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; 572d5c12070SMiao Xie u64 qgroup_reserved; 5738ea05e3aSAlexander Block uuid_le new_uuid; 574f46b5a66SChristoph Hellwig 57549a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 57649a3c4d9SDavid Sterba if (!root_item) 57749a3c4d9SDavid Sterba return -ENOMEM; 57849a3c4d9SDavid Sterba 5790b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 5802fbe8c8aSAl Viro if (ret) 58149a3c4d9SDavid Sterba goto fail_free; 5826a912213SJosef Bacik 583e09fe2d2SQu Wenruo /* 584e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 58501327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 586e09fe2d2SQu Wenruo */ 58749a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 58849a3c4d9SDavid Sterba ret = -ENOSPC; 58949a3c4d9SDavid Sterba goto fail_free; 59049a3c4d9SDavid Sterba } 591e09fe2d2SQu Wenruo 592d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 5939ed74f2dSJosef Bacik /* 594d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 595d5c12070SMiao Xie * of create_snapshot(). 5969ed74f2dSJosef Bacik */ 597d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 598dd5f9615SStefan Behrens 8, &qgroup_reserved, false); 599d5c12070SMiao Xie if (ret) 60049a3c4d9SDavid Sterba goto fail_free; 601f46b5a66SChristoph Hellwig 602d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 603d5c12070SMiao Xie if (IS_ERR(trans)) { 604d5c12070SMiao Xie ret = PTR_ERR(trans); 6057775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 60649a3c4d9SDavid Sterba goto fail_free; 607d5c12070SMiao Xie } 608d5c12070SMiao Xie trans->block_rsv = &block_rsv; 609d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 610f46b5a66SChristoph Hellwig 6110b246afaSJeff Mahoney ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); 6126f72c7e2SArne Jansen if (ret) 6136f72c7e2SArne Jansen goto fail; 6146f72c7e2SArne Jansen 6154d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 6168e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 6178e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 6188e8a1e31SJosef Bacik goto fail; 6198e8a1e31SJosef Bacik } 620f46b5a66SChristoph Hellwig 621b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 622f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 623f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 6245d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 625f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 626f46b5a66SChristoph Hellwig 6270b246afaSJeff Mahoney write_extent_buffer_fsid(leaf, fs_info->fsid); 6280b246afaSJeff Mahoney write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 629f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 630f46b5a66SChristoph Hellwig 63149a3c4d9SDavid Sterba inode_item = &root_item->inode; 6323cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 6333cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 6343cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 635da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 6360b246afaSJeff Mahoney fs_info->nodesize); 6373cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 638f46b5a66SChristoph Hellwig 63949a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 64049a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 6413cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 64208fe4db1SLi Zefan 64349a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 64449a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 64549a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 64649a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 64749a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 64849a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 649f46b5a66SChristoph Hellwig 65049a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 65149a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 6528ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 65349a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 65449a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 65549a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 65649a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 65749a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 65849a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 659f46b5a66SChristoph Hellwig 660925baeddSChris Mason btrfs_tree_unlock(leaf); 661f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 662f46b5a66SChristoph Hellwig leaf = NULL; 663f46b5a66SChristoph Hellwig 66449a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 665f46b5a66SChristoph Hellwig 666f46b5a66SChristoph Hellwig key.objectid = objectid; 6675d4f98a2SYan Zheng key.offset = 0; 668962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 6690b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 67049a3c4d9SDavid Sterba root_item); 671f46b5a66SChristoph Hellwig if (ret) 672f46b5a66SChristoph Hellwig goto fail; 673f46b5a66SChristoph Hellwig 67476dda93cSYan, Zheng key.offset = (u64)-1; 6750b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 67679787eaaSJeff Mahoney if (IS_ERR(new_root)) { 67779787eaaSJeff Mahoney ret = PTR_ERR(new_root); 67866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 67979787eaaSJeff Mahoney goto fail; 68079787eaaSJeff Mahoney } 68176dda93cSYan, Zheng 68276dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 68376dda93cSYan, Zheng 68463541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 685ce598979SMark Fasheh if (ret) { 686ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 68766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 688ce598979SMark Fasheh goto fail; 689ce598979SMark Fasheh } 690ce598979SMark Fasheh 691f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 692f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 693f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 694f32e48e9SChandan Rajendra 695f46b5a66SChristoph Hellwig /* 696f46b5a66SChristoph Hellwig * insert the directory item 697f46b5a66SChristoph Hellwig */ 698877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 69979787eaaSJeff Mahoney if (ret) { 70066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 70179787eaaSJeff Mahoney goto fail; 70279787eaaSJeff Mahoney } 7033de4586cSChris Mason 7043de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 7058e7611cfSNikolay Borisov name, namelen, BTRFS_I(dir), &key, 7063de4586cSChris Mason BTRFS_FT_DIR, index); 70779787eaaSJeff Mahoney if (ret) { 70866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 709f46b5a66SChristoph Hellwig goto fail; 71079787eaaSJeff Mahoney } 7110660b5afSChris Mason 7126ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 71352c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 71452c26179SYan Zheng BUG_ON(ret); 71552c26179SYan Zheng 7160b246afaSJeff Mahoney ret = btrfs_add_root_ref(trans, fs_info, 7174df27c4dSYan, Zheng objectid, root->root_key.objectid, 7184a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 7190660b5afSChris Mason BUG_ON(ret); 7200660b5afSChris Mason 721cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, root_item->uuid, 7226bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 723dd5f9615SStefan Behrens if (ret) 72466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 725dd5f9615SStefan Behrens 726f46b5a66SChristoph Hellwig fail: 72749a3c4d9SDavid Sterba kfree(root_item); 728d5c12070SMiao Xie trans->block_rsv = NULL; 729d5c12070SMiao Xie trans->bytes_reserved = 0; 7307775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 731de6e8200SLiu Bo 73272fd032eSSage Weil if (async_transid) { 73372fd032eSSage Weil *async_transid = trans->transid; 7343a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 73500d71c9cSMiao Xie if (err) 7363a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73772fd032eSSage Weil } else { 7383a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 73972fd032eSSage Weil } 740f46b5a66SChristoph Hellwig if (err && !ret) 741f46b5a66SChristoph Hellwig ret = err; 7421a65e24bSChris Mason 7435662344bSTsutomu Itoh if (!ret) { 7445662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 745de6e8200SLiu Bo if (IS_ERR(inode)) 746de6e8200SLiu Bo return PTR_ERR(inode); 7475662344bSTsutomu Itoh d_instantiate(dentry, inode); 7485662344bSTsutomu Itoh } 749f46b5a66SChristoph Hellwig return ret; 75049a3c4d9SDavid Sterba 75149a3c4d9SDavid Sterba fail_free: 75249a3c4d9SDavid Sterba kfree(root_item); 75349a3c4d9SDavid Sterba return ret; 754f46b5a66SChristoph Hellwig } 755f46b5a66SChristoph Hellwig 756e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 75761d7e4cbSDavid Sterba struct dentry *dentry, 758e9662f70SMiao Xie u64 *async_transid, bool readonly, 759e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 760f46b5a66SChristoph Hellwig { 7610b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 7622e4bfab9SYan, Zheng struct inode *inode; 763f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 764f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 7652e4bfab9SYan, Zheng int ret; 766f46b5a66SChristoph Hellwig 76727cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 768f46b5a66SChristoph Hellwig return -EINVAL; 769f46b5a66SChristoph Hellwig 77023269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 771a1ee7362SDavid Sterba if (!pending_snapshot) 772a1ee7362SDavid Sterba return -ENOMEM; 773a1ee7362SDavid Sterba 774b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 77523269bf5SDavid Sterba GFP_KERNEL); 7768546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 7778546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 778b0c0ea63SDavid Sterba ret = -ENOMEM; 779b0c0ea63SDavid Sterba goto free_pending; 780b0c0ea63SDavid Sterba } 781b0c0ea63SDavid Sterba 782ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 7834e857c58SPeter Zijlstra smp_mb__after_atomic(); 78445bac0f3SLiu Bo /* wait for no snapshot writes */ 78545bac0f3SLiu Bo wait_event(root->subv_writers->wait, 78645bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 7878257b2dcSMiao Xie 78876f32e24SNikolay Borisov ret = btrfs_start_delalloc_inodes(root); 7896a03843dSMiao Xie if (ret) 790a1ee7362SDavid Sterba goto dec_and_free; 7916a03843dSMiao Xie 7926374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 7936a03843dSMiao Xie 79466d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 79566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 796d5c12070SMiao Xie /* 797d5c12070SMiao Xie * 1 - parent dir inode 798d5c12070SMiao Xie * 2 - dir entries 799d5c12070SMiao Xie * 1 - root item 800d5c12070SMiao Xie * 2 - root ref/backref 801d5c12070SMiao Xie * 1 - root of snapshot 802dd5f9615SStefan Behrens * 1 - UUID item 803d5c12070SMiao Xie */ 804d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 805dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 806ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 807ee3441b4SJeff Mahoney false); 808d5c12070SMiao Xie if (ret) 809a1ee7362SDavid Sterba goto dec_and_free; 810d5c12070SMiao Xie 811a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 812a22285a6SYan, Zheng pending_snapshot->root = root; 813b83cc969SLi Zefan pending_snapshot->readonly = readonly; 814e9662f70SMiao Xie pending_snapshot->dir = dir; 8158696c533SMiao Xie pending_snapshot->inherit = inherit; 816a22285a6SYan, Zheng 817d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 818a22285a6SYan, Zheng if (IS_ERR(trans)) { 819a22285a6SYan, Zheng ret = PTR_ERR(trans); 820a22285a6SYan, Zheng goto fail; 821a22285a6SYan, Zheng } 822a22285a6SYan, Zheng 8230b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 824a22285a6SYan, Zheng list_add(&pending_snapshot->list, 825a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 8260b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 82772fd032eSSage Weil if (async_transid) { 82872fd032eSSage Weil *async_transid = trans->transid; 8293a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 83000d71c9cSMiao Xie if (ret) 8313a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 83272fd032eSSage Weil } else { 8333a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 83472fd032eSSage Weil } 835aec8030aSMiao Xie if (ret) 836c37b2b62SJosef Bacik goto fail; 837a22285a6SYan, Zheng 838a22285a6SYan, Zheng ret = pending_snapshot->error; 839f46b5a66SChristoph Hellwig if (ret) 8402e4bfab9SYan, Zheng goto fail; 841f46b5a66SChristoph Hellwig 842d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 843d3797308SChris Mason if (ret) 844d3797308SChris Mason goto fail; 845d3797308SChris Mason 8462b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 8472e4bfab9SYan, Zheng if (IS_ERR(inode)) { 8482e4bfab9SYan, Zheng ret = PTR_ERR(inode); 8492e4bfab9SYan, Zheng goto fail; 8502e4bfab9SYan, Zheng } 8515662344bSTsutomu Itoh 8522e4bfab9SYan, Zheng d_instantiate(dentry, inode); 8532e4bfab9SYan, Zheng ret = 0; 8542e4bfab9SYan, Zheng fail: 8557775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 856a1ee7362SDavid Sterba dec_and_free: 857ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 8584625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 859b0c0ea63SDavid Sterba free_pending: 860b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 8618546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 862a1ee7362SDavid Sterba kfree(pending_snapshot); 863a1ee7362SDavid Sterba 864f46b5a66SChristoph Hellwig return ret; 865f46b5a66SChristoph Hellwig } 866f46b5a66SChristoph Hellwig 8674260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 8684260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 8694260f7c7SSage Weil * whether the type of victim is right. 8704260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 8714260f7c7SSage Weil * 2. We should have write and exec permissions on dir 8724260f7c7SSage Weil * 3. We can't remove anything from append-only dir 8734260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 8744260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 8754260f7c7SSage Weil * a. be owner of dir, or 8764260f7c7SSage Weil * b. be owner of victim, or 8774260f7c7SSage Weil * c. have CAP_FOWNER capability 87801327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 8794260f7c7SSage Weil * links pointing to it. 8804260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 8814260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 8824260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 8834260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8844260f7c7SSage Weil * nfs_async_unlink(). 8854260f7c7SSage Weil */ 8864260f7c7SSage Weil 8874260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 8884260f7c7SSage Weil { 8894260f7c7SSage Weil int error; 8904260f7c7SSage Weil 8912b0143b5SDavid Howells if (d_really_is_negative(victim)) 8924260f7c7SSage Weil return -ENOENT; 8934260f7c7SSage Weil 8942b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8954fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8964260f7c7SSage Weil 8974260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 8984260f7c7SSage Weil if (error) 8994260f7c7SSage Weil return error; 9004260f7c7SSage Weil if (IS_APPEND(dir)) 9014260f7c7SSage Weil return -EPERM; 9022b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 9032b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 9044260f7c7SSage Weil return -EPERM; 9054260f7c7SSage Weil if (isdir) { 906e36cb0b8SDavid Howells if (!d_is_dir(victim)) 9074260f7c7SSage Weil return -ENOTDIR; 9084260f7c7SSage Weil if (IS_ROOT(victim)) 9094260f7c7SSage Weil return -EBUSY; 910e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 9114260f7c7SSage Weil return -EISDIR; 9124260f7c7SSage Weil if (IS_DEADDIR(dir)) 9134260f7c7SSage Weil return -ENOENT; 9144260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 9154260f7c7SSage Weil return -EBUSY; 9164260f7c7SSage Weil return 0; 9174260f7c7SSage Weil } 9184260f7c7SSage Weil 919cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 920cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 921cb8e7090SChristoph Hellwig { 9222b0143b5SDavid Howells if (d_really_is_positive(child)) 923cb8e7090SChristoph Hellwig return -EEXIST; 924cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 925cb8e7090SChristoph Hellwig return -ENOENT; 926cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 927cb8e7090SChristoph Hellwig } 928cb8e7090SChristoph Hellwig 929cb8e7090SChristoph Hellwig /* 930cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 931cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 932cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 933cb8e7090SChristoph Hellwig */ 93492872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 93552f75f4fSDavid Sterba const char *name, int namelen, 93672fd032eSSage Weil struct btrfs_root *snap_src, 9376f72c7e2SArne Jansen u64 *async_transid, bool readonly, 9388696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 939cb8e7090SChristoph Hellwig { 9402b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 9410b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 942cb8e7090SChristoph Hellwig struct dentry *dentry; 943cb8e7090SChristoph Hellwig int error; 944cb8e7090SChristoph Hellwig 94500235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 94600235411SAl Viro if (error == -EINTR) 94700235411SAl Viro return error; 948cb8e7090SChristoph Hellwig 949cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 950cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 951cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 952cb8e7090SChristoph Hellwig goto out_unlock; 953cb8e7090SChristoph Hellwig 95476dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 955cb8e7090SChristoph Hellwig if (error) 956a874a63eSLiu Bo goto out_dput; 957cb8e7090SChristoph Hellwig 9589c52057cSChris Mason /* 9599c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 9609c52057cSChris Mason * check for them now when we can safely fail 9619c52057cSChris Mason */ 9629c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 9639c52057cSChris Mason dir->i_ino, name, 9649c52057cSChris Mason namelen); 9659c52057cSChris Mason if (error) 9669c52057cSChris Mason goto out_dput; 9679c52057cSChris Mason 9680b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 96976dda93cSYan, Zheng 97076dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 97176dda93cSYan, Zheng goto out_up_read; 97276dda93cSYan, Zheng 9733de4586cSChris Mason if (snap_src) { 97461d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 9756f72c7e2SArne Jansen async_transid, readonly, inherit); 9763de4586cSChris Mason } else { 977d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 978d5c12070SMiao Xie async_transid, inherit); 9793de4586cSChris Mason } 98076dda93cSYan, Zheng if (!error) 98176dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 98276dda93cSYan, Zheng out_up_read: 9830b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 984cb8e7090SChristoph Hellwig out_dput: 985cb8e7090SChristoph Hellwig dput(dentry); 986cb8e7090SChristoph Hellwig out_unlock: 9875955102cSAl Viro inode_unlock(dir); 988cb8e7090SChristoph Hellwig return error; 989cb8e7090SChristoph Hellwig } 990cb8e7090SChristoph Hellwig 9914cb5300bSChris Mason /* 9924cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 9934cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 9944cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 9954cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 9964cb5300bSChris Mason * part of the file 9974cb5300bSChris Mason */ 998aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 9994cb5300bSChris Mason { 10004cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10014cb5300bSChris Mason struct extent_map *em = NULL; 10024cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 10034cb5300bSChris Mason u64 end; 10044cb5300bSChris Mason 10054cb5300bSChris Mason read_lock(&em_tree->lock); 100609cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 10074cb5300bSChris Mason read_unlock(&em_tree->lock); 10084cb5300bSChris Mason 10094cb5300bSChris Mason if (em) { 10104cb5300bSChris Mason end = extent_map_end(em); 10114cb5300bSChris Mason free_extent_map(em); 10124cb5300bSChris Mason if (end - offset > thresh) 10134cb5300bSChris Mason return 0; 10144cb5300bSChris Mason } 10154cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 10164cb5300bSChris Mason thresh /= 2; 10174cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 10184cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 10194cb5300bSChris Mason if (end >= thresh) 10204cb5300bSChris Mason return 0; 10214cb5300bSChris Mason return 1; 10224cb5300bSChris Mason } 10234cb5300bSChris