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; 494f7039b1dSLi Dongyang int ret; 495f7039b1dSLi Dongyang 496f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 497f7039b1dSLi Dongyang return -EPERM; 498f7039b1dSLi Dongyang 4991f78160cSXiao Guangrong rcu_read_lock(); 5001f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 5011f78160cSXiao Guangrong dev_list) { 502f7039b1dSLi Dongyang if (!device->bdev) 503f7039b1dSLi Dongyang continue; 504f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 505f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 506f7039b1dSLi Dongyang num_devices++; 50750d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 508f7039b1dSLi Dongyang minlen); 509f7039b1dSLi Dongyang } 510f7039b1dSLi Dongyang } 5111f78160cSXiao Guangrong rcu_read_unlock(); 512f4c697e6SLukas Czerner 513f7039b1dSLi Dongyang if (!num_devices) 514f7039b1dSLi Dongyang return -EOPNOTSUPP; 515f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 516f7039b1dSLi Dongyang return -EFAULT; 5176ba9fc8eSQu Wenruo 5186ba9fc8eSQu Wenruo /* 5196ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 5206ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 5216ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 5226ba9fc8eSQu Wenruo */ 5236ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 524f4c697e6SLukas Czerner return -EINVAL; 525f7039b1dSLi Dongyang 526f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5272ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 528f7039b1dSLi Dongyang if (ret < 0) 529f7039b1dSLi Dongyang return ret; 530f7039b1dSLi Dongyang 531f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 532f7039b1dSLi Dongyang return -EFAULT; 533f7039b1dSLi Dongyang 534f7039b1dSLi Dongyang return 0; 535f7039b1dSLi Dongyang } 536f7039b1dSLi Dongyang 537dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 538dd5f9615SStefan Behrens { 53946e0f66aSChris Mason int i; 54046e0f66aSChris Mason 54146e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 54246e0f66aSChris Mason if (uuid[i]) 54346e0f66aSChris Mason return 0; 54446e0f66aSChris Mason } 54546e0f66aSChris Mason return 1; 546dd5f9615SStefan Behrens } 547dd5f9615SStefan Behrens 548d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 549cb8e7090SChristoph Hellwig struct dentry *dentry, 55052f75f4fSDavid Sterba const char *name, int namelen, 5516f72c7e2SArne Jansen u64 *async_transid, 5528696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 553f46b5a66SChristoph Hellwig { 5540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 555f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 556f46b5a66SChristoph Hellwig struct btrfs_key key; 55749a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 558f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 559f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 560d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 56176dda93cSYan, Zheng struct btrfs_root *new_root; 562d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 56395582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5645662344bSTsutomu Itoh struct inode *inode; 565f46b5a66SChristoph Hellwig int ret; 566f46b5a66SChristoph Hellwig int err; 567f46b5a66SChristoph Hellwig u64 objectid; 568f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 5693de4586cSChris Mason u64 index = 0; 5708ea05e3aSAlexander Block uuid_le new_uuid; 571f46b5a66SChristoph Hellwig 57249a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 57349a3c4d9SDavid Sterba if (!root_item) 57449a3c4d9SDavid Sterba return -ENOMEM; 57549a3c4d9SDavid Sterba 5760b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5772fbe8c8aSAl Viro if (ret) 57849a3c4d9SDavid Sterba goto fail_free; 5796a912213SJosef Bacik 580e09fe2d2SQu Wenruo /* 581e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 58201327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 583e09fe2d2SQu Wenruo */ 58449a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 58549a3c4d9SDavid Sterba ret = -ENOSPC; 58649a3c4d9SDavid Sterba goto fail_free; 58749a3c4d9SDavid Sterba } 588e09fe2d2SQu Wenruo 589d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 5909ed74f2dSJosef Bacik /* 591d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 592d5c12070SMiao Xie * of create_snapshot(). 5939ed74f2dSJosef Bacik */ 594c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 595d5c12070SMiao Xie if (ret) 59649a3c4d9SDavid Sterba goto fail_free; 597f46b5a66SChristoph Hellwig 598d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 599d5c12070SMiao Xie if (IS_ERR(trans)) { 600d5c12070SMiao Xie ret = PTR_ERR(trans); 6017775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 60249a3c4d9SDavid Sterba goto fail_free; 603d5c12070SMiao Xie } 604d5c12070SMiao Xie trans->block_rsv = &block_rsv; 605d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 606f46b5a66SChristoph Hellwig 607a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 6086f72c7e2SArne Jansen if (ret) 6096f72c7e2SArne Jansen goto fail; 6106f72c7e2SArne Jansen 6114d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6128e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6138e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6148e8a1e31SJosef Bacik goto fail; 6158e8a1e31SJosef Bacik } 616f46b5a66SChristoph Hellwig 617f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 618f46b5a66SChristoph Hellwig 61949a3c4d9SDavid Sterba inode_item = &root_item->inode; 6203cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6213cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6223cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 623da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6240b246afaSJeff Mahoney fs_info->nodesize); 6253cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 626f46b5a66SChristoph Hellwig 62749a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 62849a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6293cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 63008fe4db1SLi Zefan 63149a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 63249a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 63349a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 63449a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 63549a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 63649a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 637f46b5a66SChristoph Hellwig 63849a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 63949a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6408ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 64149a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 64249a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 64349a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 64449a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 64549a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 64649a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 647f46b5a66SChristoph Hellwig 648925baeddSChris Mason btrfs_tree_unlock(leaf); 649f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 650f46b5a66SChristoph Hellwig leaf = NULL; 651f46b5a66SChristoph Hellwig 65249a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 653f46b5a66SChristoph Hellwig 654f46b5a66SChristoph Hellwig key.objectid = objectid; 6555d4f98a2SYan Zheng key.offset = 0; 656962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6570b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 65849a3c4d9SDavid Sterba root_item); 659f46b5a66SChristoph Hellwig if (ret) 660f46b5a66SChristoph Hellwig goto fail; 661f46b5a66SChristoph Hellwig 66276dda93cSYan, Zheng key.offset = (u64)-1; 6630b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 66479787eaaSJeff Mahoney if (IS_ERR(new_root)) { 66579787eaaSJeff Mahoney ret = PTR_ERR(new_root); 66666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 66779787eaaSJeff Mahoney goto fail; 66879787eaaSJeff Mahoney } 66976dda93cSYan, Zheng 67076dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 67176dda93cSYan, Zheng 67263541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 673ce598979SMark Fasheh if (ret) { 674ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 67566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 676ce598979SMark Fasheh goto fail; 677ce598979SMark Fasheh } 678ce598979SMark Fasheh 679f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 680f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 681f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 682f32e48e9SChandan Rajendra 683f46b5a66SChristoph Hellwig /* 684f46b5a66SChristoph Hellwig * insert the directory item 685f46b5a66SChristoph Hellwig */ 686877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 68779787eaaSJeff Mahoney if (ret) { 68866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 68979787eaaSJeff Mahoney goto fail; 69079787eaaSJeff Mahoney } 6913de4586cSChris Mason 692684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 6933de4586cSChris Mason BTRFS_FT_DIR, index); 69479787eaaSJeff Mahoney if (ret) { 69566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 696f46b5a66SChristoph Hellwig goto fail; 69779787eaaSJeff Mahoney } 6980660b5afSChris Mason 6996ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 70052c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 70152c26179SYan Zheng BUG_ON(ret); 70252c26179SYan Zheng 7036025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7044a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7050660b5afSChris Mason BUG_ON(ret); 7060660b5afSChris Mason 707cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7086bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 709dd5f9615SStefan Behrens if (ret) 71066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 711dd5f9615SStefan Behrens 712f46b5a66SChristoph Hellwig fail: 71349a3c4d9SDavid Sterba kfree(root_item); 714d5c12070SMiao Xie trans->block_rsv = NULL; 715d5c12070SMiao Xie trans->bytes_reserved = 0; 7167775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 717de6e8200SLiu Bo 71872fd032eSSage Weil if (async_transid) { 71972fd032eSSage Weil *async_transid = trans->transid; 7203a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 72100d71c9cSMiao Xie if (err) 7223a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 72372fd032eSSage Weil } else { 7243a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 72572fd032eSSage Weil } 726f46b5a66SChristoph Hellwig if (err && !ret) 727f46b5a66SChristoph Hellwig ret = err; 7281a65e24bSChris Mason 7295662344bSTsutomu Itoh if (!ret) { 7305662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 731de6e8200SLiu Bo if (IS_ERR(inode)) 732de6e8200SLiu Bo return PTR_ERR(inode); 7335662344bSTsutomu Itoh d_instantiate(dentry, inode); 7345662344bSTsutomu Itoh } 735f46b5a66SChristoph Hellwig return ret; 73649a3c4d9SDavid Sterba 73749a3c4d9SDavid Sterba fail_free: 73849a3c4d9SDavid Sterba kfree(root_item); 73949a3c4d9SDavid Sterba return ret; 740f46b5a66SChristoph Hellwig } 741f46b5a66SChristoph Hellwig 742e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 74361d7e4cbSDavid Sterba struct dentry *dentry, 744e9662f70SMiao Xie u64 *async_transid, bool readonly, 745e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 746f46b5a66SChristoph Hellwig { 7470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7482e4bfab9SYan, Zheng struct inode *inode; 749f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 750f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7512e4bfab9SYan, Zheng int ret; 7528ecebf4dSRobbie Ko bool snapshot_force_cow = false; 753f46b5a66SChristoph Hellwig 75427cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 755f46b5a66SChristoph Hellwig return -EINVAL; 756f46b5a66SChristoph Hellwig 75723269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 758a1ee7362SDavid Sterba if (!pending_snapshot) 759a1ee7362SDavid Sterba return -ENOMEM; 760a1ee7362SDavid Sterba 761b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 76223269bf5SDavid Sterba GFP_KERNEL); 7638546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7648546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 765b0c0ea63SDavid Sterba ret = -ENOMEM; 766b0c0ea63SDavid Sterba goto free_pending; 767b0c0ea63SDavid Sterba } 768b0c0ea63SDavid Sterba 7698ecebf4dSRobbie Ko /* 7708ecebf4dSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 7718ecebf4dSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 7728ecebf4dSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 7738ecebf4dSRobbie Ko */ 774ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7754e857c58SPeter Zijlstra smp_mb__after_atomic(); 77645bac0f3SLiu Bo /* wait for no snapshot writes */ 77745bac0f3SLiu Bo wait_event(root->subv_writers->wait, 77845bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 7798257b2dcSMiao Xie 78076f32e24SNikolay Borisov ret = btrfs_start_delalloc_inodes(root); 7816a03843dSMiao Xie if (ret) 782a1ee7362SDavid Sterba goto dec_and_free; 7836a03843dSMiao Xie 7848ecebf4dSRobbie Ko /* 7858ecebf4dSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 7868ecebf4dSRobbie Ko * we force future writes to fallback to COW mode during snapshot 7878ecebf4dSRobbie Ko * creation. 7888ecebf4dSRobbie Ko */ 7898ecebf4dSRobbie Ko atomic_inc(&root->snapshot_force_cow); 7908ecebf4dSRobbie Ko snapshot_force_cow = true; 7918ecebf4dSRobbie Ko 7926374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 7936a03843dSMiao Xie 79466d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 79566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 796d5c12070SMiao Xie /* 797d5c12070SMiao Xie * 1 - parent dir inode 798d5c12070SMiao Xie * 2 - dir entries 799d5c12070SMiao Xie * 1 - root item 800d5c12070SMiao Xie * 2 - root ref/backref 801d5c12070SMiao Xie * 1 - root of snapshot 802dd5f9615SStefan Behrens * 1 - UUID item 803d5c12070SMiao Xie */ 804d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 805dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 806ee3441b4SJeff Mahoney false); 807d5c12070SMiao Xie if (ret) 808a1ee7362SDavid Sterba goto dec_and_free; 809d5c12070SMiao Xie 810a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 811a22285a6SYan, Zheng pending_snapshot->root = root; 812b83cc969SLi Zefan pending_snapshot->readonly = readonly; 813e9662f70SMiao Xie pending_snapshot->dir = dir; 8148696c533SMiao Xie pending_snapshot->inherit = inherit; 815a22285a6SYan, Zheng 816d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 817a22285a6SYan, Zheng if (IS_ERR(trans)) { 818a22285a6SYan, Zheng ret = PTR_ERR(trans); 819a22285a6SYan, Zheng goto fail; 820a22285a6SYan, Zheng } 821a22285a6SYan, Zheng 8220b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 823a22285a6SYan, Zheng list_add(&pending_snapshot->list, 824a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8250b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 82672fd032eSSage Weil if (async_transid) { 82772fd032eSSage Weil *async_transid = trans->transid; 8283a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 82900d71c9cSMiao Xie if (ret) 8303a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 83172fd032eSSage Weil } else { 8323a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 83372fd032eSSage Weil } 834aec8030aSMiao Xie if (ret) 835c37b2b62SJosef Bacik goto fail; 836a22285a6SYan, Zheng 837a22285a6SYan, Zheng ret = pending_snapshot->error; 838f46b5a66SChristoph Hellwig if (ret) 8392e4bfab9SYan, Zheng goto fail; 840f46b5a66SChristoph Hellwig 841d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 842d3797308SChris Mason if (ret) 843d3797308SChris Mason goto fail; 844d3797308SChris Mason 8452b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8462e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8472e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8482e4bfab9SYan, Zheng goto fail; 8492e4bfab9SYan, Zheng } 8505662344bSTsutomu Itoh 8512e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8522e4bfab9SYan, Zheng ret = 0; 8532e4bfab9SYan, Zheng fail: 8547775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 855a1ee7362SDavid Sterba dec_and_free: 8568ecebf4dSRobbie Ko if (snapshot_force_cow) 8578ecebf4dSRobbie Ko atomic_dec(&root->snapshot_force_cow); 858ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8594625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 860b0c0ea63SDavid Sterba free_pending: 861b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8628546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 863a1ee7362SDavid Sterba kfree(pending_snapshot); 864a1ee7362SDavid Sterba 865f46b5a66SChristoph Hellwig return ret; 866f46b5a66SChristoph Hellwig } 867f46b5a66SChristoph Hellwig 8684260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8694260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8704260f7c7SSage Weil * whether the type of victim is right. 8714260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8724260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8734260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8744260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8754260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8764260f7c7SSage Weil * a. be owner of dir, or 8774260f7c7SSage Weil * b. be owner of victim, or 8784260f7c7SSage Weil * c. have CAP_FOWNER capability 87901327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8804260f7c7SSage Weil * links pointing to it. 8814260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8824260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8834260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8844260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8854260f7c7SSage Weil * nfs_async_unlink(). 8864260f7c7SSage Weil */ 8874260f7c7SSage Weil 8884260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 8894260f7c7SSage Weil { 8904260f7c7SSage Weil int error; 8914260f7c7SSage Weil 8922b0143b5SDavid Howells if (d_really_is_negative(victim)) 8934260f7c7SSage Weil return -ENOENT; 8944260f7c7SSage Weil 8952b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8964fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8974260f7c7SSage Weil 8984260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 8994260f7c7SSage Weil if (error) 9004260f7c7SSage Weil return error; 9014260f7c7SSage Weil if (IS_APPEND(dir)) 9024260f7c7SSage Weil return -EPERM; 9032b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9042b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9054260f7c7SSage Weil return -EPERM; 9064260f7c7SSage Weil if (isdir) { 907e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9084260f7c7SSage Weil return -ENOTDIR; 9094260f7c7SSage Weil if (IS_ROOT(victim)) 9104260f7c7SSage Weil return -EBUSY; 911e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9124260f7c7SSage Weil return -EISDIR; 9134260f7c7SSage Weil if (IS_DEADDIR(dir)) 9144260f7c7SSage Weil return -ENOENT; 9154260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9164260f7c7SSage Weil return -EBUSY; 9174260f7c7SSage Weil return 0; 9184260f7c7SSage Weil } 9194260f7c7SSage Weil 920cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 921cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 922cb8e7090SChristoph Hellwig { 9232b0143b5SDavid Howells if (d_really_is_positive(child)) 924cb8e7090SChristoph Hellwig return -EEXIST; 925cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 926cb8e7090SChristoph Hellwig return -ENOENT; 927cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 928cb8e7090SChristoph Hellwig } 929cb8e7090SChristoph Hellwig 930cb8e7090SChristoph Hellwig /* 931cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 932cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 933cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 934cb8e7090SChristoph Hellwig */ 93592872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 93652f75f4fSDavid Sterba const char *name, int namelen, 93772fd032eSSage Weil struct btrfs_root *snap_src, 9386f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9398696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 940cb8e7090SChristoph Hellwig { 9412b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9420b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 943cb8e7090SChristoph Hellwig struct dentry *dentry; 944cb8e7090SChristoph Hellwig int error; 945cb8e7090SChristoph Hellwig 94600235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 94700235411SAl Viro if (error == -EINTR) 94800235411SAl Viro return error; 949cb8e7090SChristoph Hellwig 950cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 951cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 952cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 953cb8e7090SChristoph Hellwig goto out_unlock; 954cb8e7090SChristoph Hellwig 95576dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 956cb8e7090SChristoph Hellwig if (error) 957a874a63eSLiu Bo goto out_dput; 958cb8e7090SChristoph Hellwig 9599c52057cSChris Mason /* 9609c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9619c52057cSChris Mason * check for them now when we can safely fail 9629c52057cSChris Mason */ 9639c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9649c52057cSChris Mason dir->i_ino, name, 9659c52057cSChris Mason namelen); 9669c52057cSChris Mason if (error) 9679c52057cSChris Mason goto out_dput; 9689c52057cSChris Mason 9690b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 97076dda93cSYan, Zheng 97176dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 97276dda93cSYan, Zheng goto out_up_read; 97376dda93cSYan, Zheng 9743de4586cSChris Mason if (snap_src) { 97561d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9766f72c7e2SArne Jansen async_transid, readonly, inherit); 9773de4586cSChris Mason } else { 978d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 979d5c12070SMiao Xie async_transid, inherit); 9803de4586cSChris Mason } 98176dda93cSYan, Zheng if (!error) 98276dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 98376dda93cSYan, Zheng out_up_read: 9840b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 985cb8e7090SChristoph Hellwig out_dput: 986cb8e7090SChristoph Hellwig dput(dentry); 987cb8e7090SChristoph Hellwig out_unlock: 9885955102cSAl Viro inode_unlock(dir); 989cb8e7090SChristoph Hellwig return error; 990cb8e7090SChristoph Hellwig } 991cb8e7090SChristoph Hellwig 9924cb5300bSChris Mason /* 9934cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 9944cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 9954cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 9964cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 9974cb5300bSChris Mason * part of the file 9984cb5300bSChris Mason */ 999aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 10004cb5300bSChris Mason { 10014cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10024cb5300bSChris Mason struct extent_map *em = NULL; 10034cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10044cb5300bSChris Mason u64 end; 10054cb5300bSChris Mason 10064cb5300bSChris Mason read_lock(&em_tree->lock); 100709cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10084cb5300bSChris Mason read_unlock(&em_tree->lock); 10094cb5300bSChris Mason 10104cb5300bSChris Mason if (em) { 10114cb5300bSChris Mason end = extent_map_end(em); 10124cb5300bSChris Mason free_extent_map(em); 10134cb5300bSChris Mason if (end - offset > thresh) 10144cb5300bSChris Mason return 0; 10154cb5300bSChris Mason } 10164cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10174cb5300bSChris Mason thresh /= 2; 10184cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10194cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10204cb5300bSChris Mason if (end >= thresh) 10214cb5300bSChris Mason return 0; 10224cb5300bSChris Mason return 1; 10234cb5300bSChris Mason } 10244cb5300bSChris Mason 10254cb5300bSChris Mason /* 10264cb5300bSChris Mason * helper function to walk through a file and find extents 10274cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10284cb5300bSChris Mason * 10294cb5300bSChris Mason * This is used by the defragging code to find new and small 10304cb5300bSChris Mason * extents 10314cb5300bSChris Mason */ 10324cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10334cb5300bSChris Mason struct inode *inode, u64 newer_than, 1034aab110abSDavid Sterba u64 *off, u32 thresh) 10354cb5300bSChris Mason { 10364cb5300bSChris Mason struct btrfs_path *path; 10374cb5300bSChris Mason struct btrfs_key min_key; 10384cb5300bSChris Mason struct extent_buffer *leaf; 10394cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10404cb5300bSChris Mason int type; 10414cb5300bSChris Mason int ret; 10424a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10434cb5300bSChris Mason 10444cb5300bSChris Mason path = btrfs_alloc_path(); 10454cb5300bSChris Mason if (!path) 10464cb5300bSChris Mason return -ENOMEM; 10474cb5300bSChris Mason 1048a4689d2bSDavid Sterba min_key.objectid = ino; 10494cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10504cb5300bSChris Mason min_key.offset = *off; 10514cb5300bSChris Mason 10524cb5300bSChris Mason while (1) { 10536174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10544cb5300bSChris Mason if (ret != 0) 10554cb5300bSChris Mason goto none; 1056f094c9bdSFilipe Manana process_slot: 1057a4689d2bSDavid Sterba if (min_key.objectid != ino) 10584cb5300bSChris Mason goto none; 10594cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10604cb5300bSChris Mason goto none; 10614cb5300bSChris Mason 10624cb5300bSChris Mason leaf = path->nodes[0]; 10634cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10644cb5300bSChris Mason struct btrfs_file_extent_item); 10654cb5300bSChris Mason 10664cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10674cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10684cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10694cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10704cb5300bSChris Mason *off = min_key.offset; 10714cb5300bSChris Mason btrfs_free_path(path); 10724cb5300bSChris Mason return 0; 10734cb5300bSChris Mason } 10744cb5300bSChris Mason 1075f094c9bdSFilipe Manana path->slots[0]++; 1076f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1077f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1078f094c9bdSFilipe Manana goto process_slot; 1079f094c9bdSFilipe Manana } 1080f094c9bdSFilipe Manana 10814cb5300bSChris Mason if (min_key.offset == (u64)-1) 10824cb5300bSChris Mason goto none; 10834cb5300bSChris Mason 10844cb5300bSChris Mason min_key.offset++; 10854cb5300bSChris Mason btrfs_release_path(path); 10864cb5300bSChris Mason } 10874cb5300bSChris Mason none: 10884cb5300bSChris Mason btrfs_free_path(path); 10894cb5300bSChris Mason return -ENOENT; 10904cb5300bSChris Mason } 10914cb5300bSChris Mason 10926c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 109317ce6ef8SLiu Bo { 109417ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1095940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10966c282eb4SLi Zefan struct extent_map *em; 109709cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1098940100a4SChris Mason 1099940100a4SChris Mason /* 1100940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1101940100a4SChris Mason * the full extent lock 1102940100a4SChris Mason */ 1103940100a4SChris Mason read_lock(&em_tree->lock); 1104940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1105940100a4SChris Mason read_unlock(&em_tree->lock); 1106940100a4SChris Mason 1107940100a4SChris Mason if (!em) { 1108308d9800SFilipe Manana struct extent_state *cached = NULL; 1109308d9800SFilipe Manana u64 end = start + len - 1; 1110308d9800SFilipe Manana 1111940100a4SChris Mason /* get the big lock and read metadata off disk */ 1112ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1113fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1114e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1115940100a4SChris Mason 11166cf8bfbfSDan Carpenter if (IS_ERR(em)) 11176c282eb4SLi Zefan return NULL; 1118940100a4SChris Mason } 1119940100a4SChris Mason 11206c282eb4SLi Zefan return em; 11216c282eb4SLi Zefan } 11226c282eb4SLi Zefan 11236c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11246c282eb4SLi Zefan { 11256c282eb4SLi Zefan struct extent_map *next; 11266c282eb4SLi Zefan bool ret = true; 11276c282eb4SLi Zefan 11286c282eb4SLi Zefan /* this is the last extent */ 11296c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11306c282eb4SLi Zefan return false; 11316c282eb4SLi Zefan 11326c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1133e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1134e9512d72SChris Mason ret = false; 1135e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1136ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11376c282eb4SLi Zefan ret = false; 11386c282eb4SLi Zefan 11396c282eb4SLi Zefan free_extent_map(next); 11406c282eb4SLi Zefan return ret; 11416c282eb4SLi Zefan } 11426c282eb4SLi Zefan 1143aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1144a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1145a43a2111SAndrew Mahone int compress) 11466c282eb4SLi Zefan { 11476c282eb4SLi Zefan struct extent_map *em; 11486c282eb4SLi Zefan int ret = 1; 11496c282eb4SLi Zefan bool next_mergeable = true; 11504a3560c4SLiu Bo bool prev_mergeable = true; 11516c282eb4SLi Zefan 11526c282eb4SLi Zefan /* 11536c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11546c282eb4SLi Zefan * defragging it 11556c282eb4SLi Zefan */ 11566c282eb4SLi Zefan if (start < *defrag_end) 11576c282eb4SLi Zefan return 1; 11586c282eb4SLi Zefan 11596c282eb4SLi Zefan *skip = 0; 11606c282eb4SLi Zefan 11616c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11626c282eb4SLi Zefan if (!em) 11636c282eb4SLi Zefan return 0; 11646c282eb4SLi Zefan 1165940100a4SChris Mason /* this will cover holes, and inline extents */ 116617ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1167940100a4SChris Mason ret = 0; 116817ce6ef8SLiu Bo goto out; 116917ce6ef8SLiu Bo } 117017ce6ef8SLiu Bo 11714a3560c4SLiu Bo if (!*defrag_end) 11724a3560c4SLiu Bo prev_mergeable = false; 11734a3560c4SLiu Bo 11746c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1175940100a4SChris Mason /* 11766c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11776c282eb4SLi Zefan * real extent, don't bother defragging it 1178940100a4SChris Mason */ 1179a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11804a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1181940100a4SChris Mason ret = 0; 118217ce6ef8SLiu Bo out: 1183940100a4SChris Mason /* 1184940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1185940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1186940100a4SChris Mason * so that the next tiny extent will force a defrag. 