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 293eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 294eede2bf3SOmar Sandoval ret = -ETXTBSY; 295eede2bf3SOmar Sandoval goto out_unlock; 296eede2bf3SOmar Sandoval } 297eede2bf3SOmar Sandoval 2985c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_COMPRESS; 2995c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_NOCOMPRESS; 30063541927SFilipe David Borba Manana 30193370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 30293370509SDavid Sterba if (!comp || comp[0] == 0) 30393370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 30493370509SDavid Sterba 30563541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", 30663541927SFilipe David Borba Manana comp, strlen(comp), 0); 30763541927SFilipe David Borba Manana if (ret) 30863541927SFilipe David Borba Manana goto out_drop; 30963541927SFilipe David Borba Manana 310ebcb904dSLi Zefan } else { 31178a017a2SFilipe Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 31278a017a2SFilipe Manana if (ret && ret != -ENODATA) 31378a017a2SFilipe Manana goto out_drop; 3145c57b8b6SDavid Sterba binode->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 31575e7cb7fSLiu Bo } 3166cbff00fSChristoph Hellwig 3174da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 318f062abf0SLi Zefan if (IS_ERR(trans)) { 319f062abf0SLi Zefan ret = PTR_ERR(trans); 320f062abf0SLi Zefan goto out_drop; 321f062abf0SLi Zefan } 3226cbff00fSChristoph Hellwig 3237b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3240c4d2d95SJosef Bacik inode_inc_iversion(inode); 325c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3266cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3276cbff00fSChristoph Hellwig 3283a45bb20SJeff Mahoney btrfs_end_transaction(trans); 329f062abf0SLi Zefan out_drop: 330f062abf0SLi Zefan if (ret) { 3315c57b8b6SDavid Sterba binode->flags = old_flags; 3325c57b8b6SDavid Sterba inode->i_flags = old_i_flags; 333f062abf0SLi Zefan } 3346cbff00fSChristoph Hellwig 3356cbff00fSChristoph Hellwig out_unlock: 3365955102cSAl Viro inode_unlock(inode); 337e7848683SJan Kara mnt_drop_write_file(file); 3382d4e6f6aSliubo return ret; 3396cbff00fSChristoph Hellwig } 3406cbff00fSChristoph Hellwig 34119f93b3cSDavid Sterba /* 34219f93b3cSDavid Sterba * Translate btrfs internal inode flags to xflags as expected by the 34319f93b3cSDavid Sterba * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are 34419f93b3cSDavid Sterba * silently dropped. 34519f93b3cSDavid Sterba */ 34619f93b3cSDavid Sterba static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags) 34719f93b3cSDavid Sterba { 34819f93b3cSDavid Sterba unsigned int xflags = 0; 34919f93b3cSDavid Sterba 35019f93b3cSDavid Sterba if (flags & BTRFS_INODE_APPEND) 35119f93b3cSDavid Sterba xflags |= FS_XFLAG_APPEND; 35219f93b3cSDavid Sterba if (flags & BTRFS_INODE_IMMUTABLE) 35319f93b3cSDavid Sterba xflags |= FS_XFLAG_IMMUTABLE; 35419f93b3cSDavid Sterba if (flags & BTRFS_INODE_NOATIME) 35519f93b3cSDavid Sterba xflags |= FS_XFLAG_NOATIME; 35619f93b3cSDavid Sterba if (flags & BTRFS_INODE_NODUMP) 35719f93b3cSDavid Sterba xflags |= FS_XFLAG_NODUMP; 35819f93b3cSDavid Sterba if (flags & BTRFS_INODE_SYNC) 35919f93b3cSDavid Sterba xflags |= FS_XFLAG_SYNC; 36019f93b3cSDavid Sterba 36119f93b3cSDavid Sterba return xflags; 36219f93b3cSDavid Sterba } 36319f93b3cSDavid Sterba 36419f93b3cSDavid Sterba /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */ 36519f93b3cSDavid Sterba static int check_xflags(unsigned int flags) 36619f93b3cSDavid Sterba { 36719f93b3cSDavid Sterba if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME | 36819f93b3cSDavid Sterba FS_XFLAG_NODUMP | FS_XFLAG_SYNC)) 36919f93b3cSDavid Sterba return -EOPNOTSUPP; 37019f93b3cSDavid Sterba return 0; 37119f93b3cSDavid Sterba } 37219f93b3cSDavid Sterba 373e4202ac9SDavid Sterba /* 374e4202ac9SDavid Sterba * Set the xflags from the internal inode flags. The remaining items of fsxattr 375e4202ac9SDavid Sterba * are zeroed. 376e4202ac9SDavid Sterba */ 377e4202ac9SDavid Sterba static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg) 378e4202ac9SDavid Sterba { 379e4202ac9SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 380e4202ac9SDavid Sterba struct fsxattr fa; 381e4202ac9SDavid Sterba 382e4202ac9SDavid Sterba memset(&fa, 0, sizeof(fa)); 383e4202ac9SDavid Sterba fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags); 384e4202ac9SDavid Sterba 385e4202ac9SDavid Sterba if (copy_to_user(arg, &fa, sizeof(fa))) 386e4202ac9SDavid Sterba return -EFAULT; 387e4202ac9SDavid Sterba 388e4202ac9SDavid Sterba return 0; 389e4202ac9SDavid Sterba } 390e4202ac9SDavid Sterba 391025f2121SDavid Sterba static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg) 392025f2121SDavid Sterba { 393025f2121SDavid Sterba struct inode *inode = file_inode(file); 394025f2121SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 395025f2121SDavid Sterba struct btrfs_root *root = binode->root; 396025f2121SDavid Sterba struct btrfs_trans_handle *trans; 397025f2121SDavid Sterba struct fsxattr fa; 398025f2121SDavid Sterba unsigned old_flags; 399025f2121SDavid Sterba unsigned old_i_flags; 400025f2121SDavid Sterba int ret = 0; 401025f2121SDavid Sterba 402025f2121SDavid Sterba if (!inode_owner_or_capable(inode)) 403025f2121SDavid Sterba return -EPERM; 404025f2121SDavid Sterba 405025f2121SDavid Sterba if (btrfs_root_readonly(root)) 406025f2121SDavid Sterba return -EROFS; 407025f2121SDavid Sterba 408025f2121SDavid Sterba memset(&fa, 0, sizeof(fa)); 409025f2121SDavid Sterba if (copy_from_user(&fa, arg, sizeof(fa))) 410025f2121SDavid Sterba return -EFAULT; 411025f2121SDavid Sterba 412025f2121SDavid Sterba ret = check_xflags(fa.fsx_xflags); 413025f2121SDavid Sterba if (ret) 414025f2121SDavid Sterba return ret; 415025f2121SDavid Sterba 416025f2121SDavid Sterba if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0) 417025f2121SDavid Sterba return -EOPNOTSUPP; 418025f2121SDavid Sterba 419025f2121SDavid Sterba ret = mnt_want_write_file(file); 420025f2121SDavid Sterba if (ret) 421025f2121SDavid Sterba return ret; 422025f2121SDavid Sterba 423025f2121SDavid Sterba inode_lock(inode); 424025f2121SDavid Sterba 425025f2121SDavid Sterba old_flags = binode->flags; 426025f2121SDavid Sterba old_i_flags = inode->i_flags; 427025f2121SDavid Sterba 428025f2121SDavid Sterba /* We need the capabilities to change append-only or immutable inode */ 429025f2121SDavid Sterba if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) || 430025f2121SDavid Sterba (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) && 431025f2121SDavid Sterba !capable(CAP_LINUX_IMMUTABLE)) { 432025f2121SDavid Sterba ret = -EPERM; 433025f2121SDavid Sterba goto out_unlock; 434025f2121SDavid Sterba } 435025f2121SDavid Sterba 436025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_SYNC) 437025f2121SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 438025f2121SDavid Sterba else 439025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 440025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE) 441025f2121SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 442025f2121SDavid Sterba else 443025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 444025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_APPEND) 445025f2121SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 446025f2121SDavid Sterba else 447025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 448025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NODUMP) 449025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 450025f2121SDavid Sterba else 451025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 452025f2121SDavid Sterba if (fa.fsx_xflags & FS_XFLAG_NOATIME) 453025f2121SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 454025f2121SDavid Sterba else 455025f2121SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 456025f2121SDavid Sterba 457025f2121SDavid Sterba /* 1 item for the inode */ 458025f2121SDavid Sterba trans = btrfs_start_transaction(root, 1); 459025f2121SDavid Sterba if (IS_ERR(trans)) { 460025f2121SDavid Sterba ret = PTR_ERR(trans); 461025f2121SDavid Sterba goto out_unlock; 462025f2121SDavid Sterba } 463025f2121SDavid Sterba 464025f2121SDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 465025f2121SDavid Sterba inode_inc_iversion(inode); 466025f2121SDavid Sterba inode->i_ctime = current_time(inode); 467025f2121SDavid Sterba ret = btrfs_update_inode(trans, root, inode); 468025f2121SDavid Sterba 469025f2121SDavid Sterba btrfs_end_transaction(trans); 470025f2121SDavid Sterba 471025f2121SDavid Sterba out_unlock: 472025f2121SDavid Sterba if (ret) { 473025f2121SDavid Sterba binode->flags = old_flags; 474025f2121SDavid Sterba inode->i_flags = old_i_flags; 475025f2121SDavid Sterba } 476025f2121SDavid Sterba 477025f2121SDavid Sterba inode_unlock(inode); 478025f2121SDavid Sterba mnt_drop_write_file(file); 479025f2121SDavid Sterba 480025f2121SDavid Sterba return ret; 481025f2121SDavid Sterba } 482025f2121SDavid Sterba 4836cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 4846cbff00fSChristoph Hellwig { 485496ad9aaSAl Viro struct inode *inode = file_inode(file); 4866cbff00fSChristoph Hellwig 4876cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 4886cbff00fSChristoph Hellwig } 489f46b5a66SChristoph Hellwig 490f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 491f7039b1dSLi Dongyang { 4920b246afaSJeff Mahoney struct inode *inode = file_inode(file); 4930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 494f7039b1dSLi Dongyang struct btrfs_device *device; 495f7039b1dSLi Dongyang struct request_queue *q; 496f7039b1dSLi Dongyang struct fstrim_range range; 497f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 498f7039b1dSLi Dongyang u64 num_devices = 0; 499f7039b1dSLi Dongyang int ret; 500f7039b1dSLi Dongyang 501f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 502f7039b1dSLi Dongyang return -EPERM; 503f7039b1dSLi Dongyang 504f35f06c3SFilipe Manana /* 505f35f06c3SFilipe Manana * If the fs is mounted with nologreplay, which requires it to be 506f35f06c3SFilipe Manana * mounted in RO mode as well, we can not allow discard on free space 507f35f06c3SFilipe Manana * inside block groups, because log trees refer to extents that are not 508f35f06c3SFilipe Manana * pinned in a block group's free space cache (pinning the extents is 509f35f06c3SFilipe Manana * precisely the first phase of replaying a log tree). 510f35f06c3SFilipe Manana */ 511f35f06c3SFilipe Manana if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 512f35f06c3SFilipe Manana return -EROFS; 513f35f06c3SFilipe Manana 5141f78160cSXiao Guangrong rcu_read_lock(); 5151f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 5161f78160cSXiao Guangrong dev_list) { 517f7039b1dSLi Dongyang if (!device->bdev) 518f7039b1dSLi Dongyang continue; 519f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 520f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 521f7039b1dSLi Dongyang num_devices++; 52250d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 523f7039b1dSLi Dongyang minlen); 524f7039b1dSLi Dongyang } 525f7039b1dSLi Dongyang } 5261f78160cSXiao Guangrong rcu_read_unlock(); 527f4c697e6SLukas Czerner 528f7039b1dSLi Dongyang if (!num_devices) 529f7039b1dSLi Dongyang return -EOPNOTSUPP; 530f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 531f7039b1dSLi Dongyang return -EFAULT; 5326ba9fc8eSQu Wenruo 5336ba9fc8eSQu Wenruo /* 5346ba9fc8eSQu Wenruo * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 5356ba9fc8eSQu Wenruo * block group is in the logical address space, which can be any 5366ba9fc8eSQu Wenruo * sectorsize aligned bytenr in the range [0, U64_MAX]. 5376ba9fc8eSQu Wenruo */ 5386ba9fc8eSQu Wenruo if (range.len < fs_info->sb->s_blocksize) 539f4c697e6SLukas Czerner return -EINVAL; 540f7039b1dSLi Dongyang 541f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5422ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 543f7039b1dSLi Dongyang if (ret < 0) 544f7039b1dSLi Dongyang return ret; 545f7039b1dSLi Dongyang 546f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 547f7039b1dSLi Dongyang return -EFAULT; 548f7039b1dSLi Dongyang 549f7039b1dSLi Dongyang return 0; 550f7039b1dSLi Dongyang } 551f7039b1dSLi Dongyang 552dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 553dd5f9615SStefan Behrens { 55446e0f66aSChris Mason int i; 55546e0f66aSChris Mason 55646e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 55746e0f66aSChris Mason if (uuid[i]) 55846e0f66aSChris Mason return 0; 55946e0f66aSChris Mason } 56046e0f66aSChris Mason return 1; 561dd5f9615SStefan Behrens } 562dd5f9615SStefan Behrens 563d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 564cb8e7090SChristoph Hellwig struct dentry *dentry, 56552f75f4fSDavid Sterba const char *name, int namelen, 5666f72c7e2SArne Jansen u64 *async_transid, 5678696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 568f46b5a66SChristoph Hellwig { 5690b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 570f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 571f46b5a66SChristoph Hellwig struct btrfs_key key; 57249a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 573f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 574f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 575d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 57676dda93cSYan, Zheng struct btrfs_root *new_root; 577d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 57895582b00SDeepa Dinamani struct timespec64 cur_time = current_time(dir); 5795662344bSTsutomu Itoh struct inode *inode; 580f46b5a66SChristoph Hellwig int ret; 581f46b5a66SChristoph Hellwig int err; 582f46b5a66SChristoph Hellwig u64 objectid; 583f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 5843de4586cSChris Mason u64 index = 0; 5858ea05e3aSAlexander Block uuid_le new_uuid; 586f46b5a66SChristoph Hellwig 58749a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 58849a3c4d9SDavid Sterba if (!root_item) 58949a3c4d9SDavid Sterba return -ENOMEM; 59049a3c4d9SDavid Sterba 5910b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5922fbe8c8aSAl Viro if (ret) 59349a3c4d9SDavid Sterba goto fail_free; 5946a912213SJosef Bacik 595e09fe2d2SQu Wenruo /* 596e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 59701327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 598e09fe2d2SQu Wenruo */ 59949a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 60049a3c4d9SDavid Sterba ret = -ENOSPC; 60149a3c4d9SDavid Sterba goto fail_free; 60249a3c4d9SDavid Sterba } 603e09fe2d2SQu Wenruo 604d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 6059ed74f2dSJosef Bacik /* 606d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 607d5c12070SMiao Xie * of create_snapshot(). 6089ed74f2dSJosef Bacik */ 609c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 610d5c12070SMiao Xie if (ret) 61149a3c4d9SDavid Sterba goto fail_free; 612f46b5a66SChristoph Hellwig 613d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 614d5c12070SMiao Xie if (IS_ERR(trans)) { 615d5c12070SMiao Xie ret = PTR_ERR(trans); 6167775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 61749a3c4d9SDavid Sterba goto fail_free; 618d5c12070SMiao Xie } 619d5c12070SMiao Xie trans->block_rsv = &block_rsv; 620d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 621f46b5a66SChristoph Hellwig 622a9377422SLu Fengqi ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); 6236f72c7e2SArne Jansen if (ret) 6246f72c7e2SArne Jansen goto fail; 6256f72c7e2SArne Jansen 6264d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6278e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6288e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6298e8a1e31SJosef Bacik goto fail; 6308e8a1e31SJosef Bacik } 631f46b5a66SChristoph Hellwig 632f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 633f46b5a66SChristoph Hellwig 63449a3c4d9SDavid Sterba inode_item = &root_item->inode; 6353cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6363cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6373cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 638da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6390b246afaSJeff Mahoney fs_info->nodesize); 6403cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 641f46b5a66SChristoph Hellwig 64249a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 64349a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6443cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 64508fe4db1SLi Zefan 64649a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 64749a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 64849a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 64949a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 65049a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 65149a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 652f46b5a66SChristoph Hellwig 65349a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 65449a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6558ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 65649a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 65749a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 65849a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 65949a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 66049a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 66149a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 662f46b5a66SChristoph Hellwig 663925baeddSChris Mason btrfs_tree_unlock(leaf); 664f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 665f46b5a66SChristoph Hellwig leaf = NULL; 666f46b5a66SChristoph Hellwig 66749a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 668f46b5a66SChristoph Hellwig 669f46b5a66SChristoph Hellwig key.objectid = objectid; 6705d4f98a2SYan Zheng key.offset = 0; 671962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6720b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 67349a3c4d9SDavid Sterba root_item); 674f46b5a66SChristoph Hellwig if (ret) 675f46b5a66SChristoph Hellwig goto fail; 676f46b5a66SChristoph Hellwig 67776dda93cSYan, Zheng key.offset = (u64)-1; 6780b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 67979787eaaSJeff Mahoney if (IS_ERR(new_root)) { 68079787eaaSJeff Mahoney ret = PTR_ERR(new_root); 68166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 68279787eaaSJeff Mahoney goto fail; 68379787eaaSJeff Mahoney } 68476dda93cSYan, Zheng 68576dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 68676dda93cSYan, Zheng 68763541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 688ce598979SMark Fasheh if (ret) { 689ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 69066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 691ce598979SMark Fasheh goto fail; 692ce598979SMark Fasheh } 693ce598979SMark Fasheh 694f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 695f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 696f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 697f32e48e9SChandan Rajendra 698f46b5a66SChristoph Hellwig /* 699f46b5a66SChristoph Hellwig * insert the directory item 700f46b5a66SChristoph Hellwig */ 701877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 70279787eaaSJeff Mahoney if (ret) { 70366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 70479787eaaSJeff Mahoney goto fail; 70579787eaaSJeff Mahoney } 7063de4586cSChris Mason 707684572dfSLu Fengqi ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, 7083de4586cSChris Mason BTRFS_FT_DIR, index); 70979787eaaSJeff Mahoney if (ret) { 71066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 711f46b5a66SChristoph Hellwig goto fail; 71279787eaaSJeff Mahoney } 7130660b5afSChris Mason 7146ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 71552c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 71652c26179SYan Zheng BUG_ON(ret); 71752c26179SYan Zheng 7186025c19fSLu Fengqi ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, 7194a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7200660b5afSChris Mason BUG_ON(ret); 7210660b5afSChris Mason 722cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7236bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 724dd5f9615SStefan Behrens if (ret) 72566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 726dd5f9615SStefan Behrens 727f46b5a66SChristoph Hellwig fail: 72849a3c4d9SDavid Sterba kfree(root_item); 729d5c12070SMiao Xie trans->block_rsv = NULL; 730d5c12070SMiao Xie trans->bytes_reserved = 0; 7317775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 732de6e8200SLiu Bo 73372fd032eSSage Weil if (async_transid) { 73472fd032eSSage Weil *async_transid = trans->transid; 7353a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 73600d71c9cSMiao Xie if (err) 7373a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73872fd032eSSage Weil } else { 7393a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 74072fd032eSSage Weil } 741f46b5a66SChristoph Hellwig if (err && !ret) 742f46b5a66SChristoph Hellwig ret = err; 7431a65e24bSChris Mason 7445662344bSTsutomu Itoh if (!ret) { 7455662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 746de6e8200SLiu Bo if (IS_ERR(inode)) 747de6e8200SLiu Bo return PTR_ERR(inode); 7485662344bSTsutomu Itoh d_instantiate(dentry, inode); 7495662344bSTsutomu Itoh } 750f46b5a66SChristoph Hellwig return ret; 75149a3c4d9SDavid Sterba 75249a3c4d9SDavid Sterba fail_free: 75349a3c4d9SDavid Sterba kfree(root_item); 75449a3c4d9SDavid Sterba return ret; 755f46b5a66SChristoph Hellwig } 756f46b5a66SChristoph Hellwig 757e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 75861d7e4cbSDavid Sterba struct dentry *dentry, 759e9662f70SMiao Xie u64 *async_transid, bool readonly, 760e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 761f46b5a66SChristoph Hellwig { 7620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7632e4bfab9SYan, Zheng struct inode *inode; 764f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 765f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7662e4bfab9SYan, Zheng int ret; 7678ecebf4dSRobbie Ko bool snapshot_force_cow = false; 768f46b5a66SChristoph Hellwig 76927cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 770f46b5a66SChristoph Hellwig return -EINVAL; 771f46b5a66SChristoph Hellwig 772eede2bf3SOmar Sandoval if (atomic_read(&root->nr_swapfiles)) { 773eede2bf3SOmar Sandoval btrfs_warn(fs_info, 774eede2bf3SOmar Sandoval "cannot snapshot subvolume with active swapfile"); 775eede2bf3SOmar Sandoval return -ETXTBSY; 776eede2bf3SOmar Sandoval } 777eede2bf3SOmar Sandoval 77823269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 779a1ee7362SDavid Sterba if (!pending_snapshot) 780a1ee7362SDavid Sterba return -ENOMEM; 781a1ee7362SDavid Sterba 782b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 78323269bf5SDavid Sterba GFP_KERNEL); 7848546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7858546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 786b0c0ea63SDavid Sterba ret = -ENOMEM; 787b0c0ea63SDavid Sterba goto free_pending; 788b0c0ea63SDavid Sterba } 789b0c0ea63SDavid Sterba 7908ecebf4dSRobbie Ko /* 7918ecebf4dSRobbie Ko * Force new buffered writes to reserve space even when NOCOW is 7928ecebf4dSRobbie Ko * possible. This is to avoid later writeback (running dealloc) to 7938ecebf4dSRobbie Ko * fallback to COW mode and unexpectedly fail with ENOSPC. 7948ecebf4dSRobbie Ko */ 795ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7964e857c58SPeter Zijlstra smp_mb__after_atomic(); 79745bac0f3SLiu Bo /* wait for no snapshot writes */ 79845bac0f3SLiu Bo wait_event(root->subv_writers->wait, 79945bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 8008257b2dcSMiao Xie 8013cd24c69SEthan Lien ret = btrfs_start_delalloc_snapshot(root); 8026a03843dSMiao Xie if (ret) 803a1ee7362SDavid Sterba goto dec_and_free; 8046a03843dSMiao Xie 8058ecebf4dSRobbie Ko /* 8068ecebf4dSRobbie Ko * All previous writes have started writeback in NOCOW mode, so now 8078ecebf4dSRobbie Ko * we force future writes to fallback to COW mode during snapshot 8088ecebf4dSRobbie Ko * creation. 8098ecebf4dSRobbie Ko */ 8108ecebf4dSRobbie Ko atomic_inc(&root->snapshot_force_cow); 8118ecebf4dSRobbie Ko snapshot_force_cow = true; 8128ecebf4dSRobbie Ko 8136374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 8146a03843dSMiao Xie 81566d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 81666d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 817d5c12070SMiao Xie /* 818d5c12070SMiao Xie * 1 - parent dir inode 819d5c12070SMiao Xie * 2 - dir entries 820d5c12070SMiao Xie * 1 - root item 821d5c12070SMiao Xie * 2 - root ref/backref 822d5c12070SMiao Xie * 1 - root of snapshot 823dd5f9615SStefan Behrens * 1 - UUID item 824d5c12070SMiao Xie */ 825d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 826dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 827ee3441b4SJeff Mahoney false); 828d5c12070SMiao Xie if (ret) 829a1ee7362SDavid Sterba goto dec_and_free; 830d5c12070SMiao Xie 831a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 832a22285a6SYan, Zheng pending_snapshot->root = root; 833b83cc969SLi Zefan pending_snapshot->readonly = readonly; 834e9662f70SMiao Xie pending_snapshot->dir = dir; 8358696c533SMiao Xie pending_snapshot->inherit = inherit; 836a22285a6SYan, Zheng 837d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 838a22285a6SYan, Zheng if (IS_ERR(trans)) { 839a22285a6SYan, Zheng ret = PTR_ERR(trans); 840a22285a6SYan, Zheng goto fail; 841a22285a6SYan, Zheng } 842a22285a6SYan, Zheng 8430b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 844a22285a6SYan, Zheng list_add(&pending_snapshot->list, 845a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8460b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 84772fd032eSSage Weil if (async_transid) { 84872fd032eSSage Weil *async_transid = trans->transid; 8493a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 85000d71c9cSMiao Xie if (ret) 8513a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 85272fd032eSSage Weil } else { 8533a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 85472fd032eSSage Weil } 855aec8030aSMiao Xie if (ret) 856c37b2b62SJosef Bacik goto fail; 857a22285a6SYan, Zheng 858a22285a6SYan, Zheng ret = pending_snapshot->error; 859f46b5a66SChristoph Hellwig if (ret) 8602e4bfab9SYan, Zheng goto fail; 861f46b5a66SChristoph Hellwig 862d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 863d3797308SChris Mason if (ret) 864d3797308SChris Mason goto fail; 865d3797308SChris Mason 8662b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8672e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8682e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8692e4bfab9SYan, Zheng goto fail; 8702e4bfab9SYan, Zheng } 8715662344bSTsutomu Itoh 8722e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8732e4bfab9SYan, Zheng ret = 0; 8742e4bfab9SYan, Zheng fail: 8757775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 876a1ee7362SDavid Sterba dec_and_free: 8778ecebf4dSRobbie Ko if (snapshot_force_cow) 8788ecebf4dSRobbie Ko atomic_dec(&root->snapshot_force_cow); 879ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8804625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 881b0c0ea63SDavid Sterba free_pending: 882b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8838546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 884a1ee7362SDavid Sterba kfree(pending_snapshot); 885a1ee7362SDavid Sterba 886f46b5a66SChristoph Hellwig return ret; 887f46b5a66SChristoph Hellwig } 888f46b5a66SChristoph Hellwig 8894260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8904260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8914260f7c7SSage Weil * whether the type of victim is right. 8924260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8934260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8944260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8954260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8964260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8974260f7c7SSage Weil * a. be owner of dir, or 8984260f7c7SSage Weil * b. be owner of victim, or 8994260f7c7SSage Weil * c. have CAP_FOWNER capability 90001327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 9014260f7c7SSage Weil * links pointing to it. 9024260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 9034260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 9044260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 9054260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 9064260f7c7SSage Weil * nfs_async_unlink(). 