order to avoid deadlocks 354667b07bd4STimofey Titovets * with concurrent tasks (such as starting writeback/delalloc). 354767b07bd4STimofey Titovets */ 354867b07bd4STimofey Titovets if (same_inode && dst_loff < loff) 354967b07bd4STimofey Titovets swap(loff, dst_loff); 355067b07bd4STimofey Titovets 355167b07bd4STimofey Titovets /* 355267b07bd4STimofey Titovets * We must gather up all the pages before we initiate our extent 355367b07bd4STimofey Titovets * locking. We use an array for the page pointers. Size of the array is 355467b07bd4STimofey Titovets * bounded by len, which is in turn bounded by BTRFS_MAX_DEDUPE_LEN. 355567b07bd4STimofey Titovets */ 3556bf5091c8SDavid Sterba cmp.src_pages = kvmalloc_array(num_pages, sizeof(struct page *), 3557bf5091c8SDavid Sterba GFP_KERNEL | __GFP_ZERO); 3558bf5091c8SDavid Sterba cmp.dst_pages = kvmalloc_array(num_pages, sizeof(struct page *), 3559bf5091c8SDavid Sterba GFP_KERNEL | __GFP_ZERO); 356067b07bd4STimofey Titovets if (!cmp.src_pages || !cmp.dst_pages) { 3561bf5091c8SDavid Sterba ret = -ENOMEM; 3562bf5091c8SDavid Sterba goto out_free; 356367b07bd4STimofey Titovets } 3564b6728768STimofey Titovets 3565b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3566b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 356767b07bd4STimofey Titovets dst, dst_loff, &cmp); 3568b6728768STimofey Titovets if (ret) 356922883ddcSLu Fengqi goto out_free; 3570b6728768STimofey Titovets 3571b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3572b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3573b6728768STimofey Titovets } 3574b6728768STimofey Titovets 3575b6728768STimofey Titovets if (tail_len > 0) 357667b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 357767b07bd4STimofey Titovets dst_loff, &cmp); 35783973909dSTimofey Titovets 357922883ddcSLu Fengqi out_free: 358022883ddcSLu Fengqi kvfree(cmp.src_pages); 358122883ddcSLu Fengqi kvfree(cmp.dst_pages); 358222883ddcSLu Fengqi 3583416161dbSMark Fasheh out_unlock: 35840efa9f48SMark Fasheh if (same_inode) 35855955102cSAl Viro inode_unlock(src); 35860efa9f48SMark Fasheh else 3587f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3588416161dbSMark Fasheh 3589416161dbSMark Fasheh return ret; 3590416161dbSMark Fasheh } 3591416161dbSMark Fasheh 35922b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 35932b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3594416161dbSMark Fasheh { 35952b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 35962b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3597416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 35982b3909f8SDarrick J. Wong ssize_t res; 3599416161dbSMark Fasheh 360009cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3601416161dbSMark Fasheh /* 3602416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3603416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3604416161dbSMark Fasheh * this situation without an update. 3605416161dbSMark Fasheh */ 36062b3909f8SDarrick J. Wong return -EINVAL; 3607416161dbSMark Fasheh } 3608416161dbSMark Fasheh 36092b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 36102b3909f8SDarrick J. Wong if (res) 36112b3909f8SDarrick J. Wong return res; 36122b3909f8SDarrick J. Wong return olen; 3613416161dbSMark Fasheh } 3614416161dbSMark Fasheh 3615f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3616f82a9901SFilipe Manana struct inode *inode, 3617f82a9901SFilipe Manana u64 endoff, 3618f82a9901SFilipe Manana const u64 destoff, 36191c919a5eSMark Fasheh const u64 olen, 36201c919a5eSMark Fasheh int no_time_update) 3621f82a9901SFilipe Manana { 3622f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3623f82a9901SFilipe Manana int ret; 3624f82a9901SFilipe Manana 3625f82a9901SFilipe Manana inode_inc_iversion(inode); 36261c919a5eSMark Fasheh if (!no_time_update) 3627c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3628f82a9901SFilipe Manana /* 3629f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3630f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3631f82a9901SFilipe Manana */ 3632f82a9901SFilipe Manana if (endoff > destoff + olen) 3633f82a9901SFilipe Manana endoff = destoff + olen; 3634f82a9901SFilipe Manana if (endoff > inode->i_size) 36356ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3636f82a9901SFilipe Manana 3637f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3638f82a9901SFilipe Manana if (ret) { 363966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 36403a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3641f82a9901SFilipe Manana goto out; 3642f82a9901SFilipe Manana } 36433a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3644f82a9901SFilipe Manana out: 3645f82a9901SFilipe Manana return ret; 3646f82a9901SFilipe Manana } 3647f82a9901SFilipe Manana 3648a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 36497ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 36507ffbb598SFilipe Manana const struct btrfs_path *path, 36517ffbb598SFilipe Manana const u64 hole_offset, 36527ffbb598SFilipe Manana const u64 hole_len) 36537ffbb598SFilipe Manana { 3654a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 36557ffbb598SFilipe Manana struct extent_map *em; 36567ffbb598SFilipe Manana int ret; 36577ffbb598SFilipe Manana 36587ffbb598SFilipe Manana em = alloc_extent_map(); 36597ffbb598SFilipe Manana if (!em) { 3660a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 36617ffbb598SFilipe Manana return; 36627ffbb598SFilipe Manana } 36637ffbb598SFilipe Manana 366414f59796SFilipe Manana if (path) { 366514f59796SFilipe Manana struct btrfs_file_extent_item *fi; 366614f59796SFilipe Manana 366714f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 366814f59796SFilipe Manana struct btrfs_file_extent_item); 3669a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 36707ffbb598SFilipe Manana em->generation = -1; 36717ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 36727ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 36737ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3674a2f392e4SNikolay Borisov &inode->runtime_flags); 36757ffbb598SFilipe Manana } else { 36767ffbb598SFilipe Manana em->start = hole_offset; 36777ffbb598SFilipe Manana em->len = hole_len; 36787ffbb598SFilipe Manana em->ram_bytes = em->len; 36797ffbb598SFilipe Manana em->orig_start = hole_offset; 36807ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 36817ffbb598SFilipe Manana em->block_len = 0; 36827ffbb598SFilipe Manana em->orig_block_len = 0; 36837ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 36847ffbb598SFilipe Manana em->generation = trans->transid; 36857ffbb598SFilipe Manana } 36867ffbb598SFilipe Manana 36877ffbb598SFilipe Manana while (1) { 36887ffbb598SFilipe Manana write_lock(&em_tree->lock); 36897ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 36907ffbb598SFilipe Manana write_unlock(&em_tree->lock); 36917ffbb598SFilipe Manana if (ret != -EEXIST) { 36927ffbb598SFilipe Manana free_extent_map(em); 36937ffbb598SFilipe Manana break; 36947ffbb598SFilipe Manana } 3695a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 36967ffbb598SFilipe Manana em->start + em->len - 1, 0); 36977ffbb598SFilipe Manana } 36987ffbb598SFilipe Manana 3699ee39b432SDavid Sterba if (ret) 3700a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 37017ffbb598SFilipe Manana } 37027ffbb598SFilipe Manana 37038039d87dSFilipe Manana /* 37048039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 37058039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 37068039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 37078039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 37088039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 37098039d87dSFilipe Manana * 37108039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 37118039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 37128039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 37138039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 37148039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 37158039d87dSFilipe Manana * i_mutex, which we are holding here). 37168039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 37178039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 37188039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 37198039d87dSFilipe Manana * 37208039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 37218039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 37228039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 37238039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 37248039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 37258039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 37268039d87dSFilipe Manana * the former. 37278039d87dSFilipe Manana */ 37284a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 37298039d87dSFilipe Manana struct btrfs_trans_handle *trans, 37308039d87dSFilipe Manana struct btrfs_path *path, 37318039d87dSFilipe Manana struct btrfs_key *new_key, 37328039d87dSFilipe Manana const u64 drop_start, 37338039d87dSFilipe Manana const u64 datal, 37348039d87dSFilipe Manana const u64 skip, 37358039d87dSFilipe Manana const u64 size, 37368039d87dSFilipe Manana char *inline_data) 37378039d87dSFilipe Manana { 37380b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 37398039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 37408039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 37410b246afaSJeff Mahoney fs_info->sectorsize); 37428039d87dSFilipe Manana int ret; 37438039d87dSFilipe Manana struct btrfs_key key; 37448039d87dSFilipe Manana 37458039d87dSFilipe Manana if (new_key->offset > 0) 37468039d87dSFilipe Manana return -EOPNOTSUPP; 37478039d87dSFilipe Manana 37484a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 37498039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 37508039d87dSFilipe Manana key.offset = 0; 37518039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 37528039d87dSFilipe Manana if (ret < 0) { 37538039d87dSFilipe Manana return ret; 37548039d87dSFilipe Manana } else if (ret > 0) { 37558039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 37568039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 37578039d87dSFilipe Manana if (ret < 0) 37588039d87dSFilipe Manana return ret; 37598039d87dSFilipe Manana else if (ret > 0) 37608039d87dSFilipe Manana goto copy_inline_extent; 37618039d87dSFilipe Manana } 37628039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 37634a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 37648039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 37658039d87dSFilipe Manana ASSERT(key.offset > 0); 37668039d87dSFilipe Manana return -EOPNOTSUPP; 37678039d87dSFilipe Manana } 37688039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 37698039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 37708039d87dSFilipe Manana u64 ext_len; 37718039d87dSFilipe Manana 37728039d87dSFilipe Manana /* 37738039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 37748039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 37758039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 37768039d87dSFilipe Manana */ 37778039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 37788039d87dSFilipe Manana struct btrfs_file_extent_item); 37798039d87dSFilipe Manana /* 37808039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 37818039d87dSFilipe Manana * following it. 37828039d87dSFilipe Manana */ 37838039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 37848039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 37858039d87dSFilipe Manana goto copy_inline_extent; 37868039d87dSFilipe Manana 37878039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 37888039d87dSFilipe Manana if (ext_len > aligned_end) 37898039d87dSFilipe Manana return -EOPNOTSUPP; 37908039d87dSFilipe Manana 37918039d87dSFilipe Manana ret = btrfs_next_item(root, path); 37928039d87dSFilipe Manana if (ret < 0) { 37938039d87dSFilipe Manana return ret; 37948039d87dSFilipe Manana } else if (ret == 0) { 37958039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 37968039d87dSFilipe Manana path->slots[0]); 37974a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 37988039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 37998039d87dSFilipe Manana return -EOPNOTSUPP; 38008039d87dSFilipe Manana } 38018039d87dSFilipe Manana } 38028039d87dSFilipe Manana 38038039d87dSFilipe Manana copy_inline_extent: 38048039d87dSFilipe Manana /* 38058039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 38068039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 38078039d87dSFilipe Manana */ 38088039d87dSFilipe Manana if (i_size_read(dst) > datal) { 38098039d87dSFilipe Manana /* 38108039d87dSFilipe Manana * If the destination inode has an inline extent... 38118039d87dSFilipe Manana * This would require copying the data from the source inline 38128039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 38138039d87dSFilipe Manana * But this is really complex, both extents can be compressed 38148039d87dSFilipe Manana * or just one of them, which would require decompressing and 38158039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 38168039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 38178039d87dSFilipe Manana * inline extent). 