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/buffer_head.h> 9f46b5a66SChristoph Hellwig #include <linux/file.h> 10f46b5a66SChristoph Hellwig #include <linux/fs.h> 11cb8e7090SChristoph Hellwig #include <linux/fsnotify.h> 12f46b5a66SChristoph Hellwig #include <linux/pagemap.h> 13f46b5a66SChristoph Hellwig #include <linux/highmem.h> 14f46b5a66SChristoph Hellwig #include <linux/time.h> 15f46b5a66SChristoph Hellwig #include <linux/init.h> 16f46b5a66SChristoph Hellwig #include <linux/string.h> 17f46b5a66SChristoph Hellwig #include <linux/backing-dev.h> 18cb8e7090SChristoph Hellwig #include <linux/mount.h> 19f46b5a66SChristoph Hellwig #include <linux/mpage.h> 20cb8e7090SChristoph Hellwig #include <linux/namei.h> 21f46b5a66SChristoph Hellwig #include <linux/swap.h> 22f46b5a66SChristoph Hellwig #include <linux/writeback.h> 23f46b5a66SChristoph Hellwig #include <linux/compat.h> 24f46b5a66SChristoph Hellwig #include <linux/bit_spinlock.h> 25cb8e7090SChristoph Hellwig #include <linux/security.h> 26f46b5a66SChristoph Hellwig #include <linux/xattr.h> 27f54de068SDavid Sterba #include <linux/mm.h> 285a0e3ad6STejun Heo #include <linux/slab.h> 29f7039b1dSLi Dongyang #include <linux/blkdev.h> 308ea05e3aSAlexander Block #include <linux/uuid.h> 3155e301fdSFilipe Brandenburger #include <linux/btrfs.h> 32416161dbSMark Fasheh #include <linux/uaccess.h> 33ae5e165dSJeff Layton #include <linux/iversion.h> 34f46b5a66SChristoph Hellwig #include "ctree.h" 35f46b5a66SChristoph Hellwig #include "disk-io.h" 36f46b5a66SChristoph Hellwig #include "transaction.h" 37f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 38f46b5a66SChristoph Hellwig #include "print-tree.h" 39f46b5a66SChristoph Hellwig #include "volumes.h" 40925baeddSChris Mason #include "locking.h" 41581bb050SLi Zefan #include "inode-map.h" 42d7728c96SJan Schmidt #include "backref.h" 43606686eeSJosef Bacik #include "rcu-string.h" 4431db9f7cSAlexander Block #include "send.h" 453f6bcfbdSStefan Behrens #include "dev-replace.h" 4663541927SFilipe David Borba Manana #include "props.h" 473b02a68aSJeff Mahoney #include "sysfs.h" 48fcebe456SJosef Bacik #include "qgroup.h" 491ec9a1aeSFilipe Manana #include "tree-log.h" 50ebb8765bSAnand Jain #include "compression.h" 51f46b5a66SChristoph Hellwig 52abccd00fSHugo Mills #ifdef CONFIG_64BIT 53abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 54abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 55abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 56abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 57abccd00fSHugo Mills */ 58abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 59abccd00fSHugo Mills __u64 sec; 60abccd00fSHugo Mills __u32 nsec; 61abccd00fSHugo Mills } __attribute__ ((__packed__)); 62abccd00fSHugo Mills 63abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 64abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 65abccd00fSHugo Mills __u64 stransid; /* in */ 66abccd00fSHugo Mills __u64 rtransid; /* out */ 67abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 68abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 69abccd00fSHugo Mills __u64 flags; /* in */ 70abccd00fSHugo Mills __u64 reserved[16]; /* in */ 71abccd00fSHugo Mills } __attribute__ ((__packed__)); 72abccd00fSHugo Mills 73abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 74abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 75abccd00fSHugo Mills #endif 76abccd00fSHugo Mills 772351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 782351f431SJosef Bacik struct btrfs_ioctl_send_args_32 { 792351f431SJosef Bacik __s64 send_fd; /* in */ 802351f431SJosef Bacik __u64 clone_sources_count; /* in */ 812351f431SJosef Bacik compat_uptr_t clone_sources; /* in */ 822351f431SJosef Bacik __u64 parent_root; /* in */ 832351f431SJosef Bacik __u64 flags; /* in */ 842351f431SJosef Bacik __u64 reserved[4]; /* in */ 852351f431SJosef Bacik } __attribute__ ((__packed__)); 862351f431SJosef Bacik 872351f431SJosef Bacik #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 882351f431SJosef Bacik struct btrfs_ioctl_send_args_32) 892351f431SJosef Bacik #endif 90abccd00fSHugo Mills 91416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 921c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 931c919a5eSMark Fasheh int no_time_update); 94416161dbSMark Fasheh 956cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 961905a0f7SDavid Sterba static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, 971905a0f7SDavid Sterba unsigned int flags) 986cbff00fSChristoph Hellwig { 991905a0f7SDavid Sterba if (S_ISDIR(inode->i_mode)) 1006cbff00fSChristoph Hellwig return flags; 1011905a0f7SDavid Sterba else if (S_ISREG(inode->i_mode)) 1026cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1036cbff00fSChristoph Hellwig else 1046cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1056cbff00fSChristoph Hellwig } 106f46b5a66SChristoph Hellwig 1076cbff00fSChristoph Hellwig /* 108a157d4fdSDavid Sterba * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 109a157d4fdSDavid Sterba * ioctl. 1106cbff00fSChristoph Hellwig */ 111a157d4fdSDavid Sterba static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags) 1126cbff00fSChristoph Hellwig { 1136cbff00fSChristoph Hellwig unsigned int iflags = 0; 1146cbff00fSChristoph Hellwig 1156cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1166cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1176cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1186cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1196cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1206cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1216cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1226cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1236cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1246cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1256cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1266cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 127d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 128d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 129d0092bddSLi Zefan 13013f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 131d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 13213f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13313f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1346cbff00fSChristoph Hellwig 1356cbff00fSChristoph Hellwig return iflags; 1366cbff00fSChristoph Hellwig } 1376cbff00fSChristoph Hellwig 1386cbff00fSChristoph Hellwig /* 1396cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1406cbff00fSChristoph Hellwig */ 1417b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1426cbff00fSChristoph Hellwig { 1435c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1443cc79392SFilipe Manana unsigned int new_fl = 0; 1456cbff00fSChristoph Hellwig 1465c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_SYNC) 1473cc79392SFilipe Manana new_fl |= S_SYNC; 1485c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_IMMUTABLE) 1493cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1505c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_APPEND) 1513cc79392SFilipe Manana new_fl |= S_APPEND; 1525c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_NOATIME) 1533cc79392SFilipe Manana new_fl |= S_NOATIME; 1545c57b8b6SDavid Sterba if (binode->flags & BTRFS_INODE_DIRSYNC) 1553cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1563cc79392SFilipe Manana 1573cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1583cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1593cc79392SFilipe Manana new_fl); 1606cbff00fSChristoph Hellwig } 1616cbff00fSChristoph Hellwig 1626cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1636cbff00fSChristoph Hellwig { 1645c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(file_inode(file)); 1655c57b8b6SDavid Sterba unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags); 1666cbff00fSChristoph Hellwig 1676cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1686cbff00fSChristoph Hellwig return -EFAULT; 1696cbff00fSChristoph Hellwig return 0; 1706cbff00fSChristoph Hellwig } 1716cbff00fSChristoph Hellwig 1725ba76abfSDavid Sterba /* Check if @flags are a supported and valid set of FS_*_FL flags */ 1735ba76abfSDavid Sterba static int check_fsflags(unsigned int flags) 17475e7cb7fSLiu Bo { 17575e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17675e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17775e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 178e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 179e1e8fb6aSLi Zefan FS_NOCOW_FL)) 18075e7cb7fSLiu Bo return -EOPNOTSUPP; 18175e7cb7fSLiu Bo 18275e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 18375e7cb7fSLiu Bo return -EINVAL; 18475e7cb7fSLiu Bo 18575e7cb7fSLiu Bo return 0; 18675e7cb7fSLiu Bo } 18775e7cb7fSLiu Bo 1886cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1896cbff00fSChristoph Hellwig { 190496ad9aaSAl Viro struct inode *inode = file_inode(file); 1910b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1925c57b8b6SDavid Sterba struct btrfs_inode *binode = BTRFS_I(inode); 1935c57b8b6SDavid Sterba struct btrfs_root *root = binode->root; 1946cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1955c57b8b6SDavid Sterba unsigned int fsflags, old_fsflags; 1966cbff00fSChristoph Hellwig int ret; 1975c57b8b6SDavid Sterba u64 old_flags; 1985c57b8b6SDavid Sterba unsigned int old_i_flags; 1997e97b8daSDavid Sterba umode_t mode; 2006cbff00fSChristoph Hellwig 201bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 202bd60ea0fSDavid Sterba return -EPERM; 203bd60ea0fSDavid Sterba 204b83cc969SLi Zefan if (btrfs_root_readonly(root)) 205b83cc969SLi Zefan return -EROFS; 206b83cc969SLi Zefan 2075c57b8b6SDavid Sterba if (copy_from_user(&fsflags, arg, sizeof(fsflags))) 2086cbff00fSChristoph Hellwig return -EFAULT; 2096cbff00fSChristoph Hellwig 2105c57b8b6SDavid Sterba ret = check_fsflags(fsflags); 21175e7cb7fSLiu Bo if (ret) 21275e7cb7fSLiu Bo return ret; 2136cbff00fSChristoph Hellwig 214e7848683SJan Kara ret = mnt_want_write_file(file); 215e7848683SJan Kara if (ret) 216e7848683SJan Kara return ret; 217e7848683SJan Kara 2185955102cSAl Viro inode_lock(inode); 2196cbff00fSChristoph Hellwig 2205c57b8b6SDavid Sterba old_flags = binode->flags; 2215c57b8b6SDavid Sterba old_i_flags = inode->i_flags; 2227e97b8daSDavid Sterba mode = inode->i_mode; 223f062abf0SLi Zefan 2245c57b8b6SDavid Sterba fsflags = btrfs_mask_fsflags_for_type(inode, fsflags); 2255c57b8b6SDavid Sterba old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags); 2265c57b8b6SDavid Sterba if ((fsflags ^ old_fsflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2276cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2286cbff00fSChristoph Hellwig ret = -EPERM; 2296cbff00fSChristoph Hellwig goto out_unlock; 2306cbff00fSChristoph Hellwig } 2316cbff00fSChristoph Hellwig } 2326cbff00fSChristoph Hellwig 2335c57b8b6SDavid Sterba if (fsflags & FS_SYNC_FL) 2345c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_SYNC; 2356cbff00fSChristoph Hellwig else 2365c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_SYNC; 2375c57b8b6SDavid Sterba if (fsflags & FS_IMMUTABLE_FL) 2385c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_IMMUTABLE; 2396cbff00fSChristoph Hellwig else 2405c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_IMMUTABLE; 2415c57b8b6SDavid Sterba if (fsflags & FS_APPEND_FL) 2425c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_APPEND; 2436cbff00fSChristoph Hellwig else 2445c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_APPEND; 2455c57b8b6SDavid Sterba if (fsflags & FS_NODUMP_FL) 2465c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NODUMP; 2476cbff00fSChristoph Hellwig else 2485c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_NODUMP; 2495c57b8b6SDavid Sterba if (fsflags & FS_NOATIME_FL) 2505c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NOATIME; 2516cbff00fSChristoph Hellwig else 2525c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_NOATIME; 2535c57b8b6SDavid Sterba if (fsflags & FS_DIRSYNC_FL) 2545c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_DIRSYNC; 2556cbff00fSChristoph Hellwig else 2565c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_DIRSYNC; 2575c57b8b6SDavid Sterba if (fsflags & FS_NOCOW_FL) { 2587e97b8daSDavid Sterba if (S_ISREG(mode)) { 2597e97b8daSDavid Sterba /* 2607e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2617e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2627e97b8daSDavid Sterba * status of the file and will not set it. 2637e97b8daSDavid Sterba */ 2647e97b8daSDavid Sterba if (inode->i_size == 0) 2655c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NODATACOW 2667e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM; 2677e97b8daSDavid Sterba } else { 2685c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NODATACOW; 2697e97b8daSDavid Sterba } 2707e97b8daSDavid Sterba } else { 2717e97b8daSDavid Sterba /* 27201327610SNicholas D Steeves * Revert back under same assumptions as above 2737e97b8daSDavid Sterba */ 2747e97b8daSDavid Sterba if (S_ISREG(mode)) { 2757e97b8daSDavid Sterba if (inode->i_size == 0) 2765c57b8b6SDavid Sterba binode->flags &= ~(BTRFS_INODE_NODATACOW 2777e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM); 2787e97b8daSDavid Sterba } else { 2795c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_NODATACOW; 2807e97b8daSDavid Sterba } 2817e97b8daSDavid Sterba } 2826cbff00fSChristoph Hellwig 28375e7cb7fSLiu Bo /* 28475e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 28575e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 28675e7cb7fSLiu Bo * things smaller. 28775e7cb7fSLiu Bo */ 2885c57b8b6SDavid Sterba if (fsflags & FS_NOCOMP_FL) { 2895c57b8b6SDavid Sterba binode->flags &= ~BTRFS_INODE_COMPRESS; 2905c57b8b6SDavid Sterba binode->flags |= BTRFS_INODE_NOCOMPRESS; 29163541927SFilipe David Borba Manana 29263541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 29363541927SFilipe David Borba Manana if (ret && ret != -ENODATA) 29463541927SFilipe David Borba Manana goto out_drop; 2955c57b8b6SDavid Sterba } else if (fsflags & FS_COMPR_FL) { 29663541927SFilipe David Borba Manana const char *comp; 29763541927SFilipe David Borba Manana 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; 499815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 500f7039b1dSLi Dongyang int ret; 501f7039b1dSLi Dongyang 502f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 503f7039b1dSLi Dongyang return -EPERM; 504f7039b1dSLi Dongyang 5051f78160cSXiao Guangrong rcu_read_lock(); 5061f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 5071f78160cSXiao Guangrong dev_list) { 508f7039b1dSLi Dongyang if (!device->bdev) 509f7039b1dSLi Dongyang continue; 510f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 511f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 512f7039b1dSLi Dongyang num_devices++; 51350d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 514f7039b1dSLi Dongyang minlen); 515f7039b1dSLi Dongyang } 516f7039b1dSLi Dongyang } 5171f78160cSXiao Guangrong rcu_read_unlock(); 518f4c697e6SLukas Czerner 519f7039b1dSLi Dongyang if (!num_devices) 520f7039b1dSLi Dongyang return -EOPNOTSUPP; 521f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 522f7039b1dSLi Dongyang return -EFAULT; 523e515c18bSLukas Czerner if (range.start > total_bytes || 524e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 525f4c697e6SLukas Czerner return -EINVAL; 526f7039b1dSLi Dongyang 527f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 528f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 5292ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 530f7039b1dSLi Dongyang if (ret < 0) 531f7039b1dSLi Dongyang return ret; 532f7039b1dSLi Dongyang 533f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 534f7039b1dSLi Dongyang return -EFAULT; 535f7039b1dSLi Dongyang 536f7039b1dSLi Dongyang return 0; 537f7039b1dSLi Dongyang } 538f7039b1dSLi Dongyang 539dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 540dd5f9615SStefan Behrens { 54146e0f66aSChris Mason int i; 54246e0f66aSChris Mason 54346e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 54446e0f66aSChris Mason if (uuid[i]) 54546e0f66aSChris Mason return 0; 54646e0f66aSChris Mason } 54746e0f66aSChris Mason return 1; 548dd5f9615SStefan Behrens } 549dd5f9615SStefan Behrens 550d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 551cb8e7090SChristoph Hellwig struct dentry *dentry, 55252f75f4fSDavid Sterba const char *name, int namelen, 5536f72c7e2SArne Jansen u64 *async_transid, 5548696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 555f46b5a66SChristoph Hellwig { 5560b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 557f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 558f46b5a66SChristoph Hellwig struct btrfs_key key; 55949a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 560f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 561f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 562d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 56376dda93cSYan, Zheng struct btrfs_root *new_root; 564d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 565c2050a45SDeepa Dinamani struct timespec cur_time = current_time(dir); 5665662344bSTsutomu Itoh struct inode *inode; 567f46b5a66SChristoph Hellwig int ret; 568f46b5a66SChristoph Hellwig int err; 569f46b5a66SChristoph Hellwig u64 objectid; 570f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 5713de4586cSChris Mason u64 index = 0; 5728ea05e3aSAlexander Block uuid_le new_uuid; 573f46b5a66SChristoph Hellwig 57449a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 57549a3c4d9SDavid Sterba if (!root_item) 57649a3c4d9SDavid Sterba return -ENOMEM; 57749a3c4d9SDavid Sterba 5780b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5792fbe8c8aSAl Viro if (ret) 58049a3c4d9SDavid Sterba goto fail_free; 5816a912213SJosef Bacik 582e09fe2d2SQu Wenruo /* 583e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 58401327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 585e09fe2d2SQu Wenruo */ 58649a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 58749a3c4d9SDavid Sterba ret = -ENOSPC; 58849a3c4d9SDavid Sterba goto fail_free; 58949a3c4d9SDavid Sterba } 590e09fe2d2SQu Wenruo 591d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 5929ed74f2dSJosef Bacik /* 593d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 594d5c12070SMiao Xie * of create_snapshot(). 5959ed74f2dSJosef Bacik */ 596c4c129dbSGu JinXiang ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); 597d5c12070SMiao Xie if (ret) 59849a3c4d9SDavid Sterba goto fail_free; 599f46b5a66SChristoph Hellwig 600d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 601d5c12070SMiao Xie if (IS_ERR(trans)) { 602d5c12070SMiao Xie ret = PTR_ERR(trans); 6037775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 60449a3c4d9SDavid Sterba goto fail_free; 605d5c12070SMiao Xie } 606d5c12070SMiao Xie trans->block_rsv = &block_rsv; 607d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 608f46b5a66SChristoph Hellwig 6090b246afaSJeff Mahoney ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); 6106f72c7e2SArne Jansen if (ret) 6116f72c7e2SArne Jansen goto fail; 6126f72c7e2SArne Jansen 6134d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6148e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6158e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6168e8a1e31SJosef Bacik goto fail; 6178e8a1e31SJosef Bacik } 618f46b5a66SChristoph Hellwig 619b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 620f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 621f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 6225d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 623f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 624f46b5a66SChristoph Hellwig 6250b246afaSJeff Mahoney write_extent_buffer_fsid(leaf, fs_info->fsid); 6260b246afaSJeff Mahoney write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 627f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 628f46b5a66SChristoph Hellwig 62949a3c4d9SDavid Sterba inode_item = &root_item->inode; 6303cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6313cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6323cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 633da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6340b246afaSJeff Mahoney fs_info->nodesize); 6353cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 636f46b5a66SChristoph Hellwig 63749a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 63849a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6393cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 64008fe4db1SLi Zefan 64149a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 64249a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 64349a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 64449a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 64549a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 64649a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 647f46b5a66SChristoph Hellwig 64849a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 64949a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6508ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 65149a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 65249a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 65349a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 65449a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 65549a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 65649a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 657f46b5a66SChristoph Hellwig 658925baeddSChris Mason btrfs_tree_unlock(leaf); 659f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 660f46b5a66SChristoph Hellwig leaf = NULL; 661f46b5a66SChristoph Hellwig 66249a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 663f46b5a66SChristoph Hellwig 664f46b5a66SChristoph Hellwig key.objectid = objectid; 6655d4f98a2SYan Zheng key.offset = 0; 666962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6670b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 66849a3c4d9SDavid Sterba root_item); 669f46b5a66SChristoph Hellwig if (ret) 670f46b5a66SChristoph Hellwig goto fail; 671f46b5a66SChristoph Hellwig 67276dda93cSYan, Zheng key.offset = (u64)-1; 6730b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 67479787eaaSJeff Mahoney if (IS_ERR(new_root)) { 67579787eaaSJeff Mahoney ret = PTR_ERR(new_root); 67666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 67779787eaaSJeff Mahoney goto fail; 67879787eaaSJeff Mahoney } 67976dda93cSYan, Zheng 68076dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 68176dda93cSYan, Zheng 68263541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 683ce598979SMark Fasheh if (ret) { 684ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 68566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 686ce598979SMark Fasheh goto fail; 687ce598979SMark Fasheh } 688ce598979SMark Fasheh 689f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 690f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 691f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 692f32e48e9SChandan Rajendra 693f46b5a66SChristoph Hellwig /* 694f46b5a66SChristoph Hellwig * insert the directory item 695f46b5a66SChristoph Hellwig */ 696877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 69779787eaaSJeff Mahoney if (ret) { 69866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 69979787eaaSJeff Mahoney goto fail; 70079787eaaSJeff Mahoney } 7013de4586cSChris Mason 7023de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 7038e7611cfSNikolay Borisov name, namelen, BTRFS_I(dir), &key, 7043de4586cSChris Mason BTRFS_FT_DIR, index); 70579787eaaSJeff Mahoney if (ret) { 70666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 707f46b5a66SChristoph Hellwig goto fail; 70879787eaaSJeff Mahoney } 7090660b5afSChris Mason 7106ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 71152c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 71252c26179SYan Zheng BUG_ON(ret); 71352c26179SYan Zheng 7140b246afaSJeff Mahoney ret = btrfs_add_root_ref(trans, fs_info, 7154df27c4dSYan, Zheng objectid, root->root_key.objectid, 7164a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7170660b5afSChris Mason BUG_ON(ret); 7180660b5afSChris Mason 719cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7206bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 721dd5f9615SStefan Behrens if (ret) 72266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 723dd5f9615SStefan Behrens 724f46b5a66SChristoph Hellwig fail: 72549a3c4d9SDavid Sterba kfree(root_item); 726d5c12070SMiao Xie trans->block_rsv = NULL; 727d5c12070SMiao Xie trans->bytes_reserved = 0; 7287775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 729de6e8200SLiu Bo 73072fd032eSSage Weil if (async_transid) { 73172fd032eSSage Weil *async_transid = trans->transid; 7323a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 73300d71c9cSMiao Xie if (err) 7343a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73572fd032eSSage Weil } else { 7363a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73772fd032eSSage Weil } 738f46b5a66SChristoph Hellwig if (err && !ret) 739f46b5a66SChristoph Hellwig ret = err; 7401a65e24bSChris Mason 7415662344bSTsutomu Itoh if (!ret) { 7425662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 743de6e8200SLiu Bo if (IS_ERR(inode)) 744de6e8200SLiu Bo return PTR_ERR(inode); 7455662344bSTsutomu Itoh d_instantiate(dentry, inode); 7465662344bSTsutomu Itoh } 747f46b5a66SChristoph Hellwig return ret; 74849a3c4d9SDavid Sterba 74949a3c4d9SDavid Sterba fail_free: 75049a3c4d9SDavid Sterba kfree(root_item); 75149a3c4d9SDavid Sterba return ret; 752f46b5a66SChristoph Hellwig } 753f46b5a66SChristoph Hellwig 754e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 75561d7e4cbSDavid Sterba struct dentry *dentry, 756e9662f70SMiao Xie u64 *async_transid, bool readonly, 757e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 758f46b5a66SChristoph Hellwig { 7590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7602e4bfab9SYan, Zheng struct inode *inode; 761f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 762f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7632e4bfab9SYan, Zheng int ret; 764f46b5a66SChristoph Hellwig 76527cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 766f46b5a66SChristoph Hellwig return -EINVAL; 767f46b5a66SChristoph Hellwig 76823269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 769a1ee7362SDavid Sterba if (!pending_snapshot) 770a1ee7362SDavid Sterba return -ENOMEM; 771a1ee7362SDavid Sterba 772b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 77323269bf5SDavid Sterba GFP_KERNEL); 7748546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7758546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 776b0c0ea63SDavid Sterba ret = -ENOMEM; 777b0c0ea63SDavid Sterba goto free_pending; 778b0c0ea63SDavid Sterba } 779b0c0ea63SDavid Sterba 780ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7814e857c58SPeter Zijlstra smp_mb__after_atomic(); 78245bac0f3SLiu Bo /* wait for no snapshot writes */ 78345bac0f3SLiu Bo wait_event(root->subv_writers->wait, 78445bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 7858257b2dcSMiao Xie 78676f32e24SNikolay Borisov ret = btrfs_start_delalloc_inodes(root); 7876a03843dSMiao Xie if (ret) 788a1ee7362SDavid Sterba goto dec_and_free; 7896a03843dSMiao Xie 7906374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 7916a03843dSMiao Xie 79266d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 79366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 794d5c12070SMiao Xie /* 795d5c12070SMiao Xie * 1 - parent dir inode 796d5c12070SMiao Xie * 2 - dir entries 797d5c12070SMiao Xie * 1 - root item 798d5c12070SMiao Xie * 2 - root ref/backref 799d5c12070SMiao Xie * 1 - root of snapshot 800dd5f9615SStefan Behrens * 1 - UUID item 801d5c12070SMiao Xie */ 802d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 803dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 804ee3441b4SJeff Mahoney false); 805d5c12070SMiao Xie if (ret) 806a1ee7362SDavid Sterba goto dec_and_free; 807d5c12070SMiao Xie 808a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 809a22285a6SYan, Zheng pending_snapshot->root = root; 810b83cc969SLi Zefan pending_snapshot->readonly = readonly; 811e9662f70SMiao Xie pending_snapshot->dir = dir; 8128696c533SMiao Xie pending_snapshot->inherit = inherit; 813a22285a6SYan, Zheng 814d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 815a22285a6SYan, Zheng if (IS_ERR(trans)) { 816a22285a6SYan, Zheng ret = PTR_ERR(trans); 817a22285a6SYan, Zheng goto fail; 818a22285a6SYan, Zheng } 819a22285a6SYan, Zheng 8200b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 821a22285a6SYan, Zheng list_add(&pending_snapshot->list, 822a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8230b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 82472fd032eSSage Weil if (async_transid) { 82572fd032eSSage Weil *async_transid = trans->transid; 8263a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 82700d71c9cSMiao Xie if (ret) 8283a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 82972fd032eSSage Weil } else { 8303a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 83172fd032eSSage Weil } 832aec8030aSMiao Xie if (ret) 833c37b2b62SJosef Bacik goto fail; 834a22285a6SYan, Zheng 835a22285a6SYan, Zheng ret = pending_snapshot->error; 836f46b5a66SChristoph Hellwig if (ret) 8372e4bfab9SYan, Zheng goto fail; 838f46b5a66SChristoph Hellwig 839d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 840d3797308SChris Mason if (ret) 841d3797308SChris Mason goto fail; 842d3797308SChris Mason 8432b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8442e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8452e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8462e4bfab9SYan, Zheng goto fail; 8472e4bfab9SYan, Zheng } 8485662344bSTsutomu Itoh 8492e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8502e4bfab9SYan, Zheng ret = 0; 8512e4bfab9SYan, Zheng fail: 8527775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 853a1ee7362SDavid Sterba dec_and_free: 854ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8554625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 856b0c0ea63SDavid Sterba free_pending: 857b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8588546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 859a1ee7362SDavid Sterba kfree(pending_snapshot); 860a1ee7362SDavid Sterba 861f46b5a66SChristoph Hellwig return ret; 862f46b5a66SChristoph Hellwig } 863f46b5a66SChristoph Hellwig 8644260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8654260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8664260f7c7SSage Weil * whether the type of victim is right. 8674260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8684260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8694260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8704260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8714260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8724260f7c7SSage Weil * a. be owner of dir, or 8734260f7c7SSage Weil * b. be owner of victim, or 8744260f7c7SSage Weil * c. have CAP_FOWNER capability 87501327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8764260f7c7SSage Weil * links pointing to it. 8774260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8784260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8794260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8804260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8814260f7c7SSage Weil * nfs_async_unlink(). 