9074260f7c7SSage Weil */ 9084260f7c7SSage Weil 9094260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 9104260f7c7SSage Weil { 9114260f7c7SSage Weil int error; 9124260f7c7SSage Weil 9132b0143b5SDavid Howells if (d_really_is_negative(victim)) 9144260f7c7SSage Weil return -ENOENT; 9154260f7c7SSage Weil 9162b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 9174fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 9184260f7c7SSage Weil 9194260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 9204260f7c7SSage Weil if (error) 9214260f7c7SSage Weil return error; 9224260f7c7SSage Weil if (IS_APPEND(dir)) 9234260f7c7SSage Weil return -EPERM; 9242b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9252b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9264260f7c7SSage Weil return -EPERM; 9274260f7c7SSage Weil if (isdir) { 928e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9294260f7c7SSage Weil return -ENOTDIR; 9304260f7c7SSage Weil if (IS_ROOT(victim)) 9314260f7c7SSage Weil return -EBUSY; 932e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9334260f7c7SSage Weil return -EISDIR; 9344260f7c7SSage Weil if (IS_DEADDIR(dir)) 9354260f7c7SSage Weil return -ENOENT; 9364260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9374260f7c7SSage Weil return -EBUSY; 9384260f7c7SSage Weil return 0; 9394260f7c7SSage Weil } 9404260f7c7SSage Weil 941cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 942cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 943cb8e7090SChristoph Hellwig { 9442b0143b5SDavid Howells if (d_really_is_positive(child)) 945cb8e7090SChristoph Hellwig return -EEXIST; 946cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 947cb8e7090SChristoph Hellwig return -ENOENT; 948cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 949cb8e7090SChristoph Hellwig } 950cb8e7090SChristoph Hellwig 951cb8e7090SChristoph Hellwig /* 952cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 953cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 954cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 955cb8e7090SChristoph Hellwig */ 95692872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 95752f75f4fSDavid Sterba const char *name, int namelen, 95872fd032eSSage Weil struct btrfs_root *snap_src, 9596f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9608696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 961cb8e7090SChristoph Hellwig { 9622b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9630b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 964cb8e7090SChristoph Hellwig struct dentry *dentry; 965cb8e7090SChristoph Hellwig int error; 966cb8e7090SChristoph Hellwig 96700235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 96800235411SAl Viro if (error == -EINTR) 96900235411SAl Viro return error; 970cb8e7090SChristoph Hellwig 971cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 972cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 973cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 974cb8e7090SChristoph Hellwig goto out_unlock; 975cb8e7090SChristoph Hellwig 97676dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 977cb8e7090SChristoph Hellwig if (error) 978a874a63eSLiu Bo goto out_dput; 979cb8e7090SChristoph Hellwig 9809c52057cSChris Mason /* 9819c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9829c52057cSChris Mason * check for them now when we can safely fail 9839c52057cSChris Mason */ 9849c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9859c52057cSChris Mason dir->i_ino, name, 9869c52057cSChris Mason namelen); 9879c52057cSChris Mason if (error) 9889c52057cSChris Mason goto out_dput; 9899c52057cSChris Mason 9900b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 99176dda93cSYan, Zheng 99276dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 99376dda93cSYan, Zheng goto out_up_read; 99476dda93cSYan, Zheng 9953de4586cSChris Mason if (snap_src) { 99661d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9976f72c7e2SArne Jansen async_transid, readonly, inherit); 9983de4586cSChris Mason } else { 999d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 1000d5c12070SMiao Xie async_transid, inherit); 10013de4586cSChris Mason } 100276dda93cSYan, Zheng if (!error) 100376dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 100476dda93cSYan, Zheng out_up_read: 10050b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 1006cb8e7090SChristoph Hellwig out_dput: 1007cb8e7090SChristoph Hellwig dput(dentry); 1008cb8e7090SChristoph Hellwig out_unlock: 10095955102cSAl Viro inode_unlock(dir); 1010cb8e7090SChristoph Hellwig return error; 1011cb8e7090SChristoph Hellwig } 1012cb8e7090SChristoph Hellwig 10134cb5300bSChris Mason /* 10144cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 10154cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 10164cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 10174cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 10184cb5300bSChris Mason * part of the file 10194cb5300bSChris Mason */ 1020aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 10214cb5300bSChris Mason { 10224cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10234cb5300bSChris Mason struct extent_map *em = NULL; 10244cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10254cb5300bSChris Mason u64 end; 10264cb5300bSChris Mason 10274cb5300bSChris Mason read_lock(&em_tree->lock); 102809cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10294cb5300bSChris Mason read_unlock(&em_tree->lock); 10304cb5300bSChris Mason 10314cb5300bSChris Mason if (em) { 10324cb5300bSChris Mason end = extent_map_end(em); 10334cb5300bSChris Mason free_extent_map(em); 10344cb5300bSChris Mason if (end - offset > thresh) 10354cb5300bSChris Mason return 0; 10364cb5300bSChris Mason } 10374cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10384cb5300bSChris Mason thresh /= 2; 10394cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10404cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10414cb5300bSChris Mason if (end >= thresh) 10424cb5300bSChris Mason return 0; 10434cb5300bSChris Mason return 1; 10444cb5300bSChris Mason } 10454cb5300bSChris Mason 10464cb5300bSChris Mason /* 10474cb5300bSChris Mason * helper function to walk through a file and find extents 10484cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10494cb5300bSChris Mason * 10504cb5300bSChris Mason * This is used by the defragging code to find new and small 10514cb5300bSChris Mason * extents 10524cb5300bSChris Mason */ 10534cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10544cb5300bSChris Mason struct inode *inode, u64 newer_than, 1055aab110abSDavid Sterba u64 *off, u32 thresh) 10564cb5300bSChris Mason { 10574cb5300bSChris Mason struct btrfs_path *path; 10584cb5300bSChris Mason struct btrfs_key min_key; 10594cb5300bSChris Mason struct extent_buffer *leaf; 10604cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10614cb5300bSChris Mason int type; 10624cb5300bSChris Mason int ret; 10634a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10644cb5300bSChris Mason 10654cb5300bSChris Mason path = btrfs_alloc_path(); 10664cb5300bSChris Mason if (!path) 10674cb5300bSChris Mason return -ENOMEM; 10684cb5300bSChris Mason 1069a4689d2bSDavid Sterba min_key.objectid = ino; 10704cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10714cb5300bSChris Mason min_key.offset = *off; 10724cb5300bSChris Mason 10734cb5300bSChris Mason while (1) { 10746174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10754cb5300bSChris Mason if (ret != 0) 10764cb5300bSChris Mason goto none; 1077f094c9bdSFilipe Manana process_slot: 1078a4689d2bSDavid Sterba if (min_key.objectid != ino) 10794cb5300bSChris Mason goto none; 10804cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10814cb5300bSChris Mason goto none; 10824cb5300bSChris Mason 10834cb5300bSChris Mason leaf = path->nodes[0]; 10844cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10854cb5300bSChris Mason struct btrfs_file_extent_item); 10864cb5300bSChris Mason 10874cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10884cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10894cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10904cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10914cb5300bSChris Mason *off = min_key.offset; 10924cb5300bSChris Mason btrfs_free_path(path); 10934cb5300bSChris Mason return 0; 10944cb5300bSChris Mason } 10954cb5300bSChris Mason 1096f094c9bdSFilipe Manana path->slots[0]++; 1097f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1098f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1099f094c9bdSFilipe Manana goto process_slot; 1100f094c9bdSFilipe Manana } 1101f094c9bdSFilipe Manana 11024cb5300bSChris Mason if (min_key.offset == (u64)-1) 11034cb5300bSChris Mason goto none; 11044cb5300bSChris Mason 11054cb5300bSChris Mason min_key.offset++; 11064cb5300bSChris Mason btrfs_release_path(path); 11074cb5300bSChris Mason } 11084cb5300bSChris Mason none: 11094cb5300bSChris Mason btrfs_free_path(path); 11104cb5300bSChris Mason return -ENOENT; 11114cb5300bSChris Mason } 11124cb5300bSChris Mason 11136c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 111417ce6ef8SLiu Bo { 111517ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1116940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 11176c282eb4SLi Zefan struct extent_map *em; 111809cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1119940100a4SChris Mason 1120940100a4SChris Mason /* 1121940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1122940100a4SChris Mason * the full extent lock 1123940100a4SChris Mason */ 1124940100a4SChris Mason read_lock(&em_tree->lock); 1125940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1126940100a4SChris Mason read_unlock(&em_tree->lock); 1127940100a4SChris Mason 1128940100a4SChris Mason if (!em) { 1129308d9800SFilipe Manana struct extent_state *cached = NULL; 1130308d9800SFilipe Manana u64 end = start + len - 1; 1131308d9800SFilipe Manana 1132940100a4SChris Mason /* get the big lock and read metadata off disk */ 1133ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1134fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1135e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1136940100a4SChris Mason 11376cf8bfbfSDan Carpenter if (IS_ERR(em)) 11386c282eb4SLi Zefan return NULL; 1139940100a4SChris Mason } 1140940100a4SChris Mason 11416c282eb4SLi Zefan return em; 11426c282eb4SLi Zefan } 11436c282eb4SLi Zefan 11446c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11456c282eb4SLi Zefan { 11466c282eb4SLi Zefan struct extent_map *next; 11476c282eb4SLi Zefan bool ret = true; 11486c282eb4SLi Zefan 11496c282eb4SLi Zefan /* this is the last extent */ 11506c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11516c282eb4SLi Zefan return false; 11526c282eb4SLi Zefan 11536c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1154e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1155e9512d72SChris Mason ret = false; 1156e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1157ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11586c282eb4SLi Zefan ret = false; 11596c282eb4SLi Zefan 11606c282eb4SLi Zefan free_extent_map(next); 11616c282eb4SLi Zefan return ret; 11626c282eb4SLi Zefan } 11636c282eb4SLi Zefan 1164aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1165a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1166a43a2111SAndrew Mahone int compress) 11676c282eb4SLi Zefan { 11686c282eb4SLi Zefan struct extent_map *em; 11696c282eb4SLi Zefan int ret = 1; 11706c282eb4SLi Zefan bool next_mergeable = true; 11714a3560c4SLiu Bo bool prev_mergeable = true; 11726c282eb4SLi Zefan 11736c282eb4SLi Zefan /* 11746c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11756c282eb4SLi Zefan * defragging it 11766c282eb4SLi Zefan */ 11776c282eb4SLi Zefan if (start < *defrag_end) 11786c282eb4SLi Zefan return 1; 11796c282eb4SLi Zefan 11806c282eb4SLi Zefan *skip = 0; 11816c282eb4SLi Zefan 11826c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11836c282eb4SLi Zefan if (!em) 11846c282eb4SLi Zefan return 0; 11856c282eb4SLi Zefan 1186940100a4SChris Mason /* this will cover holes, and inline extents */ 118717ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1188940100a4SChris Mason ret = 0; 118917ce6ef8SLiu Bo goto out; 119017ce6ef8SLiu Bo } 119117ce6ef8SLiu Bo 11924a3560c4SLiu Bo if (!*defrag_end) 11934a3560c4SLiu Bo prev_mergeable = false; 11944a3560c4SLiu Bo 11956c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1196940100a4SChris Mason /* 11976c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11986c282eb4SLi Zefan * real extent, don't bother defragging it 1199940100a4SChris Mason */ 1200a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 12014a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1202940100a4SChris Mason ret = 0; 120317ce6ef8SLiu Bo out: 1204940100a4SChris Mason /* 1205940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1206940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1207940100a4SChris Mason * so that the next tiny extent will force a defrag. 1208940100a4SChris Mason * 1209940100a4SChris Mason * The end result of this is that tiny extents before a single big 1210940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1211940100a4SChris Mason */ 1212940100a4SChris Mason if (ret) { 1213940100a4SChris Mason *defrag_end = extent_map_end(em); 1214940100a4SChris Mason } else { 1215940100a4SChris Mason *last_len = 0; 1216940100a4SChris Mason *skip = extent_map_end(em); 1217940100a4SChris Mason *defrag_end = 0; 1218940100a4SChris Mason } 1219940100a4SChris Mason 1220940100a4SChris Mason free_extent_map(em); 1221940100a4SChris Mason return ret; 1222940100a4SChris Mason } 1223940100a4SChris Mason 12244cb5300bSChris Mason /* 12254cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12264cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12274cb5300bSChris Mason * to COW and defrag. 12284cb5300bSChris Mason * 12294cb5300bSChris Mason * It also makes sure the delalloc code has enough 12304cb5300bSChris Mason * dirty data to avoid making new small extents as part 12314cb5300bSChris Mason * of the defrag 12324cb5300bSChris Mason * 12334cb5300bSChris Mason * It's a good idea to start RA on this range 12344cb5300bSChris Mason * before calling this. 12354cb5300bSChris Mason */ 12364cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12374cb5300bSChris Mason struct page **pages, 12384cb5300bSChris Mason unsigned long start_index, 1239c41570c9SJustin Maggard unsigned long num_pages) 1240f46b5a66SChristoph Hellwig { 12414cb5300bSChris Mason unsigned long file_end; 12424cb5300bSChris Mason u64 isize = i_size_read(inode); 1243f46b5a66SChristoph Hellwig u64 page_start; 1244f46b5a66SChristoph Hellwig u64 page_end; 12451f12bd06SLiu Bo u64 page_cnt; 12464cb5300bSChris Mason int ret; 12474cb5300bSChris Mason int i; 12484cb5300bSChris Mason int i_done; 12494cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12504cb5300bSChris Mason struct extent_state *cached_state = NULL; 1251600a45e1SMiao Xie struct extent_io_tree *tree; 1252364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12533b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12544cb5300bSChris Mason 125509cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12561f12bd06SLiu Bo if (!isize || start_index > file_end) 12571f12bd06SLiu Bo return 0; 12581f12bd06SLiu Bo 12591f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12604cb5300bSChris Mason 1261364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 126209cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 126309cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12644cb5300bSChris Mason if (ret) 12654cb5300bSChris Mason return ret; 12664cb5300bSChris Mason i_done = 0; 1267600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12684cb5300bSChris Mason 12694cb5300bSChris Mason /* step one, lock all the pages */ 12701f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12714cb5300bSChris Mason struct page *page; 1272600a45e1SMiao Xie again: 1273a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12743b16a4e3SJosef Bacik start_index + i, mask); 12754cb5300bSChris Mason if (!page) 12764cb5300bSChris Mason break; 12774cb5300bSChris Mason 1278600a45e1SMiao Xie page_start = page_offset(page); 127909cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1280600a45e1SMiao Xie while (1) { 1281308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1282ff13db41SDavid Sterba &cached_state); 1283600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1284600a45e1SMiao Xie page_start); 1285308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1286e43bbe5eSDavid Sterba &cached_state); 1287600a45e1SMiao Xie if (!ordered) 1288600a45e1SMiao Xie break; 1289600a45e1SMiao Xie 1290600a45e1SMiao Xie unlock_page(page); 1291600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1292600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1293600a45e1SMiao Xie lock_page(page); 12941f12bd06SLiu Bo /* 12951f12bd06SLiu Bo * we unlocked the page above, so we need check if 12961f12bd06SLiu Bo * it was released or not. 12971f12bd06SLiu Bo */ 12981f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12991f12bd06SLiu Bo unlock_page(page); 130009cbfeafSKirill A. Shutemov put_page(page); 13011f12bd06SLiu Bo goto again; 13021f12bd06SLiu Bo } 1303600a45e1SMiao Xie } 1304600a45e1SMiao Xie 13054cb5300bSChris Mason if (!PageUptodate(page)) { 13064cb5300bSChris Mason btrfs_readpage(NULL, page); 13074cb5300bSChris Mason lock_page(page); 13084cb5300bSChris Mason if (!PageUptodate(page)) { 13094cb5300bSChris Mason unlock_page(page); 131009cbfeafSKirill A. Shutemov put_page(page); 13114cb5300bSChris Mason ret = -EIO; 13124cb5300bSChris Mason break; 13134cb5300bSChris Mason } 13144cb5300bSChris Mason } 1315600a45e1SMiao Xie 1316600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1317600a45e1SMiao Xie unlock_page(page); 131809cbfeafSKirill A. Shutemov put_page(page); 1319600a45e1SMiao Xie goto again; 1320600a45e1SMiao Xie } 1321600a45e1SMiao Xie 13224cb5300bSChris Mason pages[i] = page; 13234cb5300bSChris Mason i_done++; 13244cb5300bSChris Mason } 13254cb5300bSChris Mason if (!i_done || ret) 13264cb5300bSChris Mason goto out; 13274cb5300bSChris Mason 13281751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13294cb5300bSChris Mason goto out; 13304cb5300bSChris Mason 13314cb5300bSChris Mason /* 13324cb5300bSChris Mason * so now we have a nice long stream of locked 13334cb5300bSChris Mason * and up to date pages, lets wait on them 13344cb5300bSChris Mason */ 13354cb5300bSChris Mason for (i = 0; i < i_done; i++) 13364cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13374cb5300bSChris Mason 13384cb5300bSChris Mason page_start = page_offset(pages[0]); 133909cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13404cb5300bSChris Mason 13414cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1342ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13434cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 13444cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 13459e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1346ae0f1625SDavid Sterba &cached_state); 13474cb5300bSChris Mason 13481f12bd06SLiu Bo if (i_done != page_cnt) { 13499e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 135028c4a3e2SSu Yue btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); 13519e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1352bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 135309cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 135443b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13554cb5300bSChris Mason } 13564cb5300bSChris Mason 13574cb5300bSChris Mason 13589e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1359018ed4f7SDavid Sterba &cached_state); 13604cb5300bSChris Mason 13614cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1362e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13634cb5300bSChris Mason 13644cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13654cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13664cb5300bSChris Mason ClearPageChecked(pages[i]); 13674cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13684cb5300bSChris Mason set_page_dirty(pages[i]); 13694cb5300bSChris Mason unlock_page(pages[i]); 137009cbfeafSKirill A. Shutemov put_page(pages[i]); 13714cb5300bSChris Mason } 137243b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 137343b18595SQu Wenruo false); 1374364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13754cb5300bSChris Mason return i_done; 13764cb5300bSChris Mason out: 13774cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13784cb5300bSChris Mason unlock_page(pages[i]); 137909cbfeafSKirill A. Shutemov put_page(pages[i]); 13804cb5300bSChris Mason } 1381bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 138209cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 138343b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 138443b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 138543b18595SQu Wenruo true); 1386364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13874cb5300bSChris Mason return ret; 13884cb5300bSChris Mason 13894cb5300bSChris Mason } 13904cb5300bSChris Mason 13914cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13924cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13934cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13944cb5300bSChris Mason { 13950b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13964cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13974cb5300bSChris Mason struct file_ra_state *ra = NULL; 13984cb5300bSChris Mason unsigned long last_index; 1399151a31b2SLi Zefan u64 isize = i_size_read(inode); 1400940100a4SChris Mason u64 last_len = 0; 1401940100a4SChris Mason u64 skip = 0; 1402940100a4SChris Mason u64 defrag_end = 0; 14034cb5300bSChris Mason u64 newer_off = range->start; 1404f46b5a66SChristoph Hellwig unsigned long i; 1405008873eaSLi Zefan unsigned long ra_index = 0; 1406f46b5a66SChristoph Hellwig int ret; 14074cb5300bSChris Mason int defrag_count = 0; 14081a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1409aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 141009cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1411c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1412ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 14134cb5300bSChris Mason struct page **pages = NULL; 14141e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 14154cb5300bSChris Mason 14160abd5b17SLiu Bo if (isize == 0) 14170abd5b17SLiu Bo return 0; 14180abd5b17SLiu Bo 14190abd5b17SLiu Bo if (range->start >= isize) 14200abd5b17SLiu Bo return -EINVAL; 14211a419d85SLi Zefan 14221e2ef46dSDavid Sterba if (do_compress) { 14231a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 14241a419d85SLi Zefan return -EINVAL; 14251a419d85SLi Zefan if (range->compress_type) 14261a419d85SLi Zefan compress_type = range->compress_type; 14271a419d85SLi Zefan } 1428f46b5a66SChristoph Hellwig 14290abd5b17SLiu Bo if (extent_thresh == 0) 1430ee22184bSByongho Lee extent_thresh = SZ_256K; 1431f46b5a66SChristoph Hellwig 14324cb5300bSChris Mason /* 14330a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14340a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14350a52d108SDavid Sterba * we don't error out. 14364cb5300bSChris Mason */ 14374cb5300bSChris Mason if (!file) { 143863e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14390a52d108SDavid Sterba if (ra) 14404cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14414cb5300bSChris Mason } else { 14424cb5300bSChris Mason ra = &file->f_ra; 14434cb5300bSChris Mason } 14444cb5300bSChris Mason 144563e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14464cb5300bSChris Mason if (!pages) { 14474cb5300bSChris Mason ret = -ENOMEM; 14484cb5300bSChris Mason goto out_ra; 14494cb5300bSChris Mason } 14504cb5300bSChris Mason 14514cb5300bSChris Mason /* find the last page to defrag */ 14521e701a32SChris Mason if (range->start + range->len > range->start) { 1453151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 145409cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14551e701a32SChris Mason } else { 145609cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14571e701a32SChris Mason } 14581e701a32SChris Mason 14594cb5300bSChris Mason if (newer_than) { 14604cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1461ee22184bSByongho Lee &newer_off, SZ_64K); 14624cb5300bSChris Mason if (!ret) { 14634cb5300bSChris Mason range->start = newer_off; 14644cb5300bSChris Mason /* 14654cb5300bSChris Mason * we always align our defrag to help keep 14664cb5300bSChris Mason * the extents in the file evenly spaced 14674cb5300bSChris Mason */ 146809cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14694cb5300bSChris Mason } else 14704cb5300bSChris Mason goto out_ra; 14714cb5300bSChris Mason } else { 147209cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14734cb5300bSChris Mason } 14744cb5300bSChris Mason if (!max_to_defrag) 1475070034bdSchandan max_to_defrag = last_index - i + 1; 14764cb5300bSChris Mason 14772a0f7f57SLi Zefan /* 14782a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14792a0f7f57SLi Zefan * written sequentially. 14802a0f7f57SLi Zefan */ 14812a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14822a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14832a0f7f57SLi Zefan 1484f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 148509cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14864cb5300bSChris Mason /* 14874cb5300bSChris Mason * make sure we stop running if someone unmounts 14884cb5300bSChris Mason * the FS 14894cb5300bSChris Mason */ 14901751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14914cb5300bSChris Mason break; 14924cb5300bSChris Mason 14930b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14940b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1495210549ebSDavid Sterba ret = -EAGAIN; 1496210549ebSDavid Sterba break; 1497210549ebSDavid Sterba } 1498210549ebSDavid Sterba 149909cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 15006c282eb4SLi Zefan extent_thresh, &last_len, &skip, 15011e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1502940100a4SChris Mason unsigned long next; 1503940100a4SChris Mason /* 1504940100a4SChris Mason * the should_defrag function tells us how much to skip 1505940100a4SChris Mason * bump our counter by the suggested amount 1506940100a4SChris Mason */ 150709cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1508940100a4SChris Mason i = max(i + 1, next); 1509940100a4SChris Mason continue; 1510940100a4SChris Mason } 1511008873eaSLi Zefan 1512008873eaSLi Zefan if (!newer_than) { 151309cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 151409cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1515008873eaSLi Zefan cluster = min(cluster, max_cluster); 1516008873eaSLi Zefan } else { 1517008873eaSLi Zefan cluster = max_cluster; 1518008873eaSLi Zefan } 1519008873eaSLi Zefan 1520008873eaSLi Zefan if (i + cluster > ra_index) { 1521008873eaSLi Zefan ra_index = max(i, ra_index); 15220a52d108SDavid Sterba if (ra) 1523d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1524d3c0bab5SDavid Sterba file, ra_index, cluster); 1525e4826a5bSchandan ra_index += cluster; 1526008873eaSLi Zefan } 15274cb5300bSChris Mason 15285955102cSAl Viro inode_lock(inode); 1529eede2bf3SOmar Sandoval if (IS_SWAPFILE(inode)) { 1530eede2bf3SOmar Sandoval ret = -ETXTBSY; 1531eede2bf3SOmar Sandoval } else { 15321e2ef46dSDavid Sterba if (do_compress) 1533eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1534008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1535eede2bf3SOmar Sandoval } 1536ecb8bea8SLiu Bo if (ret < 0) { 15375955102cSAl Viro inode_unlock(inode); 15384cb5300bSChris Mason goto out_ra; 1539ecb8bea8SLiu Bo } 15404cb5300bSChris Mason 15414cb5300bSChris Mason defrag_count += ret; 1542d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15435955102cSAl Viro inode_unlock(inode); 15444cb5300bSChris Mason 15454cb5300bSChris Mason if (newer_than) { 15464cb5300bSChris Mason if (newer_off == (u64)-1) 15474cb5300bSChris Mason break; 15484cb5300bSChris Mason 1549e1f041e1SLiu Bo if (ret > 0) 1550e1f041e1SLiu Bo i += ret; 1551e1f041e1SLiu Bo 15524cb5300bSChris Mason newer_off = max(newer_off + 1, 155309cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15544cb5300bSChris Mason 1555ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1556ee22184bSByongho Lee &newer_off, SZ_64K); 15574cb5300bSChris Mason if (!ret) { 15584cb5300bSChris Mason range->start = newer_off; 155909cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15604cb5300bSChris Mason } else { 15614cb5300bSChris Mason break; 1562940100a4SChris Mason } 15634cb5300bSChris Mason } else { 1564008873eaSLi Zefan if (ret > 0) { 1565cbcc8326SLi Zefan i += ret; 156609cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1567008873eaSLi Zefan } else { 1568940100a4SChris Mason i++; 1569008873eaSLi Zefan last_len = 0; 1570008873eaSLi Zefan } 1571f46b5a66SChristoph Hellwig } 15724cb5300bSChris Mason } 1573f46b5a66SChristoph Hellwig 1574dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15751e701a32SChris Mason filemap_flush(inode->i_mapping); 1576dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1577dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1578dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1579dec8ef90SFilipe Manana } 15801e701a32SChris Mason 15811a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15820b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15835c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15845c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15851a419d85SLi Zefan } 15861a419d85SLi Zefan 158760ccf82fSDiego Calleja ret = defrag_count; 1588940100a4SChris Mason 15894cb5300bSChris Mason out_ra: 15901e2ef46dSDavid Sterba if (do_compress) { 15915955102cSAl Viro inode_lock(inode); 1592eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15935955102cSAl Viro inode_unlock(inode); 1594633085c7SFilipe David Borba Manana } 15954cb5300bSChris Mason if (!file) 15964cb5300bSChris Mason kfree(ra); 15974cb5300bSChris Mason kfree(pages); 1598940100a4SChris Mason return ret; 1599f46b5a66SChristoph Hellwig } 1600f46b5a66SChristoph Hellwig 1601198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 160276dda93cSYan, Zheng void __user *arg) 1603f46b5a66SChristoph Hellwig { 16040b246afaSJeff Mahoney struct inode *inode = file_inode(file); 16050b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1606f46b5a66SChristoph Hellwig u64 new_size; 1607f46b5a66SChristoph Hellwig u64 old_size; 1608f46b5a66SChristoph Hellwig u64 devid = 1; 16090b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1610f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1611f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1612f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1613f46b5a66SChristoph Hellwig char *sizestr; 16149a40f122SGui Hecheng char *retptr; 1615f46b5a66SChristoph Hellwig char *devstr = NULL; 1616f46b5a66SChristoph Hellwig int ret = 0; 1617f46b5a66SChristoph Hellwig int mod = 0; 1618f46b5a66SChristoph Hellwig 1619e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1620e441d54dSChris Mason return -EPERM; 1621e441d54dSChris Mason 1622198605a8SMiao Xie ret = mnt_want_write_file(file); 1623198605a8SMiao Xie if (ret) 1624198605a8SMiao Xie return ret; 1625198605a8SMiao Xie 1626171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 162797547676SMiao Xie mnt_drop_write_file(file); 1628e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1629c9e9f97bSIlya Dryomov } 1630c9e9f97bSIlya Dryomov 1631dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1632c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1633c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1634c9e9f97bSIlya Dryomov goto out; 1635c9e9f97bSIlya Dryomov } 16365516e595SMark Fasheh 16375516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1638f46b5a66SChristoph Hellwig 1639f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1640f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1641f46b5a66SChristoph Hellwig if (devstr) { 1642f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1643f46b5a66SChristoph Hellwig *devstr = '\0'; 1644f46b5a66SChristoph Hellwig devstr = vol_args->name; 164558dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 164658dfae63SZhangZhen if (ret) 164758dfae63SZhangZhen goto out_free; 