38188039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 38198039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 38208039d87dSFilipe Manana */ 38218039d87dSFilipe Manana return -EOPNOTSUPP; 38228039d87dSFilipe Manana } 38238039d87dSFilipe Manana 38248039d87dSFilipe Manana btrfs_release_path(path); 38258039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 38268039d87dSFilipe Manana if (ret) 38278039d87dSFilipe Manana return ret; 38288039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 38298039d87dSFilipe Manana if (ret) 38308039d87dSFilipe Manana return ret; 38318039d87dSFilipe Manana 38328039d87dSFilipe Manana if (skip) { 38338039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 38348039d87dSFilipe Manana 38358039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 38368039d87dSFilipe Manana } 38378039d87dSFilipe Manana 38388039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 38398039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 38408039d87dSFilipe Manana path->slots[0]), 38418039d87dSFilipe Manana size); 38428039d87dSFilipe Manana inode_add_bytes(dst, datal); 38438039d87dSFilipe Manana 38448039d87dSFilipe Manana return 0; 38458039d87dSFilipe Manana } 38468039d87dSFilipe Manana 384732b7c687SMark Fasheh /** 384832b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 384932b7c687SMark Fasheh * 385032b7c687SMark Fasheh * @src: Inode to clone from 385132b7c687SMark Fasheh * @inode: Inode to clone to 385232b7c687SMark Fasheh * @off: Offset within source to start clone from 385332b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 38541c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 385532b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 38561c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 385732b7c687SMark Fasheh */ 385832b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3859f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 38601c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3861f46b5a66SChristoph Hellwig { 38620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3863f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 386432b7c687SMark Fasheh struct btrfs_path *path = NULL; 3865f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 386632b7c687SMark Fasheh struct btrfs_trans_handle *trans; 386732b7c687SMark Fasheh char *buf = NULL; 3868ae01a0abSYan Zheng struct btrfs_key key; 3869f46b5a66SChristoph Hellwig u32 nritems; 3870f46b5a66SChristoph Hellwig int slot; 3871ae01a0abSYan Zheng int ret; 3872f82a9901SFilipe Manana const u64 len = olen_aligned; 3873f82a9901SFilipe Manana u64 last_dest_end = destoff; 3874ae01a0abSYan Zheng 3875ae01a0abSYan Zheng ret = -ENOMEM; 3876752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3877ae01a0abSYan Zheng if (!buf) 387832b7c687SMark Fasheh return ret; 3879ae01a0abSYan Zheng 3880ae01a0abSYan Zheng path = btrfs_alloc_path(); 3881ae01a0abSYan Zheng if (!path) { 388215351955SDavid Sterba kvfree(buf); 388332b7c687SMark Fasheh return ret; 3884ae01a0abSYan Zheng } 388532b7c687SMark Fasheh 3886e4058b54SDavid Sterba path->reada = READA_FORWARD; 3887c5c9cd4dSSage Weil /* clone data */ 38884a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3889ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 38902c463823SFilipe Manana key.offset = off; 3891f46b5a66SChristoph Hellwig 3892f46b5a66SChristoph Hellwig while (1) { 3893de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3894df858e76SFilipe Manana 3895f46b5a66SChristoph Hellwig /* 3896f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3897f46b5a66SChristoph Hellwig * tree. 3898f46b5a66SChristoph Hellwig */ 3899e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3900362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3901362a20c5SDavid Sterba 0, 0); 3902f46b5a66SChristoph Hellwig if (ret < 0) 3903f46b5a66SChristoph Hellwig goto out; 39042c463823SFilipe Manana /* 39052c463823SFilipe Manana * First search, if no extent item that starts at offset off was 39062c463823SFilipe Manana * found but the previous item is an extent item, it's possible 39072c463823SFilipe Manana * it might overlap our target range, therefore process it. 39082c463823SFilipe Manana */ 39092c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 39102c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 39112c463823SFilipe Manana path->slots[0] - 1); 39122c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 39132c463823SFilipe Manana path->slots[0]--; 39142c463823SFilipe Manana } 3915f46b5a66SChristoph Hellwig 3916ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3917e4355f34SFilipe David Borba Manana process_slot: 3918ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3919362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3920f46b5a66SChristoph Hellwig if (ret < 0) 3921f46b5a66SChristoph Hellwig goto out; 3922f46b5a66SChristoph Hellwig if (ret > 0) 3923f46b5a66SChristoph Hellwig break; 3924ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3925f46b5a66SChristoph Hellwig } 3926f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3927f46b5a66SChristoph Hellwig slot = path->slots[0]; 3928f46b5a66SChristoph Hellwig 3929ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3930962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 39314a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3932f46b5a66SChristoph Hellwig break; 3933f46b5a66SChristoph Hellwig 3934962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3935c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3936c5c9cd4dSSage Weil int type; 393731840ae1SZheng Yan u32 size; 393831840ae1SZheng Yan struct btrfs_key new_key; 3939c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3940c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3941c5c9cd4dSSage Weil u8 comp; 3942f82a9901SFilipe Manana u64 drop_start; 394331840ae1SZheng Yan 3944c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3945c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3946c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3947c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3948c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3949c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3950d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3951d397712bSChris Mason extent); 3952d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3953d397712bSChris Mason extent); 3954c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3955d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3956d397712bSChris Mason extent); 3957c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3958c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3959c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3960c5c9cd4dSSage Weil extent); 3961c5c9cd4dSSage Weil } 396231840ae1SZheng Yan 39632c463823SFilipe Manana /* 39642c463823SFilipe Manana * The first search might have left us at an extent 39652c463823SFilipe Manana * item that ends before our target range's start, can 39662c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 39672c463823SFilipe Manana */ 39682c463823SFilipe Manana if (key.offset + datal <= off) { 3969e4355f34SFilipe David Borba Manana path->slots[0]++; 3970e4355f34SFilipe David Borba Manana goto process_slot; 39712c463823SFilipe Manana } else if (key.offset >= off + len) { 39722c463823SFilipe Manana break; 3973e4355f34SFilipe David Borba Manana } 3974df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3975e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3976e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3977e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3978e4355f34SFilipe David Borba Manana size); 3979e4355f34SFilipe David Borba Manana 3980e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3981e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3982c5c9cd4dSSage Weil 398331840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 39844a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 39854d728ec7SLi Zefan if (off <= key.offset) 3986c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 39874d728ec7SLi Zefan else 39884d728ec7SLi Zefan new_key.offset = destoff; 3989c5c9cd4dSSage Weil 3990b6f3409bSSage Weil /* 3991f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3992f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3993f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3994f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3995f82a9901SFilipe Manana * partially overlaps it). 3996f82a9901SFilipe Manana */ 3997f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3998f82a9901SFilipe Manana drop_start = last_dest_end; 3999f82a9901SFilipe Manana else 4000f82a9901SFilipe Manana drop_start = new_key.offset; 4001f82a9901SFilipe Manana 4002f82a9901SFilipe Manana /* 4003b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 4004b6f3409bSSage Weil * 1 - add new extent 4005b6f3409bSSage Weil * 1 - inode update 4006b6f3409bSSage Weil */ 4007b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 4008a22285a6SYan, Zheng if (IS_ERR(trans)) { 4009a22285a6SYan, Zheng ret = PTR_ERR(trans); 4010a22285a6SYan, Zheng goto out; 4011a22285a6SYan, Zheng } 4012a22285a6SYan, Zheng 4013c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 4014c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 4015d72c0842SLi Zefan /* 4016d72c0842SLi Zefan * a | --- range to clone ---| b 4017d72c0842SLi Zefan * | ------------- extent ------------- | 4018d72c0842SLi Zefan */ 4019d72c0842SLi Zefan 402093915584SAntonio Ospite /* subtract range b */ 4021d72c0842SLi Zefan if (key.offset + datal > off + len) 4022d72c0842SLi Zefan datal = off + len - key.offset; 4023d72c0842SLi Zefan 402493915584SAntonio Ospite /* subtract range a */ 4025a22285a6SYan, Zheng if (off > key.offset) { 4026a22285a6SYan, Zheng datao += off - key.offset; 4027a22285a6SYan, Zheng datal -= off - key.offset; 4028a22285a6SYan, Zheng } 4029a22285a6SYan, Zheng 40305dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 4031f82a9901SFilipe Manana drop_start, 4032a22285a6SYan, Zheng new_key.offset + datal, 40332671485dSJosef Bacik 1); 403479787eaaSJeff Mahoney if (ret) { 40353f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 403600fdf13aSLiu Bo btrfs_abort_transaction(trans, 403766642832SJeff Mahoney ret); 40383a45bb20SJeff Mahoney btrfs_end_transaction(trans); 403979787eaaSJeff Mahoney goto out; 404079787eaaSJeff Mahoney } 4041a22285a6SYan, Zheng 404231840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 404331840ae1SZheng Yan &new_key, size); 404479787eaaSJeff Mahoney if (ret) { 404566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 40463a45bb20SJeff Mahoney btrfs_end_transaction(trans); 404779787eaaSJeff Mahoney goto out; 404879787eaaSJeff Mahoney } 404931840ae1SZheng Yan 405031840ae1SZheng Yan leaf = path->nodes[0]; 405131840ae1SZheng Yan slot = path->slots[0]; 405231840ae1SZheng Yan write_extent_buffer(leaf, buf, 405331840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 405431840ae1SZheng Yan size); 4055ae01a0abSYan Zheng 4056f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 4057f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 4058c5c9cd4dSSage Weil 4059c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 4060c5c9cd4dSSage Weil if (!disko) 4061c5c9cd4dSSage Weil datao = 0; 4062c5c9cd4dSSage Weil 4063c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 4064c5c9cd4dSSage Weil datao); 4065c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 4066c5c9cd4dSSage Weil datal); 4067fcebe456SJosef Bacik 4068c5c9cd4dSSage Weil if (disko) { 4069c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 40702ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 407184f7d8e6SJosef Bacik root, 40725d4f98a2SYan Zheng disko, diskl, 0, 4073f46b5a66SChristoph Hellwig root->root_key.objectid, 40744a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 4075b06c4bf5SFilipe Manana new_key.offset - datao); 407679787eaaSJeff Mahoney if (ret) { 407779787eaaSJeff Mahoney btrfs_abort_transaction(trans, 407879787eaaSJeff Mahoney ret); 40793a45bb20SJeff Mahoney btrfs_end_transaction(trans); 408079787eaaSJeff Mahoney goto out; 408179787eaaSJeff Mahoney 408279787eaaSJeff Mahoney } 4083f46b5a66SChristoph Hellwig } 4084c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 4085c5c9cd4dSSage Weil u64 skip = 0; 4086c5c9cd4dSSage Weil u64 trim = 0; 4087ed958762SFilipe Manana 4088c5c9cd4dSSage Weil if (off > key.offset) { 4089c5c9cd4dSSage Weil skip = off - key.offset; 4090c5c9cd4dSSage Weil new_key.offset += skip; 409131840ae1SZheng Yan } 4092d397712bSChris Mason 4093c5c9cd4dSSage Weil if (key.offset + datal > off + len) 4094c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 4095d397712bSChris Mason 4096c5c9cd4dSSage Weil if (comp && (skip || trim)) { 4097c5c9cd4dSSage Weil ret = -EINVAL; 40983a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4099c5c9cd4dSSage Weil goto out; 410031840ae1SZheng Yan } 4101c5c9cd4dSSage Weil size -= skip + trim; 4102c5c9cd4dSSage Weil datal -= skip + trim; 4103a22285a6SYan, Zheng 41044a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 41058039d87dSFilipe Manana trans, path, 41068039d87dSFilipe Manana &new_key, 4107f82a9901SFilipe Manana drop_start, 41088039d87dSFilipe Manana datal, 41098039d87dSFilipe Manana skip, size, buf); 411079787eaaSJeff Mahoney if (ret) { 41113f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 41123a29bc09SChris Mason btrfs_abort_transaction(trans, 411379787eaaSJeff Mahoney ret); 41143a45bb20SJeff Mahoney btrfs_end_transaction(trans); 411579787eaaSJeff Mahoney goto out; 411679787eaaSJeff Mahoney } 4117c5c9cd4dSSage Weil leaf = path->nodes[0]; 4118c5c9cd4dSSage Weil slot = path->slots[0]; 4119c5c9cd4dSSage Weil } 4120c5c9cd4dSSage Weil 41217ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 41227ffbb598SFilipe Manana if (drop_start < new_key.offset) 4123a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 412414f59796SFilipe Manana NULL, drop_start, 41257ffbb598SFilipe Manana new_key.offset - drop_start); 41267ffbb598SFilipe Manana 4127a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 4128a2f392e4SNikolay Borisov path, 0, 