8824260f7c7SSage Weil */ 8834260f7c7SSage Weil 8844260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 8854260f7c7SSage Weil { 8864260f7c7SSage Weil int error; 8874260f7c7SSage Weil 8882b0143b5SDavid Howells if (d_really_is_negative(victim)) 8894260f7c7SSage Weil return -ENOENT; 8904260f7c7SSage Weil 8912b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8924fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8934260f7c7SSage Weil 8944260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 8954260f7c7SSage Weil if (error) 8964260f7c7SSage Weil return error; 8974260f7c7SSage Weil if (IS_APPEND(dir)) 8984260f7c7SSage Weil return -EPERM; 8992b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9002b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9014260f7c7SSage Weil return -EPERM; 9024260f7c7SSage Weil if (isdir) { 903e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9044260f7c7SSage Weil return -ENOTDIR; 9054260f7c7SSage Weil if (IS_ROOT(victim)) 9064260f7c7SSage Weil return -EBUSY; 907e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9084260f7c7SSage Weil return -EISDIR; 9094260f7c7SSage Weil if (IS_DEADDIR(dir)) 9104260f7c7SSage Weil return -ENOENT; 9114260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9124260f7c7SSage Weil return -EBUSY; 9134260f7c7SSage Weil return 0; 9144260f7c7SSage Weil } 9154260f7c7SSage Weil 916cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 917cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 918cb8e7090SChristoph Hellwig { 9192b0143b5SDavid Howells if (d_really_is_positive(child)) 920cb8e7090SChristoph Hellwig return -EEXIST; 921cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 922cb8e7090SChristoph Hellwig return -ENOENT; 923cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 924cb8e7090SChristoph Hellwig } 925cb8e7090SChristoph Hellwig 926cb8e7090SChristoph Hellwig /* 927cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 928cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 929cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 930cb8e7090SChristoph Hellwig */ 93192872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 93252f75f4fSDavid Sterba const char *name, int namelen, 93372fd032eSSage Weil struct btrfs_root *snap_src, 9346f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9358696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 936cb8e7090SChristoph Hellwig { 9372b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9380b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 939cb8e7090SChristoph Hellwig struct dentry *dentry; 940cb8e7090SChristoph Hellwig int error; 941cb8e7090SChristoph Hellwig 94200235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 94300235411SAl Viro if (error == -EINTR) 94400235411SAl Viro return error; 945cb8e7090SChristoph Hellwig 946cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 947cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 948cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 949cb8e7090SChristoph Hellwig goto out_unlock; 950cb8e7090SChristoph Hellwig 95176dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 952cb8e7090SChristoph Hellwig if (error) 953a874a63eSLiu Bo goto out_dput; 954cb8e7090SChristoph Hellwig 9559c52057cSChris Mason /* 9569c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9579c52057cSChris Mason * check for them now when we can safely fail 9589c52057cSChris Mason */ 9599c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9609c52057cSChris Mason dir->i_ino, name, 9619c52057cSChris Mason namelen); 9629c52057cSChris Mason if (error) 9639c52057cSChris Mason goto out_dput; 9649c52057cSChris Mason 9650b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 96676dda93cSYan, Zheng 96776dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 96876dda93cSYan, Zheng goto out_up_read; 96976dda93cSYan, Zheng 9703de4586cSChris Mason if (snap_src) { 97161d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9726f72c7e2SArne Jansen async_transid, readonly, inherit); 9733de4586cSChris Mason } else { 974d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 975d5c12070SMiao Xie async_transid, inherit); 9763de4586cSChris Mason } 97776dda93cSYan, Zheng if (!error) 97876dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 97976dda93cSYan, Zheng out_up_read: 9800b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 981cb8e7090SChristoph Hellwig out_dput: 982cb8e7090SChristoph Hellwig dput(dentry); 983cb8e7090SChristoph Hellwig out_unlock: 9845955102cSAl Viro inode_unlock(dir); 985cb8e7090SChristoph Hellwig return error; 986cb8e7090SChristoph Hellwig } 987cb8e7090SChristoph Hellwig 9884cb5300bSChris Mason /* 9894cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 9904cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 9914cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 9924cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 9934cb5300bSChris Mason * part of the file 9944cb5300bSChris Mason */ 995aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 9964cb5300bSChris Mason { 9974cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9984cb5300bSChris Mason struct extent_map *em = NULL; 9994cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10004cb5300bSChris Mason u64 end; 10014cb5300bSChris Mason 10024cb5300bSChris Mason read_lock(&em_tree->lock); 100309cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10044cb5300bSChris Mason read_unlock(&em_tree->lock); 10054cb5300bSChris Mason 10064cb5300bSChris Mason if (em) { 10074cb5300bSChris Mason end = extent_map_end(em); 10084cb5300bSChris Mason free_extent_map(em); 10094cb5300bSChris Mason if (end - offset > thresh) 10104cb5300bSChris Mason return 0; 10114cb5300bSChris Mason } 10124cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10134cb5300bSChris Mason thresh /= 2; 10144cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10154cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10164cb5300bSChris Mason if (end >= thresh) 10174cb5300bSChris Mason return 0; 10184cb5300bSChris Mason return 1; 10194cb5300bSChris Mason } 10204cb5300bSChris Mason 10214cb5300bSChris Mason /* 10224cb5300bSChris Mason * helper function to walk through a file and find extents 10234cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10244cb5300bSChris Mason * 10254cb5300bSChris Mason * This is used by the defragging code to find new and small 10264cb5300bSChris Mason * extents 10274cb5300bSChris Mason */ 10284cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10294cb5300bSChris Mason struct inode *inode, u64 newer_than, 1030aab110abSDavid Sterba u64 *off, u32 thresh) 10314cb5300bSChris Mason { 10324cb5300bSChris Mason struct btrfs_path *path; 10334cb5300bSChris Mason struct btrfs_key min_key; 10344cb5300bSChris Mason struct extent_buffer *leaf; 10354cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10364cb5300bSChris Mason int type; 10374cb5300bSChris Mason int ret; 10384a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10394cb5300bSChris Mason 10404cb5300bSChris Mason path = btrfs_alloc_path(); 10414cb5300bSChris Mason if (!path) 10424cb5300bSChris Mason return -ENOMEM; 10434cb5300bSChris Mason 1044a4689d2bSDavid Sterba min_key.objectid = ino; 10454cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10464cb5300bSChris Mason min_key.offset = *off; 10474cb5300bSChris Mason 10484cb5300bSChris Mason while (1) { 10496174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10504cb5300bSChris Mason if (ret != 0) 10514cb5300bSChris Mason goto none; 1052f094c9bdSFilipe Manana process_slot: 1053a4689d2bSDavid Sterba if (min_key.objectid != ino) 10544cb5300bSChris Mason goto none; 10554cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10564cb5300bSChris Mason goto none; 10574cb5300bSChris Mason 10584cb5300bSChris Mason leaf = path->nodes[0]; 10594cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10604cb5300bSChris Mason struct btrfs_file_extent_item); 10614cb5300bSChris Mason 10624cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10634cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10644cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10654cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10664cb5300bSChris Mason *off = min_key.offset; 10674cb5300bSChris Mason btrfs_free_path(path); 10684cb5300bSChris Mason return 0; 10694cb5300bSChris Mason } 10704cb5300bSChris Mason 1071f094c9bdSFilipe Manana path->slots[0]++; 1072f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1073f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1074f094c9bdSFilipe Manana goto process_slot; 1075f094c9bdSFilipe Manana } 1076f094c9bdSFilipe Manana 10774cb5300bSChris Mason if (min_key.offset == (u64)-1) 10784cb5300bSChris Mason goto none; 10794cb5300bSChris Mason 10804cb5300bSChris Mason min_key.offset++; 10814cb5300bSChris Mason btrfs_release_path(path); 10824cb5300bSChris Mason } 10834cb5300bSChris Mason none: 10844cb5300bSChris Mason btrfs_free_path(path); 10854cb5300bSChris Mason return -ENOENT; 10864cb5300bSChris Mason } 10874cb5300bSChris Mason 10886c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 108917ce6ef8SLiu Bo { 109017ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1091940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10926c282eb4SLi Zefan struct extent_map *em; 109309cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1094940100a4SChris Mason 1095940100a4SChris Mason /* 1096940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1097940100a4SChris Mason * the full extent lock 1098940100a4SChris Mason */ 1099940100a4SChris Mason read_lock(&em_tree->lock); 1100940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1101940100a4SChris Mason read_unlock(&em_tree->lock); 1102940100a4SChris Mason 1103940100a4SChris Mason if (!em) { 1104308d9800SFilipe Manana struct extent_state *cached = NULL; 1105308d9800SFilipe Manana u64 end = start + len - 1; 1106308d9800SFilipe Manana 1107940100a4SChris Mason /* get the big lock and read metadata off disk */ 1108ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1109fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1110e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1111940100a4SChris Mason 11126cf8bfbfSDan Carpenter if (IS_ERR(em)) 11136c282eb4SLi Zefan return NULL; 1114940100a4SChris Mason } 1115940100a4SChris Mason 11166c282eb4SLi Zefan return em; 11176c282eb4SLi Zefan } 11186c282eb4SLi Zefan 11196c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11206c282eb4SLi Zefan { 11216c282eb4SLi Zefan struct extent_map *next; 11226c282eb4SLi Zefan bool ret = true; 11236c282eb4SLi Zefan 11246c282eb4SLi Zefan /* this is the last extent */ 11256c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11266c282eb4SLi Zefan return false; 11276c282eb4SLi Zefan 11286c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1129e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1130e9512d72SChris Mason ret = false; 1131e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1132ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11336c282eb4SLi Zefan ret = false; 11346c282eb4SLi Zefan 11356c282eb4SLi Zefan free_extent_map(next); 11366c282eb4SLi Zefan return ret; 11376c282eb4SLi Zefan } 11386c282eb4SLi Zefan 1139aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1140a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1141a43a2111SAndrew Mahone int compress) 11426c282eb4SLi Zefan { 11436c282eb4SLi Zefan struct extent_map *em; 11446c282eb4SLi Zefan int ret = 1; 11456c282eb4SLi Zefan bool next_mergeable = true; 11464a3560c4SLiu Bo bool prev_mergeable = true; 11476c282eb4SLi Zefan 11486c282eb4SLi Zefan /* 11496c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11506c282eb4SLi Zefan * defragging it 11516c282eb4SLi Zefan */ 11526c282eb4SLi Zefan if (start < *defrag_end) 11536c282eb4SLi Zefan return 1; 11546c282eb4SLi Zefan 11556c282eb4SLi Zefan *skip = 0; 11566c282eb4SLi Zefan 11576c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11586c282eb4SLi Zefan if (!em) 11596c282eb4SLi Zefan return 0; 11606c282eb4SLi Zefan 1161940100a4SChris Mason /* this will cover holes, and inline extents */ 116217ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1163940100a4SChris Mason ret = 0; 116417ce6ef8SLiu Bo goto out; 116517ce6ef8SLiu Bo } 116617ce6ef8SLiu Bo 11674a3560c4SLiu Bo if (!*defrag_end) 11684a3560c4SLiu Bo prev_mergeable = false; 11694a3560c4SLiu Bo 11706c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1171940100a4SChris Mason /* 11726c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11736c282eb4SLi Zefan * real extent, don't bother defragging it 1174940100a4SChris Mason */ 1175a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11764a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1177940100a4SChris Mason ret = 0; 117817ce6ef8SLiu Bo out: 1179940100a4SChris Mason /* 1180940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1181940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1182940100a4SChris Mason * so that the next tiny extent will force a defrag. 1183940100a4SChris Mason * 1184940100a4SChris Mason * The end result of this is that tiny extents before a single big 1185940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1186940100a4SChris Mason */ 1187940100a4SChris Mason if (ret) { 1188940100a4SChris Mason *defrag_end = extent_map_end(em); 1189940100a4SChris Mason } else { 1190940100a4SChris Mason *last_len = 0; 1191940100a4SChris Mason *skip = extent_map_end(em); 1192940100a4SChris Mason *defrag_end = 0; 1193940100a4SChris Mason } 1194940100a4SChris Mason 1195940100a4SChris Mason free_extent_map(em); 1196940100a4SChris Mason return ret; 1197940100a4SChris Mason } 1198940100a4SChris Mason 11994cb5300bSChris Mason /* 12004cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12014cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12024cb5300bSChris Mason * to COW and defrag. 12034cb5300bSChris Mason * 12044cb5300bSChris Mason * It also makes sure the delalloc code has enough 12054cb5300bSChris Mason * dirty data to avoid making new small extents as part 12064cb5300bSChris Mason * of the defrag 12074cb5300bSChris Mason * 12084cb5300bSChris Mason * It's a good idea to start RA on this range 12094cb5300bSChris Mason * before calling this. 12104cb5300bSChris Mason */ 12114cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12124cb5300bSChris Mason struct page **pages, 12134cb5300bSChris Mason unsigned long start_index, 1214c41570c9SJustin Maggard unsigned long num_pages) 1215f46b5a66SChristoph Hellwig { 12164cb5300bSChris Mason unsigned long file_end; 12174cb5300bSChris Mason u64 isize = i_size_read(inode); 1218f46b5a66SChristoph Hellwig u64 page_start; 1219f46b5a66SChristoph Hellwig u64 page_end; 12201f12bd06SLiu Bo u64 page_cnt; 12214cb5300bSChris Mason int ret; 12224cb5300bSChris Mason int i; 12234cb5300bSChris Mason int i_done; 12244cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12254cb5300bSChris Mason struct extent_state *cached_state = NULL; 1226600a45e1SMiao Xie struct extent_io_tree *tree; 1227364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12283b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12294cb5300bSChris Mason 123009cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12311f12bd06SLiu Bo if (!isize || start_index > file_end) 12321f12bd06SLiu Bo return 0; 12331f12bd06SLiu Bo 12341f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12354cb5300bSChris Mason 1236364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 123709cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 123809cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12394cb5300bSChris Mason if (ret) 12404cb5300bSChris Mason return ret; 12414cb5300bSChris Mason i_done = 0; 1242600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12434cb5300bSChris Mason 12444cb5300bSChris Mason /* step one, lock all the pages */ 12451f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12464cb5300bSChris Mason struct page *page; 1247600a45e1SMiao Xie again: 1248a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12493b16a4e3SJosef Bacik start_index + i, mask); 12504cb5300bSChris Mason if (!page) 12514cb5300bSChris Mason break; 12524cb5300bSChris Mason 1253600a45e1SMiao Xie page_start = page_offset(page); 125409cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1255600a45e1SMiao Xie while (1) { 1256308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1257ff13db41SDavid Sterba &cached_state); 1258600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1259600a45e1SMiao Xie page_start); 1260308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1261e43bbe5eSDavid Sterba &cached_state); 1262600a45e1SMiao Xie if (!ordered) 1263600a45e1SMiao Xie break; 1264600a45e1SMiao Xie 1265600a45e1SMiao Xie unlock_page(page); 1266600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1267600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1268600a45e1SMiao Xie lock_page(page); 12691f12bd06SLiu Bo /* 12701f12bd06SLiu Bo * we unlocked the page above, so we need check if 12711f12bd06SLiu Bo * it was released or not. 12721f12bd06SLiu Bo */ 12731f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12741f12bd06SLiu Bo unlock_page(page); 127509cbfeafSKirill A. Shutemov put_page(page); 12761f12bd06SLiu Bo goto again; 12771f12bd06SLiu Bo } 1278600a45e1SMiao Xie } 1279600a45e1SMiao Xie 12804cb5300bSChris Mason if (!PageUptodate(page)) { 12814cb5300bSChris Mason btrfs_readpage(NULL, page); 12824cb5300bSChris Mason lock_page(page); 12834cb5300bSChris Mason if (!PageUptodate(page)) { 12844cb5300bSChris Mason unlock_page(page); 128509cbfeafSKirill A. Shutemov put_page(page); 12864cb5300bSChris Mason ret = -EIO; 12874cb5300bSChris Mason break; 12884cb5300bSChris Mason } 12894cb5300bSChris Mason } 1290600a45e1SMiao Xie 1291600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1292600a45e1SMiao Xie unlock_page(page); 129309cbfeafSKirill A. Shutemov put_page(page); 1294600a45e1SMiao Xie goto again; 1295600a45e1SMiao Xie } 1296600a45e1SMiao Xie 12974cb5300bSChris Mason pages[i] = page; 12984cb5300bSChris Mason i_done++; 12994cb5300bSChris Mason } 13004cb5300bSChris Mason if (!i_done || ret) 13014cb5300bSChris Mason goto out; 13024cb5300bSChris Mason 13031751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13044cb5300bSChris Mason goto out; 13054cb5300bSChris Mason 13064cb5300bSChris Mason /* 13074cb5300bSChris Mason * so now we have a nice long stream of locked 13084cb5300bSChris Mason * and up to date pages, lets wait on them 13094cb5300bSChris Mason */ 13104cb5300bSChris Mason for (i = 0; i < i_done; i++) 13114cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13124cb5300bSChris Mason 13134cb5300bSChris Mason page_start = page_offset(pages[0]); 131409cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13154cb5300bSChris Mason 13164cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1317ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13184cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 13194cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 13209e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1321ae0f1625SDavid Sterba &cached_state); 13224cb5300bSChris Mason 13231f12bd06SLiu Bo if (i_done != page_cnt) { 13249e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 13259e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 13269e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1327bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 132809cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 132943b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13304cb5300bSChris Mason } 13314cb5300bSChris Mason 13324cb5300bSChris Mason 13339e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1334018ed4f7SDavid Sterba &cached_state); 13354cb5300bSChris Mason 13364cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1337e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13384cb5300bSChris Mason 13394cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13404cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13414cb5300bSChris Mason ClearPageChecked(pages[i]); 13424cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13434cb5300bSChris Mason set_page_dirty(pages[i]); 13444cb5300bSChris Mason unlock_page(pages[i]); 134509cbfeafSKirill A. Shutemov put_page(pages[i]); 13464cb5300bSChris Mason } 134743b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 134843b18595SQu Wenruo false); 1349364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13504cb5300bSChris Mason return i_done; 13514cb5300bSChris Mason out: 13524cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13534cb5300bSChris Mason unlock_page(pages[i]); 135409cbfeafSKirill A. Shutemov put_page(pages[i]); 13554cb5300bSChris Mason } 1356bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 135709cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 135843b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 135943b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 136043b18595SQu Wenruo true); 1361364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13624cb5300bSChris Mason return ret; 13634cb5300bSChris Mason 13644cb5300bSChris Mason } 13654cb5300bSChris Mason 13664cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13674cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13684cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13694cb5300bSChris Mason { 13700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13714cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13724cb5300bSChris Mason struct file_ra_state *ra = NULL; 13734cb5300bSChris Mason unsigned long last_index; 1374151a31b2SLi Zefan u64 isize = i_size_read(inode); 1375940100a4SChris Mason u64 last_len = 0; 1376940100a4SChris Mason u64 skip = 0; 1377940100a4SChris Mason u64 defrag_end = 0; 13784cb5300bSChris Mason u64 newer_off = range->start; 1379f46b5a66SChristoph Hellwig unsigned long i; 1380008873eaSLi Zefan unsigned long ra_index = 0; 1381f46b5a66SChristoph Hellwig int ret; 13824cb5300bSChris Mason int defrag_count = 0; 13831a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1384aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 138509cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1386c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1387ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 13884cb5300bSChris Mason struct page **pages = NULL; 13891e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 13904cb5300bSChris Mason 13910abd5b17SLiu Bo if (isize == 0) 13920abd5b17SLiu Bo return 0; 13930abd5b17SLiu Bo 13940abd5b17SLiu Bo if (range->start >= isize) 13950abd5b17SLiu Bo return -EINVAL; 13961a419d85SLi Zefan 13971e2ef46dSDavid Sterba if (do_compress) { 13981a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 13991a419d85SLi Zefan return -EINVAL; 14001a419d85SLi Zefan if (range->compress_type) 14011a419d85SLi Zefan compress_type = range->compress_type; 14021a419d85SLi Zefan } 1403f46b5a66SChristoph Hellwig 14040abd5b17SLiu Bo if (extent_thresh == 0) 1405ee22184bSByongho Lee extent_thresh = SZ_256K; 1406f46b5a66SChristoph Hellwig 14074cb5300bSChris Mason /* 14080a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14090a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14100a52d108SDavid Sterba * we don't error out. 14114cb5300bSChris Mason */ 14124cb5300bSChris Mason if (!file) { 141363e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14140a52d108SDavid Sterba if (ra) 14154cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14164cb5300bSChris Mason } else { 14174cb5300bSChris Mason ra = &file->f_ra; 14184cb5300bSChris Mason } 14194cb5300bSChris Mason 142063e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14214cb5300bSChris Mason if (!pages) { 14224cb5300bSChris Mason ret = -ENOMEM; 14234cb5300bSChris Mason goto out_ra; 14244cb5300bSChris Mason } 14254cb5300bSChris Mason 14264cb5300bSChris Mason /* find the last page to defrag */ 14271e701a32SChris Mason if (range->start + range->len > range->start) { 1428151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 142909cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14301e701a32SChris Mason } else { 143109cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14321e701a32SChris Mason } 14331e701a32SChris Mason 14344cb5300bSChris Mason if (newer_than) { 14354cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1436ee22184bSByongho Lee &newer_off, SZ_64K); 14374cb5300bSChris Mason if (!ret) { 14384cb5300bSChris Mason range->start = newer_off; 14394cb5300bSChris Mason /* 14404cb5300bSChris Mason * we always align our defrag to help keep 14414cb5300bSChris Mason * the extents in the file evenly spaced 14424cb5300bSChris Mason */ 144309cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14444cb5300bSChris Mason } else 14454cb5300bSChris Mason goto out_ra; 14464cb5300bSChris Mason } else { 144709cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14484cb5300bSChris Mason } 14494cb5300bSChris Mason if (!max_to_defrag) 1450070034bdSchandan max_to_defrag = last_index - i + 1; 14514cb5300bSChris Mason 14522a0f7f57SLi Zefan /* 14532a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14542a0f7f57SLi Zefan * written sequentially. 14552a0f7f57SLi Zefan */ 14562a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14572a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14582a0f7f57SLi Zefan 1459f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 146009cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14614cb5300bSChris Mason /* 14624cb5300bSChris Mason * make sure we stop running if someone unmounts 14634cb5300bSChris Mason * the FS 14644cb5300bSChris Mason */ 14651751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14664cb5300bSChris Mason break; 14674cb5300bSChris Mason 14680b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14690b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1470210549ebSDavid Sterba ret = -EAGAIN; 1471210549ebSDavid Sterba break; 1472210549ebSDavid Sterba } 1473210549ebSDavid Sterba 147409cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14756c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14761e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1477940100a4SChris Mason unsigned long next; 1478940100a4SChris Mason /* 1479940100a4SChris Mason * the should_defrag function tells us how much to skip 1480940100a4SChris Mason * bump our counter by the suggested amount 1481940100a4SChris Mason */ 148209cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1483940100a4SChris Mason i = max(i + 1, next); 1484940100a4SChris Mason continue; 1485940100a4SChris Mason } 1486008873eaSLi Zefan 1487008873eaSLi Zefan if (!newer_than) { 148809cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 148909cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1490008873eaSLi Zefan cluster = min(cluster, max_cluster); 1491008873eaSLi Zefan } else { 1492008873eaSLi Zefan cluster = max_cluster; 1493008873eaSLi Zefan } 1494008873eaSLi Zefan 1495008873eaSLi Zefan if (i + cluster > ra_index) { 1496008873eaSLi Zefan ra_index = max(i, ra_index); 14970a52d108SDavid Sterba if (ra) 1498d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1499d3c0bab5SDavid Sterba file, ra_index, cluster); 1500e4826a5bSchandan ra_index += cluster; 1501008873eaSLi Zefan } 15024cb5300bSChris Mason 15035955102cSAl Viro inode_lock(inode); 15041e2ef46dSDavid Sterba if (do_compress) 1505eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1506008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1507ecb8bea8SLiu Bo if (ret < 0) { 15085955102cSAl Viro inode_unlock(inode); 15094cb5300bSChris Mason goto out_ra; 1510ecb8bea8SLiu Bo } 15114cb5300bSChris Mason 15124cb5300bSChris Mason defrag_count += ret; 1513d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15145955102cSAl Viro inode_unlock(inode); 15154cb5300bSChris Mason 15164cb5300bSChris Mason if (newer_than) { 15174cb5300bSChris Mason if (newer_off == (u64)-1) 15184cb5300bSChris Mason break; 15194cb5300bSChris Mason 1520e1f041e1SLiu Bo if (ret > 0) 1521e1f041e1SLiu Bo i += ret; 1522e1f041e1SLiu Bo 15234cb5300bSChris Mason newer_off = max(newer_off + 1, 152409cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15254cb5300bSChris Mason 1526ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1527ee22184bSByongho Lee &newer_off, SZ_64K); 15284cb5300bSChris Mason if (!ret) { 15294cb5300bSChris Mason range->start = newer_off; 153009cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15314cb5300bSChris Mason } else { 15324cb5300bSChris Mason break; 1533940100a4SChris Mason } 15344cb5300bSChris Mason } else { 1535008873eaSLi Zefan if (ret > 0) { 1536cbcc8326SLi Zefan i += ret; 153709cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1538008873eaSLi Zefan } else { 1539940100a4SChris Mason i++; 1540008873eaSLi Zefan last_len = 0; 1541008873eaSLi Zefan } 1542f46b5a66SChristoph Hellwig } 15434cb5300bSChris Mason } 1544f46b5a66SChristoph Hellwig 1545dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15461e701a32SChris Mason filemap_flush(inode->i_mapping); 1547dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1548dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1549dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1550dec8ef90SFilipe Manana } 15511e701a32SChris Mason 15521a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15530b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15545c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15555c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15561a419d85SLi Zefan } 15571a419d85SLi Zefan 155860ccf82fSDiego Calleja ret = defrag_count; 1559940100a4SChris Mason 15604cb5300bSChris Mason out_ra: 15611e2ef46dSDavid Sterba if (do_compress) { 15625955102cSAl Viro inode_lock(inode); 1563eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15645955102cSAl Viro inode_unlock(inode); 1565633085c7SFilipe David Borba Manana } 15664cb5300bSChris Mason if (!file) 15674cb5300bSChris Mason kfree(ra); 15684cb5300bSChris Mason kfree(pages); 1569940100a4SChris Mason return ret; 1570f46b5a66SChristoph Hellwig } 1571f46b5a66SChristoph Hellwig 1572198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 157376dda93cSYan, Zheng void __user *arg) 1574f46b5a66SChristoph Hellwig { 15750b246afaSJeff Mahoney struct inode *inode = file_inode(file); 15760b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1577f46b5a66SChristoph Hellwig u64 new_size; 1578f46b5a66SChristoph Hellwig u64 old_size; 1579f46b5a66SChristoph Hellwig u64 devid = 1; 15800b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1581f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1582f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1583f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1584f46b5a66SChristoph Hellwig char *sizestr; 15859a40f122SGui Hecheng char *retptr; 1586f46b5a66SChristoph Hellwig char *devstr = NULL; 1587f46b5a66SChristoph Hellwig int ret = 0; 1588f46b5a66SChristoph Hellwig int mod = 0; 1589f46b5a66SChristoph Hellwig 1590e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1591e441d54dSChris Mason return -EPERM; 1592e441d54dSChris Mason 1593198605a8SMiao Xie ret = mnt_want_write_file(file); 1594198605a8SMiao Xie if (ret) 1595198605a8SMiao Xie return ret; 1596198605a8SMiao Xie 1597171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 159897547676SMiao Xie mnt_drop_write_file(file); 1599e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1600c9e9f97bSIlya Dryomov } 1601c9e9f97bSIlya Dryomov 1602dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1603c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1604c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1605c9e9f97bSIlya Dryomov goto out; 1606c9e9f97bSIlya Dryomov } 16075516e595SMark Fasheh 16085516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1609f46b5a66SChristoph Hellwig 1610f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1611f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1612f46b5a66SChristoph Hellwig if (devstr) { 1613f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1614f46b5a66SChristoph Hellwig *devstr = '\0'; 1615f46b5a66SChristoph Hellwig devstr = vol_args->name; 161658dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 161758dfae63SZhangZhen if (ret) 161858dfae63SZhangZhen goto out_free; 1619dfd79829SMiao Xie if (!devid) { 1620dfd79829SMiao Xie ret = -EINVAL; 1621dfd79829SMiao Xie goto out_free; 1622dfd79829SMiao Xie } 16230b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1624f46b5a66SChristoph Hellwig } 1625dba60f3fSMiao Xie 16260b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1627f46b5a66SChristoph Hellwig if (!device) { 16280b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1629c1c9ff7cSGeert Uytterhoeven devid); 1630dfd79829SMiao Xie ret = -ENODEV; 1631c9e9f97bSIlya Dryomov goto out_free; 1632f46b5a66SChristoph Hellwig } 1633dba60f3fSMiao Xie 1634ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16350b246afaSJeff Mahoney btrfs_info(fs_info, 1636efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1637c1c9ff7cSGeert Uytterhoeven devid); 1638dfd79829SMiao Xie ret = -EPERM; 16394e42ae1bSLiu Bo goto out_free; 16404e42ae1bSLiu Bo } 16414e42ae1bSLiu Bo 1642f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1643f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1644f46b5a66SChristoph Hellwig else { 1645f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1646f46b5a66SChristoph Hellwig mod = -1; 1647f46b5a66SChristoph Hellwig sizestr++; 1648f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1649f46b5a66SChristoph Hellwig mod = 1; 1650f46b5a66SChristoph Hellwig sizestr++; 1651f46b5a66SChristoph Hellwig } 16529a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16539a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1654f46b5a66SChristoph Hellwig ret = -EINVAL; 1655c9e9f97bSIlya Dryomov goto out_free; 1656f46b5a66SChristoph