1648dfd79829SMiao Xie if (!devid) { 1649dfd79829SMiao Xie ret = -EINVAL; 1650dfd79829SMiao Xie goto out_free; 1651dfd79829SMiao Xie } 16520b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1653f46b5a66SChristoph Hellwig } 1654dba60f3fSMiao Xie 165509ba3bc9SAnand Jain device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true); 1656f46b5a66SChristoph Hellwig if (!device) { 16570b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1658c1c9ff7cSGeert Uytterhoeven devid); 1659dfd79829SMiao Xie ret = -ENODEV; 1660c9e9f97bSIlya Dryomov goto out_free; 1661f46b5a66SChristoph Hellwig } 1662dba60f3fSMiao Xie 1663ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16640b246afaSJeff Mahoney btrfs_info(fs_info, 1665efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1666c1c9ff7cSGeert Uytterhoeven devid); 1667dfd79829SMiao Xie ret = -EPERM; 16684e42ae1bSLiu Bo goto out_free; 16694e42ae1bSLiu Bo } 16704e42ae1bSLiu Bo 1671f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1672f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1673f46b5a66SChristoph Hellwig else { 1674f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1675f46b5a66SChristoph Hellwig mod = -1; 1676f46b5a66SChristoph Hellwig sizestr++; 1677f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1678f46b5a66SChristoph Hellwig mod = 1; 1679f46b5a66SChristoph Hellwig sizestr++; 1680f46b5a66SChristoph Hellwig } 16819a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16829a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1683f46b5a66SChristoph Hellwig ret = -EINVAL; 1684c9e9f97bSIlya Dryomov goto out_free; 1685f46b5a66SChristoph Hellwig } 1686f46b5a66SChristoph Hellwig } 1687f46b5a66SChristoph Hellwig 1688401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1689dfd79829SMiao Xie ret = -EPERM; 169063a212abSStefan Behrens goto out_free; 169163a212abSStefan Behrens } 169263a212abSStefan Behrens 16937cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1694f46b5a66SChristoph Hellwig 1695f46b5a66SChristoph Hellwig if (mod < 0) { 1696f46b5a66SChristoph Hellwig if (new_size > old_size) { 1697f46b5a66SChristoph Hellwig ret = -EINVAL; 1698c9e9f97bSIlya Dryomov goto out_free; 1699f46b5a66SChristoph Hellwig } 1700f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1701f46b5a66SChristoph Hellwig } else if (mod > 0) { 1702eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1703902c68a4SGui Hecheng ret = -ERANGE; 1704eb8052e0SWenliang Fan goto out_free; 1705eb8052e0SWenliang Fan } 1706f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1707f46b5a66SChristoph Hellwig } 1708f46b5a66SChristoph Hellwig 1709ee22184bSByongho Lee if (new_size < SZ_256M) { 1710f46b5a66SChristoph Hellwig ret = -EINVAL; 1711c9e9f97bSIlya Dryomov goto out_free; 1712f46b5a66SChristoph Hellwig } 1713f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1714f46b5a66SChristoph Hellwig ret = -EFBIG; 1715c9e9f97bSIlya Dryomov goto out_free; 1716f46b5a66SChristoph Hellwig } 1717f46b5a66SChristoph Hellwig 171847f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1719f46b5a66SChristoph Hellwig 17200b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1721c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1722f46b5a66SChristoph Hellwig 1723f46b5a66SChristoph Hellwig if (new_size > old_size) { 1724a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 172598d5dc13STsutomu Itoh if (IS_ERR(trans)) { 172698d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1727c9e9f97bSIlya Dryomov goto out_free; 172898d5dc13STsutomu Itoh } 1729f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17303a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1731ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1732f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17330253f40eSjeff.liu } /* equal, nothing need to do */ 1734f46b5a66SChristoph Hellwig 1735c9e9f97bSIlya Dryomov out_free: 1736f46b5a66SChristoph Hellwig kfree(vol_args); 1737c9e9f97bSIlya Dryomov out: 1738171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 173918f39c41SIlya Dryomov mnt_drop_write_file(file); 1740f46b5a66SChristoph Hellwig return ret; 1741f46b5a66SChristoph Hellwig } 1742f46b5a66SChristoph Hellwig 174372fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 174452f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17456f72c7e2SArne Jansen u64 *transid, bool readonly, 17468696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1747f46b5a66SChristoph Hellwig { 1748f46b5a66SChristoph Hellwig int namelen; 17493de4586cSChris Mason int ret = 0; 1750f46b5a66SChristoph Hellwig 1751325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1752325c50e3SJeff Mahoney return -ENOTDIR; 1753325c50e3SJeff Mahoney 1754a874a63eSLiu Bo ret = mnt_want_write_file(file); 1755a874a63eSLiu Bo if (ret) 1756a874a63eSLiu Bo goto out; 1757a874a63eSLiu Bo 175872fd032eSSage Weil namelen = strlen(name); 175972fd032eSSage Weil if (strchr(name, '/')) { 1760f46b5a66SChristoph Hellwig ret = -EINVAL; 1761a874a63eSLiu Bo goto out_drop_write; 1762f46b5a66SChristoph Hellwig } 1763f46b5a66SChristoph Hellwig 176416780cabSChris Mason if (name[0] == '.' && 176516780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 176616780cabSChris Mason ret = -EEXIST; 1767a874a63eSLiu Bo goto out_drop_write; 176816780cabSChris Mason } 176916780cabSChris Mason 17703de4586cSChris Mason if (subvol) { 177172fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17726f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1773cb8e7090SChristoph Hellwig } else { 17742903ff01SAl Viro struct fd src = fdget(fd); 17753de4586cSChris Mason struct inode *src_inode; 17762903ff01SAl Viro if (!src.file) { 17773de4586cSChris Mason ret = -EINVAL; 1778a874a63eSLiu Bo goto out_drop_write; 17793de4586cSChris Mason } 17803de4586cSChris Mason 1781496ad9aaSAl Viro src_inode = file_inode(src.file); 1782496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1783c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1784efe120a0SFrank Holton "Snapshot src from another FS"); 178523ad5b17SKusanagi Kouichi ret = -EXDEV; 1786d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1787d0242061SDavid Sterba /* 1788d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1789d0242061SDavid Sterba * are limited to own subvolumes only 1790d0242061SDavid Sterba */ 1791d0242061SDavid Sterba ret = -EPERM; 1792ecd18815SAl Viro } else { 179372fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 179472fd032eSSage Weil BTRFS_I(src_inode)->root, 17956f72c7e2SArne Jansen transid, readonly, inherit); 1796ecd18815SAl Viro } 17972903ff01SAl Viro fdput(src); 1798cb8e7090SChristoph Hellwig } 1799a874a63eSLiu Bo out_drop_write: 1800a874a63eSLiu Bo mnt_drop_write_file(file); 1801f46b5a66SChristoph Hellwig out: 180272fd032eSSage Weil return ret; 180372fd032eSSage Weil } 180472fd032eSSage Weil 180572fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1806fa0d2b9bSLi Zefan void __user *arg, int subvol) 180772fd032eSSage Weil { 1808fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 180972fd032eSSage Weil int ret; 181072fd032eSSage Weil 1811325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1812325c50e3SJeff Mahoney return -ENOTDIR; 1813325c50e3SJeff Mahoney 1814fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1815fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1816fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1817fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1818fa0d2b9bSLi Zefan 1819fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1820b83cc969SLi Zefan vol_args->fd, subvol, 18216f72c7e2SArne Jansen NULL, false, NULL); 1822fa0d2b9bSLi Zefan 1823fa0d2b9bSLi Zefan kfree(vol_args); 1824fa0d2b9bSLi Zefan return ret; 1825fa0d2b9bSLi Zefan } 1826fa0d2b9bSLi Zefan 1827fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1828fa0d2b9bSLi Zefan void __user *arg, int subvol) 1829fa0d2b9bSLi Zefan { 1830fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1831fa0d2b9bSLi Zefan int ret; 1832fdfb1e4fSLi Zefan u64 transid = 0; 1833fdfb1e4fSLi Zefan u64 *ptr = NULL; 1834b83cc969SLi Zefan bool readonly = false; 18356f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 183672fd032eSSage Weil 1837325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1838325c50e3SJeff Mahoney return -ENOTDIR; 1839325c50e3SJeff Mahoney 1840fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1841fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1842fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1843fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1844fdfb1e4fSLi Zefan 1845b83cc969SLi Zefan if (vol_args->flags & 18466f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18476f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1848b83cc969SLi Zefan ret = -EOPNOTSUPP; 1849c47ca32dSDan Carpenter goto free_args; 1850fdfb1e4fSLi Zefan } 1851fdfb1e4fSLi Zefan 1852fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1853fdfb1e4fSLi Zefan ptr = &transid; 1854b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1855b83cc969SLi Zefan readonly = true; 18566f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 185709cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18586f72c7e2SArne Jansen ret = -EINVAL; 1859c47ca32dSDan Carpenter goto free_args; 18606f72c7e2SArne Jansen } 18616f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18626f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18636f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1864c47ca32dSDan Carpenter goto free_args; 18656f72c7e2SArne Jansen } 18666f72c7e2SArne Jansen } 186775eaa0e2SSage Weil 1868fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18696f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18708696c533SMiao Xie readonly, inherit); 1871c47ca32dSDan Carpenter if (ret) 1872c47ca32dSDan Carpenter goto free_inherit; 187375eaa0e2SSage Weil 1874c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1875fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1876c47ca32dSDan Carpenter transid), 1877c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 187875eaa0e2SSage Weil ret = -EFAULT; 1879c47ca32dSDan Carpenter 1880c47ca32dSDan Carpenter free_inherit: 18816f72c7e2SArne Jansen kfree(inherit); 1882c47ca32dSDan Carpenter free_args: 1883c47ca32dSDan Carpenter kfree(vol_args); 1884f46b5a66SChristoph Hellwig return ret; 1885f46b5a66SChristoph Hellwig } 1886f46b5a66SChristoph Hellwig 18870caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18880caa102dSLi Zefan void __user *arg) 18890caa102dSLi Zefan { 1890496ad9aaSAl Viro struct inode *inode = file_inode(file); 18910b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18920caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18930caa102dSLi Zefan int ret = 0; 18940caa102dSLi Zefan u64 flags = 0; 18950caa102dSLi Zefan 18964a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18970caa102dSLi Zefan return -EINVAL; 18980caa102dSLi Zefan 18990b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 19000caa102dSLi Zefan if (btrfs_root_readonly(root)) 19010caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 19020b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 19030caa102dSLi Zefan 19040caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 19050caa102dSLi Zefan ret = -EFAULT; 19060caa102dSLi Zefan 19070caa102dSLi Zefan return ret; 19080caa102dSLi Zefan } 19090caa102dSLi Zefan 19100caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 19110caa102dSLi Zefan void __user *arg) 19120caa102dSLi Zefan { 1913496ad9aaSAl Viro struct inode *inode = file_inode(file); 19140b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 19150caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 19160caa102dSLi Zefan struct btrfs_trans_handle *trans; 19170caa102dSLi Zefan u64 root_flags; 19180caa102dSLi Zefan u64 flags; 19190caa102dSLi Zefan int ret = 0; 19200caa102dSLi Zefan 1921bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1922bd60ea0fSDavid Sterba return -EPERM; 1923bd60ea0fSDavid Sterba 1924b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1925b9ca0664SLiu Bo if (ret) 1926b9ca0664SLiu Bo goto out; 19270caa102dSLi Zefan 19284a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1929b9ca0664SLiu Bo ret = -EINVAL; 1930b9ca0664SLiu Bo goto out_drop_write; 1931b9ca0664SLiu Bo } 19320caa102dSLi Zefan 1933b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1934b9ca0664SLiu Bo ret = -EFAULT; 1935b9ca0664SLiu Bo goto out_drop_write; 1936b9ca0664SLiu Bo } 19370caa102dSLi Zefan 1938b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1939b9ca0664SLiu Bo ret = -EINVAL; 1940b9ca0664SLiu Bo goto out_drop_write; 1941b9ca0664SLiu Bo } 19420caa102dSLi Zefan 1943b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1944b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1945b9ca0664SLiu Bo goto out_drop_write; 1946b9ca0664SLiu Bo } 19470caa102dSLi Zefan 19480b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19490caa102dSLi Zefan 19500caa102dSLi Zefan /* nothing to do */ 19510caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1952b9ca0664SLiu Bo goto out_drop_sem; 19530caa102dSLi Zefan 19540caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19552c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19560caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19570caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19582c686537SDavid Sterba } else { 19592c686537SDavid Sterba /* 19602c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19612c686537SDavid Sterba * send 19622c686537SDavid Sterba */ 19632c686537SDavid Sterba spin_lock(&root->root_item_lock); 19642c686537SDavid Sterba if (root->send_in_progress == 0) { 19650caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19660caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19672c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19682c686537SDavid Sterba } else { 19692c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19700b246afaSJeff Mahoney btrfs_warn(fs_info, 19712c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19722c686537SDavid Sterba root->root_key.objectid); 19732c686537SDavid Sterba ret = -EPERM; 19742c686537SDavid Sterba goto out_drop_sem; 19752c686537SDavid Sterba } 19762c686537SDavid Sterba } 19770caa102dSLi Zefan 19780caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19790caa102dSLi Zefan if (IS_ERR(trans)) { 19800caa102dSLi Zefan ret = PTR_ERR(trans); 19810caa102dSLi Zefan goto out_reset; 19820caa102dSLi Zefan } 19830caa102dSLi Zefan 19840b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19850caa102dSLi Zefan &root->root_key, &root->root_item); 19869417ebc8SNikolay Borisov if (ret < 0) { 19879417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19889417ebc8SNikolay Borisov goto out_reset; 19899417ebc8SNikolay Borisov } 19900caa102dSLi Zefan 19919417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19929417ebc8SNikolay Borisov 19930caa102dSLi Zefan out_reset: 19940caa102dSLi Zefan if (ret) 19950caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1996b9ca0664SLiu Bo out_drop_sem: 19970b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1998b9ca0664SLiu Bo out_drop_write: 1999b9ca0664SLiu Bo mnt_drop_write_file(file); 2000b9ca0664SLiu Bo out: 20010caa102dSLi Zefan return ret; 20020caa102dSLi Zefan } 20030caa102dSLi Zefan 2004ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 2005ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 2006ac8e9819SChris Mason { 2007abc6e134SChris Mason struct btrfs_key test; 2008abc6e134SChris Mason int ret; 2009abc6e134SChris Mason 2010abc6e134SChris Mason test.objectid = sk->min_objectid; 2011abc6e134SChris Mason test.type = sk->min_type; 2012abc6e134SChris Mason test.offset = sk->min_offset; 2013abc6e134SChris Mason 2014abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2015abc6e134SChris Mason if (ret < 0) 2016ac8e9819SChris Mason return 0; 2017abc6e134SChris Mason 2018abc6e134SChris Mason test.objectid = sk->max_objectid; 2019abc6e134SChris Mason test.type = sk->max_type; 2020abc6e134SChris Mason test.offset = sk->max_offset; 2021abc6e134SChris Mason 2022abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 2023abc6e134SChris Mason if (ret > 0) 2024ac8e9819SChris Mason return 0; 2025ac8e9819SChris Mason return 1; 2026ac8e9819SChris Mason } 2027ac8e9819SChris Mason 2028df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2029ac8e9819SChris Mason struct btrfs_key *key, 2030ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20319b6e817dSGerhard Heift size_t *buf_size, 2032ba346b35SGerhard Heift char __user *ubuf, 2033ac8e9819SChris Mason unsigned long *sk_offset, 2034ac8e9819SChris Mason int *num_found) 2035ac8e9819SChris Mason { 2036ac8e9819SChris Mason u64 found_transid; 2037ac8e9819SChris Mason struct extent_buffer *leaf; 2038ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2039dd81d459SNaohiro Aota struct btrfs_key test; 2040ac8e9819SChris Mason unsigned long item_off; 2041ac8e9819SChris Mason unsigned long item_len; 2042ac8e9819SChris Mason int nritems; 2043ac8e9819SChris Mason int i; 2044ac8e9819SChris Mason int slot; 2045ac8e9819SChris Mason int ret = 0; 2046ac8e9819SChris Mason 2047ac8e9819SChris Mason leaf = path->nodes[0]; 2048ac8e9819SChris Mason slot = path->slots[0]; 2049ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2050ac8e9819SChris Mason 2051ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2052ac8e9819SChris Mason i = nritems; 2053ac8e9819SChris Mason goto advance_key; 2054ac8e9819SChris Mason } 2055ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2056ac8e9819SChris Mason 2057ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2058ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2059ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2060ac8e9819SChris Mason 206103b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 206203b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 206303b71c6cSGabriel de Perthuis continue; 206403b71c6cSGabriel de Perthuis 20659b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20668f5f6178SGerhard Heift if (*num_found) { 2067ac8e9819SChris Mason ret = 1; 20688f5f6178SGerhard Heift goto out; 20698f5f6178SGerhard Heift } 20708f5f6178SGerhard Heift 20718f5f6178SGerhard Heift /* 20728f5f6178SGerhard Heift * return one empty item back for v1, which does not 20738f5f6178SGerhard Heift * handle -EOVERFLOW 20748f5f6178SGerhard Heift */ 20758f5f6178SGerhard Heift 20769b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2077ac8e9819SChris Mason item_len = 0; 20788f5f6178SGerhard Heift ret = -EOVERFLOW; 20798f5f6178SGerhard Heift } 2080ac8e9819SChris Mason 20819b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2082ac8e9819SChris Mason ret = 1; 208325c9bc2eSGerhard Heift goto out; 2084ac8e9819SChris Mason } 2085ac8e9819SChris Mason 2086ac8e9819SChris Mason sh.objectid = key->objectid; 2087ac8e9819SChris Mason sh.offset = key->offset; 2088ac8e9819SChris Mason sh.type = key->type; 2089ac8e9819SChris Mason sh.len = item_len; 2090ac8e9819SChris Mason sh.transid = found_transid; 2091ac8e9819SChris Mason 2092ac8e9819SChris Mason /* copy search result header */ 2093ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2094ba346b35SGerhard Heift ret = -EFAULT; 2095ba346b35SGerhard Heift goto out; 2096ba346b35SGerhard Heift } 2097ba346b35SGerhard Heift 2098ac8e9819SChris Mason *sk_offset += sizeof(sh); 2099ac8e9819SChris Mason 2100ac8e9819SChris Mason if (item_len) { 2101ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2102ac8e9819SChris Mason /* copy the item */ 2103ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2104ba346b35SGerhard Heift item_off, item_len)) { 2105ba346b35SGerhard Heift ret = -EFAULT; 2106ba346b35SGerhard Heift goto out; 2107ba346b35SGerhard Heift } 2108ba346b35SGerhard Heift 2109ac8e9819SChris Mason *sk_offset += item_len; 2110ac8e9819SChris Mason } 2111e2156867SHugo Mills (*num_found)++; 2112ac8e9819SChris Mason 21138f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 21148f5f6178SGerhard Heift goto out; 21158f5f6178SGerhard Heift 211625c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 211725c9bc2eSGerhard Heift ret = 1; 211825c9bc2eSGerhard Heift goto out; 211925c9bc2eSGerhard Heift } 2120ac8e9819SChris Mason } 2121ac8e9819SChris Mason advance_key: 2122ac8e9819SChris Mason ret = 0; 2123dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2124dd81d459SNaohiro Aota test.type = sk->max_type; 2125dd81d459SNaohiro Aota test.offset = sk->max_offset; 2126dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2127dd81d459SNaohiro Aota ret = 1; 2128dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2129abc6e134SChris Mason key->offset++; 2130dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2131abc6e134SChris Mason key->offset = 0; 2132abc6e134SChris Mason key->type++; 2133dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2134abc6e134SChris Mason key->offset = 0; 2135abc6e134SChris Mason key->type = 0; 2136abc6e134SChris Mason key->objectid++; 2137abc6e134SChris Mason } else 2138abc6e134SChris Mason ret = 1; 213925c9bc2eSGerhard Heift out: 2140ba346b35SGerhard Heift /* 2141ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2142ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2143ba346b35SGerhard Heift * * all items were found 2144ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2145ba346b35SGerhard Heift * leaf 2146ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2147ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2148ba346b35SGerhard Heift */ 2149ac8e9819SChris Mason return ret; 2150ac8e9819SChris Mason } 2151ac8e9819SChris Mason 2152ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 215312544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21549b6e817dSGerhard Heift size_t *buf_size, 2155ba346b35SGerhard Heift char __user *ubuf) 2156ac8e9819SChris Mason { 21570b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2158ac8e9819SChris Mason struct btrfs_root *root; 2159ac8e9819SChris Mason struct btrfs_key key; 2160ac8e9819SChris Mason struct btrfs_path *path; 2161ac8e9819SChris Mason int ret; 2162ac8e9819SChris Mason int num_found = 0; 2163ac8e9819SChris Mason unsigned long sk_offset = 0; 2164ac8e9819SChris Mason 21659b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21669b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 216712544442SGerhard Heift return -EOVERFLOW; 21689b6e817dSGerhard Heift } 216912544442SGerhard Heift 2170ac8e9819SChris Mason path = btrfs_alloc_path(); 2171ac8e9819SChris Mason if (!path) 2172ac8e9819SChris Mason return -ENOMEM; 2173ac8e9819SChris Mason 2174ac8e9819SChris Mason if (sk->tree_id == 0) { 2175ac8e9819SChris Mason /* search the root of the inode that was passed */ 2176ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2177ac8e9819SChris Mason } else { 2178ac8e9819SChris Mason key.objectid = sk->tree_id; 2179ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2180ac8e9819SChris Mason key.offset = (u64)-1; 2181ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2182ac8e9819SChris Mason if (IS_ERR(root)) { 2183ac8e9819SChris Mason btrfs_free_path(path); 2184ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2185ac8e9819SChris Mason } 2186ac8e9819SChris Mason } 2187ac8e9819SChris Mason 2188ac8e9819SChris Mason key.objectid = sk->min_objectid; 2189ac8e9819SChris Mason key.type = sk->min_type; 2190ac8e9819SChris Mason key.offset = sk->min_offset; 2191ac8e9819SChris Mason 2192ac8e9819SChris Mason while (1) { 21936174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2194ac8e9819SChris Mason if (ret != 0) { 2195ac8e9819SChris Mason if (ret > 0) 2196ac8e9819SChris Mason ret = 0; 2197ac8e9819SChris Mason goto err; 2198ac8e9819SChris Mason } 2199df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2200ac8e9819SChris Mason &sk_offset, &num_found); 2201b3b4aa74SDavid Sterba btrfs_release_path(path); 220225c9bc2eSGerhard Heift if (ret) 2203ac8e9819SChris Mason break; 2204ac8e9819SChris Mason 2205ac8e9819SChris Mason } 22068f5f6178SGerhard Heift if (ret > 0) 2207ac8e9819SChris Mason ret = 0; 2208ac8e9819SChris Mason err: 2209ac8e9819SChris Mason sk->nr_items = num_found; 2210ac8e9819SChris Mason btrfs_free_path(path); 2211ac8e9819SChris Mason return ret; 2212ac8e9819SChris Mason } 2213ac8e9819SChris Mason 2214ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2215ac8e9819SChris Mason void __user *argp) 2216ac8e9819SChris Mason { 2217ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2218ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2219ac8e9819SChris Mason struct inode *inode; 2220ac8e9819SChris Mason int ret; 22219b6e817dSGerhard Heift size_t buf_size; 2222ac8e9819SChris Mason 2223ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2224ac8e9819SChris Mason return -EPERM; 2225ac8e9819SChris Mason 2226ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2227ac8e9819SChris Mason 2228ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2229ba346b35SGerhard Heift return -EFAULT; 2230ba346b35SGerhard Heift 2231ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2232ac8e9819SChris Mason 2233496ad9aaSAl Viro inode = file_inode(file); 2234ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22358f5f6178SGerhard Heift 22368f5f6178SGerhard Heift /* 22378f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22388f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22398f5f6178SGerhard Heift */ 22408f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22418f5f6178SGerhard Heift ret = 0; 22428f5f6178SGerhard Heift 2243ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2244ac8e9819SChris Mason ret = -EFAULT; 2245ac8e9819SChris Mason return ret; 2246ac8e9819SChris Mason } 2247ac8e9819SChris Mason 2248cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2249cc68a8a5SGerhard Heift void __user *argp) 2250cc68a8a5SGerhard Heift { 2251cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2252cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2253cc68a8a5SGerhard Heift struct inode *inode; 2254cc68a8a5SGerhard Heift int ret; 2255cc68a8a5SGerhard Heift size_t buf_size; 2256ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2257cc68a8a5SGerhard Heift 2258cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2259cc68a8a5SGerhard Heift return -EPERM; 2260cc68a8a5SGerhard Heift 2261cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2262cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2263cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2264cc68a8a5SGerhard Heift return -EFAULT; 2265cc68a8a5SGerhard Heift 2266cc68a8a5SGerhard Heift buf_size = args.buf_size; 2267cc68a8a5SGerhard Heift 2268cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2269cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2270cc68a8a5SGerhard Heift buf_size = buf_limit; 2271cc68a8a5SGerhard Heift 2272cc68a8a5SGerhard Heift inode = file_inode(file); 2273cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2274718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2275cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2276cc68a8a5SGerhard Heift ret = -EFAULT; 2277cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2278cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2279cc68a8a5SGerhard Heift ret = -EFAULT; 2280cc68a8a5SGerhard Heift 228176dda93cSYan, Zheng return ret; 228276dda93cSYan, Zheng } 228376dda93cSYan, Zheng 228498d377a0STARUISI Hiroaki /* 2285ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2286ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 228798d377a0STARUISI Hiroaki */ 228898d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 228998d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 229098d377a0STARUISI Hiroaki { 229198d377a0STARUISI Hiroaki struct btrfs_root *root; 229298d377a0STARUISI Hiroaki struct btrfs_key key; 2293ac8e9819SChris Mason char *ptr; 229498d377a0STARUISI Hiroaki int ret = -1; 229598d377a0STARUISI Hiroaki int slot; 229698d377a0STARUISI Hiroaki int len; 229798d377a0STARUISI Hiroaki int total_len = 0; 229898d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 229998d377a0STARUISI Hiroaki struct extent_buffer *l; 230098d377a0STARUISI Hiroaki struct btrfs_path *path; 230198d377a0STARUISI Hiroaki 230298d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 230398d377a0STARUISI Hiroaki name[0]='\0'; 230498d377a0STARUISI Hiroaki return 0; 230598d377a0STARUISI Hiroaki } 230698d377a0STARUISI Hiroaki 230798d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 230898d377a0STARUISI Hiroaki if (!path) 230998d377a0STARUISI Hiroaki return -ENOMEM; 231098d377a0STARUISI Hiroaki 2311c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 231298d377a0STARUISI Hiroaki 231398d377a0STARUISI Hiroaki key.objectid = tree_id; 231498d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 231598d377a0STARUISI Hiroaki key.offset = (u64)-1; 231698d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 231798d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2318ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 23198ad6fcabSChris Mason goto out; 232098d377a0STARUISI Hiroaki } 232198d377a0STARUISI Hiroaki 232298d377a0STARUISI Hiroaki key.objectid = dirid; 232398d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 23248ad6fcabSChris Mason key.offset = (u64)-1; 232598d377a0STARUISI Hiroaki 232698d377a0STARUISI Hiroaki while (1) { 232798d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 232898d377a0STARUISI Hiroaki if (ret < 0) 232998d377a0STARUISI Hiroaki goto out; 233018674c6cSFilipe David Borba Manana else if (ret > 0) { 233118674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 233218674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 233318674c6cSFilipe David Borba Manana if (ret < 0) 233418674c6cSFilipe David Borba Manana goto out; 233518674c6cSFilipe David Borba Manana else if (ret > 0) { 2336ac8e9819SChris Mason ret = -ENOENT; 233798d377a0STARUISI Hiroaki goto out; 2338ac8e9819SChris Mason } 233918674c6cSFilipe David Borba Manana } 234018674c6cSFilipe David Borba Manana 234118674c6cSFilipe David Borba Manana l = path->nodes[0]; 234218674c6cSFilipe David Borba Manana slot = path->slots[0]; 234318674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 234498d377a0STARUISI Hiroaki 234598d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 234698d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 234798d377a0STARUISI Hiroaki ptr -= len + 1; 234898d377a0STARUISI Hiroaki total_len += len + 1; 2349a696cf35SFilipe David Borba Manana if (ptr < name) { 2350a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 