1187940100a4SChris Mason * 1188940100a4SChris Mason * The end result of this is that tiny extents before a single big 1189940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1190940100a4SChris Mason */ 1191940100a4SChris Mason if (ret) { 1192940100a4SChris Mason *defrag_end = extent_map_end(em); 1193940100a4SChris Mason } else { 1194940100a4SChris Mason *last_len = 0; 1195940100a4SChris Mason *skip = extent_map_end(em); 1196940100a4SChris Mason *defrag_end = 0; 1197940100a4SChris Mason } 1198940100a4SChris Mason 1199940100a4SChris Mason free_extent_map(em); 1200940100a4SChris Mason return ret; 1201940100a4SChris Mason } 1202940100a4SChris Mason 12034cb5300bSChris Mason /* 12044cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12054cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12064cb5300bSChris Mason * to COW and defrag. 12074cb5300bSChris Mason * 12084cb5300bSChris Mason * It also makes sure the delalloc code has enough 12094cb5300bSChris Mason * dirty data to avoid making new small extents as part 12104cb5300bSChris Mason * of the defrag 12114cb5300bSChris Mason * 12124cb5300bSChris Mason * It's a good idea to start RA on this range 12134cb5300bSChris Mason * before calling this. 12144cb5300bSChris Mason */ 12154cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12164cb5300bSChris Mason struct page **pages, 12174cb5300bSChris Mason unsigned long start_index, 1218c41570c9SJustin Maggard unsigned long num_pages) 1219f46b5a66SChristoph Hellwig { 12204cb5300bSChris Mason unsigned long file_end; 12214cb5300bSChris Mason u64 isize = i_size_read(inode); 1222f46b5a66SChristoph Hellwig u64 page_start; 1223f46b5a66SChristoph Hellwig u64 page_end; 12241f12bd06SLiu Bo u64 page_cnt; 12254cb5300bSChris Mason int ret; 12264cb5300bSChris Mason int i; 12274cb5300bSChris Mason int i_done; 12284cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12294cb5300bSChris Mason struct extent_state *cached_state = NULL; 1230600a45e1SMiao Xie struct extent_io_tree *tree; 1231364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12323b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12334cb5300bSChris Mason 123409cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12351f12bd06SLiu Bo if (!isize || start_index > file_end) 12361f12bd06SLiu Bo return 0; 12371f12bd06SLiu Bo 12381f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12394cb5300bSChris Mason 1240364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 124109cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 124209cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12434cb5300bSChris Mason if (ret) 12444cb5300bSChris Mason return ret; 12454cb5300bSChris Mason i_done = 0; 1246600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12474cb5300bSChris Mason 12484cb5300bSChris Mason /* step one, lock all the pages */ 12491f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12504cb5300bSChris Mason struct page *page; 1251600a45e1SMiao Xie again: 1252a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12533b16a4e3SJosef Bacik start_index + i, mask); 12544cb5300bSChris Mason if (!page) 12554cb5300bSChris Mason break; 12564cb5300bSChris Mason 1257600a45e1SMiao Xie page_start = page_offset(page); 125809cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1259600a45e1SMiao Xie while (1) { 1260308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1261ff13db41SDavid Sterba &cached_state); 1262600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1263600a45e1SMiao Xie page_start); 1264308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1265e43bbe5eSDavid Sterba &cached_state); 1266600a45e1SMiao Xie if (!ordered) 1267600a45e1SMiao Xie break; 1268600a45e1SMiao Xie 1269600a45e1SMiao Xie unlock_page(page); 1270600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1271600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1272600a45e1SMiao Xie lock_page(page); 12731f12bd06SLiu Bo /* 12741f12bd06SLiu Bo * we unlocked the page above, so we need check if 12751f12bd06SLiu Bo * it was released or not. 12761f12bd06SLiu Bo */ 12771f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12781f12bd06SLiu Bo unlock_page(page); 127909cbfeafSKirill A. Shutemov put_page(page); 12801f12bd06SLiu Bo goto again; 12811f12bd06SLiu Bo } 1282600a45e1SMiao Xie } 1283600a45e1SMiao Xie 12844cb5300bSChris Mason if (!PageUptodate(page)) { 12854cb5300bSChris Mason btrfs_readpage(NULL, page); 12864cb5300bSChris Mason lock_page(page); 12874cb5300bSChris Mason if (!PageUptodate(page)) { 12884cb5300bSChris Mason unlock_page(page); 128909cbfeafSKirill A. Shutemov put_page(page); 12904cb5300bSChris Mason ret = -EIO; 12914cb5300bSChris Mason break; 12924cb5300bSChris Mason } 12934cb5300bSChris Mason } 1294600a45e1SMiao Xie 1295600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1296600a45e1SMiao Xie unlock_page(page); 129709cbfeafSKirill A. Shutemov put_page(page); 1298600a45e1SMiao Xie goto again; 1299600a45e1SMiao Xie } 1300600a45e1SMiao Xie 13014cb5300bSChris Mason pages[i] = page; 13024cb5300bSChris Mason i_done++; 13034cb5300bSChris Mason } 13044cb5300bSChris Mason if (!i_done || ret) 13054cb5300bSChris Mason goto out; 13064cb5300bSChris Mason 13071751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13084cb5300bSChris Mason goto out; 13094cb5300bSChris Mason 13104cb5300bSChris Mason /* 13114cb5300bSChris Mason * so now we have a nice long stream of locked 13124cb5300bSChris Mason * and up to date pages, lets wait on them 13134cb5300bSChris Mason */ 13144cb5300bSChris Mason for (i = 0; i < i_done; i++) 13154cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13164cb5300bSChris Mason 13174cb5300bSChris Mason page_start = page_offset(pages[0]); 131809cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13194cb5300bSChris Mason 13204cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1321ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13224cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 13234cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 13249e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1325ae0f1625SDavid Sterba &cached_state); 13264cb5300bSChris Mason 13271f12bd06SLiu Bo if (i_done != page_cnt) { 13289e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 132928c4a3e2SSu Yue btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); 13309e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1331bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 133209cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 133343b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13344cb5300bSChris Mason } 13354cb5300bSChris Mason 13364cb5300bSChris Mason 13379e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1338018ed4f7SDavid Sterba &cached_state); 13394cb5300bSChris Mason 13404cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1341e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13424cb5300bSChris Mason 13434cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13444cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13454cb5300bSChris Mason ClearPageChecked(pages[i]); 13464cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13474cb5300bSChris Mason set_page_dirty(pages[i]); 13484cb5300bSChris Mason unlock_page(pages[i]); 134909cbfeafSKirill A. Shutemov put_page(pages[i]); 13504cb5300bSChris Mason } 135143b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 135243b18595SQu Wenruo false); 1353364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13544cb5300bSChris Mason return i_done; 13554cb5300bSChris Mason out: 13564cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13574cb5300bSChris Mason unlock_page(pages[i]); 135809cbfeafSKirill A. Shutemov put_page(pages[i]); 13594cb5300bSChris Mason } 1360bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 136109cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 136243b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 136343b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 136443b18595SQu Wenruo true); 1365364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13664cb5300bSChris Mason return ret; 13674cb5300bSChris Mason 13684cb5300bSChris Mason } 13694cb5300bSChris Mason 13704cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13714cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13724cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13734cb5300bSChris Mason { 13740b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13754cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13764cb5300bSChris Mason struct file_ra_state *ra = NULL; 13774cb5300bSChris Mason unsigned long last_index; 1378151a31b2SLi Zefan u64 isize = i_size_read(inode); 1379940100a4SChris Mason u64 last_len = 0; 1380940100a4SChris Mason u64 skip = 0; 1381940100a4SChris Mason u64 defrag_end = 0; 13824cb5300bSChris Mason u64 newer_off = range->start; 1383f46b5a66SChristoph Hellwig unsigned long i; 1384008873eaSLi Zefan unsigned long ra_index = 0; 1385f46b5a66SChristoph Hellwig int ret; 13864cb5300bSChris Mason int defrag_count = 0; 13871a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1388aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 138909cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1390c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1391ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 13924cb5300bSChris Mason struct page **pages = NULL; 13931e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 13944cb5300bSChris Mason 13950abd5b17SLiu Bo if (isize == 0) 13960abd5b17SLiu Bo return 0; 13970abd5b17SLiu Bo 13980abd5b17SLiu Bo if (range->start >= isize) 13990abd5b17SLiu Bo return -EINVAL; 14001a419d85SLi Zefan 14011e2ef46dSDavid Sterba if (do_compress) { 14021a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 14031a419d85SLi Zefan return -EINVAL; 14041a419d85SLi Zefan if (range->compress_type) 14051a419d85SLi Zefan compress_type = range->compress_type; 14061a419d85SLi Zefan } 1407f46b5a66SChristoph Hellwig 14080abd5b17SLiu Bo if (extent_thresh == 0) 1409ee22184bSByongho Lee extent_thresh = SZ_256K; 1410f46b5a66SChristoph Hellwig 14114cb5300bSChris Mason /* 14120a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14130a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14140a52d108SDavid Sterba * we don't error out. 14154cb5300bSChris Mason */ 14164cb5300bSChris Mason if (!file) { 141763e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14180a52d108SDavid Sterba if (ra) 14194cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14204cb5300bSChris Mason } else { 14214cb5300bSChris Mason ra = &file->f_ra; 14224cb5300bSChris Mason } 14234cb5300bSChris Mason 142463e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14254cb5300bSChris Mason if (!pages) { 14264cb5300bSChris Mason ret = -ENOMEM; 14274cb5300bSChris Mason goto out_ra; 14284cb5300bSChris Mason } 14294cb5300bSChris Mason 14304cb5300bSChris Mason /* find the last page to defrag */ 14311e701a32SChris Mason if (range->start + range->len > range->start) { 1432151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 143309cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14341e701a32SChris Mason } else { 143509cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14361e701a32SChris Mason } 14371e701a32SChris Mason 14384cb5300bSChris Mason if (newer_than) { 14394cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1440ee22184bSByongho Lee &newer_off, SZ_64K); 14414cb5300bSChris Mason if (!ret) { 14424cb5300bSChris Mason range->start = newer_off; 14434cb5300bSChris Mason /* 14444cb5300bSChris Mason * we always align our defrag to help keep 14454cb5300bSChris Mason * the extents in the file evenly spaced 14464cb5300bSChris Mason */ 144709cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14484cb5300bSChris Mason } else 14494cb5300bSChris Mason goto out_ra; 14504cb5300bSChris Mason } else { 145109cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14524cb5300bSChris Mason } 14534cb5300bSChris Mason if (!max_to_defrag) 1454070034bdSchandan max_to_defrag = last_index - i + 1; 14554cb5300bSChris Mason 14562a0f7f57SLi Zefan /* 14572a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14582a0f7f57SLi Zefan * written sequentially. 14592a0f7f57SLi Zefan */ 14602a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14612a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14622a0f7f57SLi Zefan 1463f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 146409cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14654cb5300bSChris Mason /* 14664cb5300bSChris Mason * make sure we stop running if someone unmounts 14674cb5300bSChris Mason * the FS 14684cb5300bSChris Mason */ 14691751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14704cb5300bSChris Mason break; 14714cb5300bSChris Mason 14720b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14730b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1474210549ebSDavid Sterba ret = -EAGAIN; 1475210549ebSDavid Sterba break; 1476210549ebSDavid Sterba } 1477210549ebSDavid Sterba 147809cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14796c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14801e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1481940100a4SChris Mason unsigned long next; 1482940100a4SChris Mason /* 1483940100a4SChris Mason * the should_defrag function tells us how much to skip 1484940100a4SChris Mason * bump our counter by the suggested amount 1485940100a4SChris Mason */ 148609cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1487940100a4SChris Mason i = max(i + 1, next); 1488940100a4SChris Mason continue; 1489940100a4SChris Mason } 1490008873eaSLi Zefan 1491008873eaSLi Zefan if (!newer_than) { 149209cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 149309cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1494008873eaSLi Zefan cluster = min(cluster, max_cluster); 1495008873eaSLi Zefan } else { 1496008873eaSLi Zefan cluster = max_cluster; 1497008873eaSLi Zefan } 1498008873eaSLi Zefan 1499008873eaSLi Zefan if (i + cluster > ra_index) { 1500008873eaSLi Zefan ra_index = max(i, ra_index); 15010a52d108SDavid Sterba if (ra) 1502d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1503d3c0bab5SDavid Sterba file, ra_index, cluster); 1504e4826a5bSchandan ra_index += cluster; 1505008873eaSLi Zefan } 15064cb5300bSChris Mason 15075955102cSAl Viro inode_lock(inode); 15081e2ef46dSDavid Sterba if (do_compress) 1509eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1510008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1511ecb8bea8SLiu Bo if (ret < 0) { 15125955102cSAl Viro inode_unlock(inode); 15134cb5300bSChris Mason goto out_ra; 1514ecb8bea8SLiu Bo } 15154cb5300bSChris Mason 15164cb5300bSChris Mason defrag_count += ret; 1517d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15185955102cSAl Viro inode_unlock(inode); 15194cb5300bSChris Mason 15204cb5300bSChris Mason if (newer_than) { 15214cb5300bSChris Mason if (newer_off == (u64)-1) 15224cb5300bSChris Mason break; 15234cb5300bSChris Mason 1524e1f041e1SLiu Bo if (ret > 0) 1525e1f041e1SLiu Bo i += ret; 1526e1f041e1SLiu Bo 15274cb5300bSChris Mason newer_off = max(newer_off + 1, 152809cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15294cb5300bSChris Mason 1530ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1531ee22184bSByongho Lee &newer_off, SZ_64K); 15324cb5300bSChris Mason if (!ret) { 15334cb5300bSChris Mason range->start = newer_off; 153409cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15354cb5300bSChris Mason } else { 15364cb5300bSChris Mason break; 1537940100a4SChris Mason } 15384cb5300bSChris Mason } else { 1539008873eaSLi Zefan if (ret > 0) { 1540cbcc8326SLi Zefan i += ret; 154109cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1542008873eaSLi Zefan } else { 1543940100a4SChris Mason i++; 1544008873eaSLi Zefan last_len = 0; 1545008873eaSLi Zefan } 1546f46b5a66SChristoph Hellwig } 15474cb5300bSChris Mason } 1548f46b5a66SChristoph Hellwig 1549dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15501e701a32SChris Mason filemap_flush(inode->i_mapping); 1551dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1552dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1553dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1554dec8ef90SFilipe Manana } 15551e701a32SChris Mason 15561a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15570b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15585c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15595c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15601a419d85SLi Zefan } 15611a419d85SLi Zefan 156260ccf82fSDiego Calleja ret = defrag_count; 1563940100a4SChris Mason 15644cb5300bSChris Mason out_ra: 15651e2ef46dSDavid Sterba if (do_compress) { 15665955102cSAl Viro inode_lock(inode); 1567eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15685955102cSAl Viro inode_unlock(inode); 1569633085c7SFilipe David Borba Manana } 15704cb5300bSChris Mason if (!file) 15714cb5300bSChris Mason kfree(ra); 15724cb5300bSChris Mason kfree(pages); 1573940100a4SChris Mason return ret; 1574f46b5a66SChristoph Hellwig } 1575f46b5a66SChristoph Hellwig 1576198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 157776dda93cSYan, Zheng void __user *arg) 1578f46b5a66SChristoph Hellwig { 15790b246afaSJeff Mahoney struct inode *inode = file_inode(file); 15800b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1581f46b5a66SChristoph Hellwig u64 new_size; 1582f46b5a66SChristoph Hellwig u64 old_size; 1583f46b5a66SChristoph Hellwig u64 devid = 1; 15840b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1585f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1586f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1587f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1588f46b5a66SChristoph Hellwig char *sizestr; 15899a40f122SGui Hecheng char *retptr; 1590f46b5a66SChristoph Hellwig char *devstr = NULL; 1591f46b5a66SChristoph Hellwig int ret = 0; 1592f46b5a66SChristoph Hellwig int mod = 0; 1593f46b5a66SChristoph Hellwig 1594e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1595e441d54dSChris Mason return -EPERM; 1596e441d54dSChris Mason 1597198605a8SMiao Xie ret = mnt_want_write_file(file); 1598198605a8SMiao Xie if (ret) 1599198605a8SMiao Xie return ret; 1600198605a8SMiao Xie 1601171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 160297547676SMiao Xie mnt_drop_write_file(file); 1603e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1604c9e9f97bSIlya Dryomov } 1605c9e9f97bSIlya Dryomov 1606dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1607c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1608c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1609c9e9f97bSIlya Dryomov goto out; 1610c9e9f97bSIlya Dryomov } 16115516e595SMark Fasheh 16125516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1613f46b5a66SChristoph Hellwig 1614f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1615f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1616f46b5a66SChristoph Hellwig if (devstr) { 1617f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1618f46b5a66SChristoph Hellwig *devstr = '\0'; 1619f46b5a66SChristoph Hellwig devstr = vol_args->name; 162058dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 162158dfae63SZhangZhen if (ret) 162258dfae63SZhangZhen goto out_free; 1623dfd79829SMiao Xie if (!devid) { 1624dfd79829SMiao Xie ret = -EINVAL; 1625dfd79829SMiao Xie goto out_free; 1626dfd79829SMiao Xie } 16270b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1628f46b5a66SChristoph Hellwig } 1629dba60f3fSMiao Xie 16300b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1631f46b5a66SChristoph Hellwig if (!device) { 16320b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1633c1c9ff7cSGeert Uytterhoeven devid); 1634dfd79829SMiao Xie ret = -ENODEV; 1635c9e9f97bSIlya Dryomov goto out_free; 1636f46b5a66SChristoph Hellwig } 1637dba60f3fSMiao Xie 1638ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16390b246afaSJeff Mahoney btrfs_info(fs_info, 1640efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1641c1c9ff7cSGeert Uytterhoeven devid); 1642dfd79829SMiao Xie ret = -EPERM; 16434e42ae1bSLiu Bo goto out_free; 16444e42ae1bSLiu Bo } 16454e42ae1bSLiu Bo 1646f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1647f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1648f46b5a66SChristoph Hellwig else { 1649f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1650f46b5a66SChristoph Hellwig mod = -1; 1651f46b5a66SChristoph Hellwig sizestr++; 1652f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1653f46b5a66SChristoph Hellwig mod = 1; 1654f46b5a66SChristoph Hellwig sizestr++; 1655f46b5a66SChristoph Hellwig } 16569a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16579a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1658f46b5a66SChristoph Hellwig ret = -EINVAL; 1659c9e9f97bSIlya Dryomov goto out_free; 1660f46b5a66SChristoph Hellwig } 1661f46b5a66SChristoph Hellwig } 1662f46b5a66SChristoph Hellwig 1663401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1664dfd79829SMiao Xie ret = -EPERM; 166563a212abSStefan Behrens goto out_free; 166663a212abSStefan Behrens } 166763a212abSStefan Behrens 16687cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1669f46b5a66SChristoph Hellwig 1670f46b5a66SChristoph Hellwig if (mod < 0) { 1671f46b5a66SChristoph Hellwig if (new_size > old_size) { 1672f46b5a66SChristoph Hellwig ret = -EINVAL; 1673c9e9f97bSIlya Dryomov goto out_free; 1674f46b5a66SChristoph Hellwig } 1675f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1676f46b5a66SChristoph Hellwig } else if (mod > 0) { 1677eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1678902c68a4SGui Hecheng ret = -ERANGE; 1679eb8052e0SWenliang Fan goto out_free; 1680eb8052e0SWenliang Fan } 1681f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1682f46b5a66SChristoph Hellwig } 1683f46b5a66SChristoph Hellwig 1684ee22184bSByongho Lee if (new_size < SZ_256M) { 1685f46b5a66SChristoph Hellwig ret = -EINVAL; 1686c9e9f97bSIlya Dryomov goto out_free; 1687f46b5a66SChristoph Hellwig } 1688f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1689f46b5a66SChristoph Hellwig ret = -EFBIG; 1690c9e9f97bSIlya Dryomov goto out_free; 1691f46b5a66SChristoph Hellwig } 1692f46b5a66SChristoph Hellwig 169347f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1694f46b5a66SChristoph Hellwig 16950b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1696c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1697f46b5a66SChristoph Hellwig 1698f46b5a66SChristoph Hellwig if (new_size > old_size) { 1699a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 170098d5dc13STsutomu Itoh if (IS_ERR(trans)) { 170198d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1702c9e9f97bSIlya Dryomov goto out_free; 170398d5dc13STsutomu Itoh } 1704f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17053a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1706ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1707f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17080253f40eSjeff.liu } /* equal, nothing need to do */ 1709f46b5a66SChristoph Hellwig 1710c9e9f97bSIlya Dryomov out_free: 1711f46b5a66SChristoph Hellwig kfree(vol_args); 1712c9e9f97bSIlya Dryomov out: 1713171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 171418f39c41SIlya Dryomov mnt_drop_write_file(file); 1715f46b5a66SChristoph Hellwig return ret; 1716f46b5a66SChristoph Hellwig } 1717f46b5a66SChristoph Hellwig 171872fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 171952f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17206f72c7e2SArne Jansen u64 *transid, bool readonly, 17218696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1722f46b5a66SChristoph Hellwig { 1723f46b5a66SChristoph Hellwig int namelen; 17243de4586cSChris Mason int ret = 0; 1725f46b5a66SChristoph Hellwig 1726325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1727325c50e3SJeff Mahoney return -ENOTDIR; 1728325c50e3SJeff Mahoney 1729a874a63eSLiu Bo ret = mnt_want_write_file(file); 1730a874a63eSLiu Bo if (ret) 1731a874a63eSLiu Bo goto out; 1732a874a63eSLiu Bo 173372fd032eSSage Weil namelen = strlen(name); 173472fd032eSSage Weil if (strchr(name, '/')) { 1735f46b5a66SChristoph Hellwig ret = -EINVAL; 1736a874a63eSLiu Bo goto out_drop_write; 1737f46b5a66SChristoph Hellwig } 1738f46b5a66SChristoph Hellwig 173916780cabSChris Mason if (name[0] == '.' && 174016780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 174116780cabSChris Mason ret = -EEXIST; 1742a874a63eSLiu Bo goto out_drop_write; 174316780cabSChris Mason } 174416780cabSChris Mason 17453de4586cSChris Mason if (subvol) { 174672fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17476f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1748cb8e7090SChristoph Hellwig } else { 17492903ff01SAl Viro struct fd src = fdget(fd); 17503de4586cSChris Mason struct inode *src_inode; 17512903ff01SAl Viro if (!src.file) { 17523de4586cSChris Mason ret = -EINVAL; 1753a874a63eSLiu Bo goto out_drop_write; 17543de4586cSChris Mason } 17553de4586cSChris Mason 1756496ad9aaSAl Viro src_inode = file_inode(src.file); 1757496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1758c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1759efe120a0SFrank Holton "Snapshot src from another FS"); 176023ad5b17SKusanagi Kouichi ret = -EXDEV; 1761d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1762d0242061SDavid Sterba /* 1763d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1764d0242061SDavid Sterba * are limited to own subvolumes only 1765d0242061SDavid Sterba */ 1766d0242061SDavid Sterba ret = -EPERM; 1767ecd18815SAl Viro } else { 176872fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 176972fd032eSSage Weil BTRFS_I(src_inode)->root, 17706f72c7e2SArne Jansen transid, readonly, inherit); 1771ecd18815SAl Viro } 17722903ff01SAl Viro fdput(src); 1773cb8e7090SChristoph Hellwig } 1774a874a63eSLiu Bo out_drop_write: 1775a874a63eSLiu Bo mnt_drop_write_file(file); 1776f46b5a66SChristoph Hellwig out: 177772fd032eSSage Weil return ret; 177872fd032eSSage Weil } 177972fd032eSSage Weil 178072fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1781fa0d2b9bSLi Zefan void __user *arg, int subvol) 178272fd032eSSage Weil { 1783fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 178472fd032eSSage Weil int ret; 178572fd032eSSage Weil 1786325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1787325c50e3SJeff Mahoney return -ENOTDIR; 1788325c50e3SJeff Mahoney 1789fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1790fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1791fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1792fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1793fa0d2b9bSLi Zefan 1794fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1795b83cc969SLi Zefan vol_args->fd, subvol, 17966f72c7e2SArne Jansen NULL, false, NULL); 1797fa0d2b9bSLi Zefan 1798fa0d2b9bSLi Zefan kfree(vol_args); 1799fa0d2b9bSLi Zefan return ret; 1800fa0d2b9bSLi Zefan } 1801fa0d2b9bSLi Zefan 1802fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1803fa0d2b9bSLi Zefan void __user *arg, int subvol) 1804fa0d2b9bSLi Zefan { 1805fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1806fa0d2b9bSLi Zefan int ret; 1807fdfb1e4fSLi Zefan u64 transid = 0; 1808fdfb1e4fSLi Zefan u64 *ptr = NULL; 1809b83cc969SLi Zefan bool readonly = false; 18106f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 181172fd032eSSage Weil 1812325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1813325c50e3SJeff Mahoney return -ENOTDIR; 1814325c50e3SJeff Mahoney 1815fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1816fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1817fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1818fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1819fdfb1e4fSLi Zefan 1820b83cc969SLi Zefan if (vol_args->flags & 18216f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18226f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1823b83cc969SLi Zefan ret = -EOPNOTSUPP; 1824c47ca32dSDan Carpenter goto free_args; 1825fdfb1e4fSLi Zefan } 1826fdfb1e4fSLi Zefan 1827fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1828fdfb1e4fSLi Zefan ptr = &transid; 1829b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1830b83cc969SLi Zefan readonly = true; 18316f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 183209cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18336f72c7e2SArne Jansen ret = -EINVAL; 1834c47ca32dSDan Carpenter goto free_args; 18356f72c7e2SArne Jansen } 18366f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18376f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18386f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1839c47ca32dSDan Carpenter goto free_args; 18406f72c7e2SArne Jansen } 18416f72c7e2SArne Jansen } 184275eaa0e2SSage Weil 1843fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18446f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18458696c533SMiao Xie readonly, inherit); 1846c47ca32dSDan Carpenter if (ret) 1847c47ca32dSDan Carpenter goto free_inherit; 184875eaa0e2SSage Weil 1849c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1850fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1851c47ca32dSDan Carpenter transid), 1852c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 185375eaa0e2SSage Weil ret = -EFAULT; 1854c47ca32dSDan Carpenter 1855c47ca32dSDan Carpenter free_inherit: 18566f72c7e2SArne Jansen kfree(inherit); 1857c47ca32dSDan Carpenter free_args: 1858c47ca32dSDan Carpenter kfree(vol_args); 1859f46b5a66SChristoph Hellwig return ret; 1860f46b5a66SChristoph Hellwig } 1861f46b5a66SChristoph Hellwig 18620caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18630caa102dSLi Zefan void __user *arg) 18640caa102dSLi Zefan { 1865496ad9aaSAl Viro struct inode *inode = file_inode(file); 18660b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18670caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18680caa102dSLi Zefan int ret = 0; 18690caa102dSLi Zefan u64 flags = 0; 18700caa102dSLi Zefan 18714a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18720caa102dSLi Zefan return -EINVAL; 18730caa102dSLi Zefan 18740b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 18750caa102dSLi Zefan if (btrfs_root_readonly(root)) 18760caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 18770b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 18780caa102dSLi Zefan 18790caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 18800caa102dSLi Zefan ret = -EFAULT; 18810caa102dSLi Zefan 18820caa102dSLi Zefan return ret; 18830caa102dSLi Zefan } 18840caa102dSLi Zefan 18850caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 18860caa102dSLi Zefan void __user *arg) 18870caa102dSLi Zefan { 1888496ad9aaSAl Viro struct inode *inode = file_inode(file); 18890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18900caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18910caa102dSLi Zefan struct btrfs_trans_handle *trans; 18920caa102dSLi Zefan u64 root_flags; 18930caa102dSLi Zefan u64 flags; 18940caa102dSLi Zefan int ret = 0; 18950caa102dSLi Zefan 