Hellwig } 1657f46b5a66SChristoph Hellwig } 1658f46b5a66SChristoph Hellwig 1659401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1660dfd79829SMiao Xie ret = -EPERM; 166163a212abSStefan Behrens goto out_free; 166263a212abSStefan Behrens } 166363a212abSStefan Behrens 16647cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1665f46b5a66SChristoph Hellwig 1666f46b5a66SChristoph Hellwig if (mod < 0) { 1667f46b5a66SChristoph Hellwig if (new_size > old_size) { 1668f46b5a66SChristoph Hellwig ret = -EINVAL; 1669c9e9f97bSIlya Dryomov goto out_free; 1670f46b5a66SChristoph Hellwig } 1671f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1672f46b5a66SChristoph Hellwig } else if (mod > 0) { 1673eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1674902c68a4SGui Hecheng ret = -ERANGE; 1675eb8052e0SWenliang Fan goto out_free; 1676eb8052e0SWenliang Fan } 1677f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1678f46b5a66SChristoph Hellwig } 1679f46b5a66SChristoph Hellwig 1680ee22184bSByongho Lee if (new_size < SZ_256M) { 1681f46b5a66SChristoph Hellwig ret = -EINVAL; 1682c9e9f97bSIlya Dryomov goto out_free; 1683f46b5a66SChristoph Hellwig } 1684f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1685f46b5a66SChristoph Hellwig ret = -EFBIG; 1686c9e9f97bSIlya Dryomov goto out_free; 1687f46b5a66SChristoph Hellwig } 1688f46b5a66SChristoph Hellwig 168947f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1690f46b5a66SChristoph Hellwig 16910b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1692c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1693f46b5a66SChristoph Hellwig 1694f46b5a66SChristoph Hellwig if (new_size > old_size) { 1695a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 169698d5dc13STsutomu Itoh if (IS_ERR(trans)) { 169798d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1698c9e9f97bSIlya Dryomov goto out_free; 169998d5dc13STsutomu Itoh } 1700f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17013a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1702ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1703f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17040253f40eSjeff.liu } /* equal, nothing need to do */ 1705f46b5a66SChristoph Hellwig 1706c9e9f97bSIlya Dryomov out_free: 1707f46b5a66SChristoph Hellwig kfree(vol_args); 1708c9e9f97bSIlya Dryomov out: 1709171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 171018f39c41SIlya Dryomov mnt_drop_write_file(file); 1711f46b5a66SChristoph Hellwig return ret; 1712f46b5a66SChristoph Hellwig } 1713f46b5a66SChristoph Hellwig 171472fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 171552f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17166f72c7e2SArne Jansen u64 *transid, bool readonly, 17178696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1718f46b5a66SChristoph Hellwig { 1719f46b5a66SChristoph Hellwig int namelen; 17203de4586cSChris Mason int ret = 0; 1721f46b5a66SChristoph Hellwig 1722325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1723325c50e3SJeff Mahoney return -ENOTDIR; 1724325c50e3SJeff Mahoney 1725a874a63eSLiu Bo ret = mnt_want_write_file(file); 1726a874a63eSLiu Bo if (ret) 1727a874a63eSLiu Bo goto out; 1728a874a63eSLiu Bo 172972fd032eSSage Weil namelen = strlen(name); 173072fd032eSSage Weil if (strchr(name, '/')) { 1731f46b5a66SChristoph Hellwig ret = -EINVAL; 1732a874a63eSLiu Bo goto out_drop_write; 1733f46b5a66SChristoph Hellwig } 1734f46b5a66SChristoph Hellwig 173516780cabSChris Mason if (name[0] == '.' && 173616780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 173716780cabSChris Mason ret = -EEXIST; 1738a874a63eSLiu Bo goto out_drop_write; 173916780cabSChris Mason } 174016780cabSChris Mason 17413de4586cSChris Mason if (subvol) { 174272fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17436f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1744cb8e7090SChristoph Hellwig } else { 17452903ff01SAl Viro struct fd src = fdget(fd); 17463de4586cSChris Mason struct inode *src_inode; 17472903ff01SAl Viro if (!src.file) { 17483de4586cSChris Mason ret = -EINVAL; 1749a874a63eSLiu Bo goto out_drop_write; 17503de4586cSChris Mason } 17513de4586cSChris Mason 1752496ad9aaSAl Viro src_inode = file_inode(src.file); 1753496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1754c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1755efe120a0SFrank Holton "Snapshot src from another FS"); 175623ad5b17SKusanagi Kouichi ret = -EXDEV; 1757d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1758d0242061SDavid Sterba /* 1759d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1760d0242061SDavid Sterba * are limited to own subvolumes only 1761d0242061SDavid Sterba */ 1762d0242061SDavid Sterba ret = -EPERM; 1763ecd18815SAl Viro } else { 176472fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 176572fd032eSSage Weil BTRFS_I(src_inode)->root, 17666f72c7e2SArne Jansen transid, readonly, inherit); 1767ecd18815SAl Viro } 17682903ff01SAl Viro fdput(src); 1769cb8e7090SChristoph Hellwig } 1770a874a63eSLiu Bo out_drop_write: 1771a874a63eSLiu Bo mnt_drop_write_file(file); 1772f46b5a66SChristoph Hellwig out: 177372fd032eSSage Weil return ret; 177472fd032eSSage Weil } 177572fd032eSSage Weil 177672fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1777fa0d2b9bSLi Zefan void __user *arg, int subvol) 177872fd032eSSage Weil { 1779fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 178072fd032eSSage Weil int ret; 178172fd032eSSage Weil 1782325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1783325c50e3SJeff Mahoney return -ENOTDIR; 1784325c50e3SJeff Mahoney 1785fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1786fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1787fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1788fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1789fa0d2b9bSLi Zefan 1790fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1791b83cc969SLi Zefan vol_args->fd, subvol, 17926f72c7e2SArne Jansen NULL, false, NULL); 1793fa0d2b9bSLi Zefan 1794fa0d2b9bSLi Zefan kfree(vol_args); 1795fa0d2b9bSLi Zefan return ret; 1796fa0d2b9bSLi Zefan } 1797fa0d2b9bSLi Zefan 1798fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1799fa0d2b9bSLi Zefan void __user *arg, int subvol) 1800fa0d2b9bSLi Zefan { 1801fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1802fa0d2b9bSLi Zefan int ret; 1803fdfb1e4fSLi Zefan u64 transid = 0; 1804fdfb1e4fSLi Zefan u64 *ptr = NULL; 1805b83cc969SLi Zefan bool readonly = false; 18066f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 180772fd032eSSage Weil 1808325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1809325c50e3SJeff Mahoney return -ENOTDIR; 1810325c50e3SJeff Mahoney 1811fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1812fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1813fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1814fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1815fdfb1e4fSLi Zefan 1816b83cc969SLi Zefan if (vol_args->flags & 18176f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18186f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1819b83cc969SLi Zefan ret = -EOPNOTSUPP; 1820c47ca32dSDan Carpenter goto free_args; 1821fdfb1e4fSLi Zefan } 1822fdfb1e4fSLi Zefan 1823fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1824fdfb1e4fSLi Zefan ptr = &transid; 1825b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1826b83cc969SLi Zefan readonly = true; 18276f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 182809cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18296f72c7e2SArne Jansen ret = -EINVAL; 1830c47ca32dSDan Carpenter goto free_args; 18316f72c7e2SArne Jansen } 18326f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18336f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18346f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1835c47ca32dSDan Carpenter goto free_args; 18366f72c7e2SArne Jansen } 18376f72c7e2SArne Jansen } 183875eaa0e2SSage Weil 1839fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18406f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18418696c533SMiao Xie readonly, inherit); 1842c47ca32dSDan Carpenter if (ret) 1843c47ca32dSDan Carpenter goto free_inherit; 184475eaa0e2SSage Weil 1845c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1846fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1847c47ca32dSDan Carpenter transid), 1848c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 184975eaa0e2SSage Weil ret = -EFAULT; 1850c47ca32dSDan Carpenter 1851c47ca32dSDan Carpenter free_inherit: 18526f72c7e2SArne Jansen kfree(inherit); 1853c47ca32dSDan Carpenter free_args: 1854c47ca32dSDan Carpenter kfree(vol_args); 1855f46b5a66SChristoph Hellwig return ret; 1856f46b5a66SChristoph Hellwig } 1857f46b5a66SChristoph Hellwig 18580caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18590caa102dSLi Zefan void __user *arg) 18600caa102dSLi Zefan { 1861496ad9aaSAl Viro struct inode *inode = file_inode(file); 18620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18630caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18640caa102dSLi Zefan int ret = 0; 18650caa102dSLi Zefan u64 flags = 0; 18660caa102dSLi Zefan 18674a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18680caa102dSLi Zefan return -EINVAL; 18690caa102dSLi Zefan 18700b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 18710caa102dSLi Zefan if (btrfs_root_readonly(root)) 18720caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 18730b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 18740caa102dSLi Zefan 18750caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 18760caa102dSLi Zefan ret = -EFAULT; 18770caa102dSLi Zefan 18780caa102dSLi Zefan return ret; 18790caa102dSLi Zefan } 18800caa102dSLi Zefan 18810caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 18820caa102dSLi Zefan void __user *arg) 18830caa102dSLi Zefan { 1884496ad9aaSAl Viro struct inode *inode = file_inode(file); 18850b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18860caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18870caa102dSLi Zefan struct btrfs_trans_handle *trans; 18880caa102dSLi Zefan u64 root_flags; 18890caa102dSLi Zefan u64 flags; 18900caa102dSLi Zefan int ret = 0; 18910caa102dSLi Zefan 1892bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1893bd60ea0fSDavid Sterba return -EPERM; 1894bd60ea0fSDavid Sterba 1895b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1896b9ca0664SLiu Bo if (ret) 1897b9ca0664SLiu Bo goto out; 18980caa102dSLi Zefan 18994a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1900b9ca0664SLiu Bo ret = -EINVAL; 1901b9ca0664SLiu Bo goto out_drop_write; 1902b9ca0664SLiu Bo } 19030caa102dSLi Zefan 1904b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1905b9ca0664SLiu Bo ret = -EFAULT; 1906b9ca0664SLiu Bo goto out_drop_write; 1907b9ca0664SLiu Bo } 19080caa102dSLi Zefan 1909b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1910b9ca0664SLiu Bo ret = -EINVAL; 1911b9ca0664SLiu Bo goto out_drop_write; 1912b9ca0664SLiu Bo } 19130caa102dSLi Zefan 1914b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1915b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1916b9ca0664SLiu Bo goto out_drop_write; 1917b9ca0664SLiu Bo } 19180caa102dSLi Zefan 19190b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19200caa102dSLi Zefan 19210caa102dSLi Zefan /* nothing to do */ 19220caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1923b9ca0664SLiu Bo goto out_drop_sem; 19240caa102dSLi Zefan 19250caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19262c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19270caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19280caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19292c686537SDavid Sterba } else { 19302c686537SDavid Sterba /* 19312c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19322c686537SDavid Sterba * send 19332c686537SDavid Sterba */ 19342c686537SDavid Sterba spin_lock(&root->root_item_lock); 19352c686537SDavid Sterba if (root->send_in_progress == 0) { 19360caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19370caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19382c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19392c686537SDavid Sterba } else { 19402c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19410b246afaSJeff Mahoney btrfs_warn(fs_info, 19422c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19432c686537SDavid Sterba root->root_key.objectid); 19442c686537SDavid Sterba ret = -EPERM; 19452c686537SDavid Sterba goto out_drop_sem; 19462c686537SDavid Sterba } 19472c686537SDavid Sterba } 19480caa102dSLi Zefan 19490caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19500caa102dSLi Zefan if (IS_ERR(trans)) { 19510caa102dSLi Zefan ret = PTR_ERR(trans); 19520caa102dSLi Zefan goto out_reset; 19530caa102dSLi Zefan } 19540caa102dSLi Zefan 19550b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19560caa102dSLi Zefan &root->root_key, &root->root_item); 19579417ebc8SNikolay Borisov if (ret < 0) { 19589417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19599417ebc8SNikolay Borisov goto out_reset; 19609417ebc8SNikolay Borisov } 19610caa102dSLi Zefan 19629417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19639417ebc8SNikolay Borisov 19640caa102dSLi Zefan out_reset: 19650caa102dSLi Zefan if (ret) 19660caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1967b9ca0664SLiu Bo out_drop_sem: 19680b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1969b9ca0664SLiu Bo out_drop_write: 1970b9ca0664SLiu Bo mnt_drop_write_file(file); 1971b9ca0664SLiu Bo out: 19720caa102dSLi Zefan return ret; 19730caa102dSLi Zefan } 19740caa102dSLi Zefan 1975ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1976ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1977ac8e9819SChris Mason { 1978abc6e134SChris Mason struct btrfs_key test; 1979abc6e134SChris Mason int ret; 1980abc6e134SChris Mason 1981abc6e134SChris Mason test.objectid = sk->min_objectid; 1982abc6e134SChris Mason test.type = sk->min_type; 1983abc6e134SChris Mason test.offset = sk->min_offset; 1984abc6e134SChris Mason 1985abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1986abc6e134SChris Mason if (ret < 0) 1987ac8e9819SChris Mason return 0; 1988abc6e134SChris Mason 1989abc6e134SChris Mason test.objectid = sk->max_objectid; 1990abc6e134SChris Mason test.type = sk->max_type; 1991abc6e134SChris Mason test.offset = sk->max_offset; 1992abc6e134SChris Mason 1993abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1994abc6e134SChris Mason if (ret > 0) 1995ac8e9819SChris Mason return 0; 1996ac8e9819SChris Mason return 1; 1997ac8e9819SChris Mason } 1998ac8e9819SChris Mason 1999df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2000ac8e9819SChris Mason struct btrfs_key *key, 2001ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20029b6e817dSGerhard Heift size_t *buf_size, 2003ba346b35SGerhard Heift char __user *ubuf, 2004ac8e9819SChris Mason unsigned long *sk_offset, 2005ac8e9819SChris Mason int *num_found) 2006ac8e9819SChris Mason { 2007ac8e9819SChris Mason u64 found_transid; 2008ac8e9819SChris Mason struct extent_buffer *leaf; 2009ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2010dd81d459SNaohiro Aota struct btrfs_key test; 2011ac8e9819SChris Mason unsigned long item_off; 2012ac8e9819SChris Mason unsigned long item_len; 2013ac8e9819SChris Mason int nritems; 2014ac8e9819SChris Mason int i; 2015ac8e9819SChris Mason int slot; 2016ac8e9819SChris Mason int ret = 0; 2017ac8e9819SChris Mason 2018ac8e9819SChris Mason leaf = path->nodes[0]; 2019ac8e9819SChris Mason slot = path->slots[0]; 2020ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2021ac8e9819SChris Mason 2022ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2023ac8e9819SChris Mason i = nritems; 2024ac8e9819SChris Mason goto advance_key; 2025ac8e9819SChris Mason } 2026ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2027ac8e9819SChris Mason 2028ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2029ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2030ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2031ac8e9819SChris Mason 203203b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 203303b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 203403b71c6cSGabriel de Perthuis continue; 203503b71c6cSGabriel de Perthuis 20369b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20378f5f6178SGerhard Heift if (*num_found) { 2038ac8e9819SChris Mason ret = 1; 20398f5f6178SGerhard Heift goto out; 20408f5f6178SGerhard Heift } 20418f5f6178SGerhard Heift 20428f5f6178SGerhard Heift /* 20438f5f6178SGerhard Heift * return one empty item back for v1, which does not 20448f5f6178SGerhard Heift * handle -EOVERFLOW 20458f5f6178SGerhard Heift */ 20468f5f6178SGerhard Heift 20479b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2048ac8e9819SChris Mason item_len = 0; 20498f5f6178SGerhard Heift ret = -EOVERFLOW; 20508f5f6178SGerhard Heift } 2051ac8e9819SChris Mason 20529b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2053ac8e9819SChris Mason ret = 1; 205425c9bc2eSGerhard Heift goto out; 2055ac8e9819SChris Mason } 2056ac8e9819SChris Mason 2057ac8e9819SChris Mason sh.objectid = key->objectid; 2058ac8e9819SChris Mason sh.offset = key->offset; 2059ac8e9819SChris Mason sh.type = key->type; 2060ac8e9819SChris Mason sh.len = item_len; 2061ac8e9819SChris Mason sh.transid = found_transid; 2062ac8e9819SChris Mason 2063ac8e9819SChris Mason /* copy search result header */ 2064ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2065ba346b35SGerhard Heift ret = -EFAULT; 2066ba346b35SGerhard Heift goto out; 2067ba346b35SGerhard Heift } 2068ba346b35SGerhard Heift 2069ac8e9819SChris Mason *sk_offset += sizeof(sh); 2070ac8e9819SChris Mason 2071ac8e9819SChris Mason if (item_len) { 2072ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2073ac8e9819SChris Mason /* copy the item */ 2074ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2075ba346b35SGerhard Heift item_off, item_len)) { 2076ba346b35SGerhard Heift ret = -EFAULT; 2077ba346b35SGerhard Heift goto out; 2078ba346b35SGerhard Heift } 2079ba346b35SGerhard Heift 2080ac8e9819SChris Mason *sk_offset += item_len; 2081ac8e9819SChris Mason } 2082e2156867SHugo Mills (*num_found)++; 2083ac8e9819SChris Mason 20848f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20858f5f6178SGerhard Heift goto out; 20868f5f6178SGerhard Heift 208725c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 208825c9bc2eSGerhard Heift ret = 1; 208925c9bc2eSGerhard Heift goto out; 209025c9bc2eSGerhard Heift } 2091ac8e9819SChris Mason } 2092ac8e9819SChris Mason advance_key: 2093ac8e9819SChris Mason ret = 0; 2094dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2095dd81d459SNaohiro Aota test.type = sk->max_type; 2096dd81d459SNaohiro Aota test.offset = sk->max_offset; 2097dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2098dd81d459SNaohiro Aota ret = 1; 2099dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2100abc6e134SChris Mason key->offset++; 2101dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2102abc6e134SChris Mason key->offset = 0; 2103abc6e134SChris Mason key->type++; 2104dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2105abc6e134SChris Mason key->offset = 0; 2106abc6e134SChris Mason key->type = 0; 2107abc6e134SChris Mason key->objectid++; 2108abc6e134SChris Mason } else 2109abc6e134SChris Mason ret = 1; 211025c9bc2eSGerhard Heift out: 2111ba346b35SGerhard Heift /* 2112ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2113ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2114ba346b35SGerhard Heift * * all items were found 2115ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2116ba346b35SGerhard Heift * leaf 2117ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2118ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2119ba346b35SGerhard Heift */ 2120ac8e9819SChris Mason return ret; 2121ac8e9819SChris Mason } 2122ac8e9819SChris Mason 2123ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 212412544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21259b6e817dSGerhard Heift size_t *buf_size, 2126ba346b35SGerhard Heift char __user *ubuf) 2127ac8e9819SChris Mason { 21280b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2129ac8e9819SChris Mason struct btrfs_root *root; 2130ac8e9819SChris Mason struct btrfs_key key; 2131ac8e9819SChris Mason struct btrfs_path *path; 2132ac8e9819SChris Mason int ret; 2133ac8e9819SChris Mason int num_found = 0; 2134ac8e9819SChris Mason unsigned long sk_offset = 0; 2135ac8e9819SChris Mason 21369b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21379b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 213812544442SGerhard Heift return -EOVERFLOW; 21399b6e817dSGerhard Heift } 214012544442SGerhard Heift 2141ac8e9819SChris Mason path = btrfs_alloc_path(); 2142ac8e9819SChris Mason if (!path) 2143ac8e9819SChris Mason return -ENOMEM; 2144ac8e9819SChris Mason 2145ac8e9819SChris Mason if (sk->tree_id == 0) { 2146ac8e9819SChris Mason /* search the root of the inode that was passed */ 2147ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2148ac8e9819SChris Mason } else { 2149ac8e9819SChris Mason key.objectid = sk->tree_id; 2150ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2151ac8e9819SChris Mason key.offset = (u64)-1; 2152ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2153ac8e9819SChris Mason if (IS_ERR(root)) { 2154ac8e9819SChris Mason btrfs_free_path(path); 2155ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2156ac8e9819SChris Mason } 2157ac8e9819SChris Mason } 2158ac8e9819SChris Mason 2159ac8e9819SChris Mason key.objectid = sk->min_objectid; 2160ac8e9819SChris Mason key.type = sk->min_type; 2161ac8e9819SChris Mason key.offset = sk->min_offset; 2162ac8e9819SChris Mason 2163ac8e9819SChris Mason while (1) { 21646174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2165ac8e9819SChris Mason if (ret != 0) { 2166ac8e9819SChris Mason if (ret > 0) 2167ac8e9819SChris Mason ret = 0; 2168ac8e9819SChris Mason goto err; 2169ac8e9819SChris Mason } 2170df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2171ac8e9819SChris Mason &sk_offset, &num_found); 2172b3b4aa74SDavid Sterba btrfs_release_path(path); 217325c9bc2eSGerhard Heift if (ret) 2174ac8e9819SChris Mason break; 2175ac8e9819SChris Mason 2176ac8e9819SChris Mason } 21778f5f6178SGerhard Heift if (ret > 0) 2178ac8e9819SChris Mason ret = 0; 2179ac8e9819SChris Mason err: 2180ac8e9819SChris Mason sk->nr_items = num_found; 2181ac8e9819SChris Mason btrfs_free_path(path); 2182ac8e9819SChris Mason return ret; 2183ac8e9819SChris Mason } 2184ac8e9819SChris Mason 2185ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2186ac8e9819SChris Mason void __user *argp) 2187ac8e9819SChris Mason { 2188ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2189ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2190ac8e9819SChris Mason struct inode *inode; 2191ac8e9819SChris Mason int ret; 21929b6e817dSGerhard Heift size_t buf_size; 2193ac8e9819SChris Mason 2194ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2195ac8e9819SChris Mason return -EPERM; 2196ac8e9819SChris Mason 2197ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2198ac8e9819SChris Mason 2199ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2200ba346b35SGerhard Heift return -EFAULT; 2201ba346b35SGerhard Heift 2202ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2203ac8e9819SChris Mason 2204496ad9aaSAl Viro inode = file_inode(file); 2205ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22068f5f6178SGerhard Heift 22078f5f6178SGerhard Heift /* 22088f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22098f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22108f5f6178SGerhard Heift */ 22118f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22128f5f6178SGerhard Heift ret = 0; 22138f5f6178SGerhard Heift 2214ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2215ac8e9819SChris Mason ret = -EFAULT; 2216ac8e9819SChris Mason return ret; 2217ac8e9819SChris Mason } 2218ac8e9819SChris Mason 2219cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2220cc68a8a5SGerhard Heift void __user *argp) 2221cc68a8a5SGerhard Heift { 2222cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2223cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2224cc68a8a5SGerhard Heift struct inode *inode; 2225cc68a8a5SGerhard Heift int ret; 2226cc68a8a5SGerhard Heift size_t buf_size; 2227ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2228cc68a8a5SGerhard Heift 2229cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2230cc68a8a5SGerhard Heift return -EPERM; 2231cc68a8a5SGerhard Heift 2232cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2233cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2234cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2235cc68a8a5SGerhard Heift return -EFAULT; 2236cc68a8a5SGerhard Heift 2237cc68a8a5SGerhard Heift buf_size = args.buf_size; 2238cc68a8a5SGerhard Heift 2239cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2240cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2241cc68a8a5SGerhard Heift buf_size = buf_limit; 2242cc68a8a5SGerhard Heift 2243cc68a8a5SGerhard Heift inode = file_inode(file); 2244cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2245718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2246cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2247cc68a8a5SGerhard Heift ret = -EFAULT; 2248cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2249cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2250cc68a8a5SGerhard Heift ret = -EFAULT; 2251cc68a8a5SGerhard Heift 225276dda93cSYan, Zheng return ret; 225376dda93cSYan, Zheng } 225476dda93cSYan, Zheng 225598d377a0STARUISI Hiroaki /* 2256ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2257ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 225898d377a0STARUISI Hiroaki */ 225998d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 226098d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 226198d377a0STARUISI Hiroaki { 226298d377a0STARUISI Hiroaki struct btrfs_root *root; 226398d377a0STARUISI Hiroaki struct btrfs_key key; 2264ac8e9819SChris Mason char *ptr; 226598d377a0STARUISI Hiroaki int ret = -1; 226698d377a0STARUISI Hiroaki int slot; 226798d377a0STARUISI Hiroaki int len; 226898d377a0STARUISI Hiroaki int total_len = 0; 226998d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 227098d377a0STARUISI Hiroaki struct extent_buffer *l; 227198d377a0STARUISI Hiroaki struct btrfs_path *path; 227298d377a0STARUISI Hiroaki 227398d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 227498d377a0STARUISI Hiroaki name[0]='\0'; 227598d377a0STARUISI Hiroaki return 0; 227698d377a0STARUISI Hiroaki } 227798d377a0STARUISI Hiroaki 227898d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 227998d377a0STARUISI Hiroaki if (!path) 228098d377a0STARUISI Hiroaki return -ENOMEM; 228198d377a0STARUISI Hiroaki 2282c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 228398d377a0STARUISI Hiroaki 228498d377a0STARUISI Hiroaki key.objectid = tree_id; 228598d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 228698d377a0STARUISI Hiroaki key.offset = (u64)-1; 228798d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 228898d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2289ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 22908ad6fcabSChris Mason goto out; 229198d377a0STARUISI Hiroaki } 229298d377a0STARUISI Hiroaki 229398d377a0STARUISI Hiroaki key.objectid = dirid; 229498d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22958ad6fcabSChris Mason key.offset = (u64)-1; 229698d377a0STARUISI Hiroaki 229798d377a0STARUISI Hiroaki while (1) { 229898d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 229998d377a0STARUISI Hiroaki if (ret < 0) 230098d377a0STARUISI Hiroaki goto out; 230118674c6cSFilipe David Borba Manana else if (ret > 0) { 230218674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 230318674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 230418674c6cSFilipe David Borba Manana if (ret < 0) 230518674c6cSFilipe David Borba Manana goto out; 230618674c6cSFilipe David Borba Manana else if (ret > 0) { 2307ac8e9819SChris Mason ret = -ENOENT; 230898d377a0STARUISI Hiroaki goto out; 2309ac8e9819SChris Mason } 231018674c6cSFilipe David Borba Manana } 231118674c6cSFilipe David Borba Manana 231218674c6cSFilipe David Borba Manana l = path->nodes[0]; 231318674c6cSFilipe David Borba Manana slot = path->slots[0]; 231418674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 231598d377a0STARUISI Hiroaki 231698d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 231798d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 231898d377a0STARUISI Hiroaki ptr -= len + 1; 231998d377a0STARUISI Hiroaki total_len += len + 1; 2320a696cf35SFilipe David Borba Manana if (ptr < name) { 2321a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 232298d377a0STARUISI Hiroaki goto out; 2323a696cf35SFilipe David Borba Manana } 232498d377a0STARUISI Hiroaki 232598d377a0STARUISI Hiroaki *(ptr + len) = '/'; 232698d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 232798d377a0STARUISI Hiroaki 232898d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 232998d377a0STARUISI Hiroaki break; 233098d377a0STARUISI Hiroaki 2331b3b4aa74SDavid Sterba btrfs_release_path(path); 233298d377a0STARUISI Hiroaki key.objectid = key.offset; 23338ad6fcabSChris Mason key.offset = (u64)-1; 233498d377a0STARUISI Hiroaki dirid = key.objectid; 233598d377a0STARUISI Hiroaki } 233677906a50SLi Zefan memmove(name, ptr, total_len); 233798d377a0STARUISI Hiroaki name[total_len] = '\0'; 233898d377a0STARUISI Hiroaki ret = 0; 233998d377a0STARUISI Hiroaki out: 234098d377a0STARUISI Hiroaki btrfs_free_path(path); 2341ac8e9819SChris Mason return ret; 2342ac8e9819SChris Mason } 2343ac8e9819SChris Mason 234423d0b79dSTomohiro Misono static int btrfs_search_path_in_tree_user(struct inode *inode, 234523d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args) 234623d0b79dSTomohiro Misono { 234723d0b79dSTomohiro Misono struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 234823d0b79dSTomohiro Misono struct super_block *sb = inode->i_sb; 234923d0b79dSTomohiro Misono struct btrfs_key upper_limit = BTRFS_I(inode)->location; 235023d0b79dSTomohiro Misono u64 treeid = BTRFS_I(inode)->root->root_key.objectid; 235123d0b79dSTomohiro Misono u64 dirid = args->dirid; 235223d0b79dSTomohiro Misono unsigned long item_off; 235323d0b79dSTomohiro Misono unsigned long item_len; 235423d0b79dSTomohiro Misono struct btrfs_inode_ref *iref; 235523d0b79dSTomohiro Misono struct btrfs_root_ref *rref; 235623d0b79dSTomohiro Misono struct btrfs_root *root; 235723d0b79dSTomohiro Misono struct btrfs_path *path; 235823d0b79dSTomohiro Misono struct btrfs_key key, key2; 235923d0b79dSTomohiro Misono struct extent_buffer *leaf; 236023d0b79dSTomohiro Misono struct inode *temp_inode; 236123d0b79dSTomohiro Misono char *ptr; 236223d0b79dSTomohiro Misono int slot; 236323d0b79dSTomohiro Misono int len; 236423d0b79dSTomohiro Misono int total_len = 0; 236523d0b79dSTomohiro Misono int ret; 236623d0b79dSTomohiro Misono 236723d0b79dSTomohiro Misono path = btrfs_alloc_path(); 236823d0b79dSTomohiro Misono if (!path) 236923d0b79dSTomohiro Misono return -ENOMEM; 237023d0b79dSTomohiro Misono 237123d0b79dSTomohiro Misono /* 237223d0b79dSTomohiro Misono * If the bottom subvolume does not exist directly under upper_limit, 237323d0b79dSTomohiro Misono * construct the path in from the bottom up. 237423d0b79dSTomohiro Misono */ 237523d0b79dSTomohiro Misono if (dirid != upper_limit.objectid) { 237623d0b79dSTomohiro Misono ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 237723d0b79dSTomohiro Misono 237823d0b79dSTomohiro Misono key.objectid = treeid; 237923d0b79dSTomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 238023d0b79dSTomohiro Misono key.offset = (u64)-1; 238123d0b79dSTomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 238223d0b79dSTomohiro Misono if (IS_ERR(root)) { 238323d0b79dSTomohiro Misono ret = PTR_ERR(root); 238423d0b79dSTomohiro Misono goto out; 238523d0b79dSTomohiro Misono } 238623d0b79dSTomohiro Misono 238723d0b79dSTomohiro Misono key.objectid = dirid; 238823d0b79dSTomohiro Misono key.type = BTRFS_INODE_REF_KEY; 238923d0b79dSTomohiro Misono key.offset = (u64)-1; 239023d0b79dSTomohiro Misono while (1) { 239123d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 239223d0b79dSTomohiro Misono if (ret < 0) { 239323d0b79dSTomohiro Misono goto out; 239423d0b79dSTomohiro Misono } else if (ret > 0) { 239523d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 239623d0b79dSTomohiro Misono BTRFS_INODE_REF_KEY); 239723d0b79dSTomohiro Misono if (ret < 0) { 239823d0b79dSTomohiro Misono goto out; 239923d0b79dSTomohiro Misono } else if (ret > 0) { 240023d0b79dSTomohiro Misono ret = -ENOENT; 240123d0b79dSTomohiro Misono goto out; 240223d0b79dSTomohiro Misono } 240323d0b79dSTomohiro Misono } 240423d0b79dSTomohiro Misono 240523d0b79dSTomohiro Misono leaf = path->nodes[0]; 240623d0b79dSTomohiro Misono slot = path->slots[0]; 240723d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 240823d0b79dSTomohiro Misono 240923d0b79dSTomohiro Misono iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 241023d0b79dSTomohiro Misono len = btrfs_inode_ref_name_len(leaf, iref); 241123d0b79dSTomohiro Misono ptr -= len + 1; 241223d0b79dSTomohiro Misono total_len += len + 1; 241323d0b79dSTomohiro Misono if (ptr < args->path) { 241423d0b79dSTomohiro Misono ret = -ENAMETOOLONG; 241523d0b79dSTomohiro Misono goto out; 241623d0b79dSTomohiro Misono } 241723d0b79dSTomohiro Misono 241823d0b79dSTomohiro Misono *(ptr + len) = '/'; 241923d0b79dSTomohiro Misono read_extent_buffer(leaf, ptr, 242023d0b79dSTomohiro Misono (unsigned long)(iref + 1), len); 242123d0b79dSTomohiro Misono 242223d0b79dSTomohiro Misono /* Check the read+exec permission of this directory */ 242323d0b79dSTomohiro Misono ret = btrfs_previous_item(root, path, dirid, 242423d0b79dSTomohiro Misono BTRFS_INODE_ITEM_KEY); 242523d0b79dSTomohiro Misono if (ret < 0) { 242623d0b79dSTomohiro Misono goto out; 242723d0b79dSTomohiro Misono } else if (ret > 0) { 242823d0b79dSTomohiro Misono ret = -ENOENT; 242923d0b79dSTomohiro Misono goto out; 243023d0b79dSTomohiro Misono } 243123d0b79dSTomohiro Misono 243223d0b79dSTomohiro Misono leaf = path->nodes[0]; 243323d0b79dSTomohiro Misono slot = path->slots[0]; 243423d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key2, slot); 243523d0b79dSTomohiro Misono if (key2.objectid != dirid) { 243623d0b79dSTomohiro Misono ret = -ENOENT; 243723d0b79dSTomohiro Misono goto out; 243823d0b79dSTomohiro Misono } 243923d0b79dSTomohiro Misono 244023d0b79dSTomohiro Misono temp_inode = btrfs_iget(sb, &key2, root, NULL); 24413ca57bd6SMisono Tomohiro if (IS_ERR(temp_inode)) { 24423ca57bd6SMisono Tomohiro ret = PTR_ERR(temp_inode); 24433ca57bd6SMisono Tomohiro goto out; 24443ca57bd6SMisono Tomohiro } 244523d0b79dSTomohiro Misono ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC); 244623d0b79dSTomohiro Misono iput(temp_inode); 244723d0b79dSTomohiro Misono if (ret) { 244823d0b79dSTomohiro Misono ret = -EACCES; 244923d0b79dSTomohiro Misono goto out; 245023d0b79dSTomohiro Misono } 245123d0b79dSTomohiro Misono 245223d0b79dSTomohiro Misono if (key.offset == upper_limit.objectid) 245323d0b79dSTomohiro Misono break; 245423d0b79dSTomohiro Misono if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 245523d0b79dSTomohiro Misono ret = -EACCES; 245623d0b79dSTomohiro Misono goto out; 245723d0b79dSTomohiro Misono } 245823d0b79dSTomohiro Misono 245923d0b79dSTomohiro Misono btrfs_release_path(path); 246023d0b79dSTomohiro Misono key.objectid = key.offset; 246123d0b79dSTomohiro Misono key.offset = (u64)-1; 246223d0b79dSTomohiro Misono dirid = key.objectid; 246323d0b79dSTomohiro Misono } 246423d0b79dSTomohiro Misono 246523d0b79dSTomohiro Misono memmove(args->path, ptr, total_len); 246623d0b79dSTomohiro Misono args->path[total_len] = '\0'; 246723d0b79dSTomohiro Misono btrfs_release_path(path); 246823d0b79dSTomohiro Misono } 246923d0b79dSTomohiro Misono 247023d0b79dSTomohiro Misono /* Get the bottom subvolume's name from ROOT_REF */ 247123d0b79dSTomohiro Misono root = fs_info->tree_root; 247223d0b79dSTomohiro Misono key.objectid = treeid; 247323d0b79dSTomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 247423d0b79dSTomohiro Misono key.offset = args->treeid; 247523d0b79dSTomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 247623d0b79dSTomohiro Misono if (ret < 0) { 247723d0b79dSTomohiro Misono goto out; 247823d0b79dSTomohiro Misono } else if (ret > 0) { 247923d0b79dSTomohiro Misono ret = -ENOENT; 248023d0b79dSTomohiro Misono goto out; 248123d0b79dSTomohiro Misono } 248223d0b79dSTomohiro Misono 248323d0b79dSTomohiro Misono leaf = path->nodes[0]; 248423d0b79dSTomohiro Misono slot = path->slots[0]; 248523d0b79dSTomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 248623d0b79dSTomohiro Misono 248723d0b79dSTomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot); 248823d0b79dSTomohiro Misono item_len = btrfs_item_size_nr(leaf, slot); 248923d0b79dSTomohiro Misono /* Check if dirid in ROOT_REF corresponds to passed dirid */ 249023d0b79dSTomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 249123d0b79dSTomohiro Misono if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { 249223d0b79dSTomohiro Misono ret = -EINVAL; 249323d0b79dSTomohiro Misono goto out; 249423d0b79dSTomohiro Misono } 249523d0b79dSTomohiro Misono 249623d0b79dSTomohiro Misono /* Copy subvolume's name */ 249723d0b79dSTomohiro Misono item_off += sizeof(struct btrfs_root_ref); 249823d0b79dSTomohiro Misono item_len -= sizeof(struct btrfs_root_ref); 249923d0b79dSTomohiro Misono read_extent_buffer(leaf, args->name, item_off, item_len); 250023d0b79dSTomohiro Misono args->name[item_len] = 0; 250123d0b79dSTomohiro Misono 250223d0b79dSTomohiro Misono out: 250323d0b79dSTomohiro Misono btrfs_free_path(path); 250423d0b79dSTomohiro Misono return ret; 250523d0b79dSTomohiro Misono } 250623d0b79dSTomohiro Misono 2507ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2508ac8e9819SChris Mason void __user *argp) 2509ac8e9819SChris Mason { 2510ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2511ac8e9819SChris Mason struct inode *inode; 251201b810b8SDavid Sterba int ret = 0; 2513ac8e9819SChris Mason 25142354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 25152354d08fSJulia Lawall if (IS_ERR(args)) 25162354d08fSJulia Lawall return PTR_ERR(args); 2517c2b96929SDan Carpenter 2518496ad9aaSAl Viro inode = file_inode(file); 2519ac8e9819SChris Mason 252001b810b8SDavid Sterba /* 252101b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 252201b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 252301b810b8SDavid Sterba */ 25241b53ac4dSChris Mason if (args->treeid == 0) 25251b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 25261b53ac4dSChris Mason 252701b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 252801b810b8SDavid Sterba args->name[0] = 0; 252901b810b8SDavid Sterba goto out; 253001b810b8SDavid Sterba } 253101b810b8SDavid Sterba 253201b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 253301b810b8SDavid Sterba ret = -EPERM; 253401b810b8SDavid Sterba goto out; 253501b810b8SDavid Sterba } 253601b810b8SDavid Sterba 2537ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2538ac8e9819SChris Mason args->treeid, args->objectid, 2539ac8e9819SChris Mason args->name); 2540ac8e9819SChris Mason 254101b810b8SDavid Sterba out: 2542ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2543ac8e9819SChris Mason ret = -EFAULT; 2544ac8e9819SChris Mason 2545ac8e9819SChris Mason kfree(args); 254698d377a0STARUISI Hiroaki return ret; 254798d377a0STARUISI Hiroaki } 254898d377a0STARUISI Hiroaki 254923d0b79dSTomohiro Misono /* 255023d0b79dSTomohiro Misono * Version of ino_lookup ioctl (unprivileged) 255123d0b79dSTomohiro Misono * 255223d0b79dSTomohiro Misono * The main differences from ino_lookup ioctl are: 255323d0b79dSTomohiro Misono * 255423d0b79dSTomohiro Misono * 1. Read + Exec permission will be checked using inode_permission() during 255523d0b79dSTomohiro Misono * path construction. -EACCES will be returned in case of failure. 255623d0b79dSTomohiro Misono * 2. Path construction will be stopped at the inode number which corresponds 255723d0b79dSTomohiro Misono * to the fd with which this ioctl is called. If constructed path does not 255823d0b79dSTomohiro Misono * exist under fd's inode, -EACCES will be returned. 255923d0b79dSTomohiro Misono * 3. The name of bottom subvolume is also searched and filled. 256023d0b79dSTomohiro Misono */ 256123d0b79dSTomohiro Misono static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 256223d0b79dSTomohiro Misono { 256323d0b79dSTomohiro Misono struct btrfs_ioctl_ino_lookup_user_args *args; 256423d0b79dSTomohiro Misono struct inode *inode; 256523d0b79dSTomohiro Misono int ret; 256623d0b79dSTomohiro Misono 256723d0b79dSTomohiro Misono args = memdup_user(argp, sizeof(*args)); 256823d0b79dSTomohiro Misono if (IS_ERR(args)) 256923d0b79dSTomohiro Misono return PTR_ERR(args); 257023d0b79dSTomohiro Misono 257123d0b79dSTomohiro Misono inode = file_inode(file); 257223d0b79dSTomohiro Misono 257323d0b79dSTomohiro Misono if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 257423d0b79dSTomohiro Misono BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { 257523d0b79dSTomohiro Misono /* 257623d0b79dSTomohiro Misono * The subvolume does not exist under fd with which this is 257723d0b79dSTomohiro Misono * called 257823d0b79dSTomohiro Misono */ 257923d0b79dSTomohiro Misono kfree(args); 258023d0b79dSTomohiro Misono return -EACCES; 258123d0b79dSTomohiro Misono } 258223d0b79dSTomohiro Misono 258323d0b79dSTomohiro Misono ret = btrfs_search_path_in_tree_user(inode, args); 258423d0b79dSTomohiro Misono 258523d0b79dSTomohiro Misono if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 258623d0b79dSTomohiro Misono ret = -EFAULT; 258723d0b79dSTomohiro Misono 258823d0b79dSTomohiro Misono kfree(args); 258923d0b79dSTomohiro Misono return ret; 259023d0b79dSTomohiro Misono } 259123d0b79dSTomohiro Misono 2592b64ec075STomohiro Misono /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 2593b64ec075STomohiro Misono static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) 2594b64ec075STomohiro Misono { 2595b64ec075STomohiro Misono struct btrfs_ioctl_get_subvol_info_args *subvol_info; 2596b64ec075STomohiro Misono struct btrfs_fs_info *fs_info; 2597b64ec075STomohiro Misono struct btrfs_root *root; 2598b64ec075STomohiro Misono struct btrfs_path *path; 2599b64ec075STomohiro Misono struct btrfs_key key; 2600b64ec075STomohiro Misono struct btrfs_root_item *root_item; 2601b64ec075STomohiro Misono struct btrfs_root_ref *rref; 2602b64ec075STomohiro Misono struct extent_buffer *leaf; 2603b64ec075STomohiro Misono unsigned long item_off; 2604b64ec075STomohiro Misono unsigned long item_len; 2605b64ec075STomohiro Misono struct inode *inode; 2606b64ec075STomohiro Misono int slot; 2607b64ec075STomohiro Misono int ret = 0; 2608b64ec075STomohiro Misono 2609b64ec075STomohiro Misono path = btrfs_alloc_path(); 2610b64ec075STomohiro Misono if (!path) 2611b64ec075STomohiro Misono return -ENOMEM; 2612b64ec075STomohiro Misono 2613b64ec075STomohiro Misono subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); 2614b64ec075STomohiro Misono if (!subvol_info) { 2615b64ec075STomohiro Misono btrfs_free_path(path); 2616b64ec075STomohiro Misono return -ENOMEM; 2617b64ec075STomohiro Misono } 2618b64ec075STomohiro Misono 2619b64ec075STomohiro Misono inode = file_inode(file); 2620b64ec075STomohiro Misono fs_info = BTRFS_I(inode)->root->fs_info; 2621b64ec075STomohiro Misono 2622b64ec075STomohiro Misono /* Get root_item of inode's subvolume */ 2623b64ec075STomohiro Misono key.objectid = BTRFS_I(inode)->root->root_key.objectid; 2624b64ec075STomohiro Misono key.type = BTRFS_ROOT_ITEM_KEY; 2625b64ec075STomohiro Misono key.offset = (u64)-1; 2626b64ec075STomohiro Misono root = btrfs_read_fs_root_no_name(fs_info, &key); 2627b64ec075STomohiro Misono if (IS_ERR(root)) { 2628b64ec075STomohiro Misono ret = PTR_ERR(root); 2629b64ec075STomohiro Misono goto out; 2630b64ec075STomohiro Misono } 2631b64ec075STomohiro Misono root_item = &root->root_item; 2632b64ec075STomohiro Misono 2633b64ec075STomohiro Misono subvol_info->treeid = key.objectid; 2634b64ec075STomohiro Misono 2635b64ec075STomohiro Misono subvol_info->generation = btrfs_root_generation(root_item); 2636b64ec075STomohiro Misono subvol_info->flags = btrfs_root_flags(root_item); 2637b64ec075STomohiro Misono 2638b64ec075STomohiro Misono memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 2639b64ec075STomohiro Misono memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 2640b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2641b64ec075STomohiro Misono memcpy(subvol_info->received_uuid, root_item->received_uuid, 2642b64ec075STomohiro Misono BTRFS_UUID_SIZE); 2643b64ec075STomohiro Misono 2644b64ec075STomohiro Misono subvol_info->ctransid = btrfs_root_ctransid(root_item); 2645b64ec075STomohiro Misono subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 2646b64ec075STomohiro Misono subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2647b64ec075STomohiro Misono 2648b64ec075STomohiro Misono subvol_info->otransid = btrfs_root_otransid(root_item); 2649b64ec075STomohiro Misono subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2650b64ec075STomohiro Misono subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2651b64ec075STomohiro Misono 2652b64ec075STomohiro Misono subvol_info->stransid = btrfs_root_stransid(root_item); 2653b64ec075STomohiro Misono subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2654b64ec075STomohiro Misono subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2655b64ec075STomohiro Misono 2656b64ec075STomohiro Misono subvol_info->rtransid = btrfs_root_rtransid(root_item); 2657b64ec075STomohiro Misono subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2658b64ec075STomohiro Misono subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2659b64ec075STomohiro Misono 2660b64ec075STomohiro Misono if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2661b64ec075STomohiro Misono /* Search root tree for ROOT_BACKREF of this subvolume */ 2662b64ec075STomohiro Misono root = fs_info->tree_root; 2663b64ec075STomohiro Misono 2664b64ec075STomohiro Misono key.type = BTRFS_ROOT_BACKREF_KEY; 2665b64ec075STomohiro Misono key.offset = 0; 2666b64ec075STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2667b64ec075STomohiro Misono if (ret < 0) { 2668b64ec075STomohiro Misono goto out; 2669b64ec075STomohiro Misono } else if (path->slots[0] >= 2670b64ec075STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 2671b64ec075STomohiro Misono ret = btrfs_next_leaf(root, path); 2672b64ec075STomohiro Misono if (ret < 0) { 2673b64ec075STomohiro Misono goto out; 2674b64ec075STomohiro Misono } else if (ret > 0) { 2675b64ec075STomohiro Misono ret = -EUCLEAN; 2676b64ec075STomohiro Misono goto out; 2677b64ec075STomohiro Misono } 2678b64ec075STomohiro Misono } 2679b64ec075STomohiro Misono 2680b64ec075STomohiro Misono leaf = path->nodes[0]; 2681b64ec075STomohiro Misono slot = path->slots[0]; 2682b64ec075STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 2683b64ec075STomohiro Misono if (key.objectid == subvol_info->treeid && 2684b64ec075STomohiro Misono key.type == BTRFS_ROOT_BACKREF_KEY) { 2685b64ec075STomohiro Misono subvol_info->parent_id = key.offset; 2686b64ec075STomohiro Misono 2687b64ec075STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2688b64ec075STomohiro Misono subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2689b64ec075STomohiro Misono 2690b64ec075STomohiro Misono item_off = btrfs_item_ptr_offset(leaf, slot) 2691b64ec075STomohiro Misono + sizeof(struct btrfs_root_ref); 2692b64ec075STomohiro Misono item_len = btrfs_item_size_nr(leaf, slot) 2693b64ec075STomohiro Misono - sizeof(struct btrfs_root_ref); 2694b64ec075STomohiro Misono read_extent_buffer(leaf, subvol_info->name, 2695b64ec075STomohiro Misono item_off, item_len); 2696b64ec075STomohiro Misono } else { 2697b64ec075STomohiro Misono ret = -ENOENT; 2698b64ec075STomohiro Misono goto out; 2699b64ec075STomohiro Misono } 2700b64ec075STomohiro Misono } 2701b64ec075STomohiro Misono 2702b64ec075STomohiro Misono if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2703b64ec075STomohiro Misono ret = -EFAULT; 2704b64ec075STomohiro Misono 2705b64ec075STomohiro Misono out: 2706b64ec075STomohiro Misono btrfs_free_path(path); 2707b64ec075STomohiro Misono kzfree(subvol_info); 2708b64ec075STomohiro Misono return ret; 2709b64ec075STomohiro Misono } 2710b64ec075STomohiro Misono 271142e4b520STomohiro Misono /* 271242e4b520STomohiro Misono * Return ROOT_REF information of the subvolume containing this inode 271342e4b520STomohiro Misono * except the subvolume name. 271442e4b520STomohiro Misono */ 271542e4b520STomohiro Misono static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) 271642e4b520STomohiro Misono { 271742e4b520STomohiro Misono struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; 271842e4b520STomohiro Misono struct btrfs_root_ref *rref; 271942e4b520STomohiro Misono struct btrfs_root *root; 272042e4b520STomohiro Misono struct btrfs_path *path; 272142e4b520STomohiro Misono struct btrfs_key key; 272242e4b520STomohiro Misono struct extent_buffer *leaf; 272342e4b520STomohiro Misono struct inode *inode; 272442e4b520STomohiro Misono u64 objectid; 272542e4b520STomohiro Misono int slot; 272642e4b520STomohiro Misono int ret; 272742e4b520STomohiro Misono u8 found; 272842e4b520STomohiro Misono 272942e4b520STomohiro Misono path = btrfs_alloc_path(); 273042e4b520STomohiro Misono if (!path) 273142e4b520STomohiro Misono return -ENOMEM; 273242e4b520STomohiro Misono 273342e4b520STomohiro Misono rootrefs = memdup_user(argp, sizeof(*rootrefs)); 273442e4b520STomohiro Misono if (IS_ERR(rootrefs)) { 273542e4b520STomohiro Misono btrfs_free_path(path); 273642e4b520STomohiro Misono return PTR_ERR(rootrefs); 273742e4b520STomohiro Misono } 273842e4b520STomohiro Misono 273942e4b520STomohiro Misono inode = file_inode(file); 274042e4b520STomohiro Misono root = BTRFS_I(inode)->root->fs_info->tree_root; 274142e4b520STomohiro Misono objectid = BTRFS_I(inode)->root->root_key.objectid; 274242e4b520STomohiro Misono 274342e4b520STomohiro Misono key.objectid = objectid; 274442e4b520STomohiro Misono key.type = BTRFS_ROOT_REF_KEY; 274542e4b520STomohiro Misono key.offset = rootrefs->min_treeid; 274642e4b520STomohiro Misono found = 0; 274742e4b520STomohiro Misono 274842e4b520STomohiro Misono ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 274942e4b520STomohiro Misono if (ret < 0) { 275042e4b520STomohiro Misono goto out; 275142e4b520STomohiro Misono } else if (path->slots[0] >= 275242e4b520STomohiro Misono btrfs_header_nritems(path->nodes[0])) { 275342e4b520STomohiro Misono ret = btrfs_next_leaf(root, path); 275442e4b520STomohiro Misono if (ret < 0) { 275542e4b520STomohiro Misono goto out; 275642e4b520STomohiro Misono } else if (ret > 0) { 275742e4b520STomohiro Misono ret = -EUCLEAN; 275842e4b520STomohiro Misono goto out; 275942e4b520STomohiro Misono } 276042e4b520STomohiro Misono } 276142e4b520STomohiro Misono while (1) { 276242e4b520STomohiro Misono leaf = path->nodes[0]; 276342e4b520STomohiro Misono slot = path->slots[0]; 276442e4b520STomohiro Misono 276542e4b520STomohiro Misono btrfs_item_key_to_cpu(leaf, &key, slot); 276642e4b520STomohiro Misono if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 276742e4b520STomohiro Misono ret = 0; 276842e4b520STomohiro Misono goto out; 276942e4b520STomohiro Misono } 277042e4b520STomohiro Misono 277142e4b520STomohiro Misono if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 277242e4b520STomohiro Misono ret = -EOVERFLOW; 277342e4b520STomohiro Misono goto out; 277442e4b520STomohiro Misono } 277542e4b520STomohiro Misono 277642e4b520STomohiro Misono rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 277742e4b520STomohiro Misono rootrefs->rootref[found].treeid = key.offset; 277842e4b520STomohiro Misono rootrefs->rootref[found].dirid = 277942e4b520STomohiro Misono btrfs_root_ref_dirid(leaf, rref); 278042e4b520STomohiro Misono found++; 278142e4b520STomohiro Misono 278242e4b520STomohiro Misono ret = btrfs_next_item(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 279142e4b520STomohiro Misono out: 279242e4b520STomohiro Misono if (!ret || ret == -EOVERFLOW) { 279342e4b520STomohiro Misono rootrefs->num_items = found; 279442e4b520STomohiro Misono /* update min_treeid for next search */ 279542e4b520STomohiro Misono if (found) 279642e4b520STomohiro Misono rootrefs->min_treeid = 279742e4b520STomohiro Misono rootrefs->rootref[found - 1].treeid + 1; 279842e4b520STomohiro Misono if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 279942e4b520STomohiro Misono ret = -EFAULT; 280042e4b520STomohiro Misono } 280142e4b520STomohiro Misono 280242e4b520STomohiro Misono kfree(rootrefs); 280342e4b520STomohiro Misono btrfs_free_path(path); 280442e4b520STomohiro Misono 280542e4b520STomohiro Misono return ret; 280642e4b520STomohiro Misono } 280742e4b520STomohiro Misono 280876dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 280976dda93cSYan, Zheng void __user *arg) 281076dda93cSYan, Zheng { 281154563d41SAl Viro struct dentry *parent = file->f_path.dentry; 28120b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 281376dda93cSYan, Zheng struct dentry *dentry; 28142b0143b5SDavid Howells struct inode *dir = d_inode(parent); 281576dda93cSYan, Zheng struct inode *inode; 281676dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 281776dda93cSYan, Zheng struct btrfs_root *dest = NULL; 281876dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 281976dda93cSYan, Zheng int namelen; 282076dda93cSYan, Zheng int err = 0; 282176dda93cSYan, Zheng 2822325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2823325c50e3SJeff Mahoney return -ENOTDIR; 2824325c50e3SJeff Mahoney 282576dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 282676dda93cSYan, Zheng if (IS_ERR(vol_args)) 282776dda93cSYan, Zheng return PTR_ERR(vol_args); 282876dda93cSYan, Zheng 282976dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 283076dda93cSYan, Zheng namelen = strlen(vol_args->name); 283176dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 283276dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 283376dda93cSYan, Zheng err = -EINVAL; 283476dda93cSYan, Zheng goto out; 283576dda93cSYan, Zheng } 283676dda93cSYan, Zheng 2837a561be71SAl Viro err = mnt_want_write_file(file); 283876dda93cSYan, Zheng if (err) 283976dda93cSYan, Zheng goto out; 284076dda93cSYan, Zheng 2841521e0546SDavid Sterba 284200235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 284300235411SAl Viro if (err == -EINTR) 284400235411SAl Viro goto out_drop_write; 284576dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 284676dda93cSYan, Zheng if (IS_ERR(dentry)) { 284776dda93cSYan, Zheng err = PTR_ERR(dentry); 284876dda93cSYan, Zheng goto out_unlock_dir; 284976dda93cSYan, Zheng } 285076dda93cSYan, Zheng 28512b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 285276dda93cSYan, Zheng err = -ENOENT; 285376dda93cSYan, Zheng goto out_dput; 285476dda93cSYan, Zheng } 285576dda93cSYan, Zheng 28562b0143b5SDavid Howells inode = d_inode(dentry); 28574260f7c7SSage Weil dest = BTRFS_I(inode)->root; 28584260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 28594260f7c7SSage Weil /* 28604260f7c7SSage Weil * Regular user. Only allow this with a special mount 28614260f7c7SSage Weil * option, when the user has write+exec access to the 28624260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 28634260f7c7SSage Weil * allowed. 28644260f7c7SSage Weil * 28654260f7c7SSage Weil * Note that this is _not_ check that the subvol is 28664260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 28674260f7c7SSage Weil * otherwise be able to delete. 28684260f7c7SSage Weil * 28694260f7c7SSage Weil * Users who want to delete empty subvols should try 28704260f7c7SSage Weil * rmdir(2). 28714260f7c7SSage Weil */ 28724260f7c7SSage Weil err = -EPERM; 28730b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 28744260f7c7SSage Weil goto out_dput; 28754260f7c7SSage Weil 28764260f7c7SSage Weil /* 28774260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 28784260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 28794260f7c7SSage Weil * must be called on the dentry referencing the root 28804260f7c7SSage Weil * of the subvol, not a random directory contained 28814260f7c7SSage Weil * within it. 28824260f7c7SSage Weil */ 28834260f7c7SSage Weil err = -EINVAL; 28844260f7c7SSage Weil if (root == dest) 28854260f7c7SSage Weil goto out_dput; 28864260f7c7SSage Weil 28874260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 28884260f7c7SSage Weil if (err) 28894260f7c7SSage Weil goto out_dput; 28905c39da5bSMiao Xie } 28914260f7c7SSage Weil 28925c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 28934260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 28944260f7c7SSage Weil if (err) 28954260f7c7SSage Weil goto out_dput; 28964260f7c7SSage Weil 28974a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 289876dda93cSYan, Zheng err = -EINVAL; 289976dda93cSYan, Zheng goto out_dput; 290076dda93cSYan, Zheng } 290176dda93cSYan, Zheng 29025955102cSAl Viro inode_lock(inode); 2903f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 29045955102cSAl Viro inode_unlock(inode); 2905f60a2364SMisono Tomohiro if (!err) 290676dda93cSYan, Zheng d_delete(dentry); 2907fa6ac876SLiu Bo 290876dda93cSYan, Zheng out_dput: 290976dda93cSYan, Zheng dput(dentry); 291076dda93cSYan, Zheng out_unlock_dir: 29115955102cSAl Viro inode_unlock(dir); 291200235411SAl Viro out_drop_write: 29132a79f17eSAl Viro mnt_drop_write_file(file); 291476dda93cSYan, Zheng out: 291576dda93cSYan, Zheng kfree(vol_args); 291676dda93cSYan, Zheng return err; 291776dda93cSYan, Zheng } 291876dda93cSYan, Zheng 29191e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2920f46b5a66SChristoph Hellwig { 2921496ad9aaSAl Viro struct inode *inode = file_inode(file); 2922f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 29231e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2924c146afadSYan Zheng int ret; 2925c146afadSYan Zheng 292625122d15SIlya Dryomov ret = mnt_want_write_file(file); 292725122d15SIlya Dryomov if (ret) 292825122d15SIlya Dryomov return ret; 2929b83cc969SLi Zefan 293025122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 293125122d15SIlya Dryomov ret = -EROFS; 293225122d15SIlya Dryomov goto out; 29335ac00addSStefan Behrens } 2934f46b5a66SChristoph Hellwig 2935f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2936f46b5a66SChristoph Hellwig case S_IFDIR: 2937e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2938e441d54dSChris Mason ret = -EPERM; 2939e441d54dSChris Mason goto out; 2940e441d54dSChris Mason } 2941de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2942f46b5a66SChristoph Hellwig break; 2943f46b5a66SChristoph Hellwig case S_IFREG: 2944e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2945e441d54dSChris Mason ret = -EINVAL; 2946e441d54dSChris Mason goto out; 2947e441d54dSChris Mason } 29481e701a32SChris Mason 29491e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 29501e701a32SChris Mason if (!range) { 29511e701a32SChris Mason ret = -ENOMEM; 29521e701a32SChris Mason goto out; 29531e701a32SChris Mason } 29541e701a32SChris Mason 29551e701a32SChris Mason if (argp) { 29561e701a32SChris Mason if (copy_from_user(range, argp, 29571e701a32SChris Mason sizeof(*range))) { 29581e701a32SChris Mason ret = -EFAULT; 29591e701a32SChris Mason kfree(range); 2960683be16eSDan Carpenter goto out; 29611e701a32SChris Mason } 29621e701a32SChris Mason /* compression requires us to start the IO */ 29631e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 29641e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 29651e701a32SChris Mason range->extent_thresh = (u32)-1; 29661e701a32SChris Mason } 29671e701a32SChris Mason } else { 29681e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 29691e701a32SChris Mason range->len = (u64)-1; 29701e701a32SChris Mason } 2971496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 29727c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 29734cb5300bSChris Mason if (ret > 0) 29744cb5300bSChris Mason ret = 0; 29751e701a32SChris Mason kfree(range); 2976f46b5a66SChristoph Hellwig break; 29778929ecfaSYan, Zheng default: 29788929ecfaSYan, Zheng ret = -EINVAL; 2979f46b5a66SChristoph Hellwig } 2980e441d54dSChris Mason out: 298125122d15SIlya Dryomov mnt_drop_write_file(file); 2982e441d54dSChris Mason return ret; 2983f46b5a66SChristoph Hellwig } 2984f46b5a66SChristoph Hellwig 29852ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2986f46b5a66SChristoph Hellwig { 2987f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2988f46b5a66SChristoph Hellwig int ret; 2989f46b5a66SChristoph Hellwig 2990e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2991e441d54dSChris Mason return -EPERM; 2992e441d54dSChris Mason 2993171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 2994e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2995c9e9f97bSIlya Dryomov 2996dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2997c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2998c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2999c9e9f97bSIlya Dryomov goto out; 3000c9e9f97bSIlya Dryomov } 3001f46b5a66SChristoph Hellwig 30025516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30032ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 3004f46b5a66SChristoph Hellwig 300543d20761SAnand Jain if (!ret) 30060b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 300743d20761SAnand Jain 3008f46b5a66SChristoph Hellwig kfree(vol_args); 3009c9e9f97bSIlya Dryomov out: 3010171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3011f46b5a66SChristoph Hellwig return ret; 3012f46b5a66SChristoph Hellwig } 3013f46b5a66SChristoph Hellwig 30146b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 3015f46b5a66SChristoph Hellwig { 30160b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30170b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 30186b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 3019f46b5a66SChristoph Hellwig int ret; 3020f46b5a66SChristoph Hellwig 3021e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 3022e441d54dSChris Mason return -EPERM; 3023e441d54dSChris Mason 3024da24927bSMiao Xie ret = mnt_want_write_file(file); 3025da24927bSMiao Xie if (ret) 3026da24927bSMiao Xie return ret; 3027c146afadSYan Zheng 3028dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 3029c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 3030c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 3031c47ca32dSDan Carpenter goto err_drop; 3032c9e9f97bSIlya Dryomov } 3033f46b5a66SChristoph Hellwig 30346b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 3035fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 3036fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 3037fd4e994bSOmar Sandoval goto out; 3038fd4e994bSOmar Sandoval } 3039f46b5a66SChristoph Hellwig 3040171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3041183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3042183860f6SAnand Jain goto out; 3043183860f6SAnand Jain } 3044183860f6SAnand Jain 3045735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 30462ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 30476b526ed7SAnand Jain } else { 30486b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 30492ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 30506b526ed7SAnand Jain } 3051171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 3052183860f6SAnand Jain 30536b526ed7SAnand Jain if (!ret) { 3054735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 30550b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 30566b526ed7SAnand Jain vol_args->devid); 30576b526ed7SAnand Jain else 30580b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 30596b526ed7SAnand Jain vol_args->name); 30606b526ed7SAnand Jain } 3061183860f6SAnand Jain out: 3062183860f6SAnand Jain kfree(vol_args); 3063c47ca32dSDan Carpenter err_drop: 30644ac20c70SIlya Dryomov mnt_drop_write_file(file); 3065f46b5a66SChristoph Hellwig return ret; 3066f46b5a66SChristoph Hellwig } 3067f46b5a66SChristoph Hellwig 3068f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 3069f46b5a66SChristoph Hellwig { 30700b246afaSJeff Mahoney struct inode *inode = file_inode(file); 30710b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3072f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 3073f46b5a66SChristoph Hellwig int ret; 3074f46b5a66SChristoph Hellwig 3075f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3076f46b5a66SChristoph Hellwig return -EPERM; 3077f46b5a66SChristoph Hellwig 3078f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 3079f46b5a66SChristoph Hellwig if (ret) 3080f46b5a66SChristoph Hellwig return ret; 3081f46b5a66SChristoph Hellwig 3082171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 3083f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 308458d7bbf8SDavid Sterba goto out_drop_write; 308558d7bbf8SDavid Sterba } 308658d7bbf8SDavid Sterba 308758d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 308858d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 308958d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 3090f46b5a66SChristoph Hellwig goto out; 3091f46b5a66SChristoph Hellwig } 3092f46b5a66SChristoph