235198d377a0STARUISI Hiroaki goto out; 2352a696cf35SFilipe David Borba Manana } 235398d377a0STARUISI Hiroaki 235498d377a0STARUISI Hiroaki *(ptr + len) = '/'; 235598d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 235698d377a0STARUISI Hiroaki 235798d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 235898d377a0STARUISI Hiroaki break; 235998d377a0STARUISI Hiroaki 2360b3b4aa74SDavid Sterba btrfs_release_path(path); 236198d377a0STARUISI Hiroaki key.objectid = key.offset; 23628ad6fcabSChris Mason key.offset = (u64)-1; 236398d377a0STARUISI Hiroaki dirid = key.objectid; 236498d377a0STARUISI Hiroaki } 236577906a50SLi Zefan memmove(name, ptr, total_len); 236698d377a0STARUISI Hiroaki name[total_len] = '\0'; 236798d377a0STARUISI Hiroaki ret = 0; 236898d377a0STARUISI Hiroaki out: 236998d377a0STARUISI Hiroaki btrfs_free_path(path); 2370ac8e9819SChris Mason return ret; 2371ac8e9819SChris Mason } 2372ac8e9819SChris Mason 237323d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 237423d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 237523d0b79dSTomohiro Misono { 237623d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 237723d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 237823d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 237923d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 238023d0b79dSTomohiro Misono u64 dirid = args->dirid; 238123d0b79dSTomohiro Misono unsigned long item_off; 238223d0b79dSTomohiro Misono unsigned long item_len; 238323d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 238423d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 238523d0b79dSTomohiro Misono struct btrfs_root *root; 238623d0b79dSTomohiro Misono struct btrfs_path *path; 238723d0b79dSTomohiro Misono struct btrfs_key key, key2; 238823d0b79dSTomohiro Misono struct extent_buffer *leaf; 238923d0b79dSTomohiro Misono struct inode *temp_inode; 239023d0b79dSTomohiro Misono char *ptr; 239123d0b79dSTomohiro Misono int slot; 239223d0b79dSTomohiro Misono int len; 239323d0b79dSTomohiro Misono int total_len = 0; 239423d0b79dSTomohiro Misono int ret; 239523d0b79dSTomohiro Misono 239623d0b79dSTomohiro Misono path = btrfs_alloc_path(); 239723d0b79dSTomohiro Misono if (!path) 239823d0b79dSTomohiro Misono return -ENOMEM; 239923d0b79dSTomohiro Misono 240023d0b79dSTomohiro Misono /* 240123d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 240223d0b79dSTomohiro Misono * construct the path in from the bottom up. 240323d0b79dSTomohiro Misono */ 240423d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 240523d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 240623d0b79dSTomohiro Misono 240723d0b79dSTomohiro Misono key.objectid = treeid; 240823d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 240923d0b79dSTomohiro Misono key.offset = (u64)-1; 241023d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 241123d0b79dSTomohiro Misono if (IS_ERR(root)) { 241223d0b79dSTomohiro Misono ret = PTR_ERR(root); 241323d0b79dSTomohiro Misono goto out; 241423d0b79dSTomohiro Misono } 241523d0b79dSTomohiro Misono 241623d0b79dSTomohiro Misono key.objectid = dirid; 241723d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 241823d0b79dSTomohiro Misono key.offset = (u64)-1; 241923d0b79dSTomohiro Misono while (1) { 242023d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 242123d0b79dSTomohiro Misono if (ret < 0) { 242223d0b79dSTomohiro Misono goto out; 242323d0b79dSTomohiro Misono } else if (ret > 0) { 242423d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 242523d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 242623d0b79dSTomohiro Misono if (ret < 0) { 242723d0b79dSTomohiro Misono goto out; 242823d0b79dSTomohiro Misono } else if (ret > 0) { 242923d0b79dSTomohiro Misono ret = -ENOENT; 243023d0b79dSTomohiro Misono goto out; 243123d0b79dSTomohiro Misono } 243223d0b79dSTomohiro Misono } 243323d0b79dSTomohiro Misono 243423d0b79dSTomohiro Misono leaf = path->nodes[0]; 243523d0b79dSTomohiro Misono slot = path->slots[0]; 243623d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 243723d0b79dSTomohiro Misono 243823d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 243923d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 244023d0b79dSTomohiro Misono ptr -= len + 1; 244123d0b79dSTomohiro Misono total_len += len + 1; 244223d0b79dSTomohiro Misono if (ptr < args->path) { 244323d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 244423d0b79dSTomohiro Misono goto out; 244523d0b79dSTomohiro Misono } 244623d0b79dSTomohiro Misono 244723d0b79dSTomohiro Misono *(ptr + len) = '/'; 244823d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 244923d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 245023d0b79dSTomohiro Misono 245123d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 245223d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 245323d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 245423d0b79dSTomohiro Misono if (ret < 0) { 245523d0b79dSTomohiro Misono goto out; 245623d0b79dSTomohiro Misono } else if (ret > 0) { 245723d0b79dSTomohiro Misono ret = -ENOENT; 245823d0b79dSTomohiro Misono goto out; 245923d0b79dSTomohiro Misono } 246023d0b79dSTomohiro Misono 246123d0b79dSTomohiro Misono leaf = path->nodes[0]; 246223d0b79dSTomohiro Misono slot = path->slots[0]; 246323d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 246423d0b79dSTomohiro Misono if (key2.objectid != dirid) { 246523d0b79dSTomohiro Misono ret = -ENOENT; 246623d0b79dSTomohiro Misono goto out; 246723d0b79dSTomohiro Misono } 246823d0b79dSTomohiro Misono 246923d0b79dSTomohiro Misono temp_inode = btrfs_iget(sb, &key2, root, NULL); 24703ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24713ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24723ca57bd6SMisono Tomohiro goto out; 24733ca57bd6SMisono Tomohiro } 247423d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 247523d0b79dSTomohiro Misono iput(temp_inode); 247623d0b79dSTomohiro Misono if (ret) { 247723d0b79dSTomohiro Misono ret = -EACCES; 247823d0b79dSTomohiro Misono goto out; 247923d0b79dSTomohiro Misono } 248023d0b79dSTomohiro Misono 248123d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 248223d0b79dSTomohiro Misono break; 248323d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 248423d0b79dSTomohiro Misono ret = -EACCES; 248523d0b79dSTomohiro Misono goto out; 248623d0b79dSTomohiro Misono } 248723d0b79dSTomohiro Misono 248823d0b79dSTomohiro Misono btrfs_release_path(path); 248923d0b79dSTomohiro Misono key.objectid = key.offset; 249023d0b79dSTomohiro Misono key.offset = (u64)-1; 249123d0b79dSTomohiro Misono dirid = key.objectid; 249223d0b79dSTomohiro Misono } 249323d0b79dSTomohiro Misono 249423d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 249523d0b79dSTomohiro Misono args->path[total_len] = '\0'; 249623d0b79dSTomohiro Misono btrfs_release_path(path); 249723d0b79dSTomohiro Misono } 249823d0b79dSTomohiro Misono 249923d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 250023d0b79dSTomohiro Misono root = fs_info->tree_root; 250123d0b79dSTomohiro Misono key.objectid = treeid; 250223d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 250323d0b79dSTomohiro Misono key.offset = args->treeid; 250423d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 250523d0b79dSTomohiro Misono if (ret < 0) { 250623d0b79dSTomohiro Misono goto out; 250723d0b79dSTomohiro Misono } else if (ret > 0) { 250823d0b79dSTomohiro Misono ret = -ENOENT; 250923d0b79dSTomohiro Misono goto out; 251023d0b79dSTomohiro Misono } 251123d0b79dSTomohiro Misono 251223d0b79dSTomohiro Misono leaf = path->nodes[0]; 251323d0b79dSTomohiro Misono slot = path->slots[0]; 251423d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 251523d0b79dSTomohiro Misono 251623d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 251723d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 251823d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 251923d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 252023d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 252123d0b79dSTomohiro Misono ret = -EINVAL; 252223d0b79dSTomohiro Misono goto out; 252323d0b79dSTomohiro Misono } 252423d0b79dSTomohiro Misono 252523d0b79dSTomohiro Misono /* Copy subvolume's name */ 252623d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 252723d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 252823d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 252923d0b79dSTomohiro Misono args->name[item_len] = 0; 253023d0b79dSTomohiro Misono 253123d0b79dSTomohiro Misono out: 253223d0b79dSTomohiro Misono btrfs_free_path(path); 253323d0b79dSTomohiro Misono return ret; 253423d0b79dSTomohiro Misono } 253523d0b79dSTomohiro Misono 2536ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2537ac8e9819SChris Mason void __user *argp) 2538ac8e9819SChris Mason { 2539ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2540ac8e9819SChris Mason struct inode *inode; 254101b810b8SDavid Sterba int ret = 0; 2542ac8e9819SChris Mason 25432354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25442354d08fSJulia Lawall if (IS_ERR(args)) 25452354d08fSJulia Lawall return PTR_ERR(args); 2546c2b96929SDan Carpenter 2547496ad9aaSAl Viro inode = file_inode(file); 2548ac8e9819SChris Mason 254901b810b8SDavid Sterba /* 255001b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 255101b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 255201b810b8SDavid Sterba */ 25531b53ac4dSChris Mason if (args->treeid == 0) 25541b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25551b53ac4dSChris Mason 255601b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 255701b810b8SDavid Sterba args->name[0] = 0; 255801b810b8SDavid Sterba goto out; 255901b810b8SDavid Sterba } 256001b810b8SDavid Sterba 256101b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 256201b810b8SDavid Sterba ret = -EPERM; 256301b810b8SDavid Sterba goto out; 256401b810b8SDavid Sterba } 256501b810b8SDavid Sterba 2566ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2567ac8e9819SChris Mason args->treeid, args->objectid, 2568ac8e9819SChris Mason args->name); 2569ac8e9819SChris Mason 257001b810b8SDavid Sterba out: 2571ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2572ac8e9819SChris Mason ret = -EFAULT; 2573ac8e9819SChris Mason 2574ac8e9819SChris Mason kfree(args); 257598d377a0STARUISI Hiroaki return ret; 257698d377a0STARUISI Hiroaki } 257798d377a0STARUISI Hiroaki 257823d0b79dSTomohiro Misono /* 257923d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 258023d0b79dSTomohiro Misono * 258123d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 258223d0b79dSTomohiro Misono * 258323d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 258423d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 258523d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 258623d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 258723d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 258823d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 258923d0b79dSTomohiro Misono */ 259023d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 259123d0b79dSTomohiro Misono { 259223d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 259323d0b79dSTomohiro Misono struct inode *inode; 259423d0b79dSTomohiro Misono int ret; 259523d0b79dSTomohiro Misono 259623d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 259723d0b79dSTomohiro Misono if (IS_ERR(args)) 259823d0b79dSTomohiro Misono return PTR_ERR(args); 259923d0b79dSTomohiro Misono 260023d0b79dSTomohiro Misono inode = file_inode(file); 260123d0b79dSTomohiro Misono 260223d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 260323d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 260423d0b79dSTomohiro Misono /* 260523d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 260623d0b79dSTomohiro Misono * called 260723d0b79dSTomohiro Misono */ 260823d0b79dSTomohiro Misono kfree(args); 260923d0b79dSTomohiro Misono return -EACCES; 261023d0b79dSTomohiro Misono } 261123d0b79dSTomohiro Misono 261223d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 261323d0b79dSTomohiro Misono 261423d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 261523d0b79dSTomohiro Misono ret = -EFAULT; 261623d0b79dSTomohiro Misono 261723d0b79dSTomohiro Misono kfree(args); 261823d0b79dSTomohiro Misono return ret; 261923d0b79dSTomohiro Misono } 262023d0b79dSTomohiro Misono 2621b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2622b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2623b64ec075STomohiro Misono { 2624b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2625b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2626b64ec075STomohiro Misono struct btrfs_root *root; 2627b64ec075STomohiro Misono struct btrfs_path *path; 2628b64ec075STomohiro Misono struct btrfs_key key; 2629b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2630b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2631b64ec075STomohiro Misono struct extent_buffer *leaf; 2632b64ec075STomohiro Misono unsigned long item_off; 2633b64ec075STomohiro Misono unsigned long item_len; 2634b64ec075STomohiro Misono struct inode *inode; 2635b64ec075STomohiro Misono int slot; 2636b64ec075STomohiro Misono int ret = 0; 2637b64ec075STomohiro Misono 2638b64ec075STomohiro Misono path = btrfs_alloc_path(); 2639b64ec075STomohiro Misono if (!path) 2640b64ec075STomohiro Misono return -ENOMEM; 2641b64ec075STomohiro Misono 2642b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2643b64ec075STomohiro Misono if (!subvol_info) { 2644b64ec075STomohiro Misono btrfs_free_path(path); 2645b64ec075STomohiro Misono return -ENOMEM; 2646b64ec075STomohiro Misono } 2647b64ec075STomohiro Misono 2648b64ec075STomohiro Misono inode = file_inode(file); 2649b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2650b64ec075STomohiro Misono 2651b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2652b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2653b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2654b64ec075STomohiro Misono key.offset = (u64)-1; 2655b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2656b64ec075STomohiro Misono if (IS_ERR(root)) { 2657b64ec075STomohiro Misono ret = PTR_ERR(root); 2658b64ec075STomohiro Misono goto out; 2659b64ec075STomohiro Misono } 2660b64ec075STomohiro Misono root_item = &root->root_item; 2661b64ec075STomohiro Misono 2662b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2663b64ec075STomohiro Misono 2664b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2665b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2666b64ec075STomohiro Misono 2667b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2668b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2669b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2670b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2671b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2672b64ec075STomohiro Misono 2673b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2674b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2675b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2676b64ec075STomohiro Misono 2677b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2678b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2679b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2680b64ec075STomohiro Misono 2681b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2682b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2683b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2684b64ec075STomohiro Misono 2685b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2686b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2687b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2688b64ec075STomohiro Misono 2689b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2690b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2691b64ec075STomohiro Misono root = fs_info->tree_root; 2692b64ec075STomohiro Misono 2693b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2694b64ec075STomohiro Misono key.offset = 0; 2695b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2696b64ec075STomohiro Misono if (ret < 0) { 2697b64ec075STomohiro Misono goto out; 2698b64ec075STomohiro Misono } else if (path->slots[0] >= 2699b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2700b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2701b64ec075STomohiro Misono if (ret < 0) { 2702b64ec075STomohiro Misono goto out; 2703b64ec075STomohiro Misono } else if (ret > 0) { 2704b64ec075STomohiro Misono ret = -EUCLEAN; 2705b64ec075STomohiro Misono goto out; 2706b64ec075STomohiro Misono } 2707b64ec075STomohiro Misono } 2708b64ec075STomohiro Misono 2709b64ec075STomohiro Misono leaf = path->nodes[0]; 2710b64ec075STomohiro Misono slot = path->slots[0]; 2711b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2712b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2713b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2714b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2715b64ec075STomohiro Misono 2716b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2717b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2718b64ec075STomohiro Misono 2719b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2720b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2721b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2722b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2723b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2724b64ec075STomohiro Misono item_off, item_len); 2725b64ec075STomohiro Misono } else { 2726b64ec075STomohiro Misono ret = -ENOENT; 2727b64ec075STomohiro Misono goto out; 2728b64ec075STomohiro Misono } 2729b64ec075STomohiro Misono } 2730b64ec075STomohiro Misono 2731b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2732b64ec075STomohiro Misono ret = -EFAULT; 2733b64ec075STomohiro Misono 2734b64ec075STomohiro Misono out: 2735b64ec075STomohiro Misono btrfs_free_path(path); 2736b64ec075STomohiro Misono kzfree(subvol_info); 2737b64ec075STomohiro Misono return ret; 2738b64ec075STomohiro Misono } 2739b64ec075STomohiro Misono 274042e4b520STomohiro Misono /* 274142e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 274242e4b520STomohiro Misono * except the subvolume name. 274342e4b520STomohiro Misono */ 274442e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 274542e4b520STomohiro Misono { 274642e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 274742e4b520STomohiro Misono struct btrfs_root_ref *rref; 274842e4b520STomohiro Misono struct btrfs_root *root; 274942e4b520STomohiro Misono struct btrfs_path *path; 275042e4b520STomohiro Misono struct btrfs_key key; 275142e4b520STomohiro Misono struct extent_buffer *leaf; 275242e4b520STomohiro Misono struct inode *inode; 275342e4b520STomohiro Misono u64 objectid; 275442e4b520STomohiro Misono int slot; 275542e4b520STomohiro Misono int ret; 275642e4b520STomohiro Misono u8 found; 275742e4b520STomohiro Misono 275842e4b520STomohiro Misono path = btrfs_alloc_path(); 275942e4b520STomohiro Misono if (!path) 276042e4b520STomohiro Misono return -ENOMEM; 276142e4b520STomohiro Misono 276242e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 276342e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 276442e4b520STomohiro Misono btrfs_free_path(path); 276542e4b520STomohiro Misono return PTR_ERR(rootrefs); 276642e4b520STomohiro Misono } 276742e4b520STomohiro Misono 276842e4b520STomohiro Misono inode = file_inode(file); 276942e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 277042e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 277142e4b520STomohiro Misono 277242e4b520STomohiro Misono key.objectid = objectid; 277342e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 277442e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 277542e4b520STomohiro Misono found = 0; 277642e4b520STomohiro Misono 277742e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 277842e4b520STomohiro Misono if (ret < 0) { 277942e4b520STomohiro Misono goto out; 278042e4b520STomohiro Misono } else if (path->slots[0] >= 278142e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 278242e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 278342e4b520STomohiro Misono if (ret < 0) { 278442e4b520STomohiro Misono goto out; 278542e4b520STomohiro Misono } else if (ret > 0) { 278642e4b520STomohiro Misono ret = -EUCLEAN; 278742e4b520STomohiro Misono goto out; 278842e4b520STomohiro Misono } 278942e4b520STomohiro Misono } 279042e4b520STomohiro Misono while (1) { 279142e4b520STomohiro Misono leaf = path->nodes[0]; 279242e4b520STomohiro Misono slot = path->slots[0]; 279342e4b520STomohiro Misono 279442e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 279542e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 279642e4b520STomohiro Misono ret = 0; 279742e4b520STomohiro Misono goto out; 279842e4b520STomohiro Misono } 279942e4b520STomohiro Misono 280042e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 280142e4b520STomohiro Misono ret = -EOVERFLOW; 280242e4b520STomohiro Misono goto out; 280342e4b520STomohiro Misono } 280442e4b520STomohiro Misono 280542e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 280642e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 280742e4b520STomohiro Misono rootrefs->rootref[found].dirid = 280842e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 280942e4b520STomohiro Misono found++; 281042e4b520STomohiro Misono 281142e4b520STomohiro Misono ret = btrfs_next_item(root, path); 281242e4b520STomohiro Misono if (ret < 0) { 281342e4b520STomohiro Misono goto out; 281442e4b520STomohiro Misono } else if (ret > 0) { 281542e4b520STomohiro Misono ret = -EUCLEAN; 281642e4b520STomohiro Misono goto out; 281742e4b520STomohiro Misono } 281842e4b520STomohiro Misono } 281942e4b520STomohiro Misono 282042e4b520STomohiro Misono out: 282142e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 282242e4b520STomohiro Misono rootrefs->num_items = found; 282342e4b520STomohiro Misono /* update min_treeid for next search */ 282442e4b520STomohiro Misono if (found) 282542e4b520STomohiro Misono rootrefs->min_treeid = 282642e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 282742e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 282842e4b520STomohiro Misono ret = -EFAULT; 282942e4b520STomohiro Misono } 283042e4b520STomohiro Misono 283142e4b520STomohiro Misono kfree(rootrefs); 283242e4b520STomohiro Misono btrfs_free_path(path); 283342e4b520STomohiro Misono 283442e4b520STomohiro Misono return ret; 283542e4b520STomohiro Misono } 283642e4b520STomohiro Misono 283776dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 283876dda93cSYan, Zheng void __user *arg) 283976dda93cSYan, Zheng { 284054563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28410b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 284276dda93cSYan, Zheng struct dentry *dentry; 28432b0143b5SDavid Howells struct inode *dir = d_inode(parent); 284476dda93cSYan, Zheng struct inode *inode; 284576dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 284676dda93cSYan, Zheng struct btrfs_root *dest = NULL; 284776dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 284876dda93cSYan, Zheng int namelen; 284976dda93cSYan, Zheng int err = 0; 285076dda93cSYan, Zheng 2851325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2852325c50e3SJeff Mahoney return -ENOTDIR; 2853325c50e3SJeff Mahoney 285476dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 285576dda93cSYan, Zheng if (IS_ERR(vol_args)) 285676dda93cSYan, Zheng return PTR_ERR(vol_args); 285776dda93cSYan, Zheng 285876dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 285976dda93cSYan, Zheng namelen = strlen(vol_args->name); 286076dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 286176dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 286276dda93cSYan, Zheng err = -EINVAL; 286376dda93cSYan, Zheng goto out; 286476dda93cSYan, Zheng } 286576dda93cSYan, Zheng 2866a561be71SAl Viro err = mnt_want_write_file(file); 286776dda93cSYan, Zheng if (err) 286876dda93cSYan, Zheng goto out; 286976dda93cSYan, Zheng 2870521e0546SDavid Sterba 287100235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 287200235411SAl Viro if (err == -EINTR) 287300235411SAl Viro goto out_drop_write; 287476dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 287576dda93cSYan, Zheng if (IS_ERR(dentry)) { 287676dda93cSYan, Zheng err = PTR_ERR(dentry); 287776dda93cSYan, Zheng goto out_unlock_dir; 287876dda93cSYan, Zheng } 287976dda93cSYan, Zheng 28802b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 288176dda93cSYan, Zheng err = -ENOENT; 288276dda93cSYan, Zheng goto out_dput; 288376dda93cSYan, Zheng } 288476dda93cSYan, Zheng 28852b0143b5SDavid Howells inode = d_inode(dentry); 28864260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28874260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28884260f7c7SSage Weil /* 28894260f7c7SSage Weil * Regular user. Only allow this with a special mount 28904260f7c7SSage Weil * option, when the user has write+exec access to the 28914260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28924260f7c7SSage Weil * allowed. 28934260f7c7SSage Weil * 28944260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28954260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28964260f7c7SSage Weil * otherwise be able to delete. 28974260f7c7SSage Weil * 28984260f7c7SSage Weil * Users who want to delete empty subvols should try 28994260f7c7SSage Weil * rmdir(2). 29004260f7c7SSage Weil */ 29014260f7c7SSage Weil err = -EPERM; 29020b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 29034260f7c7SSage Weil goto out_dput; 29044260f7c7SSage Weil 29054260f7c7SSage Weil /* 29064260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 29074260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 29084260f7c7SSage Weil * must be called on the dentry referencing the root 29094260f7c7SSage Weil * of the subvol, not a random directory contained 29104260f7c7SSage Weil * within it. 29114260f7c7SSage Weil */ 29124260f7c7SSage Weil err = -EINVAL; 29134260f7c7SSage Weil if (root == dest) 29144260f7c7SSage Weil goto out_dput; 29154260f7c7SSage Weil 29164260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 29174260f7c7SSage Weil if (err) 29184260f7c7SSage Weil goto out_dput; 29195c39da5bSMiao Xie } 29204260f7c7SSage Weil 29215c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 29224260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 29234260f7c7SSage Weil if (err) 29244260f7c7SSage Weil goto out_dput; 29254260f7c7SSage Weil 29264a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 292776dda93cSYan, Zheng err = -EINVAL; 292876dda93cSYan, Zheng goto out_dput; 292976dda93cSYan, Zheng } 293076dda93cSYan, Zheng 29315955102cSAl Viro inode_lock(inode); 2932f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 29335955102cSAl Viro inode_unlock(inode); 2934f60a2364SMisono Tomohiro if (!err) 293576dda93cSYan, Zheng d_delete(dentry); 2936fa6ac876SLiu Bo 293776dda93cSYan, Zheng out_dput: 293876dda93cSYan, Zheng dput(dentry); 293976dda93cSYan, Zheng out_unlock_dir: 29405955102cSAl Viro inode_unlock(dir); 294100235411SAl Viro out_drop_write: 29422a79f17eSAl Viro mnt_drop_write_file(file); 294376dda93cSYan, Zheng out: 294476dda93cSYan, Zheng kfree(vol_args); 294576dda93cSYan, Zheng return err; 294676dda93cSYan, Zheng } 294776dda93cSYan, Zheng 29481e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2949f46b5a66SChristoph Hellwig { 2950496ad9aaSAl Viro struct inode *inode = file_inode(file); 2951f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29521e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2953c146afadSYan Zheng int ret; 2954c146afadSYan Zheng 295525122d15SIlya Dryomov ret = mnt_want_write_file(file); 295625122d15SIlya Dryomov if (ret) 295725122d15SIlya Dryomov return ret; 2958b83cc969SLi Zefan 295925122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 296025122d15SIlya Dryomov ret = -EROFS; 296125122d15SIlya Dryomov goto out; 29625ac00addSStefan Behrens } 2963f46b5a66SChristoph Hellwig 2964f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2965f46b5a66SChristoph Hellwig case S_IFDIR: 2966e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2967e441d54dSChris Mason ret = -EPERM; 2968e441d54dSChris Mason goto out; 2969e441d54dSChris Mason } 2970de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2971f46b5a66SChristoph Hellwig break; 2972f46b5a66SChristoph Hellwig case S_IFREG: 2973616d374eSAdam Borowski /* 2974616d374eSAdam Borowski * Note that this does not check the file descriptor for write 2975616d374eSAdam Borowski * access. This prevents defragmenting executables that are 2976616d374eSAdam Borowski * running and allows defrag on files open in read-only mode. 2977616d374eSAdam Borowski */ 2978616d374eSAdam Borowski if (!capable(CAP_SYS_ADMIN) && 2979616d374eSAdam Borowski inode_permission(inode, MAY_WRITE)) { 2980616d374eSAdam Borowski ret = -EPERM; 2981e441d54dSChris Mason goto out; 2982e441d54dSChris Mason } 29831e701a32SChris Mason 29841e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29851e701a32SChris Mason if (!range) { 29861e701a32SChris Mason ret = -ENOMEM; 29871e701a32SChris Mason goto out; 29881e701a32SChris Mason } 29891e701a32SChris Mason 29901e701a32SChris Mason if (argp) { 29911e701a32SChris Mason if (copy_from_user(range, argp, 29921e701a32SChris Mason sizeof(*range))) { 29931e701a32SChris Mason ret = -EFAULT; 29941e701a32SChris Mason kfree(range); 2995683be16eSDan Carpenter goto out; 29961e701a32SChris Mason } 29971e701a32SChris Mason /* compression requires us to start the IO */ 29981e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 29991e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 30001e701a32SChris Mason range->extent_thresh = (u32)-1; 30011e701a32SChris Mason } 30021e701a32SChris Mason } else { 30031e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 30041e701a32SChris Mason range->len = (u64)-1; 30051e701a32SChris Mason } 3006496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 30077c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 30084cb5300bSChris Mason if (ret > 0) 30094cb5300bSChris Mason ret = 0; 30101e701a32SChris Mason kfree(range); 3011f46b5a66SChristoph Hellwig break; 30128929ecfaSYan, Zheng default: 30138929ecfaSYan, Zheng ret = -EINVAL; 3014f46b5a66SChristoph Hellwig } 3015e441d54dSChris Mason out: 301625122d15SIlya Dryomov mnt_drop_write_file(file); 3017e441d54dSChris Mason return ret; 3018f46b5a66SChristoph Hellwig } 3019f46b5a66SChristoph Hellwig 30202ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 