Mason 10244cb5300bSChris Mason /* 10254cb5300bSChris Mason * helper function to walk through a file and find extents 10264cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 10274cb5300bSChris Mason * 10284cb5300bSChris Mason * This is used by the defragging code to find new and small 10294cb5300bSChris Mason * extents 10304cb5300bSChris Mason */ 10314cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 10324cb5300bSChris Mason struct inode *inode, u64 newer_than, 1033aab110abSDavid Sterba u64 *off, u32 thresh) 10344cb5300bSChris Mason { 10354cb5300bSChris Mason struct btrfs_path *path; 10364cb5300bSChris Mason struct btrfs_key min_key; 10374cb5300bSChris Mason struct extent_buffer *leaf; 10384cb5300bSChris Mason struct btrfs_file_extent_item *extent; 10394cb5300bSChris Mason int type; 10404cb5300bSChris Mason int ret; 10414a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 10424cb5300bSChris Mason 10434cb5300bSChris Mason path = btrfs_alloc_path(); 10444cb5300bSChris Mason if (!path) 10454cb5300bSChris Mason return -ENOMEM; 10464cb5300bSChris Mason 1047a4689d2bSDavid Sterba min_key.objectid = ino; 10484cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 10494cb5300bSChris Mason min_key.offset = *off; 10504cb5300bSChris Mason 10514cb5300bSChris Mason while (1) { 10526174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 10534cb5300bSChris Mason if (ret != 0) 10544cb5300bSChris Mason goto none; 1055f094c9bdSFilipe Manana process_slot: 1056a4689d2bSDavid Sterba if (min_key.objectid != ino) 10574cb5300bSChris Mason goto none; 10584cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 10594cb5300bSChris Mason goto none; 10604cb5300bSChris Mason 10614cb5300bSChris Mason leaf = path->nodes[0]; 10624cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 10634cb5300bSChris Mason struct btrfs_file_extent_item); 10644cb5300bSChris Mason 10654cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 10664cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 10674cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 10684cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 10694cb5300bSChris Mason *off = min_key.offset; 10704cb5300bSChris Mason btrfs_free_path(path); 10714cb5300bSChris Mason return 0; 10724cb5300bSChris Mason } 10734cb5300bSChris Mason 1074f094c9bdSFilipe Manana path->slots[0]++; 1075f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 1076f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1077f094c9bdSFilipe Manana goto process_slot; 1078f094c9bdSFilipe Manana } 1079f094c9bdSFilipe Manana 10804cb5300bSChris Mason if (min_key.offset == (u64)-1) 10814cb5300bSChris Mason goto none; 10824cb5300bSChris Mason 10834cb5300bSChris Mason min_key.offset++; 10844cb5300bSChris Mason btrfs_release_path(path); 10854cb5300bSChris Mason } 10864cb5300bSChris Mason none: 10874cb5300bSChris Mason btrfs_free_path(path); 10884cb5300bSChris Mason return -ENOENT; 10894cb5300bSChris Mason } 10904cb5300bSChris Mason 10916c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 109217ce6ef8SLiu Bo { 109317ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1094940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10956c282eb4SLi Zefan struct extent_map *em; 109609cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1097940100a4SChris Mason 1098940100a4SChris Mason /* 1099940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1100940100a4SChris Mason * the full extent lock 1101940100a4SChris Mason */ 1102940100a4SChris Mason read_lock(&em_tree->lock); 1103940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1104940100a4SChris Mason read_unlock(&em_tree->lock); 1105940100a4SChris Mason 1106940100a4SChris Mason if (!em) { 1107308d9800SFilipe Manana struct extent_state *cached = NULL; 1108308d9800SFilipe Manana u64 end = start + len - 1; 1109308d9800SFilipe Manana 1110940100a4SChris Mason /* get the big lock and read metadata off disk */ 1111ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1112fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1113e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 1114940100a4SChris Mason 11156cf8bfbfSDan Carpenter if (IS_ERR(em)) 11166c282eb4SLi Zefan return NULL; 1117940100a4SChris Mason } 1118940100a4SChris Mason 11196c282eb4SLi Zefan return em; 11206c282eb4SLi Zefan } 11216c282eb4SLi Zefan 11226c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 11236c282eb4SLi Zefan { 11246c282eb4SLi Zefan struct extent_map *next; 11256c282eb4SLi Zefan bool ret = true; 11266c282eb4SLi Zefan 11276c282eb4SLi Zefan /* this is the last extent */ 11286c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 11296c282eb4SLi Zefan return false; 11306c282eb4SLi Zefan 11316c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1132e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1133e9512d72SChris Mason ret = false; 1134e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1135ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 11366c282eb4SLi Zefan ret = false; 11376c282eb4SLi Zefan 11386c282eb4SLi Zefan free_extent_map(next); 11396c282eb4SLi Zefan return ret; 11406c282eb4SLi Zefan } 11416c282eb4SLi Zefan 1142aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1143a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1144a43a2111SAndrew Mahone int compress) 11456c282eb4SLi Zefan { 11466c282eb4SLi Zefan struct extent_map *em; 11476c282eb4SLi Zefan int ret = 1; 11486c282eb4SLi Zefan bool next_mergeable = true; 11494a3560c4SLiu Bo bool prev_mergeable = true; 11506c282eb4SLi Zefan 11516c282eb4SLi Zefan /* 11526c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 11536c282eb4SLi Zefan * defragging it 11546c282eb4SLi Zefan */ 11556c282eb4SLi Zefan if (start < *defrag_end) 11566c282eb4SLi Zefan return 1; 11576c282eb4SLi Zefan 11586c282eb4SLi Zefan *skip = 0; 11596c282eb4SLi Zefan 11606c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 11616c282eb4SLi Zefan if (!em) 11626c282eb4SLi Zefan return 0; 11636c282eb4SLi Zefan 1164940100a4SChris Mason /* this will cover holes, and inline extents */ 116517ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1166940100a4SChris Mason ret = 0; 116717ce6ef8SLiu Bo goto out; 116817ce6ef8SLiu Bo } 116917ce6ef8SLiu Bo 11704a3560c4SLiu Bo if (!*defrag_end) 11714a3560c4SLiu Bo prev_mergeable = false; 11724a3560c4SLiu Bo 11736c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1174940100a4SChris Mason /* 11756c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 11766c282eb4SLi Zefan * real extent, don't bother defragging it 1177940100a4SChris Mason */ 1178a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 11794a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1180940100a4SChris Mason ret = 0; 118117ce6ef8SLiu Bo out: 1182940100a4SChris Mason /* 1183940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1184940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1185940100a4SChris Mason * so that the next tiny extent will force a defrag. 1186940100a4SChris Mason * 1187940100a4SChris Mason * The end result of this is that tiny extents before a single big 1188940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1189940100a4SChris Mason */ 1190940100a4SChris Mason if (ret) { 1191940100a4SChris Mason *defrag_end = extent_map_end(em); 1192940100a4SChris Mason } else { 1193940100a4SChris Mason *last_len = 0; 1194940100a4SChris Mason *skip = extent_map_end(em); 1195940100a4SChris Mason *defrag_end = 0; 1196940100a4SChris Mason } 1197940100a4SChris Mason 1198940100a4SChris Mason free_extent_map(em); 1199940100a4SChris Mason return ret; 1200940100a4SChris Mason } 1201940100a4SChris Mason 12024cb5300bSChris Mason /* 12034cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 12044cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 12054cb5300bSChris Mason * to COW and defrag. 12064cb5300bSChris Mason * 12074cb5300bSChris Mason * It also makes sure the delalloc code has enough 12084cb5300bSChris Mason * dirty data to avoid making new small extents as part 12094cb5300bSChris Mason * of the defrag 12104cb5300bSChris Mason * 12114cb5300bSChris Mason * It's a good idea to start RA on this range 12124cb5300bSChris Mason * before calling this. 12134cb5300bSChris Mason */ 12144cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 12154cb5300bSChris Mason struct page **pages, 12164cb5300bSChris Mason unsigned long start_index, 1217c41570c9SJustin Maggard unsigned long num_pages) 1218f46b5a66SChristoph Hellwig { 12194cb5300bSChris Mason unsigned long file_end; 12204cb5300bSChris Mason u64 isize = i_size_read(inode); 1221f46b5a66SChristoph Hellwig u64 page_start; 1222f46b5a66SChristoph Hellwig u64 page_end; 12231f12bd06SLiu Bo u64 page_cnt; 12244cb5300bSChris Mason int ret; 12254cb5300bSChris Mason int i; 12264cb5300bSChris Mason int i_done; 12274cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 12284cb5300bSChris Mason struct extent_state *cached_state = NULL; 1229600a45e1SMiao Xie struct extent_io_tree *tree; 1230364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 12313b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 12324cb5300bSChris Mason 123309cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 12341f12bd06SLiu Bo if (!isize || start_index > file_end) 12351f12bd06SLiu Bo return 0; 12361f12bd06SLiu Bo 12371f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 12384cb5300bSChris Mason 1239364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 124009cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 124109cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12424cb5300bSChris Mason if (ret) 12434cb5300bSChris Mason return ret; 12444cb5300bSChris Mason i_done = 0; 1245600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 12464cb5300bSChris Mason 12474cb5300bSChris Mason /* step one, lock all the pages */ 12481f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 12494cb5300bSChris Mason struct page *page; 1250600a45e1SMiao Xie again: 1251a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 12523b16a4e3SJosef Bacik start_index + i, mask); 12534cb5300bSChris Mason if (!page) 12544cb5300bSChris Mason break; 12554cb5300bSChris Mason 1256600a45e1SMiao Xie page_start = page_offset(page); 125709cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1258600a45e1SMiao Xie while (1) { 1259308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1260ff13db41SDavid Sterba &cached_state); 1261600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1262600a45e1SMiao Xie page_start); 1263308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1264e43bbe5eSDavid Sterba &cached_state); 1265600a45e1SMiao Xie if (!ordered) 1266600a45e1SMiao Xie break; 1267600a45e1SMiao Xie 1268600a45e1SMiao Xie unlock_page(page); 1269600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1270600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1271600a45e1SMiao Xie lock_page(page); 12721f12bd06SLiu Bo /* 12731f12bd06SLiu Bo * we unlocked the page above, so we need check if 12741f12bd06SLiu Bo * it was released or not. 12751f12bd06SLiu Bo */ 12761f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 12771f12bd06SLiu Bo unlock_page(page); 127809cbfeafSKirill A. Shutemov put_page(page); 12791f12bd06SLiu Bo goto again; 12801f12bd06SLiu Bo } 1281600a45e1SMiao Xie } 1282600a45e1SMiao Xie 12834cb5300bSChris Mason if (!PageUptodate(page)) { 12844cb5300bSChris Mason btrfs_readpage(NULL, page); 12854cb5300bSChris Mason lock_page(page); 12864cb5300bSChris Mason if (!PageUptodate(page)) { 12874cb5300bSChris Mason unlock_page(page); 128809cbfeafSKirill A. Shutemov put_page(page); 12894cb5300bSChris Mason ret = -EIO; 12904cb5300bSChris Mason break; 12914cb5300bSChris Mason } 12924cb5300bSChris Mason } 1293600a45e1SMiao Xie 1294600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1295600a45e1SMiao Xie unlock_page(page); 129609cbfeafSKirill A. Shutemov put_page(page); 1297600a45e1SMiao Xie goto again; 1298600a45e1SMiao Xie } 1299600a45e1SMiao Xie 13004cb5300bSChris Mason pages[i] = page; 13014cb5300bSChris Mason i_done++; 13024cb5300bSChris Mason } 13034cb5300bSChris Mason if (!i_done || ret) 13044cb5300bSChris Mason goto out; 13054cb5300bSChris Mason 13061751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13074cb5300bSChris Mason goto out; 13084cb5300bSChris Mason 13094cb5300bSChris Mason /* 13104cb5300bSChris Mason * so now we have a nice long stream of locked 13114cb5300bSChris Mason * and up to date pages, lets wait on them 13124cb5300bSChris Mason */ 13134cb5300bSChris Mason for (i = 0; i < i_done; i++) 13144cb5300bSChris Mason wait_on_page_writeback(pages[i]); 13154cb5300bSChris Mason 13164cb5300bSChris Mason page_start = page_offset(pages[0]); 131709cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 13184cb5300bSChris Mason 13194cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1320ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 13214cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 13224cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 13239e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1324ae0f1625SDavid Sterba &cached_state); 13254cb5300bSChris Mason 13261f12bd06SLiu Bo if (i_done != page_cnt) { 13279e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 13289e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 13299e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1330bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 133109cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 133243b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 13334cb5300bSChris Mason } 13344cb5300bSChris Mason 13354cb5300bSChris Mason 13369e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1337018ed4f7SDavid Sterba &cached_state); 13384cb5300bSChris Mason 13394cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1340e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 13414cb5300bSChris Mason 13424cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13434cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 13444cb5300bSChris Mason ClearPageChecked(pages[i]); 13454cb5300bSChris Mason set_page_extent_mapped(pages[i]); 13464cb5300bSChris Mason set_page_dirty(pages[i]); 13474cb5300bSChris Mason unlock_page(pages[i]); 134809cbfeafSKirill A. Shutemov put_page(pages[i]); 13494cb5300bSChris Mason } 135043b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 135143b18595SQu Wenruo false); 1352364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13534cb5300bSChris Mason return i_done; 13544cb5300bSChris Mason out: 13554cb5300bSChris Mason for (i = 0; i < i_done; i++) { 13564cb5300bSChris Mason unlock_page(pages[i]); 135709cbfeafSKirill A. Shutemov put_page(pages[i]); 13584cb5300bSChris Mason } 1359bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 136009cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 136143b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 136243b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 136343b18595SQu Wenruo true); 1364364ecf36SQu Wenruo extent_changeset_free(data_reserved); 13654cb5300bSChris Mason return ret; 13664cb5300bSChris Mason 13674cb5300bSChris Mason } 13684cb5300bSChris Mason 13694cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 13704cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 13714cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 13724cb5300bSChris Mason { 13730b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 13744cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 13754cb5300bSChris Mason struct file_ra_state *ra = NULL; 13764cb5300bSChris Mason unsigned long last_index; 1377151a31b2SLi Zefan u64 isize = i_size_read(inode); 1378940100a4SChris Mason u64 last_len = 0; 1379940100a4SChris Mason u64 skip = 0; 1380940100a4SChris Mason u64 defrag_end = 0; 13814cb5300bSChris Mason u64 newer_off = range->start; 1382f46b5a66SChristoph Hellwig unsigned long i; 1383008873eaSLi Zefan unsigned long ra_index = 0; 1384f46b5a66SChristoph Hellwig int ret; 13854cb5300bSChris Mason int defrag_count = 0; 13861a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1387aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 138809cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1389c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1390ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 13914cb5300bSChris Mason struct page **pages = NULL; 13921e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 13934cb5300bSChris Mason 13940abd5b17SLiu Bo if (isize == 0) 13950abd5b17SLiu Bo return 0; 13960abd5b17SLiu Bo 13970abd5b17SLiu Bo if (range->start >= isize) 13980abd5b17SLiu Bo return -EINVAL; 13991a419d85SLi Zefan 14001e2ef46dSDavid Sterba if (do_compress) { 14011a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 14021a419d85SLi Zefan return -EINVAL; 14031a419d85SLi Zefan if (range->compress_type) 14041a419d85SLi Zefan compress_type = range->compress_type; 14051a419d85SLi Zefan } 1406f46b5a66SChristoph Hellwig 14070abd5b17SLiu Bo if (extent_thresh == 0) 1408ee22184bSByongho Lee extent_thresh = SZ_256K; 1409f46b5a66SChristoph Hellwig 14104cb5300bSChris Mason /* 14110a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 14120a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 14130a52d108SDavid Sterba * we don't error out. 14144cb5300bSChris Mason */ 14154cb5300bSChris Mason if (!file) { 141663e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 14170a52d108SDavid Sterba if (ra) 14184cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 14194cb5300bSChris Mason } else { 14204cb5300bSChris Mason ra = &file->f_ra; 14214cb5300bSChris Mason } 14224cb5300bSChris Mason 142363e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 14244cb5300bSChris Mason if (!pages) { 14254cb5300bSChris Mason ret = -ENOMEM; 14264cb5300bSChris Mason goto out_ra; 14274cb5300bSChris Mason } 14284cb5300bSChris Mason 14294cb5300bSChris Mason /* find the last page to defrag */ 14301e701a32SChris Mason if (range->start + range->len > range->start) { 1431151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 143209cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 14331e701a32SChris Mason } else { 143409cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 14351e701a32SChris Mason } 14361e701a32SChris Mason 14374cb5300bSChris Mason if (newer_than) { 14384cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1439ee22184bSByongho Lee &newer_off, SZ_64K); 14404cb5300bSChris Mason if (!ret) { 14414cb5300bSChris Mason range->start = newer_off; 14424cb5300bSChris Mason /* 14434cb5300bSChris Mason * we always align our defrag to help keep 14444cb5300bSChris Mason * the extents in the file evenly spaced 14454cb5300bSChris Mason */ 144609cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14474cb5300bSChris Mason } else 14484cb5300bSChris Mason goto out_ra; 14494cb5300bSChris Mason } else { 145009cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 14514cb5300bSChris Mason } 14524cb5300bSChris Mason if (!max_to_defrag) 1453070034bdSchandan max_to_defrag = last_index - i + 1; 14544cb5300bSChris Mason 14552a0f7f57SLi Zefan /* 14562a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 14572a0f7f57SLi Zefan * written sequentially. 