Hellwig 309358d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 30942ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 3095f46b5a66SChristoph Hellwig 3096f46b5a66SChristoph Hellwig if (!ret) 30970b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 3098f46b5a66SChristoph Hellwig kfree(vol_args); 309958d7bbf8SDavid Sterba out: 3100171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 310158d7bbf8SDavid Sterba out_drop_write: 3102f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 310358d7bbf8SDavid Sterba 3104f46b5a66SChristoph Hellwig return ret; 3105f46b5a66SChristoph Hellwig } 3106f46b5a66SChristoph Hellwig 31072ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 31082ff7e61eSJeff Mahoney void __user *arg) 3109475f6387SJan Schmidt { 3110027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 3111475f6387SJan Schmidt struct btrfs_device *device; 31120b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 3113027ed2f0SLi Zefan int ret = 0; 3114475f6387SJan Schmidt 3115027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 3116027ed2f0SLi Zefan if (!fi_args) 3117027ed2f0SLi Zefan return -ENOMEM; 3118027ed2f0SLi Zefan 3119d03262c7SDavid Sterba rcu_read_lock(); 3120027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 3121475f6387SJan Schmidt 3122d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 3123027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 3124027ed2f0SLi Zefan fi_args->max_id = device->devid; 3125475f6387SJan Schmidt } 3126d03262c7SDavid Sterba rcu_read_unlock(); 3127475f6387SJan Schmidt 3128d03262c7SDavid Sterba memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 3129bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 3130bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 3131bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 313280a773fbSDavid Sterba 3133027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 3134027ed2f0SLi Zefan ret = -EFAULT; 3135475f6387SJan Schmidt 3136027ed2f0SLi Zefan kfree(fi_args); 3137027ed2f0SLi Zefan return ret; 3138475f6387SJan Schmidt } 3139475f6387SJan Schmidt 31402ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 31412ff7e61eSJeff Mahoney void __user *arg) 3142475f6387SJan Schmidt { 3143475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 3144475f6387SJan Schmidt struct btrfs_device *dev; 3145475f6387SJan Schmidt int ret = 0; 3146475f6387SJan Schmidt char *s_uuid = NULL; 3147475f6387SJan Schmidt 3148475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 3149475f6387SJan Schmidt if (IS_ERR(di_args)) 3150475f6387SJan Schmidt return PTR_ERR(di_args); 3151475f6387SJan Schmidt 3152dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 3153475f6387SJan Schmidt s_uuid = di_args->uuid; 3154475f6387SJan Schmidt 3155c5593ca3SDavid Sterba rcu_read_lock(); 31560b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 3157475f6387SJan Schmidt 3158475f6387SJan Schmidt if (!dev) { 3159475f6387SJan Schmidt ret = -ENODEV; 3160475f6387SJan Schmidt goto out; 3161475f6387SJan Schmidt } 3162475f6387SJan Schmidt 3163475f6387SJan Schmidt di_args->devid = dev->devid; 31647cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 31657cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 3166475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 3167a27202fbSJim Meyering if (dev->name) { 3168606686eeSJosef Bacik struct rcu_string *name; 3169606686eeSJosef Bacik 3170606686eeSJosef Bacik name = rcu_dereference(dev->name); 31716670d4c2SXiongfeng Wang strncpy(di_args->path, name->str, sizeof(di_args->path) - 1); 3172a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 3173a27202fbSJim Meyering } else { 317499ba55adSStefan Behrens di_args->path[0] = '\0'; 3175a27202fbSJim Meyering } 3176475f6387SJan Schmidt 3177475f6387SJan Schmidt out: 3178c5593ca3SDavid Sterba rcu_read_unlock(); 3179475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 3180475f6387SJan Schmidt ret = -EFAULT; 3181475f6387SJan Schmidt 3182475f6387SJan Schmidt kfree(di_args); 3183475f6387SJan Schmidt return ret; 3184475f6387SJan Schmidt } 3185475f6387SJan Schmidt 3186f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 3187416161dbSMark Fasheh { 3188416161dbSMark Fasheh struct page *page; 3189416161dbSMark Fasheh 3190416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 3191416161dbSMark Fasheh if (!page) 319231314002SFilipe Manana return ERR_PTR(-ENOMEM); 3193416161dbSMark Fasheh 3194416161dbSMark Fasheh if (!PageUptodate(page)) { 319531314002SFilipe Manana int ret; 319631314002SFilipe Manana 319731314002SFilipe Manana ret = btrfs_readpage(NULL, page); 319831314002SFilipe Manana if (ret) 319931314002SFilipe Manana return ERR_PTR(ret); 3200416161dbSMark Fasheh lock_page(page); 3201416161dbSMark Fasheh if (!PageUptodate(page)) { 3202416161dbSMark Fasheh unlock_page(page); 320309cbfeafSKirill A. Shutemov put_page(page); 320431314002SFilipe Manana return ERR_PTR(-EIO); 3205416161dbSMark Fasheh } 320631314002SFilipe Manana if (page->mapping != inode->i_mapping) { 3207416161dbSMark Fasheh unlock_page(page); 320809cbfeafSKirill A. Shutemov put_page(page); 320931314002SFilipe Manana return ERR_PTR(-EAGAIN); 3210416161dbSMark Fasheh } 3211416161dbSMark Fasheh } 3212416161dbSMark Fasheh 3213416161dbSMark Fasheh return page; 3214416161dbSMark Fasheh } 3215416161dbSMark Fasheh 3216f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 3217f4414602SMark Fasheh int num_pages, u64 off) 3218f4414602SMark Fasheh { 3219f4414602SMark Fasheh int i; 322009cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 3221f4414602SMark Fasheh 3222f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 322331314002SFilipe Manana again: 3224f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 322531314002SFilipe Manana if (IS_ERR(pages[i])) { 322631314002SFilipe Manana int err = PTR_ERR(pages[i]); 322731314002SFilipe Manana 322831314002SFilipe Manana if (err == -EAGAIN) 322931314002SFilipe Manana goto again; 323031314002SFilipe Manana pages[i] = NULL; 323131314002SFilipe Manana return err; 323231314002SFilipe Manana } 3233f4414602SMark Fasheh } 3234f4414602SMark Fasheh return 0; 3235f4414602SMark Fasheh } 3236f4414602SMark Fasheh 3237e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 3238e0bd70c6SFilipe Manana bool retry_range_locking) 323977fe20dcSMark Fasheh { 3240e0bd70c6SFilipe Manana /* 3241e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 3242e0bd70c6SFilipe Manana * another, and lock file content. 3243e0bd70c6SFilipe Manana * The locking order is: 3244e0bd70c6SFilipe Manana * 3245e0bd70c6SFilipe Manana * 1) pages 3246e0bd70c6SFilipe Manana * 2) range in the inode's io tree 3247e0bd70c6SFilipe Manana */ 324877fe20dcSMark Fasheh while (1) { 324977fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 325077fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 325177fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 325277fe20dcSMark Fasheh off + len - 1); 3253ff5df9b8SFilipe Manana if ((!ordered || 3254ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 3255ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 325677fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 3257ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 3258ff5df9b8SFilipe Manana if (ordered) 3259ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 326077fe20dcSMark Fasheh break; 3261ff5df9b8SFilipe Manana } 326277fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 326377fe20dcSMark Fasheh if (ordered) 326477fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 3265e0bd70c6SFilipe Manana if (!retry_range_locking) 3266e0bd70c6SFilipe Manana return -EAGAIN; 326777fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 326877fe20dcSMark Fasheh } 3269e0bd70c6SFilipe Manana return 0; 327077fe20dcSMark Fasheh } 327177fe20dcSMark Fasheh 3272f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 3273416161dbSMark Fasheh { 32745955102cSAl Viro inode_unlock(inode1); 32755955102cSAl Viro inode_unlock(inode2); 3276416161dbSMark Fasheh } 3277416161dbSMark Fasheh 3278f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 3279f4414602SMark Fasheh { 3280f4414602SMark Fasheh if (inode1 < inode2) 3281f4414602SMark Fasheh swap(inode1, inode2); 3282f4414602SMark Fasheh 32835955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 32845955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 3285f4414602SMark Fasheh } 3286f4414602SMark Fasheh 3287f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 3288f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 3289f4414602SMark Fasheh { 3290f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 3291f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 3292f4414602SMark Fasheh } 3293f4414602SMark Fasheh 3294e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 3295e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 3296e0bd70c6SFilipe Manana bool retry_range_locking) 3297416161dbSMark Fasheh { 3298e0bd70c6SFilipe Manana int ret; 3299e0bd70c6SFilipe Manana 3300416161dbSMark Fasheh if (inode1 < inode2) { 3301416161dbSMark Fasheh swap(inode1, inode2); 3302416161dbSMark Fasheh swap(loff1, loff2); 3303416161dbSMark Fasheh } 3304e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 3305e0bd70c6SFilipe Manana if (ret) 3306e0bd70c6SFilipe Manana return ret; 3307e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 3308e0bd70c6SFilipe Manana if (ret) 3309e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 3310e0bd70c6SFilipe Manana loff1 + len - 1); 3311e0bd70c6SFilipe Manana return ret; 3312416161dbSMark Fasheh } 3313f4414602SMark Fasheh 3314f4414602SMark Fasheh struct cmp_pages { 3315f4414602SMark Fasheh int num_pages; 3316f4414602SMark Fasheh struct page **src_pages; 3317f4414602SMark Fasheh struct page **dst_pages; 3318f4414602SMark Fasheh }; 3319f4414602SMark Fasheh 3320f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 3321f4414602SMark Fasheh { 3322f4414602SMark Fasheh int i; 3323f4414602SMark Fasheh struct page *pg; 3324f4414602SMark Fasheh 3325f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 3326f4414602SMark Fasheh pg = cmp->src_pages[i]; 3327e0bd70c6SFilipe Manana if (pg) { 3328e0bd70c6SFilipe Manana unlock_page(pg); 332909cbfeafSKirill A. Shutemov put_page(pg); 3330e0bd70c6SFilipe Manana } 3331f4414602SMark Fasheh pg = cmp->dst_pages[i]; 3332e0bd70c6SFilipe Manana if (pg) { 3333e0bd70c6SFilipe Manana unlock_page(pg); 333409cbfeafSKirill A. Shutemov put_page(pg); 3335f4414602SMark Fasheh } 3336e0bd70c6SFilipe Manana } 3337f4414602SMark Fasheh } 3338f4414602SMark Fasheh 3339f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 3340f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 3341f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 3342f4414602SMark Fasheh { 3343f4414602SMark Fasheh int ret; 334409cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 3345f4414602SMark Fasheh 3346f4414602SMark Fasheh cmp->num_pages = num_pages; 3347f4414602SMark Fasheh 334867b07bd4STimofey Titovets ret = gather_extent_pages(src, cmp->src_pages, num_pages, loff); 3349f4414602SMark Fasheh if (ret) 3350f4414602SMark Fasheh goto out; 3351f4414602SMark Fasheh 335267b07bd4STimofey Titovets ret = gather_extent_pages(dst, cmp->dst_pages, num_pages, dst_loff); 3353f4414602SMark Fasheh 3354f4414602SMark Fasheh out: 3355f4414602SMark Fasheh if (ret) 3356f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 335778ad4ce0SNaohiro Aota return ret; 3358416161dbSMark Fasheh } 3359416161dbSMark Fasheh 33601a287cfeSDavid Sterba static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp) 3361416161dbSMark Fasheh { 3362416161dbSMark Fasheh int ret = 0; 3363f4414602SMark Fasheh int i; 3364416161dbSMark Fasheh struct page *src_page, *dst_page; 336509cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 3366416161dbSMark Fasheh void *addr, *dst_addr; 3367416161dbSMark Fasheh 3368f4414602SMark Fasheh i = 0; 3369416161dbSMark Fasheh while (len) { 337009cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 3371416161dbSMark Fasheh cmp_len = len; 3372416161dbSMark Fasheh 3373f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 3374f4414602SMark Fasheh 3375f4414602SMark Fasheh src_page = cmp->src_pages[i]; 3376f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 3377e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 3378e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 3379f4414602SMark Fasheh 3380416161dbSMark Fasheh addr = kmap_atomic(src_page); 3381416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 3382416161dbSMark Fasheh 3383416161dbSMark Fasheh flush_dcache_page(src_page); 3384416161dbSMark Fasheh flush_dcache_page(dst_page); 3385416161dbSMark Fasheh 3386416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 33872b3909f8SDarrick J. Wong ret = -EBADE; 3388416161dbSMark Fasheh 3389416161dbSMark Fasheh kunmap_atomic(addr); 3390416161dbSMark Fasheh kunmap_atomic(dst_addr); 3391416161dbSMark Fasheh 3392416161dbSMark Fasheh if (ret) 3393416161dbSMark Fasheh break; 3394416161dbSMark Fasheh 3395416161dbSMark Fasheh len -= cmp_len; 3396f4414602SMark Fasheh i++; 3397416161dbSMark Fasheh } 3398416161dbSMark Fasheh 3399416161dbSMark Fasheh return ret; 3400416161dbSMark Fasheh } 3401416161dbSMark Fasheh 3402e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3403e1d227a4SMark Fasheh u64 olen) 3404416161dbSMark Fasheh { 3405e1d227a4SMark Fasheh u64 len = *plen; 3406416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3407416161dbSMark Fasheh 3408e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 3409416161dbSMark Fasheh return -EINVAL; 3410e1d227a4SMark Fasheh 3411e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 3412e1d227a4SMark Fasheh if (off + len == inode->i_size) 3413e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 3414e1d227a4SMark Fasheh 3415416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 3416416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3417416161dbSMark Fasheh return -EINVAL; 3418416161dbSMark Fasheh 3419416161dbSMark Fasheh return 0; 3420416161dbSMark Fasheh } 3421416161dbSMark Fasheh 34223973909dSTimofey Titovets static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 olen, 342367b07bd4STimofey Titovets struct inode *dst, u64 dst_loff, 342467b07bd4STimofey Titovets struct cmp_pages *cmp) 3425416161dbSMark Fasheh { 3426416161dbSMark Fasheh int ret; 3427e1d227a4SMark Fasheh u64 len = olen; 3428fc4badd9SOmar Sandoval bool same_inode = (src == dst); 34290efa9f48SMark Fasheh u64 same_lock_start = 0; 34300efa9f48SMark Fasheh u64 same_lock_len = 0; 3431416161dbSMark Fasheh 34320efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 34330efa9f48SMark Fasheh if (ret) 34343973909dSTimofey Titovets return ret; 3435fc4badd9SOmar Sandoval 3436fc4badd9SOmar Sandoval ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3437f4dfe687SFilipe Manana if (ret) 34383973909dSTimofey Titovets return ret; 34390efa9f48SMark Fasheh 3440fc4badd9SOmar Sandoval if (same_inode) { 34410efa9f48SMark Fasheh /* 34420efa9f48SMark Fasheh * Single inode case wants the same checks, except we 34430efa9f48SMark Fasheh * don't want our length pushed out past i_size as 34440efa9f48SMark Fasheh * comparing that data range makes no sense. 34450efa9f48SMark Fasheh * 34460efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 34470efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 34480efa9f48SMark Fasheh * reject the request. 34490efa9f48SMark Fasheh * 34500efa9f48SMark Fasheh * This effectively means we require aligned extents 34510efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 34520efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 34530efa9f48SMark Fasheh * i_size. 34540efa9f48SMark Fasheh */ 34553973909dSTimofey Titovets if (len != olen) 34563973909dSTimofey Titovets return -EINVAL; 34570efa9f48SMark Fasheh 34580efa9f48SMark Fasheh /* Check for overlapping ranges */ 34593973909dSTimofey Titovets if (dst_loff + len > loff && dst_loff < loff + len) 34603973909dSTimofey Titovets return -EINVAL; 34610efa9f48SMark Fasheh 34620efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 34630efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 34640efa9f48SMark Fasheh } 3465416161dbSMark Fasheh 3466e0bd70c6SFilipe Manana again: 346767b07bd4STimofey Titovets ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, cmp); 3468f4414602SMark Fasheh if (ret) 34693973909dSTimofey Titovets return ret; 3470f4414602SMark Fasheh 34710efa9f48SMark Fasheh if (same_inode) 3472e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3473e0bd70c6SFilipe Manana false); 34740efa9f48SMark Fasheh else 3475e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3476e0bd70c6SFilipe Manana false); 3477e0bd70c6SFilipe Manana /* 3478e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3479e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3480e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3481e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3482e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3483e0bd70c6SFilipe Manana * range locking). 3484e0bd70c6SFilipe Manana */ 3485e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3486e0bd70c6SFilipe Manana /* 3487e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3488e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3489e0bd70c6SFilipe Manana */ 349067b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 3491e0bd70c6SFilipe Manana if (same_inode) { 3492e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3493e0bd70c6SFilipe Manana same_lock_len); 3494e0bd70c6SFilipe Manana } else { 3495e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3496e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3497e0bd70c6SFilipe Manana } 3498e0bd70c6SFilipe Manana goto again; 3499e0bd70c6SFilipe Manana } 3500e0bd70c6SFilipe Manana ASSERT(ret == 0); 3501e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3502e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 350367b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 3504e0bd70c6SFilipe Manana return ret; 3505e0bd70c6SFilipe Manana } 3506f4414602SMark Fasheh 3507207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 350867b07bd4STimofey Titovets ret = btrfs_cmp_data(olen, cmp); 3509416161dbSMark Fasheh if (ret == 0) 35101c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3511416161dbSMark Fasheh 35120efa9f48SMark Fasheh if (same_inode) 35130efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 35140efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 35150efa9f48SMark Fasheh else 3516f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3517f4414602SMark Fasheh 351867b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 35193973909dSTimofey Titovets 35203973909dSTimofey Titovets return ret; 35213973909dSTimofey Titovets } 35223973909dSTimofey Titovets 3523b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3524b6728768STimofey Titovets 35253973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 35263973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 35273973909dSTimofey Titovets { 35283973909dSTimofey Titovets int ret; 352967b07bd4STimofey Titovets struct cmp_pages cmp; 353067b07bd4STimofey Titovets int num_pages = PAGE_ALIGN(BTRFS_MAX_DEDUPE_LEN) >> PAGE_SHIFT; 35313973909dSTimofey Titovets bool same_inode = (src == dst); 3532b6728768STimofey Titovets u64 i, tail_len, chunk_count; 35333973909dSTimofey Titovets 35343973909dSTimofey Titovets if (olen == 0) 35353973909dSTimofey Titovets return 0; 35363973909dSTimofey Titovets 35373973909dSTimofey Titovets if (same_inode) 35383973909dSTimofey Titovets inode_lock(src); 35393973909dSTimofey Titovets else 35403973909dSTimofey Titovets btrfs_double_inode_lock(src, dst); 35413973909dSTimofey Titovets 35423973909dSTimofey Titovets /* don't make the dst file partly checksummed */ 35433973909dSTimofey Titovets if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 35443973909dSTimofey Titovets (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 35453973909dSTimofey Titovets ret = -EINVAL; 35463973909dSTimofey Titovets goto out_unlock; 35473973909dSTimofey Titovets } 35483973909dSTimofey Titovets 3549b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3550b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 355167b07bd4STimofey Titovets if (chunk_count == 0) 355267b07bd4STimofey Titovets num_pages = PAGE_ALIGN(tail_len) >> PAGE_SHIFT; 355367b07bd4STimofey Titovets 355467b07bd4STimofey Titovets /* 355567b07bd4STimofey Titovets * If deduping ranges in the same inode, locking rules make it 355667b07bd4STimofey Titovets * mandatory to always lock pages in ascending order to avoid deadlocks 355767b07bd4STimofey Titovets * with concurrent tasks (such as starting writeback/delalloc). 355867b07bd4STimofey Titovets */ 355967b07bd4STimofey Titovets if (same_inode && dst_loff < loff) 356067b07bd4STimofey Titovets swap(loff, dst_loff); 356167b07bd4STimofey Titovets 356267b07bd4STimofey Titovets /* 356367b07bd4STimofey Titovets * We must gather up all the pages before we initiate our extent 356467b07bd4STimofey Titovets * locking. We use an array for the page pointers. Size of the array is 356567b07bd4STimofey Titovets * bounded by len, which is in turn bounded by BTRFS_MAX_DEDUPE_LEN. 356667b07bd4STimofey Titovets */ 3567bf5091c8SDavid Sterba cmp.src_pages = kvmalloc_array(num_pages, sizeof(struct page *), 3568bf5091c8SDavid Sterba GFP_KERNEL | __GFP_ZERO); 3569bf5091c8SDavid Sterba cmp.dst_pages = kvmalloc_array(num_pages, sizeof(struct page *), 3570bf5091c8SDavid Sterba GFP_KERNEL | __GFP_ZERO); 357167b07bd4STimofey Titovets if (!cmp.src_pages || !cmp.dst_pages) { 3572bf5091c8SDavid Sterba ret = -ENOMEM; 3573bf5091c8SDavid Sterba goto out_free; 357467b07bd4STimofey Titovets } 3575b6728768STimofey Titovets 3576b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3577b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 357867b07bd4STimofey Titovets dst, dst_loff, &cmp); 3579b6728768STimofey Titovets if (ret) 3580b6728768STimofey Titovets goto out_unlock; 3581b6728768STimofey Titovets 3582b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3583b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3584b6728768STimofey Titovets } 3585b6728768STimofey Titovets 3586b6728768STimofey Titovets if (tail_len > 0) 358767b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 358867b07bd4STimofey Titovets dst_loff, &cmp); 35893973909dSTimofey Titovets 3590416161dbSMark Fasheh out_unlock: 35910efa9f48SMark Fasheh if (same_inode) 35925955102cSAl Viro inode_unlock(src); 35930efa9f48SMark Fasheh else 3594f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3595416161dbSMark Fasheh 3596bf5091c8SDavid Sterba out_free: 3597bf5091c8SDavid Sterba kvfree(cmp.src_pages); 3598bf5091c8SDavid Sterba kvfree(cmp.dst_pages); 359967b07bd4STimofey Titovets 3600416161dbSMark Fasheh return ret; 3601416161dbSMark Fasheh } 3602416161dbSMark Fasheh 36032b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 36042b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3605416161dbSMark Fasheh { 36062b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 36072b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3608416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 36092b3909f8SDarrick J. Wong ssize_t res; 3610416161dbSMark Fasheh 361109cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3612416161dbSMark Fasheh /* 3613416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3614416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3615416161dbSMark Fasheh * this situation without an update. 3616416161dbSMark Fasheh */ 36172b3909f8SDarrick J. Wong return -EINVAL; 3618416161dbSMark Fasheh } 3619416161dbSMark Fasheh 36202b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 36212b3909f8SDarrick J. Wong if (res) 36222b3909f8SDarrick J. Wong return res; 36232b3909f8SDarrick J. Wong return olen; 3624416161dbSMark Fasheh } 3625416161dbSMark Fasheh 3626f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3627f82a9901SFilipe Manana struct inode *inode, 3628f82a9901SFilipe Manana u64 endoff, 3629f82a9901SFilipe Manana const u64 destoff, 36301c919a5eSMark Fasheh const u64 olen, 36311c919a5eSMark Fasheh int no_time_update) 3632f82a9901SFilipe Manana { 3633f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3634f82a9901SFilipe Manana int ret; 3635f82a9901SFilipe Manana 3636f82a9901SFilipe Manana inode_inc_iversion(inode); 36371c919a5eSMark Fasheh if (!no_time_update) 3638c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3639f82a9901SFilipe Manana /* 3640f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3641f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3642f82a9901SFilipe Manana */ 3643f82a9901SFilipe Manana if (endoff > destoff + olen) 3644f82a9901SFilipe Manana endoff = destoff + olen; 3645f82a9901SFilipe Manana if (endoff > inode->i_size) 36466ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3647f82a9901SFilipe Manana 3648f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3649f82a9901SFilipe Manana if (ret) { 365066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 36513a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3652f82a9901SFilipe Manana goto out; 3653f82a9901SFilipe Manana } 36543a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3655f82a9901SFilipe Manana out: 3656f82a9901SFilipe Manana return ret; 3657f82a9901SFilipe Manana } 3658f82a9901SFilipe Manana 3659a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 36607ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 36617ffbb598SFilipe Manana const struct btrfs_path *path, 36627ffbb598SFilipe Manana const u64 hole_offset, 36637ffbb598SFilipe Manana const u64 hole_len) 36647ffbb598SFilipe Manana { 3665a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 36667ffbb598SFilipe Manana struct extent_map *em; 36677ffbb598SFilipe Manana int ret; 36687ffbb598SFilipe Manana 36697ffbb598SFilipe Manana em = alloc_extent_map(); 36707ffbb598SFilipe Manana if (!em) { 3671a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 36727ffbb598SFilipe Manana return; 36737ffbb598SFilipe Manana } 36747ffbb598SFilipe Manana 367514f59796SFilipe Manana if (path) { 367614f59796SFilipe Manana struct btrfs_file_extent_item *fi; 367714f59796SFilipe Manana 367814f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 367914f59796SFilipe Manana struct btrfs_file_extent_item); 3680a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 36817ffbb598SFilipe Manana em->generation = -1; 36827ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 36837ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 36847ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3685a2f392e4SNikolay Borisov &inode->runtime_flags); 36867ffbb598SFilipe Manana } else { 36877ffbb598SFilipe Manana em->start = hole_offset; 36887ffbb598SFilipe Manana em->len = hole_len; 36897ffbb598SFilipe Manana em->ram_bytes = em->len; 36907ffbb598SFilipe Manana em->orig_start = hole_offset; 36917ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 36927ffbb598SFilipe Manana em->block_len = 0; 36937ffbb598SFilipe Manana em->orig_block_len = 0; 36947ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 36957ffbb598SFilipe Manana em->generation = trans->transid; 36967ffbb598SFilipe Manana } 36977ffbb598SFilipe Manana 36987ffbb598SFilipe Manana while (1) { 36997ffbb598SFilipe Manana write_lock(&em_tree->lock); 37007ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 37017ffbb598SFilipe Manana write_unlock(&em_tree->lock); 37027ffbb598SFilipe Manana if (ret != -EEXIST) { 37037ffbb598SFilipe Manana free_extent_map(em); 37047ffbb598SFilipe Manana break; 37057ffbb598SFilipe Manana } 3706a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 37077ffbb598SFilipe Manana em->start + em->len - 1, 0); 37087ffbb598SFilipe Manana } 37097ffbb598SFilipe Manana 3710ee39b432SDavid Sterba if (ret) 3711a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 37127ffbb598SFilipe Manana } 37137ffbb598SFilipe Manana 37148039d87dSFilipe Manana /* 37158039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 37168039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 37178039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 37188039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 37198039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 37208039d87dSFilipe Manana * 37218039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 37228039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 37238039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 37248039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 37258039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 37268039d87dSFilipe Manana * i_mutex, which we are holding here). 37278039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 37288039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 37298039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 37308039d87dSFilipe Manana * 37318039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 37328039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 37338039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 37348039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 37358039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 37368039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 37378039d87dSFilipe Manana * the former. 37388039d87dSFilipe Manana */ 37394a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 37408039d87dSFilipe Manana struct btrfs_trans_handle *trans, 37418039d87dSFilipe Manana struct btrfs_path *path, 37428039d87dSFilipe Manana struct btrfs_key *new_key, 37438039d87dSFilipe Manana const u64 drop_start, 37448039d87dSFilipe Manana const u64 datal, 37458039d87dSFilipe Manana const u64 skip, 37468039d87dSFilipe Manana const u64 size, 37478039d87dSFilipe Manana char *inline_data) 37488039d87dSFilipe Manana { 37490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 37508039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 37518039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 37520b246afaSJeff Mahoney fs_info->sectorsize); 37538039d87dSFilipe Manana int ret; 37548039d87dSFilipe Manana struct btrfs_key key; 37558039d87dSFilipe Manana 37568039d87dSFilipe Manana if (new_key->offset > 0) 37578039d87dSFilipe Manana return -EOPNOTSUPP; 37588039d87dSFilipe Manana 37594a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 37608039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 37618039d87dSFilipe Manana key.offset = 0; 37628039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 37638039d87dSFilipe Manana if (ret < 0) { 37648039d87dSFilipe Manana return ret; 37658039d87dSFilipe Manana } else if (ret > 0) { 37668039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 37678039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 37688039d87dSFilipe Manana if (ret < 0) 37698039d87dSFilipe Manana return ret; 37708039d87dSFilipe Manana else if (ret > 0) 37718039d87dSFilipe Manana goto copy_inline_extent; 37728039d87dSFilipe Manana } 37738039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 37744a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 37758039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 37768039d87dSFilipe Manana ASSERT(key.offset > 0); 37778039d87dSFilipe Manana return -EOPNOTSUPP; 37788039d87dSFilipe Manana } 37798039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 37808039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 37818039d87dSFilipe Manana u64 ext_len; 37828039d87dSFilipe Manana 37838039d87dSFilipe Manana /* 37848039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 37858039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 37868039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 37878039d87dSFilipe Manana */ 37888039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 37898039d87dSFilipe Manana struct btrfs_file_extent_item); 37908039d87dSFilipe Manana /* 37918039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 37928039d87dSFilipe Manana * following it. 37938039d87dSFilipe Manana */ 37948039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 37958039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 37968039d87dSFilipe Manana goto copy_inline_extent; 37978039d87dSFilipe Manana 37988039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 37998039d87dSFilipe Manana if (ext_len > aligned_end) 38008039d87dSFilipe Manana return -EOPNOTSUPP; 38018039d87dSFilipe Manana 38028039d87dSFilipe Manana ret = btrfs_next_item(root, path); 38038039d87dSFilipe Manana if (ret < 0) { 38048039d87dSFilipe Manana return ret; 38058039d87dSFilipe Manana } else if (ret == 0) { 38068039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 38078039d87dSFilipe Manana path->slots[0]); 38084a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 38098039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 38108039d87dSFilipe Manana return -EOPNOTSUPP; 38118039d87dSFilipe Manana } 38128039d87dSFilipe Manana } 38138039d87dSFilipe Manana 38148039d87dSFilipe Manana copy_inline_extent: 38158039d87dSFilipe Manana /* 38168039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 38178039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 38188039d87dSFilipe Manana */ 38198039d87dSFilipe Manana if (i_size_read(dst) > datal) { 38208039d87dSFilipe Manana /* 38218039d87dSFilipe Manana * If the destination inode has an inline extent... 