3021f46b5a66SChristoph Hellwig { 3022f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3023f46b5a66SChristoph Hellwig int ret; 3024f46b5a66SChristoph Hellwig 3025e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3026e441d54dSChris Mason return -EPERM; 3027e441d54dSChris Mason 3028171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 3029e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3030c9e9f97bSIlya Dryomov 3031dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3032c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3033c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3034c9e9f97bSIlya Dryomov goto out; 3035c9e9f97bSIlya Dryomov } 3036f46b5a66SChristoph Hellwig 30375516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30382ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3039f46b5a66SChristoph Hellwig 304043d20761SAnand Jain if (!ret) 30410b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 304243d20761SAnand Jain 3043f46b5a66SChristoph Hellwig kfree(vol_args); 3044c9e9f97bSIlya Dryomov out: 3045171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3046f46b5a66SChristoph Hellwig return ret; 3047f46b5a66SChristoph Hellwig } 3048f46b5a66SChristoph Hellwig 30496b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3050f46b5a66SChristoph Hellwig { 30510b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30520b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30536b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3054f46b5a66SChristoph Hellwig int ret; 3055f46b5a66SChristoph Hellwig 3056e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3057e441d54dSChris Mason return -EPERM; 3058e441d54dSChris Mason 3059da24927bSMiao Xie ret = mnt_want_write_file(file); 3060da24927bSMiao Xie if (ret) 3061da24927bSMiao Xie return ret; 3062c146afadSYan Zheng 3063dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3064c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3065c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3066c47ca32dSDan Carpenter goto err_drop; 3067c9e9f97bSIlya Dryomov } 3068f46b5a66SChristoph Hellwig 30696b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3070fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3071fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3072fd4e994bSOmar Sandoval goto out; 3073fd4e994bSOmar Sandoval } 3074f46b5a66SChristoph Hellwig 3075171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3076183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3077183860f6SAnand Jain goto out; 3078183860f6SAnand Jain } 3079183860f6SAnand Jain 3080735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30812ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30826b526ed7SAnand Jain } else { 30836b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30842ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30856b526ed7SAnand Jain } 3086171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3087183860f6SAnand Jain 30886b526ed7SAnand Jain if (!ret) { 3089735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30900b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30916b526ed7SAnand Jain vol_args->devid); 30926b526ed7SAnand Jain else 30930b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30946b526ed7SAnand Jain vol_args->name); 30956b526ed7SAnand Jain } 3096183860f6SAnand Jain out: 3097183860f6SAnand Jain kfree(vol_args); 3098c47ca32dSDan Carpenter err_drop: 30994ac20c70SIlya Dryomov mnt_drop_write_file(file); 3100f46b5a66SChristoph Hellwig return ret; 3101f46b5a66SChristoph Hellwig } 3102f46b5a66SChristoph Hellwig 3103f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3104f46b5a66SChristoph Hellwig { 31050b246afaSJeff Mahoney struct inode *inode = file_inode(file); 31060b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3107f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3108f46b5a66SChristoph Hellwig int ret; 3109f46b5a66SChristoph Hellwig 3110f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3111f46b5a66SChristoph Hellwig return -EPERM; 3112f46b5a66SChristoph Hellwig 3113f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3114f46b5a66SChristoph Hellwig if (ret) 3115f46b5a66SChristoph Hellwig return ret; 3116f46b5a66SChristoph Hellwig 3117171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3118f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 311958d7bbf8SDavid Sterba goto out_drop_write; 312058d7bbf8SDavid Sterba } 312158d7bbf8SDavid Sterba 312258d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 312358d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 312458d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3125f46b5a66SChristoph Hellwig goto out; 3126f46b5a66SChristoph Hellwig } 3127f46b5a66SChristoph Hellwig 312858d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 31292ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3130f46b5a66SChristoph Hellwig 3131f46b5a66SChristoph Hellwig if (!ret) 31320b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3133f46b5a66SChristoph Hellwig kfree(vol_args); 313458d7bbf8SDavid Sterba out: 3135171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 313658d7bbf8SDavid Sterba out_drop_write: 3137f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 313858d7bbf8SDavid Sterba 3139f46b5a66SChristoph Hellwig return ret; 3140f46b5a66SChristoph Hellwig } 3141f46b5a66SChristoph Hellwig 31422ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 31432ff7e61eSJeff Mahoney void __user *arg) 3144475f6387SJan Schmidt { 3145027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3146475f6387SJan Schmidt struct btrfs_device *device; 31470b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3148027ed2f0SLi Zefan int ret = 0; 3149475f6387SJan Schmidt 3150027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3151027ed2f0SLi Zefan if (!fi_args) 3152027ed2f0SLi Zefan return -ENOMEM; 3153027ed2f0SLi Zefan 3154d03262c7SDavid Sterba rcu_read_lock(); 3155027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3156475f6387SJan Schmidt 3157d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3158027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3159027ed2f0SLi Zefan fi_args->max_id = device->devid; 3160475f6387SJan Schmidt } 3161d03262c7SDavid Sterba rcu_read_unlock(); 3162475f6387SJan Schmidt 3163de37aa51SNikolay Borisov memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 3164bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3165bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3166bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 316780a773fbSDavid Sterba 3168027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3169027ed2f0SLi Zefan ret = -EFAULT; 3170475f6387SJan Schmidt 3171027ed2f0SLi Zefan kfree(fi_args); 3172027ed2f0SLi Zefan return ret; 3173475f6387SJan Schmidt } 3174475f6387SJan Schmidt 31752ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31762ff7e61eSJeff Mahoney void __user *arg) 3177475f6387SJan Schmidt { 3178475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3179475f6387SJan Schmidt struct btrfs_device *dev; 3180475f6387SJan Schmidt int ret = 0; 3181475f6387SJan Schmidt char *s_uuid = NULL; 3182475f6387SJan Schmidt 3183475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3184475f6387SJan Schmidt if (IS_ERR(di_args)) 3185475f6387SJan Schmidt return PTR_ERR(di_args); 3186475f6387SJan Schmidt 3187dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3188475f6387SJan Schmidt s_uuid = di_args->uuid; 3189475f6387SJan Schmidt 3190c5593ca3SDavid Sterba rcu_read_lock(); 3191e4319cd9SAnand Jain dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid, 319209ba3bc9SAnand Jain NULL, true); 3193475f6387SJan Schmidt 3194475f6387SJan Schmidt if (!dev) { 3195475f6387SJan Schmidt ret = -ENODEV; 3196475f6387SJan Schmidt goto out; 3197475f6387SJan Schmidt } 3198475f6387SJan Schmidt 3199475f6387SJan Schmidt di_args->devid = dev->devid; 32007cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 32017cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3202475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3203a27202fbSJim Meyering if (dev->name) { 3204672d5990SMisono Tomohiro strncpy(di_args->path, rcu_str_deref(dev->name), 3205672d5990SMisono Tomohiro sizeof(di_args->path) - 1); 3206a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3207a27202fbSJim Meyering } else { 320899ba55adSStefan Behrens di_args->path[0] = '\0'; 3209a27202fbSJim Meyering } 3210475f6387SJan Schmidt 3211475f6387SJan Schmidt out: 3212c5593ca3SDavid Sterba rcu_read_unlock(); 3213475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3214475f6387SJan Schmidt ret = -EFAULT; 3215475f6387SJan Schmidt 3216475f6387SJan Schmidt kfree(di_args); 3217475f6387SJan Schmidt return ret; 3218475f6387SJan Schmidt } 3219475f6387SJan Schmidt 3220d8b55242SFilipe Manana static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3221d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3222d8b55242SFilipe Manana { 3223d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3224d8b55242SFilipe Manana unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3225d8b55242SFilipe Manana } 3226d8b55242SFilipe Manana 3227d8b55242SFilipe Manana static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3228d8b55242SFilipe Manana struct inode *inode2, u64 loff2, u64 len) 3229d8b55242SFilipe Manana { 3230d8b55242SFilipe Manana if (inode1 < inode2) { 3231d8b55242SFilipe Manana swap(inode1, inode2); 3232d8b55242SFilipe Manana swap(loff1, loff2); 3233d8b55242SFilipe Manana } else if (inode1 == inode2 && loff2 < loff1) { 3234d8b55242SFilipe Manana swap(loff1, loff2); 3235d8b55242SFilipe Manana } 3236d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3237d8b55242SFilipe Manana lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3238d8b55242SFilipe Manana } 3239d8b55242SFilipe Manana 324057a50e25SFilipe Manana static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len, 324134a28e3dSFilipe Manana struct inode *dst, u64 dst_loff) 3242416161dbSMark Fasheh { 3243416161dbSMark Fasheh int ret; 32440efa9f48SMark Fasheh 3245de02b9f6SFilipe Manana /* 3246d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3247d8b55242SFilipe Manana * source range to serialize with relocation. 3248de02b9f6SFilipe Manana */ 3249d8b55242SFilipe Manana btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 325057a50e25SFilipe Manana ret = btrfs_clone(src, dst, loff, len, len, dst_loff, 1); 3251d8b55242SFilipe Manana btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 32523973909dSTimofey Titovets 32533973909dSTimofey Titovets return ret; 32543973909dSTimofey Titovets } 32553973909dSTimofey Titovets 3256b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3257b6728768STimofey Titovets 32583973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 32593973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 32603973909dSTimofey Titovets { 32613973909dSTimofey Titovets int ret; 3262b6728768STimofey Titovets u64 i, tail_len, chunk_count; 32633973909dSTimofey Titovets 3264b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3265b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 326667b07bd4STimofey Titovets 3267b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3268b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 326934a28e3dSFilipe Manana dst, dst_loff); 3270b6728768STimofey Titovets if (ret) 327134a28e3dSFilipe Manana return ret; 3272b6728768STimofey Titovets 3273b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3274b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3275b6728768STimofey Titovets } 3276b6728768STimofey Titovets 3277b6728768STimofey Titovets if (tail_len > 0) 327867b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 327934a28e3dSFilipe Manana dst_loff); 3280416161dbSMark Fasheh 3281416161dbSMark Fasheh return ret; 3282416161dbSMark Fasheh } 3283416161dbSMark Fasheh 3284f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3285f82a9901SFilipe Manana struct inode *inode, 3286f82a9901SFilipe Manana u64 endoff, 3287f82a9901SFilipe Manana const u64 destoff, 32881c919a5eSMark Fasheh const u64 olen, 32891c919a5eSMark Fasheh int no_time_update) 3290f82a9901SFilipe Manana { 3291f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3292f82a9901SFilipe Manana int ret; 3293f82a9901SFilipe Manana 3294f82a9901SFilipe Manana inode_inc_iversion(inode); 32951c919a5eSMark Fasheh if (!no_time_update) 3296c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3297f82a9901SFilipe Manana /* 3298f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3299f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3300f82a9901SFilipe Manana */ 3301f82a9901SFilipe Manana if (endoff > destoff + olen) 3302f82a9901SFilipe Manana endoff = destoff + olen; 3303f82a9901SFilipe Manana if (endoff > inode->i_size) 33046ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3305f82a9901SFilipe Manana 3306f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3307f82a9901SFilipe Manana if (ret) { 330866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 33093a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3310f82a9901SFilipe Manana goto out; 3311f82a9901SFilipe Manana } 33123a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3313f82a9901SFilipe Manana out: 3314f82a9901SFilipe Manana return ret; 3315f82a9901SFilipe Manana } 3316f82a9901SFilipe Manana 3317a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 33187ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 33197ffbb598SFilipe Manana const struct btrfs_path *path, 33207ffbb598SFilipe Manana const u64 hole_offset, 33217ffbb598SFilipe Manana const u64 hole_len) 33227ffbb598SFilipe Manana { 3323a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 33247ffbb598SFilipe Manana struct extent_map *em; 33257ffbb598SFilipe Manana int ret; 33267ffbb598SFilipe Manana 33277ffbb598SFilipe Manana em = alloc_extent_map(); 33287ffbb598SFilipe Manana if (!em) { 3329a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33307ffbb598SFilipe Manana return; 33317ffbb598SFilipe Manana } 33327ffbb598SFilipe Manana 333314f59796SFilipe Manana if (path) { 333414f59796SFilipe Manana struct btrfs_file_extent_item *fi; 333514f59796SFilipe Manana 333614f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 333714f59796SFilipe Manana struct btrfs_file_extent_item); 3338a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 33397ffbb598SFilipe Manana em->generation = -1; 33407ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 33417ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33427ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3343a2f392e4SNikolay Borisov &inode->runtime_flags); 33447ffbb598SFilipe Manana } else { 33457ffbb598SFilipe Manana em->start = hole_offset; 33467ffbb598SFilipe Manana em->len = hole_len; 33477ffbb598SFilipe Manana em->ram_bytes = em->len; 33487ffbb598SFilipe Manana em->orig_start = hole_offset; 33497ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33507ffbb598SFilipe Manana em->block_len = 0; 33517ffbb598SFilipe Manana em->orig_block_len = 0; 33527ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33537ffbb598SFilipe Manana em->generation = trans->transid; 33547ffbb598SFilipe Manana } 33557ffbb598SFilipe Manana 33567ffbb598SFilipe Manana while (1) { 33577ffbb598SFilipe Manana write_lock(&em_tree->lock); 33587ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33597ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33607ffbb598SFilipe Manana if (ret != -EEXIST) { 33617ffbb598SFilipe Manana free_extent_map(em); 33627ffbb598SFilipe Manana break; 33637ffbb598SFilipe Manana } 3364a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 33657ffbb598SFilipe Manana em->start + em->len - 1, 0); 33667ffbb598SFilipe Manana } 33677ffbb598SFilipe Manana 3368ee39b432SDavid Sterba if (ret) 3369a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33707ffbb598SFilipe Manana } 33717ffbb598SFilipe Manana 33728039d87dSFilipe Manana /* 33738039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33748039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33758039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33768039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33778039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33788039d87dSFilipe Manana * 33798039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33808039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33818039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33828039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33838039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33848039d87dSFilipe Manana * i_mutex, which we are holding here). 33858039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33868039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33878039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33888039d87dSFilipe Manana * 33898039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33908039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33918039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33928039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33938039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33948039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33958039d87dSFilipe Manana * the former. 33968039d87dSFilipe Manana */ 33974a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 33988039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33998039d87dSFilipe Manana struct btrfs_path *path, 34008039d87dSFilipe Manana struct btrfs_key *new_key, 34018039d87dSFilipe Manana const u64 drop_start, 34028039d87dSFilipe Manana const u64 datal, 34038039d87dSFilipe Manana const u64 skip, 34048039d87dSFilipe Manana const u64 size, 34058039d87dSFilipe Manana char *inline_data) 34068039d87dSFilipe Manana { 34070b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 34088039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 34098039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 34100b246afaSJeff Mahoney fs_info->sectorsize); 34118039d87dSFilipe Manana int ret; 34128039d87dSFilipe Manana struct btrfs_key key; 34138039d87dSFilipe Manana 34148039d87dSFilipe Manana if (new_key->offset > 0) 34158039d87dSFilipe Manana return -EOPNOTSUPP; 34168039d87dSFilipe Manana 34174a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 34188039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 34198039d87dSFilipe Manana key.offset = 0; 34208039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 34218039d87dSFilipe Manana if (ret < 0) { 34228039d87dSFilipe Manana return ret; 34238039d87dSFilipe Manana } else if (ret > 0) { 34248039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 34258039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 34268039d87dSFilipe Manana if (ret < 0) 34278039d87dSFilipe Manana return ret; 34288039d87dSFilipe Manana else if (ret > 0) 34298039d87dSFilipe Manana goto copy_inline_extent; 34308039d87dSFilipe Manana } 34318039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 34324a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34338039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 34348039d87dSFilipe Manana ASSERT(key.offset > 0); 34358039d87dSFilipe Manana return -EOPNOTSUPP; 34368039d87dSFilipe Manana } 34378039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34388039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34398039d87dSFilipe Manana u64 ext_len; 34408039d87dSFilipe Manana 34418039d87dSFilipe Manana /* 34428039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34438039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34448039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34458039d87dSFilipe Manana */ 34468039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34478039d87dSFilipe Manana struct btrfs_file_extent_item); 34488039d87dSFilipe Manana /* 34498039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34508039d87dSFilipe Manana * following it. 34518039d87dSFilipe Manana */ 34528039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34538039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34548039d87dSFilipe Manana goto copy_inline_extent; 34558039d87dSFilipe Manana 34568039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34578039d87dSFilipe Manana if (ext_len > aligned_end) 34588039d87dSFilipe Manana return -EOPNOTSUPP; 34598039d87dSFilipe Manana 34608039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34618039d87dSFilipe Manana if (ret < 0) { 34628039d87dSFilipe Manana return ret; 34638039d87dSFilipe Manana } else if (ret == 0) { 34648039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34658039d87dSFilipe Manana path->slots[0]); 34664a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34678039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34688039d87dSFilipe Manana return -EOPNOTSUPP; 34698039d87dSFilipe Manana } 34708039d87dSFilipe Manana } 34718039d87dSFilipe Manana 34728039d87dSFilipe Manana copy_inline_extent: 34738039d87dSFilipe Manana /* 34748039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34758039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34768039d87dSFilipe Manana */ 34778039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34788039d87dSFilipe Manana /* 34798039d87dSFilipe Manana * If the destination inode has an inline extent... 34808039d87dSFilipe Manana * This would require copying the data from the source inline 34818039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34828039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34838039d87dSFilipe Manana * or just one of them, which would require decompressing and 34848039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34858039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34868039d87dSFilipe Manana * inline extent). 34878039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34888039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34898039d87dSFilipe Manana */ 34908039d87dSFilipe Manana return -EOPNOTSUPP; 34918039d87dSFilipe Manana } 34928039d87dSFilipe Manana 34938039d87dSFilipe Manana btrfs_release_path(path); 34948039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34958039d87dSFilipe Manana if (ret) 34968039d87dSFilipe Manana return ret; 34978039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34988039d87dSFilipe Manana if (ret) 34998039d87dSFilipe Manana return ret; 35008039d87dSFilipe Manana 35018039d87dSFilipe Manana if (skip) { 35028039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 35038039d87dSFilipe Manana 35048039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 35058039d87dSFilipe Manana } 35068039d87dSFilipe Manana 35078039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 35088039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 35098039d87dSFilipe Manana path->slots[0]), 35108039d87dSFilipe Manana size); 35118039d87dSFilipe Manana inode_add_bytes(dst, datal); 35128039d87dSFilipe Manana 35138039d87dSFilipe Manana return 0; 35148039d87dSFilipe Manana } 35158039d87dSFilipe Manana 351632b7c687SMark Fasheh /** 351732b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 351832b7c687SMark Fasheh * 351932b7c687SMark Fasheh * @src: Inode to clone from 352032b7c687SMark Fasheh * @inode: Inode to clone to 352132b7c687SMark Fasheh * @off: Offset within source to start clone from 352232b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 35231c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 352432b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 35251c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 352632b7c687SMark Fasheh */ 352732b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3528f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 35291c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3530f46b5a66SChristoph Hellwig { 35310b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3532f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 353332b7c687SMark Fasheh struct btrfs_path *path = NULL; 3534f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 353532b7c687SMark Fasheh struct btrfs_trans_handle *trans; 353632b7c687SMark Fasheh char *buf = NULL; 3537ae01a0abSYan Zheng struct btrfs_key key; 3538f46b5a66SChristoph Hellwig u32 nritems; 3539f46b5a66SChristoph Hellwig int slot; 3540ae01a0abSYan Zheng int ret; 3541f82a9901SFilipe Manana const u64 len = olen_aligned; 3542f82a9901SFilipe Manana u64 last_dest_end = destoff; 3543ae01a0abSYan Zheng 3544ae01a0abSYan Zheng ret = -ENOMEM; 3545752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3546ae01a0abSYan Zheng if (!buf) 354732b7c687SMark Fasheh return ret; 3548ae01a0abSYan Zheng 3549ae01a0abSYan Zheng path = btrfs_alloc_path(); 3550ae01a0abSYan Zheng if (!path) { 355115351955SDavid Sterba kvfree(buf); 355232b7c687SMark Fasheh return ret; 3553ae01a0abSYan Zheng } 355432b7c687SMark Fasheh 3555e4058b54SDavid Sterba path->reada = READA_FORWARD; 3556c5c9cd4dSSage Weil /* clone data */ 35574a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3558ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35592c463823SFilipe Manana key.offset = off; 3560f46b5a66SChristoph Hellwig 3561f46b5a66SChristoph Hellwig while (1) { 3562de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3563df858e76SFilipe Manana 3564f46b5a66SChristoph Hellwig /* 3565f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3566f46b5a66SChristoph Hellwig * tree. 3567f46b5a66SChristoph Hellwig */ 3568e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3569362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3570362a20c5SDavid Sterba 0, 0); 3571f46b5a66SChristoph Hellwig if (ret < 0) 3572f46b5a66SChristoph Hellwig goto out; 35732c463823SFilipe Manana /* 35742c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35752c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35762c463823SFilipe Manana * it might overlap our target range, therefore process it. 35772c463823SFilipe Manana */ 35782c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35792c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35802c463823SFilipe Manana path->slots[0] - 1); 35812c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35822c463823SFilipe Manana path->slots[0]--; 35832c463823SFilipe Manana } 3584f46b5a66SChristoph Hellwig 3585ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3586e4355f34SFilipe David Borba Manana process_slot: 3587ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3588362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3589f46b5a66SChristoph Hellwig if (ret < 0) 3590f46b5a66SChristoph Hellwig goto out; 3591f46b5a66SChristoph Hellwig if (ret > 0) 3592f46b5a66SChristoph Hellwig break; 3593ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3594f46b5a66SChristoph Hellwig } 3595f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3596f46b5a66SChristoph Hellwig slot = path->slots[0]; 3597f46b5a66SChristoph Hellwig 3598ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3599962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 36004a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3601f46b5a66SChristoph Hellwig break; 3602f46b5a66SChristoph Hellwig 3603962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3604c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3605c5c9cd4dSSage Weil int type; 360631840ae1SZheng Yan u32 size; 360731840ae1SZheng Yan struct btrfs_key new_key; 3608c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3609c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3610c5c9cd4dSSage Weil u8 comp; 3611f82a9901SFilipe Manana u64 drop_start; 361231840ae1SZheng Yan 3613c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3614c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3615c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3616c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3617c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3618c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3619d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3620d397712bSChris Mason extent); 3621d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3622d397712bSChris Mason extent); 3623c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3624d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3625d397712bSChris Mason extent); 3626c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3627c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3628c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3629c5c9cd4dSSage Weil extent); 3630c5c9cd4dSSage Weil } 363131840ae1SZheng Yan 36322c463823SFilipe Manana /* 36332c463823SFilipe Manana * The first search might have left us at an extent 36342c463823SFilipe Manana * item that ends before our target range's start, can 36352c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 36362c463823SFilipe Manana */ 36372c463823SFilipe Manana if (key.offset + datal <= off) { 3638e4355f34SFilipe David Borba Manana path->slots[0]++; 3639e4355f34SFilipe David Borba Manana goto process_slot; 36402c463823SFilipe Manana } else if (key.offset >= off + len) { 36412c463823SFilipe Manana break; 3642e4355f34SFilipe David Borba Manana } 3643df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3644e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3645e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3646e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3647e4355f34SFilipe David Borba Manana size); 3648e4355f34SFilipe David Borba Manana 3649e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3650e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3651c5c9cd4dSSage Weil 365231840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36534a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36544d728ec7SLi Zefan if (off <= key.offset) 3655c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36564d728ec7SLi Zefan else 36574d728ec7SLi Zefan new_key.offset = destoff; 3658c5c9cd4dSSage Weil 3659b6f3409bSSage Weil /* 3660f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3661f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3662f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3663f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3664f82a9901SFilipe Manana * partially overlaps it). 