1896bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1897bd60ea0fSDavid Sterba return -EPERM; 1898bd60ea0fSDavid Sterba 1899b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1900b9ca0664SLiu Bo if (ret) 1901b9ca0664SLiu Bo goto out; 19020caa102dSLi Zefan 19034a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1904b9ca0664SLiu Bo ret = -EINVAL; 1905b9ca0664SLiu Bo goto out_drop_write; 1906b9ca0664SLiu Bo } 19070caa102dSLi Zefan 1908b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1909b9ca0664SLiu Bo ret = -EFAULT; 1910b9ca0664SLiu Bo goto out_drop_write; 1911b9ca0664SLiu Bo } 19120caa102dSLi Zefan 1913b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1914b9ca0664SLiu Bo ret = -EINVAL; 1915b9ca0664SLiu Bo goto out_drop_write; 1916b9ca0664SLiu Bo } 19170caa102dSLi Zefan 1918b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1919b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1920b9ca0664SLiu Bo goto out_drop_write; 1921b9ca0664SLiu Bo } 19220caa102dSLi Zefan 19230b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19240caa102dSLi Zefan 19250caa102dSLi Zefan /* nothing to do */ 19260caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1927b9ca0664SLiu Bo goto out_drop_sem; 19280caa102dSLi Zefan 19290caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19302c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19310caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19320caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19332c686537SDavid Sterba } else { 19342c686537SDavid Sterba /* 19352c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19362c686537SDavid Sterba * send 19372c686537SDavid Sterba */ 19382c686537SDavid Sterba spin_lock(&root->root_item_lock); 19392c686537SDavid Sterba if (root->send_in_progress == 0) { 19400caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19410caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19422c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19432c686537SDavid Sterba } else { 19442c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19450b246afaSJeff Mahoney btrfs_warn(fs_info, 19462c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19472c686537SDavid Sterba root->root_key.objectid); 19482c686537SDavid Sterba ret = -EPERM; 19492c686537SDavid Sterba goto out_drop_sem; 19502c686537SDavid Sterba } 19512c686537SDavid Sterba } 19520caa102dSLi Zefan 19530caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19540caa102dSLi Zefan if (IS_ERR(trans)) { 19550caa102dSLi Zefan ret = PTR_ERR(trans); 19560caa102dSLi Zefan goto out_reset; 19570caa102dSLi Zefan } 19580caa102dSLi Zefan 19590b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19600caa102dSLi Zefan &root->root_key, &root->root_item); 19619417ebc8SNikolay Borisov if (ret < 0) { 19629417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19639417ebc8SNikolay Borisov goto out_reset; 19649417ebc8SNikolay Borisov } 19650caa102dSLi Zefan 19669417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19679417ebc8SNikolay Borisov 19680caa102dSLi Zefan out_reset: 19690caa102dSLi Zefan if (ret) 19700caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1971b9ca0664SLiu Bo out_drop_sem: 19720b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1973b9ca0664SLiu Bo out_drop_write: 1974b9ca0664SLiu Bo mnt_drop_write_file(file); 1975b9ca0664SLiu Bo out: 19760caa102dSLi Zefan return ret; 19770caa102dSLi Zefan } 19780caa102dSLi Zefan 1979ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1980ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1981ac8e9819SChris Mason { 1982abc6e134SChris Mason struct btrfs_key test; 1983abc6e134SChris Mason int ret; 1984abc6e134SChris Mason 1985abc6e134SChris Mason test.objectid = sk->min_objectid; 1986abc6e134SChris Mason test.type = sk->min_type; 1987abc6e134SChris Mason test.offset = sk->min_offset; 1988abc6e134SChris Mason 1989abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1990abc6e134SChris Mason if (ret < 0) 1991ac8e9819SChris Mason return 0; 1992abc6e134SChris Mason 1993abc6e134SChris Mason test.objectid = sk->max_objectid; 1994abc6e134SChris Mason test.type = sk->max_type; 1995abc6e134SChris Mason test.offset = sk->max_offset; 1996abc6e134SChris Mason 1997abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1998abc6e134SChris Mason if (ret > 0) 1999ac8e9819SChris Mason return 0; 2000ac8e9819SChris Mason return 1; 2001ac8e9819SChris Mason } 2002ac8e9819SChris Mason 2003df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2004ac8e9819SChris Mason struct btrfs_key *key, 2005ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20069b6e817dSGerhard Heift size_t *buf_size, 2007ba346b35SGerhard Heift char __user *ubuf, 2008ac8e9819SChris Mason unsigned long *sk_offset, 2009ac8e9819SChris Mason int *num_found) 2010ac8e9819SChris Mason { 2011ac8e9819SChris Mason u64 found_transid; 2012ac8e9819SChris Mason struct extent_buffer *leaf; 2013ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2014dd81d459SNaohiro Aota struct btrfs_key test; 2015ac8e9819SChris Mason unsigned long item_off; 2016ac8e9819SChris Mason unsigned long item_len; 2017ac8e9819SChris Mason int nritems; 2018ac8e9819SChris Mason int i; 2019ac8e9819SChris Mason int slot; 2020ac8e9819SChris Mason int ret = 0; 2021ac8e9819SChris Mason 2022ac8e9819SChris Mason leaf = path->nodes[0]; 2023ac8e9819SChris Mason slot = path->slots[0]; 2024ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2025ac8e9819SChris Mason 2026ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2027ac8e9819SChris Mason i = nritems; 2028ac8e9819SChris Mason goto advance_key; 2029ac8e9819SChris Mason } 2030ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2031ac8e9819SChris Mason 2032ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2033ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2034ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2035ac8e9819SChris Mason 203603b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 203703b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 203803b71c6cSGabriel de Perthuis continue; 203903b71c6cSGabriel de Perthuis 20409b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20418f5f6178SGerhard Heift if (*num_found) { 2042ac8e9819SChris Mason ret = 1; 20438f5f6178SGerhard Heift goto out; 20448f5f6178SGerhard Heift } 20458f5f6178SGerhard Heift 20468f5f6178SGerhard Heift /* 20478f5f6178SGerhard Heift * return one empty item back for v1, which does not 20488f5f6178SGerhard Heift * handle -EOVERFLOW 20498f5f6178SGerhard Heift */ 20508f5f6178SGerhard Heift 20519b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2052ac8e9819SChris Mason item_len = 0; 20538f5f6178SGerhard Heift ret = -EOVERFLOW; 20548f5f6178SGerhard Heift } 2055ac8e9819SChris Mason 20569b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2057ac8e9819SChris Mason ret = 1; 205825c9bc2eSGerhard Heift goto out; 2059ac8e9819SChris Mason } 2060ac8e9819SChris Mason 2061ac8e9819SChris Mason sh.objectid = key->objectid; 2062ac8e9819SChris Mason sh.offset = key->offset; 2063ac8e9819SChris Mason sh.type = key->type; 2064ac8e9819SChris Mason sh.len = item_len; 2065ac8e9819SChris Mason sh.transid = found_transid; 2066ac8e9819SChris Mason 2067ac8e9819SChris Mason /* copy search result header */ 2068ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2069ba346b35SGerhard Heift ret = -EFAULT; 2070ba346b35SGerhard Heift goto out; 2071ba346b35SGerhard Heift } 2072ba346b35SGerhard Heift 2073ac8e9819SChris Mason *sk_offset += sizeof(sh); 2074ac8e9819SChris Mason 2075ac8e9819SChris Mason if (item_len) { 2076ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2077ac8e9819SChris Mason /* copy the item */ 2078ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2079ba346b35SGerhard Heift item_off, item_len)) { 2080ba346b35SGerhard Heift ret = -EFAULT; 2081ba346b35SGerhard Heift goto out; 2082ba346b35SGerhard Heift } 2083ba346b35SGerhard Heift 2084ac8e9819SChris Mason *sk_offset += item_len; 2085ac8e9819SChris Mason } 2086e2156867SHugo Mills (*num_found)++; 2087ac8e9819SChris Mason 20888f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20898f5f6178SGerhard Heift goto out; 20908f5f6178SGerhard Heift 209125c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 209225c9bc2eSGerhard Heift ret = 1; 209325c9bc2eSGerhard Heift goto out; 209425c9bc2eSGerhard Heift } 2095ac8e9819SChris Mason } 2096ac8e9819SChris Mason advance_key: 2097ac8e9819SChris Mason ret = 0; 2098dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2099dd81d459SNaohiro Aota test.type = sk->max_type; 2100dd81d459SNaohiro Aota test.offset = sk->max_offset; 2101dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2102dd81d459SNaohiro Aota ret = 1; 2103dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2104abc6e134SChris Mason key->offset++; 2105dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2106abc6e134SChris Mason key->offset = 0; 2107abc6e134SChris Mason key->type++; 2108dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2109abc6e134SChris Mason key->offset = 0; 2110abc6e134SChris Mason key->type = 0; 2111abc6e134SChris Mason key->objectid++; 2112abc6e134SChris Mason } else 2113abc6e134SChris Mason ret = 1; 211425c9bc2eSGerhard Heift out: 2115ba346b35SGerhard Heift /* 2116ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2117ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2118ba346b35SGerhard Heift * * all items were found 2119ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2120ba346b35SGerhard Heift * leaf 2121ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2122ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2123ba346b35SGerhard Heift */ 2124ac8e9819SChris Mason return ret; 2125ac8e9819SChris Mason } 2126ac8e9819SChris Mason 2127ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 212812544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21299b6e817dSGerhard Heift size_t *buf_size, 2130ba346b35SGerhard Heift char __user *ubuf) 2131ac8e9819SChris Mason { 21320b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2133ac8e9819SChris Mason struct btrfs_root *root; 2134ac8e9819SChris Mason struct btrfs_key key; 2135ac8e9819SChris Mason struct btrfs_path *path; 2136ac8e9819SChris Mason int ret; 2137ac8e9819SChris Mason int num_found = 0; 2138ac8e9819SChris Mason unsigned long sk_offset = 0; 2139ac8e9819SChris Mason 21409b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21419b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 214212544442SGerhard Heift return -EOVERFLOW; 21439b6e817dSGerhard Heift } 214412544442SGerhard Heift 2145ac8e9819SChris Mason path = btrfs_alloc_path(); 2146ac8e9819SChris Mason if (!path) 2147ac8e9819SChris Mason return -ENOMEM; 2148ac8e9819SChris Mason 2149ac8e9819SChris Mason if (sk->tree_id == 0) { 2150ac8e9819SChris Mason /* search the root of the inode that was passed */ 2151ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2152ac8e9819SChris Mason } else { 2153ac8e9819SChris Mason key.objectid = sk->tree_id; 2154ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2155ac8e9819SChris Mason key.offset = (u64)-1; 2156ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2157ac8e9819SChris Mason if (IS_ERR(root)) { 2158ac8e9819SChris Mason btrfs_free_path(path); 2159ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2160ac8e9819SChris Mason } 2161ac8e9819SChris Mason } 2162ac8e9819SChris Mason 2163ac8e9819SChris Mason key.objectid = sk->min_objectid; 2164ac8e9819SChris Mason key.type = sk->min_type; 2165ac8e9819SChris Mason key.offset = sk->min_offset; 2166ac8e9819SChris Mason 2167ac8e9819SChris Mason while (1) { 21686174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2169ac8e9819SChris Mason if (ret != 0) { 2170ac8e9819SChris Mason if (ret > 0) 2171ac8e9819SChris Mason ret = 0; 2172ac8e9819SChris Mason goto err; 2173ac8e9819SChris Mason } 2174df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2175ac8e9819SChris Mason &sk_offset, &num_found); 2176b3b4aa74SDavid Sterba btrfs_release_path(path); 217725c9bc2eSGerhard Heift if (ret) 2178ac8e9819SChris Mason break; 2179ac8e9819SChris Mason 2180ac8e9819SChris Mason } 21818f5f6178SGerhard Heift if (ret > 0) 2182ac8e9819SChris Mason ret = 0; 2183ac8e9819SChris Mason err: 2184ac8e9819SChris Mason sk->nr_items = num_found; 2185ac8e9819SChris Mason btrfs_free_path(path); 2186ac8e9819SChris Mason return ret; 2187ac8e9819SChris Mason } 2188ac8e9819SChris Mason 2189ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2190ac8e9819SChris Mason void __user *argp) 2191ac8e9819SChris Mason { 2192ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2193ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2194ac8e9819SChris Mason struct inode *inode; 2195ac8e9819SChris Mason int ret; 21969b6e817dSGerhard Heift size_t buf_size; 2197ac8e9819SChris Mason 2198ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2199ac8e9819SChris Mason return -EPERM; 2200ac8e9819SChris Mason 2201ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2202ac8e9819SChris Mason 2203ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2204ba346b35SGerhard Heift return -EFAULT; 2205ba346b35SGerhard Heift 2206ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2207ac8e9819SChris Mason 2208496ad9aaSAl Viro inode = file_inode(file); 2209ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22108f5f6178SGerhard Heift 22118f5f6178SGerhard Heift /* 22128f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22138f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22148f5f6178SGerhard Heift */ 22158f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22168f5f6178SGerhard Heift ret = 0; 22178f5f6178SGerhard Heift 2218ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2219ac8e9819SChris Mason ret = -EFAULT; 2220ac8e9819SChris Mason return ret; 2221ac8e9819SChris Mason } 2222ac8e9819SChris Mason 2223cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2224cc68a8a5SGerhard Heift void __user *argp) 2225cc68a8a5SGerhard Heift { 2226cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2227cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2228cc68a8a5SGerhard Heift struct inode *inode; 2229cc68a8a5SGerhard Heift int ret; 2230cc68a8a5SGerhard Heift size_t buf_size; 2231ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2232cc68a8a5SGerhard Heift 2233cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2234cc68a8a5SGerhard Heift return -EPERM; 2235cc68a8a5SGerhard Heift 2236cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2237cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2238cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2239cc68a8a5SGerhard Heift return -EFAULT; 2240cc68a8a5SGerhard Heift 2241cc68a8a5SGerhard Heift buf_size = args.buf_size; 2242cc68a8a5SGerhard Heift 2243cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2244cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2245cc68a8a5SGerhard Heift buf_size = buf_limit; 2246cc68a8a5SGerhard Heift 2247cc68a8a5SGerhard Heift inode = file_inode(file); 2248cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2249718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2250cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2251cc68a8a5SGerhard Heift ret = -EFAULT; 2252cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2253cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2254cc68a8a5SGerhard Heift ret = -EFAULT; 2255cc68a8a5SGerhard Heift 225676dda93cSYan, Zheng return ret; 225776dda93cSYan, Zheng } 225876dda93cSYan, Zheng 225998d377a0STARUISI Hiroaki /* 2260ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2261ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 226298d377a0STARUISI Hiroaki */ 226398d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 226498d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 226598d377a0STARUISI Hiroaki { 226698d377a0STARUISI Hiroaki struct btrfs_root *root; 226798d377a0STARUISI Hiroaki struct btrfs_key key; 2268ac8e9819SChris Mason char *ptr; 226998d377a0STARUISI Hiroaki int ret = -1; 227098d377a0STARUISI Hiroaki int slot; 227198d377a0STARUISI Hiroaki int len; 227298d377a0STARUISI Hiroaki int total_len = 0; 227398d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 227498d377a0STARUISI Hiroaki struct extent_buffer *l; 227598d377a0STARUISI Hiroaki struct btrfs_path *path; 227698d377a0STARUISI Hiroaki 227798d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 227898d377a0STARUISI Hiroaki name[0]='\0'; 227998d377a0STARUISI Hiroaki return 0; 228098d377a0STARUISI Hiroaki } 228198d377a0STARUISI Hiroaki 228298d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 228398d377a0STARUISI Hiroaki if (!path) 228498d377a0STARUISI Hiroaki return -ENOMEM; 228598d377a0STARUISI Hiroaki 2286c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 228798d377a0STARUISI Hiroaki 228898d377a0STARUISI Hiroaki key.objectid = tree_id; 228998d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 229098d377a0STARUISI Hiroaki key.offset = (u64)-1; 229198d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 229298d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2293ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 22948ad6fcabSChris Mason goto out; 229598d377a0STARUISI Hiroaki } 229698d377a0STARUISI Hiroaki 229798d377a0STARUISI Hiroaki key.objectid = dirid; 229898d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22998ad6fcabSChris Mason key.offset = (u64)-1; 230098d377a0STARUISI Hiroaki 230198d377a0STARUISI Hiroaki while (1) { 230298d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 230398d377a0STARUISI Hiroaki if (ret < 0) 230498d377a0STARUISI Hiroaki goto out; 230518674c6cSFilipe David Borba Manana else if (ret > 0) { 230618674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 230718674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 230818674c6cSFilipe David Borba Manana if (ret < 0) 230918674c6cSFilipe David Borba Manana goto out; 231018674c6cSFilipe David Borba Manana else if (ret > 0) { 2311ac8e9819SChris Mason ret = -ENOENT; 231298d377a0STARUISI Hiroaki goto out; 2313ac8e9819SChris Mason } 231418674c6cSFilipe David Borba Manana } 231518674c6cSFilipe David Borba Manana 231618674c6cSFilipe David Borba Manana l = path->nodes[0]; 231718674c6cSFilipe David Borba Manana slot = path->slots[0]; 231818674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 231998d377a0STARUISI Hiroaki 232098d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 232198d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 232298d377a0STARUISI Hiroaki ptr -= len + 1; 232398d377a0STARUISI Hiroaki total_len += len + 1; 2324a696cf35SFilipe David Borba Manana if (ptr < name) { 2325a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 232698d377a0STARUISI Hiroaki goto out; 2327a696cf35SFilipe David Borba Manana } 232898d377a0STARUISI Hiroaki 232998d377a0STARUISI Hiroaki *(ptr + len) = '/'; 233098d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 233198d377a0STARUISI Hiroaki 233298d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 233398d377a0STARUISI Hiroaki break; 233498d377a0STARUISI Hiroaki 2335b3b4aa74SDavid Sterba btrfs_release_path(path); 233698d377a0STARUISI Hiroaki key.objectid = key.offset; 23378ad6fcabSChris Mason key.offset = (u64)-1; 233898d377a0STARUISI Hiroaki dirid = key.objectid; 233998d377a0STARUISI Hiroaki } 234077906a50SLi Zefan memmove(name, ptr, total_len); 234198d377a0STARUISI Hiroaki name[total_len] = '\0'; 234298d377a0STARUISI Hiroaki ret = 0; 234398d377a0STARUISI Hiroaki out: 234498d377a0STARUISI Hiroaki btrfs_free_path(path); 2345ac8e9819SChris Mason return ret; 2346ac8e9819SChris Mason } 2347ac8e9819SChris Mason 234823d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 234923d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 235023d0b79dSTomohiro Misono { 235123d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 235223d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 235323d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 235423d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 235523d0b79dSTomohiro Misono u64 dirid = args->dirid; 235623d0b79dSTomohiro Misono unsigned long item_off; 235723d0b79dSTomohiro Misono unsigned long item_len; 235823d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 235923d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 236023d0b79dSTomohiro Misono struct btrfs_root *root; 236123d0b79dSTomohiro Misono struct btrfs_path *path; 236223d0b79dSTomohiro Misono struct btrfs_key key, key2; 236323d0b79dSTomohiro Misono struct extent_buffer *leaf; 236423d0b79dSTomohiro Misono struct inode *temp_inode; 236523d0b79dSTomohiro Misono char *ptr; 236623d0b79dSTomohiro Misono int slot; 236723d0b79dSTomohiro Misono int len; 236823d0b79dSTomohiro Misono int total_len = 0; 236923d0b79dSTomohiro Misono int ret; 237023d0b79dSTomohiro Misono 237123d0b79dSTomohiro Misono path = btrfs_alloc_path(); 237223d0b79dSTomohiro Misono if (!path) 237323d0b79dSTomohiro Misono return -ENOMEM; 237423d0b79dSTomohiro Misono 237523d0b79dSTomohiro Misono /* 237623d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 237723d0b79dSTomohiro Misono * construct the path in from the bottom up. 237823d0b79dSTomohiro Misono */ 237923d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 238023d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 238123d0b79dSTomohiro Misono 238223d0b79dSTomohiro Misono key.objectid = treeid; 238323d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 238423d0b79dSTomohiro Misono key.offset = (u64)-1; 238523d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 238623d0b79dSTomohiro Misono if (IS_ERR(root)) { 238723d0b79dSTomohiro Misono ret = PTR_ERR(root); 238823d0b79dSTomohiro Misono goto out; 238923d0b79dSTomohiro Misono } 239023d0b79dSTomohiro Misono 239123d0b79dSTomohiro Misono key.objectid = dirid; 239223d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 239323d0b79dSTomohiro Misono key.offset = (u64)-1; 239423d0b79dSTomohiro Misono while (1) { 239523d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 239623d0b79dSTomohiro Misono if (ret < 0) { 239723d0b79dSTomohiro Misono goto out; 239823d0b79dSTomohiro Misono } else if (ret > 0) { 239923d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 240023d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 240123d0b79dSTomohiro Misono if (ret < 0) { 240223d0b79dSTomohiro Misono goto out; 240323d0b79dSTomohiro Misono } else if (ret > 0) { 240423d0b79dSTomohiro Misono ret = -ENOENT; 240523d0b79dSTomohiro Misono goto out; 240623d0b79dSTomohiro Misono } 240723d0b79dSTomohiro Misono } 240823d0b79dSTomohiro Misono 240923d0b79dSTomohiro Misono leaf = path->nodes[0]; 241023d0b79dSTomohiro Misono slot = path->slots[0]; 241123d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 241223d0b79dSTomohiro Misono 241323d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 241423d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 241523d0b79dSTomohiro Misono ptr -= len + 1; 241623d0b79dSTomohiro Misono total_len += len + 1; 241723d0b79dSTomohiro Misono if (ptr < args->path) { 241823d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 241923d0b79dSTomohiro Misono goto out; 242023d0b79dSTomohiro Misono } 242123d0b79dSTomohiro Misono 242223d0b79dSTomohiro Misono *(ptr + len) = '/'; 242323d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 242423d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 242523d0b79dSTomohiro Misono 242623d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 242723d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 242823d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 242923d0b79dSTomohiro Misono if (ret < 0) { 243023d0b79dSTomohiro Misono goto out; 243123d0b79dSTomohiro Misono } else if (ret > 0) { 243223d0b79dSTomohiro Misono ret = -ENOENT; 243323d0b79dSTomohiro Misono goto out; 243423d0b79dSTomohiro Misono } 243523d0b79dSTomohiro Misono 243623d0b79dSTomohiro Misono leaf = path->nodes[0]; 243723d0b79dSTomohiro Misono slot = path->slots[0]; 243823d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 243923d0b79dSTomohiro Misono if (key2.objectid != dirid) { 244023d0b79dSTomohiro Misono ret = -ENOENT; 244123d0b79dSTomohiro Misono goto out; 244223d0b79dSTomohiro Misono } 244323d0b79dSTomohiro Misono 244423d0b79dSTomohiro Misono temp_inode = btrfs_iget(sb, &key2, root, NULL); 24453ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24463ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24473ca57bd6SMisono Tomohiro goto out; 24483ca57bd6SMisono Tomohiro } 244923d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 245023d0b79dSTomohiro Misono iput(temp_inode); 245123d0b79dSTomohiro Misono if (ret) { 245223d0b79dSTomohiro Misono ret = -EACCES; 245323d0b79dSTomohiro Misono goto out; 245423d0b79dSTomohiro Misono } 245523d0b79dSTomohiro Misono 245623d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 245723d0b79dSTomohiro Misono break; 245823d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 245923d0b79dSTomohiro Misono ret = -EACCES; 246023d0b79dSTomohiro Misono goto out; 246123d0b79dSTomohiro Misono } 246223d0b79dSTomohiro Misono 246323d0b79dSTomohiro Misono btrfs_release_path(path); 246423d0b79dSTomohiro Misono key.objectid = key.offset; 246523d0b79dSTomohiro Misono key.offset = (u64)-1; 246623d0b79dSTomohiro Misono dirid = key.objectid; 246723d0b79dSTomohiro Misono } 246823d0b79dSTomohiro Misono 246923d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 247023d0b79dSTomohiro Misono args->path[total_len] = '\0'; 247123d0b79dSTomohiro Misono btrfs_release_path(path); 247223d0b79dSTomohiro Misono } 247323d0b79dSTomohiro Misono 247423d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 247523d0b79dSTomohiro Misono root = fs_info->tree_root; 247623d0b79dSTomohiro Misono key.objectid = treeid; 247723d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 247823d0b79dSTomohiro Misono key.offset = args->treeid; 247923d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 248023d0b79dSTomohiro Misono if (ret < 0) { 248123d0b79dSTomohiro Misono goto out; 248223d0b79dSTomohiro Misono } else if (ret > 0) { 248323d0b79dSTomohiro Misono ret = -ENOENT; 248423d0b79dSTomohiro Misono goto out; 248523d0b79dSTomohiro Misono } 248623d0b79dSTomohiro Misono 248723d0b79dSTomohiro Misono leaf = path->nodes[0]; 248823d0b79dSTomohiro Misono slot = path->slots[0]; 248923d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 249023d0b79dSTomohiro Misono 249123d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 249223d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 249323d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 249423d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 249523d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 249623d0b79dSTomohiro Misono ret = -EINVAL; 249723d0b79dSTomohiro Misono goto out; 249823d0b79dSTomohiro Misono } 249923d0b79dSTomohiro Misono 250023d0b79dSTomohiro Misono /* Copy subvolume's name */ 250123d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 250223d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 250323d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 250423d0b79dSTomohiro Misono args->name[item_len] = 0; 250523d0b79dSTomohiro Misono 250623d0b79dSTomohiro Misono out: 250723d0b79dSTomohiro Misono btrfs_free_path(path); 250823d0b79dSTomohiro Misono return ret; 250923d0b79dSTomohiro Misono } 251023d0b79dSTomohiro Misono 2511ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2512ac8e9819SChris Mason void __user *argp) 2513ac8e9819SChris Mason { 2514ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2515ac8e9819SChris Mason struct inode *inode; 251601b810b8SDavid Sterba int ret = 0; 2517ac8e9819SChris Mason 25182354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25192354d08fSJulia Lawall if (IS_ERR(args)) 25202354d08fSJulia Lawall return PTR_ERR(args); 2521c2b96929SDan Carpenter 2522496ad9aaSAl Viro inode = file_inode(file); 2523ac8e9819SChris Mason 252401b810b8SDavid Sterba /* 252501b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 252601b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 252701b810b8SDavid Sterba */ 25281b53ac4dSChris Mason if (args->treeid == 0) 25291b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25301b53ac4dSChris Mason 253101b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 253201b810b8SDavid Sterba args->name[0] = 0; 253301b810b8SDavid Sterba goto out; 253401b810b8SDavid Sterba } 253501b810b8SDavid Sterba 253601b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 253701b810b8SDavid Sterba ret = -EPERM; 253801b810b8SDavid Sterba goto out; 253901b810b8SDavid Sterba } 254001b810b8SDavid Sterba 2541ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2542ac8e9819SChris Mason args->treeid, args->objectid, 2543ac8e9819SChris Mason args->name); 2544ac8e9819SChris Mason 254501b810b8SDavid Sterba out: 2546ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2547ac8e9819SChris Mason ret = -EFAULT; 2548ac8e9819SChris Mason 2549ac8e9819SChris Mason kfree(args); 255098d377a0STARUISI Hiroaki return ret; 255198d377a0STARUISI Hiroaki } 255298d377a0STARUISI Hiroaki 255323d0b79dSTomohiro Misono /* 255423d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 255523d0b79dSTomohiro Misono * 255623d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 255723d0b79dSTomohiro Misono * 255823d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 255923d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 256023d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 256123d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 256223d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 256323d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 256423d0b79dSTomohiro Misono */ 256523d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 256623d0b79dSTomohiro Misono { 256723d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 256823d0b79dSTomohiro Misono struct inode *inode; 256923d0b79dSTomohiro Misono int ret; 257023d0b79dSTomohiro Misono 257123d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 257223d0b79dSTomohiro Misono if (IS_ERR(args)) 257323d0b79dSTomohiro Misono return PTR_ERR(args); 257423d0b79dSTomohiro Misono 257523d0b79dSTomohiro Misono inode = file_inode(file); 257623d0b79dSTomohiro Misono 257723d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 257823d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 257923d0b79dSTomohiro Misono /* 258023d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 258123d0b79dSTomohiro Misono * called 258223d0b79dSTomohiro Misono */ 258323d0b79dSTomohiro Misono kfree(args); 258423d0b79dSTomohiro Misono return -EACCES; 258523d0b79dSTomohiro Misono } 258623d0b79dSTomohiro Misono 258723d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 258823d0b79dSTomohiro Misono 258923d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 259023d0b79dSTomohiro Misono ret = -EFAULT; 259123d0b79dSTomohiro Misono 259223d0b79dSTomohiro Misono