38228039d87dSFilipe Manana * This would require copying the data from the source inline 38238039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 38248039d87dSFilipe Manana * But this is really complex, both extents can be compressed 38258039d87dSFilipe Manana * or just one of them, which would require decompressing and 38268039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 38278039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 38288039d87dSFilipe Manana * inline extent). 38298039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 38308039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 38318039d87dSFilipe Manana */ 38328039d87dSFilipe Manana return -EOPNOTSUPP; 38338039d87dSFilipe Manana } 38348039d87dSFilipe Manana 38358039d87dSFilipe Manana btrfs_release_path(path); 38368039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 38378039d87dSFilipe Manana if (ret) 38388039d87dSFilipe Manana return ret; 38398039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 38408039d87dSFilipe Manana if (ret) 38418039d87dSFilipe Manana return ret; 38428039d87dSFilipe Manana 38438039d87dSFilipe Manana if (skip) { 38448039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 38458039d87dSFilipe Manana 38468039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 38478039d87dSFilipe Manana } 38488039d87dSFilipe Manana 38498039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 38508039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 38518039d87dSFilipe Manana path->slots[0]), 38528039d87dSFilipe Manana size); 38538039d87dSFilipe Manana inode_add_bytes(dst, datal); 38548039d87dSFilipe Manana 38558039d87dSFilipe Manana return 0; 38568039d87dSFilipe Manana } 38578039d87dSFilipe Manana 385832b7c687SMark Fasheh /** 385932b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 386032b7c687SMark Fasheh * 386132b7c687SMark Fasheh * @src: Inode to clone from 386232b7c687SMark Fasheh * @inode: Inode to clone to 386332b7c687SMark Fasheh * @off: Offset within source to start clone from 386432b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 38651c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 386632b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 38671c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 386832b7c687SMark Fasheh */ 386932b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3870f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 38711c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3872f46b5a66SChristoph Hellwig { 38730b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3874f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 387532b7c687SMark Fasheh struct btrfs_path *path = NULL; 3876f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 387732b7c687SMark Fasheh struct btrfs_trans_handle *trans; 387832b7c687SMark Fasheh char *buf = NULL; 3879ae01a0abSYan Zheng struct btrfs_key key; 3880f46b5a66SChristoph Hellwig u32 nritems; 3881f46b5a66SChristoph Hellwig int slot; 3882ae01a0abSYan Zheng int ret; 3883f82a9901SFilipe Manana const u64 len = olen_aligned; 3884f82a9901SFilipe Manana u64 last_dest_end = destoff; 3885ae01a0abSYan Zheng 3886ae01a0abSYan Zheng ret = -ENOMEM; 3887752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3888ae01a0abSYan Zheng if (!buf) 388932b7c687SMark Fasheh return ret; 3890ae01a0abSYan Zheng 3891ae01a0abSYan Zheng path = btrfs_alloc_path(); 3892ae01a0abSYan Zheng if (!path) { 389315351955SDavid Sterba kvfree(buf); 389432b7c687SMark Fasheh return ret; 3895ae01a0abSYan Zheng } 389632b7c687SMark Fasheh 3897e4058b54SDavid Sterba path->reada = READA_FORWARD; 3898c5c9cd4dSSage Weil /* clone data */ 38994a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3900ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 39012c463823SFilipe Manana key.offset = off; 3902f46b5a66SChristoph Hellwig 3903f46b5a66SChristoph Hellwig while (1) { 3904de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3905df858e76SFilipe Manana 3906f46b5a66SChristoph Hellwig /* 3907f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3908f46b5a66SChristoph Hellwig * tree. 3909f46b5a66SChristoph Hellwig */ 3910e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3911362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3912362a20c5SDavid Sterba 0, 0); 3913f46b5a66SChristoph Hellwig if (ret < 0) 3914f46b5a66SChristoph Hellwig goto out; 39152c463823SFilipe Manana /* 39162c463823SFilipe Manana * First search, if no extent item that starts at offset off was 39172c463823SFilipe Manana * found but the previous item is an extent item, it's possible 39182c463823SFilipe Manana * it might overlap our target range, therefore process it. 39192c463823SFilipe Manana */ 39202c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 39212c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 39222c463823SFilipe Manana path->slots[0] - 1); 39232c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 39242c463823SFilipe Manana path->slots[0]--; 39252c463823SFilipe Manana } 3926f46b5a66SChristoph Hellwig 3927ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3928e4355f34SFilipe David Borba Manana process_slot: 3929ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3930362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3931f46b5a66SChristoph Hellwig if (ret < 0) 3932f46b5a66SChristoph Hellwig goto out; 3933f46b5a66SChristoph Hellwig if (ret > 0) 3934f46b5a66SChristoph Hellwig break; 3935ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3936f46b5a66SChristoph Hellwig } 3937f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3938f46b5a66SChristoph Hellwig slot = path->slots[0]; 3939f46b5a66SChristoph Hellwig 3940ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3941962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 39424a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3943f46b5a66SChristoph Hellwig break; 3944f46b5a66SChristoph Hellwig 3945962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3946c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3947c5c9cd4dSSage Weil int type; 394831840ae1SZheng Yan u32 size; 394931840ae1SZheng Yan struct btrfs_key new_key; 3950c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3951c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3952c5c9cd4dSSage Weil u8 comp; 3953f82a9901SFilipe Manana u64 drop_start; 395431840ae1SZheng Yan 3955c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3956c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3957c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3958c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3959c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3960c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3961d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3962d397712bSChris Mason extent); 3963d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3964d397712bSChris Mason extent); 3965c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3966d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3967d397712bSChris Mason extent); 3968c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3969c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3970c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3971c5c9cd4dSSage Weil extent); 3972c5c9cd4dSSage Weil } 397331840ae1SZheng Yan 39742c463823SFilipe Manana /* 39752c463823SFilipe Manana * The first search might have left us at an extent 39762c463823SFilipe Manana * item that ends before our target range's start, can 39772c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 39782c463823SFilipe Manana */ 39792c463823SFilipe Manana if (key.offset + datal <= off) { 3980e4355f34SFilipe David Borba Manana path->slots[0]++; 3981e4355f34SFilipe David Borba Manana goto process_slot; 39822c463823SFilipe Manana } else if (key.offset >= off + len) { 39832c463823SFilipe Manana break; 3984e4355f34SFilipe David Borba Manana } 3985df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3986e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3987e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3988e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3989e4355f34SFilipe David Borba Manana size); 3990e4355f34SFilipe David Borba Manana 3991e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3992e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3993c5c9cd4dSSage Weil 399431840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 39954a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 39964d728ec7SLi Zefan if (off <= key.offset) 3997c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 39984d728ec7SLi Zefan else 39994d728ec7SLi Zefan new_key.offset = destoff; 4000c5c9cd4dSSage Weil 4001b6f3409bSSage Weil /* 4002f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 4003f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 4004f82a9901SFilipe Manana * This hole is either in the middle of the cloning 4005f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 4006f82a9901SFilipe Manana * partially overlaps it). 4007f82a9901SFilipe Manana */ 4008f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 4009f82a9901SFilipe Manana drop_start = last_dest_end; 4010f82a9901SFilipe Manana else 4011f82a9901SFilipe Manana drop_start = new_key.offset; 4012f82a9901SFilipe Manana 4013f82a9901SFilipe Manana /* 4014b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 4015b6f3409bSSage Weil * 1 - add new extent 4016b6f3409bSSage Weil * 1 - inode update 4017b6f3409bSSage Weil */ 4018b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 4019a22285a6SYan, Zheng if (IS_ERR(trans)) { 4020a22285a6SYan, Zheng ret = PTR_ERR(trans); 4021a22285a6SYan, Zheng goto out; 4022a22285a6SYan, Zheng } 4023a22285a6SYan, Zheng 4024c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 4025c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 4026d72c0842SLi Zefan /* 4027d72c0842SLi Zefan * a | --- range to clone ---| b 4028d72c0842SLi Zefan * | ------------- extent ------------- | 4029d72c0842SLi Zefan */ 4030d72c0842SLi Zefan 403193915584SAntonio Ospite /* subtract range b */ 4032d72c0842SLi Zefan if (key.offset + datal > off + len) 4033d72c0842SLi Zefan datal = off + len - key.offset; 4034d72c0842SLi Zefan 403593915584SAntonio Ospite /* subtract range a */ 4036a22285a6SYan, Zheng if (off > key.offset) { 4037a22285a6SYan, Zheng datao += off - key.offset; 4038a22285a6SYan, Zheng datal -= off - key.offset; 4039a22285a6SYan, Zheng } 4040a22285a6SYan, Zheng 40415dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 4042f82a9901SFilipe Manana drop_start, 4043a22285a6SYan, Zheng new_key.offset + datal, 40442671485dSJosef Bacik 1); 404579787eaaSJeff Mahoney if (ret) { 40463f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 404700fdf13aSLiu Bo btrfs_abort_transaction(trans, 404866642832SJeff Mahoney ret); 40493a45bb20SJeff Mahoney btrfs_end_transaction(trans); 405079787eaaSJeff Mahoney goto out; 405179787eaaSJeff Mahoney } 4052a22285a6SYan, Zheng 405331840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 405431840ae1SZheng Yan &new_key, size); 405579787eaaSJeff Mahoney if (ret) { 405666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 40573a45bb20SJeff Mahoney btrfs_end_transaction(trans); 405879787eaaSJeff Mahoney goto out; 405979787eaaSJeff Mahoney } 406031840ae1SZheng Yan 406131840ae1SZheng Yan leaf = path->nodes[0]; 406231840ae1SZheng Yan slot = path->slots[0]; 406331840ae1SZheng Yan write_extent_buffer(leaf, buf, 406431840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 406531840ae1SZheng Yan size); 4066ae01a0abSYan Zheng 4067f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 4068f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 4069c5c9cd4dSSage Weil 4070c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 4071c5c9cd4dSSage Weil if (!disko) 4072c5c9cd4dSSage Weil datao = 0; 4073c5c9cd4dSSage Weil 4074c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 4075c5c9cd4dSSage Weil datao); 4076c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 4077c5c9cd4dSSage Weil datal); 4078fcebe456SJosef Bacik 4079c5c9cd4dSSage Weil if (disko) { 4080c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 40812ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 408284f7d8e6SJosef Bacik root, 40835d4f98a2SYan Zheng disko, diskl, 0, 4084f46b5a66SChristoph Hellwig root->root_key.objectid, 40854a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 4086b06c4bf5SFilipe Manana new_key.offset - datao); 408779787eaaSJeff Mahoney if (ret) { 408879787eaaSJeff Mahoney btrfs_abort_transaction(trans, 408979787eaaSJeff Mahoney ret); 40903a45bb20SJeff Mahoney btrfs_end_transaction(trans); 409179787eaaSJeff Mahoney goto out; 409279787eaaSJeff Mahoney 409379787eaaSJeff Mahoney } 4094f46b5a66SChristoph Hellwig } 4095c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 4096c5c9cd4dSSage Weil u64 skip = 0; 4097c5c9cd4dSSage Weil u64 trim = 0; 4098ed958762SFilipe Manana 4099c5c9cd4dSSage Weil if (off > key.offset) { 4100c5c9cd4dSSage Weil skip = off - key.offset; 4101c5c9cd4dSSage Weil new_key.offset += skip; 410231840ae1SZheng Yan } 4103d397712bSChris Mason 4104c5c9cd4dSSage Weil if (key.offset + datal > off + len) 4105c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 4106d397712bSChris Mason 4107c5c9cd4dSSage Weil if (comp && (skip || trim)) { 4108c5c9cd4dSSage Weil ret = -EINVAL; 41093a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4110c5c9cd4dSSage Weil goto out; 411131840ae1SZheng Yan } 4112c5c9cd4dSSage Weil size -= skip + trim; 4113c5c9cd4dSSage Weil datal -= skip + trim; 4114a22285a6SYan, Zheng 41154a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 41168039d87dSFilipe Manana trans, path, 41178039d87dSFilipe Manana &new_key, 4118f82a9901SFilipe Manana drop_start, 41198039d87dSFilipe Manana datal, 41208039d87dSFilipe Manana skip, size, buf); 412179787eaaSJeff Mahoney if (ret) { 41223f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 41233a29bc09SChris Mason btrfs_abort_transaction(trans, 412479787eaaSJeff Mahoney ret); 41253a45bb20SJeff Mahoney btrfs_end_transaction(trans); 412679787eaaSJeff Mahoney goto out; 412779787eaaSJeff Mahoney } 4128c5c9cd4dSSage Weil leaf = path->nodes[0]; 4129c5c9cd4dSSage Weil slot = path->slots[0]; 4130c5c9cd4dSSage Weil } 4131c5c9cd4dSSage Weil 41327ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 41337ffbb598SFilipe Manana if (drop_start < new_key.offset) 4134a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 413514f59796SFilipe Manana NULL, drop_start, 41367ffbb598SFilipe Manana new_key.offset - drop_start); 41377ffbb598SFilipe Manana 4138a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 4139a2f392e4SNikolay Borisov path, 0, 0); 41407ffbb598SFilipe Manana 4141c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 4142b3b4aa74SDavid Sterba btrfs_release_path(path); 4143c5c9cd4dSSage Weil 414462e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 41450b246afaSJeff Mahoney fs_info->sectorsize); 4146f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 4147f82a9901SFilipe Manana last_dest_end, 41481c919a5eSMark Fasheh destoff, olen, 41491c919a5eSMark Fasheh no_time_update); 4150f82a9901SFilipe Manana if (ret) 415179787eaaSJeff Mahoney goto out; 41522c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 41532c463823SFilipe Manana break; 4154a22285a6SYan, Zheng } 4155b3b4aa74SDavid Sterba btrfs_release_path(path); 4156df858e76SFilipe Manana key.offset = next_key_min_offset; 415769ae5e44SWang Xiaoguang 415869ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 415969ae5e44SWang Xiaoguang ret = -EINTR; 416069ae5e44SWang Xiaoguang goto out; 416169ae5e44SWang Xiaoguang } 4162ae01a0abSYan Zheng } 4163f46b5a66SChristoph Hellwig ret = 0; 416432b7c687SMark Fasheh 4165f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 4166f82a9901SFilipe Manana /* 4167f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 4168f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 4169f82a9901SFilipe Manana */ 4170b3b4aa74SDavid Sterba btrfs_release_path(path); 4171f82a9901SFilipe Manana 4172f82a9901SFilipe Manana /* 4173f82a9901SFilipe Manana * 1 - remove extent(s) 4174f82a9901SFilipe Manana * 1 - inode update 4175f82a9901SFilipe Manana */ 4176f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 4177f82a9901SFilipe Manana if (IS_ERR(trans)) { 4178f82a9901SFilipe Manana ret = PTR_ERR(trans); 4179f82a9901SFilipe Manana goto out; 4180f82a9901SFilipe Manana } 4181f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 4182f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 4183f82a9901SFilipe Manana if (ret) { 4184f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 418566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 41863a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4187f82a9901SFilipe Manana goto out; 4188f82a9901SFilipe Manana } 4189a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 4190a2f392e4SNikolay Borisov last_dest_end, 419114f59796SFilipe Manana destoff + len - last_dest_end); 4192f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 41931c919a5eSMark Fasheh destoff, olen, no_time_update); 4194f82a9901SFilipe Manana } 4195f82a9901SFilipe Manana 4196f46b5a66SChristoph Hellwig out: 419732b7c687SMark Fasheh btrfs_free_path(path); 419815351955SDavid Sterba kvfree(buf); 419932b7c687SMark Fasheh return ret; 420032b7c687SMark Fasheh } 420132b7c687SMark Fasheh 42023db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 420332b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 420432b7c687SMark Fasheh { 420554563d41SAl Viro struct inode *inode = file_inode(file); 42063db11b2eSZach Brown struct inode *src = file_inode(file_src); 42070b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 420832b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 420932b7c687SMark Fasheh int ret; 421032b7c687SMark Fasheh u64 len = olen; 42110b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 42123db11b2eSZach Brown int same_inode = src == inode; 421332b7c687SMark Fasheh 421432b7c687SMark Fasheh /* 421532b7c687SMark Fasheh * TODO: 421632b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 421732b7c687SMark Fasheh * decompress into destination's address_space (the file offset 421832b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 421932b7c687SMark Fasheh * otherwise reinsert) a subrange. 422000fdf13aSLiu Bo * 422100fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 422200fdf13aSLiu Bo * be either compressed or non-compressed. 422332b7c687SMark Fasheh */ 422432b7c687SMark Fasheh 422532b7c687SMark Fasheh if (btrfs_root_readonly(root)) 422632b7c687SMark Fasheh return -EROFS; 422732b7c687SMark Fasheh 42283db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 42293db11b2eSZach Brown src->i_sb != inode->i_sb) 42303db11b2eSZach Brown return -EXDEV; 423132b7c687SMark Fasheh 423232b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 42333db11b2eSZach Brown return -EISDIR; 423432b7c687SMark Fasheh 423532b7c687SMark Fasheh if (!same_inode) { 4236293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 423732b7c687SMark Fasheh } else { 42385955102cSAl Viro inode_lock(src); 423932b7c687SMark Fasheh } 424032b7c687SMark Fasheh 4241b5c40d59SOmar Sandoval /* don't make the dst file partly checksummed */ 4242b5c40d59SOmar Sandoval if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 4243b5c40d59SOmar Sandoval (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) { 4244b5c40d59SOmar Sandoval ret = -EINVAL; 4245b5c40d59SOmar Sandoval goto out_unlock; 4246b5c40d59SOmar Sandoval } 4247b5c40d59SOmar Sandoval 424832b7c687SMark Fasheh /* determine range to clone */ 424932b7c687SMark Fasheh ret = -EINVAL; 425032b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 425132b7c687SMark Fasheh goto out_unlock; 425232b7c687SMark Fasheh if (len == 0) 425332b7c687SMark Fasheh olen = len = src->i_size - off; 425432b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 425532b7c687SMark Fasheh if (off + len == src->i_size) 425632b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 425732b7c687SMark Fasheh 4258ccccf3d6SFilipe Manana if (len == 0) { 4259ccccf3d6SFilipe Manana ret = 0; 4260ccccf3d6SFilipe Manana goto out_unlock; 4261ccccf3d6SFilipe Manana } 4262ccccf3d6SFilipe Manana 426332b7c687SMark Fasheh /* verify the end result is block aligned */ 426432b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 426532b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 426632b7c687SMark Fasheh goto out_unlock; 426732b7c687SMark Fasheh 426832b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 426932b7c687SMark Fasheh if (same_inode) { 427032b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 427132b7c687SMark Fasheh goto out_unlock; 427232b7c687SMark Fasheh } 427332b7c687SMark Fasheh 427432b7c687SMark Fasheh if (destoff > inode->i_size) { 427532b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 427632b7c687SMark Fasheh if (ret) 427732b7c687SMark Fasheh goto out_unlock; 427832b7c687SMark Fasheh } 427932b7c687SMark Fasheh 4280c125b8bfSFilipe Manana /* 4281c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 4282c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 4283c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 4284c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 4285c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 4286c125b8bfSFilipe Manana */ 4287c125b8bfSFilipe Manana if (same_inode) { 4288c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 4289c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 429032b7c687SMark Fasheh 4291e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 4292c125b8bfSFilipe Manana } else { 4293e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 4294e0bd70c6SFilipe Manana true); 4295e0bd70c6SFilipe Manana } 4296e0bd70c6SFilipe Manana ASSERT(ret == 0); 4297e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 4298e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 4299e0bd70c6SFilipe Manana goto out_unlock; 4300c125b8bfSFilipe Manana } 430132b7c687SMark Fasheh 43021c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 430332b7c687SMark Fasheh 4304c125b8bfSFilipe Manana if (same_inode) { 4305c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 4306c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 4307c125b8bfSFilipe Manana 4308c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 4309c125b8bfSFilipe Manana } else { 4310293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 4311c125b8bfSFilipe Manana } 4312c125b8bfSFilipe Manana /* 4313c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 4314c125b8bfSFilipe Manana * data immediately and not the previous data. 4315c125b8bfSFilipe Manana */ 431665bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 431709cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 431809cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 4319f46b5a66SChristoph Hellwig out_unlock: 4320293a8489SMark Fasheh if (!same_inode) 4321293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 4322293a8489SMark Fasheh else 43235955102cSAl Viro inode_unlock(src); 43243db11b2eSZach Brown return ret; 43253db11b2eSZach Brown } 43263db11b2eSZach Brown 432704b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 432804b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 43293db11b2eSZach Brown { 433004b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 4331c5c9cd4dSSage Weil } 4332c5c9cd4dSSage Weil 43336ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 43346ef5ed0dSJosef Bacik { 4335496ad9aaSAl Viro struct inode *inode = file_inode(file); 43360b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 43376ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 43386ef5ed0dSJosef Bacik struct btrfs_root *new_root; 43396ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 43406ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 43416ef5ed0dSJosef Bacik struct btrfs_path *path; 43426ef5ed0dSJosef Bacik struct btrfs_key location; 43436ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 43446ef5ed0dSJosef Bacik u64 objectid = 0; 43456ef5ed0dSJosef Bacik u64 dir_id; 43463c04ce01SMiao Xie int ret; 43476ef5ed0dSJosef Bacik 43486ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 43496ef5ed0dSJosef Bacik return -EPERM; 43506ef5ed0dSJosef Bacik 43513c04ce01SMiao Xie ret = mnt_want_write_file(file); 43523c04ce01SMiao Xie if (ret) 43533c04ce01SMiao Xie return ret; 43543c04ce01SMiao Xie 43553c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 43563c04ce01SMiao Xie ret = -EFAULT; 43573c04ce01SMiao Xie goto out; 43583c04ce01SMiao Xie } 43596ef5ed0dSJosef Bacik 43606ef5ed0dSJosef Bacik if (!objectid) 43611cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 43626ef5ed0dSJosef Bacik 43636ef5ed0dSJosef Bacik location.objectid = objectid; 43646ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 43656ef5ed0dSJosef Bacik location.offset = (u64)-1; 43666ef5ed0dSJosef Bacik 43670b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 43683c04ce01SMiao Xie if (IS_ERR(new_root)) { 43693c04ce01SMiao Xie ret = PTR_ERR(new_root); 43703c04ce01SMiao Xie goto out; 43713c04ce01SMiao Xie } 43726d6d2829Ssatoru takeuchi if (!is_fstree(new_root->objectid)) { 43736d6d2829Ssatoru takeuchi ret = -ENOENT; 43746d6d2829Ssatoru takeuchi goto out; 43756d6d2829Ssatoru takeuchi } 43766ef5ed0dSJosef Bacik 43776ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 43783c04ce01SMiao Xie if (!path) { 43793c04ce01SMiao Xie ret = -ENOMEM; 43803c04ce01SMiao Xie goto out; 43813c04ce01SMiao Xie } 43826ef5ed0dSJosef Bacik path->leave_spinning = 1; 43836ef5ed0dSJosef Bacik 43846ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 438598d5dc13STsutomu Itoh if (IS_ERR(trans)) { 43866ef5ed0dSJosef Bacik btrfs_free_path(path); 43873c04ce01SMiao Xie ret = PTR_ERR(trans); 43883c04ce01SMiao Xie goto out; 43896ef5ed0dSJosef Bacik } 43906ef5ed0dSJosef Bacik 43910b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 43920b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 43936ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4394cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 43956ef5ed0dSJosef Bacik btrfs_free_path(path); 43963a45bb20SJeff Mahoney btrfs_end_transaction(trans); 43970b246afaSJeff Mahoney btrfs_err(fs_info, 43985d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 43993c04ce01SMiao Xie ret = -ENOENT; 44003c04ce01SMiao Xie goto out; 44016ef5ed0dSJosef Bacik } 44026ef5ed0dSJosef Bacik 44036ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 44046ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 44056ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 44066ef5ed0dSJosef Bacik btrfs_free_path(path); 44076ef5ed0dSJosef Bacik 44080b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 44093a45bb20SJeff Mahoney btrfs_end_transaction(trans); 44103c04ce01SMiao Xie out: 44113c04ce01SMiao Xie mnt_drop_write_file(file); 44123c04ce01SMiao Xie return ret; 44136ef5ed0dSJosef Bacik } 44146ef5ed0dSJosef Bacik 4415c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 4416bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4417bf5fc093SJosef Bacik { 4418bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4419bf5fc093SJosef Bacik 4420bf5fc093SJosef Bacik space->total_bytes = 0; 4421bf5fc093SJosef Bacik space->used_bytes = 0; 4422bf5fc093SJosef Bacik space->flags = 0; 4423bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4424bf5fc093SJosef Bacik space->flags = block_group->flags; 4425bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4426bf5fc093SJosef Bacik space->used_bytes += 4427bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4428bf5fc093SJosef Bacik } 4429bf5fc093SJosef Bacik } 4430bf5fc093SJosef Bacik 44312ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 44322ff7e61eSJeff Mahoney void __user *arg) 44331406e432SJosef Bacik { 44341406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 44351406e432SJosef Bacik struct btrfs_ioctl_space_info space; 44361406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 44377fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 443813f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 44391406e432SJosef Bacik struct btrfs_space_info *info; 4440315d8e98SColin Ian King static const u64 types[] = { 4441315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4442bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4443bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4444315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4445315d8e98SColin Ian King }; 4446bf5fc093SJosef Bacik int num_types = 4; 44477fde62bfSChris Mason int alloc_size; 44481406e432SJosef Bacik int ret = 0; 444951788b1bSDan Rosenberg u64 slot_count = 0; 4450bf5fc093SJosef Bacik int i, c; 44511406e432SJosef Bacik 44521406e432SJosef Bacik if (copy_from_user(&space_args, 44531406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 44541406e432SJosef Bacik sizeof(space_args))) 44551406e432SJosef Bacik return -EFAULT; 44561406e432SJosef Bacik 4457bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4458bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4459bf5fc093SJosef Bacik 4460bf5fc093SJosef Bacik info = NULL; 44617fde62bfSChris Mason rcu_read_lock(); 44620b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4463bf5fc093SJosef Bacik list) { 4464bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4465bf5fc093SJosef Bacik info = tmp; 4466bf5fc093SJosef Bacik break; 4467bf5fc093SJosef Bacik } 4468bf5fc093SJosef Bacik } 44697fde62bfSChris Mason rcu_read_unlock(); 44701406e432SJosef Bacik 4471bf5fc093SJosef Bacik if (!info) 4472bf5fc093SJosef Bacik continue; 4473bf5fc093SJosef Bacik 4474bf5fc093SJosef Bacik down_read(&info->groups_sem); 4475bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4476bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4477bf5fc093SJosef Bacik slot_count++; 4478bf5fc093SJosef Bacik } 4479bf5fc093SJosef Bacik up_read(&info->groups_sem); 4480bf5fc093SJosef Bacik } 4481bf5fc093SJosef Bacik 448236523e95SDavid Sterba /* 448336523e95SDavid Sterba * Global block reserve, exported as a space_info 448436523e95SDavid Sterba */ 448536523e95SDavid Sterba slot_count++; 448636523e95SDavid Sterba 44877fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 44887fde62bfSChris Mason if (space_args.space_slots == 0) { 44897fde62bfSChris Mason space_args.total_spaces = slot_count; 44907fde62bfSChris Mason goto out; 44917fde62bfSChris Mason } 4492bf5fc093SJosef Bacik 449351788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4494bf5fc093SJosef Bacik 44957fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4496bf5fc093SJosef Bacik 44977fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 44987fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 44997fde62bfSChris Mason */ 450009cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 45017fde62bfSChris Mason return -ENOMEM; 45027fde62bfSChris Mason 45037fde62bfSChris Mason space_args.total_spaces = 0; 45048d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 45057fde62bfSChris Mason if (!dest) 45067fde62bfSChris Mason return -ENOMEM; 45077fde62bfSChris Mason dest_orig = dest; 45087fde62bfSChris Mason 45097fde62bfSChris Mason /* now we have a buffer to copy into */ 4510bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4511bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4512bf5fc093SJosef Bacik 451351788b1bSDan Rosenberg if (!slot_count) 451451788b1bSDan Rosenberg break; 451551788b1bSDan Rosenberg 4516bf5fc093SJosef Bacik info = NULL; 45171406e432SJosef Bacik rcu_read_lock(); 45180b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4519bf5fc093SJosef Bacik list) { 4520bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4521bf5fc093SJosef Bacik info = tmp; 45227fde62bfSChris Mason break; 4523bf5fc093SJosef Bacik } 4524bf5fc093SJosef Bacik } 4525bf5fc093SJosef Bacik rcu_read_unlock(); 45267fde62bfSChris Mason 4527bf5fc093SJosef Bacik if (!info) 4528bf5fc093SJosef Bacik continue; 4529bf5fc093SJosef Bacik down_read(&info->groups_sem); 4530bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4531bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4532c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4533c065f5b1SSu Yue &space); 45347fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 