3665f82a9901SFilipe Manana */ 3666f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3667f82a9901SFilipe Manana drop_start = last_dest_end; 3668f82a9901SFilipe Manana else 3669f82a9901SFilipe Manana drop_start = new_key.offset; 3670f82a9901SFilipe Manana 3671f82a9901SFilipe Manana /* 3672b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3673b6f3409bSSage Weil * 1 - add new extent 3674b6f3409bSSage Weil * 1 - inode update 3675b6f3409bSSage Weil */ 3676b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3677a22285a6SYan, Zheng if (IS_ERR(trans)) { 3678a22285a6SYan, Zheng ret = PTR_ERR(trans); 3679a22285a6SYan, Zheng goto out; 3680a22285a6SYan, Zheng } 3681a22285a6SYan, Zheng 3682c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3683c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3684d72c0842SLi Zefan /* 3685d72c0842SLi Zefan * a | --- range to clone ---| b 3686d72c0842SLi Zefan * | ------------- extent ------------- | 3687d72c0842SLi Zefan */ 3688d72c0842SLi Zefan 368993915584SAntonio Ospite /* subtract range b */ 3690d72c0842SLi Zefan if (key.offset + datal > off + len) 3691d72c0842SLi Zefan datal = off + len - key.offset; 3692d72c0842SLi Zefan 369393915584SAntonio Ospite /* subtract range a */ 3694a22285a6SYan, Zheng if (off > key.offset) { 3695a22285a6SYan, Zheng datao += off - key.offset; 3696a22285a6SYan, Zheng datal -= off - key.offset; 3697a22285a6SYan, Zheng } 3698a22285a6SYan, Zheng 36995dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3700f82a9901SFilipe Manana drop_start, 3701a22285a6SYan, Zheng new_key.offset + datal, 37022671485dSJosef Bacik 1); 370379787eaaSJeff Mahoney if (ret) { 37043f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 370500fdf13aSLiu Bo btrfs_abort_transaction(trans, 370666642832SJeff Mahoney ret); 37073a45bb20SJeff Mahoney btrfs_end_transaction(trans); 370879787eaaSJeff Mahoney goto out; 370979787eaaSJeff Mahoney } 3710a22285a6SYan, Zheng 371131840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 371231840ae1SZheng Yan &new_key, size); 371379787eaaSJeff Mahoney if (ret) { 371466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 37153a45bb20SJeff Mahoney btrfs_end_transaction(trans); 371679787eaaSJeff Mahoney goto out; 371779787eaaSJeff Mahoney } 371831840ae1SZheng Yan 371931840ae1SZheng Yan leaf = path->nodes[0]; 372031840ae1SZheng Yan slot = path->slots[0]; 372131840ae1SZheng Yan write_extent_buffer(leaf, buf, 372231840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 372331840ae1SZheng Yan size); 3724ae01a0abSYan Zheng 3725f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3726f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3727c5c9cd4dSSage Weil 3728c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3729c5c9cd4dSSage Weil if (!disko) 3730c5c9cd4dSSage Weil datao = 0; 3731c5c9cd4dSSage Weil 3732c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3733c5c9cd4dSSage Weil datao); 3734c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3735c5c9cd4dSSage Weil datal); 3736fcebe456SJosef Bacik 3737c5c9cd4dSSage Weil if (disko) { 3738c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 37392ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 374084f7d8e6SJosef Bacik root, 37415d4f98a2SYan Zheng disko, diskl, 0, 3742f46b5a66SChristoph Hellwig root->root_key.objectid, 37434a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 3744b06c4bf5SFilipe Manana new_key.offset - datao); 374579787eaaSJeff Mahoney if (ret) { 374679787eaaSJeff Mahoney btrfs_abort_transaction(trans, 374779787eaaSJeff Mahoney ret); 37483a45bb20SJeff Mahoney btrfs_end_transaction(trans); 374979787eaaSJeff Mahoney goto out; 375079787eaaSJeff Mahoney 375179787eaaSJeff Mahoney } 3752f46b5a66SChristoph Hellwig } 3753c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3754c5c9cd4dSSage Weil u64 skip = 0; 3755c5c9cd4dSSage Weil u64 trim = 0; 3756ed958762SFilipe Manana 3757c5c9cd4dSSage Weil if (off > key.offset) { 3758c5c9cd4dSSage Weil skip = off - key.offset; 3759c5c9cd4dSSage Weil new_key.offset += skip; 376031840ae1SZheng Yan } 3761d397712bSChris Mason 3762c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3763c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3764d397712bSChris Mason 3765c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3766c5c9cd4dSSage Weil ret = -EINVAL; 37673a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3768c5c9cd4dSSage Weil goto out; 376931840ae1SZheng Yan } 3770c5c9cd4dSSage Weil size -= skip + trim; 3771c5c9cd4dSSage Weil datal -= skip + trim; 3772a22285a6SYan, Zheng 37734a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 37748039d87dSFilipe Manana trans, path, 37758039d87dSFilipe Manana &new_key, 3776f82a9901SFilipe Manana drop_start, 37778039d87dSFilipe Manana datal, 37788039d87dSFilipe Manana skip, size, buf); 377979787eaaSJeff Mahoney if (ret) { 37803f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37813a29bc09SChris Mason btrfs_abort_transaction(trans, 378279787eaaSJeff Mahoney ret); 37833a45bb20SJeff Mahoney btrfs_end_transaction(trans); 378479787eaaSJeff Mahoney goto out; 378579787eaaSJeff Mahoney } 3786c5c9cd4dSSage Weil leaf = path->nodes[0]; 3787c5c9cd4dSSage Weil slot = path->slots[0]; 3788c5c9cd4dSSage Weil } 3789c5c9cd4dSSage Weil 37907ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 37917ffbb598SFilipe Manana if (drop_start < new_key.offset) 3792a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 379314f59796SFilipe Manana NULL, drop_start, 37947ffbb598SFilipe Manana new_key.offset - drop_start); 37957ffbb598SFilipe Manana 3796a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 3797a2f392e4SNikolay Borisov path, 0, 0); 37987ffbb598SFilipe Manana 3799c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3800b3b4aa74SDavid Sterba btrfs_release_path(path); 3801c5c9cd4dSSage Weil 380262e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 38030b246afaSJeff Mahoney fs_info->sectorsize); 3804f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3805f82a9901SFilipe Manana last_dest_end, 38061c919a5eSMark Fasheh destoff, olen, 38071c919a5eSMark Fasheh no_time_update); 3808f82a9901SFilipe Manana if (ret) 380979787eaaSJeff Mahoney goto out; 38102c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 38112c463823SFilipe Manana break; 3812a22285a6SYan, Zheng } 3813b3b4aa74SDavid Sterba btrfs_release_path(path); 3814df858e76SFilipe Manana key.offset = next_key_min_offset; 381569ae5e44SWang Xiaoguang 381669ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 381769ae5e44SWang Xiaoguang ret = -EINTR; 381869ae5e44SWang Xiaoguang goto out; 381969ae5e44SWang Xiaoguang } 3820ae01a0abSYan Zheng } 3821f46b5a66SChristoph Hellwig ret = 0; 382232b7c687SMark Fasheh 3823f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3824f82a9901SFilipe Manana /* 3825f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3826f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3827f82a9901SFilipe Manana */ 3828b3b4aa74SDavid Sterba btrfs_release_path(path); 3829f82a9901SFilipe Manana 3830f82a9901SFilipe Manana /* 3831f82a9901SFilipe Manana * 1 - remove extent(s) 3832f82a9901SFilipe Manana * 1 - inode update 3833f82a9901SFilipe Manana */ 3834f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3835f82a9901SFilipe Manana if (IS_ERR(trans)) { 3836f82a9901SFilipe Manana ret = PTR_ERR(trans); 3837f82a9901SFilipe Manana goto out; 3838f82a9901SFilipe Manana } 3839f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3840f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3841f82a9901SFilipe Manana if (ret) { 3842f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 384366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 38443a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3845f82a9901SFilipe Manana goto out; 3846f82a9901SFilipe Manana } 3847a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 3848a2f392e4SNikolay Borisov last_dest_end, 384914f59796SFilipe Manana destoff + len - last_dest_end); 3850f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38511c919a5eSMark Fasheh destoff, olen, no_time_update); 3852f82a9901SFilipe Manana } 3853f82a9901SFilipe Manana 3854f46b5a66SChristoph Hellwig out: 385532b7c687SMark Fasheh btrfs_free_path(path); 385615351955SDavid Sterba kvfree(buf); 385732b7c687SMark Fasheh return ret; 385832b7c687SMark Fasheh } 385932b7c687SMark Fasheh 38603db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 386132b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 386232b7c687SMark Fasheh { 386354563d41SAl Viro struct inode *inode = file_inode(file); 38643db11b2eSZach Brown struct inode *src = file_inode(file_src); 38650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 386632b7c687SMark Fasheh int ret; 386732b7c687SMark Fasheh u64 len = olen; 38680b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 386932b7c687SMark Fasheh 387032b7c687SMark Fasheh /* 387132b7c687SMark Fasheh * TODO: 387232b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 387332b7c687SMark Fasheh * decompress into destination's address_space (the file offset 387432b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 387532b7c687SMark Fasheh * otherwise reinsert) a subrange. 387600fdf13aSLiu Bo * 387700fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 387800fdf13aSLiu Bo * be either compressed or non-compressed. 387932b7c687SMark Fasheh */ 388032b7c687SMark Fasheh 3881ac765f83SFilipe Manana /* 388234a28e3dSFilipe Manana * VFS's generic_remap_file_range_prep() protects us from cloning the 388334a28e3dSFilipe Manana * eof block into the middle of a file, which would result in corruption 388434a28e3dSFilipe Manana * if the file size is not blocksize aligned. So we don't need to check 388534a28e3dSFilipe Manana * for that case here. 3886ac765f83SFilipe Manana */ 388734a28e3dSFilipe Manana if (off + len == src->i_size) 388832b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 388932b7c687SMark Fasheh 389032b7c687SMark Fasheh if (destoff > inode->i_size) { 3891f7fa1107SFilipe Manana const u64 wb_start = ALIGN_DOWN(inode->i_size, bs); 3892f7fa1107SFilipe Manana 389332b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 389432b7c687SMark Fasheh if (ret) 389534a28e3dSFilipe Manana return ret; 3896f7fa1107SFilipe Manana /* 3897f7fa1107SFilipe Manana * We may have truncated the last block if the inode's size is 3898f7fa1107SFilipe Manana * not sector size aligned, so we need to wait for writeback to 3899f7fa1107SFilipe Manana * complete before proceeding further, otherwise we can race 3900f7fa1107SFilipe Manana * with cloning and attempt to increment a reference to an 3901f7fa1107SFilipe Manana * extent that no longer exists (writeback completed right after 3902f7fa1107SFilipe Manana * we found the previous extent covering eof and before we 3903f7fa1107SFilipe Manana * attempted to increment its reference count). 3904f7fa1107SFilipe Manana */ 3905f7fa1107SFilipe Manana ret = btrfs_wait_ordered_range(inode, wb_start, 3906f7fa1107SFilipe Manana destoff - wb_start); 3907f7fa1107SFilipe Manana if (ret) 3908f7fa1107SFilipe Manana return ret; 390932b7c687SMark Fasheh } 391032b7c687SMark Fasheh 3911c125b8bfSFilipe Manana /* 3912d8b55242SFilipe Manana * Lock destination range to serialize with concurrent readpages() and 3913d8b55242SFilipe Manana * source range to serialize with relocation. 3914c125b8bfSFilipe Manana */ 3915d8b55242SFilipe Manana btrfs_double_extent_lock(src, off, inode, destoff, len); 39161c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 3917d8b55242SFilipe Manana btrfs_double_extent_unlock(src, off, inode, destoff, len); 3918c125b8bfSFilipe Manana /* 3919c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3920c125b8bfSFilipe Manana * data immediately and not the previous data. 3921c125b8bfSFilipe Manana */ 392265bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 392309cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 392409cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 392534a28e3dSFilipe Manana 392634a28e3dSFilipe Manana return ret; 392734a28e3dSFilipe Manana } 392834a28e3dSFilipe Manana 392934a28e3dSFilipe Manana static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in, 393034a28e3dSFilipe Manana struct file *file_out, loff_t pos_out, 393134a28e3dSFilipe Manana loff_t *len, unsigned int remap_flags) 393234a28e3dSFilipe Manana { 393334a28e3dSFilipe Manana struct inode *inode_in = file_inode(file_in); 393434a28e3dSFilipe Manana struct inode *inode_out = file_inode(file_out); 393534a28e3dSFilipe Manana u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize; 393634a28e3dSFilipe Manana bool same_inode = inode_out == inode_in; 393734a28e3dSFilipe Manana u64 wb_len; 393834a28e3dSFilipe Manana int ret; 393934a28e3dSFilipe Manana 394034a28e3dSFilipe Manana if (!(remap_flags & REMAP_FILE_DEDUP)) { 394134a28e3dSFilipe Manana struct btrfs_root *root_out = BTRFS_I(inode_out)->root; 394234a28e3dSFilipe Manana 394334a28e3dSFilipe Manana if (btrfs_root_readonly(root_out)) 394434a28e3dSFilipe Manana return -EROFS; 394534a28e3dSFilipe Manana 394634a28e3dSFilipe Manana if (file_in->f_path.mnt != file_out->f_path.mnt || 394734a28e3dSFilipe Manana inode_in->i_sb != inode_out->i_sb) 394834a28e3dSFilipe Manana return -EXDEV; 394934a28e3dSFilipe Manana } 395034a28e3dSFilipe Manana 395134a28e3dSFilipe Manana if (same_inode) 395234a28e3dSFilipe Manana inode_lock(inode_in); 3953293a8489SMark Fasheh else 39544ea748e1SFilipe Manana lock_two_nondirectories(inode_in, inode_out); 395534a28e3dSFilipe Manana 3956500710d3SFilipe Manana /* don't make the dst file partly checksummed */ 3957500710d3SFilipe Manana if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) != 3958500710d3SFilipe Manana (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) { 3959500710d3SFilipe Manana ret = -EINVAL; 3960500710d3SFilipe Manana goto out_unlock; 3961500710d3SFilipe Manana } 3962500710d3SFilipe Manana 396334a28e3dSFilipe Manana /* 396434a28e3dSFilipe Manana * Now that the inodes are locked, we need to start writeback ourselves 396534a28e3dSFilipe Manana * and can not rely on the writeback from the VFS's generic helper 396634a28e3dSFilipe Manana * generic_remap_file_range_prep() because: 396734a28e3dSFilipe Manana * 396834a28e3dSFilipe Manana * 1) For compression we must call filemap_fdatawrite_range() range 396934a28e3dSFilipe Manana * twice (btrfs_fdatawrite_range() does it for us), and the generic 397034a28e3dSFilipe Manana * helper only calls it once; 397134a28e3dSFilipe Manana * 397234a28e3dSFilipe Manana * 2) filemap_fdatawrite_range(), called by the generic helper only 397334a28e3dSFilipe Manana * waits for the writeback to complete, i.e. for IO to be done, and 397434a28e3dSFilipe Manana * not for the ordered extents to complete. We need to wait for them 397534a28e3dSFilipe Manana * to complete so that new file extent items are in the fs tree. 397634a28e3dSFilipe Manana */ 397734a28e3dSFilipe Manana if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP)) 397834a28e3dSFilipe Manana wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs); 397934a28e3dSFilipe Manana else 398034a28e3dSFilipe Manana wb_len = ALIGN(*len, bs); 398134a28e3dSFilipe Manana 398234a28e3dSFilipe Manana /* 398334a28e3dSFilipe Manana * Since we don't lock ranges, wait for ongoing lockless dio writes (as 398434a28e3dSFilipe Manana * any in progress could create its ordered extents after we wait for 398534a28e3dSFilipe Manana * existing ordered extents below). 398634a28e3dSFilipe Manana */ 398734a28e3dSFilipe Manana inode_dio_wait(inode_in); 398834a28e3dSFilipe Manana if (!same_inode) 398934a28e3dSFilipe Manana inode_dio_wait(inode_out); 399034a28e3dSFilipe Manana 399134a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs), 399234a28e3dSFilipe Manana wb_len); 399334a28e3dSFilipe Manana if (ret < 0) 399434a28e3dSFilipe Manana goto out_unlock; 399534a28e3dSFilipe Manana ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs), 399634a28e3dSFilipe Manana wb_len); 399734a28e3dSFilipe Manana if (ret < 0) 399834a28e3dSFilipe Manana goto out_unlock; 399934a28e3dSFilipe Manana 400034a28e3dSFilipe Manana ret = generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out, 400134a28e3dSFilipe Manana len, remap_flags); 400234a28e3dSFilipe Manana if (ret < 0 || *len == 0) 400334a28e3dSFilipe Manana goto out_unlock; 400434a28e3dSFilipe Manana 400534a28e3dSFilipe Manana return 0; 400634a28e3dSFilipe Manana 400734a28e3dSFilipe Manana out_unlock: 400834a28e3dSFilipe Manana if (same_inode) 400934a28e3dSFilipe Manana inode_unlock(inode_in); 401034a28e3dSFilipe Manana else 40114ea748e1SFilipe Manana unlock_two_nondirectories(inode_in, inode_out); 401234a28e3dSFilipe Manana 40133db11b2eSZach Brown return ret; 40143db11b2eSZach Brown } 40153db11b2eSZach Brown 401642ec3d4cSDarrick J. Wong loff_t btrfs_remap_file_range(struct file *src_file, loff_t off, 401742ec3d4cSDarrick J. Wong struct file *dst_file, loff_t destoff, loff_t len, 40182e5dfc99SDarrick J. Wong unsigned int remap_flags) 40193db11b2eSZach Brown { 402034a28e3dSFilipe Manana struct inode *src_inode = file_inode(src_file); 402134a28e3dSFilipe Manana struct inode *dst_inode = file_inode(dst_file); 402234a28e3dSFilipe Manana bool same_inode = dst_inode == src_inode; 402342ec3d4cSDarrick J. Wong int ret; 402442ec3d4cSDarrick J. Wong 40252e5dfc99SDarrick J. Wong if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY)) 40262e5dfc99SDarrick J. Wong return -EINVAL; 40272e5dfc99SDarrick J. Wong 402834a28e3dSFilipe Manana ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff, 402934a28e3dSFilipe Manana &len, remap_flags); 403034a28e3dSFilipe Manana if (ret < 0 || len == 0) 403134a28e3dSFilipe Manana return ret; 40322e5dfc99SDarrick J. Wong 403334a28e3dSFilipe Manana if (remap_flags & REMAP_FILE_DEDUP) 403434a28e3dSFilipe Manana ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff); 403534a28e3dSFilipe Manana else 403642ec3d4cSDarrick J. Wong ret = btrfs_clone_files(dst_file, src_file, off, len, destoff); 403734a28e3dSFilipe Manana 403834a28e3dSFilipe Manana if (same_inode) 403934a28e3dSFilipe Manana inode_unlock(src_inode); 404034a28e3dSFilipe Manana else 40414ea748e1SFilipe Manana unlock_two_nondirectories(src_inode, dst_inode); 404234a28e3dSFilipe Manana 404342ec3d4cSDarrick J. Wong return ret < 0 ? ret : len; 4044c5c9cd4dSSage Weil } 4045c5c9cd4dSSage Weil 40466ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40476ef5ed0dSJosef Bacik { 4048496ad9aaSAl Viro struct inode *inode = file_inode(file); 40490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 40506ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40516ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40526ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40536ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40546ef5ed0dSJosef Bacik struct btrfs_path *path; 40556ef5ed0dSJosef Bacik struct btrfs_key location; 40566ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40576ef5ed0dSJosef Bacik u64 objectid = 0; 40586ef5ed0dSJosef Bacik u64 dir_id; 40593c04ce01SMiao Xie int ret; 40606ef5ed0dSJosef Bacik 40616ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40626ef5ed0dSJosef Bacik return -EPERM; 40636ef5ed0dSJosef Bacik 40643c04ce01SMiao Xie ret = mnt_want_write_file(file); 40653c04ce01SMiao Xie if (ret) 40663c04ce01SMiao Xie return ret; 40673c04ce01SMiao Xie 40683c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40693c04ce01SMiao Xie ret = -EFAULT; 40703c04ce01SMiao Xie goto out; 40713c04ce01SMiao Xie } 40726ef5ed0dSJosef Bacik 40736ef5ed0dSJosef Bacik if (!objectid) 40741cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40756ef5ed0dSJosef Bacik 40766ef5ed0dSJosef Bacik location.objectid = objectid; 40776ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40786ef5ed0dSJosef Bacik location.offset = (u64)-1; 40796ef5ed0dSJosef Bacik 40800b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 40813c04ce01SMiao Xie if (IS_ERR(new_root)) { 40823c04ce01SMiao Xie ret = PTR_ERR(new_root); 40833c04ce01SMiao Xie goto out; 40843c04ce01SMiao Xie } 40854fd786e6SMisono Tomohiro if (!is_fstree(new_root->root_key.objectid)) { 40866d6d2829Ssatoru takeuchi ret = -ENOENT; 40876d6d2829Ssatoru takeuchi goto out; 40886d6d2829Ssatoru takeuchi } 40896ef5ed0dSJosef Bacik 40906ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40913c04ce01SMiao Xie if (!path) { 40923c04ce01SMiao Xie ret = -ENOMEM; 40933c04ce01SMiao Xie goto out; 40943c04ce01SMiao Xie } 40956ef5ed0dSJosef Bacik path->leave_spinning = 1; 40966ef5ed0dSJosef Bacik 40976ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 409898d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40996ef5ed0dSJosef Bacik btrfs_free_path(path); 41003c04ce01SMiao Xie ret = PTR_ERR(trans); 41013c04ce01SMiao Xie goto out; 41026ef5ed0dSJosef Bacik } 41036ef5ed0dSJosef Bacik 41040b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 41050b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 41066ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4107cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 41086ef5ed0dSJosef Bacik btrfs_free_path(path); 41093a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41100b246afaSJeff Mahoney btrfs_err(fs_info, 41115d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 41123c04ce01SMiao Xie ret = -ENOENT; 41133c04ce01SMiao Xie goto out; 41146ef5ed0dSJosef Bacik } 41156ef5ed0dSJosef Bacik 41166ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 41176ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 41186ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 41196ef5ed0dSJosef Bacik btrfs_free_path(path); 41206ef5ed0dSJosef Bacik 41210b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 41223a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41233c04ce01SMiao Xie out: 41243c04ce01SMiao Xie mnt_drop_write_file(file); 41253c04ce01SMiao Xie return ret; 41266ef5ed0dSJosef Bacik } 41276ef5ed0dSJosef Bacik 4128c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4129bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4130bf5fc093SJosef Bacik { 4131bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4132bf5fc093SJosef Bacik 4133bf5fc093SJosef Bacik space->total_bytes = 0; 4134bf5fc093SJosef Bacik space->used_bytes = 0; 4135bf5fc093SJosef Bacik space->flags = 0; 4136bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4137bf5fc093SJosef Bacik space->flags = block_group->flags; 4138bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4139bf5fc093SJosef Bacik space->used_bytes += 4140bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4141bf5fc093SJosef Bacik } 4142bf5fc093SJosef Bacik } 4143bf5fc093SJosef Bacik 41442ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 41452ff7e61eSJeff Mahoney void __user *arg) 41461406e432SJosef Bacik { 41471406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41481406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41491406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41507fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 415113f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41521406e432SJosef Bacik struct btrfs_space_info *info; 4153315d8e98SColin Ian King static const u64 types[] = { 4154315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4155bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4156bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4157315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4158315d8e98SColin Ian King }; 4159bf5fc093SJosef Bacik int num_types = 4; 41607fde62bfSChris Mason int alloc_size; 41611406e432SJosef Bacik int ret = 0; 416251788b1bSDan Rosenberg u64 slot_count = 0; 4163bf5fc093SJosef Bacik int i, c; 41641406e432SJosef Bacik 41651406e432SJosef Bacik if (copy_from_user(&space_args, 41661406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41671406e432SJosef Bacik sizeof(space_args))) 41681406e432SJosef Bacik return -EFAULT; 41691406e432SJosef Bacik 4170bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4171bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4172bf5fc093SJosef Bacik 4173bf5fc093SJosef Bacik info = NULL; 41747fde62bfSChris Mason rcu_read_lock(); 41750b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4176bf5fc093SJosef Bacik list) { 4177bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4178bf5fc093SJosef Bacik info = tmp; 4179bf5fc093SJosef Bacik break; 4180bf5fc093SJosef Bacik } 4181bf5fc093SJosef Bacik } 41827fde62bfSChris Mason rcu_read_unlock(); 41831406e432SJosef Bacik 4184bf5fc093SJosef Bacik if (!info) 4185bf5fc093SJosef Bacik continue; 4186bf5fc093SJosef Bacik 4187bf5fc093SJosef Bacik down_read(&info->groups_sem); 4188bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4189bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4190bf5fc093SJosef Bacik slot_count++; 4191bf5fc093SJosef Bacik } 4192bf5fc093SJosef Bacik up_read(&info->groups_sem); 4193bf5fc093SJosef Bacik } 4194bf5fc093SJosef Bacik 419536523e95SDavid Sterba /* 419636523e95SDavid Sterba * Global block reserve, exported as a space_info 419736523e95SDavid Sterba */ 419836523e95SDavid Sterba slot_count++; 419936523e95SDavid Sterba 42007fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 42017fde62bfSChris Mason if (space_args.space_slots == 0) { 42027fde62bfSChris Mason space_args.total_spaces = slot_count; 42037fde62bfSChris Mason goto out; 42047fde62bfSChris Mason } 4205bf5fc093SJosef Bacik 420651788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4207bf5fc093SJosef Bacik 42087fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4209bf5fc093SJosef Bacik 42107fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 42117fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 42127fde62bfSChris Mason */ 421309cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 42147fde62bfSChris Mason return -ENOMEM; 42157fde62bfSChris Mason 42167fde62bfSChris Mason space_args.total_spaces = 0; 42178d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 42187fde62bfSChris Mason if (!dest) 42197fde62bfSChris Mason return -ENOMEM; 42207fde62bfSChris Mason dest_orig = dest; 42217fde62bfSChris Mason 42227fde62bfSChris Mason /* now we have a buffer to copy into */ 4223bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4224bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4225bf5fc093SJosef Bacik 422651788b1bSDan Rosenberg if (!slot_count) 422751788b1bSDan Rosenberg break; 422851788b1bSDan Rosenberg 4229bf5fc093SJosef Bacik info = NULL; 42301406e432SJosef Bacik rcu_read_lock(); 42310b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4232bf5fc093SJosef Bacik list) { 4233bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4234bf5fc093SJosef Bacik info = tmp; 42357fde62bfSChris Mason break; 4236bf5fc093SJosef Bacik } 4237bf5fc093SJosef Bacik } 4238bf5fc093SJosef Bacik rcu_read_unlock(); 42397fde62bfSChris Mason 4240bf5fc093SJosef Bacik if (!info) 4241bf5fc093SJosef Bacik continue; 4242bf5fc093SJosef Bacik down_read(&info->groups_sem); 4243bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4244bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4245c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4246c065f5b1SSu Yue &space); 42477fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42481406e432SJosef Bacik dest++; 42491406e432SJosef Bacik space_args.total_spaces++; 425051788b1bSDan Rosenberg slot_count--; 42511406e432SJosef Bacik } 425251788b1bSDan Rosenberg if (!slot_count) 425351788b1bSDan Rosenberg break; 4254bf5fc093SJosef Bacik } 4255bf5fc093SJosef Bacik up_read(&info->groups_sem); 4256bf5fc093SJosef Bacik } 42571406e432SJosef Bacik 425836523e95SDavid Sterba /* 425936523e95SDavid Sterba * Add global block reserve 426036523e95SDavid Sterba */ 426136523e95SDavid Sterba if (slot_count) { 42620b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 426336523e95SDavid Sterba 426436523e95SDavid Sterba spin_lock(&block_rsv->lock); 426536523e95SDavid Sterba space.total_bytes = block_rsv->size; 426636523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 426736523e95SDavid Sterba spin_unlock(&block_rsv->lock); 426836523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 426936523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 427036523e95SDavid Sterba space_args.total_spaces++; 427136523e95SDavid Sterba } 427236523e95SDavid Sterba 42732eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42747fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42757fde62bfSChris Mason 42767fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42777fde62bfSChris Mason ret = -EFAULT; 42787fde62bfSChris Mason 42797fde62bfSChris Mason kfree(dest_orig); 42807fde62bfSChris Mason out: 42817fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42821406e432SJosef Bacik ret = -EFAULT; 42831406e432SJosef Bacik 42841406e432SJosef Bacik return ret; 42851406e432SJosef Bacik } 42861406e432SJosef Bacik 42879a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42889a8c28beSMiao Xie void __user *argp) 428946204592SSage Weil { 429046204592SSage Weil struct btrfs_trans_handle *trans; 429146204592SSage Weil u64 transid; 4292db5b493aSTsutomu Itoh int ret; 429346204592SSage Weil 4294d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4295ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4296ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 429798d5dc13STsutomu Itoh return PTR_ERR(trans); 4298ff7c1d33SMiao Xie 4299ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4300ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4301ff7c1d33SMiao Xie goto out; 4302ff7c1d33SMiao Xie } 430346204592SSage Weil transid = trans->transid; 43043a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 43058b2b2d3cSTsutomu Itoh if (ret) { 43063a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4307db5b493aSTsutomu Itoh return ret; 43088b2b2d3cSTsutomu Itoh } 4309ff7c1d33SMiao Xie out: 431046204592SSage Weil if (argp) 431146204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 431246204592SSage Weil return -EFAULT; 431346204592SSage Weil return 0; 431446204592SSage Weil } 431546204592SSage Weil 43162ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 43179a8c28beSMiao Xie void __user *argp) 431846204592SSage Weil { 431946204592SSage Weil u64 transid; 432046204592SSage Weil 432146204592SSage Weil if (argp) { 432246204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 432346204592SSage Weil return -EFAULT; 432446204592SSage Weil } else { 432546204592SSage Weil transid = 0; /* current trans */ 432646204592SSage Weil } 43272ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 432846204592SSage Weil } 432946204592SSage Weil 4330b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4331475f6387SJan Schmidt { 43320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4333475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4334b8e95489SMiao Xie int ret; 4335475f6387SJan Schmidt 