kfree(args); 259323d0b79dSTomohiro Misono return ret; 259423d0b79dSTomohiro Misono } 259523d0b79dSTomohiro Misono 2596b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2597b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2598b64ec075STomohiro Misono { 2599b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2600b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2601b64ec075STomohiro Misono struct btrfs_root *root; 2602b64ec075STomohiro Misono struct btrfs_path *path; 2603b64ec075STomohiro Misono struct btrfs_key key; 2604b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2605b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2606b64ec075STomohiro Misono struct extent_buffer *leaf; 2607b64ec075STomohiro Misono unsigned long item_off; 2608b64ec075STomohiro Misono unsigned long item_len; 2609b64ec075STomohiro Misono struct inode *inode; 2610b64ec075STomohiro Misono int slot; 2611b64ec075STomohiro Misono int ret = 0; 2612b64ec075STomohiro Misono 2613b64ec075STomohiro Misono path = btrfs_alloc_path(); 2614b64ec075STomohiro Misono if (!path) 2615b64ec075STomohiro Misono return -ENOMEM; 2616b64ec075STomohiro Misono 2617b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2618b64ec075STomohiro Misono if (!subvol_info) { 2619b64ec075STomohiro Misono btrfs_free_path(path); 2620b64ec075STomohiro Misono return -ENOMEM; 2621b64ec075STomohiro Misono } 2622b64ec075STomohiro Misono 2623b64ec075STomohiro Misono inode = file_inode(file); 2624b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2625b64ec075STomohiro Misono 2626b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2627b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2628b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2629b64ec075STomohiro Misono key.offset = (u64)-1; 2630b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2631b64ec075STomohiro Misono if (IS_ERR(root)) { 2632b64ec075STomohiro Misono ret = PTR_ERR(root); 2633b64ec075STomohiro Misono goto out; 2634b64ec075STomohiro Misono } 2635b64ec075STomohiro Misono root_item = &root->root_item; 2636b64ec075STomohiro Misono 2637b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2638b64ec075STomohiro Misono 2639b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2640b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2641b64ec075STomohiro Misono 2642b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2643b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2644b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2645b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2646b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2647b64ec075STomohiro Misono 2648b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2649b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2650b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2651b64ec075STomohiro Misono 2652b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2653b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2654b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2655b64ec075STomohiro Misono 2656b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2657b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2658b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2659b64ec075STomohiro Misono 2660b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2661b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2662b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2663b64ec075STomohiro Misono 2664b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2665b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2666b64ec075STomohiro Misono root = fs_info->tree_root; 2667b64ec075STomohiro Misono 2668b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2669b64ec075STomohiro Misono key.offset = 0; 2670b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2671b64ec075STomohiro Misono if (ret < 0) { 2672b64ec075STomohiro Misono goto out; 2673b64ec075STomohiro Misono } else if (path->slots[0] >= 2674b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2675b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2676b64ec075STomohiro Misono if (ret < 0) { 2677b64ec075STomohiro Misono goto out; 2678b64ec075STomohiro Misono } else if (ret > 0) { 2679b64ec075STomohiro Misono ret = -EUCLEAN; 2680b64ec075STomohiro Misono goto out; 2681b64ec075STomohiro Misono } 2682b64ec075STomohiro Misono } 2683b64ec075STomohiro Misono 2684b64ec075STomohiro Misono leaf = path->nodes[0]; 2685b64ec075STomohiro Misono slot = path->slots[0]; 2686b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2687b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2688b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2689b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2690b64ec075STomohiro Misono 2691b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2692b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2693b64ec075STomohiro Misono 2694b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2695b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2696b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2697b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2698b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2699b64ec075STomohiro Misono item_off, item_len); 2700b64ec075STomohiro Misono } else { 2701b64ec075STomohiro Misono ret = -ENOENT; 2702b64ec075STomohiro Misono goto out; 2703b64ec075STomohiro Misono } 2704b64ec075STomohiro Misono } 2705b64ec075STomohiro Misono 2706b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2707b64ec075STomohiro Misono ret = -EFAULT; 2708b64ec075STomohiro Misono 2709b64ec075STomohiro Misono out: 2710b64ec075STomohiro Misono btrfs_free_path(path); 2711b64ec075STomohiro Misono kzfree(subvol_info); 2712b64ec075STomohiro Misono return ret; 2713b64ec075STomohiro Misono } 2714b64ec075STomohiro Misono 271542e4b520STomohiro Misono /* 271642e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 271742e4b520STomohiro Misono * except the subvolume name. 271842e4b520STomohiro Misono */ 271942e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 272042e4b520STomohiro Misono { 272142e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 272242e4b520STomohiro Misono struct btrfs_root_ref *rref; 272342e4b520STomohiro Misono struct btrfs_root *root; 272442e4b520STomohiro Misono struct btrfs_path *path; 272542e4b520STomohiro Misono struct btrfs_key key; 272642e4b520STomohiro Misono struct extent_buffer *leaf; 272742e4b520STomohiro Misono struct inode *inode; 272842e4b520STomohiro Misono u64 objectid; 272942e4b520STomohiro Misono int slot; 273042e4b520STomohiro Misono int ret; 273142e4b520STomohiro Misono u8 found; 273242e4b520STomohiro Misono 273342e4b520STomohiro Misono path = btrfs_alloc_path(); 273442e4b520STomohiro Misono if (!path) 273542e4b520STomohiro Misono return -ENOMEM; 273642e4b520STomohiro Misono 273742e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 273842e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 273942e4b520STomohiro Misono btrfs_free_path(path); 274042e4b520STomohiro Misono return PTR_ERR(rootrefs); 274142e4b520STomohiro Misono } 274242e4b520STomohiro Misono 274342e4b520STomohiro Misono inode = file_inode(file); 274442e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 274542e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 274642e4b520STomohiro Misono 274742e4b520STomohiro Misono key.objectid = objectid; 274842e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 274942e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 275042e4b520STomohiro Misono found = 0; 275142e4b520STomohiro Misono 275242e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 275342e4b520STomohiro Misono if (ret < 0) { 275442e4b520STomohiro Misono goto out; 275542e4b520STomohiro Misono } else if (path->slots[0] >= 275642e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 275742e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 275842e4b520STomohiro Misono if (ret < 0) { 275942e4b520STomohiro Misono goto out; 276042e4b520STomohiro Misono } else if (ret > 0) { 276142e4b520STomohiro Misono ret = -EUCLEAN; 276242e4b520STomohiro Misono goto out; 276342e4b520STomohiro Misono } 276442e4b520STomohiro Misono } 276542e4b520STomohiro Misono while (1) { 276642e4b520STomohiro Misono leaf = path->nodes[0]; 276742e4b520STomohiro Misono slot = path->slots[0]; 276842e4b520STomohiro Misono 276942e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 277042e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 277142e4b520STomohiro Misono ret = 0; 277242e4b520STomohiro Misono goto out; 277342e4b520STomohiro Misono } 277442e4b520STomohiro Misono 277542e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 277642e4b520STomohiro Misono ret = -EOVERFLOW; 277742e4b520STomohiro Misono goto out; 277842e4b520STomohiro Misono } 277942e4b520STomohiro Misono 278042e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 278142e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 278242e4b520STomohiro Misono rootrefs->rootref[found].dirid = 278342e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 278442e4b520STomohiro Misono found++; 278542e4b520STomohiro Misono 278642e4b520STomohiro Misono ret = btrfs_next_item(root, path); 278742e4b520STomohiro Misono if (ret < 0) { 278842e4b520STomohiro Misono goto out; 278942e4b520STomohiro Misono } else if (ret > 0) { 279042e4b520STomohiro Misono ret = -EUCLEAN; 279142e4b520STomohiro Misono goto out; 279242e4b520STomohiro Misono } 279342e4b520STomohiro Misono } 279442e4b520STomohiro Misono 279542e4b520STomohiro Misono out: 279642e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 279742e4b520STomohiro Misono rootrefs->num_items = found; 279842e4b520STomohiro Misono /* update min_treeid for next search */ 279942e4b520STomohiro Misono if (found) 280042e4b520STomohiro Misono rootrefs->min_treeid = 280142e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 280242e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 280342e4b520STomohiro Misono ret = -EFAULT; 280442e4b520STomohiro Misono } 280542e4b520STomohiro Misono 280642e4b520STomohiro Misono kfree(rootrefs); 280742e4b520STomohiro Misono btrfs_free_path(path); 280842e4b520STomohiro Misono 280942e4b520STomohiro Misono return ret; 281042e4b520STomohiro Misono } 281142e4b520STomohiro Misono 281276dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 281376dda93cSYan, Zheng void __user *arg) 281476dda93cSYan, Zheng { 281554563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28160b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 281776dda93cSYan, Zheng struct dentry *dentry; 28182b0143b5SDavid Howells struct inode *dir = d_inode(parent); 281976dda93cSYan, Zheng struct inode *inode; 282076dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 282176dda93cSYan, Zheng struct btrfs_root *dest = NULL; 282276dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 282376dda93cSYan, Zheng int namelen; 282476dda93cSYan, Zheng int err = 0; 282576dda93cSYan, Zheng 2826325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2827325c50e3SJeff Mahoney return -ENOTDIR; 2828325c50e3SJeff Mahoney 282976dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 283076dda93cSYan, Zheng if (IS_ERR(vol_args)) 283176dda93cSYan, Zheng return PTR_ERR(vol_args); 283276dda93cSYan, Zheng 283376dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 283476dda93cSYan, Zheng namelen = strlen(vol_args->name); 283576dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 283676dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 283776dda93cSYan, Zheng err = -EINVAL; 283876dda93cSYan, Zheng goto out; 283976dda93cSYan, Zheng } 284076dda93cSYan, Zheng 2841a561be71SAl Viro err = mnt_want_write_file(file); 284276dda93cSYan, Zheng if (err) 284376dda93cSYan, Zheng goto out; 284476dda93cSYan, Zheng 2845521e0546SDavid Sterba 284600235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 284700235411SAl Viro if (err == -EINTR) 284800235411SAl Viro goto out_drop_write; 284976dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 285076dda93cSYan, Zheng if (IS_ERR(dentry)) { 285176dda93cSYan, Zheng err = PTR_ERR(dentry); 285276dda93cSYan, Zheng goto out_unlock_dir; 285376dda93cSYan, Zheng } 285476dda93cSYan, Zheng 28552b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 285676dda93cSYan, Zheng err = -ENOENT; 285776dda93cSYan, Zheng goto out_dput; 285876dda93cSYan, Zheng } 285976dda93cSYan, Zheng 28602b0143b5SDavid Howells inode = d_inode(dentry); 28614260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28624260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28634260f7c7SSage Weil /* 28644260f7c7SSage Weil * Regular user. Only allow this with a special mount 28654260f7c7SSage Weil * option, when the user has write+exec access to the 28664260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28674260f7c7SSage Weil * allowed. 28684260f7c7SSage Weil * 28694260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28704260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28714260f7c7SSage Weil * otherwise be able to delete. 28724260f7c7SSage Weil * 28734260f7c7SSage Weil * Users who want to delete empty subvols should try 28744260f7c7SSage Weil * rmdir(2). 28754260f7c7SSage Weil */ 28764260f7c7SSage Weil err = -EPERM; 28770b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 28784260f7c7SSage Weil goto out_dput; 28794260f7c7SSage Weil 28804260f7c7SSage Weil /* 28814260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 28824260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 28834260f7c7SSage Weil * must be called on the dentry referencing the root 28844260f7c7SSage Weil * of the subvol, not a random directory contained 28854260f7c7SSage Weil * within it. 28864260f7c7SSage Weil */ 28874260f7c7SSage Weil err = -EINVAL; 28884260f7c7SSage Weil if (root == dest) 28894260f7c7SSage Weil goto out_dput; 28904260f7c7SSage Weil 28914260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 28924260f7c7SSage Weil if (err) 28934260f7c7SSage Weil goto out_dput; 28945c39da5bSMiao Xie } 28954260f7c7SSage Weil 28965c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 28974260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 28984260f7c7SSage Weil if (err) 28994260f7c7SSage Weil goto out_dput; 29004260f7c7SSage Weil 29014a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 290276dda93cSYan, Zheng err = -EINVAL; 290376dda93cSYan, Zheng goto out_dput; 290476dda93cSYan, Zheng } 290576dda93cSYan, Zheng 29065955102cSAl Viro inode_lock(inode); 2907f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 29085955102cSAl Viro inode_unlock(inode); 2909f60a2364SMisono Tomohiro if (!err) 291076dda93cSYan, Zheng d_delete(dentry); 2911fa6ac876SLiu Bo 291276dda93cSYan, Zheng out_dput: 291376dda93cSYan, Zheng dput(dentry); 291476dda93cSYan, Zheng out_unlock_dir: 29155955102cSAl Viro inode_unlock(dir); 291600235411SAl Viro out_drop_write: 29172a79f17eSAl Viro mnt_drop_write_file(file); 291876dda93cSYan, Zheng out: 291976dda93cSYan, Zheng kfree(vol_args); 292076dda93cSYan, Zheng return err; 292176dda93cSYan, Zheng } 292276dda93cSYan, Zheng 29231e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2924f46b5a66SChristoph Hellwig { 2925496ad9aaSAl Viro struct inode *inode = file_inode(file); 2926f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29271e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2928c146afadSYan Zheng int ret; 2929c146afadSYan Zheng 293025122d15SIlya Dryomov ret = mnt_want_write_file(file); 293125122d15SIlya Dryomov if (ret) 293225122d15SIlya Dryomov return ret; 2933b83cc969SLi Zefan 293425122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 293525122d15SIlya Dryomov ret = -EROFS; 293625122d15SIlya Dryomov goto out; 29375ac00addSStefan Behrens } 2938f46b5a66SChristoph Hellwig 2939f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2940f46b5a66SChristoph Hellwig case S_IFDIR: 2941e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2942e441d54dSChris Mason ret = -EPERM; 2943e441d54dSChris Mason goto out; 2944e441d54dSChris Mason } 2945de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2946f46b5a66SChristoph Hellwig break; 2947f46b5a66SChristoph Hellwig case S_IFREG: 2948616d374eSAdam Borowski /* 2949616d374eSAdam Borowski * Note that this does not check the file descriptor for write 2950616d374eSAdam Borowski * access. This prevents defragmenting executables that are 2951616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 2952616d374eSAdam Borowski */ 2953616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 2954616d374eSAdam Borowski inode_permission(inode, MAY_WRITE)) { 2955616d374eSAdam Borowski ret = -EPERM; 2956e441d54dSChris Mason goto out; 2957e441d54dSChris Mason } 29581e701a32SChris Mason 29591e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29601e701a32SChris Mason if (!range) { 29611e701a32SChris Mason ret = -ENOMEM; 29621e701a32SChris Mason goto out; 29631e701a32SChris Mason } 29641e701a32SChris Mason 29651e701a32SChris Mason if (argp) { 29661e701a32SChris Mason if (copy_from_user(range, argp, 29671e701a32SChris Mason sizeof(*range))) { 29681e701a32SChris Mason ret = -EFAULT; 29691e701a32SChris Mason kfree(range); 2970683be16eSDan Carpenter goto out; 29711e701a32SChris Mason } 29721e701a32SChris Mason /* compression requires us to start the IO */ 29731e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 29741e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 29751e701a32SChris Mason range->extent_thresh = (u32)-1; 29761e701a32SChris Mason } 29771e701a32SChris Mason } else { 29781e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 29791e701a32SChris Mason range->len = (u64)-1; 29801e701a32SChris Mason } 2981496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 29827c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 29834cb5300bSChris Mason if (ret > 0) 29844cb5300bSChris Mason ret = 0; 29851e701a32SChris Mason kfree(range); 2986f46b5a66SChristoph Hellwig break; 29878929ecfaSYan, Zheng default: 29888929ecfaSYan, Zheng ret = -EINVAL; 2989f46b5a66SChristoph Hellwig } 2990e441d54dSChris Mason out: 299125122d15SIlya Dryomov mnt_drop_write_file(file); 2992e441d54dSChris Mason return ret; 2993f46b5a66SChristoph Hellwig } 2994f46b5a66SChristoph Hellwig 29952ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2996f46b5a66SChristoph Hellwig { 2997f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2998f46b5a66SChristoph Hellwig int ret; 2999f46b5a66SChristoph Hellwig 3000e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3001e441d54dSChris Mason return -EPERM; 3002e441d54dSChris Mason 3003171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 3004e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3005c9e9f97bSIlya Dryomov 3006dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3007c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3008c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3009c9e9f97bSIlya Dryomov goto out; 3010c9e9f97bSIlya Dryomov } 3011f46b5a66SChristoph Hellwig 30125516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30132ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3014f46b5a66SChristoph Hellwig 301543d20761SAnand Jain if (!ret) 30160b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 301743d20761SAnand Jain 3018f46b5a66SChristoph Hellwig kfree(vol_args); 3019c9e9f97bSIlya Dryomov out: 3020171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3021f46b5a66SChristoph Hellwig return ret; 3022f46b5a66SChristoph Hellwig } 3023f46b5a66SChristoph Hellwig 30246b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3025f46b5a66SChristoph Hellwig { 30260b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30270b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30286b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3029f46b5a66SChristoph Hellwig int ret; 3030f46b5a66SChristoph Hellwig 3031e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3032e441d54dSChris Mason return -EPERM; 3033e441d54dSChris Mason 3034da24927bSMiao Xie ret = mnt_want_write_file(file); 3035da24927bSMiao Xie if (ret) 3036da24927bSMiao Xie return ret; 3037c146afadSYan Zheng 3038dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3039c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3040c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3041c47ca32dSDan Carpenter goto err_drop; 3042c9e9f97bSIlya Dryomov } 3043f46b5a66SChristoph Hellwig 30446b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3045fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3046fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3047fd4e994bSOmar Sandoval goto out; 3048fd4e994bSOmar Sandoval } 3049f46b5a66SChristoph Hellwig 3050171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3051183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3052183860f6SAnand Jain goto out; 3053183860f6SAnand Jain } 3054183860f6SAnand Jain 3055735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30562ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30576b526ed7SAnand Jain } else { 30586b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30592ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30606b526ed7SAnand Jain } 3061171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3062183860f6SAnand Jain 30636b526ed7SAnand Jain if (!ret) { 3064735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30650b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30666b526ed7SAnand Jain vol_args->devid); 30676b526ed7SAnand Jain else 30680b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30696b526ed7SAnand Jain vol_args->name); 30706b526ed7SAnand Jain } 3071183860f6SAnand Jain out: 3072183860f6SAnand Jain kfree(vol_args); 3073c47ca32dSDan Carpenter err_drop: 30744ac20c70SIlya Dryomov mnt_drop_write_file(file); 3075f46b5a66SChristoph Hellwig return ret; 3076f46b5a66SChristoph Hellwig } 3077f46b5a66SChristoph Hellwig 3078f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3079f46b5a66SChristoph Hellwig { 30800b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30810b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3082f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3083f46b5a66SChristoph Hellwig int ret; 3084f46b5a66SChristoph Hellwig 3085f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3086f46b5a66SChristoph Hellwig return -EPERM; 3087f46b5a66SChristoph Hellwig 3088f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3089f46b5a66SChristoph Hellwig if (ret) 3090f46b5a66SChristoph Hellwig return ret; 3091f46b5a66SChristoph Hellwig 3092171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3093f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 309458d7bbf8SDavid Sterba goto out_drop_write; 309558d7bbf8SDavid Sterba } 309658d7bbf8SDavid Sterba 309758d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 309858d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 309958d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3100f46b5a66SChristoph Hellwig goto out; 3101f46b5a66SChristoph Hellwig } 3102f46b5a66SChristoph Hellwig 310358d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31042ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3105f46b5a66SChristoph Hellwig 3106f46b5a66SChristoph Hellwig if (!ret) 31070b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3108f46b5a66SChristoph Hellwig kfree(vol_args); 310958d7bbf8SDavid Sterba out: 3110171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 311158d7bbf8SDavid Sterba out_drop_write: 3112f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 311358d7bbf8SDavid Sterba 3114f46b5a66SChristoph Hellwig return ret; 3115f46b5a66SChristoph Hellwig } 3116f46b5a66SChristoph Hellwig 31172ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 31182ff7e61eSJeff Mahoney void __user *arg) 3119475f6387SJan Schmidt { 3120027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3121475f6387SJan Schmidt struct btrfs_device *device; 31220b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3123027ed2f0SLi Zefan int ret = 0; 3124475f6387SJan Schmidt 3125027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3126027ed2f0SLi Zefan if (!fi_args) 3127027ed2f0SLi Zefan return -ENOMEM; 3128027ed2f0SLi Zefan 3129d03262c7SDavid Sterba rcu_read_lock(); 3130027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3131475f6387SJan Schmidt 3132d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3133027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3134027ed2f0SLi Zefan fi_args->max_id = device->devid; 3135475f6387SJan Schmidt } 3136d03262c7SDavid Sterba rcu_read_unlock(); 3137475f6387SJan Schmidt 3138d03262c7SDavid Sterba memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 3139bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3140bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3141bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 314280a773fbSDavid Sterba 3143027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3144027ed2f0SLi Zefan ret = -EFAULT; 3145475f6387SJan Schmidt 3146027ed2f0SLi Zefan kfree(fi_args); 3147027ed2f0SLi Zefan return ret; 3148475f6387SJan Schmidt } 3149475f6387SJan Schmidt 31502ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31512ff7e61eSJeff Mahoney void __user *arg) 3152475f6387SJan Schmidt { 3153475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3154475f6387SJan Schmidt struct btrfs_device *dev; 3155475f6387SJan Schmidt int ret = 0; 3156475f6387SJan Schmidt char *s_uuid = NULL; 3157475f6387SJan Schmidt 3158475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3159475f6387SJan Schmidt if (IS_ERR(di_args)) 3160475f6387SJan Schmidt return PTR_ERR(di_args); 3161475f6387SJan Schmidt 3162dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3163475f6387SJan Schmidt s_uuid = di_args->uuid; 3164475f6387SJan Schmidt 3165c5593ca3SDavid Sterba rcu_read_lock(); 31660b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 3167475f6387SJan Schmidt 3168475f6387SJan Schmidt if (!dev) { 3169475f6387SJan Schmidt ret = -ENODEV; 3170475f6387SJan Schmidt goto out; 3171475f6387SJan Schmidt } 3172475f6387SJan Schmidt 3173475f6387SJan Schmidt di_args->devid = dev->devid; 31747cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 31757cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3176475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3177a27202fbSJim Meyering if (dev->name) { 3178672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3179672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3180a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3181a27202fbSJim Meyering } else { 318299ba55adSStefan Behrens di_args->path[0] = '\0'; 3183a27202fbSJim Meyering } 3184475f6387SJan Schmidt 3185475f6387SJan Schmidt out: 3186c5593ca3SDavid Sterba rcu_read_unlock(); 3187475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3188475f6387SJan Schmidt ret = -EFAULT; 3189475f6387SJan Schmidt 3190475f6387SJan Schmidt kfree(di_args); 3191475f6387SJan Schmidt return ret; 3192475f6387SJan Schmidt } 3193475f6387SJan Schmidt 3194f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 3195416161dbSMark Fasheh { 3196416161dbSMark Fasheh struct page *page; 3197416161dbSMark Fasheh 3198416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 3199416161dbSMark Fasheh if (!page) 320031314002SFilipe Manana return ERR_PTR(-ENOMEM); 3201416161dbSMark Fasheh 3202416161dbSMark Fasheh if (!PageUptodate(page)) { 320331314002SFilipe Manana int ret; 320431314002SFilipe Manana 320531314002SFilipe Manana ret = btrfs_readpage(NULL, page); 320631314002SFilipe Manana if (ret) 320731314002SFilipe Manana return ERR_PTR(ret); 3208416161dbSMark Fasheh lock_page(page); 3209416161dbSMark Fasheh if (!PageUptodate(page)) { 3210416161dbSMark Fasheh unlock_page(page); 321109cbfeafSKirill A. Shutemov put_page(page); 321231314002SFilipe Manana return ERR_PTR(-EIO); 3213416161dbSMark Fasheh } 321431314002SFilipe Manana if (page->mapping != inode->i_mapping) { 3215416161dbSMark Fasheh unlock_page(page); 321609cbfeafSKirill A. Shutemov put_page(page); 321731314002SFilipe Manana return ERR_PTR(-EAGAIN); 3218416161dbSMark Fasheh } 3219416161dbSMark Fasheh } 3220416161dbSMark Fasheh 3221416161dbSMark Fasheh return page; 3222416161dbSMark Fasheh } 3223416161dbSMark Fasheh 3224f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 3225f4414602SMark Fasheh int num_pages, u64 off) 3226f4414602SMark Fasheh { 3227f4414602SMark Fasheh int i; 322809cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 3229f4414602SMark Fasheh 3230f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 323131314002SFilipe Manana again: 3232f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 323331314002SFilipe Manana if (IS_ERR(pages[i])) { 323431314002SFilipe Manana int err = PTR_ERR(pages[i]); 323531314002SFilipe Manana 323631314002SFilipe Manana if (err == -EAGAIN) 323731314002SFilipe Manana goto again; 323831314002SFilipe Manana pages[i] = NULL; 323931314002SFilipe Manana return err; 324031314002SFilipe Manana } 3241f4414602SMark Fasheh } 3242f4414602SMark Fasheh return 0; 3243f4414602SMark Fasheh } 3244f4414602SMark Fasheh 3245e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 3246e0bd70c6SFilipe Manana bool retry_range_locking) 324777fe20dcSMark Fasheh { 3248e0bd70c6SFilipe Manana /* 3249e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 3250e0bd70c6SFilipe Manana * another, and lock file content. 