0); 41297ffbb598SFilipe Manana 4130c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 4131b3b4aa74SDavid Sterba btrfs_release_path(path); 4132c5c9cd4dSSage Weil 413362e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 41340b246afaSJeff Mahoney fs_info->sectorsize); 4135f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 4136f82a9901SFilipe Manana last_dest_end, 41371c919a5eSMark Fasheh destoff, olen, 41381c919a5eSMark Fasheh no_time_update); 4139f82a9901SFilipe Manana if (ret) 414079787eaaSJeff Mahoney goto out; 41412c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 41422c463823SFilipe Manana break; 4143a22285a6SYan, Zheng } 4144b3b4aa74SDavid Sterba btrfs_release_path(path); 4145df858e76SFilipe Manana key.offset = next_key_min_offset; 414669ae5e44SWang Xiaoguang 414769ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 414869ae5e44SWang Xiaoguang ret = -EINTR; 414969ae5e44SWang Xiaoguang goto out; 415069ae5e44SWang Xiaoguang } 4151ae01a0abSYan Zheng } 4152f46b5a66SChristoph Hellwig ret = 0; 415332b7c687SMark Fasheh 4154f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 4155f82a9901SFilipe Manana /* 4156f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 4157f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 4158f82a9901SFilipe Manana */ 4159b3b4aa74SDavid Sterba btrfs_release_path(path); 4160f82a9901SFilipe Manana 4161f82a9901SFilipe Manana /* 4162f82a9901SFilipe Manana * 1 - remove extent(s) 4163f82a9901SFilipe Manana * 1 - inode update 4164f82a9901SFilipe Manana */ 4165f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 4166f82a9901SFilipe Manana if (IS_ERR(trans)) { 4167f82a9901SFilipe Manana ret = PTR_ERR(trans); 4168f82a9901SFilipe Manana goto out; 4169f82a9901SFilipe Manana } 4170f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 4171f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 4172f82a9901SFilipe Manana if (ret) { 4173f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 417466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 41753a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4176f82a9901SFilipe Manana goto out; 4177f82a9901SFilipe Manana } 4178a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 4179a2f392e4SNikolay Borisov last_dest_end, 418014f59796SFilipe Manana destoff + len - last_dest_end); 4181f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 41821c919a5eSMark Fasheh destoff, olen, no_time_update); 4183f82a9901SFilipe Manana } 4184f82a9901SFilipe Manana 4185f46b5a66SChristoph Hellwig out: 418632b7c687SMark Fasheh btrfs_free_path(path); 418715351955SDavid Sterba kvfree(buf); 418832b7c687SMark Fasheh return ret; 418932b7c687SMark Fasheh } 419032b7c687SMark Fasheh 41913db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 419232b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 419332b7c687SMark Fasheh { 419454563d41SAl Viro struct inode *inode = file_inode(file); 41953db11b2eSZach Brown struct inode *src = file_inode(file_src); 41960b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 419732b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 419832b7c687SMark Fasheh int ret; 419932b7c687SMark Fasheh u64 len = olen; 42000b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 42013db11b2eSZach Brown int same_inode = src == inode; 420232b7c687SMark Fasheh 420332b7c687SMark Fasheh /* 420432b7c687SMark Fasheh * TODO: 420532b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 420632b7c687SMark Fasheh * decompress into destination's address_space (the file offset 420732b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 420832b7c687SMark Fasheh * otherwise reinsert) a subrange. 420900fdf13aSLiu Bo * 421000fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 421100fdf13aSLiu Bo * be either compressed or non-compressed. 421232b7c687SMark Fasheh */ 421332b7c687SMark Fasheh 421432b7c687SMark Fasheh if (btrfs_root_readonly(root)) 421532b7c687SMark Fasheh return -EROFS; 421632b7c687SMark Fasheh 42173db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 42183db11b2eSZach Brown src->i_sb != inode->i_sb) 42193db11b2eSZach Brown return -EXDEV; 422032b7c687SMark Fasheh 422132b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 42223db11b2eSZach Brown return -EISDIR; 422332b7c687SMark Fasheh 422432b7c687SMark Fasheh if (!same_inode) { 4225293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 422632b7c687SMark Fasheh } else { 42275955102cSAl Viro inode_lock(src); 422832b7c687SMark Fasheh } 422932b7c687SMark Fasheh 4230b5c40d59SOmar Sandoval /* don't make the dst file partly checksummed */ 4231b5c40d59SOmar Sandoval if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 4232b5c40d59SOmar Sandoval (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) { 4233b5c40d59SOmar Sandoval ret = -EINVAL; 4234b5c40d59SOmar Sandoval goto out_unlock; 4235b5c40d59SOmar Sandoval } 4236b5c40d59SOmar Sandoval 423732b7c687SMark Fasheh /* determine range to clone */ 423832b7c687SMark Fasheh ret = -EINVAL; 423932b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 424032b7c687SMark Fasheh goto out_unlock; 424132b7c687SMark Fasheh if (len == 0) 424232b7c687SMark Fasheh olen = len = src->i_size - off; 424332b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 424432b7c687SMark Fasheh if (off + len == src->i_size) 424532b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 424632b7c687SMark Fasheh 4247ccccf3d6SFilipe Manana if (len == 0) { 4248ccccf3d6SFilipe Manana ret = 0; 4249ccccf3d6SFilipe Manana goto out_unlock; 4250ccccf3d6SFilipe Manana } 4251ccccf3d6SFilipe Manana 425232b7c687SMark Fasheh /* verify the end result is block aligned */ 425332b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 425432b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 425532b7c687SMark Fasheh goto out_unlock; 425632b7c687SMark Fasheh 425732b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 425832b7c687SMark Fasheh if (same_inode) { 425932b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 426032b7c687SMark Fasheh goto out_unlock; 426132b7c687SMark Fasheh } 426232b7c687SMark Fasheh 426332b7c687SMark Fasheh if (destoff > inode->i_size) { 426432b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 426532b7c687SMark Fasheh if (ret) 426632b7c687SMark Fasheh goto out_unlock; 426732b7c687SMark Fasheh } 426832b7c687SMark Fasheh 4269c125b8bfSFilipe Manana /* 4270c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 4271c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 4272c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 4273c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 4274c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 4275c125b8bfSFilipe Manana */ 4276c125b8bfSFilipe Manana if (same_inode) { 4277c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 4278c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 427932b7c687SMark Fasheh 4280e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 4281c125b8bfSFilipe Manana } else { 4282e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 4283e0bd70c6SFilipe Manana true); 4284e0bd70c6SFilipe Manana } 4285e0bd70c6SFilipe Manana ASSERT(ret == 0); 4286e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 4287e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 4288e0bd70c6SFilipe Manana goto out_unlock; 4289c125b8bfSFilipe Manana } 429032b7c687SMark Fasheh 42911c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 429232b7c687SMark Fasheh 4293c125b8bfSFilipe Manana if (same_inode) { 4294c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 4295c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 4296c125b8bfSFilipe Manana 4297c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 4298c125b8bfSFilipe Manana } else { 4299293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 4300c125b8bfSFilipe Manana } 4301c125b8bfSFilipe Manana /* 4302c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 4303c125b8bfSFilipe Manana * data immediately and not the previous data. 4304c125b8bfSFilipe Manana */ 430565bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 430609cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 430709cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 4308f46b5a66SChristoph Hellwig out_unlock: 4309293a8489SMark Fasheh if (!same_inode) 4310293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 4311293a8489SMark Fasheh else 43125955102cSAl Viro inode_unlock(src); 43133db11b2eSZach Brown return ret; 43143db11b2eSZach Brown } 43153db11b2eSZach Brown 431604b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 431704b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 43183db11b2eSZach Brown { 431904b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 4320c5c9cd4dSSage Weil } 4321c5c9cd4dSSage Weil 43226ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 43236ef5ed0dSJosef Bacik { 4324496ad9aaSAl Viro struct inode *inode = file_inode(file); 43250b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 43266ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 43276ef5ed0dSJosef Bacik struct btrfs_root *new_root; 43286ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 43296ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 43306ef5ed0dSJosef Bacik struct btrfs_path *path; 43316ef5ed0dSJosef Bacik struct btrfs_key location; 43326ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 43336ef5ed0dSJosef Bacik u64 objectid = 0; 43346ef5ed0dSJosef Bacik u64 dir_id; 43353c04ce01SMiao Xie int ret; 43366ef5ed0dSJosef Bacik 43376ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 43386ef5ed0dSJosef Bacik return -EPERM; 43396ef5ed0dSJosef Bacik 43403c04ce01SMiao Xie ret = mnt_want_write_file(file); 43413c04ce01SMiao Xie if (ret) 43423c04ce01SMiao Xie return ret; 43433c04ce01SMiao Xie 43443c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 43453c04ce01SMiao Xie ret = -EFAULT; 43463c04ce01SMiao Xie goto out; 43473c04ce01SMiao Xie } 43486ef5ed0dSJosef Bacik 43496ef5ed0dSJosef Bacik if (!objectid) 43501cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 43516ef5ed0dSJosef Bacik 43526ef5ed0dSJosef Bacik location.objectid = objectid; 43536ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 43546ef5ed0dSJosef Bacik location.offset = (u64)-1; 43556ef5ed0dSJosef Bacik 43560b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 43573c04ce01SMiao Xie if (IS_ERR(new_root)) { 43583c04ce01SMiao Xie ret = PTR_ERR(new_root); 43593c04ce01SMiao Xie goto out; 43603c04ce01SMiao Xie } 43616d6d2829Ssatoru takeuchi if (!is_fstree(new_root->objectid)) { 43626d6d2829Ssatoru takeuchi ret = -ENOENT; 43636d6d2829Ssatoru takeuchi goto out; 43646d6d2829Ssatoru takeuchi } 43656ef5ed0dSJosef Bacik 43666ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 43673c04ce01SMiao Xie if (!path) { 43683c04ce01SMiao Xie ret = -ENOMEM; 43693c04ce01SMiao Xie goto out; 43703c04ce01SMiao Xie } 43716ef5ed0dSJosef Bacik path->leave_spinning = 1; 43726ef5ed0dSJosef Bacik 43736ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 437498d5dc13STsutomu Itoh if (IS_ERR(trans)) { 43756ef5ed0dSJosef Bacik btrfs_free_path(path); 43763c04ce01SMiao Xie ret = PTR_ERR(trans); 43773c04ce01SMiao Xie goto out; 43786ef5ed0dSJosef Bacik } 43796ef5ed0dSJosef Bacik 43800b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 43810b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 43826ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4383cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 43846ef5ed0dSJosef Bacik btrfs_free_path(path); 43853a45bb20SJeff Mahoney btrfs_end_transaction(trans); 43860b246afaSJeff Mahoney btrfs_err(fs_info, 43875d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 43883c04ce01SMiao Xie ret = -ENOENT; 43893c04ce01SMiao Xie goto out; 43906ef5ed0dSJosef Bacik } 43916ef5ed0dSJosef Bacik 43926ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 43936ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 43946ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 43956ef5ed0dSJosef Bacik btrfs_free_path(path); 43966ef5ed0dSJosef Bacik 43970b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 43983a45bb20SJeff Mahoney btrfs_end_transaction(trans); 43993c04ce01SMiao Xie out: 44003c04ce01SMiao Xie mnt_drop_write_file(file); 44013c04ce01SMiao Xie return ret; 44026ef5ed0dSJosef Bacik } 44036ef5ed0dSJosef Bacik 4404c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4405bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4406bf5fc093SJosef Bacik { 4407bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4408bf5fc093SJosef Bacik 4409bf5fc093SJosef Bacik space->total_bytes = 0; 4410bf5fc093SJosef Bacik space->used_bytes = 0; 4411bf5fc093SJosef Bacik space->flags = 0; 4412bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4413bf5fc093SJosef Bacik space->flags = block_group->flags; 4414bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4415bf5fc093SJosef Bacik space->used_bytes += 4416bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4417bf5fc093SJosef Bacik } 4418bf5fc093SJosef Bacik } 4419bf5fc093SJosef Bacik 44202ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 44212ff7e61eSJeff Mahoney void __user *arg) 44221406e432SJosef Bacik { 44231406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 44241406e432SJosef Bacik struct btrfs_ioctl_space_info space; 44251406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 44267fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 442713f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 44281406e432SJosef Bacik struct btrfs_space_info *info; 4429315d8e98SColin Ian King static const u64 types[] = { 4430315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4431bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4432bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4433315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4434315d8e98SColin