45351406e432SJosef Bacik dest++; 45361406e432SJosef Bacik space_args.total_spaces++; 453751788b1bSDan Rosenberg slot_count--; 45381406e432SJosef Bacik } 453951788b1bSDan Rosenberg if (!slot_count) 454051788b1bSDan Rosenberg break; 4541bf5fc093SJosef Bacik } 4542bf5fc093SJosef Bacik up_read(&info->groups_sem); 4543bf5fc093SJosef Bacik } 45441406e432SJosef Bacik 454536523e95SDavid Sterba /* 454636523e95SDavid Sterba * Add global block reserve 454736523e95SDavid Sterba */ 454836523e95SDavid Sterba if (slot_count) { 45490b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 455036523e95SDavid Sterba 455136523e95SDavid Sterba spin_lock(&block_rsv->lock); 455236523e95SDavid Sterba space.total_bytes = block_rsv->size; 455336523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 455436523e95SDavid Sterba spin_unlock(&block_rsv->lock); 455536523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 455636523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 455736523e95SDavid Sterba space_args.total_spaces++; 455836523e95SDavid Sterba } 455936523e95SDavid Sterba 45602eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 45617fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 45627fde62bfSChris Mason 45637fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 45647fde62bfSChris Mason ret = -EFAULT; 45657fde62bfSChris Mason 45667fde62bfSChris Mason kfree(dest_orig); 45677fde62bfSChris Mason out: 45687fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 45691406e432SJosef Bacik ret = -EFAULT; 45701406e432SJosef Bacik 45711406e432SJosef Bacik return ret; 45721406e432SJosef Bacik } 45731406e432SJosef Bacik 45749a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 45759a8c28beSMiao Xie void __user *argp) 457646204592SSage Weil { 457746204592SSage Weil struct btrfs_trans_handle *trans; 457846204592SSage Weil u64 transid; 4579db5b493aSTsutomu Itoh int ret; 458046204592SSage Weil 4581d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4582ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4583ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 458498d5dc13STsutomu Itoh return PTR_ERR(trans); 4585ff7c1d33SMiao Xie 4586ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4587ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4588ff7c1d33SMiao Xie goto out; 4589ff7c1d33SMiao Xie } 459046204592SSage Weil transid = trans->transid; 45913a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 45928b2b2d3cSTsutomu Itoh if (ret) { 45933a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4594db5b493aSTsutomu Itoh return ret; 45958b2b2d3cSTsutomu Itoh } 4596ff7c1d33SMiao Xie out: 459746204592SSage Weil if (argp) 459846204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 459946204592SSage Weil return -EFAULT; 460046204592SSage Weil return 0; 460146204592SSage Weil } 460246204592SSage Weil 46032ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 46049a8c28beSMiao Xie void __user *argp) 460546204592SSage Weil { 460646204592SSage Weil u64 transid; 460746204592SSage Weil 460846204592SSage Weil if (argp) { 460946204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 461046204592SSage Weil return -EFAULT; 461146204592SSage Weil } else { 461246204592SSage Weil transid = 0; /* current trans */ 461346204592SSage Weil } 46142ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 461546204592SSage Weil } 461646204592SSage Weil 4617b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4618475f6387SJan Schmidt { 46190b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4620475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4621b8e95489SMiao Xie int ret; 4622475f6387SJan Schmidt 4623475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4624475f6387SJan Schmidt return -EPERM; 4625475f6387SJan Schmidt 4626475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4627475f6387SJan Schmidt if (IS_ERR(sa)) 4628475f6387SJan Schmidt return PTR_ERR(sa); 4629475f6387SJan Schmidt 4630b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4631b8e95489SMiao Xie ret = mnt_want_write_file(file); 4632b8e95489SMiao Xie if (ret) 4633b8e95489SMiao Xie goto out; 4634b8e95489SMiao Xie } 4635b8e95489SMiao Xie 46360b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 463763a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 463863a212abSStefan Behrens 0); 4639475f6387SJan Schmidt 4640475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4641475f6387SJan Schmidt ret = -EFAULT; 4642475f6387SJan Schmidt 4643b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4644b8e95489SMiao Xie mnt_drop_write_file(file); 4645b8e95489SMiao Xie out: 4646475f6387SJan Schmidt kfree(sa); 4647475f6387SJan Schmidt return ret; 4648475f6387SJan Schmidt } 4649475f6387SJan Schmidt 46502ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4651475f6387SJan Schmidt { 4652475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4653475f6387SJan Schmidt return -EPERM; 4654475f6387SJan Schmidt 46552ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4656475f6387SJan Schmidt } 4657475f6387SJan Schmidt 46582ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4659475f6387SJan Schmidt void __user *arg) 4660475f6387SJan Schmidt { 4661475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4662475f6387SJan Schmidt int ret; 4663475f6387SJan Schmidt 4664475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4665475f6387SJan Schmidt return -EPERM; 4666475f6387SJan Schmidt 4667475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4668475f6387SJan Schmidt if (IS_ERR(sa)) 4669475f6387SJan Schmidt return PTR_ERR(sa); 4670475f6387SJan Schmidt 46712ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4672475f6387SJan Schmidt 4673475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4674475f6387SJan Schmidt ret = -EFAULT; 4675475f6387SJan Schmidt 4676475f6387SJan Schmidt kfree(sa); 4677475f6387SJan Schmidt return ret; 4678475f6387SJan Schmidt } 4679475f6387SJan Schmidt 46802ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4681b27f7c0cSDavid Sterba void __user *arg) 4682c11d2c23SStefan Behrens { 4683c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4684c11d2c23SStefan Behrens int ret; 4685c11d2c23SStefan Behrens 4686c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4687c11d2c23SStefan Behrens if (IS_ERR(sa)) 4688c11d2c23SStefan Behrens return PTR_ERR(sa); 4689c11d2c23SStefan Behrens 4690b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4691b27f7c0cSDavid Sterba kfree(sa); 4692b27f7c0cSDavid Sterba return -EPERM; 4693b27f7c0cSDavid Sterba } 4694b27f7c0cSDavid Sterba 46952ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4696c11d2c23SStefan Behrens 4697c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4698c11d2c23SStefan Behrens ret = -EFAULT; 4699c11d2c23SStefan Behrens 4700c11d2c23SStefan Behrens kfree(sa); 4701c11d2c23SStefan Behrens return ret; 4702c11d2c23SStefan Behrens } 4703c11d2c23SStefan Behrens 47042ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 47052ff7e61eSJeff Mahoney void __user *arg) 47063f6bcfbdSStefan Behrens { 47073f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 47083f6bcfbdSStefan Behrens int ret; 47093f6bcfbdSStefan Behrens 47103f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 47113f6bcfbdSStefan Behrens return -EPERM; 47123f6bcfbdSStefan Behrens 47133f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 47143f6bcfbdSStefan Behrens if (IS_ERR(p)) 47153f6bcfbdSStefan Behrens return PTR_ERR(p); 47163f6bcfbdSStefan Behrens 47173f6bcfbdSStefan Behrens switch (p->cmd) { 47183f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4719bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4720adfa97cbSIlya Dryomov ret = -EROFS; 4721adfa97cbSIlya Dryomov goto out; 4722adfa97cbSIlya Dryomov } 4723171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4724e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 47253f6bcfbdSStefan Behrens } else { 47262ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4727171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 47283f6bcfbdSStefan Behrens } 47293f6bcfbdSStefan Behrens break; 47303f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 47310b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 47323f6bcfbdSStefan Behrens ret = 0; 47333f6bcfbdSStefan Behrens break; 47343f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 473517d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 473697282031SAnand Jain ret = 0; 47373f6bcfbdSStefan Behrens break; 47383f6bcfbdSStefan Behrens default: 47393f6bcfbdSStefan Behrens ret = -EINVAL; 47403f6bcfbdSStefan Behrens break; 47413f6bcfbdSStefan Behrens } 47423f6bcfbdSStefan Behrens 47433f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 47443f6bcfbdSStefan Behrens ret = -EFAULT; 4745adfa97cbSIlya Dryomov out: 47463f6bcfbdSStefan Behrens kfree(p); 47473f6bcfbdSStefan Behrens return ret; 47483f6bcfbdSStefan Behrens } 47493f6bcfbdSStefan Behrens 4750d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4751d7728c96SJan Schmidt { 4752d7728c96SJan Schmidt int ret = 0; 4753d7728c96SJan Schmidt int i; 4754740c3d22SChris Mason u64 rel_ptr; 4755d7728c96SJan Schmidt int size; 4756806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4757d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4758d7728c96SJan Schmidt struct btrfs_path *path; 4759d7728c96SJan Schmidt 476082b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4761d7728c96SJan Schmidt return -EPERM; 4762d7728c96SJan Schmidt 4763d7728c96SJan Schmidt path = btrfs_alloc_path(); 4764d7728c96SJan Schmidt if (!path) { 4765d7728c96SJan Schmidt ret = -ENOMEM; 4766d7728c96SJan Schmidt goto out; 4767d7728c96SJan Schmidt } 4768d7728c96SJan Schmidt 4769d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4770d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4771d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4772d7728c96SJan Schmidt ipa = NULL; 4773d7728c96SJan Schmidt goto out; 4774d7728c96SJan Schmidt } 4775d7728c96SJan Schmidt 4776d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4777d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4778d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4779d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4780d7728c96SJan Schmidt ipath = NULL; 4781d7728c96SJan Schmidt goto out; 4782d7728c96SJan Schmidt } 4783d7728c96SJan Schmidt 4784d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4785d7728c96SJan Schmidt if (ret < 0) 4786d7728c96SJan Schmidt goto out; 4787d7728c96SJan Schmidt 4788d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4789745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4790745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4791740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4792d7728c96SJan Schmidt } 4793d7728c96SJan Schmidt 4794718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4795718dc5faSOmar Sandoval ipath->fspath, size); 4796d7728c96SJan Schmidt if (ret) { 4797d7728c96SJan Schmidt ret = -EFAULT; 4798d7728c96SJan Schmidt goto out; 4799d7728c96SJan Schmidt } 4800d7728c96SJan Schmidt 4801d7728c96SJan Schmidt out: 4802d7728c96SJan Schmidt btrfs_free_path(path); 4803d7728c96SJan Schmidt free_ipath(ipath); 4804d7728c96SJan Schmidt kfree(ipa); 4805d7728c96SJan Schmidt 4806d7728c96SJan Schmidt return ret; 4807d7728c96SJan Schmidt } 4808d7728c96SJan Schmidt 4809d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4810d7728c96SJan Schmidt { 4811d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4812d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4813d7728c96SJan Schmidt 4814d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4815d7728c96SJan Schmidt inodes->bytes_left -= c; 4816d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4817d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4818d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4819d7728c96SJan Schmidt inodes->elem_cnt += 3; 4820d7728c96SJan Schmidt } else { 4821d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4822d7728c96SJan Schmidt inodes->bytes_left = 0; 4823d7728c96SJan Schmidt inodes->elem_missed += 3; 4824d7728c96SJan Schmidt } 4825d7728c96SJan Schmidt 4826d7728c96SJan Schmidt return 0; 4827d7728c96SJan Schmidt } 4828d7728c96SJan Schmidt 48292ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4830d24a67b2SZygo Blaxell void __user *arg, int version) 4831d7728c96SJan Schmidt { 4832d7728c96SJan Schmidt int ret = 0; 4833d7728c96SJan Schmidt int size; 4834d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4835d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4836d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4837d24a67b2SZygo Blaxell bool ignore_offset; 4838d7728c96SJan Schmidt 4839d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4840d7728c96SJan Schmidt return -EPERM; 4841d7728c96SJan Schmidt 4842d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 48437b9ea627SShailendra Verma if (IS_ERR(loi)) 48447b9ea627SShailendra Verma return PTR_ERR(loi); 4845d7728c96SJan Schmidt 4846d24a67b2SZygo Blaxell if (version == 1) { 4847d24a67b2SZygo Blaxell ignore_offset = false; 4848b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4849d24a67b2SZygo Blaxell } else { 4850d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4851d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4852d24a67b2SZygo Blaxell ret = -EINVAL; 4853d24a67b2SZygo Blaxell goto out_loi; 4854d24a67b2SZygo Blaxell } 4855d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4856d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4857d24a67b2SZygo Blaxell ret = -EINVAL; 4858d24a67b2SZygo Blaxell goto out_loi; 4859d24a67b2SZygo Blaxell } 4860d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4861b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4862d24a67b2SZygo Blaxell } 4863d24a67b2SZygo Blaxell 4864d7728c96SJan Schmidt path = btrfs_alloc_path(); 4865d7728c96SJan Schmidt if (!path) { 4866d7728c96SJan Schmidt ret = -ENOMEM; 4867d7728c96SJan Schmidt goto out; 4868d7728c96SJan Schmidt } 4869d7728c96SJan Schmidt 4870d7728c96SJan Schmidt inodes = init_data_container(size); 4871d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4872d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4873d7728c96SJan Schmidt inodes = NULL; 4874d7728c96SJan Schmidt goto out; 4875d7728c96SJan Schmidt } 4876d7728c96SJan Schmidt 48772ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4878d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4879df031f07SLiu Bo if (ret == -EINVAL) 4880d7728c96SJan Schmidt ret = -ENOENT; 4881d7728c96SJan Schmidt if (ret < 0) 4882d7728c96SJan Schmidt goto out; 4883d7728c96SJan Schmidt 4884718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4885718dc5faSOmar Sandoval size); 4886d7728c96SJan Schmidt if (ret) 4887d7728c96SJan Schmidt ret = -EFAULT; 4888d7728c96SJan Schmidt 4889d7728c96SJan Schmidt out: 4890d7728c96SJan Schmidt btrfs_free_path(path); 4891f54de068SDavid Sterba kvfree(inodes); 4892d24a67b2SZygo Blaxell out_loi: 4893d7728c96SJan Schmidt kfree(loi); 4894d7728c96SJan Schmidt 4895d7728c96SJan Schmidt return ret; 4896d7728c96SJan Schmidt } 4897d7728c96SJan Schmidt 4898008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4899c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4900c9e9f97bSIlya Dryomov { 4901c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4902c9e9f97bSIlya Dryomov 4903c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4904c9e9f97bSIlya Dryomov 49053009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4906837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4907837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4908837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4909a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4910a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4911837d5b6eSIlya Dryomov 4912c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4913c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4914c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 491519a39dceSIlya Dryomov 491619a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 491719a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 491819a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4919c9e9f97bSIlya Dryomov } 4920c9e9f97bSIlya Dryomov 49219ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4922c9e9f97bSIlya Dryomov { 4923496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4924c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4925c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4926c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4927ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4928c9e9f97bSIlya Dryomov int ret; 4929c9e9f97bSIlya Dryomov 4930c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4931c9e9f97bSIlya Dryomov return -EPERM; 4932c9e9f97bSIlya Dryomov 4933e54bfa31SLiu Bo ret = mnt_want_write_file(file); 49349ba1f6e4SLiu Bo if (ret) 49359ba1f6e4SLiu Bo return ret; 49369ba1f6e4SLiu Bo 4937ed0fb78fSIlya Dryomov again: 4938171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4939c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4940ed0fb78fSIlya Dryomov need_unlock = true; 4941ed0fb78fSIlya Dryomov goto locked; 4942ed0fb78fSIlya Dryomov } 4943ed0fb78fSIlya Dryomov 4944ed0fb78fSIlya Dryomov /* 494501327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4946ed0fb78fSIlya Dryomov * (1) some other op is running 4947ed0fb78fSIlya Dryomov * (2) balance is running 4948ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4949ed0fb78fSIlya Dryomov */ 4950ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4951ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4952ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 49533009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4954ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4955dccdb07bSDavid Sterba /* 4956dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4957dccdb07bSDavid Sterba * we'll reexamine the status again. 4958dccdb07bSDavid Sterba */ 4959ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4960ed0fb78fSIlya Dryomov 4961ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 49623009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4963ed0fb78fSIlya Dryomov /* this is (3) */ 4964ed0fb78fSIlya Dryomov need_unlock = false; 4965ed0fb78fSIlya Dryomov goto locked; 4966ed0fb78fSIlya Dryomov } 4967ed0fb78fSIlya Dryomov 4968ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4969ed0fb78fSIlya Dryomov goto again; 4970ed0fb78fSIlya Dryomov } else { 4971ed0fb78fSIlya Dryomov /* this is (2) */ 4972ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4973ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4974ed0fb78fSIlya Dryomov goto out; 4975ed0fb78fSIlya Dryomov } 4976ed0fb78fSIlya Dryomov } else { 4977ed0fb78fSIlya Dryomov /* this is (1) */ 4978ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4979e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4980ed0fb78fSIlya Dryomov goto out; 4981ed0fb78fSIlya Dryomov } 4982ed0fb78fSIlya Dryomov 4983ed0fb78fSIlya Dryomov locked: 4984171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4985c9e9f97bSIlya Dryomov 4986c9e9f97bSIlya Dryomov if (arg) { 4987c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4988c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4989c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4990ed0fb78fSIlya Dryomov goto out_unlock; 4991c9e9f97bSIlya Dryomov } 4992de322263SIlya Dryomov 4993de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4994de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4995de322263SIlya Dryomov ret = -ENOTCONN; 4996de322263SIlya Dryomov goto out_bargs; 4997de322263SIlya Dryomov } 4998de322263SIlya Dryomov 4999de322263SIlya Dryomov bctl = fs_info->balance_ctl; 5000de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 5001de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 5002de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 5003de322263SIlya Dryomov 5004de322263SIlya Dryomov goto do_balance; 5005de322263SIlya Dryomov } 5006c9e9f97bSIlya Dryomov } else { 5007c9e9f97bSIlya Dryomov bargs = NULL; 5008c9e9f97bSIlya Dryomov } 5009c9e9f97bSIlya Dryomov 5010ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 5011837d5b6eSIlya Dryomov ret = -EINPROGRESS; 5012837d5b6eSIlya Dryomov goto out_bargs; 5013837d5b6eSIlya Dryomov } 5014837d5b6eSIlya Dryomov 50158d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 5016c9e9f97bSIlya Dryomov if (!bctl) { 5017c9e9f97bSIlya Dryomov ret = -ENOMEM; 5018c9e9f97bSIlya Dryomov goto out_bargs; 5019c9e9f97bSIlya Dryomov } 5020c9e9f97bSIlya Dryomov 5021c9e9f97bSIlya Dryomov if (arg) { 5022c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 5023c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 5024c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 5025c9e9f97bSIlya Dryomov 5026c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 5027f43ffb60SIlya Dryomov } else { 5028f43ffb60SIlya Dryomov /* balance everything - no filters */ 5029f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 5030c9e9f97bSIlya Dryomov } 5031c9e9f97bSIlya Dryomov 50328eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 50338eb93459SDavid Sterba ret = -EINVAL; 50340f89abf5SChristian Engelmayer goto out_bctl; 50358eb93459SDavid Sterba } 50368eb93459SDavid Sterba 5037de322263SIlya Dryomov do_balance: 5038c9e9f97bSIlya Dryomov /* 5039149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 5040149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 5041149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 5042149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 5043c9e9f97bSIlya Dryomov */ 5044ed0fb78fSIlya Dryomov need_unlock = false; 5045ed0fb78fSIlya Dryomov 50466fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 50470f89abf5SChristian Engelmayer bctl = NULL; 5048ed0fb78fSIlya Dryomov 5049c9e9f97bSIlya Dryomov if (arg) { 5050c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 5051c9e9f97bSIlya Dryomov ret = -EFAULT; 5052c9e9f97bSIlya Dryomov } 5053c9e9f97bSIlya Dryomov 50540f89abf5SChristian Engelmayer out_bctl: 50550f89abf5SChristian Engelmayer kfree(bctl); 5056c9e9f97bSIlya Dryomov out_bargs: 5057c9e9f97bSIlya Dryomov kfree(bargs); 5058ed0fb78fSIlya Dryomov out_unlock: 5059c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 5060ed0fb78fSIlya Dryomov if (need_unlock) 5061171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 5062ed0fb78fSIlya Dryomov out: 5063e54bfa31SLiu Bo mnt_drop_write_file(file); 5064c9e9f97bSIlya Dryomov return ret; 5065c9e9f97bSIlya Dryomov } 5066c9e9f97bSIlya Dryomov 50672ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 5068837d5b6eSIlya Dryomov { 5069837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 5070837d5b6eSIlya Dryomov return -EPERM; 5071837d5b6eSIlya Dryomov 5072837d5b6eSIlya Dryomov switch (cmd) { 5073837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 50740b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 5075a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 50760b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 5077837d5b6eSIlya Dryomov } 5078837d5b6eSIlya Dryomov 5079837d5b6eSIlya Dryomov return -EINVAL; 5080837d5b6eSIlya Dryomov } 5081837d5b6eSIlya Dryomov 50822ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 508319a39dceSIlya Dryomov void __user *arg) 508419a39dceSIlya Dryomov { 508519a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 508619a39dceSIlya Dryomov int ret = 0; 508719a39dceSIlya Dryomov 508819a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 508919a39dceSIlya Dryomov return -EPERM; 509019a39dceSIlya Dryomov 509119a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 509219a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 509319a39dceSIlya Dryomov ret = -ENOTCONN; 509419a39dceSIlya Dryomov goto out; 509519a39dceSIlya Dryomov } 509619a39dceSIlya Dryomov 50978d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 509819a39dceSIlya Dryomov if (!bargs) { 509919a39dceSIlya Dryomov ret = -ENOMEM; 510019a39dceSIlya Dryomov goto out; 510119a39dceSIlya Dryomov } 510219a39dceSIlya Dryomov 5103008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 510419a39dceSIlya Dryomov 510519a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 510619a39dceSIlya Dryomov ret = -EFAULT; 510719a39dceSIlya Dryomov 510819a39dceSIlya Dryomov kfree(bargs); 510919a39dceSIlya Dryomov out: 511019a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 511119a39dceSIlya Dryomov return ret; 511219a39dceSIlya Dryomov } 511319a39dceSIlya Dryomov 5114905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 51155d13a37bSArne Jansen { 51160b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51170b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51185d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 51195d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 51205d13a37bSArne Jansen int ret; 51215d13a37bSArne Jansen int err; 51225d13a37bSArne Jansen 51235d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 51245d13a37bSArne Jansen return -EPERM; 51255d13a37bSArne Jansen 5126905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5127905b0ddaSMiao Xie if (ret) 5128905b0ddaSMiao Xie return ret; 51295d13a37bSArne Jansen 51305d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5131905b0ddaSMiao Xie if (IS_ERR(sa)) { 5132905b0ddaSMiao Xie ret = PTR_ERR(sa); 5133905b0ddaSMiao Xie goto drop_write; 5134905b0ddaSMiao Xie } 51355d13a37bSArne Jansen 51360b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51370b246afaSJeff Mahoney trans = btrfs_start_transaction(fs_info->tree_root, 2); 51385d13a37bSArne Jansen if (IS_ERR(trans)) { 51395d13a37bSArne Jansen ret = PTR_ERR(trans); 51405d13a37bSArne Jansen goto out; 51415d13a37bSArne Jansen } 51425d13a37bSArne Jansen 51435d13a37bSArne Jansen switch (sa->cmd) { 51445d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 51450b246afaSJeff Mahoney ret = btrfs_quota_enable(trans, fs_info); 51465d13a37bSArne Jansen break; 51475d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 51480b246afaSJeff Mahoney ret = btrfs_quota_disable(trans, fs_info); 51495d13a37bSArne Jansen break; 51505d13a37bSArne Jansen default: 51515d13a37bSArne Jansen ret = -EINVAL; 51525d13a37bSArne Jansen break; 51535d13a37bSArne Jansen } 51545d13a37bSArne Jansen 51553a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 51565d13a37bSArne Jansen if (err && !ret) 51575d13a37bSArne Jansen ret = err; 51585d13a37bSArne Jansen out: 51595d13a37bSArne Jansen kfree(sa); 51600b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5161905b0ddaSMiao Xie drop_write: 5162905b0ddaSMiao Xie mnt_drop_write_file(file); 51635d13a37bSArne Jansen return ret; 51645d13a37bSArne Jansen } 51655d13a37bSArne Jansen 5166905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 51675d13a37bSArne Jansen { 51680b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51690b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51700b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51715d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 51725d13a37bSArne Jansen struct btrfs_trans_handle *trans; 51735d13a37bSArne Jansen int ret; 51745d13a37bSArne Jansen int err; 51755d13a37bSArne Jansen 51765d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 51775d13a37bSArne Jansen return -EPERM; 51785d13a37bSArne Jansen 5179905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5180905b0ddaSMiao Xie if (ret) 5181905b0ddaSMiao Xie return ret; 51825d13a37bSArne Jansen 51835d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5184905b0ddaSMiao Xie if (IS_ERR(sa)) { 5185905b0ddaSMiao Xie ret = PTR_ERR(sa); 5186905b0ddaSMiao Xie goto drop_write; 5187905b0ddaSMiao Xie } 51885d13a37bSArne Jansen 51895d13a37bSArne Jansen trans = btrfs_join_transaction(root); 51905d13a37bSArne Jansen if (IS_ERR(trans)) { 51915d13a37bSArne Jansen ret = PTR_ERR(trans); 51925d13a37bSArne Jansen goto out; 51935d13a37bSArne Jansen } 51945d13a37bSArne Jansen 51955d13a37bSArne Jansen if (sa->assign) { 51960b246afaSJeff Mahoney ret = btrfs_add_qgroup_relation(trans, fs_info, 51975d13a37bSArne Jansen sa->src, sa->dst); 51985d13a37bSArne Jansen } else { 51990b246afaSJeff Mahoney ret = btrfs_del_qgroup_relation(trans, fs_info, 52005d13a37bSArne Jansen sa->src, sa->dst); 52015d13a37bSArne Jansen } 52025d13a37bSArne Jansen 5203e082f563SQu Wenruo /* update qgroup status and info */ 52040b246afaSJeff Mahoney err = btrfs_run_qgroups(trans, fs_info); 5205e082f563SQu Wenruo if (err < 0) 52060b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 5207ad8403dfSAnand Jain "failed to update qgroup status and info"); 52083a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 52095d13a37bSArne Jansen if (err && !ret) 52105d13a37bSArne Jansen ret = err; 52115d13a37bSArne Jansen 52125d13a37bSArne Jansen out: 52135d13a37bSArne Jansen kfree(sa); 5214905b0ddaSMiao Xie drop_write: 5215905b0ddaSMiao Xie mnt_drop_write_file(file); 52165d13a37bSArne Jansen return ret; 52175d13a37bSArne Jansen } 52185d13a37bSArne Jansen 5219905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 52205d13a37bSArne Jansen { 52210b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52220b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52230b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52245d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 52255d13a37bSArne Jansen struct btrfs_trans_handle *trans; 52265d13a37bSArne Jansen int ret; 52275d13a37bSArne Jansen int err; 52285d13a37bSArne Jansen 52295d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 52305d13a37bSArne Jansen return -EPERM; 52315d13a37bSArne Jansen 5232905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5233905b0ddaSMiao Xie if (ret) 5234905b0ddaSMiao Xie return ret; 52355d13a37bSArne Jansen 52365d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5237905b0ddaSMiao Xie if (IS_ERR(sa)) { 5238905b0ddaSMiao Xie ret = PTR_ERR(sa); 5239905b0ddaSMiao Xie goto drop_write; 5240905b0ddaSMiao Xie } 52415d13a37bSArne Jansen 5242d86e56cfSMiao Xie if (!sa->qgroupid) { 5243d86e56cfSMiao Xie ret = -EINVAL; 5244d86e56cfSMiao Xie goto out; 5245d86e56cfSMiao Xie } 5246d86e56cfSMiao Xie 52475d13a37bSArne Jansen trans = btrfs_join_transaction(root); 52485d13a37bSArne Jansen if (IS_ERR(trans)) { 52495d13a37bSArne Jansen ret = PTR_ERR(trans); 52505d13a37bSArne Jansen goto out; 52515d13a37bSArne Jansen } 52525d13a37bSArne Jansen 52535d13a37bSArne Jansen if (sa->create) { 52540b246afaSJeff Mahoney ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); 52555d13a37bSArne Jansen } else { 52560b246afaSJeff Mahoney ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); 52575d13a37bSArne Jansen } 52585d13a37bSArne Jansen 52593a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 52605d13a37bSArne Jansen if (err && !ret) 52615d13a37bSArne Jansen ret = err; 52625d13a37bSArne Jansen 52635d13a37bSArne Jansen out: 52645d13a37bSArne Jansen kfree(sa); 5265905b0ddaSMiao Xie drop_write: 5266905b0ddaSMiao Xie mnt_drop_write_file(file); 52675d13a37bSArne Jansen return ret; 52685d13a37bSArne Jansen } 52695d13a37bSArne Jansen 5270905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 52715d13a37bSArne Jansen { 52720b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52730b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52740b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52755d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 52765d13a37bSArne Jansen struct btrfs_trans_handle *trans; 52775d13a37bSArne Jansen int ret; 52785d13a37bSArne Jansen int err; 52795d13a37bSArne Jansen u64 qgroupid; 52805d13a37bSArne Jansen 52815d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 52825d13a37bSArne Jansen return -EPERM; 52835d13a37bSArne Jansen 5284905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5285905b0ddaSMiao Xie if (ret) 5286905b0ddaSMiao Xie return ret; 52875d13a37bSArne Jansen 52885d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5289905b0ddaSMiao Xie if (IS_ERR(sa)) { 5290905b0ddaSMiao Xie ret = PTR_ERR(sa); 5291905b0ddaSMiao Xie goto drop_write; 5292905b0ddaSMiao Xie } 52935d13a37bSArne Jansen 52945d13a37bSArne Jansen trans = btrfs_join_transaction(root); 52955d13a37bSArne Jansen if (IS_ERR(trans)) { 52965d13a37bSArne Jansen ret = PTR_ERR(trans); 52975d13a37bSArne Jansen goto out; 52985d13a37bSArne Jansen } 52995d13a37bSArne Jansen 53005d13a37bSArne Jansen qgroupid = sa->qgroupid; 53015d13a37bSArne Jansen if (!qgroupid) { 53025d13a37bSArne Jansen /* take the current subvol as qgroup */ 53035d13a37bSArne Jansen qgroupid = root->root_key.objectid; 53045d13a37bSArne Jansen } 53055d13a37bSArne Jansen 53060b246afaSJeff Mahoney ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); 53075d13a37bSArne Jansen 53083a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 53095d13a37bSArne Jansen if (err && !ret) 53105d13a37bSArne Jansen ret = err; 53115d13a37bSArne Jansen 53125d13a37bSArne Jansen out: 53135d13a37bSArne Jansen kfree(sa); 5314905b0ddaSMiao Xie drop_write: 5315905b0ddaSMiao Xie mnt_drop_write_file(file); 53165d13a37bSArne Jansen return ret; 53175d13a37bSArne Jansen } 53185d13a37bSArne Jansen 53192f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 53202f232036SJan Schmidt { 53210b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53220b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53232f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 53242f232036SJan Schmidt int ret; 53252f232036SJan Schmidt 53262f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 53272f232036SJan Schmidt return -EPERM; 53282f232036SJan Schmidt 53292f232036SJan Schmidt ret = mnt_want_write_file(file); 53302f232036SJan Schmidt if (ret) 53312f232036SJan Schmidt return ret; 53322f232036SJan Schmidt 53332f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 53342f232036SJan Schmidt if (IS_ERR(qsa)) { 53352f232036SJan Schmidt ret = PTR_ERR(qsa); 53362f232036SJan Schmidt goto drop_write; 53372f232036SJan Schmidt } 53382f232036SJan Schmidt 53392f232036SJan Schmidt if (qsa->flags) { 53402f232036SJan Schmidt ret = -EINVAL; 53412f232036SJan Schmidt goto out; 53422f232036SJan Schmidt } 53432f232036SJan Schmidt 53440b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 53452f232036SJan Schmidt 53462f232036SJan Schmidt out: 53472f232036SJan Schmidt kfree(qsa); 53482f232036SJan Schmidt drop_write: 53492f232036SJan Schmidt mnt_drop_write_file(file); 53502f232036SJan Schmidt return ret; 53512f232036SJan Schmidt } 53522f232036SJan Schmidt 53532f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 53542f232036SJan Schmidt { 53550b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53560b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53572f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 53582f232036SJan Schmidt int ret = 0; 53592f232036SJan Schmidt 53602f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 53612f232036SJan Schmidt return -EPERM; 53622f232036SJan Schmidt 53638d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 53642f232036SJan Schmidt if (!qsa) 53652f232036SJan Schmidt return -ENOMEM; 53662f232036SJan Schmidt 53670b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 53682f232036SJan Schmidt qsa->flags = 1; 53690b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 53702f232036SJan Schmidt } 53712f232036SJan Schmidt 53722f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 53732f232036SJan Schmidt ret = -EFAULT; 53742f232036SJan Schmidt 53752f232036SJan Schmidt kfree(qsa); 53762f232036SJan Schmidt return ret; 53772f232036SJan Schmidt } 53782f232036SJan Schmidt 537957254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 538057254b6eSJan Schmidt { 53810b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53820b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 538357254b6eSJan Schmidt 538457254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 538557254b6eSJan Schmidt return -EPERM; 538657254b6eSJan Schmidt 53870b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 538857254b6eSJan Schmidt } 538957254b6eSJan Schmidt 5390abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5391abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 53928ea05e3aSAlexander Block { 5393496ad9aaSAl Viro struct inode *inode = file_inode(file); 53940b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53958ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 53968ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 53978ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 5398c2050a45SDeepa Dinamani struct timespec ct = current_time(inode); 53998ea05e3aSAlexander Block int ret = 0; 5400dd5f9615SStefan Behrens int received_uuid_changed; 54018ea05e3aSAlexander Block 5402bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5403bd60ea0fSDavid Sterba return -EPERM; 5404bd60ea0fSDavid Sterba 54058ea05e3aSAlexander Block ret = mnt_want_write_file(file); 54068ea05e3aSAlexander Block if (ret < 0) 54078ea05e3aSAlexander Block return ret; 54088ea05e3aSAlexander Block 54090b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 54108ea05e3aSAlexander Block 54114a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 54128ea05e3aSAlexander Block ret = -EINVAL; 54138ea05e3aSAlexander Block goto out; 54148ea05e3aSAlexander Block } 54158ea05e3aSAlexander Block 54168ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 54178ea05e3aSAlexander Block ret = -EROFS; 54188ea05e3aSAlexander Block goto out; 54198ea05e3aSAlexander Block } 54208ea05e3aSAlexander Block 5421dd5f9615SStefan Behrens /* 5422dd5f9615SStefan Behrens * 1 - root item 5423dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5424dd5f9615SStefan Behrens */ 5425dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 54268ea05e3aSAlexander Block if (IS_ERR(trans)) { 54278ea05e3aSAlexander Block ret = PTR_ERR(trans); 54288ea05e3aSAlexander Block trans = NULL; 54298ea05e3aSAlexander Block goto out; 54308ea05e3aSAlexander Block } 54318ea05e3aSAlexander Block 54328ea05e3aSAlexander Block sa->rtransid = trans->transid; 54338ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 54348ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 54358ea05e3aSAlexander Block 5436dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5437dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5438dd5f9615SStefan Behrens if (received_uuid_changed && 5439d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5440d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5441dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5442dd5f9615SStefan Behrens root->root_key.objectid); 5443d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5444d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5445d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5446d87ff758SNikolay Borisov goto out; 5447d87ff758SNikolay Borisov } 5448d87ff758SNikolay Borisov } 54498ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 54508ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 54518ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 54523cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 54533cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 54543cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 54553cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 54568ea05e3aSAlexander Block 54570b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 54588ea05e3aSAlexander Block &root->root_key, &root->root_item); 54598ea05e3aSAlexander Block if (ret < 0) { 54603a45bb20SJeff Mahoney btrfs_end_transaction(trans); 54618ea05e3aSAlexander Block goto out; 5462dd5f9615SStefan Behrens } 5463dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5464cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5465dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5466dd5f9615SStefan Behrens root->root_key.objectid); 5467dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 546866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5469efd38150SNikolay Borisov btrfs_end_transaction(trans); 5470dd5f9615SStefan Behrens goto out; 5471dd5f9615SStefan Behrens } 5472dd5f9615SStefan Behrens } 54733a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5474abccd00fSHugo Mills out: 54750b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5476abccd00fSHugo Mills mnt_drop_write_file(file); 5477abccd00fSHugo Mills return ret; 5478abccd00fSHugo Mills } 5479abccd00fSHugo Mills 5480abccd00fSHugo Mills #ifdef CONFIG_64BIT 5481abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5482abccd00fSHugo Mills void __user *arg) 5483abccd00fSHugo Mills { 5484abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5485abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5486abccd00fSHugo Mills int ret = 0; 5487abccd00fSHugo Mills 5488abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 54897b9ea627SShailendra Verma if (IS_ERR(args32)) 54907b9ea627SShailendra Verma return PTR_ERR(args32); 5491abccd00fSHugo Mills 54928d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 549384dbeb87SDan Carpenter if (!args64) { 549484dbeb87SDan Carpenter ret = -ENOMEM; 5495abccd00fSHugo Mills goto out; 5496abccd00fSHugo Mills } 5497abccd00fSHugo Mills 5498abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5499abccd00fSHugo Mills args64->stransid = args32->stransid; 5500abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5501abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5502abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5503abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5504abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5505abccd00fSHugo Mills args64->flags = args32->flags; 5506abccd00fSHugo Mills 5507abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5508abccd00fSHugo Mills if (ret) 5509abccd00fSHugo Mills goto out; 5510abccd00fSHugo Mills 5511abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5512abccd00fSHugo Mills args32->stransid = args64->stransid; 5513abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5514abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5515abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5516abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5517abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5518abccd00fSHugo Mills args32->flags = args64->flags; 5519abccd00fSHugo Mills 5520abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5521abccd00fSHugo Mills if (ret) 5522abccd00fSHugo Mills ret = -EFAULT; 5523abccd00fSHugo Mills 5524abccd00fSHugo Mills out: 5525abccd00fSHugo Mills kfree(args32); 5526abccd00fSHugo Mills kfree(args64); 5527abccd00fSHugo Mills return ret; 5528abccd00fSHugo Mills } 5529abccd00fSHugo Mills #endif 5530abccd00fSHugo Mills 5531abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5532abccd00fSHugo Mills void __user *arg) 5533abccd00fSHugo Mills { 5534abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5535abccd00fSHugo Mills int ret = 0; 5536abccd00fSHugo Mills 5537abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 55387b9ea627SShailendra Verma if (IS_ERR(sa)) 55397b9ea627SShailendra Verma return PTR_ERR(sa); 5540abccd00fSHugo Mills 5541abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5542abccd00fSHugo Mills 5543abccd00fSHugo Mills if (ret) 5544abccd00fSHugo Mills goto out; 5545abccd00fSHugo Mills 55468ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 55478ea05e3aSAlexander Block if (ret) 55488ea05e3aSAlexander Block ret = -EFAULT; 55498ea05e3aSAlexander Block 55508ea05e3aSAlexander Block out: 55518ea05e3aSAlexander Block kfree(sa); 55528ea05e3aSAlexander Block return ret; 55538ea05e3aSAlexander Block } 55548ea05e3aSAlexander Block 5555867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5556867ab667Sjeff.liu { 55570b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5559a1b83ac5SAnand Jain size_t len; 5560867ab667Sjeff.liu int ret; 5561a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5562a1b83ac5SAnand Jain 55630b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 55640b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 55650b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5566a1b83ac5SAnand Jain 5567a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5568867ab667Sjeff.liu 5569867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 55700b246afaSJeff Mahoney btrfs_warn(fs_info, 55710b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 55720b246afaSJeff Mahoney --len); 5573867ab667Sjeff.liu } 5574867ab667Sjeff.liu 5575867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5576867ab667Sjeff.liu 5577867ab667Sjeff.liu return ret ? -EFAULT : 0; 5578867ab667Sjeff.liu } 5579867ab667Sjeff.liu 5580a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5581a8bfd4abSjeff.liu { 55820b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55830b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55840b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55850b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5586a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5587a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5588a8bfd4abSjeff.liu int ret; 5589a8bfd4abSjeff.liu 5590a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5591a8bfd4abSjeff.liu return -EPERM; 5592a8bfd4abSjeff.liu 5593a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5594a8bfd4abSjeff.liu return -EFAULT; 5595a8bfd4abSjeff.liu 5596a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 55970b246afaSJeff Mahoney btrfs_err(fs_info, 55985d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5599a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5600a8bfd4abSjeff.liu return -EINVAL; 5601a8bfd4abSjeff.liu } 5602a8bfd4abSjeff.liu 5603a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5604a8bfd4abSjeff.liu if (ret) 5605a8bfd4abSjeff.liu return ret; 5606a8bfd4abSjeff.liu 5607a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5608a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5609a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5610a8bfd4abSjeff.liu goto out_unlock; 5611a8bfd4abSjeff.liu } 5612a8bfd4abSjeff.liu 56130b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5614a8bfd4abSjeff.liu strcpy(super_block->label, label); 56150b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 56163a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5617a8bfd4abSjeff.liu 5618a8bfd4abSjeff.liu out_unlock: 5619a8bfd4abSjeff.liu mnt_drop_write_file(file); 5620a8bfd4abSjeff.liu return ret; 5621a8bfd4abSjeff.liu } 5622a8bfd4abSjeff.liu 56232eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 56242eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 56252eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 56262eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 56272eaa055fSJeff Mahoney 5628d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 56292eaa055fSJeff Mahoney { 56304d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 56312eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 56322eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 56332eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 56342eaa055fSJeff Mahoney }; 56352eaa055fSJeff Mahoney 56362eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 56372eaa055fSJeff Mahoney return -EFAULT; 56382eaa055fSJeff Mahoney 56392eaa055fSJeff Mahoney return 0; 56402eaa055fSJeff Mahoney } 56412eaa055fSJeff Mahoney 56422eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 56432eaa055fSJeff Mahoney { 56440b246afaSJeff Mahoney struct inode *inode = file_inode(file); 56450b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 56460b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 56472eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 56482eaa055fSJeff Mahoney 56492eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 56502eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 56512eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 56522eaa055fSJeff Mahoney 56532eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 56542eaa055fSJeff Mahoney return -EFAULT; 56552eaa055fSJeff Mahoney 56562eaa055fSJeff Mahoney return 0; 56572eaa055fSJeff Mahoney } 56582eaa055fSJeff Mahoney 56592ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 56603b02a68aSJeff Mahoney enum btrfs_feature_set set, 56612eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 56622eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 56632eaa055fSJeff Mahoney { 56643b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 56653b02a68aSJeff Mahoney char *names; 56662eaa055fSJeff Mahoney u64 disallowed, unsupported; 56672eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 56682eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 56692eaa055fSJeff Mahoney 56702eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 56712eaa055fSJeff Mahoney if (unsupported) { 56723b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 56733b02a68aSJeff Mahoney if (names) { 56740b246afaSJeff Mahoney btrfs_warn(fs_info, 56753b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 56763b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 56773b02a68aSJeff Mahoney kfree(names); 56783b02a68aSJeff Mahoney } else 56790b246afaSJeff Mahoney btrfs_warn(fs_info, 56802eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 56812eaa055fSJeff Mahoney type, unsupported); 56822eaa055fSJeff Mahoney return -EOPNOTSUPP; 56832eaa055fSJeff Mahoney } 56842eaa055fSJeff Mahoney 56852eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 56862eaa055fSJeff Mahoney if (disallowed) { 56873b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 56883b02a68aSJeff Mahoney if (names) { 56890b246afaSJeff Mahoney btrfs_warn(fs_info, 56903b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 56913b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 56923b02a68aSJeff Mahoney kfree(names); 56933b02a68aSJeff Mahoney } else 56940b246afaSJeff Mahoney btrfs_warn(fs_info, 56952eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 56962eaa055fSJeff Mahoney type, disallowed); 56972eaa055fSJeff Mahoney return -EPERM; 56982eaa055fSJeff Mahoney } 56992eaa055fSJeff Mahoney 57002eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 57012eaa055fSJeff Mahoney if (disallowed) { 57023b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 57033b02a68aSJeff Mahoney if (names) { 57040b246afaSJeff Mahoney btrfs_warn(fs_info, 57053b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 57063b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 57073b02a68aSJeff Mahoney kfree(names); 57083b02a68aSJeff Mahoney } else 57090b246afaSJeff Mahoney btrfs_warn(fs_info, 57102eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 57112eaa055fSJeff Mahoney type, disallowed); 57122eaa055fSJeff Mahoney return -EPERM; 57132eaa055fSJeff Mahoney } 57142eaa055fSJeff Mahoney 57152eaa055fSJeff Mahoney return 0; 57162eaa055fSJeff Mahoney } 57172eaa055fSJeff Mahoney 57182ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 57192ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 57202eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 57212eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 57222eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 57232eaa055fSJeff Mahoney 57242eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 57252eaa055fSJeff Mahoney { 57260b246afaSJeff Mahoney struct inode *inode = file_inode(file); 57270b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 57280b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 57290b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 57302eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 57312eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 57322eaa055fSJeff Mahoney u64 newflags; 57332eaa055fSJeff Mahoney int ret; 57342eaa055fSJeff Mahoney 57352eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 57362eaa055fSJeff Mahoney return -EPERM; 57372eaa055fSJeff Mahoney 57382eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 57392eaa055fSJeff Mahoney return -EFAULT; 57402eaa055fSJeff Mahoney 57412eaa055fSJeff Mahoney /* Nothing to do */ 57422eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 57432eaa055fSJeff Mahoney !flags[0].incompat_flags) 57442eaa055fSJeff Mahoney return 0; 57452eaa055fSJeff Mahoney 57462ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 57472eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 57482eaa055fSJeff Mahoney if (ret) 57492eaa055fSJeff Mahoney return ret; 57502eaa055fSJeff Mahoney 57512ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 57522eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 57532eaa055fSJeff Mahoney if (ret) 57542eaa055fSJeff Mahoney return ret; 57552eaa055fSJeff Mahoney 57562ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 57572eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 57582eaa055fSJeff Mahoney if (ret) 57592eaa055fSJeff Mahoney return ret; 57602eaa055fSJeff Mahoney 57617ab19625SDavid Sterba ret = mnt_want_write_file(file); 57627ab19625SDavid Sterba if (ret) 57637ab19625SDavid Sterba return ret; 57647ab19625SDavid Sterba 57658051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 57667ab19625SDavid Sterba if (IS_ERR(trans)) { 57677ab19625SDavid Sterba ret = PTR_ERR(trans); 57687ab19625SDavid Sterba goto out_drop_write; 57697ab19625SDavid Sterba } 57702eaa055fSJeff Mahoney 57710b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 57722eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 57732eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 57742eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 57752eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 57762eaa055fSJeff Mahoney 57772eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 57782eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 57792eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 57802eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 57812eaa055fSJeff Mahoney 57822eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 57832eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 57842eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 57852eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 57860b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 57872eaa055fSJeff Mahoney 57883a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 57897ab19625SDavid Sterba out_drop_write: 57907ab19625SDavid Sterba mnt_drop_write_file(file); 57917ab19625SDavid Sterba 57927ab19625SDavid Sterba return ret; 57932eaa055fSJeff Mahoney } 57942eaa055fSJeff Mahoney 57952351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 57962351f431SJosef Bacik { 57972351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 57982351f431SJosef Bacik int ret; 57992351f431SJosef Bacik 58002351f431SJosef Bacik if (compat) { 58012351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 58022351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 58032351f431SJosef Bacik 58042351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 58052351f431SJosef Bacik if (ret) 58062351f431SJosef Bacik return -EFAULT; 58072351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 58082351f431SJosef Bacik if (!arg) 58092351f431SJosef Bacik return -ENOMEM; 58102351f431SJosef Bacik arg->send_fd = args32.send_fd; 58112351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 58122351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 58132351f431SJosef Bacik arg->parent_root = args32.parent_root; 58142351f431SJosef Bacik arg->flags = args32.flags; 58152351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 58162351f431SJosef Bacik sizeof(args32.reserved)); 58172351f431SJosef Bacik #else 58182351f431SJosef Bacik return -ENOTTY; 58192351f431SJosef Bacik #endif 58202351f431SJosef Bacik } else { 58212351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 58222351f431SJosef Bacik if (IS_ERR(arg)) 58232351f431SJosef Bacik return PTR_ERR(arg); 58242351f431SJosef Bacik } 58252351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 58262351f431SJosef Bacik kfree(arg); 58272351f431SJosef Bacik return ret; 58282351f431SJosef Bacik } 58292351f431SJosef Bacik 5830f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5831f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5832f46b5a66SChristoph Hellwig { 58330b246afaSJeff Mahoney struct inode *inode = file_inode(file); 58340b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 58350b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 58364bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5837f46b5a66SChristoph Hellwig 5838f46b5a66SChristoph Hellwig switch (cmd) { 58396cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 58406cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 58416cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 58426cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 58436cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 58446cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5845f7039b1dSLi Dongyang case FITRIM: 5846f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5847f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5848fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5849fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5850fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 58513de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5852fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 58536f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 58546f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 585576dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 585676dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 58570caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 58580caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 58590caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 58600caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 58616ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 58626ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5863f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 58641e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 58651e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 58661e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5867f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5868198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5869f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 58702ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5871f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5872da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 58736b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 58746b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5875475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 58762ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5877475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 58782ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5879f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 58809ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5881ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5882ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5883cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5884cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5885ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5886ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5887d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5888d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5889d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5890d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5891d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5892d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 58931406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 58942ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 58959b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 58969b199859SFilipe David Borba Manana int ret; 58979b199859SFilipe David Borba Manana 589882b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 58999b199859SFilipe David Borba Manana if (ret) 59009b199859SFilipe David Borba Manana return ret; 59010b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 59022fad4e83SDavid Sterba /* 59032fad4e83SDavid Sterba * The transaction thread may want to do more work, 590401327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 59052fad4e83SDavid Sterba * processing uncleaned subvols. 59062fad4e83SDavid Sterba */ 59070b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 59089b199859SFilipe David Borba Manana return ret; 59099b199859SFilipe David Borba Manana } 591046204592SSage Weil case BTRFS_IOC_START_SYNC: 59119a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 591246204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 59132ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5914475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5915b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5916475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 59172ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5918475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 59192ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5920c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 59219ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5922837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 59232ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 592419a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 59252ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 59268ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 59278ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5928abccd00fSHugo Mills #ifdef CONFIG_64BIT 5929abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5930abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5931abccd00fSHugo Mills #endif 593231db9f7cSAlexander Block case BTRFS_IOC_SEND: 59332351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 59342351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 59352351f431SJosef Bacik case BTRFS_IOC_SEND_32: 59362351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 59372351f431SJosef Bacik #endif 5938c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 59392ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 59405d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5941905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 59425d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5943905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 59445d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5945905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 59465d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5947905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 59482f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 59492f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 59502f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 59512f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 595257254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 595357254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 59543f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 59552ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5956867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5957867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5958a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5959a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 59602eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5961d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 59622eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 59632eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 59642eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 59652eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5966e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5967e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5968025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5969025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5970b64ec075STomohiro Misono case BTRFS_IOC_GET_SUBVOL_INFO: 5971b64ec075STomohiro Misono return btrfs_ioctl_get_subvol_info(file, argp); 597242e4b520STomohiro Misono case BTRFS_IOC_GET_SUBVOL_ROOTREF: 597342e4b520STomohiro Misono return btrfs_ioctl_get_subvol_rootref(file, argp); 597423d0b79dSTomohiro Misono case BTRFS_IOC_INO_LOOKUP_USER: 597523d0b79dSTomohiro Misono return btrfs_ioctl_ino_lookup_user(file, argp); 5976f46b5a66SChristoph Hellwig } 5977f46b5a66SChristoph Hellwig 5978f46b5a66SChristoph Hellwig return -ENOTTY; 5979f46b5a66SChristoph Hellwig } 59804c63c245SLuke Dashjr 59814c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 59824c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 59834c63c245SLuke Dashjr { 59842a362249SJeff Mahoney /* 59852a362249SJeff Mahoney * These all access 32-bit values anyway so no further 59862a362249SJeff Mahoney * handling is necessary. 59872a362249SJeff Mahoney */ 59884c63c245SLuke Dashjr switch (cmd) { 59894c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 59904c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 59914c63c245SLuke Dashjr break; 59924c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 59934c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 59944c63c245SLuke Dashjr break; 59954c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 59964c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 59974c63c245SLuke Dashjr break; 59984c63c245SLuke Dashjr } 59994c63c245SLuke Dashjr 60004c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 60014c63c245SLuke Dashjr } 60024c63c245SLuke Dashjr #endif 6003