3251e0bd70c6SFilipe Manana * The locking order is: 3252e0bd70c6SFilipe Manana * 3253e0bd70c6SFilipe Manana * 1) pages 3254e0bd70c6SFilipe Manana * 2) range in the inode's io tree 3255e0bd70c6SFilipe Manana */ 325677fe20dcSMark Fasheh while (1) { 325777fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 325877fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 325977fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 326077fe20dcSMark Fasheh off + len - 1); 3261ff5df9b8SFilipe Manana if ((!ordered || 3262ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 3263ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 326477fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 3265ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 3266ff5df9b8SFilipe Manana if (ordered) 3267ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 326877fe20dcSMark Fasheh break; 3269ff5df9b8SFilipe Manana } 327077fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 327177fe20dcSMark Fasheh if (ordered) 327277fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 3273e0bd70c6SFilipe Manana if (!retry_range_locking) 3274e0bd70c6SFilipe Manana return -EAGAIN; 327577fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 327677fe20dcSMark Fasheh } 3277e0bd70c6SFilipe Manana return 0; 327877fe20dcSMark Fasheh } 327977fe20dcSMark Fasheh 3280f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 3281416161dbSMark Fasheh { 32825955102cSAl Viro inode_unlock(inode1); 32835955102cSAl Viro inode_unlock(inode2); 3284416161dbSMark Fasheh } 3285416161dbSMark Fasheh 3286f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 3287f4414602SMark Fasheh { 3288f4414602SMark Fasheh if (inode1 < inode2) 3289f4414602SMark Fasheh swap(inode1, inode2); 3290f4414602SMark Fasheh 32915955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 32925955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 3293f4414602SMark Fasheh } 3294f4414602SMark Fasheh 3295f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3296f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 3297f4414602SMark Fasheh { 3298f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3299f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3300f4414602SMark Fasheh } 3301f4414602SMark Fasheh 3302e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3303e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 3304e0bd70c6SFilipe Manana bool retry_range_locking) 3305416161dbSMark Fasheh { 3306e0bd70c6SFilipe Manana int ret; 3307e0bd70c6SFilipe Manana 3308416161dbSMark Fasheh if (inode1 < inode2) { 3309416161dbSMark Fasheh swap(inode1, inode2); 3310416161dbSMark Fasheh swap(loff1, loff2); 3311416161dbSMark Fasheh } 3312e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 3313e0bd70c6SFilipe Manana if (ret) 3314e0bd70c6SFilipe Manana return ret; 3315e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 3316e0bd70c6SFilipe Manana if (ret) 3317e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 3318e0bd70c6SFilipe Manana loff1 + len - 1); 3319e0bd70c6SFilipe Manana return ret; 3320416161dbSMark Fasheh } 3321f4414602SMark Fasheh 3322f4414602SMark Fasheh struct cmp_pages { 3323f4414602SMark Fasheh int num_pages; 3324f4414602SMark Fasheh struct page **src_pages; 3325f4414602SMark Fasheh struct page **dst_pages; 3326f4414602SMark Fasheh }; 3327f4414602SMark Fasheh 3328f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 3329f4414602SMark Fasheh { 3330f4414602SMark Fasheh int i; 3331f4414602SMark Fasheh struct page *pg; 3332f4414602SMark Fasheh 3333f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 3334f4414602SMark Fasheh pg = cmp->src_pages[i]; 3335e0bd70c6SFilipe Manana if (pg) { 3336e0bd70c6SFilipe Manana unlock_page(pg); 333709cbfeafSKirill A. Shutemov put_page(pg); 333897b19170SNaohiro Aota cmp->src_pages[i] = NULL; 3339e0bd70c6SFilipe Manana } 3340f4414602SMark Fasheh pg = cmp->dst_pages[i]; 3341e0bd70c6SFilipe Manana if (pg) { 3342e0bd70c6SFilipe Manana unlock_page(pg); 334309cbfeafSKirill A. Shutemov put_page(pg); 334497b19170SNaohiro Aota cmp->dst_pages[i] = NULL; 3345f4414602SMark Fasheh } 3346e0bd70c6SFilipe Manana } 3347f4414602SMark Fasheh } 3348f4414602SMark Fasheh 3349f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 3350f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 3351f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 3352f4414602SMark Fasheh { 3353f4414602SMark Fasheh int ret; 335409cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 3355f4414602SMark Fasheh 3356f4414602SMark Fasheh cmp->num_pages = num_pages; 3357f4414602SMark Fasheh 335867b07bd4STimofey Titovets ret = gather_extent_pages(src, cmp->src_pages, num_pages, loff); 3359f4414602SMark Fasheh if (ret) 3360f4414602SMark Fasheh goto out; 3361f4414602SMark Fasheh 336267b07bd4STimofey Titovets ret = gather_extent_pages(dst, cmp->dst_pages, num_pages, dst_loff); 3363f4414602SMark Fasheh 3364f4414602SMark Fasheh out: 3365f4414602SMark Fasheh if (ret) 3366f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 336778ad4ce0SNaohiro Aota return ret; 3368416161dbSMark Fasheh } 3369416161dbSMark Fasheh 33701a287cfeSDavid Sterba static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp) 3371416161dbSMark Fasheh { 3372416161dbSMark Fasheh int ret = 0; 3373f4414602SMark Fasheh int i; 3374416161dbSMark Fasheh struct page *src_page, *dst_page; 337509cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 3376416161dbSMark Fasheh void *addr, *dst_addr; 3377416161dbSMark Fasheh 3378f4414602SMark Fasheh i = 0; 3379416161dbSMark Fasheh while (len) { 338009cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 3381416161dbSMark Fasheh cmp_len = len; 3382416161dbSMark Fasheh 3383f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 3384f4414602SMark Fasheh 3385f4414602SMark Fasheh src_page = cmp->src_pages[i]; 3386f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 3387e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 3388e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 3389f4414602SMark Fasheh 3390416161dbSMark Fasheh addr = kmap_atomic(src_page); 3391416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 3392416161dbSMark Fasheh 3393416161dbSMark Fasheh flush_dcache_page(src_page); 3394416161dbSMark Fasheh flush_dcache_page(dst_page); 3395416161dbSMark Fasheh 3396416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 33972b3909f8SDarrick J. Wong ret = -EBADE; 3398416161dbSMark Fasheh 3399416161dbSMark Fasheh kunmap_atomic(addr); 3400416161dbSMark Fasheh kunmap_atomic(dst_addr); 3401416161dbSMark Fasheh 3402416161dbSMark Fasheh if (ret) 3403416161dbSMark Fasheh break; 3404416161dbSMark Fasheh 3405416161dbSMark Fasheh len -= cmp_len; 3406f4414602SMark Fasheh i++; 3407416161dbSMark Fasheh } 3408416161dbSMark Fasheh 3409416161dbSMark Fasheh return ret; 3410416161dbSMark Fasheh } 3411416161dbSMark Fasheh 3412e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3413e1d227a4SMark Fasheh u64 olen) 3414416161dbSMark Fasheh { 3415e1d227a4SMark Fasheh u64 len = *plen; 3416416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3417416161dbSMark Fasheh 3418e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 3419416161dbSMark Fasheh return -EINVAL; 3420e1d227a4SMark Fasheh 3421e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 3422e1d227a4SMark Fasheh if (off + len == inode->i_size) 3423e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 3424e1d227a4SMark Fasheh 3425416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 3426416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3427416161dbSMark Fasheh return -EINVAL; 3428416161dbSMark Fasheh 3429416161dbSMark Fasheh return 0; 3430416161dbSMark Fasheh } 3431416161dbSMark Fasheh 34323973909dSTimofey Titovets static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 olen, 343367b07bd4STimofey Titovets struct inode *dst, u64 dst_loff, 343467b07bd4STimofey Titovets struct cmp_pages *cmp) 3435416161dbSMark Fasheh { 3436416161dbSMark Fasheh int ret; 3437e1d227a4SMark Fasheh u64 len = olen; 3438fc4badd9SOmar Sandoval bool same_inode = (src == dst); 34390efa9f48SMark Fasheh u64 same_lock_start = 0; 34400efa9f48SMark Fasheh u64 same_lock_len = 0; 3441416161dbSMark Fasheh 34420efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 34430efa9f48SMark Fasheh if (ret) 34443973909dSTimofey Titovets return ret; 3445fc4badd9SOmar Sandoval 3446fc4badd9SOmar Sandoval ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3447f4dfe687SFilipe Manana if (ret) 34483973909dSTimofey Titovets return ret; 34490efa9f48SMark Fasheh 3450fc4badd9SOmar Sandoval if (same_inode) { 34510efa9f48SMark Fasheh /* 34520efa9f48SMark Fasheh * Single inode case wants the same checks, except we 34530efa9f48SMark Fasheh * don't want our length pushed out past i_size as 34540efa9f48SMark Fasheh * comparing that data range makes no sense. 34550efa9f48SMark Fasheh * 34560efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 34570efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 34580efa9f48SMark Fasheh * reject the request. 34590efa9f48SMark Fasheh * 34600efa9f48SMark Fasheh * This effectively means we require aligned extents 34610efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 34620efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 34630efa9f48SMark Fasheh * i_size. 34640efa9f48SMark Fasheh */ 34653973909dSTimofey Titovets if (len != olen) 34663973909dSTimofey Titovets return -EINVAL; 34670efa9f48SMark Fasheh 34680efa9f48SMark Fasheh /* Check for overlapping ranges */ 34693973909dSTimofey Titovets if (dst_loff + len > loff && dst_loff < loff + len) 34703973909dSTimofey Titovets return -EINVAL; 34710efa9f48SMark Fasheh 34720efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 34730efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 3474de02b9f6SFilipe Manana } else { 3475de02b9f6SFilipe Manana /* 3476de02b9f6SFilipe Manana * If the source and destination inodes are different, the 3477de02b9f6SFilipe Manana * source's range end offset matches the source's i_size, that 3478de02b9f6SFilipe Manana * i_size is not a multiple of the sector size, and the 3479de02b9f6SFilipe Manana * destination range does not go past the destination's i_size, 3480de02b9f6SFilipe Manana * we must round down the length to the nearest sector size 3481de02b9f6SFilipe Manana * multiple. If we don't do this adjustment we end replacing 3482de02b9f6SFilipe Manana * with zeroes the bytes in the range that starts at the 3483de02b9f6SFilipe Manana * deduplication range's end offset and ends at the next sector 3484de02b9f6SFilipe Manana * size multiple. 3485de02b9f6SFilipe Manana */ 3486de02b9f6SFilipe Manana if (loff + olen == i_size_read(src) && 3487de02b9f6SFilipe Manana dst_loff + len < i_size_read(dst)) { 3488de02b9f6SFilipe Manana const u64 sz = BTRFS_I(src)->root->fs_info->sectorsize; 3489de02b9f6SFilipe Manana 3490de02b9f6SFilipe Manana len = round_down(i_size_read(src), sz) - loff; 3491de02b9f6SFilipe Manana olen = len; 3492de02b9f6SFilipe Manana } 34930efa9f48SMark Fasheh } 3494416161dbSMark Fasheh 3495e0bd70c6SFilipe Manana again: 349667b07bd4STimofey Titovets ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, cmp); 3497f4414602SMark Fasheh if (ret) 34983973909dSTimofey Titovets return ret; 3499f4414602SMark Fasheh 35000efa9f48SMark Fasheh if (same_inode) 3501e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3502e0bd70c6SFilipe Manana false); 35030efa9f48SMark Fasheh else 3504e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3505e0bd70c6SFilipe Manana false); 3506e0bd70c6SFilipe Manana /* 3507e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3508e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3509e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3510e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3511e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3512e0bd70c6SFilipe Manana * range locking). 3513e0bd70c6SFilipe Manana */ 3514e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3515e0bd70c6SFilipe Manana /* 3516e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3517e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3518e0bd70c6SFilipe Manana */ 351967b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 3520e0bd70c6SFilipe Manana if (same_inode) { 3521e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3522e0bd70c6SFilipe Manana same_lock_len); 3523e0bd70c6SFilipe Manana } else { 3524e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3525e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3526e0bd70c6SFilipe Manana } 3527e0bd70c6SFilipe Manana goto again; 3528e0bd70c6SFilipe Manana } 3529e0bd70c6SFilipe Manana ASSERT(ret == 0); 3530e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3531e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 353267b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 3533e0bd70c6SFilipe Manana return ret; 3534e0bd70c6SFilipe Manana } 3535f4414602SMark Fasheh 3536207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 353767b07bd4STimofey Titovets ret = btrfs_cmp_data(olen, cmp); 3538416161dbSMark Fasheh if (ret == 0) 35391c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3540416161dbSMark Fasheh 35410efa9f48SMark Fasheh if (same_inode) 35420efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 35430efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 35440efa9f48SMark Fasheh else 3545f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3546f4414602SMark Fasheh 354767b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 35483973909dSTimofey Titovets 35493973909dSTimofey Titovets return ret; 35503973909dSTimofey Titovets } 35513973909dSTimofey Titovets 3552b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3553b6728768STimofey Titovets 35543973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 35553973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 35563973909dSTimofey Titovets { 35573973909dSTimofey Titovets int ret; 355867b07bd4STimofey Titovets struct cmp_pages cmp; 355967b07bd4STimofey Titovets int num_pages = PAGE_ALIGN(BTRFS_MAX_DEDUPE_LEN) >> PAGE_SHIFT; 35603973909dSTimofey Titovets bool same_inode = (src == dst); 3561b6728768STimofey Titovets u64 i, tail_len, chunk_count; 35623973909dSTimofey Titovets 35633973909dSTimofey Titovets if (olen == 0) 35643973909dSTimofey Titovets return 0; 35653973909dSTimofey Titovets 35663973909dSTimofey Titovets if (same_inode) 35673973909dSTimofey Titovets inode_lock(src); 35683973909dSTimofey Titovets else 35693973909dSTimofey Titovets btrfs_double_inode_lock(src, dst); 35703973909dSTimofey Titovets 35713973909dSTimofey Titovets /* don't make the dst file partly checksummed */ 35723973909dSTimofey Titovets if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 35733973909dSTimofey Titovets (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 35743973909dSTimofey Titovets ret = -EINVAL; 35753973909dSTimofey Titovets goto out_unlock; 35763973909dSTimofey Titovets } 35773973909dSTimofey Titovets 3578b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3579b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 358067b07bd4STimofey Titovets if (chunk_count == 0) 358167b07bd4STimofey Titovets num_pages = PAGE_ALIGN(tail_len) >> PAGE_SHIFT; 358267b07bd4STimofey Titovets 358367b07bd4STimofey Titovets /* 358467b07bd4STimofey Titovets * If deduping ranges in the same inode, locking rules make it 358567b07bd4STimofey Titovets * mandatory to always lock pages in ascending order to avoid deadlocks 358667b07bd4STimofey Titovets * with concurrent tasks (such as starting writeback/delalloc). 358767b07bd4STimofey Titovets */ 358867b07bd4STimofey Titovets if (same_inode && dst_loff < loff) 358967b07bd4STimofey Titovets swap(loff, dst_loff); 359067b07bd4STimofey Titovets 359167b07bd4STimofey Titovets /* 359267b07bd4STimofey Titovets * We must gather up all the pages before we initiate our extent 359367b07bd4STimofey Titovets * locking. We use an array for the page pointers. Size of the array is 359467b07bd4STimofey Titovets * bounded by len, which is in turn bounded by BTRFS_MAX_DEDUPE_LEN. 359567b07bd4STimofey Titovets */ 3596bf5091c8SDavid Sterba cmp.src_pages = kvmalloc_array(num_pages, sizeof(struct page *), 3597bf5091c8SDavid Sterba GFP_KERNEL | __GFP_ZERO); 3598bf5091c8SDavid Sterba cmp.dst_pages = kvmalloc_array(num_pages, sizeof(struct page *), 3599bf5091c8SDavid Sterba GFP_KERNEL | __GFP_ZERO); 360067b07bd4STimofey Titovets if (!cmp.src_pages || !cmp.dst_pages) { 3601bf5091c8SDavid Sterba ret = -ENOMEM; 3602bf5091c8SDavid Sterba goto out_free; 360367b07bd4STimofey Titovets } 3604b6728768STimofey Titovets 3605b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3606b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 360767b07bd4STimofey Titovets dst, dst_loff, &cmp); 3608b6728768STimofey Titovets if (ret) 360922883ddcSLu Fengqi goto out_free; 3610b6728768STimofey Titovets 3611b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3612b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3613b6728768STimofey Titovets } 3614b6728768STimofey Titovets 3615b6728768STimofey Titovets if (tail_len > 0) 361667b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 361767b07bd4STimofey Titovets dst_loff, &cmp); 36183973909dSTimofey Titovets 361922883ddcSLu Fengqi out_free: 362022883ddcSLu Fengqi kvfree(cmp.src_pages); 362122883ddcSLu Fengqi kvfree(cmp.dst_pages); 362222883ddcSLu Fengqi 3623416161dbSMark Fasheh out_unlock: 36240efa9f48SMark Fasheh if (same_inode) 36255955102cSAl Viro inode_unlock(src); 36260efa9f48SMark Fasheh else 3627f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3628416161dbSMark Fasheh 3629416161dbSMark Fasheh return ret; 3630416161dbSMark Fasheh } 3631416161dbSMark Fasheh 363287eb5eb2SMiklos Szeredi int btrfs_dedupe_file_range(struct file *src_file, loff_t src_loff, 363387eb5eb2SMiklos Szeredi struct file *dst_file, loff_t dst_loff, 363487eb5eb2SMiklos Szeredi u64 olen) 3635416161dbSMark Fasheh { 36362b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 36372b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3638416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 3639416161dbSMark Fasheh 364009cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3641416161dbSMark Fasheh /* 3642416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3643416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3644416161dbSMark Fasheh * this situation without an update. 3645416161dbSMark Fasheh */ 36462b3909f8SDarrick J. Wong return -EINVAL; 3647416161dbSMark Fasheh } 3648416161dbSMark Fasheh 364987eb5eb2SMiklos Szeredi return btrfs_extent_same(src, src_loff, olen, dst, dst_loff); 3650416161dbSMark Fasheh } 3651416161dbSMark Fasheh 3652f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3653f82a9901SFilipe Manana struct inode *inode, 3654f82a9901SFilipe Manana u64 endoff, 3655f82a9901SFilipe Manana const u64 destoff, 36561c919a5eSMark Fasheh const u64 olen, 36571c919a5eSMark Fasheh int no_time_update) 3658f82a9901SFilipe Manana { 3659f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3660f82a9901SFilipe Manana int ret; 3661f82a9901SFilipe Manana 3662f82a9901SFilipe Manana inode_inc_iversion(inode); 36631c919a5eSMark Fasheh if (!no_time_update) 3664c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3665f82a9901SFilipe Manana /* 3666f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3667f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3668f82a9901SFilipe Manana */ 3669f82a9901SFilipe Manana if (endoff > destoff + olen) 3670f82a9901SFilipe Manana endoff = destoff + olen; 3671f82a9901SFilipe Manana if (endoff > inode->i_size) 36726ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3673f82a9901SFilipe Manana 3674f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3675f82a9901SFilipe Manana if (ret) { 367666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 36773a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3678f82a9901SFilipe Manana goto out; 3679f82a9901SFilipe Manana } 36803a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3681f82a9901SFilipe Manana out: 3682f82a9901SFilipe Manana return ret; 3683f82a9901SFilipe Manana } 3684f82a9901SFilipe Manana 3685a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 36867ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 36877ffbb598SFilipe Manana const struct btrfs_path *path, 36887ffbb598SFilipe Manana const u64 hole_offset, 36897ffbb598SFilipe Manana const u64 hole_len) 36907ffbb598SFilipe Manana { 3691a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 36927ffbb598SFilipe Manana struct extent_map *em; 36937ffbb598SFilipe Manana int ret; 36947ffbb598SFilipe Manana 36957ffbb598SFilipe Manana em = alloc_extent_map(); 36967ffbb598SFilipe Manana if (!em) { 3697a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 36987ffbb598SFilipe Manana return; 36997ffbb598SFilipe Manana } 37007ffbb598SFilipe Manana 370114f59796SFilipe Manana if (path) { 370214f59796SFilipe Manana struct btrfs_file_extent_item *fi; 370314f59796SFilipe Manana 370414f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 370514f59796SFilipe Manana struct btrfs_file_extent_item); 3706a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 37077ffbb598SFilipe Manana em->generation = -1; 37087ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 37097ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 37107ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3711a2f392e4SNikolay Borisov &inode->runtime_flags); 37127ffbb598SFilipe Manana } else { 37137ffbb598SFilipe Manana em->start = hole_offset; 37147ffbb598SFilipe Manana em->len = hole_len; 37157ffbb598SFilipe Manana em->ram_bytes = em->len; 37167ffbb598SFilipe Manana em->orig_start = hole_offset; 37177ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 37187ffbb598SFilipe Manana em->block_len = 0; 37197ffbb598SFilipe Manana em->orig_block_len = 0; 37207ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 37217ffbb598SFilipe Manana em->generation = trans->transid; 37227ffbb598SFilipe Manana } 37237ffbb598SFilipe Manana 37247ffbb598SFilipe Manana while (1) { 37257ffbb598SFilipe Manana write_lock(&em_tree->lock); 37267ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 37277ffbb598SFilipe Manana write_unlock(&em_tree->lock); 37287ffbb598SFilipe Manana if (ret != -EEXIST) { 37297ffbb598SFilipe Manana free_extent_map(em); 37307ffbb598SFilipe Manana break; 37317ffbb598SFilipe Manana } 3732a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 37337ffbb598SFilipe Manana em->start + em->len - 1, 0); 37347ffbb598SFilipe Manana } 37357ffbb598SFilipe Manana 3736ee39b432SDavid Sterba if (ret) 3737a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 37387ffbb598SFilipe Manana } 37397ffbb598SFilipe Manana 37408039d87dSFilipe Manana /* 37418039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 37428039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 37438039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 37448039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 37458039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 37468039d87dSFilipe Manana * 37478039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 37488039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 37498039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 37508039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 37518039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 37528039d87dSFilipe Manana * i_mutex, which we are holding here). 37538039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 37548039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 37558039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 37568039d87dSFilipe Manana * 37578039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 37588039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 37598039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 37608039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 37618039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 37628039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 37638039d87dSFilipe Manana * the former. 37648039d87dSFilipe Manana */ 37654a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 37668039d87dSFilipe Manana struct btrfs_trans_handle *trans, 37678039d87dSFilipe Manana struct btrfs_path *path, 37688039d87dSFilipe Manana struct btrfs_key *new_key, 37698039d87dSFilipe Manana const u64 drop_start, 37708039d87dSFilipe Manana const u64 datal, 37718039d87dSFilipe Manana const u64 skip, 37728039d87dSFilipe Manana const u64 size, 37738039d87dSFilipe Manana char *inline_data) 37748039d87dSFilipe Manana { 37750b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 37768039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 37778039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 37780b246afaSJeff Mahoney fs_info->sectorsize); 37798039d87dSFilipe Manana int ret; 37808039d87dSFilipe Manana struct btrfs_key key; 37818039d87dSFilipe Manana 37828039d87dSFilipe Manana if (new_key->offset > 0) 37838039d87dSFilipe Manana return -EOPNOTSUPP; 37848039d87dSFilipe Manana 37854a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 37868039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 37878039d87dSFilipe Manana key.offset = 0; 37888039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 37898039d87dSFilipe Manana if (ret < 0) { 37908039d87dSFilipe Manana return ret; 37918039d87dSFilipe Manana } else if (ret > 0) { 37928039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 37938039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 37948039d87dSFilipe Manana if (ret < 0) 37958039d87dSFilipe Manana return ret; 37968039d87dSFilipe Manana else if (ret > 0) 37978039d87dSFilipe Manana goto copy_inline_extent; 37988039d87dSFilipe Manana } 37998039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 38004a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 38018039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 38028039d87dSFilipe Manana ASSERT(key.offset > 0); 38038039d87dSFilipe Manana return -EOPNOTSUPP; 38048039d87dSFilipe Manana } 38058039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 38068039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 38078039d87dSFilipe Manana u64 ext_len; 38088039d87dSFilipe Manana 38098039d87dSFilipe Manana /* 38108039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 38118039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 38128039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 38138039d87dSFilipe Manana */ 38148039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 38158039d87dSFilipe Manana struct btrfs_file_extent_item); 38168039d87dSFilipe Manana /* 38178039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 38188039d87dSFilipe Manana * following it. 38198039d87dSFilipe Manana */ 38208039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 38218039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 38228039d87dSFilipe Manana goto copy_inline_extent; 38238039d87dSFilipe Manana 38248039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 38258039d87dSFilipe Manana if (ext_len > aligned_end) 38268039d87dSFilipe Manana return -EOPNOTSUPP; 38278039d87dSFilipe Manana 38288039d87dSFilipe Manana ret = btrfs_next_item(root, path); 38298039d87dSFilipe Manana if (ret < 0) { 38308039d87dSFilipe Manana return ret; 38318039d87dSFilipe Manana } else if (ret == 0) { 38328039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 38338039d87dSFilipe Manana path->slots[0]); 38344a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 38358039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 38368039d87dSFilipe Manana return -EOPNOTSUPP; 38378039d87dSFilipe Manana } 38388039d87dSFilipe Manana } 38398039d87dSFilipe Manana 38408039d87dSFilipe Manana copy_inline_extent: 38418039d87dSFilipe Manana /* 38428039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 38438039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 38448039d87dSFilipe Manana */ 38458039d87dSFilipe Manana if (i_size_read(dst) > datal) { 38468039d87dSFilipe Manana /* 38478039d87dSFilipe Manana * If the destination inode has an inline extent... 38488039d87dSFilipe Manana * This would require copying the data from the source inline 38498039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 38508039d87dSFilipe Manana * But this is really complex, both extents can be compressed 38518039d87dSFilipe Manana * or just one of them, which would require decompressing and 38528039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 38538039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 38548039d87dSFilipe Manana * inline extent). 