Ian King }; 4435bf5fc093SJosef Bacik int num_types = 4; 44367fde62bfSChris Mason int alloc_size; 44371406e432SJosef Bacik int ret = 0; 443851788b1bSDan Rosenberg u64 slot_count = 0; 4439bf5fc093SJosef Bacik int i, c; 44401406e432SJosef Bacik 44411406e432SJosef Bacik if (copy_from_user(&space_args, 44421406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 44431406e432SJosef Bacik sizeof(space_args))) 44441406e432SJosef Bacik return -EFAULT; 44451406e432SJosef Bacik 4446bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4447bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4448bf5fc093SJosef Bacik 4449bf5fc093SJosef Bacik info = NULL; 44507fde62bfSChris Mason rcu_read_lock(); 44510b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4452bf5fc093SJosef Bacik list) { 4453bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4454bf5fc093SJosef Bacik info = tmp; 4455bf5fc093SJosef Bacik break; 4456bf5fc093SJosef Bacik } 4457bf5fc093SJosef Bacik } 44587fde62bfSChris Mason rcu_read_unlock(); 44591406e432SJosef Bacik 4460bf5fc093SJosef Bacik if (!info) 4461bf5fc093SJosef Bacik continue; 4462bf5fc093SJosef Bacik 4463bf5fc093SJosef Bacik down_read(&info->groups_sem); 4464bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4465bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4466bf5fc093SJosef Bacik slot_count++; 4467bf5fc093SJosef Bacik } 4468bf5fc093SJosef Bacik up_read(&info->groups_sem); 4469bf5fc093SJosef Bacik } 4470bf5fc093SJosef Bacik 447136523e95SDavid Sterba /* 447236523e95SDavid Sterba * Global block reserve, exported as a space_info 447336523e95SDavid Sterba */ 447436523e95SDavid Sterba slot_count++; 447536523e95SDavid Sterba 44767fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 44777fde62bfSChris Mason if (space_args.space_slots == 0) { 44787fde62bfSChris Mason space_args.total_spaces = slot_count; 44797fde62bfSChris Mason goto out; 44807fde62bfSChris Mason } 4481bf5fc093SJosef Bacik 448251788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4483bf5fc093SJosef Bacik 44847fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4485bf5fc093SJosef Bacik 44867fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 44877fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 44887fde62bfSChris Mason */ 448909cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 44907fde62bfSChris Mason return -ENOMEM; 44917fde62bfSChris Mason 44927fde62bfSChris Mason space_args.total_spaces = 0; 44938d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 44947fde62bfSChris Mason if (!dest) 44957fde62bfSChris Mason return -ENOMEM; 44967fde62bfSChris Mason dest_orig = dest; 44977fde62bfSChris Mason 44987fde62bfSChris Mason /* now we have a buffer to copy into */ 4499bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4500bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4501bf5fc093SJosef Bacik 450251788b1bSDan Rosenberg if (!slot_count) 450351788b1bSDan Rosenberg break; 450451788b1bSDan Rosenberg 4505bf5fc093SJosef Bacik info = NULL; 45061406e432SJosef Bacik rcu_read_lock(); 45070b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4508bf5fc093SJosef Bacik list) { 4509bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4510bf5fc093SJosef Bacik info = tmp; 45117fde62bfSChris Mason break; 4512bf5fc093SJosef Bacik } 4513bf5fc093SJosef Bacik } 4514bf5fc093SJosef Bacik rcu_read_unlock(); 45157fde62bfSChris Mason 4516bf5fc093SJosef Bacik if (!info) 4517bf5fc093SJosef Bacik continue; 4518bf5fc093SJosef Bacik down_read(&info->groups_sem); 4519bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4520bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4521c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4522c065f5b1SSu Yue &space); 45237fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 45241406e432SJosef Bacik dest++; 45251406e432SJosef Bacik space_args.total_spaces++; 452651788b1bSDan Rosenberg slot_count--; 45271406e432SJosef Bacik } 452851788b1bSDan Rosenberg if (!slot_count) 452951788b1bSDan Rosenberg break; 4530bf5fc093SJosef Bacik } 4531bf5fc093SJosef Bacik up_read(&info->groups_sem); 4532bf5fc093SJosef Bacik } 45331406e432SJosef Bacik 453436523e95SDavid Sterba /* 453536523e95SDavid Sterba * Add global block reserve 453636523e95SDavid Sterba */ 453736523e95SDavid Sterba if (slot_count) { 45380b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 453936523e95SDavid Sterba 454036523e95SDavid Sterba spin_lock(&block_rsv->lock); 454136523e95SDavid Sterba space.total_bytes = block_rsv->size; 454236523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 454336523e95SDavid Sterba spin_unlock(&block_rsv->lock); 454436523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 454536523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 454636523e95SDavid Sterba space_args.total_spaces++; 454736523e95SDavid Sterba } 454836523e95SDavid Sterba 45492eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 45507fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 45517fde62bfSChris Mason 45527fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 45537fde62bfSChris Mason ret = -EFAULT; 45547fde62bfSChris Mason 45557fde62bfSChris Mason kfree(dest_orig); 45567fde62bfSChris Mason out: 45577fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 45581406e432SJosef Bacik ret = -EFAULT; 45591406e432SJosef Bacik 45601406e432SJosef Bacik return ret; 45611406e432SJosef Bacik } 45621406e432SJosef Bacik 45639a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 45649a8c28beSMiao Xie void __user *argp) 456546204592SSage Weil { 456646204592SSage Weil struct btrfs_trans_handle *trans; 456746204592SSage Weil u64 transid; 4568db5b493aSTsutomu Itoh int ret; 456946204592SSage Weil 4570d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4571ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4572ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 457398d5dc13STsutomu Itoh return PTR_ERR(trans); 4574ff7c1d33SMiao Xie 4575ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4576ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4577ff7c1d33SMiao Xie goto out; 4578ff7c1d33SMiao Xie } 457946204592SSage Weil transid = trans->transid; 45803a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 45818b2b2d3cSTsutomu Itoh if (ret) { 45823a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4583db5b493aSTsutomu Itoh return ret; 45848b2b2d3cSTsutomu Itoh } 4585ff7c1d33SMiao Xie out: 458646204592SSage Weil if (argp) 458746204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 458846204592SSage Weil return -EFAULT; 458946204592SSage Weil return 0; 459046204592SSage Weil } 459146204592SSage Weil 45922ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 45939a8c28beSMiao Xie void __user *argp) 459446204592SSage Weil { 459546204592SSage Weil u64 transid; 459646204592SSage Weil 459746204592SSage Weil if (argp) { 459846204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 459946204592SSage Weil return -EFAULT; 460046204592SSage Weil } else { 460146204592SSage Weil transid = 0; /* current trans */ 460246204592SSage Weil } 46032ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 460446204592SSage Weil } 460546204592SSage Weil 4606b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4607475f6387SJan Schmidt { 46080b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4609475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4610b8e95489SMiao Xie int ret; 4611475f6387SJan Schmidt 4612475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4613475f6387SJan Schmidt return -EPERM; 4614475f6387SJan Schmidt 4615475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4616475f6387SJan Schmidt if (IS_ERR(sa)) 4617475f6387SJan Schmidt return PTR_ERR(sa); 4618475f6387SJan Schmidt 4619b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4620b8e95489SMiao Xie ret = mnt_want_write_file(file); 4621b8e95489SMiao Xie if (ret) 4622b8e95489SMiao Xie goto out; 4623b8e95489SMiao Xie } 4624b8e95489SMiao Xie 46250b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 462663a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 462763a212abSStefan Behrens 0); 4628475f6387SJan Schmidt 4629475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4630475f6387SJan Schmidt ret = -EFAULT; 4631475f6387SJan Schmidt 4632b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4633b8e95489SMiao Xie mnt_drop_write_file(file); 4634b8e95489SMiao Xie out: 4635475f6387SJan Schmidt kfree(sa); 4636475f6387SJan Schmidt return ret; 4637475f6387SJan Schmidt } 4638475f6387SJan Schmidt 46392ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4640475f6387SJan Schmidt { 4641475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4642475f6387SJan Schmidt return -EPERM; 4643475f6387SJan Schmidt 46442ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4645475f6387SJan Schmidt } 4646475f6387SJan Schmidt 46472ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4648475f6387SJan Schmidt void __user *arg) 4649475f6387SJan Schmidt { 4650475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4651475f6387SJan Schmidt int ret; 4652475f6387SJan Schmidt 4653475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4654475f6387SJan Schmidt return -EPERM; 4655475f6387SJan Schmidt 4656475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4657475f6387SJan Schmidt if (IS_ERR(sa)) 4658475f6387SJan Schmidt return PTR_ERR(sa); 4659475f6387SJan Schmidt 46602ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4661475f6387SJan Schmidt 4662475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4663475f6387SJan Schmidt ret = -EFAULT; 4664475f6387SJan Schmidt 4665475f6387SJan Schmidt kfree(sa); 4666475f6387SJan Schmidt return ret; 4667475f6387SJan Schmidt } 4668475f6387SJan Schmidt 46692ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4670b27f7c0cSDavid Sterba void __user *arg) 4671c11d2c23SStefan Behrens { 4672c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4673c11d2c23SStefan Behrens int ret; 4674c11d2c23SStefan Behrens 4675c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4676c11d2c23SStefan Behrens if (IS_ERR(sa)) 4677c11d2c23SStefan Behrens return PTR_ERR(sa); 4678c11d2c23SStefan Behrens 4679b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4680b27f7c0cSDavid Sterba kfree(sa); 4681b27f7c0cSDavid Sterba return -EPERM; 4682b27f7c0cSDavid Sterba } 4683b27f7c0cSDavid Sterba 46842ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4685c11d2c23SStefan Behrens 4686c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4687c11d2c23SStefan Behrens ret = -EFAULT; 4688c11d2c23SStefan Behrens 4689c11d2c23SStefan Behrens kfree(sa); 4690c11d2c23SStefan Behrens return ret; 4691c11d2c23SStefan Behrens } 4692c11d2c23SStefan Behrens 46932ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 46942ff7e61eSJeff Mahoney void __user *arg) 46953f6bcfbdSStefan Behrens { 46963f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 46973f6bcfbdSStefan Behrens int ret; 46983f6bcfbdSStefan Behrens 46993f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 47003f6bcfbdSStefan Behrens return -EPERM; 47013f6bcfbdSStefan Behrens 47023f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 47033f6bcfbdSStefan Behrens if (IS_ERR(p)) 47043f6bcfbdSStefan Behrens return PTR_ERR(p); 47053f6bcfbdSStefan Behrens 47063f6bcfbdSStefan Behrens switch (p->cmd) { 47073f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4708bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4709adfa97cbSIlya Dryomov ret = -EROFS; 4710adfa97cbSIlya Dryomov goto out; 4711adfa97cbSIlya Dryomov } 4712171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4713e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 47143f6bcfbdSStefan Behrens } else { 47152ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4716171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 47173f6bcfbdSStefan Behrens } 47183f6bcfbdSStefan Behrens break; 47193f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 47200b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 47213f6bcfbdSStefan Behrens ret = 0; 47223f6bcfbdSStefan Behrens break; 47233f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 472417d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 472597282031SAnand Jain ret = 0; 47263f6bcfbdSStefan Behrens break; 47273f6bcfbdSStefan Behrens default: 47283f6bcfbdSStefan Behrens ret = -EINVAL; 47293f6bcfbdSStefan Behrens break; 47303f6bcfbdSStefan Behrens } 47313f6bcfbdSStefan Behrens 47323f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 47333f6bcfbdSStefan Behrens ret = -EFAULT; 4734adfa97cbSIlya Dryomov out: 47353f6bcfbdSStefan Behrens kfree(p); 47363f6bcfbdSStefan Behrens return ret; 47373f6bcfbdSStefan Behrens } 47383f6bcfbdSStefan Behrens 4739d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4740d7728c96SJan Schmidt { 4741d7728c96SJan Schmidt int ret = 0; 4742d7728c96SJan Schmidt int i; 4743740c3d22SChris Mason u64 rel_ptr; 4744d7728c96SJan Schmidt int size; 4745806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4746d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4747d7728c96SJan Schmidt struct btrfs_path *path; 4748d7728c96SJan Schmidt 474982b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4750d7728c96SJan Schmidt return -EPERM; 4751d7728c96SJan Schmidt 4752d7728c96SJan Schmidt path = btrfs_alloc_path(); 4753d7728c96SJan Schmidt if (!path) { 4754d7728c96SJan Schmidt ret = -ENOMEM; 4755d7728c96SJan Schmidt goto out; 4756d7728c96SJan Schmidt } 4757d7728c96SJan Schmidt 4758d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4759d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4760d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4761d7728c96SJan