4336475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4337475f6387SJan Schmidt return -EPERM; 4338475f6387SJan Schmidt 4339475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4340475f6387SJan Schmidt if (IS_ERR(sa)) 4341475f6387SJan Schmidt return PTR_ERR(sa); 4342475f6387SJan Schmidt 4343b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4344b8e95489SMiao Xie ret = mnt_want_write_file(file); 4345b8e95489SMiao Xie if (ret) 4346b8e95489SMiao Xie goto out; 4347b8e95489SMiao Xie } 4348b8e95489SMiao Xie 43490b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 435063a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 435163a212abSStefan Behrens 0); 4352475f6387SJan Schmidt 435306fe39abSFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4354475f6387SJan Schmidt ret = -EFAULT; 4355475f6387SJan Schmidt 4356b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4357b8e95489SMiao Xie mnt_drop_write_file(file); 4358b8e95489SMiao Xie out: 4359475f6387SJan Schmidt kfree(sa); 4360475f6387SJan Schmidt return ret; 4361475f6387SJan Schmidt } 4362475f6387SJan Schmidt 43632ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4364475f6387SJan Schmidt { 4365475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4366475f6387SJan Schmidt return -EPERM; 4367475f6387SJan Schmidt 43682ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4369475f6387SJan Schmidt } 4370475f6387SJan Schmidt 43712ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4372475f6387SJan Schmidt void __user *arg) 4373475f6387SJan Schmidt { 4374475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4375475f6387SJan Schmidt int ret; 4376475f6387SJan Schmidt 4377475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4378475f6387SJan Schmidt return -EPERM; 4379475f6387SJan Schmidt 4380475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4381475f6387SJan Schmidt if (IS_ERR(sa)) 4382475f6387SJan Schmidt return PTR_ERR(sa); 4383475f6387SJan Schmidt 43842ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4385475f6387SJan Schmidt 43864fa99b00SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4387475f6387SJan Schmidt ret = -EFAULT; 4388475f6387SJan Schmidt 4389475f6387SJan Schmidt kfree(sa); 4390475f6387SJan Schmidt return ret; 4391475f6387SJan Schmidt } 4392475f6387SJan Schmidt 43932ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4394b27f7c0cSDavid Sterba void __user *arg) 4395c11d2c23SStefan Behrens { 4396c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4397c11d2c23SStefan Behrens int ret; 4398c11d2c23SStefan Behrens 4399c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4400c11d2c23SStefan Behrens if (IS_ERR(sa)) 4401c11d2c23SStefan Behrens return PTR_ERR(sa); 4402c11d2c23SStefan Behrens 4403b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4404b27f7c0cSDavid Sterba kfree(sa); 4405b27f7c0cSDavid Sterba return -EPERM; 4406b27f7c0cSDavid Sterba } 4407b27f7c0cSDavid Sterba 44082ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4409c11d2c23SStefan Behrens 4410eee99577SFilipe Manana if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 4411c11d2c23SStefan Behrens ret = -EFAULT; 4412c11d2c23SStefan Behrens 4413c11d2c23SStefan Behrens kfree(sa); 4414c11d2c23SStefan Behrens return ret; 4415c11d2c23SStefan Behrens } 4416c11d2c23SStefan Behrens 44172ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 44182ff7e61eSJeff Mahoney void __user *arg) 44193f6bcfbdSStefan Behrens { 44203f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 44213f6bcfbdSStefan Behrens int ret; 44223f6bcfbdSStefan Behrens 44233f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44243f6bcfbdSStefan Behrens return -EPERM; 44253f6bcfbdSStefan Behrens 44263f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44273f6bcfbdSStefan Behrens if (IS_ERR(p)) 44283f6bcfbdSStefan Behrens return PTR_ERR(p); 44293f6bcfbdSStefan Behrens 44303f6bcfbdSStefan Behrens switch (p->cmd) { 44313f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4432bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4433adfa97cbSIlya Dryomov ret = -EROFS; 4434adfa97cbSIlya Dryomov goto out; 4435adfa97cbSIlya Dryomov } 4436171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4437e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44383f6bcfbdSStefan Behrens } else { 44392ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4440171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 44413f6bcfbdSStefan Behrens } 44423f6bcfbdSStefan Behrens break; 44433f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44440b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 44453f6bcfbdSStefan Behrens ret = 0; 44463f6bcfbdSStefan Behrens break; 44473f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 444817d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 444997282031SAnand Jain ret = 0; 44503f6bcfbdSStefan Behrens break; 44513f6bcfbdSStefan Behrens default: 44523f6bcfbdSStefan Behrens ret = -EINVAL; 44533f6bcfbdSStefan Behrens break; 44543f6bcfbdSStefan Behrens } 44553f6bcfbdSStefan Behrens 4456d3a53286SFilipe Manana if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 44573f6bcfbdSStefan Behrens ret = -EFAULT; 4458adfa97cbSIlya Dryomov out: 44593f6bcfbdSStefan Behrens kfree(p); 44603f6bcfbdSStefan Behrens return ret; 44613f6bcfbdSStefan Behrens } 44623f6bcfbdSStefan Behrens 4463d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4464d7728c96SJan Schmidt { 4465d7728c96SJan Schmidt int ret = 0; 4466d7728c96SJan Schmidt int i; 4467740c3d22SChris Mason u64 rel_ptr; 4468d7728c96SJan Schmidt int size; 4469806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4470d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4471d7728c96SJan Schmidt struct btrfs_path *path; 4472d7728c96SJan Schmidt 447382b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4474d7728c96SJan Schmidt return -EPERM; 4475d7728c96SJan Schmidt 4476d7728c96SJan Schmidt path = btrfs_alloc_path(); 4477d7728c96SJan Schmidt if (!path) { 4478d7728c96SJan Schmidt ret = -ENOMEM; 4479d7728c96SJan Schmidt goto out; 4480d7728c96SJan Schmidt } 4481d7728c96SJan Schmidt 4482d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4483d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4484d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4485d7728c96SJan Schmidt ipa = NULL; 4486d7728c96SJan Schmidt goto out; 4487d7728c96SJan Schmidt } 4488d7728c96SJan Schmidt 4489d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4490d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4491d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4492d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4493d7728c96SJan Schmidt ipath = NULL; 4494d7728c96SJan Schmidt goto out; 4495d7728c96SJan Schmidt } 4496d7728c96SJan Schmidt 4497d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4498d7728c96SJan Schmidt if (ret < 0) 4499d7728c96SJan Schmidt goto out; 4500d7728c96SJan Schmidt 4501d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4502745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4503745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4504740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4505d7728c96SJan Schmidt } 4506d7728c96SJan Schmidt 4507718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4508718dc5faSOmar Sandoval ipath->fspath, size); 4509d7728c96SJan Schmidt if (ret) { 4510d7728c96SJan Schmidt ret = -EFAULT; 4511d7728c96SJan Schmidt goto out; 4512d7728c96SJan Schmidt } 4513d7728c96SJan Schmidt 4514d7728c96SJan Schmidt out: 4515d7728c96SJan Schmidt btrfs_free_path(path); 4516d7728c96SJan Schmidt free_ipath(ipath); 4517d7728c96SJan Schmidt kfree(ipa); 4518d7728c96SJan Schmidt 4519d7728c96SJan Schmidt return ret; 4520d7728c96SJan Schmidt } 4521d7728c96SJan Schmidt 4522d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4523d7728c96SJan Schmidt { 4524d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4525d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4526d7728c96SJan Schmidt 4527d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4528d7728c96SJan Schmidt inodes->bytes_left -= c; 4529d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4530d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4531d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4532d7728c96SJan Schmidt inodes->elem_cnt += 3; 4533d7728c96SJan Schmidt } else { 4534d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4535d7728c96SJan Schmidt inodes->bytes_left = 0; 4536d7728c96SJan Schmidt inodes->elem_missed += 3; 4537d7728c96SJan Schmidt } 4538d7728c96SJan Schmidt 4539d7728c96SJan Schmidt return 0; 4540d7728c96SJan Schmidt } 4541d7728c96SJan Schmidt 45422ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4543d24a67b2SZygo Blaxell void __user *arg, int version) 4544d7728c96SJan Schmidt { 4545d7728c96SJan Schmidt int ret = 0; 4546d7728c96SJan Schmidt int size; 4547d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4548d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4549d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4550d24a67b2SZygo Blaxell bool ignore_offset; 4551d7728c96SJan Schmidt 4552d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4553d7728c96SJan Schmidt return -EPERM; 4554d7728c96SJan Schmidt 4555d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 45567b9ea627SShailendra Verma if (IS_ERR(loi)) 45577b9ea627SShailendra Verma return PTR_ERR(loi); 4558d7728c96SJan Schmidt 4559d24a67b2SZygo Blaxell if (version == 1) { 4560d24a67b2SZygo Blaxell ignore_offset = false; 4561b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4562d24a67b2SZygo Blaxell } else { 4563d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4564d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4565d24a67b2SZygo Blaxell ret = -EINVAL; 4566d24a67b2SZygo Blaxell goto out_loi; 4567d24a67b2SZygo Blaxell } 4568d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4569d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4570d24a67b2SZygo Blaxell ret = -EINVAL; 4571d24a67b2SZygo Blaxell goto out_loi; 4572d24a67b2SZygo Blaxell } 4573d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4574b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4575d24a67b2SZygo Blaxell } 4576d24a67b2SZygo Blaxell 4577d7728c96SJan Schmidt path = btrfs_alloc_path(); 4578d7728c96SJan Schmidt if (!path) { 4579d7728c96SJan Schmidt ret = -ENOMEM; 4580d7728c96SJan Schmidt goto out; 4581d7728c96SJan Schmidt } 4582d7728c96SJan Schmidt 4583d7728c96SJan Schmidt inodes = init_data_container(size); 4584d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4585d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4586d7728c96SJan Schmidt inodes = NULL; 4587d7728c96SJan Schmidt goto out; 4588d7728c96SJan Schmidt } 4589d7728c96SJan Schmidt 45902ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4591d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4592df031f07SLiu Bo if (ret == -EINVAL) 4593d7728c96SJan Schmidt ret = -ENOENT; 4594d7728c96SJan Schmidt if (ret < 0) 4595d7728c96SJan Schmidt goto out; 4596d7728c96SJan Schmidt 4597718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4598718dc5faSOmar Sandoval size); 4599d7728c96SJan Schmidt if (ret) 4600d7728c96SJan Schmidt ret = -EFAULT; 4601d7728c96SJan Schmidt 4602d7728c96SJan Schmidt out: 4603d7728c96SJan Schmidt btrfs_free_path(path); 4604f54de068SDavid Sterba kvfree(inodes); 4605d24a67b2SZygo Blaxell out_loi: 4606d7728c96SJan Schmidt kfree(loi); 4607d7728c96SJan Schmidt 4608d7728c96SJan Schmidt return ret; 4609d7728c96SJan Schmidt } 4610d7728c96SJan Schmidt 4611008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4612c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4613c9e9f97bSIlya Dryomov { 4614c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4615c9e9f97bSIlya Dryomov 4616c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4617c9e9f97bSIlya Dryomov 46183009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4619837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4620837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4621837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4622a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4623a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4624837d5b6eSIlya Dryomov 4625c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4626c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4627c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 462819a39dceSIlya Dryomov 462919a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 463019a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 463119a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4632c9e9f97bSIlya Dryomov } 4633c9e9f97bSIlya Dryomov 46349ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4635c9e9f97bSIlya Dryomov { 4636496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4637c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4638c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4639c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4640ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4641c9e9f97bSIlya Dryomov int ret; 4642c9e9f97bSIlya Dryomov 4643c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4644c9e9f97bSIlya Dryomov return -EPERM; 4645c9e9f97bSIlya Dryomov 4646e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46479ba1f6e4SLiu Bo if (ret) 46489ba1f6e4SLiu Bo return ret; 46499ba1f6e4SLiu Bo 4650ed0fb78fSIlya Dryomov again: 4651171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4652c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4653ed0fb78fSIlya Dryomov need_unlock = true; 4654ed0fb78fSIlya Dryomov goto locked; 4655ed0fb78fSIlya Dryomov } 4656ed0fb78fSIlya Dryomov 4657ed0fb78fSIlya Dryomov /* 465801327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4659ed0fb78fSIlya Dryomov * (1) some other op is running 4660ed0fb78fSIlya Dryomov * (2) balance is running 4661ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4662ed0fb78fSIlya Dryomov */ 4663ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4664ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4665ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 46663009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4667ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4668dccdb07bSDavid Sterba /* 4669dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4670dccdb07bSDavid Sterba * we'll reexamine the status again. 4671dccdb07bSDavid Sterba */ 4672ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4673ed0fb78fSIlya Dryomov 4674ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 46753009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4676ed0fb78fSIlya Dryomov /* this is (3) */ 4677ed0fb78fSIlya Dryomov need_unlock = false; 4678ed0fb78fSIlya Dryomov goto locked; 4679ed0fb78fSIlya Dryomov } 4680ed0fb78fSIlya Dryomov 4681ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4682ed0fb78fSIlya Dryomov goto again; 4683ed0fb78fSIlya Dryomov } else { 4684ed0fb78fSIlya Dryomov /* this is (2) */ 4685ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4686ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4687ed0fb78fSIlya Dryomov goto out; 4688ed0fb78fSIlya Dryomov } 4689ed0fb78fSIlya Dryomov } else { 4690ed0fb78fSIlya Dryomov /* this is (1) */ 4691ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4692e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4693ed0fb78fSIlya Dryomov goto out; 4694ed0fb78fSIlya Dryomov } 4695ed0fb78fSIlya Dryomov 4696ed0fb78fSIlya Dryomov locked: 4697171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4698c9e9f97bSIlya Dryomov 4699c9e9f97bSIlya Dryomov if (arg) { 4700c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4701c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4702c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4703ed0fb78fSIlya Dryomov goto out_unlock; 4704c9e9f97bSIlya Dryomov } 4705de322263SIlya Dryomov 4706de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4707de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4708de322263SIlya Dryomov ret = -ENOTCONN; 4709de322263SIlya Dryomov goto out_bargs; 4710de322263SIlya Dryomov } 4711de322263SIlya Dryomov 4712de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4713de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4714de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4715de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4716de322263SIlya Dryomov 4717de322263SIlya Dryomov goto do_balance; 4718de322263SIlya Dryomov } 4719c9e9f97bSIlya Dryomov } else { 4720c9e9f97bSIlya Dryomov bargs = NULL; 4721c9e9f97bSIlya Dryomov } 4722c9e9f97bSIlya Dryomov 4723ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4724837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4725837d5b6eSIlya Dryomov goto out_bargs; 4726837d5b6eSIlya Dryomov } 4727837d5b6eSIlya Dryomov 47288d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4729c9e9f97bSIlya Dryomov if (!bctl) { 4730c9e9f97bSIlya Dryomov ret = -ENOMEM; 4731c9e9f97bSIlya Dryomov goto out_bargs; 4732c9e9f97bSIlya Dryomov } 4733c9e9f97bSIlya Dryomov 4734c9e9f97bSIlya Dryomov if (arg) { 4735c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4736c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4737c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4738c9e9f97bSIlya Dryomov 4739c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4740f43ffb60SIlya Dryomov } else { 4741f43ffb60SIlya Dryomov /* balance everything - no filters */ 4742f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4743c9e9f97bSIlya Dryomov } 4744c9e9f97bSIlya Dryomov 47458eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47468eb93459SDavid Sterba ret = -EINVAL; 47470f89abf5SChristian Engelmayer goto out_bctl; 47488eb93459SDavid Sterba } 47498eb93459SDavid Sterba 4750de322263SIlya Dryomov do_balance: 4751c9e9f97bSIlya Dryomov /* 4752149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4753149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4754149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4755149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4756c9e9f97bSIlya Dryomov */ 4757ed0fb78fSIlya Dryomov need_unlock = false; 4758ed0fb78fSIlya Dryomov 47596fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 47600f89abf5SChristian Engelmayer bctl = NULL; 4761ed0fb78fSIlya Dryomov 4762d00c2d9cSFilipe Manana if ((ret == 0 || ret == -ECANCELED) && arg) { 4763c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4764c9e9f97bSIlya Dryomov ret = -EFAULT; 4765c9e9f97bSIlya Dryomov } 4766c9e9f97bSIlya Dryomov 47670f89abf5SChristian Engelmayer out_bctl: 47680f89abf5SChristian Engelmayer kfree(bctl); 4769c9e9f97bSIlya Dryomov out_bargs: 4770c9e9f97bSIlya Dryomov kfree(bargs); 4771ed0fb78fSIlya Dryomov out_unlock: 4772c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4773ed0fb78fSIlya Dryomov if (need_unlock) 4774171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4775ed0fb78fSIlya Dryomov out: 4776e54bfa31SLiu Bo mnt_drop_write_file(file); 4777c9e9f97bSIlya Dryomov return ret; 4778c9e9f97bSIlya Dryomov } 4779c9e9f97bSIlya Dryomov 47802ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4781837d5b6eSIlya Dryomov { 4782837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4783837d5b6eSIlya Dryomov return -EPERM; 4784837d5b6eSIlya Dryomov 4785837d5b6eSIlya Dryomov switch (cmd) { 4786837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 47870b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4788a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47890b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4790837d5b6eSIlya Dryomov } 4791837d5b6eSIlya Dryomov 4792837d5b6eSIlya Dryomov return -EINVAL; 4793837d5b6eSIlya Dryomov } 4794837d5b6eSIlya Dryomov 47952ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 479619a39dceSIlya Dryomov void __user *arg) 479719a39dceSIlya Dryomov { 479819a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 479919a39dceSIlya Dryomov int ret = 0; 480019a39dceSIlya Dryomov 480119a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 480219a39dceSIlya Dryomov return -EPERM; 480319a39dceSIlya Dryomov 480419a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 480519a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 480619a39dceSIlya Dryomov ret = -ENOTCONN; 480719a39dceSIlya Dryomov goto out; 480819a39dceSIlya Dryomov } 480919a39dceSIlya Dryomov 48108d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 481119a39dceSIlya Dryomov if (!bargs) { 481219a39dceSIlya Dryomov ret = -ENOMEM; 481319a39dceSIlya Dryomov goto out; 481419a39dceSIlya Dryomov } 481519a39dceSIlya Dryomov 4816008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 481719a39dceSIlya Dryomov 481819a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 481919a39dceSIlya Dryomov ret = -EFAULT; 482019a39dceSIlya Dryomov 482119a39dceSIlya Dryomov kfree(bargs); 482219a39dceSIlya Dryomov out: 482319a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 482419a39dceSIlya Dryomov return ret; 482519a39dceSIlya Dryomov } 482619a39dceSIlya Dryomov 4827905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 48285d13a37bSArne Jansen { 48290b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48300b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48315d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 48325d13a37bSArne Jansen int ret; 48335d13a37bSArne Jansen 48345d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48355d13a37bSArne Jansen return -EPERM; 48365d13a37bSArne Jansen 4837905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4838905b0ddaSMiao Xie if (ret) 4839905b0ddaSMiao Xie return ret; 48405d13a37bSArne Jansen 48415d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4842905b0ddaSMiao Xie if (IS_ERR(sa)) { 4843905b0ddaSMiao Xie ret = PTR_ERR(sa); 4844905b0ddaSMiao Xie goto drop_write; 4845905b0ddaSMiao Xie } 48465d13a37bSArne Jansen 48470b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48485d13a37bSArne Jansen 48495d13a37bSArne Jansen switch (sa->cmd) { 48505d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 4851340f1aa2SNikolay Borisov ret = btrfs_quota_enable(fs_info); 48525d13a37bSArne Jansen break; 48535d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 4854340f1aa2SNikolay Borisov ret = btrfs_quota_disable(fs_info); 48555d13a37bSArne Jansen break; 48565d13a37bSArne Jansen default: 48575d13a37bSArne Jansen ret = -EINVAL; 48585d13a37bSArne Jansen break; 48595d13a37bSArne Jansen } 48605d13a37bSArne Jansen 48615d13a37bSArne Jansen kfree(sa); 48620b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4863905b0ddaSMiao Xie drop_write: 4864905b0ddaSMiao Xie mnt_drop_write_file(file); 48655d13a37bSArne Jansen return ret; 48665d13a37bSArne Jansen } 48675d13a37bSArne Jansen 4868905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48695d13a37bSArne Jansen { 48700b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48710b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48720b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48735d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48745d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48755d13a37bSArne Jansen int ret; 48765d13a37bSArne Jansen int err; 48775d13a37bSArne Jansen 48785d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48795d13a37bSArne Jansen return -EPERM; 48805d13a37bSArne Jansen 4881905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4882905b0ddaSMiao Xie if (ret) 4883905b0ddaSMiao Xie return ret; 48845d13a37bSArne Jansen 48855d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4886905b0ddaSMiao Xie if (IS_ERR(sa)) { 4887905b0ddaSMiao Xie ret = PTR_ERR(sa); 4888905b0ddaSMiao Xie goto drop_write; 4889905b0ddaSMiao Xie } 48905d13a37bSArne Jansen 48915d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48925d13a37bSArne Jansen if (IS_ERR(trans)) { 48935d13a37bSArne Jansen ret = PTR_ERR(trans); 48945d13a37bSArne Jansen goto out; 48955d13a37bSArne Jansen } 48965d13a37bSArne Jansen 48975d13a37bSArne Jansen if (sa->assign) { 48989f8a6ce6SLu Fengqi ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); 48995d13a37bSArne Jansen } else { 490039616c27SLu Fengqi ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 49015d13a37bSArne Jansen } 49025d13a37bSArne Jansen 4903e082f563SQu Wenruo /* update qgroup status and info */ 4904280f8bd2SLu Fengqi err = btrfs_run_qgroups(trans); 4905e082f563SQu Wenruo if (err < 0) 49060b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4907ad8403dfSAnand Jain "failed to update qgroup status and info"); 49083a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49095d13a37bSArne Jansen if (err && !ret) 49105d13a37bSArne Jansen ret = err; 49115d13a37bSArne Jansen 49125d13a37bSArne Jansen out: 49135d13a37bSArne Jansen kfree(sa); 4914905b0ddaSMiao Xie drop_write: 4915905b0ddaSMiao Xie mnt_drop_write_file(file); 49165d13a37bSArne Jansen return ret; 49175d13a37bSArne Jansen } 49185d13a37bSArne Jansen 4919905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 49205d13a37bSArne Jansen { 49210b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49220b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49235d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49245d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49255d13a37bSArne Jansen int ret; 49265d13a37bSArne Jansen int err; 49275d13a37bSArne Jansen 49285d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49295d13a37bSArne Jansen return -EPERM; 49305d13a37bSArne Jansen 4931905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4932905b0ddaSMiao Xie if (ret) 4933905b0ddaSMiao Xie return ret; 49345d13a37bSArne Jansen 49355d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4936905b0ddaSMiao Xie if (IS_ERR(sa)) { 4937905b0ddaSMiao Xie ret = PTR_ERR(sa); 4938905b0ddaSMiao Xie goto drop_write; 4939905b0ddaSMiao Xie } 49405d13a37bSArne Jansen 4941d86e56cfSMiao Xie if (!sa->qgroupid) { 4942d86e56cfSMiao Xie ret = -EINVAL; 4943d86e56cfSMiao Xie goto out; 4944d86e56cfSMiao Xie } 4945d86e56cfSMiao Xie 49465d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49475d13a37bSArne Jansen if (IS_ERR(trans)) { 49485d13a37bSArne Jansen ret = PTR_ERR(trans); 49495d13a37bSArne Jansen goto out; 49505d13a37bSArne Jansen } 49515d13a37bSArne Jansen 49525d13a37bSArne Jansen if (sa->create) { 495349a05ecdSLu Fengqi ret = btrfs_create_qgroup(trans, sa->qgroupid); 49545d13a37bSArne Jansen } else { 49553efbee1dSLu Fengqi ret = btrfs_remove_qgroup(trans, sa->qgroupid); 49565d13a37bSArne Jansen } 49575d13a37bSArne Jansen 49583a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49595d13a37bSArne Jansen if (err && !ret) 49605d13a37bSArne Jansen ret = err; 49615d13a37bSArne Jansen 49625d13a37bSArne Jansen out: 49635d13a37bSArne Jansen kfree(sa); 4964905b0ddaSMiao Xie drop_write: 4965905b0ddaSMiao Xie mnt_drop_write_file(file); 49665d13a37bSArne Jansen return ret; 49675d13a37bSArne Jansen } 49685d13a37bSArne Jansen 4969905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49705d13a37bSArne Jansen { 49710b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49720b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49735d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49745d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49755d13a37bSArne Jansen int ret; 49765d13a37bSArne Jansen int err; 49775d13a37bSArne Jansen u64 qgroupid; 49785d13a37bSArne Jansen 49795d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49805d13a37bSArne Jansen return -EPERM; 49815d13a37bSArne Jansen 4982905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4983905b0ddaSMiao Xie if (ret) 4984905b0ddaSMiao Xie return ret; 49855d13a37bSArne Jansen 49865d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4987905b0ddaSMiao Xie if (IS_ERR(sa)) { 4988905b0ddaSMiao Xie ret = PTR_ERR(sa); 4989905b0ddaSMiao Xie goto drop_write; 4990905b0ddaSMiao Xie } 49915d13a37bSArne Jansen 49925d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49935d13a37bSArne Jansen if (IS_ERR(trans)) { 49945d13a37bSArne Jansen ret = PTR_ERR(trans); 49955d13a37bSArne Jansen goto out; 49965d13a37bSArne Jansen } 49975d13a37bSArne Jansen 49985d13a37bSArne Jansen qgroupid = sa->qgroupid; 49995d13a37bSArne Jansen if (!qgroupid) { 50005d13a37bSArne Jansen /* take the current subvol as qgroup */ 50015d13a37bSArne Jansen qgroupid = root->root_key.objectid; 50025d13a37bSArne Jansen } 50035d13a37bSArne Jansen 5004f0042d5eSLu Fengqi ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 50055d13a37bSArne Jansen 50063a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 50075d13a37bSArne Jansen if (err && !ret) 50085d13a37bSArne Jansen ret = err; 50095d13a37bSArne Jansen 50105d13a37bSArne Jansen out: 50115d13a37bSArne Jansen kfree(sa); 5012905b0ddaSMiao Xie drop_write: 5013905b0ddaSMiao Xie mnt_drop_write_file(file); 50145d13a37bSArne Jansen return ret; 50155d13a37bSArne Jansen } 50165d13a37bSArne Jansen 50172f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 50182f232036SJan Schmidt { 50190b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50200b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50212f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50222f232036SJan Schmidt int ret; 50232f232036SJan Schmidt 50242f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50252f232036SJan Schmidt return -EPERM; 50262f232036SJan Schmidt 50272f232036SJan Schmidt ret = mnt_want_write_file(file); 50282f232036SJan Schmidt if (ret) 50292f232036SJan Schmidt return ret; 50302f232036SJan Schmidt 50312f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50322f232036SJan Schmidt if (IS_ERR(qsa)) { 50332f232036SJan Schmidt ret = PTR_ERR(qsa); 50342f232036SJan Schmidt goto drop_write; 50352f232036SJan Schmidt } 50362f232036SJan Schmidt 50372f232036SJan Schmidt if (qsa->flags) { 50382f232036SJan Schmidt ret = -EINVAL; 50392f232036SJan Schmidt goto out; 50402f232036SJan Schmidt } 50412f232036SJan Schmidt 50420b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 50432f232036SJan Schmidt 50442f232036SJan Schmidt out: 50452f232036SJan Schmidt kfree(qsa); 50462f232036SJan Schmidt drop_write: 50472f232036SJan Schmidt mnt_drop_write_file(file); 50482f232036SJan Schmidt return ret; 50492f232036SJan Schmidt } 50502f232036SJan Schmidt 50512f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50522f232036SJan Schmidt { 50530b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50552f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50562f232036SJan Schmidt int ret = 0; 50572f232036SJan Schmidt 50582f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50592f232036SJan Schmidt return -EPERM; 50602f232036SJan Schmidt 