38558039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 38568039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 38578039d87dSFilipe Manana */ 38588039d87dSFilipe Manana return -EOPNOTSUPP; 38598039d87dSFilipe Manana } 38608039d87dSFilipe Manana 38618039d87dSFilipe Manana btrfs_release_path(path); 38628039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 38638039d87dSFilipe Manana if (ret) 38648039d87dSFilipe Manana return ret; 38658039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 38668039d87dSFilipe Manana if (ret) 38678039d87dSFilipe Manana return ret; 38688039d87dSFilipe Manana 38698039d87dSFilipe Manana if (skip) { 38708039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 38718039d87dSFilipe Manana 38728039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 38738039d87dSFilipe Manana } 38748039d87dSFilipe Manana 38758039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 38768039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 38778039d87dSFilipe Manana path->slots[0]), 38788039d87dSFilipe Manana size); 38798039d87dSFilipe Manana inode_add_bytes(dst, datal); 38808039d87dSFilipe Manana 38818039d87dSFilipe Manana return 0; 38828039d87dSFilipe Manana } 38838039d87dSFilipe Manana 388432b7c687SMark Fasheh /** 388532b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 388632b7c687SMark Fasheh * 388732b7c687SMark Fasheh * @src: Inode to clone from 388832b7c687SMark Fasheh * @inode: Inode to clone to 388932b7c687SMark Fasheh * @off: Offset within source to start clone from 389032b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 38911c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 389232b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 38931c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 389432b7c687SMark Fasheh */ 389532b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3896f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 38971c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3898f46b5a66SChristoph Hellwig { 38990b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3900f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 390132b7c687SMark Fasheh struct btrfs_path *path = NULL; 3902f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 390332b7c687SMark Fasheh struct btrfs_trans_handle *trans; 390432b7c687SMark Fasheh char *buf = NULL; 3905ae01a0abSYan Zheng struct btrfs_key key; 3906f46b5a66SChristoph Hellwig u32 nritems; 3907f46b5a66SChristoph Hellwig int slot; 3908ae01a0abSYan Zheng int ret; 3909f82a9901SFilipe Manana const u64 len = olen_aligned; 3910f82a9901SFilipe Manana u64 last_dest_end = destoff; 3911ae01a0abSYan Zheng 3912ae01a0abSYan Zheng ret = -ENOMEM; 3913752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3914ae01a0abSYan Zheng if (!buf) 391532b7c687SMark Fasheh return ret; 3916ae01a0abSYan Zheng 3917ae01a0abSYan Zheng path = btrfs_alloc_path(); 3918ae01a0abSYan Zheng if (!path) { 391915351955SDavid Sterba kvfree(buf); 392032b7c687SMark Fasheh return ret; 3921ae01a0abSYan Zheng } 392232b7c687SMark Fasheh 3923e4058b54SDavid Sterba path->reada = READA_FORWARD; 3924c5c9cd4dSSage Weil /* clone data */ 39254a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3926ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 39272c463823SFilipe Manana key.offset = off; 3928f46b5a66SChristoph Hellwig 3929f46b5a66SChristoph Hellwig while (1) { 3930de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3931df858e76SFilipe Manana 3932f46b5a66SChristoph Hellwig /* 3933f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3934f46b5a66SChristoph Hellwig * tree. 3935f46b5a66SChristoph Hellwig */ 3936e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3937362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3938362a20c5SDavid Sterba 0, 0); 3939f46b5a66SChristoph Hellwig if (ret < 0) 3940f46b5a66SChristoph Hellwig goto out; 39412c463823SFilipe Manana /* 39422c463823SFilipe Manana * First search, if no extent item that starts at offset off was 39432c463823SFilipe Manana * found but the previous item is an extent item, it's possible 39442c463823SFilipe Manana * it might overlap our target range, therefore process it. 39452c463823SFilipe Manana */ 39462c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 39472c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 39482c463823SFilipe Manana path->slots[0] - 1); 39492c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 39502c463823SFilipe Manana path->slots[0]--; 39512c463823SFilipe Manana } 3952f46b5a66SChristoph Hellwig 3953ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3954e4355f34SFilipe David Borba Manana process_slot: 3955ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3956362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3957f46b5a66SChristoph Hellwig if (ret < 0) 3958f46b5a66SChristoph Hellwig goto out; 3959f46b5a66SChristoph Hellwig if (ret > 0) 3960f46b5a66SChristoph Hellwig break; 3961ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3962f46b5a66SChristoph Hellwig } 3963f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3964f46b5a66SChristoph Hellwig slot = path->slots[0]; 3965f46b5a66SChristoph Hellwig 3966ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3967962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 39684a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3969f46b5a66SChristoph Hellwig break; 3970f46b5a66SChristoph Hellwig 3971962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3972c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3973c5c9cd4dSSage Weil int type; 397431840ae1SZheng Yan u32 size; 397531840ae1SZheng Yan struct btrfs_key new_key; 3976c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3977c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3978c5c9cd4dSSage Weil u8 comp; 3979f82a9901SFilipe Manana u64 drop_start; 398031840ae1SZheng Yan 3981c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3982c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3983c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3984c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3985c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3986c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3987d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3988d397712bSChris Mason extent); 3989d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3990d397712bSChris Mason extent); 3991c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3992d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3993d397712bSChris Mason extent); 3994c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3995c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3996c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3997c5c9cd4dSSage Weil extent); 3998c5c9cd4dSSage Weil } 399931840ae1SZheng Yan 40002c463823SFilipe Manana /* 40012c463823SFilipe Manana * The first search might have left us at an extent 40022c463823SFilipe Manana * item that ends before our target range's start, can 40032c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 40042c463823SFilipe Manana */ 40052c463823SFilipe Manana if (key.offset + datal <= off) { 4006e4355f34SFilipe David Borba Manana path->slots[0]++; 4007e4355f34SFilipe David Borba Manana goto process_slot; 40082c463823SFilipe Manana } else if (key.offset >= off + len) { 40092c463823SFilipe Manana break; 4010e4355f34SFilipe David Borba Manana } 4011df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 4012e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 4013e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 4014e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 4015e4355f34SFilipe David Borba Manana size); 4016e4355f34SFilipe David Borba Manana 4017e4355f34SFilipe David Borba Manana btrfs_release_path(path); 4018e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 4019c5c9cd4dSSage Weil 402031840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 40214a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 40224d728ec7SLi Zefan if (off <= key.offset) 4023c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 40244d728ec7SLi Zefan else 40254d728ec7SLi Zefan new_key.offset = destoff; 4026c5c9cd4dSSage Weil 4027b6f3409bSSage Weil /* 4028f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 4029f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 4030f82a9901SFilipe Manana * This hole is either in the middle of the cloning 4031f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 4032f82a9901SFilipe Manana * partially overlaps it). 4033f82a9901SFilipe Manana */ 4034f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 4035f82a9901SFilipe Manana drop_start = last_dest_end; 4036f82a9901SFilipe Manana else 4037f82a9901SFilipe Manana drop_start = new_key.offset; 4038f82a9901SFilipe Manana 4039f82a9901SFilipe Manana /* 4040b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 4041b6f3409bSSage Weil * 1 - add new extent 4042b6f3409bSSage Weil * 1 - inode update 4043b6f3409bSSage Weil */ 4044b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 4045a22285a6SYan, Zheng if (IS_ERR(trans)) { 4046a22285a6SYan, Zheng ret = PTR_ERR(trans); 4047a22285a6SYan, Zheng goto out; 4048a22285a6SYan, Zheng } 4049a22285a6SYan, Zheng 4050c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 4051c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 4052d72c0842SLi Zefan /* 4053d72c0842SLi Zefan * a | --- range to clone ---| b 4054d72c0842SLi Zefan * | ------------- extent ------------- | 4055d72c0842SLi Zefan */ 4056d72c0842SLi Zefan 405793915584SAntonio Ospite /* subtract range b */ 4058d72c0842SLi Zefan if (key.offset + datal > off + len) 4059d72c0842SLi Zefan datal = off + len - key.offset; 4060d72c0842SLi Zefan 406193915584SAntonio Ospite /* subtract range a */ 4062a22285a6SYan, Zheng if (off > key.offset) { 4063a22285a6SYan, Zheng datao += off - key.offset; 4064a22285a6SYan, Zheng datal -= off - key.offset; 4065a22285a6SYan, Zheng } 4066a22285a6SYan, Zheng 40675dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 4068f82a9901SFilipe Manana drop_start, 4069a22285a6SYan, Zheng new_key.offset + datal, 40702671485dSJosef Bacik 1); 407179787eaaSJeff Mahoney if (ret) { 40723f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 407300fdf13aSLiu Bo btrfs_abort_transaction(trans, 407466642832SJeff Mahoney ret); 40753a45bb20SJeff Mahoney btrfs_end_transaction(trans); 407679787eaaSJeff Mahoney goto out; 407779787eaaSJeff Mahoney } 4078a22285a6SYan, Zheng 407931840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 408031840ae1SZheng Yan &new_key, size); 408179787eaaSJeff Mahoney if (ret) { 408266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 40833a45bb20SJeff Mahoney btrfs_end_transaction(trans); 408479787eaaSJeff Mahoney goto out; 408579787eaaSJeff Mahoney } 408631840ae1SZheng Yan 408731840ae1SZheng Yan leaf = path->nodes[0]; 408831840ae1SZheng Yan slot = path->slots[0]; 408931840ae1SZheng Yan write_extent_buffer(leaf, buf, 409031840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 409131840ae1SZheng Yan size); 4092ae01a0abSYan Zheng 4093f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 4094f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 4095c5c9cd4dSSage Weil 4096c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 4097c5c9cd4dSSage Weil if (!disko) 4098c5c9cd4dSSage Weil datao = 0; 4099c5c9cd4dSSage Weil 4100c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 4101c5c9cd4dSSage Weil datao); 4102c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 4103c5c9cd4dSSage Weil datal); 4104fcebe456SJosef Bacik 4105c5c9cd4dSSage Weil if (disko) { 4106c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 41072ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 410884f7d8e6SJosef Bacik root, 41095d4f98a2SYan Zheng disko, diskl, 0, 4110f46b5a66SChristoph Hellwig root->root_key.objectid, 41114a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 4112b06c4bf5SFilipe Manana new_key.offset - datao); 411379787eaaSJeff Mahoney if (ret) { 411479787eaaSJeff Mahoney btrfs_abort_transaction(trans, 411579787eaaSJeff Mahoney ret); 41163a45bb20SJeff Mahoney btrfs_end_transaction(trans); 411779787eaaSJeff Mahoney goto out; 411879787eaaSJeff Mahoney 411979787eaaSJeff Mahoney } 4120f46b5a66SChristoph Hellwig } 4121c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 4122c5c9cd4dSSage Weil u64 skip = 0; 4123c5c9cd4dSSage Weil u64 trim = 0; 4124ed958762SFilipe Manana 4125c5c9cd4dSSage Weil if (off > key.offset) { 4126c5c9cd4dSSage Weil skip = off - key.offset; 4127c5c9cd4dSSage Weil new_key.offset += skip; 412831840ae1SZheng Yan } 4129d397712bSChris Mason 4130c5c9cd4dSSage Weil if (key.offset + datal > off + len) 4131c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 4132d397712bSChris Mason 4133c5c9cd4dSSage Weil if (comp && (skip || trim)) { 4134c5c9cd4dSSage Weil ret = -EINVAL; 41353a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4136c5c9cd4dSSage Weil goto out; 413731840ae1SZheng Yan } 4138c5c9cd4dSSage Weil size -= skip + trim; 4139c5c9cd4dSSage Weil datal -= skip + trim; 4140a22285a6SYan, Zheng 41414a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 41428039d87dSFilipe Manana trans, path, 41438039d87dSFilipe Manana &new_key, 4144f82a9901SFilipe Manana drop_start, 41458039d87dSFilipe Manana datal, 41468039d87dSFilipe Manana skip, size, buf); 414779787eaaSJeff Mahoney if (ret) { 41483f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 41493a29bc09SChris Mason btrfs_abort_transaction(trans, 415079787eaaSJeff Mahoney ret); 41513a45bb20SJeff Mahoney btrfs_end_transaction(trans); 415279787eaaSJeff Mahoney goto out; 415379787eaaSJeff Mahoney } 4154c5c9cd4dSSage Weil leaf = path->nodes[0]; 4155c5c9cd4dSSage Weil slot = path->slots[0]; 4156c5c9cd4dSSage Weil } 4157c5c9cd4dSSage Weil 41587ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 41597ffbb598SFilipe Manana if (drop_start < new_key.offset) 4160a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 416114f59796SFilipe Manana NULL, drop_start, 41627ffbb598SFilipe Manana new_key.offset - drop_start); 41637ffbb598SFilipe Manana 4164a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 4165a2f392e4SNikolay Borisov path, 0, 0); 41667ffbb598SFilipe Manana 4167c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 4168b3b4aa74SDavid Sterba btrfs_release_path(path); 4169c5c9cd4dSSage Weil 417062e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 41710b246afaSJeff Mahoney fs_info->sectorsize); 4172f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 4173f82a9901SFilipe Manana last_dest_end, 41741c919a5eSMark Fasheh destoff, olen, 41751c919a5eSMark Fasheh no_time_update); 4176f82a9901SFilipe Manana if (ret) 417779787eaaSJeff Mahoney goto out; 41782c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 41792c463823SFilipe Manana break; 4180a22285a6SYan, Zheng } 4181b3b4aa74SDavid Sterba btrfs_release_path(path); 4182df858e76SFilipe Manana key.offset = next_key_min_offset; 418369ae5e44SWang Xiaoguang 418469ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 418569ae5e44SWang Xiaoguang ret = -EINTR; 418669ae5e44SWang Xiaoguang goto out; 418769ae5e44SWang Xiaoguang } 4188ae01a0abSYan Zheng } 4189f46b5a66SChristoph Hellwig ret = 0; 419032b7c687SMark Fasheh 4191f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 4192f82a9901SFilipe Manana /* 4193f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 4194f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 4195f82a9901SFilipe Manana */ 4196b3b4aa74SDavid Sterba btrfs_release_path(path); 4197f82a9901SFilipe Manana 4198f82a9901SFilipe Manana /* 4199f82a9901SFilipe Manana * 1 - remove extent(s) 4200f82a9901SFilipe Manana * 1 - inode update 4201f82a9901SFilipe Manana */ 4202f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 4203f82a9901SFilipe Manana if (IS_ERR(trans)) { 4204f82a9901SFilipe Manana ret = PTR_ERR(trans); 4205f82a9901SFilipe Manana goto out; 4206f82a9901SFilipe Manana } 4207f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 4208f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 4209f82a9901SFilipe Manana if (ret) { 4210f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 421166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 42123a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4213f82a9901SFilipe Manana goto out; 4214f82a9901SFilipe Manana } 4215a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 4216a2f392e4SNikolay Borisov last_dest_end, 421714f59796SFilipe Manana destoff + len - last_dest_end); 4218f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 42191c919a5eSMark Fasheh destoff, olen, no_time_update); 4220f82a9901SFilipe Manana } 4221f82a9901SFilipe Manana 4222f46b5a66SChristoph Hellwig out: 422332b7c687SMark Fasheh btrfs_free_path(path); 422415351955SDavid Sterba kvfree(buf); 422532b7c687SMark Fasheh return ret; 422632b7c687SMark Fasheh } 422732b7c687SMark Fasheh 42283db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 422932b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 423032b7c687SMark Fasheh { 423154563d41SAl Viro struct inode *inode = file_inode(file); 42323db11b2eSZach Brown struct inode *src = file_inode(file_src); 42330b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 423432b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 423532b7c687SMark Fasheh int ret; 423632b7c687SMark Fasheh u64 len = olen; 42370b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 42383db11b2eSZach Brown int same_inode = src == inode; 423932b7c687SMark Fasheh 424032b7c687SMark Fasheh /* 424132b7c687SMark Fasheh * TODO: 424232b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 424332b7c687SMark Fasheh * decompress into destination's address_space (the file offset 424432b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 424532b7c687SMark Fasheh * otherwise reinsert) a subrange. 424600fdf13aSLiu Bo * 424700fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 424800fdf13aSLiu Bo * be either compressed or non-compressed. 424932b7c687SMark Fasheh */ 425032b7c687SMark Fasheh 425132b7c687SMark Fasheh if (btrfs_root_readonly(root)) 425232b7c687SMark Fasheh return -EROFS; 425332b7c687SMark Fasheh 42543db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 42553db11b2eSZach Brown src->i_sb != inode->i_sb) 42563db11b2eSZach Brown return -EXDEV; 425732b7c687SMark Fasheh 425832b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 42593db11b2eSZach Brown return -EISDIR; 426032b7c687SMark Fasheh 426132b7c687SMark Fasheh if (!same_inode) { 4262293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 426332b7c687SMark Fasheh } else { 42645955102cSAl Viro inode_lock(src); 426532b7c687SMark Fasheh } 426632b7c687SMark Fasheh 4267b5c40d59SOmar Sandoval /* don't make the dst file partly checksummed */ 4268b5c40d59SOmar Sandoval if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 4269b5c40d59SOmar Sandoval (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) { 4270b5c40d59SOmar Sandoval ret = -EINVAL; 4271b5c40d59SOmar Sandoval goto out_unlock; 4272b5c40d59SOmar Sandoval } 4273b5c40d59SOmar Sandoval 427432b7c687SMark Fasheh /* determine range to clone */ 427532b7c687SMark Fasheh ret = -EINVAL; 427632b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 427732b7c687SMark Fasheh goto out_unlock; 427832b7c687SMark Fasheh if (len == 0) 427932b7c687SMark Fasheh olen = len = src->i_size - off; 428032b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 428132b7c687SMark Fasheh if (off + len == src->i_size) 428232b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 428332b7c687SMark Fasheh 4284ccccf3d6SFilipe Manana if (len == 0) { 4285ccccf3d6SFilipe Manana ret = 0; 4286ccccf3d6SFilipe Manana goto out_unlock; 4287ccccf3d6SFilipe Manana } 4288ccccf3d6SFilipe Manana 428932b7c687SMark Fasheh /* verify the end result is block aligned */ 429032b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 429132b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 429232b7c687SMark Fasheh goto out_unlock; 429332b7c687SMark Fasheh 429432b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 429532b7c687SMark Fasheh if (same_inode) { 429632b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 429732b7c687SMark Fasheh goto out_unlock; 429832b7c687SMark Fasheh } 429932b7c687SMark Fasheh 430032b7c687SMark Fasheh if (destoff > inode->i_size) { 430132b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 430232b7c687SMark Fasheh if (ret) 430332b7c687SMark Fasheh goto out_unlock; 430432b7c687SMark Fasheh } 430532b7c687SMark Fasheh 4306c125b8bfSFilipe Manana /* 4307c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 4308c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 4309c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 4310c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 4311c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 4312c125b8bfSFilipe Manana */ 4313c125b8bfSFilipe Manana if (same_inode) { 4314c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 4315c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 431632b7c687SMark Fasheh 4317e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 4318c125b8bfSFilipe Manana } else { 4319e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 4320e0bd70c6SFilipe Manana true); 4321e0bd70c6SFilipe Manana } 4322e0bd70c6SFilipe Manana ASSERT(ret == 0); 4323e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 4324e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 4325e0bd70c6SFilipe Manana goto out_unlock; 4326c125b8bfSFilipe Manana } 432732b7c687SMark Fasheh 43281c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 432932b7c687SMark Fasheh 4330c125b8bfSFilipe Manana if (same_inode) { 4331c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 4332c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 4333c125b8bfSFilipe Manana 4334c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 4335c125b8bfSFilipe Manana } else { 4336293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 4337c125b8bfSFilipe Manana } 4338c125b8bfSFilipe Manana /* 4339c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 4340c125b8bfSFilipe Manana * data immediately and not the previous data. 4341c125b8bfSFilipe Manana */ 434265bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 434309cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 434409cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 4345f46b5a66SChristoph Hellwig out_unlock: 4346293a8489SMark Fasheh if (!same_inode) 4347293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 4348293a8489SMark Fasheh else 43495955102cSAl Viro inode_unlock(src); 43503db11b2eSZach Brown return ret; 43513db11b2eSZach Brown } 43523db11b2eSZach Brown 435304b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 435404b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 43553db11b2eSZach Brown { 435604b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 4357c5c9cd4dSSage Weil } 4358c5c9cd4dSSage Weil 43596ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 43606ef5ed0dSJosef Bacik { 4361496ad9aaSAl Viro struct inode *inode = file_inode(file); 43620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 43636ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 43646ef5ed0dSJosef Bacik struct btrfs_root *new_root; 43656ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 43666ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 43676ef5ed0dSJosef Bacik struct btrfs_path *path; 43686ef5ed0dSJosef Bacik struct btrfs_key location; 43696ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 43706ef5ed0dSJosef Bacik u64 objectid = 0; 43716ef5ed0dSJosef Bacik u64 dir_id; 43723c04ce01SMiao Xie int ret; 43736ef5ed0dSJosef Bacik 43746ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 43756ef5ed0dSJosef Bacik return -EPERM; 43766ef5ed0dSJosef Bacik 43773c04ce01SMiao Xie ret = mnt_want_write_file(file); 43783c04ce01SMiao Xie if (ret) 43793c04ce01SMiao Xie return ret; 43803c04ce01SMiao Xie 43813c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 43823c04ce01SMiao Xie ret = -EFAULT; 43833c04ce01SMiao Xie goto out; 43843c04ce01SMiao Xie } 43856ef5ed0dSJosef Bacik 43866ef5ed0dSJosef Bacik if (!objectid) 43871cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 43886ef5ed0dSJosef Bacik 43896ef5ed0dSJosef Bacik location.objectid = objectid; 43906ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 43916ef5ed0dSJosef Bacik location.offset = (u64)-1; 43926ef5ed0dSJosef Bacik 43930b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 43943c04ce01SMiao Xie if (IS_ERR(new_root)) { 43953c04ce01SMiao Xie ret = PTR_ERR(new_root); 43963c04ce01SMiao Xie goto out; 43973c04ce01SMiao Xie } 43984fd786e6SMisono Tomohiro if (!is_fstree(new_root->root_key.objectid)) { 43996d6d2829Ssatoru takeuchi ret = -ENOENT; 44006d6d2829Ssatoru takeuchi goto out; 44016d6d2829Ssatoru takeuchi } 44026ef5ed0dSJosef Bacik 44036ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 44043c04ce01SMiao Xie if (!path) { 44053c04ce01SMiao Xie ret = -ENOMEM; 44063c04ce01SMiao Xie goto out; 44073c04ce01SMiao Xie } 44086ef5ed0dSJosef Bacik path->leave_spinning = 1; 44096ef5ed0dSJosef Bacik 44106ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 441198d5dc13STsutomu Itoh if (IS_ERR(trans)) { 44126ef5ed0dSJosef Bacik btrfs_free_path(path); 44133c04ce01SMiao Xie ret = PTR_ERR(trans); 44143c04ce01SMiao Xie goto out; 44156ef5ed0dSJosef Bacik } 44166ef5ed0dSJosef Bacik 44170b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 44180b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 44196ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4420cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 44216ef5ed0dSJosef Bacik btrfs_free_path(path); 44223a45bb20SJeff Mahoney btrfs_end_transaction(trans); 44230b246afaSJeff Mahoney btrfs_err(fs_info, 44245d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 44253c04ce01SMiao Xie ret = -ENOENT; 44263c04ce01SMiao Xie goto out; 44276ef5ed0dSJosef Bacik } 44286ef5ed0dSJosef Bacik 44296ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 44306ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 44316ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 44326ef5ed0dSJosef Bacik btrfs_free_path(path); 44336ef5ed0dSJosef Bacik 44340b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 44353a45bb20SJeff Mahoney btrfs_end_transaction(trans); 44363c04ce01SMiao Xie out: 44373c04ce01SMiao Xie mnt_drop_write_file(file); 44383c04ce01SMiao Xie return ret; 44396ef5ed0dSJosef Bacik } 44406ef5ed0dSJosef Bacik 4441c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4442bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4443bf5fc093SJosef Bacik { 4444bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4445bf5fc093SJosef Bacik 4446bf5fc093SJosef Bacik space->total_bytes = 0; 4447bf5fc093SJosef Bacik space->used_bytes = 0; 4448bf5fc093SJosef Bacik space->flags = 0; 4449bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4450bf5fc093SJosef Bacik space->flags = block_group->flags; 4451bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4452bf5fc093SJosef Bacik space->used_bytes += 4453bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4454bf5fc093SJosef Bacik } 4455bf5fc093SJosef Bacik } 4456bf5fc093SJosef Bacik 44572ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 44582ff7e61eSJeff Mahoney void __user *arg) 44591406e432SJosef Bacik { 44601406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 44611406e432SJosef Bacik struct btrfs_ioctl_space_info space; 44621406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 44637fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 446413f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 44651406e432SJosef Bacik struct btrfs_space_info *info; 4466315d8e98SColin Ian King static const u64 types[] = { 4467315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4468bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4469bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4470315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4471315d8e98SColin Ian King }; 4472bf5fc093SJosef Bacik int num_types = 4; 44737fde62bfSChris Mason int alloc_size; 44741406e432SJosef Bacik int ret = 0; 447551788b1bSDan Rosenberg u64 slot_count = 0; 4476bf5fc093SJosef Bacik int i, c; 44771406e432SJosef Bacik 44781406e432SJosef Bacik if (copy_from_user(&space_args, 44791406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 44801406e432SJosef Bacik sizeof(space_args))) 44811406e432SJosef Bacik return -EFAULT; 44821406e432SJosef Bacik 4483bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4484bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4485bf5fc093SJosef Bacik 4486bf5fc093SJosef Bacik info = NULL; 44877fde62bfSChris Mason rcu_read_lock(); 44880b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4489bf5fc093SJosef Bacik list) { 4490bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4491bf5fc093SJosef Bacik info = tmp; 4492bf5fc093SJosef Bacik break; 4493bf5fc093SJosef Bacik } 4494bf5fc093SJosef Bacik } 44957fde62bfSChris Mason rcu_read_unlock(); 44961406e432SJosef Bacik 4497bf5fc093SJosef Bacik if (!info) 4498bf5fc093SJosef Bacik continue; 4499bf5fc093SJosef Bacik 4500bf5fc093SJosef Bacik down_read(&info->groups_sem); 4501bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4502bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4503bf5fc093SJosef Bacik slot_count++; 4504bf5fc093SJosef Bacik } 4505bf5fc093SJosef Bacik up_read(&info->groups_sem); 4506bf5fc093SJosef Bacik } 4507bf5fc093SJosef Bacik 450836523e95SDavid Sterba /* 450936523e95SDavid Sterba * Global block reserve, exported as a space_info 451036523e95SDavid Sterba */ 451136523e95SDavid Sterba slot_count++; 451236523e95SDavid Sterba 45137fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 45147fde62bfSChris Mason if (space_args.space_slots == 0) { 45157fde62bfSChris Mason space_args.total_spaces = slot_count; 45167fde62bfSChris Mason goto out; 45177fde62bfSChris Mason } 4518bf5fc093SJosef Bacik 451951788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4520bf5fc093SJosef Bacik 45217fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4522bf5fc093SJosef Bacik 45237fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 45247fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 45257fde62bfSChris Mason */ 452609cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 45277fde62bfSChris Mason return -ENOMEM; 45287fde62bfSChris Mason 45297fde62bfSChris Mason space_args.total_spaces = 0; 45308d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 45317fde62bfSChris Mason if (!dest) 45327fde62bfSChris Mason return -ENOMEM; 45337fde62bfSChris Mason dest_orig = dest; 45347fde62bfSChris Mason 45357fde62bfSChris Mason /* now we have a buffer to copy into */ 4536bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4537bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4538bf5fc093SJosef Bacik 453951788b1bSDan Rosenberg if (!slot_count) 454051788b1bSDan Rosenberg break; 454151788b1bSDan Rosenberg 4542bf5fc093SJosef Bacik info = NULL; 45431406e432SJosef Bacik rcu_read_lock(); 45440b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4545bf5fc093SJosef Bacik list) { 4546bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4547bf5fc093SJosef Bacik info = tmp; 45487fde62bfSChris Mason break; 4549bf5fc093SJosef Bacik } 4550bf5fc093SJosef Bacik } 4551bf5fc093SJosef Bacik rcu_read_unlock(); 45527fde62bfSChris Mason 4553bf5fc093SJosef Bacik if (!info) 4554bf5fc093SJosef Bacik continue; 4555bf5fc093SJosef Bacik down_read(&info->groups_sem); 4556bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4557bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4558c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4559c065f5b1SSu Yue &space); 45607fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 45611406e432SJosef Bacik dest++; 45621406e432SJosef Bacik space_args.total_spaces++; 456351788b1bSDan Rosenberg slot_count--; 45641406e432SJosef Bacik } 456551788b1bSDan Rosenberg if (!slot_count) 456651788b1bSDan Rosenberg break; 4567bf5fc093SJosef Bacik } 4568bf5fc093SJosef Bacik up_read(&info->groups_sem); 4569bf5fc093SJosef Bacik } 45701406e432SJosef Bacik 457136523e95SDavid Sterba /* 457236523e95SDavid Sterba * Add global block reserve 457336523e95SDavid Sterba */ 457436523e95SDavid Sterba if (slot_count) { 45750b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 457636523e95SDavid Sterba 457736523e95SDavid Sterba spin_lock(&block_rsv->lock); 457836523e95SDavid Sterba space.total_bytes = block_rsv->size; 457936523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 458036523e95SDavid Sterba spin_unlock(&block_rsv->lock); 458136523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 458236523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 458336523e95SDavid Sterba space_args.total_spaces++; 458436523e95SDavid Sterba } 458536523e95SDavid Sterba 45862eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 45877fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 45887fde62bfSChris Mason 45897fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 45907fde62bfSChris Mason ret = -EFAULT; 45917fde62bfSChris Mason 45927fde62bfSChris Mason kfree(dest_orig); 45937fde62bfSChris Mason out: 45947fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 45951406e432SJosef Bacik ret = -EFAULT; 45961406e432SJosef Bacik 45971406e432SJosef Bacik return ret; 45981406e432SJosef Bacik } 45991406e432SJosef Bacik 46009a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 46019a8c28beSMiao Xie void __user *argp) 460246204592SSage Weil { 460346204592SSage Weil struct btrfs_trans_handle *trans; 460446204592SSage Weil u64 transid; 4605db5b493aSTsutomu Itoh int ret; 460646204592SSage Weil 4607d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4608ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4609ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 461098d5dc13STsutomu Itoh return PTR_ERR(trans); 4611ff7c1d33SMiao Xie 4612ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4613ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4614ff7c1d33SMiao Xie goto out; 4615ff7c1d33SMiao Xie } 461646204592SSage Weil transid = trans->transid; 46173a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 46188b2b2d3cSTsutomu Itoh if (ret) { 46193a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4620db5b493aSTsutomu Itoh return ret; 46218b2b2d3cSTsutomu Itoh } 4622ff7c1d33SMiao Xie out: 462346204592SSage Weil if (argp) 462446204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 462546204592SSage Weil return -EFAULT; 462646204592SSage Weil return 0; 462746204592SSage Weil } 462846204592SSage Weil 46292ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 46309a8c28beSMiao Xie void __user *argp) 