14582a0f7f57SLi Zefan */ 14592a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 14602a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 14612a0f7f57SLi Zefan 1462f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 146309cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 14644cb5300bSChris Mason /* 14654cb5300bSChris Mason * make sure we stop running if someone unmounts 14664cb5300bSChris Mason * the FS 14674cb5300bSChris Mason */ 14681751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 14694cb5300bSChris Mason break; 14704cb5300bSChris Mason 14710b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 14720b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1473210549ebSDavid Sterba ret = -EAGAIN; 1474210549ebSDavid Sterba break; 1475210549ebSDavid Sterba } 1476210549ebSDavid Sterba 147709cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 14786c282eb4SLi Zefan extent_thresh, &last_len, &skip, 14791e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1480940100a4SChris Mason unsigned long next; 1481940100a4SChris Mason /* 1482940100a4SChris Mason * the should_defrag function tells us how much to skip 1483940100a4SChris Mason * bump our counter by the suggested amount 1484940100a4SChris Mason */ 148509cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1486940100a4SChris Mason i = max(i + 1, next); 1487940100a4SChris Mason continue; 1488940100a4SChris Mason } 1489008873eaSLi Zefan 1490008873eaSLi Zefan if (!newer_than) { 149109cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 149209cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1493008873eaSLi Zefan cluster = min(cluster, max_cluster); 1494008873eaSLi Zefan } else { 1495008873eaSLi Zefan cluster = max_cluster; 1496008873eaSLi Zefan } 1497008873eaSLi Zefan 1498008873eaSLi Zefan if (i + cluster > ra_index) { 1499008873eaSLi Zefan ra_index = max(i, ra_index); 15000a52d108SDavid Sterba if (ra) 1501d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1502d3c0bab5SDavid Sterba file, ra_index, cluster); 1503e4826a5bSchandan ra_index += cluster; 1504008873eaSLi Zefan } 15054cb5300bSChris Mason 15065955102cSAl Viro inode_lock(inode); 15071e2ef46dSDavid Sterba if (do_compress) 1508eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1509008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1510ecb8bea8SLiu Bo if (ret < 0) { 15115955102cSAl Viro inode_unlock(inode); 15124cb5300bSChris Mason goto out_ra; 1513ecb8bea8SLiu Bo } 15144cb5300bSChris Mason 15154cb5300bSChris Mason defrag_count += ret; 1516d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 15175955102cSAl Viro inode_unlock(inode); 15184cb5300bSChris Mason 15194cb5300bSChris Mason if (newer_than) { 15204cb5300bSChris Mason if (newer_off == (u64)-1) 15214cb5300bSChris Mason break; 15224cb5300bSChris Mason 1523e1f041e1SLiu Bo if (ret > 0) 1524e1f041e1SLiu Bo i += ret; 1525e1f041e1SLiu Bo 15264cb5300bSChris Mason newer_off = max(newer_off + 1, 152709cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 15284cb5300bSChris Mason 1529ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1530ee22184bSByongho Lee &newer_off, SZ_64K); 15314cb5300bSChris Mason if (!ret) { 15324cb5300bSChris Mason range->start = newer_off; 153309cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 15344cb5300bSChris Mason } else { 15354cb5300bSChris Mason break; 1536940100a4SChris Mason } 15374cb5300bSChris Mason } else { 1538008873eaSLi Zefan if (ret > 0) { 1539cbcc8326SLi Zefan i += ret; 154009cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1541008873eaSLi Zefan } else { 1542940100a4SChris Mason i++; 1543008873eaSLi Zefan last_len = 0; 1544008873eaSLi Zefan } 1545f46b5a66SChristoph Hellwig } 15464cb5300bSChris Mason } 1547f46b5a66SChristoph Hellwig 1548dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 15491e701a32SChris Mason filemap_flush(inode->i_mapping); 1550dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1551dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1552dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1553dec8ef90SFilipe Manana } 15541e701a32SChris Mason 15551a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 15560b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 15575c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 15585c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 15591a419d85SLi Zefan } 15601a419d85SLi Zefan 156160ccf82fSDiego Calleja ret = defrag_count; 1562940100a4SChris Mason 15634cb5300bSChris Mason out_ra: 15641e2ef46dSDavid Sterba if (do_compress) { 15655955102cSAl Viro inode_lock(inode); 1566eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 15675955102cSAl Viro inode_unlock(inode); 1568633085c7SFilipe David Borba Manana } 15694cb5300bSChris Mason if (!file) 15704cb5300bSChris Mason kfree(ra); 15714cb5300bSChris Mason kfree(pages); 1572940100a4SChris Mason return ret; 1573f46b5a66SChristoph Hellwig } 1574f46b5a66SChristoph Hellwig 1575198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 157676dda93cSYan, Zheng void __user *arg) 1577f46b5a66SChristoph Hellwig { 15780b246afaSJeff Mahoney struct inode *inode = file_inode(file); 15790b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1580f46b5a66SChristoph Hellwig u64 new_size; 1581f46b5a66SChristoph Hellwig u64 old_size; 1582f46b5a66SChristoph Hellwig u64 devid = 1; 15830b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1584f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1585f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1586f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1587f46b5a66SChristoph Hellwig char *sizestr; 15889a40f122SGui Hecheng char *retptr; 1589f46b5a66SChristoph Hellwig char *devstr = NULL; 1590f46b5a66SChristoph Hellwig int ret = 0; 1591f46b5a66SChristoph Hellwig int mod = 0; 1592f46b5a66SChristoph Hellwig 1593e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1594e441d54dSChris Mason return -EPERM; 1595e441d54dSChris Mason 1596198605a8SMiao Xie ret = mnt_want_write_file(file); 1597198605a8SMiao Xie if (ret) 1598198605a8SMiao Xie return ret; 1599198605a8SMiao Xie 1600171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 160197547676SMiao Xie mnt_drop_write_file(file); 1602e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1603c9e9f97bSIlya Dryomov } 1604c9e9f97bSIlya Dryomov 1605dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1606c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1607c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1608c9e9f97bSIlya Dryomov goto out; 1609c9e9f97bSIlya Dryomov } 16105516e595SMark Fasheh 16115516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1612f46b5a66SChristoph Hellwig 1613f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1614f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1615f46b5a66SChristoph Hellwig if (devstr) { 1616f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1617f46b5a66SChristoph Hellwig *devstr = '\0'; 1618f46b5a66SChristoph Hellwig devstr = vol_args->name; 161958dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 162058dfae63SZhangZhen if (ret) 162158dfae63SZhangZhen goto out_free; 1622dfd79829SMiao Xie if (!devid) { 1623dfd79829SMiao Xie ret = -EINVAL; 1624dfd79829SMiao Xie goto out_free; 1625dfd79829SMiao Xie } 16260b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1627f46b5a66SChristoph Hellwig } 1628dba60f3fSMiao Xie 16290b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1630f46b5a66SChristoph Hellwig if (!device) { 16310b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1632c1c9ff7cSGeert Uytterhoeven devid); 1633dfd79829SMiao Xie ret = -ENODEV; 1634c9e9f97bSIlya Dryomov goto out_free; 1635f46b5a66SChristoph Hellwig } 1636dba60f3fSMiao Xie 1637ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 16380b246afaSJeff Mahoney btrfs_info(fs_info, 1639efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1640c1c9ff7cSGeert Uytterhoeven devid); 1641dfd79829SMiao Xie ret = -EPERM; 16424e42ae1bSLiu Bo goto out_free; 16434e42ae1bSLiu Bo } 16444e42ae1bSLiu Bo 1645f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1646f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1647f46b5a66SChristoph Hellwig else { 1648f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1649f46b5a66SChristoph Hellwig mod = -1; 1650f46b5a66SChristoph Hellwig sizestr++; 1651f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1652f46b5a66SChristoph Hellwig mod = 1; 1653f46b5a66SChristoph Hellwig sizestr++; 1654f46b5a66SChristoph Hellwig } 16559a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 16569a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1657f46b5a66SChristoph Hellwig ret = -EINVAL; 1658c9e9f97bSIlya Dryomov goto out_free; 1659f46b5a66SChristoph Hellwig } 1660f46b5a66SChristoph Hellwig } 1661f46b5a66SChristoph Hellwig 1662401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1663dfd79829SMiao Xie ret = -EPERM; 166463a212abSStefan Behrens goto out_free; 166563a212abSStefan Behrens } 166663a212abSStefan Behrens 16677cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1668f46b5a66SChristoph Hellwig 1669f46b5a66SChristoph Hellwig if (mod < 0) { 1670f46b5a66SChristoph Hellwig if (new_size > old_size) { 1671f46b5a66SChristoph Hellwig ret = -EINVAL; 1672c9e9f97bSIlya Dryomov goto out_free; 1673f46b5a66SChristoph Hellwig } 1674f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1675f46b5a66SChristoph Hellwig } else if (mod > 0) { 1676eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1677902c68a4SGui Hecheng ret = -ERANGE; 1678eb8052e0SWenliang Fan goto out_free; 1679eb8052e0SWenliang Fan } 1680f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1681f46b5a66SChristoph Hellwig } 1682f46b5a66SChristoph Hellwig 1683ee22184bSByongho Lee if (new_size < SZ_256M) { 1684f46b5a66SChristoph Hellwig ret = -EINVAL; 1685c9e9f97bSIlya Dryomov goto out_free; 1686f46b5a66SChristoph Hellwig } 1687f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1688f46b5a66SChristoph Hellwig ret = -EFBIG; 1689c9e9f97bSIlya Dryomov goto out_free; 1690f46b5a66SChristoph Hellwig } 1691f46b5a66SChristoph Hellwig 169247f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1693f46b5a66SChristoph Hellwig 16940b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1695c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1696f46b5a66SChristoph Hellwig 1697f46b5a66SChristoph Hellwig if (new_size > old_size) { 1698a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 169998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 170098d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1701c9e9f97bSIlya Dryomov goto out_free; 170298d5dc13STsutomu Itoh } 1703f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 17043a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1705ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1706f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 17070253f40eSjeff.liu } /* equal, nothing need to do */ 1708f46b5a66SChristoph Hellwig 1709c9e9f97bSIlya Dryomov out_free: 1710f46b5a66SChristoph Hellwig kfree(vol_args); 1711c9e9f97bSIlya Dryomov out: 1712171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 171318f39c41SIlya Dryomov mnt_drop_write_file(file); 1714f46b5a66SChristoph Hellwig return ret; 1715f46b5a66SChristoph Hellwig } 1716f46b5a66SChristoph Hellwig 171772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 171852f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 17196f72c7e2SArne Jansen u64 *transid, bool readonly, 17208696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1721f46b5a66SChristoph Hellwig { 1722f46b5a66SChristoph Hellwig int namelen; 17233de4586cSChris Mason int ret = 0; 1724f46b5a66SChristoph Hellwig 1725325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1726325c50e3SJeff Mahoney return -ENOTDIR; 1727325c50e3SJeff Mahoney 1728a874a63eSLiu Bo ret = mnt_want_write_file(file); 1729a874a63eSLiu Bo if (ret) 1730a874a63eSLiu Bo goto out; 1731a874a63eSLiu Bo 173272fd032eSSage Weil namelen = strlen(name); 173372fd032eSSage Weil if (strchr(name, '/')) { 1734f46b5a66SChristoph Hellwig ret = -EINVAL; 1735a874a63eSLiu Bo goto out_drop_write; 1736f46b5a66SChristoph Hellwig } 1737f46b5a66SChristoph Hellwig 173816780cabSChris Mason if (name[0] == '.' && 173916780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 174016780cabSChris Mason ret = -EEXIST; 1741a874a63eSLiu Bo goto out_drop_write; 174216780cabSChris Mason } 174316780cabSChris Mason 17443de4586cSChris Mason if (subvol) { 174572fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 17466f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1747cb8e7090SChristoph Hellwig } else { 17482903ff01SAl Viro struct fd src = fdget(fd); 17493de4586cSChris Mason struct inode *src_inode; 17502903ff01SAl Viro if (!src.file) { 17513de4586cSChris Mason ret = -EINVAL; 1752a874a63eSLiu Bo goto out_drop_write; 17533de4586cSChris Mason } 17543de4586cSChris Mason 1755496ad9aaSAl Viro src_inode = file_inode(src.file); 1756496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1757c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1758efe120a0SFrank Holton "Snapshot src from another FS"); 175923ad5b17SKusanagi Kouichi ret = -EXDEV; 1760d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1761d0242061SDavid Sterba /* 1762d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1763d0242061SDavid Sterba * are limited to own subvolumes only 1764d0242061SDavid Sterba */ 1765d0242061SDavid Sterba ret = -EPERM; 1766ecd18815SAl Viro } else { 176772fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 176872fd032eSSage Weil BTRFS_I(src_inode)->root, 17696f72c7e2SArne Jansen transid, readonly, inherit); 1770ecd18815SAl Viro } 17712903ff01SAl Viro fdput(src); 1772cb8e7090SChristoph Hellwig } 1773a874a63eSLiu Bo out_drop_write: 1774a874a63eSLiu Bo mnt_drop_write_file(file); 1775f46b5a66SChristoph Hellwig out: 177672fd032eSSage Weil return ret; 177772fd032eSSage Weil } 177872fd032eSSage Weil 177972fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1780fa0d2b9bSLi Zefan void __user *arg, int subvol) 178172fd032eSSage Weil { 1782fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 178372fd032eSSage Weil int ret; 178472fd032eSSage Weil 1785325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1786325c50e3SJeff Mahoney return -ENOTDIR; 1787325c50e3SJeff Mahoney 1788fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1789fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1790fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1791fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1792fa0d2b9bSLi Zefan 1793fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1794b83cc969SLi Zefan vol_args->fd, subvol, 17956f72c7e2SArne Jansen NULL, false, NULL); 1796fa0d2b9bSLi Zefan 1797fa0d2b9bSLi Zefan kfree(vol_args); 1798fa0d2b9bSLi Zefan return ret; 1799fa0d2b9bSLi Zefan } 1800fa0d2b9bSLi Zefan 1801fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1802fa0d2b9bSLi Zefan void __user *arg, int subvol) 1803fa0d2b9bSLi Zefan { 1804fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1805fa0d2b9bSLi Zefan int ret; 1806fdfb1e4fSLi Zefan u64 transid = 0; 1807fdfb1e4fSLi Zefan u64 *ptr = NULL; 1808b83cc969SLi Zefan bool readonly = false; 18096f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 181072fd032eSSage Weil 1811325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1812325c50e3SJeff Mahoney return -ENOTDIR; 1813325c50e3SJeff Mahoney 1814fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1815fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1816fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1817fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1818fdfb1e4fSLi Zefan 1819b83cc969SLi Zefan if (vol_args->flags & 18206f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 18216f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1822b83cc969SLi Zefan ret = -EOPNOTSUPP; 1823c47ca32dSDan Carpenter goto free_args; 1824fdfb1e4fSLi Zefan } 1825fdfb1e4fSLi Zefan 1826fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1827fdfb1e4fSLi Zefan ptr = &transid; 1828b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1829b83cc969SLi Zefan readonly = true; 18306f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 183109cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 18326f72c7e2SArne Jansen ret = -EINVAL; 1833c47ca32dSDan Carpenter goto free_args; 18346f72c7e2SArne Jansen } 18356f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 18366f72c7e2SArne Jansen if (IS_ERR(inherit)) { 18376f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1838c47ca32dSDan Carpenter goto free_args; 18396f72c7e2SArne Jansen } 18406f72c7e2SArne Jansen } 184175eaa0e2SSage Weil 1842fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 18436f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 18448696c533SMiao Xie readonly, inherit); 1845c47ca32dSDan Carpenter if (ret) 1846c47ca32dSDan Carpenter goto free_inherit; 184775eaa0e2SSage Weil 1848c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1849fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1850c47ca32dSDan Carpenter transid), 1851c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 185275eaa0e2SSage Weil ret = -EFAULT; 1853c47ca32dSDan Carpenter 1854c47ca32dSDan Carpenter free_inherit: 18556f72c7e2SArne Jansen kfree(inherit); 1856c47ca32dSDan Carpenter free_args: 1857c47ca32dSDan Carpenter kfree(vol_args); 1858f46b5a66SChristoph Hellwig return ret; 1859f46b5a66SChristoph Hellwig } 1860f46b5a66SChristoph Hellwig 18610caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 18620caa102dSLi Zefan void __user *arg) 18630caa102dSLi Zefan { 1864496ad9aaSAl Viro struct inode *inode = file_inode(file); 18650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18660caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18670caa102dSLi Zefan int ret = 0; 18680caa102dSLi Zefan u64 flags = 0; 18690caa102dSLi Zefan 18704a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 18710caa102dSLi Zefan return -EINVAL; 18720caa102dSLi Zefan 18730b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 18740caa102dSLi Zefan if (btrfs_root_readonly(root)) 18750caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 18760b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 18770caa102dSLi Zefan 18780caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 18790caa102dSLi Zefan ret = -EFAULT; 18800caa102dSLi Zefan 18810caa102dSLi Zefan return ret; 18820caa102dSLi Zefan } 18830caa102dSLi Zefan 18840caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 18850caa102dSLi Zefan void __user *arg) 18860caa102dSLi Zefan { 1887496ad9aaSAl Viro struct inode *inode = file_inode(file); 18880b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18890caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18900caa102dSLi Zefan struct btrfs_trans_handle *trans; 18910caa102dSLi Zefan u64 root_flags; 18920caa102dSLi Zefan u64 flags; 18930caa102dSLi Zefan int ret = 0; 18940caa102dSLi Zefan 1895bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1896bd60ea0fSDavid Sterba return -EPERM; 1897bd60ea0fSDavid Sterba 1898b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1899b9ca0664SLiu Bo if (ret) 1900b9ca0664SLiu Bo goto out; 19010caa102dSLi Zefan 19024a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1903b9ca0664SLiu Bo ret = -EINVAL; 1904b9ca0664SLiu Bo goto out_drop_write; 1905b9ca0664SLiu Bo } 19060caa102dSLi Zefan 1907b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1908b9ca0664SLiu Bo ret = -EFAULT; 1909b9ca0664SLiu Bo goto out_drop_write; 1910b9ca0664SLiu Bo } 19110caa102dSLi Zefan 1912b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1913b9ca0664SLiu Bo ret = -EINVAL; 1914b9ca0664SLiu Bo goto out_drop_write; 1915b9ca0664SLiu Bo } 19160caa102dSLi Zefan 1917b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1918b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1919b9ca0664SLiu Bo goto out_drop_write; 1920b9ca0664SLiu Bo } 19210caa102dSLi Zefan 19220b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 19230caa102dSLi Zefan 19240caa102dSLi Zefan /* nothing to do */ 19250caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1926b9ca0664SLiu Bo goto out_drop_sem; 19270caa102dSLi Zefan 19280caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 19292c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 19300caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19310caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 19322c686537SDavid Sterba } else { 19332c686537SDavid Sterba /* 19342c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 19352c686537SDavid Sterba * send 19362c686537SDavid Sterba */ 19372c686537SDavid Sterba spin_lock(&root->root_item_lock); 19382c686537SDavid Sterba if (root->send_in_progress == 0) { 19390caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 19400caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 19412c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19422c686537SDavid Sterba } else { 19432c686537SDavid Sterba spin_unlock(&root->root_item_lock); 19440b246afaSJeff Mahoney btrfs_warn(fs_info, 19452c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 19462c686537SDavid Sterba root->root_key.objectid); 19472c686537SDavid Sterba ret = -EPERM; 19482c686537SDavid Sterba goto out_drop_sem; 19492c686537SDavid Sterba } 19502c686537SDavid Sterba } 19510caa102dSLi Zefan 19520caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 19530caa102dSLi Zefan if (IS_ERR(trans)) { 19540caa102dSLi Zefan ret = PTR_ERR(trans); 19550caa102dSLi Zefan goto out_reset; 19560caa102dSLi Zefan } 19570caa102dSLi Zefan 19580b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 19590caa102dSLi Zefan &root->root_key, &root->root_item); 19609417ebc8SNikolay Borisov if (ret < 0) { 19619417ebc8SNikolay Borisov btrfs_end_transaction(trans); 19629417ebc8SNikolay Borisov goto out_reset; 19639417ebc8SNikolay Borisov } 19640caa102dSLi Zefan 