Schmidt ipa = NULL; 4762d7728c96SJan Schmidt goto out; 4763d7728c96SJan Schmidt } 4764d7728c96SJan Schmidt 4765d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4766d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4767d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4768d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4769d7728c96SJan Schmidt ipath = NULL; 4770d7728c96SJan Schmidt goto out; 4771d7728c96SJan Schmidt } 4772d7728c96SJan Schmidt 4773d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4774d7728c96SJan Schmidt if (ret < 0) 4775d7728c96SJan Schmidt goto out; 4776d7728c96SJan Schmidt 4777d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4778745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4779745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4780740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4781d7728c96SJan Schmidt } 4782d7728c96SJan Schmidt 4783718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4784718dc5faSOmar Sandoval ipath->fspath, size); 4785d7728c96SJan Schmidt if (ret) { 4786d7728c96SJan Schmidt ret = -EFAULT; 4787d7728c96SJan Schmidt goto out; 4788d7728c96SJan Schmidt } 4789d7728c96SJan Schmidt 4790d7728c96SJan Schmidt out: 4791d7728c96SJan Schmidt btrfs_free_path(path); 4792d7728c96SJan Schmidt free_ipath(ipath); 4793d7728c96SJan Schmidt kfree(ipa); 4794d7728c96SJan Schmidt 4795d7728c96SJan Schmidt return ret; 4796d7728c96SJan Schmidt } 4797d7728c96SJan Schmidt 4798d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4799d7728c96SJan Schmidt { 4800d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4801d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4802d7728c96SJan Schmidt 4803d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4804d7728c96SJan Schmidt inodes->bytes_left -= c; 4805d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4806d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4807d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4808d7728c96SJan Schmidt inodes->elem_cnt += 3; 4809d7728c96SJan Schmidt } else { 4810d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4811d7728c96SJan Schmidt inodes->bytes_left = 0; 4812d7728c96SJan Schmidt inodes->elem_missed += 3; 4813d7728c96SJan Schmidt } 4814d7728c96SJan Schmidt 4815d7728c96SJan Schmidt return 0; 4816d7728c96SJan Schmidt } 4817d7728c96SJan Schmidt 48182ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4819d24a67b2SZygo Blaxell void __user *arg, int version) 4820d7728c96SJan Schmidt { 4821d7728c96SJan Schmidt int ret = 0; 4822d7728c96SJan Schmidt int size; 4823d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4824d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4825d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4826d24a67b2SZygo Blaxell bool ignore_offset; 4827d7728c96SJan Schmidt 4828d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4829d7728c96SJan Schmidt return -EPERM; 4830d7728c96SJan Schmidt 4831d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 48327b9ea627SShailendra Verma if (IS_ERR(loi)) 48337b9ea627SShailendra Verma return PTR_ERR(loi); 4834d7728c96SJan Schmidt 4835d24a67b2SZygo Blaxell if (version == 1) { 4836d24a67b2SZygo Blaxell ignore_offset = false; 4837b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4838d24a67b2SZygo Blaxell } else { 4839d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4840d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4841d24a67b2SZygo Blaxell ret = -EINVAL; 4842d24a67b2SZygo Blaxell goto out_loi; 4843d24a67b2SZygo Blaxell } 4844d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4845d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4846d24a67b2SZygo Blaxell ret = -EINVAL; 4847d24a67b2SZygo Blaxell goto out_loi; 4848d24a67b2SZygo Blaxell } 4849d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4850b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4851d24a67b2SZygo Blaxell } 4852d24a67b2SZygo Blaxell 4853d7728c96SJan Schmidt path = btrfs_alloc_path(); 4854d7728c96SJan Schmidt if (!path) { 4855d7728c96SJan Schmidt ret = -ENOMEM; 4856d7728c96SJan Schmidt goto out; 4857d7728c96SJan Schmidt } 4858d7728c96SJan Schmidt 4859d7728c96SJan Schmidt inodes = init_data_container(size); 4860d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4861d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4862d7728c96SJan Schmidt inodes = NULL; 4863d7728c96SJan Schmidt goto out; 4864d7728c96SJan Schmidt } 4865d7728c96SJan Schmidt 48662ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4867d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4868df031f07SLiu Bo if (ret == -EINVAL) 4869d7728c96SJan Schmidt ret = -ENOENT; 4870d7728c96SJan Schmidt if (ret < 0) 4871d7728c96SJan Schmidt goto out; 4872d7728c96SJan Schmidt 4873718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4874718dc5faSOmar Sandoval size); 4875d7728c96SJan Schmidt if (ret) 4876d7728c96SJan Schmidt ret = -EFAULT; 4877d7728c96SJan Schmidt 4878d7728c96SJan Schmidt out: 4879d7728c96SJan Schmidt btrfs_free_path(path); 4880f54de068SDavid Sterba kvfree(inodes); 4881d24a67b2SZygo Blaxell out_loi: 4882d7728c96SJan Schmidt kfree(loi); 4883d7728c96SJan Schmidt 4884d7728c96SJan Schmidt return ret; 4885d7728c96SJan Schmidt } 4886d7728c96SJan Schmidt 4887008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4888c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4889c9e9f97bSIlya Dryomov { 4890c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4891c9e9f97bSIlya Dryomov 4892c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4893c9e9f97bSIlya Dryomov 48943009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4895837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4896837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4897837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4898a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4899a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4900837d5b6eSIlya Dryomov 4901c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4902c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4903c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 490419a39dceSIlya Dryomov 490519a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 490619a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 490719a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4908c9e9f97bSIlya Dryomov } 4909c9e9f97bSIlya Dryomov 49109ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4911c9e9f97bSIlya Dryomov { 4912496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4913c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4914c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4915c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4916ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4917c9e9f97bSIlya Dryomov int ret; 4918c9e9f97bSIlya Dryomov 4919c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4920c9e9f97bSIlya Dryomov return -EPERM; 4921c9e9f97bSIlya Dryomov 4922e54bfa31SLiu Bo ret = mnt_want_write_file(file); 49239ba1f6e4SLiu Bo if (ret) 49249ba1f6e4SLiu Bo return ret; 49259ba1f6e4SLiu Bo 4926ed0fb78fSIlya Dryomov again: 4927171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4928c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4929ed0fb78fSIlya Dryomov need_unlock = true; 4930ed0fb78fSIlya Dryomov goto locked; 4931ed0fb78fSIlya Dryomov } 4932ed0fb78fSIlya Dryomov 4933ed0fb78fSIlya Dryomov /* 493401327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4935ed0fb78fSIlya Dryomov * (1) some other op is running 4936ed0fb78fSIlya Dryomov * (2) balance is running 4937ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4938ed0fb78fSIlya Dryomov */ 4939ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4940ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4941ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 49423009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4943ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4944dccdb07bSDavid Sterba /* 4945dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4946dccdb07bSDavid Sterba * we'll reexamine the status again. 4947dccdb07bSDavid Sterba */ 4948ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4949ed0fb78fSIlya Dryomov 4950ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 49513009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4952ed0fb78fSIlya Dryomov /* this is (3) */ 4953ed0fb78fSIlya Dryomov need_unlock = false; 4954ed0fb78fSIlya Dryomov goto locked; 4955ed0fb78fSIlya Dryomov } 4956ed0fb78fSIlya Dryomov 4957ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4958ed0fb78fSIlya Dryomov goto again; 4959ed0fb78fSIlya Dryomov } else { 4960ed0fb78fSIlya Dryomov /* this is (2) */ 4961ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4962ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4963ed0fb78fSIlya Dryomov goto out; 4964ed0fb78fSIlya Dryomov } 4965ed0fb78fSIlya Dryomov } else { 4966ed0fb78fSIlya Dryomov /* this is (1) */ 4967ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4968e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4969ed0fb78fSIlya Dryomov goto out; 4970ed0fb78fSIlya Dryomov } 4971ed0fb78fSIlya Dryomov 4972ed0fb78fSIlya Dryomov locked: 4973171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4974c9e9f97bSIlya Dryomov 4975c9e9f97bSIlya Dryomov if (arg) { 4976c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4977c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4978c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4979ed0fb78fSIlya Dryomov goto out_unlock; 4980c9e9f97bSIlya Dryomov } 4981de322263SIlya Dryomov 4982de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4983de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4984de322263SIlya Dryomov ret = -ENOTCONN; 4985de322263SIlya Dryomov goto out_bargs; 4986de322263SIlya Dryomov } 4987de322263SIlya Dryomov 4988de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4989de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4990de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4991de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4992de322263SIlya Dryomov 4993de322263SIlya Dryomov goto do_balance; 4994de322263SIlya Dryomov } 4995c9e9f97bSIlya Dryomov } else { 4996c9e9f97bSIlya Dryomov bargs = NULL; 4997c9e9f97bSIlya Dryomov } 4998c9e9f97bSIlya Dryomov 4999ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 5000837d5b6eSIlya Dryomov ret = -EINPROGRESS; 5001837d5b6eSIlya Dryomov goto out_bargs; 5002837d5b6eSIlya Dryomov } 5003837d5b6eSIlya Dryomov 50048d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 5005c9e9f97bSIlya Dryomov if (!bctl) { 5006c9e9f97bSIlya Dryomov ret = -ENOMEM; 5007c9e9f97bSIlya Dryomov goto out_bargs; 5008c9e9f97bSIlya Dryomov } 5009c9e9f97bSIlya Dryomov 5010c9e9f97bSIlya Dryomov if (arg) { 5011c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 5012c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 5013c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 5014c9e9f97bSIlya Dryomov 5015c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 5016f43ffb60SIlya Dryomov } else { 5017f43ffb60SIlya Dryomov /* balance everything - no filters */ 5018f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 5019c9e9f97bSIlya Dryomov } 5020c9e9f97bSIlya Dryomov 50218eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 50228eb93459SDavid Sterba ret = -EINVAL; 50230f89abf5SChristian Engelmayer goto out_bctl; 50248eb93459SDavid Sterba } 50258eb93459SDavid Sterba 5026de322263SIlya Dryomov do_balance: 5027c9e9f97bSIlya Dryomov /* 5028149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 5029149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 5030149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 5031149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 5032c9e9f97bSIlya Dryomov */ 5033ed0fb78fSIlya Dryomov need_unlock = false; 5034ed0fb78fSIlya Dryomov 50356fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 50360f89abf5SChristian Engelmayer bctl = NULL; 5037ed0fb78fSIlya Dryomov 5038c9e9f97bSIlya Dryomov if (arg) { 5039c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 5040c9e9f97bSIlya Dryomov ret = -EFAULT; 5041c9e9f97bSIlya Dryomov } 5042c9e9f97bSIlya Dryomov 50430f89abf5SChristian Engelmayer out_bctl: 50440f89abf5SChristian Engelmayer kfree(bctl); 5045c9e9f97bSIlya Dryomov out_bargs: 5046c9e9f97bSIlya Dryomov kfree(bargs); 5047ed0fb78fSIlya Dryomov out_unlock: 5048c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 5049ed0fb78fSIlya Dryomov if (need_unlock) 5050171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 5051ed0fb78fSIlya Dryomov out: 5052e54bfa31SLiu Bo mnt_drop_write_file(file); 5053c9e9f97bSIlya Dryomov return ret; 5054c9e9f97bSIlya Dryomov } 5055c9e9f97bSIlya Dryomov 50562ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 5057837d5b6eSIlya Dryomov { 5058837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 5059837d5b6eSIlya Dryomov return -EPERM; 5060837d5b6eSIlya Dryomov 5061837d5b6eSIlya Dryomov switch (cmd) { 5062837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 50630b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 5064a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 50650b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 5066837d5b6eSIlya Dryomov } 5067837d5b6eSIlya Dryomov 5068837d5b6eSIlya Dryomov return -EINVAL; 5069837d5b6eSIlya Dryomov } 5070837d5b6eSIlya Dryomov 50712ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 507219a39dceSIlya Dryomov void __user *arg) 507319a39dceSIlya Dryomov { 507419a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 507519a39dceSIlya Dryomov int ret = 0; 507619a39dceSIlya Dryomov 