463146204592SSage Weil { 463246204592SSage Weil u64 transid; 463346204592SSage Weil 463446204592SSage Weil if (argp) { 463546204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 463646204592SSage Weil return -EFAULT; 463746204592SSage Weil } else { 463846204592SSage Weil transid = 0; /* current trans */ 463946204592SSage Weil } 46402ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 464146204592SSage Weil } 464246204592SSage Weil 4643b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4644475f6387SJan Schmidt { 46450b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4646475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4647b8e95489SMiao Xie int ret; 4648475f6387SJan Schmidt 4649475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4650475f6387SJan Schmidt return -EPERM; 4651475f6387SJan Schmidt 4652475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4653475f6387SJan Schmidt if (IS_ERR(sa)) 4654475f6387SJan Schmidt return PTR_ERR(sa); 4655475f6387SJan Schmidt 4656b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4657b8e95489SMiao Xie ret = mnt_want_write_file(file); 4658b8e95489SMiao Xie if (ret) 4659b8e95489SMiao Xie goto out; 4660b8e95489SMiao Xie } 4661b8e95489SMiao Xie 46620b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 466363a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 466463a212abSStefan Behrens 0); 4665475f6387SJan Schmidt 4666475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4667475f6387SJan Schmidt ret = -EFAULT; 4668475f6387SJan Schmidt 4669b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4670b8e95489SMiao Xie mnt_drop_write_file(file); 4671b8e95489SMiao Xie out: 4672475f6387SJan Schmidt kfree(sa); 4673475f6387SJan Schmidt return ret; 4674475f6387SJan Schmidt } 4675475f6387SJan Schmidt 46762ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4677475f6387SJan Schmidt { 4678475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4679475f6387SJan Schmidt return -EPERM; 4680475f6387SJan Schmidt 46812ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4682475f6387SJan Schmidt } 4683475f6387SJan Schmidt 46842ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4685475f6387SJan Schmidt void __user *arg) 4686475f6387SJan Schmidt { 4687475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4688475f6387SJan Schmidt int ret; 4689475f6387SJan Schmidt 4690475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4691475f6387SJan Schmidt return -EPERM; 4692475f6387SJan Schmidt 4693475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4694475f6387SJan Schmidt if (IS_ERR(sa)) 4695475f6387SJan Schmidt return PTR_ERR(sa); 4696475f6387SJan Schmidt 46972ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4698475f6387SJan Schmidt 4699475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4700475f6387SJan Schmidt ret = -EFAULT; 4701475f6387SJan Schmidt 4702475f6387SJan Schmidt kfree(sa); 4703475f6387SJan Schmidt return ret; 4704475f6387SJan Schmidt } 4705475f6387SJan Schmidt 47062ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4707b27f7c0cSDavid Sterba void __user *arg) 4708c11d2c23SStefan Behrens { 4709c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4710c11d2c23SStefan Behrens int ret; 4711c11d2c23SStefan Behrens 4712c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4713c11d2c23SStefan Behrens if (IS_ERR(sa)) 4714c11d2c23SStefan Behrens return PTR_ERR(sa); 4715c11d2c23SStefan Behrens 4716b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4717b27f7c0cSDavid Sterba kfree(sa); 4718b27f7c0cSDavid Sterba return -EPERM; 4719b27f7c0cSDavid Sterba } 4720b27f7c0cSDavid Sterba 47212ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4722c11d2c23SStefan Behrens 4723c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4724c11d2c23SStefan Behrens ret = -EFAULT; 4725c11d2c23SStefan Behrens 4726c11d2c23SStefan Behrens kfree(sa); 4727c11d2c23SStefan Behrens return ret; 4728c11d2c23SStefan Behrens } 4729c11d2c23SStefan Behrens 47302ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 47312ff7e61eSJeff Mahoney void __user *arg) 47323f6bcfbdSStefan Behrens { 47333f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 47343f6bcfbdSStefan Behrens int ret; 47353f6bcfbdSStefan Behrens 47363f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 47373f6bcfbdSStefan Behrens return -EPERM; 47383f6bcfbdSStefan Behrens 47393f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 47403f6bcfbdSStefan Behrens if (IS_ERR(p)) 47413f6bcfbdSStefan Behrens return PTR_ERR(p); 47423f6bcfbdSStefan Behrens 47433f6bcfbdSStefan Behrens switch (p->cmd) { 47443f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4745bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4746adfa97cbSIlya Dryomov ret = -EROFS; 4747adfa97cbSIlya Dryomov goto out; 4748adfa97cbSIlya Dryomov } 4749171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4750e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 47513f6bcfbdSStefan Behrens } else { 47522ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4753171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 47543f6bcfbdSStefan Behrens } 47553f6bcfbdSStefan Behrens break; 47563f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 47570b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 47583f6bcfbdSStefan Behrens ret = 0; 47593f6bcfbdSStefan Behrens break; 47603f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 476117d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 476297282031SAnand Jain ret = 0; 47633f6bcfbdSStefan Behrens break; 47643f6bcfbdSStefan Behrens default: 47653f6bcfbdSStefan Behrens ret = -EINVAL; 47663f6bcfbdSStefan Behrens break; 47673f6bcfbdSStefan Behrens } 47683f6bcfbdSStefan Behrens 47693f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 47703f6bcfbdSStefan Behrens ret = -EFAULT; 4771adfa97cbSIlya Dryomov out: 47723f6bcfbdSStefan Behrens kfree(p); 47733f6bcfbdSStefan Behrens return ret; 47743f6bcfbdSStefan Behrens } 47753f6bcfbdSStefan Behrens 4776d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4777d7728c96SJan Schmidt { 4778d7728c96SJan Schmidt int ret = 0; 4779d7728c96SJan Schmidt int i; 4780740c3d22SChris Mason u64 rel_ptr; 4781d7728c96SJan Schmidt int size; 4782806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4783d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4784d7728c96SJan Schmidt struct btrfs_path *path; 4785d7728c96SJan Schmidt 478682b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4787d7728c96SJan Schmidt return -EPERM; 4788d7728c96SJan Schmidt 4789d7728c96SJan Schmidt path = btrfs_alloc_path(); 4790d7728c96SJan Schmidt if (!path) { 4791d7728c96SJan Schmidt ret = -ENOMEM; 4792d7728c96SJan Schmidt goto out; 4793d7728c96SJan Schmidt } 4794d7728c96SJan Schmidt 4795d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4796d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4797d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4798d7728c96SJan Schmidt ipa = NULL; 4799d7728c96SJan Schmidt goto out; 4800d7728c96SJan Schmidt } 4801d7728c96SJan Schmidt 4802d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4803d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4804d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4805d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4806d7728c96SJan Schmidt ipath = NULL; 4807d7728c96SJan Schmidt goto out; 4808d7728c96SJan Schmidt } 4809d7728c96SJan Schmidt 4810d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4811d7728c96SJan Schmidt if (ret < 0) 4812d7728c96SJan Schmidt goto out; 4813d7728c96SJan Schmidt 4814d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4815745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4816745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4817740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4818d7728c96SJan Schmidt } 4819d7728c96SJan Schmidt 4820718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4821718dc5faSOmar Sandoval ipath->fspath, size); 4822d7728c96SJan Schmidt if (ret) { 4823d7728c96SJan Schmidt ret = -EFAULT; 4824d7728c96SJan Schmidt goto out; 4825d7728c96SJan Schmidt } 4826d7728c96SJan Schmidt 4827d7728c96SJan Schmidt out: 4828d7728c96SJan Schmidt btrfs_free_path(path); 4829d7728c96SJan Schmidt free_ipath(ipath); 4830d7728c96SJan Schmidt kfree(ipa); 4831d7728c96SJan Schmidt 4832d7728c96SJan Schmidt return ret; 4833d7728c96SJan Schmidt } 4834d7728c96SJan Schmidt 4835d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4836d7728c96SJan Schmidt { 4837d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4838d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4839d7728c96SJan Schmidt 4840d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4841d7728c96SJan Schmidt inodes->bytes_left -= c; 4842d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4843d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4844d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4845d7728c96SJan Schmidt inodes->elem_cnt += 3; 4846d7728c96SJan Schmidt } else { 4847d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4848d7728c96SJan Schmidt inodes->bytes_left = 0; 4849d7728c96SJan Schmidt inodes->elem_missed += 3; 4850d7728c96SJan Schmidt } 4851d7728c96SJan Schmidt 4852d7728c96SJan Schmidt return 0; 4853d7728c96SJan Schmidt } 4854d7728c96SJan Schmidt 48552ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4856d24a67b2SZygo Blaxell void __user *arg, int version) 4857d7728c96SJan Schmidt { 4858d7728c96SJan Schmidt int ret = 0; 4859d7728c96SJan Schmidt int size; 4860d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4861d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4862d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4863d24a67b2SZygo Blaxell bool ignore_offset; 4864d7728c96SJan Schmidt 4865d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4866d7728c96SJan Schmidt return -EPERM; 4867d7728c96SJan Schmidt 4868d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 48697b9ea627SShailendra Verma if (IS_ERR(loi)) 48707b9ea627SShailendra Verma return PTR_ERR(loi); 4871d7728c96SJan Schmidt 4872d24a67b2SZygo Blaxell if (version == 1) { 4873d24a67b2SZygo Blaxell ignore_offset = false; 4874b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4875d24a67b2SZygo Blaxell } else { 4876d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4877d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4878d24a67b2SZygo Blaxell ret = -EINVAL; 4879d24a67b2SZygo Blaxell goto out_loi; 4880d24a67b2SZygo Blaxell } 4881d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4882d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4883d24a67b2SZygo Blaxell ret = -EINVAL; 4884d24a67b2SZygo Blaxell goto out_loi; 4885d24a67b2SZygo Blaxell } 4886d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4887b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4888d24a67b2SZygo Blaxell } 4889d24a67b2SZygo Blaxell 4890d7728c96SJan Schmidt path = btrfs_alloc_path(); 4891d7728c96SJan Schmidt if (!path) { 4892d7728c96SJan Schmidt ret = -ENOMEM; 4893d7728c96SJan Schmidt goto out; 4894d7728c96SJan Schmidt } 4895d7728c96SJan Schmidt 4896d7728c96SJan Schmidt inodes = init_data_container(size); 4897d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4898d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4899d7728c96SJan Schmidt inodes = NULL; 4900d7728c96SJan Schmidt goto out; 4901d7728c96SJan Schmidt } 4902d7728c96SJan Schmidt 49032ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4904d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4905df031f07SLiu Bo if (ret == -EINVAL) 4906d7728c96SJan Schmidt ret = -ENOENT; 4907d7728c96SJan Schmidt if (ret < 0) 4908d7728c96SJan Schmidt goto out; 4909d7728c96SJan Schmidt 4910718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4911718dc5faSOmar Sandoval size); 4912d7728c96SJan Schmidt if (ret) 4913d7728c96SJan Schmidt ret = -EFAULT; 4914d7728c96SJan Schmidt 4915d7728c96SJan Schmidt out: 4916d7728c96SJan Schmidt btrfs_free_path(path); 4917f54de068SDavid Sterba kvfree(inodes); 4918d24a67b2SZygo Blaxell out_loi: 4919d7728c96SJan Schmidt kfree(loi); 4920d7728c96SJan Schmidt 4921d7728c96SJan Schmidt return ret; 4922d7728c96SJan Schmidt } 4923d7728c96SJan Schmidt 4924008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4925c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4926c9e9f97bSIlya Dryomov { 4927c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4928c9e9f97bSIlya Dryomov 4929c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4930c9e9f97bSIlya Dryomov 49313009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4932837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4933837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4934837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4935a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4936a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4937837d5b6eSIlya Dryomov 4938c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4939c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4940c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 494119a39dceSIlya Dryomov 494219a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 494319a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 494419a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4945c9e9f97bSIlya Dryomov } 4946c9e9f97bSIlya Dryomov 49479ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4948c9e9f97bSIlya Dryomov { 4949496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4950c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4951c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4952c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4953ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4954c9e9f97bSIlya Dryomov int ret; 4955c9e9f97bSIlya Dryomov 4956c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4957c9e9f97bSIlya Dryomov return -EPERM; 4958c9e9f97bSIlya Dryomov 4959e54bfa31SLiu Bo ret = mnt_want_write_file(file); 49609ba1f6e4SLiu Bo if (ret) 49619ba1f6e4SLiu Bo return ret; 49629ba1f6e4SLiu Bo 4963ed0fb78fSIlya Dryomov again: 4964171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4965c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4966ed0fb78fSIlya Dryomov need_unlock = true; 4967ed0fb78fSIlya Dryomov goto locked; 4968ed0fb78fSIlya Dryomov } 4969ed0fb78fSIlya Dryomov 4970ed0fb78fSIlya Dryomov /* 497101327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4972ed0fb78fSIlya Dryomov * (1) some other op is running 4973ed0fb78fSIlya Dryomov * (2) balance is running 4974ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4975ed0fb78fSIlya Dryomov */ 4976ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4977ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4978ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 49793009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4980ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4981dccdb07bSDavid Sterba /* 4982dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4983dccdb07bSDavid Sterba * we'll reexamine the status again. 4984dccdb07bSDavid Sterba */ 4985ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4986ed0fb78fSIlya Dryomov 4987ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 49883009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4989ed0fb78fSIlya Dryomov /* this is (3) */ 4990ed0fb78fSIlya Dryomov need_unlock = false; 4991ed0fb78fSIlya Dryomov goto locked; 4992ed0fb78fSIlya Dryomov } 4993ed0fb78fSIlya Dryomov 4994ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4995ed0fb78fSIlya Dryomov goto again; 4996ed0fb78fSIlya Dryomov } else { 4997ed0fb78fSIlya Dryomov /* this is (2) */ 4998ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4999ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 5000ed0fb78fSIlya Dryomov goto out; 5001ed0fb78fSIlya Dryomov } 5002ed0fb78fSIlya Dryomov } else { 5003ed0fb78fSIlya Dryomov /* this is (1) */ 5004ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 5005e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 5006ed0fb78fSIlya Dryomov goto out; 5007ed0fb78fSIlya Dryomov } 5008ed0fb78fSIlya Dryomov 5009ed0fb78fSIlya Dryomov locked: 5010171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 5011c9e9f97bSIlya Dryomov 5012c9e9f97bSIlya Dryomov if (arg) { 5013c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 5014c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 5015c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 5016ed0fb78fSIlya Dryomov goto out_unlock; 5017c9e9f97bSIlya Dryomov } 5018de322263SIlya Dryomov 5019de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 5020de322263SIlya Dryomov if (!fs_info->balance_ctl) { 5021de322263SIlya Dryomov ret = -ENOTCONN; 5022de322263SIlya Dryomov goto out_bargs; 5023de322263SIlya Dryomov } 5024de322263SIlya Dryomov 5025de322263SIlya Dryomov bctl = fs_info->balance_ctl; 5026de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 5027de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 5028de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 5029de322263SIlya Dryomov 5030de322263SIlya Dryomov goto do_balance; 5031de322263SIlya Dryomov } 5032c9e9f97bSIlya Dryomov } else { 5033c9e9f97bSIlya Dryomov bargs = NULL; 5034c9e9f97bSIlya Dryomov } 5035c9e9f97bSIlya Dryomov 5036ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 5037837d5b6eSIlya Dryomov ret = -EINPROGRESS; 5038837d5b6eSIlya Dryomov goto out_bargs; 5039837d5b6eSIlya Dryomov } 5040837d5b6eSIlya Dryomov 50418d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 5042c9e9f97bSIlya Dryomov if (!bctl) { 5043c9e9f97bSIlya Dryomov ret = -ENOMEM; 5044c9e9f97bSIlya Dryomov goto out_bargs; 5045c9e9f97bSIlya Dryomov } 5046c9e9f97bSIlya Dryomov 5047c9e9f97bSIlya Dryomov if (arg) { 5048c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 5049c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 5050c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 5051c9e9f97bSIlya Dryomov 5052c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 5053f43ffb60SIlya Dryomov } else { 5054f43ffb60SIlya Dryomov /* balance everything - no filters */ 5055f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 5056c9e9f97bSIlya Dryomov } 5057c9e9f97bSIlya Dryomov 50588eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 50598eb93459SDavid Sterba ret = -EINVAL; 50600f89abf5SChristian Engelmayer goto out_bctl; 50618eb93459SDavid Sterba } 50628eb93459SDavid Sterba 5063de322263SIlya Dryomov do_balance: 5064c9e9f97bSIlya Dryomov /* 5065149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 5066149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 5067149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 5068149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 5069c9e9f97bSIlya Dryomov */ 5070ed0fb78fSIlya Dryomov need_unlock = false; 5071ed0fb78fSIlya Dryomov 50726fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 50730f89abf5SChristian Engelmayer bctl = NULL; 5074ed0fb78fSIlya Dryomov 5075c9e9f97bSIlya Dryomov if (arg) { 5076c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 5077c9e9f97bSIlya Dryomov ret = -EFAULT; 5078c9e9f97bSIlya Dryomov } 5079c9e9f97bSIlya Dryomov 50800f89abf5SChristian Engelmayer out_bctl: 50810f89abf5SChristian Engelmayer kfree(bctl); 5082c9e9f97bSIlya Dryomov out_bargs: 5083c9e9f97bSIlya Dryomov kfree(bargs); 5084ed0fb78fSIlya Dryomov out_unlock: 5085c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 5086ed0fb78fSIlya Dryomov if (need_unlock) 5087171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 5088ed0fb78fSIlya Dryomov out: 5089e54bfa31SLiu Bo mnt_drop_write_file(file); 5090c9e9f97bSIlya Dryomov return ret; 5091c9e9f97bSIlya Dryomov } 5092c9e9f97bSIlya Dryomov 50932ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 5094837d5b6eSIlya Dryomov { 5095837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 5096837d5b6eSIlya Dryomov return -EPERM; 5097837d5b6eSIlya Dryomov 5098837d5b6eSIlya Dryomov switch (cmd) { 5099837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 51000b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 5101a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 51020b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 5103837d5b6eSIlya Dryomov } 5104837d5b6eSIlya Dryomov 5105837d5b6eSIlya Dryomov return -EINVAL; 5106837d5b6eSIlya Dryomov } 5107837d5b6eSIlya Dryomov 51082ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 510919a39dceSIlya Dryomov void __user *arg) 511019a39dceSIlya Dryomov { 511119a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 511219a39dceSIlya Dryomov int ret = 0; 511319a39dceSIlya Dryomov 511419a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 511519a39dceSIlya Dryomov return -EPERM; 511619a39dceSIlya Dryomov 511719a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 511819a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 511919a39dceSIlya Dryomov ret = -ENOTCONN; 512019a39dceSIlya Dryomov goto out; 512119a39dceSIlya Dryomov } 512219a39dceSIlya Dryomov 51238d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 512419a39dceSIlya Dryomov if (!bargs) { 512519a39dceSIlya Dryomov ret = -ENOMEM; 512619a39dceSIlya Dryomov goto out; 512719a39dceSIlya Dryomov } 512819a39dceSIlya Dryomov 5129008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 513019a39dceSIlya Dryomov 513119a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 513219a39dceSIlya Dryomov ret = -EFAULT; 513319a39dceSIlya Dryomov 513419a39dceSIlya Dryomov kfree(bargs); 513519a39dceSIlya Dryomov out: 513619a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 513719a39dceSIlya Dryomov return ret; 513819a39dceSIlya Dryomov } 513919a39dceSIlya Dryomov 5140905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 51415d13a37bSArne Jansen { 51420b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51430b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51445d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 51455d13a37bSArne Jansen int ret; 51465d13a37bSArne Jansen 51475d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 51485d13a37bSArne Jansen return -EPERM; 51495d13a37bSArne Jansen 5150905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5151905b0ddaSMiao Xie if (ret) 5152905b0ddaSMiao Xie return ret; 51535d13a37bSArne Jansen 51545d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5155905b0ddaSMiao Xie if (IS_ERR(sa)) { 5156905b0ddaSMiao Xie ret = PTR_ERR(sa); 5157905b0ddaSMiao Xie goto drop_write; 5158905b0ddaSMiao Xie } 51595d13a37bSArne Jansen 51600b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51615d13a37bSArne Jansen 51625d13a37bSArne Jansen switch (sa->cmd) { 51635d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 5164340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 51655d13a37bSArne Jansen break; 51665d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 5167340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 51685d13a37bSArne Jansen break; 51695d13a37bSArne Jansen default: 51705d13a37bSArne Jansen ret = -EINVAL; 51715d13a37bSArne Jansen break; 51725d13a37bSArne Jansen } 51735d13a37bSArne Jansen 51745d13a37bSArne Jansen kfree(sa); 51750b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5176905b0ddaSMiao Xie drop_write: 5177905b0ddaSMiao Xie mnt_drop_write_file(file); 51785d13a37bSArne Jansen return ret; 51795d13a37bSArne Jansen } 51805d13a37bSArne Jansen 5181905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 51825d13a37bSArne Jansen { 51830b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51840b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51850b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51865d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 51875d13a37bSArne Jansen struct btrfs_trans_handle *trans; 51885d13a37bSArne Jansen int ret; 51895d13a37bSArne Jansen int err; 51905d13a37bSArne Jansen 51915d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 51925d13a37bSArne Jansen return -EPERM; 51935d13a37bSArne Jansen 5194905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5195905b0ddaSMiao Xie if (ret) 5196905b0ddaSMiao Xie return ret; 51975d13a37bSArne Jansen 51985d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5199905b0ddaSMiao Xie if (IS_ERR(sa)) { 5200905b0ddaSMiao Xie ret = PTR_ERR(sa); 5201905b0ddaSMiao Xie goto drop_write; 5202905b0ddaSMiao Xie } 52035d13a37bSArne Jansen 52045d13a37bSArne Jansen trans = btrfs_join_transaction(root); 52055d13a37bSArne Jansen if (IS_ERR(trans)) { 52065d13a37bSArne Jansen ret = PTR_ERR(trans); 52075d13a37bSArne Jansen goto out; 52085d13a37bSArne Jansen } 52095d13a37bSArne Jansen 52105d13a37bSArne Jansen if (sa->assign) { 52119f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 52125d13a37bSArne Jansen } else { 521339616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 52145d13a37bSArne Jansen } 52155d13a37bSArne Jansen 5216e082f563SQu Wenruo /* update qgroup status and info */ 5217280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 5218e082f563SQu Wenruo if (err < 0) 52190b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 5220ad8403dfSAnand Jain "failed to update qgroup status and info"); 52213a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 52225d13a37bSArne Jansen if (err && !ret) 52235d13a37bSArne Jansen ret = err; 52245d13a37bSArne Jansen 52255d13a37bSArne Jansen out: 52265d13a37bSArne Jansen kfree(sa); 5227905b0ddaSMiao Xie drop_write: 5228905b0ddaSMiao Xie mnt_drop_write_file(file); 52295d13a37bSArne Jansen return ret; 52305d13a37bSArne Jansen } 52315d13a37bSArne Jansen 5232905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 52335d13a37bSArne Jansen { 52340b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52350b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52365d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 52375d13a37bSArne Jansen struct btrfs_trans_handle *trans; 52385d13a37bSArne Jansen int ret; 52395d13a37bSArne Jansen int err; 52405d13a37bSArne Jansen 52415d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 52425d13a37bSArne Jansen return -EPERM; 52435d13a37bSArne Jansen 5244905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5245905b0ddaSMiao Xie if (ret) 5246905b0ddaSMiao Xie return ret; 52475d13a37bSArne Jansen 52485d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5249905b0ddaSMiao Xie if (IS_ERR(sa)) { 5250905b0ddaSMiao Xie ret = PTR_ERR(sa); 5251905b0ddaSMiao Xie goto drop_write; 5252905b0ddaSMiao Xie } 52535d13a37bSArne Jansen 5254d86e56cfSMiao Xie if (!sa->qgroupid) { 5255d86e56cfSMiao Xie ret = -EINVAL; 5256d86e56cfSMiao Xie goto out; 5257d86e56cfSMiao Xie } 5258d86e56cfSMiao Xie 52595d13a37bSArne Jansen trans = btrfs_join_transaction(root); 52605d13a37bSArne Jansen if (IS_ERR(trans)) { 52615d13a37bSArne Jansen ret = PTR_ERR(trans); 52625d13a37bSArne Jansen goto out; 52635d13a37bSArne Jansen } 52645d13a37bSArne Jansen 52655d13a37bSArne Jansen if (sa->create) { 526649a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 52675d13a37bSArne Jansen } else { 52683efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 52695d13a37bSArne Jansen } 52705d13a37bSArne Jansen 52713a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 52725d13a37bSArne Jansen if (err && !ret) 52735d13a37bSArne Jansen ret = err; 52745d13a37bSArne Jansen 52755d13a37bSArne Jansen out: 52765d13a37bSArne Jansen kfree(sa); 5277905b0ddaSMiao Xie drop_write: 5278905b0ddaSMiao Xie mnt_drop_write_file(file); 52795d13a37bSArne Jansen return ret; 52805d13a37bSArne Jansen } 52815d13a37bSArne Jansen 5282905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 52835d13a37bSArne Jansen { 52840b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52850b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52865d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 52875d13a37bSArne Jansen struct btrfs_trans_handle *trans; 52885d13a37bSArne Jansen int ret; 52895d13a37bSArne Jansen int err; 52905d13a37bSArne Jansen u64 qgroupid; 52915d13a37bSArne Jansen 52925d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 52935d13a37bSArne Jansen return -EPERM; 52945d13a37bSArne Jansen 5295905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5296905b0ddaSMiao Xie if (ret) 5297905b0ddaSMiao Xie return ret; 52985d13a37bSArne Jansen 52995d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5300905b0ddaSMiao Xie if (IS_ERR(sa)) { 5301905b0ddaSMiao Xie ret = PTR_ERR(sa); 5302905b0ddaSMiao Xie goto drop_write; 5303905b0ddaSMiao Xie } 53045d13a37bSArne Jansen 53055d13a37bSArne Jansen trans = btrfs_join_transaction(root); 53065d13a37bSArne Jansen if (IS_ERR(trans)) { 53075d13a37bSArne Jansen ret = PTR_ERR(trans); 53085d13a37bSArne Jansen goto out; 53095d13a37bSArne Jansen } 53105d13a37bSArne Jansen 53115d13a37bSArne Jansen qgroupid = sa->qgroupid; 53125d13a37bSArne Jansen if (!qgroupid) { 53135d13a37bSArne Jansen /* take the current subvol as qgroup */ 53145d13a37bSArne Jansen qgroupid = root->root_key.objectid; 53155d13a37bSArne Jansen } 53165d13a37bSArne Jansen 5317f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 53185d13a37bSArne Jansen 53193a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 53205d13a37bSArne Jansen if (err && !ret) 53215d13a37bSArne Jansen ret = err; 53225d13a37bSArne Jansen 53235d13a37bSArne Jansen out: 53245d13a37bSArne Jansen kfree(sa); 5325905b0ddaSMiao Xie drop_write: 5326905b0ddaSMiao Xie mnt_drop_write_file(file); 53275d13a37bSArne Jansen return ret; 53285d13a37bSArne Jansen } 53295d13a37bSArne Jansen 53302f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 53312f232036SJan Schmidt { 53320b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53330b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53342f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 53352f232036SJan Schmidt int ret; 53362f232036SJan Schmidt 53372f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 53382f232036SJan Schmidt return -EPERM; 53392f232036SJan Schmidt 53402f232036SJan Schmidt ret = mnt_want_write_file(file); 53412f232036SJan Schmidt if (ret) 53422f232036SJan Schmidt return ret; 53432f232036SJan Schmidt 53442f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 53452f232036SJan Schmidt if (IS_ERR(qsa)) { 53462f232036SJan Schmidt ret = PTR_ERR(qsa); 53472f232036SJan Schmidt goto drop_write; 53482f232036SJan Schmidt } 53492f232036SJan Schmidt 53502f232036SJan Schmidt if (qsa->flags) { 53512f232036SJan Schmidt ret = -EINVAL; 53522f232036SJan Schmidt goto out; 53532f232036SJan Schmidt } 53542f232036SJan Schmidt 53550b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 53562f232036SJan Schmidt 53572f232036SJan Schmidt out: 53582f232036SJan Schmidt kfree(qsa); 53592f232036SJan Schmidt drop_write: 53602f232036SJan Schmidt mnt_drop_write_file(file); 53612f232036SJan Schmidt return ret; 53622f232036SJan Schmidt } 53632f232036SJan Schmidt 53642f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 53652f232036SJan Schmidt { 53660b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53670b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53682f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 