50618d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 50622f232036SJan Schmidt if (!qsa) 50632f232036SJan Schmidt return -ENOMEM; 50642f232036SJan Schmidt 50650b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50662f232036SJan Schmidt qsa->flags = 1; 50670b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 50682f232036SJan Schmidt } 50692f232036SJan Schmidt 50702f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50712f232036SJan Schmidt ret = -EFAULT; 50722f232036SJan Schmidt 50732f232036SJan Schmidt kfree(qsa); 50742f232036SJan Schmidt return ret; 50752f232036SJan Schmidt } 50762f232036SJan Schmidt 507757254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 507857254b6eSJan Schmidt { 50790b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50800b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 508157254b6eSJan Schmidt 508257254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 508357254b6eSJan Schmidt return -EPERM; 508457254b6eSJan Schmidt 50850b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 508657254b6eSJan Schmidt } 508757254b6eSJan Schmidt 5088abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5089abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50908ea05e3aSAlexander Block { 5091496ad9aaSAl Viro struct inode *inode = file_inode(file); 50920b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50938ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50948ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50958ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 509695582b00SDeepa Dinamani struct timespec64 ct = current_time(inode); 50978ea05e3aSAlexander Block int ret = 0; 5098dd5f9615SStefan Behrens int received_uuid_changed; 50998ea05e3aSAlexander Block 5100bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5101bd60ea0fSDavid Sterba return -EPERM; 5102bd60ea0fSDavid Sterba 51038ea05e3aSAlexander Block ret = mnt_want_write_file(file); 51048ea05e3aSAlexander Block if (ret < 0) 51058ea05e3aSAlexander Block return ret; 51068ea05e3aSAlexander Block 51070b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51088ea05e3aSAlexander Block 51094a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 51108ea05e3aSAlexander Block ret = -EINVAL; 51118ea05e3aSAlexander Block goto out; 51128ea05e3aSAlexander Block } 51138ea05e3aSAlexander Block 51148ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 51158ea05e3aSAlexander Block ret = -EROFS; 51168ea05e3aSAlexander Block goto out; 51178ea05e3aSAlexander Block } 51188ea05e3aSAlexander Block 5119dd5f9615SStefan Behrens /* 5120dd5f9615SStefan Behrens * 1 - root item 5121dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5122dd5f9615SStefan Behrens */ 5123dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51248ea05e3aSAlexander Block if (IS_ERR(trans)) { 51258ea05e3aSAlexander Block ret = PTR_ERR(trans); 51268ea05e3aSAlexander Block trans = NULL; 51278ea05e3aSAlexander Block goto out; 51288ea05e3aSAlexander Block } 51298ea05e3aSAlexander Block 51308ea05e3aSAlexander Block sa->rtransid = trans->transid; 51318ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51328ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51338ea05e3aSAlexander Block 5134dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5135dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5136dd5f9615SStefan Behrens if (received_uuid_changed && 5137d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5138d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5139dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5140dd5f9615SStefan Behrens root->root_key.objectid); 5141d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5142d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5143d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5144d87ff758SNikolay Borisov goto out; 5145d87ff758SNikolay Borisov } 5146d87ff758SNikolay Borisov } 51478ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51488ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51498ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51503cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51513cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51523cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51533cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51548ea05e3aSAlexander Block 51550b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 51568ea05e3aSAlexander Block &root->root_key, &root->root_item); 51578ea05e3aSAlexander Block if (ret < 0) { 51583a45bb20SJeff Mahoney btrfs_end_transaction(trans); 51598ea05e3aSAlexander Block goto out; 5160dd5f9615SStefan Behrens } 5161dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5162cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5163dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5164dd5f9615SStefan Behrens root->root_key.objectid); 5165dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 516666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5167efd38150SNikolay Borisov btrfs_end_transaction(trans); 5168dd5f9615SStefan Behrens goto out; 5169dd5f9615SStefan Behrens } 5170dd5f9615SStefan Behrens } 51713a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5172abccd00fSHugo Mills out: 51730b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5174abccd00fSHugo Mills mnt_drop_write_file(file); 5175abccd00fSHugo Mills return ret; 5176abccd00fSHugo Mills } 5177abccd00fSHugo Mills 5178abccd00fSHugo Mills #ifdef CONFIG_64BIT 5179abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5180abccd00fSHugo Mills void __user *arg) 5181abccd00fSHugo Mills { 5182abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5183abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5184abccd00fSHugo Mills int ret = 0; 5185abccd00fSHugo Mills 5186abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 51877b9ea627SShailendra Verma if (IS_ERR(args32)) 51887b9ea627SShailendra Verma return PTR_ERR(args32); 5189abccd00fSHugo Mills 51908d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 519184dbeb87SDan Carpenter if (!args64) { 519284dbeb87SDan Carpenter ret = -ENOMEM; 5193abccd00fSHugo Mills goto out; 5194abccd00fSHugo Mills } 5195abccd00fSHugo Mills 5196abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5197abccd00fSHugo Mills args64->stransid = args32->stransid; 5198abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5199abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5200abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5201abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5202abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5203abccd00fSHugo Mills args64->flags = args32->flags; 5204abccd00fSHugo Mills 5205abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5206abccd00fSHugo Mills if (ret) 5207abccd00fSHugo Mills goto out; 5208abccd00fSHugo Mills 5209abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5210abccd00fSHugo Mills args32->stransid = args64->stransid; 5211abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5212abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5213abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5214abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5215abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5216abccd00fSHugo Mills args32->flags = args64->flags; 5217abccd00fSHugo Mills 5218abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5219abccd00fSHugo Mills if (ret) 5220abccd00fSHugo Mills ret = -EFAULT; 5221abccd00fSHugo Mills 5222abccd00fSHugo Mills out: 5223abccd00fSHugo Mills kfree(args32); 5224abccd00fSHugo Mills kfree(args64); 5225abccd00fSHugo Mills return ret; 5226abccd00fSHugo Mills } 5227abccd00fSHugo Mills #endif 5228abccd00fSHugo Mills 5229abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5230abccd00fSHugo Mills void __user *arg) 5231abccd00fSHugo Mills { 5232abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5233abccd00fSHugo Mills int ret = 0; 5234abccd00fSHugo Mills 5235abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 52367b9ea627SShailendra Verma if (IS_ERR(sa)) 52377b9ea627SShailendra Verma return PTR_ERR(sa); 5238abccd00fSHugo Mills 5239abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5240abccd00fSHugo Mills 5241abccd00fSHugo Mills if (ret) 5242abccd00fSHugo Mills goto out; 5243abccd00fSHugo Mills 52448ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52458ea05e3aSAlexander Block if (ret) 52468ea05e3aSAlexander Block ret = -EFAULT; 52478ea05e3aSAlexander Block 52488ea05e3aSAlexander Block out: 52498ea05e3aSAlexander Block kfree(sa); 52508ea05e3aSAlexander Block return ret; 52518ea05e3aSAlexander Block } 52528ea05e3aSAlexander Block 5253867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5254867ab667Sjeff.liu { 52550b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52560b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5257a1b83ac5SAnand Jain size_t len; 5258867ab667Sjeff.liu int ret; 5259a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5260a1b83ac5SAnand Jain 52610b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 52620b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 52630b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5264a1b83ac5SAnand Jain 5265a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5266867ab667Sjeff.liu 5267867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 52680b246afaSJeff Mahoney btrfs_warn(fs_info, 52690b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 52700b246afaSJeff Mahoney --len); 5271867ab667Sjeff.liu } 5272867ab667Sjeff.liu 5273867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5274867ab667Sjeff.liu 5275867ab667Sjeff.liu return ret ? -EFAULT : 0; 5276867ab667Sjeff.liu } 5277867ab667Sjeff.liu 5278a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5279a8bfd4abSjeff.liu { 52800b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52810b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52820b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52830b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5284a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5285a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5286a8bfd4abSjeff.liu int ret; 5287a8bfd4abSjeff.liu 5288a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5289a8bfd4abSjeff.liu return -EPERM; 5290a8bfd4abSjeff.liu 5291a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5292a8bfd4abSjeff.liu return -EFAULT; 5293a8bfd4abSjeff.liu 5294a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 52950b246afaSJeff Mahoney btrfs_err(fs_info, 52965d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5297a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5298a8bfd4abSjeff.liu return -EINVAL; 5299a8bfd4abSjeff.liu } 5300a8bfd4abSjeff.liu 5301a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5302a8bfd4abSjeff.liu if (ret) 5303a8bfd4abSjeff.liu return ret; 5304a8bfd4abSjeff.liu 5305a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5306a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5307a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5308a8bfd4abSjeff.liu goto out_unlock; 5309a8bfd4abSjeff.liu } 5310a8bfd4abSjeff.liu 53110b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5312a8bfd4abSjeff.liu strcpy(super_block->label, label); 53130b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53143a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5315a8bfd4abSjeff.liu 5316a8bfd4abSjeff.liu out_unlock: 5317a8bfd4abSjeff.liu mnt_drop_write_file(file); 5318a8bfd4abSjeff.liu return ret; 5319a8bfd4abSjeff.liu } 5320a8bfd4abSjeff.liu 53212eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 53222eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 53232eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53242eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53252eaa055fSJeff Mahoney 5326d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53272eaa055fSJeff Mahoney { 53284d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53292eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53302eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53312eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53322eaa055fSJeff Mahoney }; 53332eaa055fSJeff Mahoney 53342eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53352eaa055fSJeff Mahoney return -EFAULT; 53362eaa055fSJeff Mahoney 53372eaa055fSJeff Mahoney return 0; 53382eaa055fSJeff Mahoney } 53392eaa055fSJeff Mahoney 53402eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53412eaa055fSJeff Mahoney { 53420b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53430b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53440b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53452eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53462eaa055fSJeff Mahoney 53472eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53482eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53492eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53502eaa055fSJeff Mahoney 53512eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53522eaa055fSJeff Mahoney return -EFAULT; 53532eaa055fSJeff Mahoney 53542eaa055fSJeff Mahoney return 0; 53552eaa055fSJeff Mahoney } 53562eaa055fSJeff Mahoney 53572ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 53583b02a68aSJeff Mahoney enum btrfs_feature_set set, 53592eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53602eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53612eaa055fSJeff Mahoney { 53623b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 53633b02a68aSJeff Mahoney char *names; 53642eaa055fSJeff Mahoney u64 disallowed, unsupported; 53652eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53662eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53672eaa055fSJeff Mahoney 53682eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53692eaa055fSJeff Mahoney if (unsupported) { 53703b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53713b02a68aSJeff Mahoney if (names) { 53720b246afaSJeff Mahoney btrfs_warn(fs_info, 53733b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53743b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53753b02a68aSJeff Mahoney kfree(names); 53763b02a68aSJeff Mahoney } else 53770b246afaSJeff Mahoney btrfs_warn(fs_info, 53782eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53792eaa055fSJeff Mahoney type, unsupported); 53802eaa055fSJeff Mahoney return -EOPNOTSUPP; 53812eaa055fSJeff Mahoney } 53822eaa055fSJeff Mahoney 53832eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53842eaa055fSJeff Mahoney if (disallowed) { 53853b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53863b02a68aSJeff Mahoney if (names) { 53870b246afaSJeff Mahoney btrfs_warn(fs_info, 53883b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53893b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53903b02a68aSJeff Mahoney kfree(names); 53913b02a68aSJeff Mahoney } else 53920b246afaSJeff Mahoney btrfs_warn(fs_info, 53932eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53942eaa055fSJeff Mahoney type, disallowed); 53952eaa055fSJeff Mahoney return -EPERM; 53962eaa055fSJeff Mahoney } 53972eaa055fSJeff Mahoney 53982eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 53992eaa055fSJeff Mahoney if (disallowed) { 54003b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54013b02a68aSJeff Mahoney if (names) { 54020b246afaSJeff Mahoney btrfs_warn(fs_info, 54033b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 54043b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54053b02a68aSJeff Mahoney kfree(names); 54063b02a68aSJeff Mahoney } else 54070b246afaSJeff Mahoney btrfs_warn(fs_info, 54082eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 54092eaa055fSJeff Mahoney type, disallowed); 54102eaa055fSJeff Mahoney return -EPERM; 54112eaa055fSJeff Mahoney } 54122eaa055fSJeff Mahoney 54132eaa055fSJeff Mahoney return 0; 54142eaa055fSJeff Mahoney } 54152eaa055fSJeff Mahoney 54162ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 54172ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 54182eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 54192eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 54202eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 54212eaa055fSJeff Mahoney 54222eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 54232eaa055fSJeff Mahoney { 54240b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54250b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54260b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54270b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 54282eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54292eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54302eaa055fSJeff Mahoney u64 newflags; 54312eaa055fSJeff Mahoney int ret; 54322eaa055fSJeff Mahoney 54332eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54342eaa055fSJeff Mahoney return -EPERM; 54352eaa055fSJeff Mahoney 54362eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54372eaa055fSJeff Mahoney return -EFAULT; 54382eaa055fSJeff Mahoney 54392eaa055fSJeff Mahoney /* Nothing to do */ 54402eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54412eaa055fSJeff Mahoney !flags[0].incompat_flags) 54422eaa055fSJeff Mahoney return 0; 54432eaa055fSJeff Mahoney 54442ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 54452eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54462eaa055fSJeff Mahoney if (ret) 54472eaa055fSJeff Mahoney return ret; 54482eaa055fSJeff Mahoney 54492ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 54502eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54512eaa055fSJeff Mahoney if (ret) 54522eaa055fSJeff Mahoney return ret; 54532eaa055fSJeff Mahoney 54542ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 54552eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54562eaa055fSJeff Mahoney if (ret) 54572eaa055fSJeff Mahoney return ret; 54582eaa055fSJeff Mahoney 54597ab19625SDavid Sterba ret = mnt_want_write_file(file); 54607ab19625SDavid Sterba if (ret) 54617ab19625SDavid Sterba return ret; 54627ab19625SDavid Sterba 54638051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54647ab19625SDavid Sterba if (IS_ERR(trans)) { 54657ab19625SDavid Sterba ret = PTR_ERR(trans); 54667ab19625SDavid Sterba goto out_drop_write; 54677ab19625SDavid Sterba } 54682eaa055fSJeff Mahoney 54690b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 54702eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54712eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54722eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54732eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54742eaa055fSJeff Mahoney 54752eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54762eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54772eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54782eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54792eaa055fSJeff Mahoney 54802eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54812eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54822eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54832eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54840b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 54852eaa055fSJeff Mahoney 54863a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 54877ab19625SDavid Sterba out_drop_write: 54887ab19625SDavid Sterba mnt_drop_write_file(file); 54897ab19625SDavid Sterba 54907ab19625SDavid Sterba return ret; 54912eaa055fSJeff Mahoney } 54922eaa055fSJeff Mahoney 54932351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 54942351f431SJosef Bacik { 54952351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 54962351f431SJosef Bacik int ret; 54972351f431SJosef Bacik 54982351f431SJosef Bacik if (compat) { 54992351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55002351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 55012351f431SJosef Bacik 55022351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 55032351f431SJosef Bacik if (ret) 55042351f431SJosef Bacik return -EFAULT; 55052351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 55062351f431SJosef Bacik if (!arg) 55072351f431SJosef Bacik return -ENOMEM; 55082351f431SJosef Bacik arg->send_fd = args32.send_fd; 55092351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 55102351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 55112351f431SJosef Bacik arg->parent_root = args32.parent_root; 55122351f431SJosef Bacik arg->flags = args32.flags; 55132351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 55142351f431SJosef Bacik sizeof(args32.reserved)); 55152351f431SJosef Bacik #else 55162351f431SJosef Bacik return -ENOTTY; 55172351f431SJosef Bacik #endif 55182351f431SJosef Bacik } else { 55192351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 55202351f431SJosef Bacik if (IS_ERR(arg)) 55212351f431SJosef Bacik return PTR_ERR(arg); 55222351f431SJosef Bacik } 55232351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 55242351f431SJosef Bacik kfree(arg); 55252351f431SJosef Bacik return ret; 55262351f431SJosef Bacik } 55272351f431SJosef Bacik 5528f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5529f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5530f46b5a66SChristoph Hellwig { 55310b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55330b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55344bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5535f46b5a66SChristoph Hellwig 5536f46b5a66SChristoph Hellwig switch (cmd) { 55376cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 55386cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 55396cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 55406cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 55416cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 55426cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5543f7039b1dSLi Dongyang case FITRIM: 5544f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5545f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5546fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5547fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5548fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 55493de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5550fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 55516f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 55526f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 555376dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 555476dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 55550caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 55560caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 55570caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 55580caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55596ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55606ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5561f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55621e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55631e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55641e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5565f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5566198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5567f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55682ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5569f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5570da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55716b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55726b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5573475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 55742ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5575475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 55762ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5577f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55789ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5579ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5580ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5581cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5582cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5583ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5584ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5585d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5586d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5587d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5588d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5589d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5590d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 55911406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55922ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 55939b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55949b199859SFilipe David Borba Manana int ret; 55959b199859SFilipe David Borba Manana 559682b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 55979b199859SFilipe David Borba Manana if (ret) 55989b199859SFilipe David Borba Manana return ret; 55990b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 56002fad4e83SDavid Sterba /* 56012fad4e83SDavid Sterba * The transaction thread may want to do more work, 560201327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 56032fad4e83SDavid Sterba * processing uncleaned subvols. 56042fad4e83SDavid Sterba */ 56050b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 56069b199859SFilipe David Borba Manana return ret; 56079b199859SFilipe David Borba Manana } 560846204592SSage Weil case BTRFS_IOC_START_SYNC: 56099a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 561046204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 56112ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5612475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5613b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5614475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 56152ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5616475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 56172ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5618c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 56199ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5620837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 56212ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 562219a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 56232ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 56248ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 56258ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5626abccd00fSHugo Mills #ifdef CONFIG_64BIT 5627abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5628abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5629abccd00fSHugo Mills #endif 563031db9f7cSAlexander Block case BTRFS_IOC_SEND: 56312351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 56322351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 56332351f431SJosef Bacik case BTRFS_IOC_SEND_32: 56342351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 56352351f431SJosef Bacik #endif 5636c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 56372ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 56385d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5639905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 56405d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5641905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 56425d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5643905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 56445d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5645905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 56462f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 56472f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 56482f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 56492f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 565057254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 565157254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 56523f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 56532ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5654867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5655867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5656a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5657a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 56582eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5659d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 56602eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 56612eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 56622eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56632eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5664e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5665e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5666025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5667025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5668b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5669b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 567042e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 567142e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 567223d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 567323d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5674f46b5a66SChristoph Hellwig } 5675f46b5a66SChristoph Hellwig 5676f46b5a66SChristoph Hellwig return -ENOTTY; 5677f46b5a66SChristoph Hellwig } 56784c63c245SLuke Dashjr 56794c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56804c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56814c63c245SLuke Dashjr { 56822a362249SJeff Mahoney /* 56832a362249SJeff Mahoney * These all access 32-bit values anyway so no further 56842a362249SJeff Mahoney * handling is necessary. 56852a362249SJeff Mahoney */ 56864c63c245SLuke Dashjr switch (cmd) { 56874c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56884c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56894c63c245SLuke Dashjr break; 56904c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56914c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56924c63c245SLuke Dashjr break; 56934c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56944c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56954c63c245SLuke Dashjr break; 56964c63c245SLuke Dashjr } 56974c63c245SLuke Dashjr 56984c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56994c63c245SLuke Dashjr } 57004c63c245SLuke Dashjr #endif 5701