507719a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 507819a39dceSIlya Dryomov return -EPERM; 507919a39dceSIlya Dryomov 508019a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 508119a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 508219a39dceSIlya Dryomov ret = -ENOTCONN; 508319a39dceSIlya Dryomov goto out; 508419a39dceSIlya Dryomov } 508519a39dceSIlya Dryomov 50868d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 508719a39dceSIlya Dryomov if (!bargs) { 508819a39dceSIlya Dryomov ret = -ENOMEM; 508919a39dceSIlya Dryomov goto out; 509019a39dceSIlya Dryomov } 509119a39dceSIlya Dryomov 5092008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 509319a39dceSIlya Dryomov 509419a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 509519a39dceSIlya Dryomov ret = -EFAULT; 509619a39dceSIlya Dryomov 509719a39dceSIlya Dryomov kfree(bargs); 509819a39dceSIlya Dryomov out: 509919a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 510019a39dceSIlya Dryomov return ret; 510119a39dceSIlya Dryomov } 510219a39dceSIlya Dryomov 5103905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 51045d13a37bSArne Jansen { 51050b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51060b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51075d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 51085d13a37bSArne Jansen int ret; 51095d13a37bSArne Jansen 51105d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 51115d13a37bSArne Jansen return -EPERM; 51125d13a37bSArne Jansen 5113905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5114905b0ddaSMiao Xie if (ret) 5115905b0ddaSMiao Xie return ret; 51165d13a37bSArne Jansen 51175d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5118905b0ddaSMiao Xie if (IS_ERR(sa)) { 5119905b0ddaSMiao Xie ret = PTR_ERR(sa); 5120905b0ddaSMiao Xie goto drop_write; 5121905b0ddaSMiao Xie } 51225d13a37bSArne Jansen 51230b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51245d13a37bSArne Jansen 51255d13a37bSArne Jansen switch (sa->cmd) { 51265d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 5127340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 51285d13a37bSArne Jansen break; 51295d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 5130340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 51315d13a37bSArne Jansen break; 51325d13a37bSArne Jansen default: 51335d13a37bSArne Jansen ret = -EINVAL; 51345d13a37bSArne Jansen break; 51355d13a37bSArne Jansen } 51365d13a37bSArne Jansen 51375d13a37bSArne Jansen kfree(sa); 51380b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5139905b0ddaSMiao Xie drop_write: 5140905b0ddaSMiao Xie mnt_drop_write_file(file); 51415d13a37bSArne Jansen return ret; 51425d13a37bSArne Jansen } 51435d13a37bSArne Jansen 5144905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 51455d13a37bSArne Jansen { 51460b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51480b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51495d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 51505d13a37bSArne Jansen struct btrfs_trans_handle *trans; 51515d13a37bSArne Jansen int ret; 51525d13a37bSArne Jansen int err; 51535d13a37bSArne Jansen 51545d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 51555d13a37bSArne Jansen return -EPERM; 51565d13a37bSArne Jansen 5157905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5158905b0ddaSMiao Xie if (ret) 5159905b0ddaSMiao Xie return ret; 51605d13a37bSArne Jansen 51615d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5162905b0ddaSMiao Xie if (IS_ERR(sa)) { 5163905b0ddaSMiao Xie ret = PTR_ERR(sa); 5164905b0ddaSMiao Xie goto drop_write; 5165905b0ddaSMiao Xie } 51665d13a37bSArne Jansen 51675d13a37bSArne Jansen trans = btrfs_join_transaction(root); 51685d13a37bSArne Jansen if (IS_ERR(trans)) { 51695d13a37bSArne Jansen ret = PTR_ERR(trans); 51705d13a37bSArne Jansen goto out; 51715d13a37bSArne Jansen } 51725d13a37bSArne Jansen 51735d13a37bSArne Jansen if (sa->assign) { 51749f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 51755d13a37bSArne Jansen } else { 517639616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 51775d13a37bSArne Jansen } 51785d13a37bSArne Jansen 5179e082f563SQu Wenruo /* update qgroup status and info */ 51800b246afaSJeff Mahoney err = btrfs_run_qgroups(trans, fs_info); 5181e082f563SQu Wenruo if (err < 0) 51820b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 5183ad8403dfSAnand Jain "failed to update qgroup status and info"); 51843a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 51855d13a37bSArne Jansen if (err && !ret) 51865d13a37bSArne Jansen ret = err; 51875d13a37bSArne Jansen 51885d13a37bSArne Jansen out: 51895d13a37bSArne Jansen kfree(sa); 5190905b0ddaSMiao Xie drop_write: 5191905b0ddaSMiao Xie mnt_drop_write_file(file); 51925d13a37bSArne Jansen return ret; 51935d13a37bSArne Jansen } 51945d13a37bSArne Jansen 5195905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 51965d13a37bSArne Jansen { 51970b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51990b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52005d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 52015d13a37bSArne Jansen struct btrfs_trans_handle *trans; 52025d13a37bSArne Jansen int ret; 52035d13a37bSArne Jansen int err; 52045d13a37bSArne Jansen 52055d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 52065d13a37bSArne Jansen return -EPERM; 52075d13a37bSArne Jansen 5208905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5209905b0ddaSMiao Xie if (ret) 5210905b0ddaSMiao Xie return ret; 52115d13a37bSArne Jansen 52125d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5213905b0ddaSMiao Xie if (IS_ERR(sa)) { 5214905b0ddaSMiao Xie ret = PTR_ERR(sa); 5215905b0ddaSMiao Xie goto drop_write; 5216905b0ddaSMiao Xie } 52175d13a37bSArne Jansen 5218d86e56cfSMiao Xie if (!sa->qgroupid) { 5219d86e56cfSMiao Xie ret = -EINVAL; 5220d86e56cfSMiao Xie goto out; 5221d86e56cfSMiao Xie } 5222d86e56cfSMiao Xie 52235d13a37bSArne Jansen trans = btrfs_join_transaction(root); 52245d13a37bSArne Jansen if (IS_ERR(trans)) { 52255d13a37bSArne Jansen ret = PTR_ERR(trans); 52265d13a37bSArne Jansen goto out; 52275d13a37bSArne Jansen } 52285d13a37bSArne Jansen 52295d13a37bSArne Jansen if (sa->create) { 52300b246afaSJeff Mahoney ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); 52315d13a37bSArne Jansen } else { 52320b246afaSJeff Mahoney ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); 52335d13a37bSArne Jansen } 52345d13a37bSArne Jansen 52353a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 52365d13a37bSArne Jansen if (err && !ret) 52375d13a37bSArne Jansen ret = err; 52385d13a37bSArne Jansen 52395d13a37bSArne Jansen out: 52405d13a37bSArne Jansen kfree(sa); 5241905b0ddaSMiao Xie drop_write: 5242905b0ddaSMiao Xie mnt_drop_write_file(file); 52435d13a37bSArne Jansen return ret; 52445d13a37bSArne Jansen } 52455d13a37bSArne Jansen 5246905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 52475d13a37bSArne Jansen { 52480b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52500b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52515d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 52525d13a37bSArne Jansen struct btrfs_trans_handle *trans; 52535d13a37bSArne Jansen int ret; 52545d13a37bSArne Jansen int err; 52555d13a37bSArne Jansen u64 qgroupid; 52565d13a37bSArne Jansen 52575d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 52585d13a37bSArne Jansen return -EPERM; 52595d13a37bSArne Jansen 5260905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5261905b0ddaSMiao Xie if (ret) 5262905b0ddaSMiao Xie return ret; 52635d13a37bSArne Jansen 52645d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5265905b0ddaSMiao Xie if (IS_ERR(sa)) { 5266905b0ddaSMiao Xie ret = PTR_ERR(sa); 5267905b0ddaSMiao Xie goto drop_write; 5268905b0ddaSMiao Xie } 52695d13a37bSArne Jansen 52705d13a37bSArne Jansen trans = btrfs_join_transaction(root); 52715d13a37bSArne Jansen if (IS_ERR(trans)) { 52725d13a37bSArne Jansen ret = PTR_ERR(trans); 52735d13a37bSArne Jansen goto out; 52745d13a37bSArne Jansen } 52755d13a37bSArne Jansen 52765d13a37bSArne Jansen qgroupid = sa->qgroupid; 52775d13a37bSArne Jansen if (!qgroupid) { 52785d13a37bSArne Jansen /* take the current subvol as qgroup */ 52795d13a37bSArne Jansen qgroupid = root->root_key.objectid; 52805d13a37bSArne Jansen } 52815d13a37bSArne Jansen 52820b246afaSJeff Mahoney ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); 52835d13a37bSArne Jansen 52843a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 52855d13a37bSArne Jansen if (err && !ret) 52865d13a37bSArne Jansen ret = err; 52875d13a37bSArne Jansen 52885d13a37bSArne Jansen out: 52895d13a37bSArne Jansen kfree(sa); 5290905b0ddaSMiao Xie drop_write: 5291905b0ddaSMiao Xie mnt_drop_write_file(file); 52925d13a37bSArne Jansen return ret; 52935d13a37bSArne Jansen } 52945d13a37bSArne Jansen 52952f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 52962f232036SJan Schmidt { 52970b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52992f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 53002f232036SJan Schmidt int ret; 53012f232036SJan Schmidt 53022f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 53032f232036SJan Schmidt return -EPERM; 53042f232036SJan Schmidt 53052f232036SJan Schmidt ret = mnt_want_write_file(file); 53062f232036SJan Schmidt if (ret) 53072f232036SJan Schmidt return ret; 53082f232036SJan Schmidt 53092f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 53102f232036SJan Schmidt if (IS_ERR(qsa)) { 53112f232036SJan Schmidt ret = PTR_ERR(qsa); 53122f232036SJan Schmidt goto drop_write; 53132f232036SJan Schmidt } 53142f232036SJan Schmidt 53152f232036SJan Schmidt if (qsa->flags) { 53162f232036SJan Schmidt ret = -EINVAL; 53172f232036SJan Schmidt goto out; 53182f232036SJan Schmidt } 53192f232036SJan Schmidt 53200b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 53212f232036SJan Schmidt 53222f232036SJan Schmidt out: 53232f232036SJan Schmidt kfree(qsa); 53242f232036SJan Schmidt drop_write: 53252f232036SJan Schmidt mnt_drop_write_file(file); 53262f232036SJan Schmidt return ret; 53272f232036SJan Schmidt } 53282f232036SJan Schmidt 53292f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 53302f232036SJan Schmidt { 53310b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53332f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 53342f232036SJan Schmidt int ret = 0; 53352f232036SJan Schmidt 53362f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 53372f232036SJan Schmidt return -EPERM; 53382f232036SJan Schmidt 53398d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 53402f232036SJan Schmidt if (!qsa) 53412f232036SJan Schmidt return -ENOMEM; 53422f232036SJan Schmidt 53430b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 53442f232036SJan Schmidt qsa->flags = 1; 53450b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 53462f232036SJan Schmidt } 53472f232036SJan Schmidt 53482f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 53492f232036SJan Schmidt ret = -EFAULT; 53502f232036SJan Schmidt 53512f232036SJan Schmidt kfree(qsa); 53522f232036SJan Schmidt return ret; 53532f232036SJan Schmidt } 53542f232036SJan Schmidt 535557254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 535657254b6eSJan Schmidt { 53570b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 535957254b6eSJan Schmidt 536057254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 536157254b6eSJan Schmidt return -EPERM; 536257254b6eSJan Schmidt 53630b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 536457254b6eSJan Schmidt } 536557254b6eSJan Schmidt 5366abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5367abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 53688ea05e3aSAlexander Block { 5369496ad9aaSAl Viro struct inode *inode = file_inode(file); 53700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53718ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 53728ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 53738ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 537495582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 53758ea05e3aSAlexander Block int ret = 0; 5376dd5f9615SStefan Behrens int received_uuid_changed; 53778ea05e3aSAlexander Block 5378bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5379bd60ea0fSDavid Sterba return -EPERM; 5380bd60ea0fSDavid Sterba 53818ea05e3aSAlexander Block ret = mnt_want_write_file(file); 53828ea05e3aSAlexander Block if (ret < 0) 53838ea05e3aSAlexander Block return ret; 53848ea05e3aSAlexander Block 53850b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 53868ea05e3aSAlexander Block 53874a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 53888ea05e3aSAlexander Block ret = -EINVAL; 53898ea05e3aSAlexander Block goto out; 53908ea05e3aSAlexander Block } 53918ea05e3aSAlexander Block 53928ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 53938ea05e3aSAlexander Block ret = -EROFS; 53948ea05e3aSAlexander Block goto out; 53958ea05e3aSAlexander Block } 53968ea05e3aSAlexander Block 5397dd5f9615SStefan Behrens /* 5398dd5f9615SStefan Behrens * 1 - root item 5399dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5400dd5f9615SStefan Behrens */ 5401dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 54028ea05e3aSAlexander Block if (IS_ERR(trans)) { 54038ea05e3aSAlexander Block ret = PTR_ERR(trans); 54048ea05e3aSAlexander Block trans = NULL; 54058ea05e3aSAlexander Block goto out; 54068ea05e3aSAlexander Block } 54078ea05e3aSAlexander Block 54088ea05e3aSAlexander Block sa->rtransid = trans->transid; 54098ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 54108ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 54118ea05e3aSAlexander Block 