19659417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 19669417ebc8SNikolay Borisov 19670caa102dSLi Zefan out_reset: 19680caa102dSLi Zefan if (ret) 19690caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1970b9ca0664SLiu Bo out_drop_sem: 19710b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1972b9ca0664SLiu Bo out_drop_write: 1973b9ca0664SLiu Bo mnt_drop_write_file(file); 1974b9ca0664SLiu Bo out: 19750caa102dSLi Zefan return ret; 19760caa102dSLi Zefan } 19770caa102dSLi Zefan 1978ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1979ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1980ac8e9819SChris Mason { 1981abc6e134SChris Mason struct btrfs_key test; 1982abc6e134SChris Mason int ret; 1983abc6e134SChris Mason 1984abc6e134SChris Mason test.objectid = sk->min_objectid; 1985abc6e134SChris Mason test.type = sk->min_type; 1986abc6e134SChris Mason test.offset = sk->min_offset; 1987abc6e134SChris Mason 1988abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1989abc6e134SChris Mason if (ret < 0) 1990ac8e9819SChris Mason return 0; 1991abc6e134SChris Mason 1992abc6e134SChris Mason test.objectid = sk->max_objectid; 1993abc6e134SChris Mason test.type = sk->max_type; 1994abc6e134SChris Mason test.offset = sk->max_offset; 1995abc6e134SChris Mason 1996abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1997abc6e134SChris Mason if (ret > 0) 1998ac8e9819SChris Mason return 0; 1999ac8e9819SChris Mason return 1; 2000ac8e9819SChris Mason } 2001ac8e9819SChris Mason 2002df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 2003ac8e9819SChris Mason struct btrfs_key *key, 2004ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 20059b6e817dSGerhard Heift size_t *buf_size, 2006ba346b35SGerhard Heift char __user *ubuf, 2007ac8e9819SChris Mason unsigned long *sk_offset, 2008ac8e9819SChris Mason int *num_found) 2009ac8e9819SChris Mason { 2010ac8e9819SChris Mason u64 found_transid; 2011ac8e9819SChris Mason struct extent_buffer *leaf; 2012ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 2013dd81d459SNaohiro Aota struct btrfs_key test; 2014ac8e9819SChris Mason unsigned long item_off; 2015ac8e9819SChris Mason unsigned long item_len; 2016ac8e9819SChris Mason int nritems; 2017ac8e9819SChris Mason int i; 2018ac8e9819SChris Mason int slot; 2019ac8e9819SChris Mason int ret = 0; 2020ac8e9819SChris Mason 2021ac8e9819SChris Mason leaf = path->nodes[0]; 2022ac8e9819SChris Mason slot = path->slots[0]; 2023ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 2024ac8e9819SChris Mason 2025ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 2026ac8e9819SChris Mason i = nritems; 2027ac8e9819SChris Mason goto advance_key; 2028ac8e9819SChris Mason } 2029ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 2030ac8e9819SChris Mason 2031ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 2032ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 2033ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 2034ac8e9819SChris Mason 203503b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 203603b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 203703b71c6cSGabriel de Perthuis continue; 203803b71c6cSGabriel de Perthuis 20399b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20408f5f6178SGerhard Heift if (*num_found) { 2041ac8e9819SChris Mason ret = 1; 20428f5f6178SGerhard Heift goto out; 20438f5f6178SGerhard Heift } 20448f5f6178SGerhard Heift 20458f5f6178SGerhard Heift /* 20468f5f6178SGerhard Heift * return one empty item back for v1, which does not 20478f5f6178SGerhard Heift * handle -EOVERFLOW 20488f5f6178SGerhard Heift */ 20498f5f6178SGerhard Heift 20509b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2051ac8e9819SChris Mason item_len = 0; 20528f5f6178SGerhard Heift ret = -EOVERFLOW; 20538f5f6178SGerhard Heift } 2054ac8e9819SChris Mason 20559b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2056ac8e9819SChris Mason ret = 1; 205725c9bc2eSGerhard Heift goto out; 2058ac8e9819SChris Mason } 2059ac8e9819SChris Mason 2060ac8e9819SChris Mason sh.objectid = key->objectid; 2061ac8e9819SChris Mason sh.offset = key->offset; 2062ac8e9819SChris Mason sh.type = key->type; 2063ac8e9819SChris Mason sh.len = item_len; 2064ac8e9819SChris Mason sh.transid = found_transid; 2065ac8e9819SChris Mason 2066ac8e9819SChris Mason /* copy search result header */ 2067ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2068ba346b35SGerhard Heift ret = -EFAULT; 2069ba346b35SGerhard Heift goto out; 2070ba346b35SGerhard Heift } 2071ba346b35SGerhard Heift 2072ac8e9819SChris Mason *sk_offset += sizeof(sh); 2073ac8e9819SChris Mason 2074ac8e9819SChris Mason if (item_len) { 2075ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2076ac8e9819SChris Mason /* copy the item */ 2077ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2078ba346b35SGerhard Heift item_off, item_len)) { 2079ba346b35SGerhard Heift ret = -EFAULT; 2080ba346b35SGerhard Heift goto out; 2081ba346b35SGerhard Heift } 2082ba346b35SGerhard Heift 2083ac8e9819SChris Mason *sk_offset += item_len; 2084ac8e9819SChris Mason } 2085e2156867SHugo Mills (*num_found)++; 2086ac8e9819SChris Mason 20878f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20888f5f6178SGerhard Heift goto out; 20898f5f6178SGerhard Heift 209025c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 209125c9bc2eSGerhard Heift ret = 1; 209225c9bc2eSGerhard Heift goto out; 209325c9bc2eSGerhard Heift } 2094ac8e9819SChris Mason } 2095ac8e9819SChris Mason advance_key: 2096ac8e9819SChris Mason ret = 0; 2097dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2098dd81d459SNaohiro Aota test.type = sk->max_type; 2099dd81d459SNaohiro Aota test.offset = sk->max_offset; 2100dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2101dd81d459SNaohiro Aota ret = 1; 2102dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2103abc6e134SChris Mason key->offset++; 2104dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2105abc6e134SChris Mason key->offset = 0; 2106abc6e134SChris Mason key->type++; 2107dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2108abc6e134SChris Mason key->offset = 0; 2109abc6e134SChris Mason key->type = 0; 2110abc6e134SChris Mason key->objectid++; 2111abc6e134SChris Mason } else 2112abc6e134SChris Mason ret = 1; 211325c9bc2eSGerhard Heift out: 2114ba346b35SGerhard Heift /* 2115ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2116ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2117ba346b35SGerhard Heift * * all items were found 2118ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2119ba346b35SGerhard Heift * leaf 2120ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2121ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2122ba346b35SGerhard Heift */ 2123ac8e9819SChris Mason return ret; 2124ac8e9819SChris Mason } 2125ac8e9819SChris Mason 2126ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 212712544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 21289b6e817dSGerhard Heift size_t *buf_size, 2129ba346b35SGerhard Heift char __user *ubuf) 2130ac8e9819SChris Mason { 21310b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2132ac8e9819SChris Mason struct btrfs_root *root; 2133ac8e9819SChris Mason struct btrfs_key key; 2134ac8e9819SChris Mason struct btrfs_path *path; 2135ac8e9819SChris Mason int ret; 2136ac8e9819SChris Mason int num_found = 0; 2137ac8e9819SChris Mason unsigned long sk_offset = 0; 2138ac8e9819SChris Mason 21399b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21409b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 214112544442SGerhard Heift return -EOVERFLOW; 21429b6e817dSGerhard Heift } 214312544442SGerhard Heift 2144ac8e9819SChris Mason path = btrfs_alloc_path(); 2145ac8e9819SChris Mason if (!path) 2146ac8e9819SChris Mason return -ENOMEM; 2147ac8e9819SChris Mason 2148ac8e9819SChris Mason if (sk->tree_id == 0) { 2149ac8e9819SChris Mason /* search the root of the inode that was passed */ 2150ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2151ac8e9819SChris Mason } else { 2152ac8e9819SChris Mason key.objectid = sk->tree_id; 2153ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2154ac8e9819SChris Mason key.offset = (u64)-1; 2155ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2156ac8e9819SChris Mason if (IS_ERR(root)) { 2157ac8e9819SChris Mason btrfs_free_path(path); 2158ad1e3d56SMisono Tomohiro return PTR_ERR(root); 2159ac8e9819SChris Mason } 2160ac8e9819SChris Mason } 2161ac8e9819SChris Mason 2162ac8e9819SChris Mason key.objectid = sk->min_objectid; 2163ac8e9819SChris Mason key.type = sk->min_type; 2164ac8e9819SChris Mason key.offset = sk->min_offset; 2165ac8e9819SChris Mason 2166ac8e9819SChris Mason while (1) { 21676174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2168ac8e9819SChris Mason if (ret != 0) { 2169ac8e9819SChris Mason if (ret > 0) 2170ac8e9819SChris Mason ret = 0; 2171ac8e9819SChris Mason goto err; 2172ac8e9819SChris Mason } 2173df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2174ac8e9819SChris Mason &sk_offset, &num_found); 2175b3b4aa74SDavid Sterba btrfs_release_path(path); 217625c9bc2eSGerhard Heift if (ret) 2177ac8e9819SChris Mason break; 2178ac8e9819SChris Mason 2179ac8e9819SChris Mason } 21808f5f6178SGerhard Heift if (ret > 0) 2181ac8e9819SChris Mason ret = 0; 2182ac8e9819SChris Mason err: 2183ac8e9819SChris Mason sk->nr_items = num_found; 2184ac8e9819SChris Mason btrfs_free_path(path); 2185ac8e9819SChris Mason return ret; 2186ac8e9819SChris Mason } 2187ac8e9819SChris Mason 2188ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2189ac8e9819SChris Mason void __user *argp) 2190ac8e9819SChris Mason { 2191ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2192ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2193ac8e9819SChris Mason struct inode *inode; 2194ac8e9819SChris Mason int ret; 21959b6e817dSGerhard Heift size_t buf_size; 2196ac8e9819SChris Mason 2197ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2198ac8e9819SChris Mason return -EPERM; 2199ac8e9819SChris Mason 2200ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2201ac8e9819SChris Mason 2202ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2203ba346b35SGerhard Heift return -EFAULT; 2204ba346b35SGerhard Heift 2205ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2206ac8e9819SChris Mason 2207496ad9aaSAl Viro inode = file_inode(file); 2208ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 22098f5f6178SGerhard Heift 22108f5f6178SGerhard Heift /* 22118f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 22128f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 22138f5f6178SGerhard Heift */ 22148f5f6178SGerhard Heift if (ret == -EOVERFLOW) 22158f5f6178SGerhard Heift ret = 0; 22168f5f6178SGerhard Heift 2217ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2218ac8e9819SChris Mason ret = -EFAULT; 2219ac8e9819SChris Mason return ret; 2220ac8e9819SChris Mason } 2221ac8e9819SChris Mason 2222cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2223cc68a8a5SGerhard Heift void __user *argp) 2224cc68a8a5SGerhard Heift { 2225cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2226cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2227cc68a8a5SGerhard Heift struct inode *inode; 2228cc68a8a5SGerhard Heift int ret; 2229cc68a8a5SGerhard Heift size_t buf_size; 2230ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2231cc68a8a5SGerhard Heift 2232cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2233cc68a8a5SGerhard Heift return -EPERM; 2234cc68a8a5SGerhard Heift 2235cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2236cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2237cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2238cc68a8a5SGerhard Heift return -EFAULT; 2239cc68a8a5SGerhard Heift 2240cc68a8a5SGerhard Heift buf_size = args.buf_size; 2241cc68a8a5SGerhard Heift 2242cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2243cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2244cc68a8a5SGerhard Heift buf_size = buf_limit; 2245cc68a8a5SGerhard Heift 2246cc68a8a5SGerhard Heift inode = file_inode(file); 2247cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2248718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2249cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2250cc68a8a5SGerhard Heift ret = -EFAULT; 2251cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2252cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2253cc68a8a5SGerhard Heift ret = -EFAULT; 2254cc68a8a5SGerhard Heift 225576dda93cSYan, Zheng return ret; 225676dda93cSYan, Zheng } 225776dda93cSYan, Zheng 225898d377a0STARUISI Hiroaki /* 2259ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2260ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 226198d377a0STARUISI Hiroaki */ 226298d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 226398d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 226498d377a0STARUISI Hiroaki { 226598d377a0STARUISI Hiroaki struct btrfs_root *root; 226698d377a0STARUISI Hiroaki struct btrfs_key key; 2267ac8e9819SChris Mason char *ptr; 226898d377a0STARUISI Hiroaki int ret = -1; 226998d377a0STARUISI Hiroaki int slot; 227098d377a0STARUISI Hiroaki int len; 227198d377a0STARUISI Hiroaki int total_len = 0; 227298d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 227398d377a0STARUISI Hiroaki struct extent_buffer *l; 227498d377a0STARUISI Hiroaki struct btrfs_path *path; 227598d377a0STARUISI Hiroaki 227698d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 227798d377a0STARUISI Hiroaki name[0]='\0'; 227898d377a0STARUISI Hiroaki return 0; 227998d377a0STARUISI Hiroaki } 228098d377a0STARUISI Hiroaki 228198d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 228298d377a0STARUISI Hiroaki if (!path) 228398d377a0STARUISI Hiroaki return -ENOMEM; 228498d377a0STARUISI Hiroaki 2285c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 228698d377a0STARUISI Hiroaki 228798d377a0STARUISI Hiroaki key.objectid = tree_id; 228898d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 228998d377a0STARUISI Hiroaki key.offset = (u64)-1; 229098d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 229198d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2292ad1e3d56SMisono Tomohiro ret = PTR_ERR(root); 22938ad6fcabSChris Mason goto out; 229498d377a0STARUISI Hiroaki } 229598d377a0STARUISI Hiroaki 229698d377a0STARUISI Hiroaki key.objectid = dirid; 229798d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22988ad6fcabSChris Mason key.offset = (u64)-1; 229998d377a0STARUISI Hiroaki 230098d377a0STARUISI Hiroaki while (1) { 230198d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 230298d377a0STARUISI Hiroaki if (ret < 0) 230398d377a0STARUISI Hiroaki goto out; 230418674c6cSFilipe David Borba Manana else if (ret > 0) { 230518674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 230618674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 230718674c6cSFilipe David Borba Manana if (ret < 0) 230818674c6cSFilipe David Borba Manana goto out; 230918674c6cSFilipe David Borba Manana else if (ret > 0) { 2310ac8e9819SChris Mason ret = -ENOENT; 231198d377a0STARUISI Hiroaki goto out; 2312ac8e9819SChris Mason } 231318674c6cSFilipe David Borba Manana } 231418674c6cSFilipe David Borba Manana 231518674c6cSFilipe David Borba Manana l = path->nodes[0]; 231618674c6cSFilipe David Borba Manana slot = path->slots[0]; 231718674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 231898d377a0STARUISI Hiroaki 231998d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 232098d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 232198d377a0STARUISI Hiroaki ptr -= len + 1; 232298d377a0STARUISI Hiroaki total_len += len + 1; 2323a696cf35SFilipe David Borba Manana if (ptr < name) { 2324a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 232598d377a0STARUISI Hiroaki goto out; 2326a696cf35SFilipe David Borba Manana } 232798d377a0STARUISI Hiroaki 232898d377a0STARUISI Hiroaki *(ptr + len) = '/'; 232998d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 233098d377a0STARUISI Hiroaki 233198d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 233298d377a0STARUISI Hiroaki break; 233398d377a0STARUISI Hiroaki 2334b3b4aa74SDavid Sterba btrfs_release_path(path); 233598d377a0STARUISI Hiroaki key.objectid = key.offset; 23368ad6fcabSChris Mason key.offset = (u64)-1; 233798d377a0STARUISI Hiroaki dirid = key.objectid; 233898d377a0STARUISI Hiroaki } 233977906a50SLi Zefan memmove(name, ptr, total_len); 234098d377a0STARUISI Hiroaki name[total_len] = '\0'; 234198d377a0STARUISI Hiroaki ret = 0; 234298d377a0STARUISI Hiroaki out: 234398d377a0STARUISI Hiroaki btrfs_free_path(path); 2344ac8e9819SChris Mason return ret; 2345ac8e9819SChris Mason } 2346ac8e9819SChris Mason 2347ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2348ac8e9819SChris Mason void __user *argp) 2349ac8e9819SChris Mason { 2350ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2351ac8e9819SChris Mason struct inode *inode; 235201b810b8SDavid Sterba int ret = 0; 2353ac8e9819SChris Mason 23542354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 23552354d08fSJulia Lawall if (IS_ERR(args)) 23562354d08fSJulia Lawall return PTR_ERR(args); 2357c2b96929SDan Carpenter 2358496ad9aaSAl Viro inode = file_inode(file); 2359ac8e9819SChris Mason 236001b810b8SDavid Sterba /* 236101b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 236201b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 236301b810b8SDavid Sterba */ 23641b53ac4dSChris Mason if (args->treeid == 0) 23651b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 23661b53ac4dSChris Mason 236701b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 236801b810b8SDavid Sterba args->name[0] = 0; 236901b810b8SDavid Sterba goto out; 237001b810b8SDavid Sterba } 237101b810b8SDavid Sterba 237201b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 237301b810b8SDavid Sterba ret = -EPERM; 237401b810b8SDavid Sterba goto out; 237501b810b8SDavid Sterba } 237601b810b8SDavid Sterba 2377ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2378ac8e9819SChris Mason args->treeid, args->objectid, 2379ac8e9819SChris Mason args->name); 2380ac8e9819SChris Mason 238101b810b8SDavid Sterba out: 2382ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2383ac8e9819SChris Mason ret = -EFAULT; 2384ac8e9819SChris Mason 2385ac8e9819SChris Mason kfree(args); 238698d377a0STARUISI Hiroaki return ret; 238798d377a0STARUISI Hiroaki } 238898d377a0STARUISI Hiroaki 238976dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 239076dda93cSYan, Zheng void __user *arg) 239176dda93cSYan, Zheng { 239254563d41SAl Viro struct dentry *parent = file->f_path.dentry; 23930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 239476dda93cSYan, Zheng struct dentry *dentry; 23952b0143b5SDavid Howells struct inode *dir = d_inode(parent); 239676dda93cSYan, Zheng struct inode *inode; 239776dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 239876dda93cSYan, Zheng struct btrfs_root *dest = NULL; 239976dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 240076dda93cSYan, Zheng int namelen; 240176dda93cSYan, Zheng int err = 0; 240276dda93cSYan, Zheng 2403325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2404325c50e3SJeff Mahoney return -ENOTDIR; 2405325c50e3SJeff Mahoney 240676dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 240776dda93cSYan, Zheng if (IS_ERR(vol_args)) 240876dda93cSYan, Zheng return PTR_ERR(vol_args); 240976dda93cSYan, Zheng 241076dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 241176dda93cSYan, Zheng namelen = strlen(vol_args->name); 241276dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 241376dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 241476dda93cSYan, Zheng err = -EINVAL; 241576dda93cSYan, Zheng goto out; 241676dda93cSYan, Zheng } 241776dda93cSYan, Zheng 2418a561be71SAl Viro err = mnt_want_write_file(file); 241976dda93cSYan, Zheng if (err) 242076dda93cSYan, Zheng goto out; 242176dda93cSYan, Zheng 2422521e0546SDavid Sterba 242300235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 242400235411SAl Viro if (err == -EINTR) 242500235411SAl Viro goto out_drop_write; 242676dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 242776dda93cSYan, Zheng if (IS_ERR(dentry)) { 242876dda93cSYan, Zheng err = PTR_ERR(dentry); 242976dda93cSYan, Zheng goto out_unlock_dir; 243076dda93cSYan, Zheng } 243176dda93cSYan, Zheng 24322b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 243376dda93cSYan, Zheng err = -ENOENT; 243476dda93cSYan, Zheng goto out_dput; 243576dda93cSYan, Zheng } 243676dda93cSYan, Zheng 24372b0143b5SDavid Howells inode = d_inode(dentry); 24384260f7c7SSage Weil dest = BTRFS_I(inode)->root; 24394260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 24404260f7c7SSage Weil /* 24414260f7c7SSage Weil * Regular user. Only allow this with a special mount 24424260f7c7SSage Weil * option, when the user has write+exec access to the 24434260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 24444260f7c7SSage Weil * allowed. 