53692f232036SJan Schmidt int ret = 0; 53702f232036SJan Schmidt 53712f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 53722f232036SJan Schmidt return -EPERM; 53732f232036SJan Schmidt 53748d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 53752f232036SJan Schmidt if (!qsa) 53762f232036SJan Schmidt return -ENOMEM; 53772f232036SJan Schmidt 53780b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 53792f232036SJan Schmidt qsa->flags = 1; 53800b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 53812f232036SJan Schmidt } 53822f232036SJan Schmidt 53832f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 53842f232036SJan Schmidt ret = -EFAULT; 53852f232036SJan Schmidt 53862f232036SJan Schmidt kfree(qsa); 53872f232036SJan Schmidt return ret; 53882f232036SJan Schmidt } 53892f232036SJan Schmidt 539057254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 539157254b6eSJan Schmidt { 53920b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 539457254b6eSJan Schmidt 539557254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 539657254b6eSJan Schmidt return -EPERM; 539757254b6eSJan Schmidt 53980b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 539957254b6eSJan Schmidt } 540057254b6eSJan Schmidt 5401abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5402abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 54038ea05e3aSAlexander Block { 5404496ad9aaSAl Viro struct inode *inode = file_inode(file); 54050b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54068ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 54078ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 54088ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 540995582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 54108ea05e3aSAlexander Block int ret = 0; 5411dd5f9615SStefan Behrens int received_uuid_changed; 54128ea05e3aSAlexander Block 5413bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5414bd60ea0fSDavid Sterba return -EPERM; 5415bd60ea0fSDavid Sterba 54168ea05e3aSAlexander Block ret = mnt_want_write_file(file); 54178ea05e3aSAlexander Block if (ret < 0) 54188ea05e3aSAlexander Block return ret; 54198ea05e3aSAlexander Block 54200b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 54218ea05e3aSAlexander Block 54224a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 54238ea05e3aSAlexander Block ret = -EINVAL; 54248ea05e3aSAlexander Block goto out; 54258ea05e3aSAlexander Block } 54268ea05e3aSAlexander Block 54278ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 54288ea05e3aSAlexander Block ret = -EROFS; 54298ea05e3aSAlexander Block goto out; 54308ea05e3aSAlexander Block } 54318ea05e3aSAlexander Block 5432dd5f9615SStefan Behrens /* 5433dd5f9615SStefan Behrens * 1 - root item 5434dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5435dd5f9615SStefan Behrens */ 5436dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 54378ea05e3aSAlexander Block if (IS_ERR(trans)) { 54388ea05e3aSAlexander Block ret = PTR_ERR(trans); 54398ea05e3aSAlexander Block trans = NULL; 54408ea05e3aSAlexander Block goto out; 54418ea05e3aSAlexander Block } 54428ea05e3aSAlexander Block 54438ea05e3aSAlexander Block sa->rtransid = trans->transid; 54448ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 54458ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 54468ea05e3aSAlexander Block 5447dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5448dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5449dd5f9615SStefan Behrens if (received_uuid_changed && 5450d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5451d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5452dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5453dd5f9615SStefan Behrens root->root_key.objectid); 5454d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5455d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5456d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5457d87ff758SNikolay Borisov goto out; 5458d87ff758SNikolay Borisov } 5459d87ff758SNikolay Borisov } 54608ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 54618ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 54628ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 54633cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 54643cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 54653cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 54663cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 54678ea05e3aSAlexander Block 54680b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 54698ea05e3aSAlexander Block &root->root_key, &root->root_item); 54708ea05e3aSAlexander Block if (ret < 0) { 54713a45bb20SJeff Mahoney btrfs_end_transaction(trans); 54728ea05e3aSAlexander Block goto out; 5473dd5f9615SStefan Behrens } 5474dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5475cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5476dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5477dd5f9615SStefan Behrens root->root_key.objectid); 5478dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 547966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5480efd38150SNikolay Borisov btrfs_end_transaction(trans); 5481dd5f9615SStefan Behrens goto out; 5482dd5f9615SStefan Behrens } 5483dd5f9615SStefan Behrens } 54843a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5485abccd00fSHugo Mills out: 54860b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5487abccd00fSHugo Mills mnt_drop_write_file(file); 5488abccd00fSHugo Mills return ret; 5489abccd00fSHugo Mills } 5490abccd00fSHugo Mills 5491abccd00fSHugo Mills #ifdef CONFIG_64BIT 5492abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5493abccd00fSHugo Mills void __user *arg) 5494abccd00fSHugo Mills { 5495abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5496abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5497abccd00fSHugo Mills int ret = 0; 5498abccd00fSHugo Mills 5499abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 55007b9ea627SShailendra Verma if (IS_ERR(args32)) 55017b9ea627SShailendra Verma return PTR_ERR(args32); 5502abccd00fSHugo Mills 55038d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 550484dbeb87SDan Carpenter if (!args64) { 550584dbeb87SDan Carpenter ret = -ENOMEM; 5506abccd00fSHugo Mills goto out; 5507abccd00fSHugo Mills } 5508abccd00fSHugo Mills 5509abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5510abccd00fSHugo Mills args64->stransid = args32->stransid; 5511abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5512abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5513abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5514abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5515abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5516abccd00fSHugo Mills args64->flags = args32->flags; 5517abccd00fSHugo Mills 5518abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5519abccd00fSHugo Mills if (ret) 5520abccd00fSHugo Mills goto out; 5521abccd00fSHugo Mills 5522abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5523abccd00fSHugo Mills args32->stransid = args64->stransid; 5524abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5525abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5526abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5527abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5528abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5529abccd00fSHugo Mills args32->flags = args64->flags; 5530abccd00fSHugo Mills 5531abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5532abccd00fSHugo Mills if (ret) 5533abccd00fSHugo Mills ret = -EFAULT; 5534abccd00fSHugo Mills 5535abccd00fSHugo Mills out: 5536abccd00fSHugo Mills kfree(args32); 5537abccd00fSHugo Mills kfree(args64); 5538abccd00fSHugo Mills return ret; 5539abccd00fSHugo Mills } 5540abccd00fSHugo Mills #endif 5541abccd00fSHugo Mills 5542abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5543abccd00fSHugo Mills void __user *arg) 5544abccd00fSHugo Mills { 5545abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5546abccd00fSHugo Mills int ret = 0; 5547abccd00fSHugo Mills 5548abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 55497b9ea627SShailendra Verma if (IS_ERR(sa)) 55507b9ea627SShailendra Verma return PTR_ERR(sa); 5551abccd00fSHugo Mills 5552abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5553abccd00fSHugo Mills 5554abccd00fSHugo Mills if (ret) 5555abccd00fSHugo Mills goto out; 5556abccd00fSHugo Mills 55578ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 55588ea05e3aSAlexander Block if (ret) 55598ea05e3aSAlexander Block ret = -EFAULT; 55608ea05e3aSAlexander Block 55618ea05e3aSAlexander Block out: 55628ea05e3aSAlexander Block kfree(sa); 55638ea05e3aSAlexander Block return ret; 55648ea05e3aSAlexander Block } 55658ea05e3aSAlexander Block 5566867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5567867ab667Sjeff.liu { 55680b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55690b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5570a1b83ac5SAnand Jain size_t len; 5571867ab667Sjeff.liu int ret; 5572a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5573a1b83ac5SAnand Jain 55740b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 55750b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 55760b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5577a1b83ac5SAnand Jain 5578a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5579867ab667Sjeff.liu 5580867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 55810b246afaSJeff Mahoney btrfs_warn(fs_info, 55820b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 55830b246afaSJeff Mahoney --len); 5584867ab667Sjeff.liu } 5585867ab667Sjeff.liu 5586867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5587867ab667Sjeff.liu 5588867ab667Sjeff.liu return ret ? -EFAULT : 0; 5589867ab667Sjeff.liu } 5590867ab667Sjeff.liu 5591a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5592a8bfd4abSjeff.liu { 55930b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55940b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55950b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55960b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5597a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5598a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5599a8bfd4abSjeff.liu int ret; 5600a8bfd4abSjeff.liu 5601a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5602a8bfd4abSjeff.liu return -EPERM; 5603a8bfd4abSjeff.liu 5604a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5605a8bfd4abSjeff.liu return -EFAULT; 5606a8bfd4abSjeff.liu 5607a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 56080b246afaSJeff Mahoney btrfs_err(fs_info, 56095d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5610a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5611a8bfd4abSjeff.liu return -EINVAL; 5612a8bfd4abSjeff.liu } 5613a8bfd4abSjeff.liu 5614a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5615a8bfd4abSjeff.liu if (ret) 5616a8bfd4abSjeff.liu return ret; 5617a8bfd4abSjeff.liu 5618a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5619a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5620a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5621a8bfd4abSjeff.liu goto out_unlock; 5622a8bfd4abSjeff.liu } 5623a8bfd4abSjeff.liu 56240b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5625a8bfd4abSjeff.liu strcpy(super_block->label, label); 56260b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 56273a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5628a8bfd4abSjeff.liu 5629a8bfd4abSjeff.liu out_unlock: 5630a8bfd4abSjeff.liu mnt_drop_write_file(file); 5631a8bfd4abSjeff.liu return ret; 5632a8bfd4abSjeff.liu } 5633a8bfd4abSjeff.liu 56342eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 56352eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 56362eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 56372eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 56382eaa055fSJeff Mahoney 5639d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 56402eaa055fSJeff Mahoney { 56414d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 56422eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 56432eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 56442eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 56452eaa055fSJeff Mahoney }; 56462eaa055fSJeff Mahoney 56472eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 56482eaa055fSJeff Mahoney return -EFAULT; 56492eaa055fSJeff Mahoney 56502eaa055fSJeff Mahoney return 0; 56512eaa055fSJeff Mahoney } 56522eaa055fSJeff Mahoney 56532eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 56542eaa055fSJeff Mahoney { 56550b246afaSJeff Mahoney struct inode *inode = file_inode(file); 56560b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 56570b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 56582eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 56592eaa055fSJeff Mahoney 56602eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 56612eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 56622eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 56632eaa055fSJeff Mahoney 56642eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 56652eaa055fSJeff Mahoney return -EFAULT; 56662eaa055fSJeff Mahoney 56672eaa055fSJeff Mahoney return 0; 56682eaa055fSJeff Mahoney } 56692eaa055fSJeff Mahoney 56702ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 56713b02a68aSJeff Mahoney enum btrfs_feature_set set, 56722eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 56732eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 56742eaa055fSJeff Mahoney { 56753b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 56763b02a68aSJeff Mahoney char *names; 56772eaa055fSJeff Mahoney u64 disallowed, unsupported; 56782eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 56792eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 56802eaa055fSJeff Mahoney 56812eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 56822eaa055fSJeff Mahoney if (unsupported) { 56833b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 56843b02a68aSJeff Mahoney if (names) { 56850b246afaSJeff Mahoney btrfs_warn(fs_info, 56863b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 56873b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 56883b02a68aSJeff Mahoney kfree(names); 56893b02a68aSJeff Mahoney } else 56900b246afaSJeff Mahoney btrfs_warn(fs_info, 56912eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 56922eaa055fSJeff Mahoney type, unsupported); 56932eaa055fSJeff Mahoney return -EOPNOTSUPP; 56942eaa055fSJeff Mahoney } 56952eaa055fSJeff Mahoney 56962eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 56972eaa055fSJeff Mahoney if (disallowed) { 56983b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 56993b02a68aSJeff Mahoney if (names) { 57000b246afaSJeff Mahoney btrfs_warn(fs_info, 57013b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 57023b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 57033b02a68aSJeff Mahoney kfree(names); 57043b02a68aSJeff Mahoney } else 57050b246afaSJeff Mahoney btrfs_warn(fs_info, 57062eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 57072eaa055fSJeff Mahoney type, disallowed); 57082eaa055fSJeff Mahoney return -EPERM; 57092eaa055fSJeff Mahoney } 57102eaa055fSJeff Mahoney 57112eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 57122eaa055fSJeff Mahoney if (disallowed) { 57133b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 57143b02a68aSJeff Mahoney if (names) { 57150b246afaSJeff Mahoney btrfs_warn(fs_info, 57163b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 57173b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 57183b02a68aSJeff Mahoney kfree(names); 57193b02a68aSJeff Mahoney } else 57200b246afaSJeff Mahoney btrfs_warn(fs_info, 57212eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 57222eaa055fSJeff Mahoney type, disallowed); 57232eaa055fSJeff Mahoney return -EPERM; 57242eaa055fSJeff Mahoney } 57252eaa055fSJeff Mahoney 57262eaa055fSJeff Mahoney return 0; 57272eaa055fSJeff Mahoney } 57282eaa055fSJeff Mahoney 57292ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 57302ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 57312eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 57322eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 57332eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 57342eaa055fSJeff Mahoney 57352eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 57362eaa055fSJeff Mahoney { 57370b246afaSJeff Mahoney struct inode *inode = file_inode(file); 57380b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 57390b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 57400b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 57412eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 57422eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 57432eaa055fSJeff Mahoney u64 newflags; 57442eaa055fSJeff Mahoney int ret; 57452eaa055fSJeff Mahoney 57462eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 57472eaa055fSJeff Mahoney return -EPERM; 57482eaa055fSJeff Mahoney 57492eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 57502eaa055fSJeff Mahoney return -EFAULT; 57512eaa055fSJeff Mahoney 57522eaa055fSJeff Mahoney /* Nothing to do */ 57532eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 57542eaa055fSJeff Mahoney !flags[0].incompat_flags) 57552eaa055fSJeff Mahoney return 0; 57562eaa055fSJeff Mahoney 57572ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 57582eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 57592eaa055fSJeff Mahoney if (ret) 57602eaa055fSJeff Mahoney return ret; 57612eaa055fSJeff Mahoney 57622ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 57632eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 57642eaa055fSJeff Mahoney if (ret) 57652eaa055fSJeff Mahoney return ret; 57662eaa055fSJeff Mahoney 57672ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 57682eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 57692eaa055fSJeff Mahoney if (ret) 57702eaa055fSJeff Mahoney return ret; 57712eaa055fSJeff Mahoney 57727ab19625SDavid Sterba ret = mnt_want_write_file(file); 57737ab19625SDavid Sterba if (ret) 57747ab19625SDavid Sterba return ret; 57757ab19625SDavid Sterba 57768051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 57777ab19625SDavid Sterba if (IS_ERR(trans)) { 57787ab19625SDavid Sterba ret = PTR_ERR(trans); 57797ab19625SDavid Sterba goto out_drop_write; 57807ab19625SDavid Sterba } 57812eaa055fSJeff Mahoney 57820b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 57832eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 57842eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 57852eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 57862eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 57872eaa055fSJeff Mahoney 57882eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 57892eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 57902eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 57912eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 57922eaa055fSJeff Mahoney 57932eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 57942eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 57952eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 57962eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 57970b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 57982eaa055fSJeff Mahoney 57993a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 58007ab19625SDavid Sterba out_drop_write: 58017ab19625SDavid Sterba mnt_drop_write_file(file); 58027ab19625SDavid Sterba 58037ab19625SDavid Sterba return ret; 58042eaa055fSJeff Mahoney } 58052eaa055fSJeff Mahoney 58062351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 58072351f431SJosef Bacik { 58082351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 58092351f431SJosef Bacik int ret; 58102351f431SJosef Bacik 58112351f431SJosef Bacik if (compat) { 58122351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 58132351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 58142351f431SJosef Bacik 58152351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 58162351f431SJosef Bacik if (ret) 58172351f431SJosef Bacik return -EFAULT; 58182351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 58192351f431SJosef Bacik if (!arg) 58202351f431SJosef Bacik return -ENOMEM; 58212351f431SJosef Bacik arg->send_fd = args32.send_fd; 58222351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 58232351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 58242351f431SJosef Bacik arg->parent_root = args32.parent_root; 58252351f431SJosef Bacik arg->flags = args32.flags; 58262351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 58272351f431SJosef Bacik sizeof(args32.reserved)); 58282351f431SJosef Bacik #else 58292351f431SJosef Bacik return -ENOTTY; 58302351f431SJosef Bacik #endif 58312351f431SJosef Bacik } else { 58322351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 58332351f431SJosef Bacik if (IS_ERR(arg)) 58342351f431SJosef Bacik return PTR_ERR(arg); 58352351f431SJosef Bacik } 58362351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 58372351f431SJosef Bacik kfree(arg); 58382351f431SJosef Bacik return ret; 58392351f431SJosef Bacik } 58402351f431SJosef Bacik 5841f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5842f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5843f46b5a66SChristoph Hellwig { 58440b246afaSJeff Mahoney struct inode *inode = file_inode(file); 58450b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 58460b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 58474bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5848f46b5a66SChristoph Hellwig 5849f46b5a66SChristoph Hellwig switch (cmd) { 58506cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 58516cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 58526cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 58536cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 58546cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 58556cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5856f7039b1dSLi Dongyang case FITRIM: 5857f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5858f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5859fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5860fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5861fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 58623de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5863fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 58646f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 58656f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 586676dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 586776dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 58680caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 58690caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 58700caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 58710caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 58726ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 58736ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5874f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 58751e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 58761e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 58771e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5878f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5879198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5880f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 58812ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5882f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5883da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 58846b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 58856b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5886475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 58872ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5888475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 58892ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5890f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 58919ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5892ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5893ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5894cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5895cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5896ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5897ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5898d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5899d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5900d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5901d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5902d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5903d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 59041406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 59052ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 59069b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 59079b199859SFilipe David Borba Manana int ret; 59089b199859SFilipe David Borba Manana 590982b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 59109b199859SFilipe David Borba Manana if (ret) 59119b199859SFilipe David Borba Manana return ret; 59120b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 59132fad4e83SDavid Sterba /* 59142fad4e83SDavid Sterba * The transaction thread may want to do more work, 591501327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 59162fad4e83SDavid Sterba * processing uncleaned subvols. 59172fad4e83SDavid Sterba */ 59180b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 59199b199859SFilipe David Borba Manana return ret; 59209b199859SFilipe David Borba Manana } 592146204592SSage Weil case BTRFS_IOC_START_SYNC: 59229a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 592346204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 59242ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5925475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5926b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5927475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 59282ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5929475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 59302ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5931c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 59329ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5933837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 59342ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 593519a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 59362ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 59378ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 59388ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5939abccd00fSHugo Mills #ifdef CONFIG_64BIT 5940abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5941abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5942abccd00fSHugo Mills #endif 594331db9f7cSAlexander Block case BTRFS_IOC_SEND: 59442351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 59452351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 59462351f431SJosef Bacik case BTRFS_IOC_SEND_32: 59472351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 59482351f431SJosef Bacik #endif 5949c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 59502ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 59515d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5952905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 59535d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5954905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 59555d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5956905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 59575d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5958905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 59592f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 59602f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 59612f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 59622f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 596357254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 596457254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 59653f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 59662ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5967867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5968867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5969a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5970a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 59712eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5972d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 59732eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 59742eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 59752eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 59762eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5977e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5978e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5979025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5980025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5981b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5982b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 598342e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 598442e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 598523d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 598623d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5987f46b5a66SChristoph Hellwig } 5988f46b5a66SChristoph Hellwig 5989f46b5a66SChristoph Hellwig return -ENOTTY; 5990f46b5a66SChristoph Hellwig } 59914c63c245SLuke Dashjr 59924c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 59934c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 59944c63c245SLuke Dashjr { 59952a362249SJeff Mahoney /* 59962a362249SJeff Mahoney * These all access 32-bit values anyway so no further 59972a362249SJeff Mahoney * handling is necessary. 59982a362249SJeff Mahoney */ 59994c63c245SLuke Dashjr switch (cmd) { 60004c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 60014c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 60024c63c245SLuke Dashjr break; 60034c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 60044c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 60054c63c245SLuke Dashjr break; 60064c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 60074c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 60084c63c245SLuke Dashjr break; 60094c63c245SLuke Dashjr } 60104c63c245SLuke Dashjr 60114c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 60124c63c245SLuke Dashjr } 60134c63c245SLuke Dashjr #endif 6014