5412dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5413dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5414dd5f9615SStefan Behrens if (received_uuid_changed && 5415d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5416d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5417dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5418dd5f9615SStefan Behrens root->root_key.objectid); 5419d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5420d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5421d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5422d87ff758SNikolay Borisov goto out; 5423d87ff758SNikolay Borisov } 5424d87ff758SNikolay Borisov } 54258ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 54268ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 54278ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 54283cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 54293cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 54303cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 54313cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 54328ea05e3aSAlexander Block 54330b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 54348ea05e3aSAlexander Block &root->root_key, &root->root_item); 54358ea05e3aSAlexander Block if (ret < 0) { 54363a45bb20SJeff Mahoney btrfs_end_transaction(trans); 54378ea05e3aSAlexander Block goto out; 5438dd5f9615SStefan Behrens } 5439dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5440cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5441dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5442dd5f9615SStefan Behrens root->root_key.objectid); 5443dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 544466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5445efd38150SNikolay Borisov btrfs_end_transaction(trans); 5446dd5f9615SStefan Behrens goto out; 5447dd5f9615SStefan Behrens } 5448dd5f9615SStefan Behrens } 54493a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5450abccd00fSHugo Mills out: 54510b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5452abccd00fSHugo Mills mnt_drop_write_file(file); 5453abccd00fSHugo Mills return ret; 5454abccd00fSHugo Mills } 5455abccd00fSHugo Mills 5456abccd00fSHugo Mills #ifdef CONFIG_64BIT 5457abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5458abccd00fSHugo Mills void __user *arg) 5459abccd00fSHugo Mills { 5460abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5461abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5462abccd00fSHugo Mills int ret = 0; 5463abccd00fSHugo Mills 5464abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 54657b9ea627SShailendra Verma if (IS_ERR(args32)) 54667b9ea627SShailendra Verma return PTR_ERR(args32); 5467abccd00fSHugo Mills 54688d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 546984dbeb87SDan Carpenter if (!args64) { 547084dbeb87SDan Carpenter ret = -ENOMEM; 5471abccd00fSHugo Mills goto out; 5472abccd00fSHugo Mills } 5473abccd00fSHugo Mills 5474abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5475abccd00fSHugo Mills args64->stransid = args32->stransid; 5476abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5477abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5478abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5479abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5480abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5481abccd00fSHugo Mills args64->flags = args32->flags; 5482abccd00fSHugo Mills 5483abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5484abccd00fSHugo Mills if (ret) 5485abccd00fSHugo Mills goto out; 5486abccd00fSHugo Mills 5487abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5488abccd00fSHugo Mills args32->stransid = args64->stransid; 5489abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5490abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5491abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5492abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5493abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5494abccd00fSHugo Mills args32->flags = args64->flags; 5495abccd00fSHugo Mills 5496abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5497abccd00fSHugo Mills if (ret) 5498abccd00fSHugo Mills ret = -EFAULT; 5499abccd00fSHugo Mills 5500abccd00fSHugo Mills out: 5501abccd00fSHugo Mills kfree(args32); 5502abccd00fSHugo Mills kfree(args64); 5503abccd00fSHugo Mills return ret; 5504abccd00fSHugo Mills } 5505abccd00fSHugo Mills #endif 5506abccd00fSHugo Mills 5507abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5508abccd00fSHugo Mills void __user *arg) 5509abccd00fSHugo Mills { 5510abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5511abccd00fSHugo Mills int ret = 0; 5512abccd00fSHugo Mills 5513abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 55147b9ea627SShailendra Verma if (IS_ERR(sa)) 55157b9ea627SShailendra Verma return PTR_ERR(sa); 5516abccd00fSHugo Mills 5517abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5518abccd00fSHugo Mills 5519abccd00fSHugo Mills if (ret) 5520abccd00fSHugo Mills goto out; 5521abccd00fSHugo Mills 55228ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 55238ea05e3aSAlexander Block if (ret) 55248ea05e3aSAlexander Block ret = -EFAULT; 55258ea05e3aSAlexander Block 55268ea05e3aSAlexander Block out: 55278ea05e3aSAlexander Block kfree(sa); 55288ea05e3aSAlexander Block return ret; 55298ea05e3aSAlexander Block } 55308ea05e3aSAlexander Block 5531867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5532867ab667Sjeff.liu { 55330b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55340b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5535a1b83ac5SAnand Jain size_t len; 5536867ab667Sjeff.liu int ret; 5537a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5538a1b83ac5SAnand Jain 55390b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 55400b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 55410b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5542a1b83ac5SAnand Jain 5543a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5544867ab667Sjeff.liu 5545867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 55460b246afaSJeff Mahoney btrfs_warn(fs_info, 55470b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 55480b246afaSJeff Mahoney --len); 5549867ab667Sjeff.liu } 5550867ab667Sjeff.liu 5551867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5552867ab667Sjeff.liu 5553867ab667Sjeff.liu return ret ? -EFAULT : 0; 5554867ab667Sjeff.liu } 5555867ab667Sjeff.liu 5556a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5557a8bfd4abSjeff.liu { 55580b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55600b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55610b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5562a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5563a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5564a8bfd4abSjeff.liu int ret; 5565a8bfd4abSjeff.liu 5566a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5567a8bfd4abSjeff.liu return -EPERM; 5568a8bfd4abSjeff.liu 5569a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5570a8bfd4abSjeff.liu return -EFAULT; 5571a8bfd4abSjeff.liu 5572a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 55730b246afaSJeff Mahoney btrfs_err(fs_info, 55745d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5575a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5576a8bfd4abSjeff.liu return -EINVAL; 5577a8bfd4abSjeff.liu } 5578a8bfd4abSjeff.liu 5579a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5580a8bfd4abSjeff.liu if (ret) 5581a8bfd4abSjeff.liu return ret; 5582a8bfd4abSjeff.liu 5583a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5584a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5585a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5586a8bfd4abSjeff.liu goto out_unlock; 5587a8bfd4abSjeff.liu } 5588a8bfd4abSjeff.liu 55890b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5590a8bfd4abSjeff.liu strcpy(super_block->label, label); 55910b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 55923a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5593a8bfd4abSjeff.liu 5594a8bfd4abSjeff.liu out_unlock: 5595a8bfd4abSjeff.liu mnt_drop_write_file(file); 5596a8bfd4abSjeff.liu return ret; 5597a8bfd4abSjeff.liu } 5598a8bfd4abSjeff.liu 55992eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 56002eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 56012eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 56022eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 56032eaa055fSJeff Mahoney 5604d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 56052eaa055fSJeff Mahoney { 56064d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 56072eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 56082eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 56092eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 56102eaa055fSJeff Mahoney }; 56112eaa055fSJeff Mahoney 56122eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 56132eaa055fSJeff Mahoney return -EFAULT; 56142eaa055fSJeff Mahoney 56152eaa055fSJeff Mahoney return 0; 56162eaa055fSJeff Mahoney } 56172eaa055fSJeff Mahoney 56182eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 56192eaa055fSJeff Mahoney { 56200b246afaSJeff Mahoney struct inode *inode = file_inode(file); 56210b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 56220b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 56232eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 56242eaa055fSJeff Mahoney 56252eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 56262eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 56272eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 56282eaa055fSJeff Mahoney 56292eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 56302eaa055fSJeff Mahoney return -EFAULT; 56312eaa055fSJeff Mahoney 56322eaa055fSJeff Mahoney return 0; 56332eaa055fSJeff Mahoney } 56342eaa055fSJeff Mahoney 56352ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 56363b02a68aSJeff Mahoney enum btrfs_feature_set set, 56372eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 56382eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 56392eaa055fSJeff Mahoney { 56403b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 56413b02a68aSJeff Mahoney char *names; 56422eaa055fSJeff Mahoney u64 disallowed, unsupported; 56432eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 56442eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 56452eaa055fSJeff Mahoney 56462eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 56472eaa055fSJeff Mahoney if (unsupported) { 56483b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 56493b02a68aSJeff Mahoney if (names) { 56500b246afaSJeff Mahoney btrfs_warn(fs_info, 56513b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 56523b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 56533b02a68aSJeff Mahoney kfree(names); 56543b02a68aSJeff Mahoney } else 56550b246afaSJeff Mahoney btrfs_warn(fs_info, 56562eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 56572eaa055fSJeff Mahoney type, unsupported); 56582eaa055fSJeff Mahoney return -EOPNOTSUPP; 56592eaa055fSJeff Mahoney } 56602eaa055fSJeff Mahoney 56612eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 56622eaa055fSJeff Mahoney if (disallowed) { 56633b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 56643b02a68aSJeff Mahoney if (names) { 56650b246afaSJeff Mahoney btrfs_warn(fs_info, 56663b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 56673b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 56683b02a68aSJeff Mahoney kfree(names); 56693b02a68aSJeff Mahoney } else 56700b246afaSJeff Mahoney btrfs_warn(fs_info, 56712eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 56722eaa055fSJeff Mahoney type, disallowed); 56732eaa055fSJeff Mahoney return -EPERM; 56742eaa055fSJeff Mahoney } 56752eaa055fSJeff Mahoney 56762eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 56772eaa055fSJeff Mahoney if (disallowed) { 56783b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 56793b02a68aSJeff Mahoney if (names) { 56800b246afaSJeff Mahoney btrfs_warn(fs_info, 56813b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 56823b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 56833b02a68aSJeff Mahoney kfree(names); 56843b02a68aSJeff Mahoney } else 56850b246afaSJeff Mahoney btrfs_warn(fs_info, 56862eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 56872eaa055fSJeff Mahoney type, disallowed); 56882eaa055fSJeff Mahoney return -EPERM; 56892eaa055fSJeff Mahoney } 56902eaa055fSJeff Mahoney 56912eaa055fSJeff Mahoney return 0; 56922eaa055fSJeff Mahoney } 56932eaa055fSJeff Mahoney 56942ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 56952ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 56962eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 56972eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 56982eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 56992eaa055fSJeff Mahoney 57002eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 57012eaa055fSJeff Mahoney { 57020b246afaSJeff Mahoney struct inode *inode = file_inode(file); 57030b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 57040b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 57050b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 57062eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 57072eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 57082eaa055fSJeff Mahoney u64 newflags; 57092eaa055fSJeff Mahoney int ret; 57102eaa055fSJeff Mahoney 57112eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 57122eaa055fSJeff Mahoney return -EPERM; 57132eaa055fSJeff Mahoney 57142eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 