24454260f7c7SSage Weil * 24464260f7c7SSage Weil * Note that this is _not_ check that the subvol is 24474260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 24484260f7c7SSage Weil * otherwise be able to delete. 24494260f7c7SSage Weil * 24504260f7c7SSage Weil * Users who want to delete empty subvols should try 24514260f7c7SSage Weil * rmdir(2). 24524260f7c7SSage Weil */ 24534260f7c7SSage Weil err = -EPERM; 24540b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 24554260f7c7SSage Weil goto out_dput; 24564260f7c7SSage Weil 24574260f7c7SSage Weil /* 24584260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 24594260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 24604260f7c7SSage Weil * must be called on the dentry referencing the root 24614260f7c7SSage Weil * of the subvol, not a random directory contained 24624260f7c7SSage Weil * within it. 24634260f7c7SSage Weil */ 24644260f7c7SSage Weil err = -EINVAL; 24654260f7c7SSage Weil if (root == dest) 24664260f7c7SSage Weil goto out_dput; 24674260f7c7SSage Weil 24684260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 24694260f7c7SSage Weil if (err) 24704260f7c7SSage Weil goto out_dput; 24715c39da5bSMiao Xie } 24724260f7c7SSage Weil 24735c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24744260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24754260f7c7SSage Weil if (err) 24764260f7c7SSage Weil goto out_dput; 24774260f7c7SSage Weil 24784a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 247976dda93cSYan, Zheng err = -EINVAL; 248076dda93cSYan, Zheng goto out_dput; 248176dda93cSYan, Zheng } 248276dda93cSYan, Zheng 24835955102cSAl Viro inode_lock(inode); 2484f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 24855955102cSAl Viro inode_unlock(inode); 2486f60a2364SMisono Tomohiro if (!err) 248776dda93cSYan, Zheng d_delete(dentry); 2488fa6ac876SLiu Bo 248976dda93cSYan, Zheng out_dput: 249076dda93cSYan, Zheng dput(dentry); 249176dda93cSYan, Zheng out_unlock_dir: 24925955102cSAl Viro inode_unlock(dir); 249300235411SAl Viro out_drop_write: 24942a79f17eSAl Viro mnt_drop_write_file(file); 249576dda93cSYan, Zheng out: 249676dda93cSYan, Zheng kfree(vol_args); 249776dda93cSYan, Zheng return err; 249876dda93cSYan, Zheng } 249976dda93cSYan, Zheng 25001e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2501f46b5a66SChristoph Hellwig { 2502496ad9aaSAl Viro struct inode *inode = file_inode(file); 2503f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25041e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2505c146afadSYan Zheng int ret; 2506c146afadSYan Zheng 250725122d15SIlya Dryomov ret = mnt_want_write_file(file); 250825122d15SIlya Dryomov if (ret) 250925122d15SIlya Dryomov return ret; 2510b83cc969SLi Zefan 251125122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 251225122d15SIlya Dryomov ret = -EROFS; 251325122d15SIlya Dryomov goto out; 25145ac00addSStefan Behrens } 2515f46b5a66SChristoph Hellwig 2516f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2517f46b5a66SChristoph Hellwig case S_IFDIR: 2518e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2519e441d54dSChris Mason ret = -EPERM; 2520e441d54dSChris Mason goto out; 2521e441d54dSChris Mason } 2522de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2523f46b5a66SChristoph Hellwig break; 2524f46b5a66SChristoph Hellwig case S_IFREG: 2525e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2526e441d54dSChris Mason ret = -EINVAL; 2527e441d54dSChris Mason goto out; 2528e441d54dSChris Mason } 25291e701a32SChris Mason 25301e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 25311e701a32SChris Mason if (!range) { 25321e701a32SChris Mason ret = -ENOMEM; 25331e701a32SChris Mason goto out; 25341e701a32SChris Mason } 25351e701a32SChris Mason 25361e701a32SChris Mason if (argp) { 25371e701a32SChris Mason if (copy_from_user(range, argp, 25381e701a32SChris Mason sizeof(*range))) { 25391e701a32SChris Mason ret = -EFAULT; 25401e701a32SChris Mason kfree(range); 2541683be16eSDan Carpenter goto out; 25421e701a32SChris Mason } 25431e701a32SChris Mason /* compression requires us to start the IO */ 25441e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 25451e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 25461e701a32SChris Mason range->extent_thresh = (u32)-1; 25471e701a32SChris Mason } 25481e701a32SChris Mason } else { 25491e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 25501e701a32SChris Mason range->len = (u64)-1; 25511e701a32SChris Mason } 2552496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 25537c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 25544cb5300bSChris Mason if (ret > 0) 25554cb5300bSChris Mason ret = 0; 25561e701a32SChris Mason kfree(range); 2557f46b5a66SChristoph Hellwig break; 25588929ecfaSYan, Zheng default: 25598929ecfaSYan, Zheng ret = -EINVAL; 2560f46b5a66SChristoph Hellwig } 2561e441d54dSChris Mason out: 256225122d15SIlya Dryomov mnt_drop_write_file(file); 2563e441d54dSChris Mason return ret; 2564f46b5a66SChristoph Hellwig } 2565f46b5a66SChristoph Hellwig 25662ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2567f46b5a66SChristoph Hellwig { 2568f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2569f46b5a66SChristoph Hellwig int ret; 2570f46b5a66SChristoph Hellwig 2571e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2572e441d54dSChris Mason return -EPERM; 2573e441d54dSChris Mason 2574171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 2575e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2576c9e9f97bSIlya Dryomov 2577dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2578c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2579c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2580c9e9f97bSIlya Dryomov goto out; 2581c9e9f97bSIlya Dryomov } 2582f46b5a66SChristoph Hellwig 25835516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 25842ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 2585f46b5a66SChristoph Hellwig 258643d20761SAnand Jain if (!ret) 25870b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 258843d20761SAnand Jain 2589f46b5a66SChristoph Hellwig kfree(vol_args); 2590c9e9f97bSIlya Dryomov out: 2591171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2592f46b5a66SChristoph Hellwig return ret; 2593f46b5a66SChristoph Hellwig } 2594f46b5a66SChristoph Hellwig 25956b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2596f46b5a66SChristoph Hellwig { 25970b246afaSJeff Mahoney struct inode *inode = file_inode(file); 25980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 25996b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 2600f46b5a66SChristoph Hellwig int ret; 2601f46b5a66SChristoph Hellwig 2602e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2603e441d54dSChris Mason return -EPERM; 2604e441d54dSChris Mason 2605da24927bSMiao Xie ret = mnt_want_write_file(file); 2606da24927bSMiao Xie if (ret) 2607da24927bSMiao Xie return ret; 2608c146afadSYan Zheng 2609dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2610c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2611c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2612c47ca32dSDan Carpenter goto err_drop; 2613c9e9f97bSIlya Dryomov } 2614f46b5a66SChristoph Hellwig 26156b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 2616fd4e994bSOmar Sandoval if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) { 2617fd4e994bSOmar Sandoval ret = -EOPNOTSUPP; 2618fd4e994bSOmar Sandoval goto out; 2619fd4e994bSOmar Sandoval } 2620f46b5a66SChristoph Hellwig 2621171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 2622183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2623183860f6SAnand Jain goto out; 2624183860f6SAnand Jain } 2625183860f6SAnand Jain 2626735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 26272ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 26286b526ed7SAnand Jain } else { 26296b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 26302ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 26316b526ed7SAnand Jain } 2632171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2633183860f6SAnand Jain 26346b526ed7SAnand Jain if (!ret) { 2635735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 26360b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 26376b526ed7SAnand Jain vol_args->devid); 26386b526ed7SAnand Jain else 26390b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 26406b526ed7SAnand Jain vol_args->name); 26416b526ed7SAnand Jain } 2642183860f6SAnand Jain out: 2643183860f6SAnand Jain kfree(vol_args); 2644c47ca32dSDan Carpenter err_drop: 26454ac20c70SIlya Dryomov mnt_drop_write_file(file); 2646f46b5a66SChristoph Hellwig return ret; 2647f46b5a66SChristoph Hellwig } 2648f46b5a66SChristoph Hellwig 2649f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2650f46b5a66SChristoph Hellwig { 26510b246afaSJeff Mahoney struct inode *inode = file_inode(file); 26520b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2653f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2654f46b5a66SChristoph Hellwig int ret; 2655f46b5a66SChristoph Hellwig 2656f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2657f46b5a66SChristoph Hellwig return -EPERM; 2658f46b5a66SChristoph Hellwig 2659f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 2660f46b5a66SChristoph Hellwig if (ret) 2661f46b5a66SChristoph Hellwig return ret; 2662f46b5a66SChristoph Hellwig 2663171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 2664f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 266558d7bbf8SDavid Sterba goto out_drop_write; 266658d7bbf8SDavid Sterba } 266758d7bbf8SDavid Sterba 266858d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 266958d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 267058d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 2671f46b5a66SChristoph Hellwig goto out; 2672f46b5a66SChristoph Hellwig } 2673f46b5a66SChristoph Hellwig 267458d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 26752ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 2676f46b5a66SChristoph Hellwig 2677f46b5a66SChristoph Hellwig if (!ret) 26780b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 2679f46b5a66SChristoph Hellwig kfree(vol_args); 268058d7bbf8SDavid Sterba out: 2681171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 268258d7bbf8SDavid Sterba out_drop_write: 2683f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 268458d7bbf8SDavid Sterba 2685f46b5a66SChristoph Hellwig return ret; 2686f46b5a66SChristoph Hellwig } 2687f46b5a66SChristoph Hellwig 26882ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 26892ff7e61eSJeff Mahoney void __user *arg) 2690475f6387SJan Schmidt { 2691027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2692475f6387SJan Schmidt struct btrfs_device *device; 26930b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2694027ed2f0SLi Zefan int ret = 0; 2695475f6387SJan Schmidt 2696027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2697027ed2f0SLi Zefan if (!fi_args) 2698027ed2f0SLi Zefan return -ENOMEM; 2699027ed2f0SLi Zefan 2700d03262c7SDavid Sterba rcu_read_lock(); 2701027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2702475f6387SJan Schmidt 2703d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 2704027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2705027ed2f0SLi Zefan fi_args->max_id = device->devid; 2706475f6387SJan Schmidt } 2707d03262c7SDavid Sterba rcu_read_unlock(); 2708475f6387SJan Schmidt 2709d03262c7SDavid Sterba memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 2710bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 2711bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 2712bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 271380a773fbSDavid Sterba 2714027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2715027ed2f0SLi Zefan ret = -EFAULT; 2716475f6387SJan Schmidt 2717027ed2f0SLi Zefan kfree(fi_args); 2718027ed2f0SLi Zefan return ret; 2719475f6387SJan Schmidt } 2720475f6387SJan Schmidt 27212ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 27222ff7e61eSJeff Mahoney void __user *arg) 2723475f6387SJan Schmidt { 2724475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2725475f6387SJan Schmidt struct btrfs_device *dev; 2726475f6387SJan Schmidt int ret = 0; 2727475f6387SJan Schmidt char *s_uuid = NULL; 2728475f6387SJan Schmidt 2729475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2730475f6387SJan Schmidt if (IS_ERR(di_args)) 2731475f6387SJan Schmidt return PTR_ERR(di_args); 2732475f6387SJan Schmidt 2733dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2734475f6387SJan Schmidt s_uuid = di_args->uuid; 2735475f6387SJan Schmidt 2736c5593ca3SDavid Sterba rcu_read_lock(); 27370b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 2738475f6387SJan Schmidt 2739475f6387SJan Schmidt if (!dev) { 2740475f6387SJan Schmidt ret = -ENODEV; 2741475f6387SJan Schmidt goto out; 2742475f6387SJan Schmidt } 2743475f6387SJan Schmidt 2744475f6387SJan Schmidt di_args->devid = dev->devid; 27457cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 27467cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2747475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2748a27202fbSJim Meyering if (dev->name) { 2749606686eeSJosef Bacik struct rcu_string *name; 2750606686eeSJosef Bacik 2751606686eeSJosef Bacik name = rcu_dereference(dev->name); 27526670d4c2SXiongfeng Wang strncpy(di_args->path, name->str, sizeof(di_args->path) - 1); 2753a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2754a27202fbSJim Meyering } else { 275599ba55adSStefan Behrens di_args->path[0] = '\0'; 2756a27202fbSJim Meyering } 2757475f6387SJan Schmidt 2758475f6387SJan Schmidt out: 2759c5593ca3SDavid Sterba rcu_read_unlock(); 2760475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2761475f6387SJan Schmidt ret = -EFAULT; 2762475f6387SJan Schmidt 2763475f6387SJan Schmidt kfree(di_args); 2764475f6387SJan Schmidt return ret; 2765475f6387SJan Schmidt } 2766475f6387SJan Schmidt 2767f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2768416161dbSMark Fasheh { 2769416161dbSMark Fasheh struct page *page; 2770416161dbSMark Fasheh 2771416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2772416161dbSMark Fasheh if (!page) 277331314002SFilipe Manana return ERR_PTR(-ENOMEM); 2774416161dbSMark Fasheh 2775416161dbSMark Fasheh if (!PageUptodate(page)) { 277631314002SFilipe Manana int ret; 277731314002SFilipe Manana 277831314002SFilipe Manana ret = btrfs_readpage(NULL, page); 277931314002SFilipe Manana if (ret) 278031314002SFilipe Manana return ERR_PTR(ret); 2781416161dbSMark Fasheh lock_page(page); 2782416161dbSMark Fasheh if (!PageUptodate(page)) { 2783416161dbSMark Fasheh unlock_page(page); 278409cbfeafSKirill A. Shutemov put_page(page); 278531314002SFilipe Manana return ERR_PTR(-EIO); 2786416161dbSMark Fasheh } 278731314002SFilipe Manana if (page->mapping != inode->i_mapping) { 2788416161dbSMark Fasheh unlock_page(page); 278909cbfeafSKirill A. Shutemov put_page(page); 279031314002SFilipe Manana return ERR_PTR(-EAGAIN); 2791416161dbSMark Fasheh } 2792416161dbSMark Fasheh } 2793416161dbSMark Fasheh 2794416161dbSMark Fasheh return page; 2795416161dbSMark Fasheh } 2796416161dbSMark Fasheh 2797f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2798f4414602SMark Fasheh int num_pages, u64 off) 2799f4414602SMark Fasheh { 2800f4414602SMark Fasheh int i; 280109cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 2802f4414602SMark Fasheh 2803f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 280431314002SFilipe Manana again: 2805f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 280631314002SFilipe Manana if (IS_ERR(pages[i])) { 280731314002SFilipe Manana int err = PTR_ERR(pages[i]); 280831314002SFilipe Manana 280931314002SFilipe Manana if (err == -EAGAIN) 281031314002SFilipe Manana goto again; 281131314002SFilipe Manana pages[i] = NULL; 281231314002SFilipe Manana return err; 281331314002SFilipe Manana } 2814f4414602SMark Fasheh } 2815f4414602SMark Fasheh return 0; 2816f4414602SMark Fasheh } 2817f4414602SMark Fasheh 2818e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 2819e0bd70c6SFilipe Manana bool retry_range_locking) 282077fe20dcSMark Fasheh { 2821e0bd70c6SFilipe Manana /* 2822e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 2823e0bd70c6SFilipe Manana * another, and lock file content. 2824e0bd70c6SFilipe Manana * The locking order is: 2825e0bd70c6SFilipe Manana * 2826e0bd70c6SFilipe Manana * 1) pages 2827e0bd70c6SFilipe Manana * 2) range in the inode's io tree 2828e0bd70c6SFilipe Manana */ 282977fe20dcSMark Fasheh while (1) { 283077fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 283177fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 283277fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 283377fe20dcSMark Fasheh off + len - 1); 2834ff5df9b8SFilipe Manana if ((!ordered || 2835ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2836ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 283777fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2838ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2839ff5df9b8SFilipe Manana if (ordered) 2840ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 284177fe20dcSMark Fasheh break; 2842ff5df9b8SFilipe Manana } 284377fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 284477fe20dcSMark Fasheh if (ordered) 284577fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 2846e0bd70c6SFilipe Manana if (!retry_range_locking) 2847e0bd70c6SFilipe Manana return -EAGAIN; 284877fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 284977fe20dcSMark Fasheh } 2850e0bd70c6SFilipe Manana return 0; 285177fe20dcSMark Fasheh } 285277fe20dcSMark Fasheh 2853f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2854416161dbSMark Fasheh { 28555955102cSAl Viro inode_unlock(inode1); 28565955102cSAl Viro inode_unlock(inode2); 2857416161dbSMark Fasheh } 2858416161dbSMark Fasheh 2859f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2860f4414602SMark Fasheh { 2861f4414602SMark Fasheh if (inode1 < inode2) 2862f4414602SMark Fasheh swap(inode1, inode2); 2863f4414602SMark Fasheh 28645955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 28655955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 2866f4414602SMark Fasheh } 2867f4414602SMark Fasheh 2868f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2869f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2870f4414602SMark Fasheh { 2871f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2872f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2873f4414602SMark Fasheh } 2874f4414602SMark Fasheh 2875e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2876e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 2877e0bd70c6SFilipe Manana bool retry_range_locking) 2878416161dbSMark Fasheh { 2879e0bd70c6SFilipe Manana int ret; 2880e0bd70c6SFilipe Manana 2881416161dbSMark Fasheh if (inode1 < inode2) { 2882416161dbSMark Fasheh swap(inode1, inode2); 2883416161dbSMark Fasheh swap(loff1, loff2); 2884416161dbSMark Fasheh } 2885e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 2886e0bd70c6SFilipe Manana if (ret) 2887e0bd70c6SFilipe Manana return ret; 2888e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 2889e0bd70c6SFilipe Manana if (ret) 2890e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 2891e0bd70c6SFilipe Manana loff1 + len - 1); 2892e0bd70c6SFilipe Manana return ret; 2893416161dbSMark Fasheh } 2894f4414602SMark Fasheh 2895f4414602SMark Fasheh struct cmp_pages { 2896f4414602SMark Fasheh int num_pages; 2897f4414602SMark Fasheh struct page **src_pages; 2898f4414602SMark Fasheh struct page **dst_pages; 2899f4414602SMark Fasheh }; 2900f4414602SMark Fasheh 2901f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 2902f4414602SMark Fasheh { 2903f4414602SMark Fasheh int i; 2904f4414602SMark Fasheh struct page *pg; 2905f4414602SMark Fasheh 2906f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 2907f4414602SMark Fasheh pg = cmp->src_pages[i]; 2908e0bd70c6SFilipe Manana if (pg) { 2909e0bd70c6SFilipe Manana unlock_page(pg); 291009cbfeafSKirill A. Shutemov put_page(pg); 2911e0bd70c6SFilipe Manana } 2912f4414602SMark Fasheh pg = cmp->dst_pages[i]; 2913e0bd70c6SFilipe Manana if (pg) { 2914e0bd70c6SFilipe Manana unlock_page(pg); 291509cbfeafSKirill A. Shutemov put_page(pg); 2916f4414602SMark Fasheh } 2917e0bd70c6SFilipe Manana } 2918f4414602SMark Fasheh } 2919f4414602SMark Fasheh 2920f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 2921f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 2922f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 2923f4414602SMark Fasheh { 2924f4414602SMark Fasheh int ret; 292509cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 2926f4414602SMark Fasheh 2927f4414602SMark Fasheh cmp->num_pages = num_pages; 2928f4414602SMark Fasheh 292967b07bd4STimofey Titovets ret = gather_extent_pages(src, cmp->src_pages, num_pages, loff); 2930f4414602SMark Fasheh if (ret) 2931f4414602SMark Fasheh goto out; 2932f4414602SMark Fasheh 293367b07bd4STimofey Titovets ret = gather_extent_pages(dst, cmp->dst_pages, num_pages, dst_loff); 2934f4414602SMark Fasheh 2935f4414602SMark Fasheh out: 2936f4414602SMark Fasheh if (ret) 2937f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 293878ad4ce0SNaohiro Aota return ret; 2939416161dbSMark Fasheh } 2940416161dbSMark Fasheh 29411a287cfeSDavid Sterba static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp) 2942416161dbSMark Fasheh { 2943416161dbSMark Fasheh int ret = 0; 2944f4414602SMark Fasheh int i; 2945416161dbSMark Fasheh struct page *src_page, *dst_page; 294609cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 2947416161dbSMark Fasheh void *addr, *dst_addr; 2948416161dbSMark Fasheh 2949f4414602SMark Fasheh i = 0; 2950416161dbSMark Fasheh while (len) { 295109cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 2952416161dbSMark Fasheh cmp_len = len; 2953416161dbSMark Fasheh 2954f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 2955f4414602SMark Fasheh 2956f4414602SMark Fasheh src_page = cmp->src_pages[i]; 2957f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 2958e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 2959e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 2960f4414602SMark Fasheh 2961416161dbSMark Fasheh addr = kmap_atomic(src_page); 2962416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 2963416161dbSMark Fasheh 2964416161dbSMark Fasheh flush_dcache_page(src_page); 2965416161dbSMark Fasheh flush_dcache_page(dst_page); 2966416161dbSMark Fasheh 2967416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 29682b3909f8SDarrick J. Wong ret = -EBADE; 2969416161dbSMark Fasheh 2970416161dbSMark Fasheh kunmap_atomic(addr); 2971416161dbSMark Fasheh kunmap_atomic(dst_addr); 2972416161dbSMark Fasheh 2973416161dbSMark Fasheh if (ret) 2974416161dbSMark Fasheh break; 2975416161dbSMark Fasheh 2976416161dbSMark Fasheh len -= cmp_len; 2977f4414602SMark Fasheh i++; 2978416161dbSMark Fasheh } 2979416161dbSMark Fasheh 2980416161dbSMark Fasheh return ret; 2981416161dbSMark Fasheh } 2982416161dbSMark Fasheh 2983e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 2984e1d227a4SMark Fasheh u64 olen) 2985416161dbSMark Fasheh { 2986e1d227a4SMark Fasheh u64 len = *plen; 2987416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 2988416161dbSMark Fasheh 2989e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 2990416161dbSMark Fasheh return -EINVAL; 2991e1d227a4SMark Fasheh 2992e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 2993e1d227a4SMark Fasheh if (off + len == inode->i_size) 2994e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 2995e1d227a4SMark Fasheh 2996416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 2997416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 2998416161dbSMark Fasheh return -EINVAL; 2999416161dbSMark Fasheh 3000416161dbSMark Fasheh return 0; 3001416161dbSMark Fasheh } 3002416161dbSMark Fasheh 30033973909dSTimofey Titovets static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 olen, 300467b07bd4STimofey Titovets struct inode *dst, u64 dst_loff, 300567b07bd4STimofey Titovets struct cmp_pages *cmp) 3006416161dbSMark Fasheh { 3007416161dbSMark Fasheh int ret; 3008e1d227a4SMark Fasheh u64 len = olen; 3009fc4badd9SOmar Sandoval bool same_inode = (src == dst); 30100efa9f48SMark Fasheh u64 same_lock_start = 0; 30110efa9f48SMark Fasheh u64 same_lock_len = 0; 3012416161dbSMark Fasheh 30130efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 30140efa9f48SMark Fasheh if (ret) 30153973909dSTimofey Titovets return ret; 3016fc4badd9SOmar Sandoval 3017fc4badd9SOmar Sandoval ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3018f4dfe687SFilipe Manana if (ret) 30193973909dSTimofey Titovets return ret; 30200efa9f48SMark Fasheh 3021fc4badd9SOmar Sandoval if (same_inode) { 30220efa9f48SMark Fasheh /* 30230efa9f48SMark Fasheh * Single inode case wants the same checks, except we 30240efa9f48SMark Fasheh * don't want our length pushed out past i_size as 30250efa9f48SMark Fasheh * comparing that data range makes no sense. 30260efa9f48SMark Fasheh * 30270efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 30280efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 30290efa9f48SMark Fasheh * reject the request. 30300efa9f48SMark Fasheh * 30310efa9f48SMark Fasheh * This effectively means we require aligned extents 30320efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 30330efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 30340efa9f48SMark Fasheh * i_size. 30350efa9f48SMark Fasheh */ 30363973909dSTimofey Titovets if (len != olen) 30373973909dSTimofey Titovets return -EINVAL; 30380efa9f48SMark Fasheh 30390efa9f48SMark Fasheh /* Check for overlapping ranges */ 30403973909dSTimofey Titovets if (dst_loff + len > loff && dst_loff < loff + len) 30413973909dSTimofey Titovets return -EINVAL; 30420efa9f48SMark Fasheh 30430efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 30440efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 30450efa9f48SMark Fasheh } 3046416161dbSMark Fasheh 3047e0bd70c6SFilipe Manana again: 304867b07bd4STimofey Titovets ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, cmp); 3049f4414602SMark Fasheh if (ret) 30503973909dSTimofey Titovets return ret; 3051f4414602SMark Fasheh 30520efa9f48SMark Fasheh if (same_inode) 3053e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3054e0bd70c6SFilipe Manana false); 30550efa9f48SMark Fasheh else 3056e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3057e0bd70c6SFilipe Manana false); 3058e0bd70c6SFilipe Manana /* 3059e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3060e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3061e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3062e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3063e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3064e0bd70c6SFilipe Manana * range locking). 3065e0bd70c6SFilipe Manana */ 3066e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3067e0bd70c6SFilipe Manana /* 3068e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3069e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3070e0bd70c6SFilipe Manana */ 307167b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 3072e0bd70c6SFilipe Manana if (same_inode) { 3073e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3074e0bd70c6SFilipe Manana same_lock_len); 3075e0bd70c6SFilipe Manana } else { 3076e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3077e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3078e0bd70c6SFilipe Manana } 3079e0bd70c6SFilipe Manana goto again; 3080e0bd70c6SFilipe Manana } 3081e0bd70c6SFilipe Manana ASSERT(ret == 0); 3082e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3083e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 308467b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 3085e0bd70c6SFilipe Manana return ret; 3086e0bd70c6SFilipe Manana } 3087f4414602SMark Fasheh 3088207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 308967b07bd4STimofey Titovets ret = btrfs_cmp_data(olen, cmp); 3090416161dbSMark Fasheh if (ret == 0) 30911c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3092416161dbSMark Fasheh 30930efa9f48SMark Fasheh if (same_inode) 30940efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 30950efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 30960efa9f48SMark Fasheh else 3097f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3098f4414602SMark Fasheh 309967b07bd4STimofey Titovets btrfs_cmp_data_free(cmp); 31003973909dSTimofey Titovets 31013973909dSTimofey Titovets return ret; 31023973909dSTimofey Titovets } 31033973909dSTimofey Titovets 3104b6728768STimofey Titovets #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3105b6728768STimofey Titovets 31063973909dSTimofey Titovets static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 31073973909dSTimofey Titovets struct inode *dst, u64 dst_loff) 31083973909dSTimofey Titovets { 31093973909dSTimofey Titovets int ret; 311067b07bd4STimofey Titovets struct cmp_pages cmp; 311167b07bd4STimofey Titovets int num_pages = PAGE_ALIGN(BTRFS_MAX_DEDUPE_LEN) >> PAGE_SHIFT; 31123973909dSTimofey Titovets bool same_inode = (src == dst); 3113b6728768STimofey Titovets u64 i, tail_len, chunk_count; 31143973909dSTimofey Titovets 31153973909dSTimofey Titovets if (olen == 0) 31163973909dSTimofey Titovets return 0; 31173973909dSTimofey Titovets 31183973909dSTimofey Titovets if (same_inode) 31193973909dSTimofey Titovets inode_lock(src); 31203973909dSTimofey Titovets else 31213973909dSTimofey Titovets btrfs_double_inode_lock(src, dst); 31223973909dSTimofey Titovets 31233973909dSTimofey Titovets /* don't make the dst file partly checksummed */ 31243973909dSTimofey Titovets if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 31253973909dSTimofey Titovets (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 31263973909dSTimofey Titovets ret = -EINVAL; 31273973909dSTimofey Titovets goto out_unlock; 31283973909dSTimofey Titovets } 31293973909dSTimofey Titovets 3130b6728768STimofey Titovets tail_len = olen % BTRFS_MAX_DEDUPE_LEN; 3131b6728768STimofey Titovets chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN); 313267b07bd4STimofey Titovets if (chunk_count == 0) 313367b07bd4STimofey Titovets num_pages = PAGE_ALIGN(tail_len) >> PAGE_SHIFT; 313467b07bd4STimofey Titovets 313567b07bd4STimofey Titovets /* 313667b07bd4STimofey Titovets * If deduping ranges in the same inode, locking rules make it 313767b07bd4STimofey Titovets * mandatory to always lock pages in ascending order to avoid deadlocks 313867b07bd4STimofey Titovets * with concurrent tasks (such as starting writeback/delalloc). 313967b07bd4STimofey Titovets */ 314067b07bd4STimofey Titovets if (same_inode && dst_loff < loff) 314167b07bd4STimofey Titovets swap(loff, dst_loff); 314267b07bd4STimofey Titovets 314367b07bd4STimofey Titovets /* 314467b07bd4STimofey Titovets * We must gather up all the pages before we initiate our extent 314567b07bd4STimofey Titovets * locking. We use an array for the page pointers. Size of the array is 314667b07bd4STimofey Titovets * bounded by len, which is in turn bounded by BTRFS_MAX_DEDUPE_LEN. 314767b07bd4STimofey Titovets */ 3148bf5091c8SDavid Sterba cmp.src_pages = kvmalloc_array(num_pages, sizeof(struct page *), 3149bf5091c8SDavid Sterba GFP_KERNEL | __GFP_ZERO); 3150bf5091c8SDavid Sterba cmp.dst_pages = kvmalloc_array(num_pages, sizeof(struct page *), 3151bf5091c8SDavid Sterba GFP_KERNEL | __GFP_ZERO); 315267b07bd4STimofey Titovets if (!cmp.src_pages || !cmp.dst_pages) { 3153bf5091c8SDavid Sterba ret = -ENOMEM; 3154bf5091c8SDavid Sterba goto out_free; 315567b07bd4STimofey Titovets } 3156b6728768STimofey Titovets 3157b6728768STimofey Titovets for (i = 0; i < chunk_count; i++) { 3158b6728768STimofey Titovets ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN, 315967b07bd4STimofey Titovets dst, dst_loff, &cmp); 3160b6728768STimofey Titovets if (ret) 3161b6728768STimofey Titovets goto out_unlock; 3162b6728768STimofey Titovets 3163b6728768STimofey Titovets loff += BTRFS_MAX_DEDUPE_LEN; 3164b6728768STimofey Titovets dst_loff += BTRFS_MAX_DEDUPE_LEN; 3165b6728768STimofey Titovets } 3166b6728768STimofey Titovets 3167b6728768STimofey Titovets if (tail_len > 0) 316867b07bd4STimofey Titovets ret = btrfs_extent_same_range(src, loff, tail_len, dst, 316967b07bd4STimofey Titovets dst_loff, &cmp); 31703973909dSTimofey Titovets 3171416161dbSMark Fasheh out_unlock: 31720efa9f48SMark Fasheh if (same_inode) 31735955102cSAl Viro inode_unlock(src); 31740efa9f48SMark Fasheh else 3175f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3176416161dbSMark Fasheh 3177bf5091c8SDavid Sterba out_free: 3178bf5091c8SDavid Sterba kvfree(cmp.src_pages); 3179bf5091c8SDavid Sterba kvfree(cmp.dst_pages); 318067b07bd4STimofey Titovets 3181416161dbSMark Fasheh return ret; 3182416161dbSMark Fasheh } 3183416161dbSMark Fasheh 31842b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 31852b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3186416161dbSMark Fasheh { 31872b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 31882b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3189416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 31902b3909f8SDarrick J. Wong ssize_t res; 3191416161dbSMark Fasheh 319209cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3193416161dbSMark Fasheh /* 3194416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3195416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3196416161dbSMark Fasheh * this situation without an update. 3197416161dbSMark Fasheh */ 31982b3909f8SDarrick J. Wong return -EINVAL; 3199416161dbSMark Fasheh } 3200416161dbSMark Fasheh 32012b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 32022b3909f8SDarrick J. Wong if (res) 32032b3909f8SDarrick J. Wong return res; 32042b3909f8SDarrick J. Wong return olen; 3205416161dbSMark Fasheh } 3206416161dbSMark Fasheh 3207f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3208f82a9901SFilipe Manana struct inode *inode, 3209f82a9901SFilipe Manana u64 endoff, 3210f82a9901SFilipe Manana const u64 destoff, 32111c919a5eSMark Fasheh const u64 olen, 32121c919a5eSMark Fasheh int no_time_update) 3213f82a9901SFilipe Manana { 3214f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3215f82a9901SFilipe Manana int ret; 3216f82a9901SFilipe Manana 3217f82a9901SFilipe Manana inode_inc_iversion(inode); 32181c919a5eSMark Fasheh if (!no_time_update) 3219c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3220f82a9901SFilipe Manana /* 3221f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3222f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3223f82a9901SFilipe Manana */ 3224f82a9901SFilipe Manana if (endoff > destoff + olen) 3225f82a9901SFilipe Manana endoff = destoff + olen; 3226f82a9901SFilipe Manana if (endoff > inode->i_size) 32276ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3228f82a9901SFilipe Manana 3229f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3230f82a9901SFilipe Manana if (ret) { 323166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 32323a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3233f82a9901SFilipe Manana goto out; 3234f82a9901SFilipe Manana } 32353a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3236f82a9901SFilipe Manana out: 3237f82a9901SFilipe Manana return ret; 3238f82a9901SFilipe Manana } 3239f82a9901SFilipe Manana 3240a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 32417ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 32427ffbb598SFilipe Manana const struct btrfs_path *path, 32437ffbb598SFilipe Manana const u64 hole_offset, 32447ffbb598SFilipe Manana const u64 hole_len) 32457ffbb598SFilipe Manana { 3246a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 32477ffbb598SFilipe Manana struct extent_map *em; 32487ffbb598SFilipe Manana int ret; 32497ffbb598SFilipe Manana 32507ffbb598SFilipe Manana em = alloc_extent_map(); 32517ffbb598SFilipe Manana if (!em) { 3252a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 32537ffbb598SFilipe Manana return; 32547ffbb598SFilipe Manana } 32557ffbb598SFilipe Manana 325614f59796SFilipe Manana if (path) { 325714f59796SFilipe Manana struct btrfs_file_extent_item *fi; 325814f59796SFilipe Manana 325914f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 326014f59796SFilipe Manana struct btrfs_file_extent_item); 3261a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 32627ffbb598SFilipe Manana em->generation = -1; 32637ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 32647ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 32657ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3266a2f392e4SNikolay Borisov &inode->runtime_flags); 32677ffbb598SFilipe Manana } else { 32687ffbb598SFilipe Manana em->start = hole_offset; 32697ffbb598SFilipe Manana em->len = hole_len; 32707ffbb598SFilipe Manana em->ram_bytes = em->len; 32717ffbb598SFilipe Manana em->orig_start = hole_offset; 32727ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 32737ffbb598SFilipe Manana em->block_len = 0; 32747ffbb598SFilipe Manana em->orig_block_len = 0; 32757ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 32767ffbb598SFilipe Manana em->generation = trans->transid; 32777ffbb598SFilipe Manana } 32787ffbb598SFilipe Manana 32797ffbb598SFilipe Manana while (1) { 32807ffbb598SFilipe Manana write_lock(&em_tree->lock); 32817ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 32827ffbb598SFilipe Manana write_unlock(&em_tree->lock); 32837ffbb598SFilipe Manana if (ret != -EEXIST) { 32847ffbb598SFilipe Manana free_extent_map(em); 32857ffbb598SFilipe Manana break; 32867ffbb598SFilipe Manana } 3287a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 32887ffbb598SFilipe Manana em->start + em->len - 1, 0); 32897ffbb598SFilipe Manana } 32907ffbb598SFilipe Manana 3291ee39b432SDavid Sterba if (ret) 3292a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 32937ffbb598SFilipe Manana } 32947ffbb598SFilipe Manana 32958039d87dSFilipe Manana /* 32968039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 32978039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 32988039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 32998039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33008039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33018039d87dSFilipe Manana * 33028039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33038039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33048039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33058039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33068039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33078039d87dSFilipe Manana * i_mutex, which we are holding here). 33088039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33098039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33108039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33118039d87dSFilipe Manana * 33128039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33138039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33148039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33158039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33168039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33178039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33188039d87dSFilipe Manana * the former. 33198039d87dSFilipe Manana */ 33204a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 33218039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33228039d87dSFilipe Manana struct btrfs_path *path, 33238039d87dSFilipe Manana struct btrfs_key *new_key, 33248039d87dSFilipe Manana const u64 drop_start, 33258039d87dSFilipe Manana const u64 datal, 33268039d87dSFilipe Manana const u64 skip, 33278039d87dSFilipe Manana const u64 size, 33288039d87dSFilipe Manana char *inline_data) 33298039d87dSFilipe Manana { 33300b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 33318039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 33328039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 33330b246afaSJeff Mahoney fs_info->sectorsize); 33348039d87dSFilipe Manana int ret; 33358039d87dSFilipe Manana struct btrfs_key key; 33368039d87dSFilipe Manana 33378039d87dSFilipe Manana if (new_key->offset > 0) 33388039d87dSFilipe Manana return -EOPNOTSUPP; 33398039d87dSFilipe Manana 33404a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 33418039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 33428039d87dSFilipe Manana key.offset = 0; 33438039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 33448039d87dSFilipe Manana if (ret < 0) { 33458039d87dSFilipe Manana return ret; 33468039d87dSFilipe Manana } else if (ret > 0) { 33478039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 33488039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 33498039d87dSFilipe Manana if (ret < 0) 33508039d87dSFilipe Manana return ret; 33518039d87dSFilipe Manana else if (ret > 0) 33528039d87dSFilipe Manana goto copy_inline_extent; 33538039d87dSFilipe Manana } 33548039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 33554a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 33568039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 33578039d87dSFilipe Manana ASSERT(key.offset > 0); 33588039d87dSFilipe Manana return -EOPNOTSUPP; 33598039d87dSFilipe Manana } 33608039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 33618039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 33628039d87dSFilipe Manana u64 ext_len; 33638039d87dSFilipe Manana 33648039d87dSFilipe Manana /* 33658039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 33668039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 33678039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 33688039d87dSFilipe Manana */ 33698039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 33708039d87dSFilipe Manana struct btrfs_file_extent_item); 33718039d87dSFilipe Manana /* 33728039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 33738039d87dSFilipe Manana * following it. 33748039d87dSFilipe Manana */ 33758039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 33768039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33778039d87dSFilipe Manana goto copy_inline_extent; 33788039d87dSFilipe Manana 33798039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 33808039d87dSFilipe Manana if (ext_len > aligned_end) 33818039d87dSFilipe Manana return -EOPNOTSUPP; 33828039d87dSFilipe Manana 33838039d87dSFilipe Manana ret = btrfs_next_item(root, path); 33848039d87dSFilipe Manana if (ret < 0) { 33858039d87dSFilipe Manana return ret; 33868039d87dSFilipe Manana } else if (ret == 0) { 33878039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 33888039d87dSFilipe Manana path->slots[0]); 33894a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 33908039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 33918039d87dSFilipe Manana return -EOPNOTSUPP; 33928039d87dSFilipe Manana } 33938039d87dSFilipe Manana } 33948039d87dSFilipe Manana 33958039d87dSFilipe Manana copy_inline_extent: 33968039d87dSFilipe Manana /* 33978039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 33988039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 33998039d87dSFilipe Manana */ 34008039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34018039d87dSFilipe Manana /* 34028039d87dSFilipe Manana * If the destination inode has an inline extent... 34038039d87dSFilipe Manana * This would require copying the data from the source inline 34048039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34058039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34068039d87dSFilipe Manana * or just one of them, which would require decompressing and 34078039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34088039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34098039d87dSFilipe Manana * inline extent). 34108039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34118039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34128039d87dSFilipe Manana */ 34138039d87dSFilipe Manana return -EOPNOTSUPP; 34148039d87dSFilipe Manana } 34158039d87dSFilipe Manana 34168039d87dSFilipe Manana btrfs_release_path(path); 34178039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34188039d87dSFilipe Manana if (ret) 34198039d87dSFilipe Manana return ret; 34208039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34218039d87dSFilipe Manana if (ret) 34228039d87dSFilipe Manana return ret; 34238039d87dSFilipe Manana 34248039d87dSFilipe Manana if (skip) { 34258039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 34268039d87dSFilipe Manana 34278039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 34288039d87dSFilipe Manana } 34298039d87dSFilipe Manana 34308039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 34318039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 34328039d87dSFilipe Manana path->slots[0]), 34338039d87dSFilipe Manana size); 34348039d87dSFilipe Manana inode_add_bytes(dst, datal); 34358039d87dSFilipe Manana 34368039d87dSFilipe Manana return 0; 34378039d87dSFilipe Manana } 34388039d87dSFilipe Manana 343932b7c687SMark Fasheh /** 344032b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 344132b7c687SMark Fasheh * 344232b7c687SMark Fasheh * @src: Inode to clone from 344332b7c687SMark Fasheh * @inode: Inode to clone to 344432b7c687SMark Fasheh * @off: Offset within source to start clone from 344532b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 34461c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 344732b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 34481c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 344932b7c687SMark Fasheh */ 345032b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3451f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 34521c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3453f46b5a66SChristoph Hellwig { 34540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3455f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 345632b7c687SMark Fasheh struct btrfs_path *path = NULL; 3457f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 345832b7c687SMark Fasheh struct btrfs_trans_handle *trans; 345932b7c687SMark Fasheh char *buf = NULL; 3460ae01a0abSYan Zheng struct btrfs_key key; 3461f46b5a66SChristoph Hellwig u32 nritems; 3462f46b5a66SChristoph Hellwig int slot; 3463ae01a0abSYan Zheng int ret; 3464f82a9901SFilipe Manana const u64 len = olen_aligned; 3465f82a9901SFilipe Manana u64 last_dest_end = destoff; 3466ae01a0abSYan Zheng 3467ae01a0abSYan Zheng ret = -ENOMEM; 3468752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3469ae01a0abSYan Zheng if (!buf) 347032b7c687SMark Fasheh return ret; 3471ae01a0abSYan Zheng 3472ae01a0abSYan Zheng path = btrfs_alloc_path(); 3473ae01a0abSYan Zheng if (!path) { 347415351955SDavid Sterba kvfree(buf); 347532b7c687SMark Fasheh return ret; 3476ae01a0abSYan Zheng } 347732b7c687SMark Fasheh 3478e4058b54SDavid Sterba path->reada = READA_FORWARD; 3479c5c9cd4dSSage Weil /* clone data */ 34804a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3481ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 34822c463823SFilipe Manana key.offset = off; 3483f46b5a66SChristoph Hellwig 3484f46b5a66SChristoph Hellwig while (1) { 3485de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3486df858e76SFilipe Manana 3487f46b5a66SChristoph Hellwig /* 3488f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3489f46b5a66SChristoph Hellwig * tree. 3490f46b5a66SChristoph Hellwig */ 3491e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3492362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3493362a20c5SDavid Sterba 0, 0); 3494f46b5a66SChristoph Hellwig if (ret < 0) 3495f46b5a66SChristoph Hellwig goto out; 34962c463823SFilipe Manana /* 34972c463823SFilipe Manana * First search, if no extent item that starts at offset off was 34982c463823SFilipe Manana * found but the previous item is an extent item, it's possible 34992c463823SFilipe Manana * it might overlap our target range, therefore process it. 35002c463823SFilipe Manana */ 35012c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35022c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35032c463823SFilipe Manana path->slots[0] - 1); 35042c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35052c463823SFilipe Manana path->slots[0]--; 35062c463823SFilipe Manana } 3507f46b5a66SChristoph Hellwig 3508ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3509e4355f34SFilipe David Borba Manana process_slot: 3510ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3511362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3512f46b5a66SChristoph Hellwig if (ret < 0) 3513f46b5a66SChristoph Hellwig goto out; 3514f46b5a66SChristoph Hellwig if (ret > 0) 3515f46b5a66SChristoph Hellwig break; 3516ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3517f46b5a66SChristoph Hellwig } 3518f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3519f46b5a66SChristoph Hellwig slot = path->slots[0]; 3520f46b5a66SChristoph Hellwig 3521ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3522962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 35234a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3524f46b5a66SChristoph Hellwig break; 3525f46b5a66SChristoph Hellwig 3526962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3527c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3528c5c9cd4dSSage Weil int type; 352931840ae1SZheng Yan u32 size; 353031840ae1SZheng Yan struct btrfs_key new_key; 3531c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3532c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3533c5c9cd4dSSage Weil u8 comp; 3534f82a9901SFilipe Manana u64 drop_start; 353531840ae1SZheng Yan 3536c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3537c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3538c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3539c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3540c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3541c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3542d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3543d397712bSChris Mason extent); 3544d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3545d397712bSChris Mason extent); 3546c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3547d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3548d397712bSChris Mason extent); 3549c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3550c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3551c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3552c5c9cd4dSSage Weil extent); 3553c5c9cd4dSSage Weil } 355431840ae1SZheng Yan 35552c463823SFilipe Manana /* 35562c463823SFilipe Manana * The first search might have left us at an extent 35572c463823SFilipe Manana * item that ends before our target range's start, can 35582c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 35592c463823SFilipe Manana */ 35602c463823SFilipe Manana if (key.offset + datal <= off) { 3561e4355f34SFilipe David Borba Manana path->slots[0]++; 3562e4355f34SFilipe David Borba Manana goto process_slot; 35632c463823SFilipe Manana } else if (key.offset >= off + len) { 35642c463823SFilipe Manana break; 3565e4355f34SFilipe David Borba Manana } 3566df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3567e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3568e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3569e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3570e4355f34SFilipe David Borba Manana size); 3571e4355f34SFilipe David Borba Manana 3572e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3573e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3574c5c9cd4dSSage Weil 357531840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 35764a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 35774d728ec7SLi Zefan if (off <= key.offset) 3578c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 35794d728ec7SLi Zefan else 35804d728ec7SLi Zefan new_key.offset = destoff; 3581c5c9cd4dSSage Weil 3582b6f3409bSSage Weil /* 3583f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3584f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3585f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3586f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3587f82a9901SFilipe Manana * partially overlaps it). 3588f82a9901SFilipe Manana */ 3589f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3590f82a9901SFilipe Manana drop_start = last_dest_end; 3591f82a9901SFilipe Manana else 3592f82a9901SFilipe Manana drop_start = new_key.offset; 3593f82a9901SFilipe Manana 3594f82a9901SFilipe Manana /* 3595b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3596b6f3409bSSage Weil * 1 - add new extent 3597b6f3409bSSage Weil * 1 - inode update 3598b6f3409bSSage Weil */ 3599b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3600a22285a6SYan, Zheng if (IS_ERR(trans)) { 3601a22285a6SYan, Zheng ret = PTR_ERR(trans); 3602a22285a6SYan, Zheng goto out; 3603a22285a6SYan, Zheng } 3604a22285a6SYan, Zheng 3605c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3606c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3607d72c0842SLi Zefan /* 3608d72c0842SLi Zefan * a | --- range to clone ---| b 3609d72c0842SLi Zefan * | ------------- extent ------------- | 3610d72c0842SLi Zefan */ 3611d72c0842SLi Zefan 361293915584SAntonio Ospite /* subtract range b */ 3613d72c0842SLi Zefan if (key.offset + datal > off + len) 3614d72c0842SLi Zefan datal = off + len - key.offset; 3615d72c0842SLi Zefan 361693915584SAntonio Ospite /* subtract range a */ 3617a22285a6SYan, Zheng if (off > key.offset) { 3618a22285a6SYan, Zheng datao += off - key.offset; 3619a22285a6SYan, Zheng datal -= off - key.offset; 3620a22285a6SYan, Zheng } 3621a22285a6SYan, Zheng 36225dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3623f82a9901SFilipe Manana drop_start, 3624a22285a6SYan, Zheng new_key.offset + datal, 36252671485dSJosef Bacik 1); 362679787eaaSJeff Mahoney if (ret) { 36273f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 362800fdf13aSLiu Bo btrfs_abort_transaction(trans, 362966642832SJeff Mahoney ret); 36303a45bb20SJeff Mahoney btrfs_end_transaction(trans); 363179787eaaSJeff Mahoney goto out; 363279787eaaSJeff Mahoney } 3633a22285a6SYan, Zheng 363431840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 363531840ae1SZheng Yan &new_key, size); 363679787eaaSJeff Mahoney if (ret) { 363766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 36383a45bb20SJeff Mahoney btrfs_end_transaction(trans); 363979787eaaSJeff Mahoney goto out; 364079787eaaSJeff Mahoney } 364131840ae1SZheng Yan 364231840ae1SZheng Yan leaf = path->nodes[0]; 364331840ae1SZheng Yan slot = path->slots[0]; 364431840ae1SZheng Yan write_extent_buffer(leaf, buf, 364531840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 364631840ae1SZheng Yan size); 3647ae01a0abSYan Zheng 3648f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3649f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3650c5c9cd4dSSage Weil 3651c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3652c5c9cd4dSSage Weil if (!disko) 3653c5c9cd4dSSage Weil datao = 0; 3654c5c9cd4dSSage Weil 3655c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3656c5c9cd4dSSage Weil datao); 3657c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3658c5c9cd4dSSage Weil datal); 3659fcebe456SJosef Bacik 3660c5c9cd4dSSage Weil if (disko) { 3661c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 36622ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 366384f7d8e6SJosef Bacik root, 36645d4f98a2SYan Zheng disko, diskl, 0, 3665f46b5a66SChristoph Hellwig root->root_key.objectid, 36664a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 3667b06c4bf5SFilipe Manana new_key.offset - datao); 366879787eaaSJeff Mahoney if (ret) { 366979787eaaSJeff Mahoney btrfs_abort_transaction(trans, 367079787eaaSJeff Mahoney ret); 36713a45bb20SJeff Mahoney btrfs_end_transaction(trans); 367279787eaaSJeff Mahoney goto out; 367379787eaaSJeff Mahoney 367479787eaaSJeff Mahoney } 3675f46b5a66SChristoph Hellwig } 3676c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3677c5c9cd4dSSage Weil u64 skip = 0; 3678c5c9cd4dSSage Weil u64 trim = 0; 3679ed958762SFilipe Manana 3680c5c9cd4dSSage Weil if (off > key.offset) { 3681c5c9cd4dSSage Weil skip = off - key.offset; 3682c5c9cd4dSSage Weil new_key.offset += skip; 368331840ae1SZheng Yan } 3684d397712bSChris Mason 3685c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3686c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3687d397712bSChris Mason 3688c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3689c5c9cd4dSSage Weil ret = -EINVAL; 36903a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3691c5c9cd4dSSage Weil goto out; 369231840ae1SZheng Yan } 3693c5c9cd4dSSage Weil size -= skip + trim; 3694c5c9cd4dSSage Weil datal -= skip + trim; 3695a22285a6SYan, Zheng 36964a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 36978039d87dSFilipe Manana trans, path, 36988039d87dSFilipe Manana &new_key, 3699f82a9901SFilipe Manana drop_start, 37008039d87dSFilipe Manana datal, 37018039d87dSFilipe Manana skip, size, buf); 370279787eaaSJeff Mahoney if (ret) { 37033f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37043a29bc09SChris Mason btrfs_abort_transaction(trans, 370579787eaaSJeff Mahoney ret); 37063a45bb20SJeff Mahoney btrfs_end_transaction(trans); 370779787eaaSJeff Mahoney goto out; 370879787eaaSJeff Mahoney } 3709c5c9cd4dSSage Weil leaf = path->nodes[0]; 3710c5c9cd4dSSage Weil slot = path->slots[0]; 3711c5c9cd4dSSage Weil } 3712c5c9cd4dSSage Weil 37137ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 37147ffbb598SFilipe Manana if (drop_start < new_key.offset) 3715a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 371614f59796SFilipe Manana NULL, drop_start, 37177ffbb598SFilipe Manana new_key.offset - drop_start); 37187ffbb598SFilipe Manana 3719a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 3720a2f392e4SNikolay Borisov path, 0, 0); 37217ffbb598SFilipe Manana 3722c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3723b3b4aa74SDavid Sterba btrfs_release_path(path); 3724c5c9cd4dSSage Weil 372562e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 37260b246afaSJeff Mahoney fs_info->sectorsize); 3727f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3728f82a9901SFilipe Manana last_dest_end, 37291c919a5eSMark Fasheh destoff, olen, 37301c919a5eSMark Fasheh no_time_update); 3731f82a9901SFilipe Manana if (ret) 373279787eaaSJeff Mahoney goto out; 37332c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 37342c463823SFilipe Manana break; 3735a22285a6SYan, Zheng } 3736b3b4aa74SDavid Sterba btrfs_release_path(path); 3737df858e76SFilipe Manana key.offset = next_key_min_offset; 373869ae5e44SWang Xiaoguang 373969ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 374069ae5e44SWang Xiaoguang ret = -EINTR; 374169ae5e44SWang Xiaoguang goto out; 374269ae5e44SWang Xiaoguang } 3743ae01a0abSYan Zheng } 3744f46b5a66SChristoph Hellwig ret = 0; 374532b7c687SMark Fasheh 3746f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3747f82a9901SFilipe Manana /* 3748f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3749f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3750f82a9901SFilipe Manana */ 3751b3b4aa74SDavid Sterba btrfs_release_path(path); 3752f82a9901SFilipe Manana 3753f82a9901SFilipe Manana /* 3754f82a9901SFilipe Manana * 1 - remove extent(s) 3755f82a9901SFilipe Manana * 1 - inode update 3756f82a9901SFilipe Manana */ 3757f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3758f82a9901SFilipe Manana if (IS_ERR(trans)) { 3759f82a9901SFilipe Manana ret = PTR_ERR(trans); 3760f82a9901SFilipe Manana goto out; 3761f82a9901SFilipe Manana } 3762f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3763f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3764f82a9901SFilipe Manana if (ret) { 3765f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 376666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 37673a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3768f82a9901SFilipe Manana goto out; 3769f82a9901SFilipe Manana } 3770a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 3771a2f392e4SNikolay Borisov last_dest_end, 377214f59796SFilipe Manana destoff + len - last_dest_end); 3773f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 37741c919a5eSMark Fasheh destoff, olen, no_time_update); 3775f82a9901SFilipe Manana } 3776f82a9901SFilipe Manana 3777f46b5a66SChristoph Hellwig out: 377832b7c687SMark Fasheh btrfs_free_path(path); 377915351955SDavid Sterba kvfree(buf); 378032b7c687SMark Fasheh return ret; 378132b7c687SMark Fasheh } 378232b7c687SMark Fasheh 37833db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 378432b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 378532b7c687SMark Fasheh { 378654563d41SAl Viro struct inode *inode = file_inode(file); 37873db11b2eSZach Brown struct inode *src = file_inode(file_src); 37880b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 378932b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 379032b7c687SMark Fasheh int ret; 379132b7c687SMark Fasheh u64 len = olen; 37920b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 37933db11b2eSZach Brown int same_inode = src == inode; 379432b7c687SMark Fasheh 379532b7c687SMark Fasheh /* 379632b7c687SMark Fasheh * TODO: 379732b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 379832b7c687SMark Fasheh * decompress into destination's address_space (the file offset 379932b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 380032b7c687SMark Fasheh * otherwise reinsert) a subrange. 