57152eaa055fSJeff Mahoney return -EFAULT; 57162eaa055fSJeff Mahoney 57172eaa055fSJeff Mahoney /* Nothing to do */ 57182eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 57192eaa055fSJeff Mahoney !flags[0].incompat_flags) 57202eaa055fSJeff Mahoney return 0; 57212eaa055fSJeff Mahoney 57222ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 57232eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 57242eaa055fSJeff Mahoney if (ret) 57252eaa055fSJeff Mahoney return ret; 57262eaa055fSJeff Mahoney 57272ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 57282eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 57292eaa055fSJeff Mahoney if (ret) 57302eaa055fSJeff Mahoney return ret; 57312eaa055fSJeff Mahoney 57322ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 57332eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 57342eaa055fSJeff Mahoney if (ret) 57352eaa055fSJeff Mahoney return ret; 57362eaa055fSJeff Mahoney 57377ab19625SDavid Sterba ret = mnt_want_write_file(file); 57387ab19625SDavid Sterba if (ret) 57397ab19625SDavid Sterba return ret; 57407ab19625SDavid Sterba 57418051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 57427ab19625SDavid Sterba if (IS_ERR(trans)) { 57437ab19625SDavid Sterba ret = PTR_ERR(trans); 57447ab19625SDavid Sterba goto out_drop_write; 57457ab19625SDavid Sterba } 57462eaa055fSJeff Mahoney 57470b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 57482eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 57492eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 57502eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 57512eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 57522eaa055fSJeff Mahoney 57532eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 57542eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 57552eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 57562eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 57572eaa055fSJeff Mahoney 57582eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 57592eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 57602eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 57612eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 57620b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 57632eaa055fSJeff Mahoney 57643a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 57657ab19625SDavid Sterba out_drop_write: 57667ab19625SDavid Sterba mnt_drop_write_file(file); 57677ab19625SDavid Sterba 57687ab19625SDavid Sterba return ret; 57692eaa055fSJeff Mahoney } 57702eaa055fSJeff Mahoney 57712351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 57722351f431SJosef Bacik { 57732351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 57742351f431SJosef Bacik int ret; 57752351f431SJosef Bacik 57762351f431SJosef Bacik if (compat) { 57772351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 57782351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 57792351f431SJosef Bacik 57802351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 57812351f431SJosef Bacik if (ret) 57822351f431SJosef Bacik return -EFAULT; 57832351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 57842351f431SJosef Bacik if (!arg) 57852351f431SJosef Bacik return -ENOMEM; 57862351f431SJosef Bacik arg->send_fd = args32.send_fd; 57872351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 57882351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 57892351f431SJosef Bacik arg->parent_root = args32.parent_root; 57902351f431SJosef Bacik arg->flags = args32.flags; 57912351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 57922351f431SJosef Bacik sizeof(args32.reserved)); 57932351f431SJosef Bacik #else 57942351f431SJosef Bacik return -ENOTTY; 57952351f431SJosef Bacik #endif 57962351f431SJosef Bacik } else { 57972351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 57982351f431SJosef Bacik if (IS_ERR(arg)) 57992351f431SJosef Bacik return PTR_ERR(arg); 58002351f431SJosef Bacik } 58012351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 58022351f431SJosef Bacik kfree(arg); 58032351f431SJosef Bacik return ret; 58042351f431SJosef Bacik } 58052351f431SJosef Bacik 5806f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5807f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5808f46b5a66SChristoph Hellwig { 58090b246afaSJeff Mahoney struct inode *inode = file_inode(file); 58100b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 58110b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 58124bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5813f46b5a66SChristoph Hellwig 5814f46b5a66SChristoph Hellwig switch (cmd) { 58156cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 58166cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 58176cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 58186cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 58196cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 58206cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5821f7039b1dSLi Dongyang case FITRIM: 5822f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5823f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5824fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5825fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5826fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 58273de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5828fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 58296f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 58306f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 583176dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 583276dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 58330caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 58340caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 58350caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 58360caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 58376ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 58386ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5839f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 58401e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 58411e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 58421e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5843f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5844198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5845f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 58462ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5847f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5848da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 58496b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 58506b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5851475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 58522ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5853475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 58542ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5855f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 58569ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5857ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5858ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5859cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5860cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5861ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5862ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5863d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5864d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5865d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5866d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5867d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5868d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 58691406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 58702ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 58719b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 58729b199859SFilipe David Borba Manana int ret; 58739b199859SFilipe David Borba Manana 587482b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 58759b199859SFilipe David Borba Manana if (ret) 58769b199859SFilipe David Borba Manana return ret; 58770b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 58782fad4e83SDavid Sterba /* 58792fad4e83SDavid Sterba * The transaction thread may want to do more work, 588001327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 58812fad4e83SDavid Sterba * processing uncleaned subvols. 58822fad4e83SDavid Sterba */ 58830b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 58849b199859SFilipe David Borba Manana return ret; 58859b199859SFilipe David Borba Manana } 588646204592SSage Weil case BTRFS_IOC_START_SYNC: 58879a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 588846204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 58892ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5890475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5891b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5892475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 58932ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5894475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 58952ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5896c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 58979ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5898837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 58992ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 590019a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 59012ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 59028ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 59038ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5904abccd00fSHugo Mills #ifdef CONFIG_64BIT 5905abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5906abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5907abccd00fSHugo Mills #endif 590831db9f7cSAlexander Block case BTRFS_IOC_SEND: 59092351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 59102351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 59112351f431SJosef Bacik case BTRFS_IOC_SEND_32: 59122351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 59132351f431SJosef Bacik #endif 5914c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 59152ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 59165d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5917905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 59185d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5919905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 59205d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5921905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 59225d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5923905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 59242f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 59252f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 59262f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 59272f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 592857254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 592957254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 59303f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 59312ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5932867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5933867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5934a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5935a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 59362eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5937d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 59382eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 59392eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 59402eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 59412eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5942e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5943e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5944025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5945025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5946b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5947b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 594842e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 594942e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 595023d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 595123d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5952f46b5a66SChristoph Hellwig } 5953f46b5a66SChristoph Hellwig 5954f46b5a66SChristoph Hellwig return -ENOTTY; 5955f46b5a66SChristoph Hellwig } 59564c63c245SLuke Dashjr 59574c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 59584c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 59594c63c245SLuke Dashjr { 59602a362249SJeff Mahoney /* 59612a362249SJeff Mahoney * These all access 32-bit values anyway so no further 59622a362249SJeff Mahoney * handling is necessary. 59632a362249SJeff Mahoney */ 59644c63c245SLuke Dashjr switch (cmd) { 59654c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 59664c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 59674c63c245SLuke Dashjr break; 59684c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 59694c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 59704c63c245SLuke Dashjr break; 59714c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 59724c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 59734c63c245SLuke Dashjr break; 59744c63c245SLuke Dashjr } 59754c63c245SLuke Dashjr 59764c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 59774c63c245SLuke Dashjr } 59784c63c245SLuke Dashjr #endif 5979