380100fdf13aSLiu Bo * 380200fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 380300fdf13aSLiu Bo * be either compressed or non-compressed. 380432b7c687SMark Fasheh */ 380532b7c687SMark Fasheh 380632b7c687SMark Fasheh if (btrfs_root_readonly(root)) 380732b7c687SMark Fasheh return -EROFS; 380832b7c687SMark Fasheh 38093db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 38103db11b2eSZach Brown src->i_sb != inode->i_sb) 38113db11b2eSZach Brown return -EXDEV; 381232b7c687SMark Fasheh 381332b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 38143db11b2eSZach Brown return -EISDIR; 381532b7c687SMark Fasheh 381632b7c687SMark Fasheh if (!same_inode) { 3817293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 381832b7c687SMark Fasheh } else { 38195955102cSAl Viro inode_lock(src); 382032b7c687SMark Fasheh } 382132b7c687SMark Fasheh 3822b5c40d59SOmar Sandoval /* don't make the dst file partly checksummed */ 3823b5c40d59SOmar Sandoval if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3824b5c40d59SOmar Sandoval (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) { 3825b5c40d59SOmar Sandoval ret = -EINVAL; 3826b5c40d59SOmar Sandoval goto out_unlock; 3827b5c40d59SOmar Sandoval } 3828b5c40d59SOmar Sandoval 382932b7c687SMark Fasheh /* determine range to clone */ 383032b7c687SMark Fasheh ret = -EINVAL; 383132b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 383232b7c687SMark Fasheh goto out_unlock; 383332b7c687SMark Fasheh if (len == 0) 383432b7c687SMark Fasheh olen = len = src->i_size - off; 383532b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 383632b7c687SMark Fasheh if (off + len == src->i_size) 383732b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 383832b7c687SMark Fasheh 3839ccccf3d6SFilipe Manana if (len == 0) { 3840ccccf3d6SFilipe Manana ret = 0; 3841ccccf3d6SFilipe Manana goto out_unlock; 3842ccccf3d6SFilipe Manana } 3843ccccf3d6SFilipe Manana 384432b7c687SMark Fasheh /* verify the end result is block aligned */ 384532b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 384632b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 384732b7c687SMark Fasheh goto out_unlock; 384832b7c687SMark Fasheh 384932b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 385032b7c687SMark Fasheh if (same_inode) { 385132b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 385232b7c687SMark Fasheh goto out_unlock; 385332b7c687SMark Fasheh } 385432b7c687SMark Fasheh 385532b7c687SMark Fasheh if (destoff > inode->i_size) { 385632b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 385732b7c687SMark Fasheh if (ret) 385832b7c687SMark Fasheh goto out_unlock; 385932b7c687SMark Fasheh } 386032b7c687SMark Fasheh 3861c125b8bfSFilipe Manana /* 3862c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3863c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3864c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3865c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3866c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3867c125b8bfSFilipe Manana */ 3868c125b8bfSFilipe Manana if (same_inode) { 3869c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3870c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 387132b7c687SMark Fasheh 3872e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 3873c125b8bfSFilipe Manana } else { 3874e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 3875e0bd70c6SFilipe Manana true); 3876e0bd70c6SFilipe Manana } 3877e0bd70c6SFilipe Manana ASSERT(ret == 0); 3878e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3879e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3880e0bd70c6SFilipe Manana goto out_unlock; 3881c125b8bfSFilipe Manana } 388232b7c687SMark Fasheh 38831c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 388432b7c687SMark Fasheh 3885c125b8bfSFilipe Manana if (same_inode) { 3886c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3887c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3888c125b8bfSFilipe Manana 3889c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3890c125b8bfSFilipe Manana } else { 3891293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 3892c125b8bfSFilipe Manana } 3893c125b8bfSFilipe Manana /* 3894c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3895c125b8bfSFilipe Manana * data immediately and not the previous data. 3896c125b8bfSFilipe Manana */ 389765bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 389809cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 389909cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 3900f46b5a66SChristoph Hellwig out_unlock: 3901293a8489SMark Fasheh if (!same_inode) 3902293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 3903293a8489SMark Fasheh else 39045955102cSAl Viro inode_unlock(src); 39053db11b2eSZach Brown return ret; 39063db11b2eSZach Brown } 39073db11b2eSZach Brown 390804b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 390904b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 39103db11b2eSZach Brown { 391104b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 3912c5c9cd4dSSage Weil } 3913c5c9cd4dSSage Weil 39146ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 39156ef5ed0dSJosef Bacik { 3916496ad9aaSAl Viro struct inode *inode = file_inode(file); 39170b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 39186ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 39196ef5ed0dSJosef Bacik struct btrfs_root *new_root; 39206ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 39216ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 39226ef5ed0dSJosef Bacik struct btrfs_path *path; 39236ef5ed0dSJosef Bacik struct btrfs_key location; 39246ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 39256ef5ed0dSJosef Bacik u64 objectid = 0; 39266ef5ed0dSJosef Bacik u64 dir_id; 39273c04ce01SMiao Xie int ret; 39286ef5ed0dSJosef Bacik 39296ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 39306ef5ed0dSJosef Bacik return -EPERM; 39316ef5ed0dSJosef Bacik 39323c04ce01SMiao Xie ret = mnt_want_write_file(file); 39333c04ce01SMiao Xie if (ret) 39343c04ce01SMiao Xie return ret; 39353c04ce01SMiao Xie 39363c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 39373c04ce01SMiao Xie ret = -EFAULT; 39383c04ce01SMiao Xie goto out; 39393c04ce01SMiao Xie } 39406ef5ed0dSJosef Bacik 39416ef5ed0dSJosef Bacik if (!objectid) 39421cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 39436ef5ed0dSJosef Bacik 39446ef5ed0dSJosef Bacik location.objectid = objectid; 39456ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 39466ef5ed0dSJosef Bacik location.offset = (u64)-1; 39476ef5ed0dSJosef Bacik 39480b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 39493c04ce01SMiao Xie if (IS_ERR(new_root)) { 39503c04ce01SMiao Xie ret = PTR_ERR(new_root); 39513c04ce01SMiao Xie goto out; 39523c04ce01SMiao Xie } 39536d6d2829Ssatoru takeuchi if (!is_fstree(new_root->objectid)) { 39546d6d2829Ssatoru takeuchi ret = -ENOENT; 39556d6d2829Ssatoru takeuchi goto out; 39566d6d2829Ssatoru takeuchi } 39576ef5ed0dSJosef Bacik 39586ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 39593c04ce01SMiao Xie if (!path) { 39603c04ce01SMiao Xie ret = -ENOMEM; 39613c04ce01SMiao Xie goto out; 39623c04ce01SMiao Xie } 39636ef5ed0dSJosef Bacik path->leave_spinning = 1; 39646ef5ed0dSJosef Bacik 39656ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 396698d5dc13STsutomu Itoh if (IS_ERR(trans)) { 39676ef5ed0dSJosef Bacik btrfs_free_path(path); 39683c04ce01SMiao Xie ret = PTR_ERR(trans); 39693c04ce01SMiao Xie goto out; 39706ef5ed0dSJosef Bacik } 39716ef5ed0dSJosef Bacik 39720b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 39730b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 39746ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 3975cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 39766ef5ed0dSJosef Bacik btrfs_free_path(path); 39773a45bb20SJeff Mahoney btrfs_end_transaction(trans); 39780b246afaSJeff Mahoney btrfs_err(fs_info, 39795d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 39803c04ce01SMiao Xie ret = -ENOENT; 39813c04ce01SMiao Xie goto out; 39826ef5ed0dSJosef Bacik } 39836ef5ed0dSJosef Bacik 39846ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 39856ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 39866ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 39876ef5ed0dSJosef Bacik btrfs_free_path(path); 39886ef5ed0dSJosef Bacik 39890b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 39903a45bb20SJeff Mahoney btrfs_end_transaction(trans); 39913c04ce01SMiao Xie out: 39923c04ce01SMiao Xie mnt_drop_write_file(file); 39933c04ce01SMiao Xie return ret; 39946ef5ed0dSJosef Bacik } 39956ef5ed0dSJosef Bacik 3996c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 3997bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 3998bf5fc093SJosef Bacik { 3999bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4000bf5fc093SJosef Bacik 4001bf5fc093SJosef Bacik space->total_bytes = 0; 4002bf5fc093SJosef Bacik space->used_bytes = 0; 4003bf5fc093SJosef Bacik space->flags = 0; 4004bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4005bf5fc093SJosef Bacik space->flags = block_group->flags; 4006bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4007bf5fc093SJosef Bacik space->used_bytes += 4008bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4009bf5fc093SJosef Bacik } 4010bf5fc093SJosef Bacik } 4011bf5fc093SJosef Bacik 40122ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 40132ff7e61eSJeff Mahoney void __user *arg) 40141406e432SJosef Bacik { 40151406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 40161406e432SJosef Bacik struct btrfs_ioctl_space_info space; 40171406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 40187fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 401913f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 40201406e432SJosef Bacik struct btrfs_space_info *info; 4021315d8e98SColin Ian King static const u64 types[] = { 4022315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4023bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4024bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4025315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4026315d8e98SColin Ian King }; 4027bf5fc093SJosef Bacik int num_types = 4; 40287fde62bfSChris Mason int alloc_size; 40291406e432SJosef Bacik int ret = 0; 403051788b1bSDan Rosenberg u64 slot_count = 0; 4031bf5fc093SJosef Bacik int i, c; 40321406e432SJosef Bacik 40331406e432SJosef Bacik if (copy_from_user(&space_args, 40341406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 40351406e432SJosef Bacik sizeof(space_args))) 40361406e432SJosef Bacik return -EFAULT; 40371406e432SJosef Bacik 4038bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4039bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4040bf5fc093SJosef Bacik 4041bf5fc093SJosef Bacik info = NULL; 40427fde62bfSChris Mason rcu_read_lock(); 40430b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4044bf5fc093SJosef Bacik list) { 4045bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4046bf5fc093SJosef Bacik info = tmp; 4047bf5fc093SJosef Bacik break; 4048bf5fc093SJosef Bacik } 4049bf5fc093SJosef Bacik } 40507fde62bfSChris Mason rcu_read_unlock(); 40511406e432SJosef Bacik 4052bf5fc093SJosef Bacik if (!info) 4053bf5fc093SJosef Bacik continue; 4054bf5fc093SJosef Bacik 4055bf5fc093SJosef Bacik down_read(&info->groups_sem); 4056bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4057bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4058bf5fc093SJosef Bacik slot_count++; 4059bf5fc093SJosef Bacik } 4060bf5fc093SJosef Bacik up_read(&info->groups_sem); 4061bf5fc093SJosef Bacik } 4062bf5fc093SJosef Bacik 406336523e95SDavid Sterba /* 406436523e95SDavid Sterba * Global block reserve, exported as a space_info 406536523e95SDavid Sterba */ 406636523e95SDavid Sterba slot_count++; 406736523e95SDavid Sterba 40687fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 40697fde62bfSChris Mason if (space_args.space_slots == 0) { 40707fde62bfSChris Mason space_args.total_spaces = slot_count; 40717fde62bfSChris Mason goto out; 40727fde62bfSChris Mason } 4073bf5fc093SJosef Bacik 407451788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4075bf5fc093SJosef Bacik 40767fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4077bf5fc093SJosef Bacik 40787fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 40797fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 40807fde62bfSChris Mason */ 408109cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 40827fde62bfSChris Mason return -ENOMEM; 40837fde62bfSChris Mason 40847fde62bfSChris Mason space_args.total_spaces = 0; 40858d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 40867fde62bfSChris Mason if (!dest) 40877fde62bfSChris Mason return -ENOMEM; 40887fde62bfSChris Mason dest_orig = dest; 40897fde62bfSChris Mason 40907fde62bfSChris Mason /* now we have a buffer to copy into */ 4091bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4092bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4093bf5fc093SJosef Bacik 409451788b1bSDan Rosenberg if (!slot_count) 409551788b1bSDan Rosenberg break; 409651788b1bSDan Rosenberg 4097bf5fc093SJosef Bacik info = NULL; 40981406e432SJosef Bacik rcu_read_lock(); 40990b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4100bf5fc093SJosef Bacik list) { 4101bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4102bf5fc093SJosef Bacik info = tmp; 41037fde62bfSChris Mason break; 4104bf5fc093SJosef Bacik } 4105bf5fc093SJosef Bacik } 4106bf5fc093SJosef Bacik rcu_read_unlock(); 41077fde62bfSChris Mason 4108bf5fc093SJosef Bacik if (!info) 4109bf5fc093SJosef Bacik continue; 4110bf5fc093SJosef Bacik down_read(&info->groups_sem); 4111bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4112bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 4113c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 4114c065f5b1SSu Yue &space); 41157fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 41161406e432SJosef Bacik dest++; 41171406e432SJosef Bacik space_args.total_spaces++; 411851788b1bSDan Rosenberg slot_count--; 41191406e432SJosef Bacik } 412051788b1bSDan Rosenberg if (!slot_count) 412151788b1bSDan Rosenberg break; 4122bf5fc093SJosef Bacik } 4123bf5fc093SJosef Bacik up_read(&info->groups_sem); 4124bf5fc093SJosef Bacik } 41251406e432SJosef Bacik 412636523e95SDavid Sterba /* 412736523e95SDavid Sterba * Add global block reserve 412836523e95SDavid Sterba */ 412936523e95SDavid Sterba if (slot_count) { 41300b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 413136523e95SDavid Sterba 413236523e95SDavid Sterba spin_lock(&block_rsv->lock); 413336523e95SDavid Sterba space.total_bytes = block_rsv->size; 413436523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 413536523e95SDavid Sterba spin_unlock(&block_rsv->lock); 413636523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 413736523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 413836523e95SDavid Sterba space_args.total_spaces++; 413936523e95SDavid Sterba } 414036523e95SDavid Sterba 41412eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 41427fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 41437fde62bfSChris Mason 41447fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 41457fde62bfSChris Mason ret = -EFAULT; 41467fde62bfSChris Mason 41477fde62bfSChris Mason kfree(dest_orig); 41487fde62bfSChris Mason out: 41497fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 41501406e432SJosef Bacik ret = -EFAULT; 41511406e432SJosef Bacik 41521406e432SJosef Bacik return ret; 41531406e432SJosef Bacik } 41541406e432SJosef Bacik 41559a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 41569a8c28beSMiao Xie void __user *argp) 415746204592SSage Weil { 415846204592SSage Weil struct btrfs_trans_handle *trans; 415946204592SSage Weil u64 transid; 4160db5b493aSTsutomu Itoh int ret; 416146204592SSage Weil 4162d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4163ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4164ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 416598d5dc13STsutomu Itoh return PTR_ERR(trans); 4166ff7c1d33SMiao Xie 4167ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4168ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4169ff7c1d33SMiao Xie goto out; 4170ff7c1d33SMiao Xie } 417146204592SSage Weil transid = trans->transid; 41723a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 41738b2b2d3cSTsutomu Itoh if (ret) { 41743a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4175db5b493aSTsutomu Itoh return ret; 41768b2b2d3cSTsutomu Itoh } 4177ff7c1d33SMiao Xie out: 417846204592SSage Weil if (argp) 417946204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 418046204592SSage Weil return -EFAULT; 418146204592SSage Weil return 0; 418246204592SSage Weil } 418346204592SSage Weil 41842ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 41859a8c28beSMiao Xie void __user *argp) 418646204592SSage Weil { 418746204592SSage Weil u64 transid; 418846204592SSage Weil 418946204592SSage Weil if (argp) { 419046204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 419146204592SSage Weil return -EFAULT; 419246204592SSage Weil } else { 419346204592SSage Weil transid = 0; /* current trans */ 419446204592SSage Weil } 41952ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 419646204592SSage Weil } 419746204592SSage Weil 4198b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4199475f6387SJan Schmidt { 42000b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4201475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4202b8e95489SMiao Xie int ret; 4203475f6387SJan Schmidt 4204475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4205475f6387SJan Schmidt return -EPERM; 4206475f6387SJan Schmidt 4207475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4208475f6387SJan Schmidt if (IS_ERR(sa)) 4209475f6387SJan Schmidt return PTR_ERR(sa); 4210475f6387SJan Schmidt 4211b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4212b8e95489SMiao Xie ret = mnt_want_write_file(file); 4213b8e95489SMiao Xie if (ret) 4214b8e95489SMiao Xie goto out; 4215b8e95489SMiao Xie } 4216b8e95489SMiao Xie 42170b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 421863a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 421963a212abSStefan Behrens 0); 4220475f6387SJan Schmidt 4221475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4222475f6387SJan Schmidt ret = -EFAULT; 4223475f6387SJan Schmidt 4224b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4225b8e95489SMiao Xie mnt_drop_write_file(file); 4226b8e95489SMiao Xie out: 4227475f6387SJan Schmidt kfree(sa); 4228475f6387SJan Schmidt return ret; 4229475f6387SJan Schmidt } 4230475f6387SJan Schmidt 42312ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4232475f6387SJan Schmidt { 4233475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4234475f6387SJan Schmidt return -EPERM; 4235475f6387SJan Schmidt 42362ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4237475f6387SJan Schmidt } 4238475f6387SJan Schmidt 42392ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4240475f6387SJan Schmidt void __user *arg) 4241475f6387SJan Schmidt { 4242475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4243475f6387SJan Schmidt int ret; 4244475f6387SJan Schmidt 4245475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4246475f6387SJan Schmidt return -EPERM; 4247475f6387SJan Schmidt 4248475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4249475f6387SJan Schmidt if (IS_ERR(sa)) 4250475f6387SJan Schmidt return PTR_ERR(sa); 4251475f6387SJan Schmidt 42522ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4253475f6387SJan Schmidt 4254475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4255475f6387SJan Schmidt ret = -EFAULT; 4256475f6387SJan Schmidt 4257475f6387SJan Schmidt kfree(sa); 4258475f6387SJan Schmidt return ret; 4259475f6387SJan Schmidt } 4260475f6387SJan Schmidt 42612ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4262b27f7c0cSDavid Sterba void __user *arg) 4263c11d2c23SStefan Behrens { 4264c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4265c11d2c23SStefan Behrens int ret; 4266c11d2c23SStefan Behrens 4267c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4268c11d2c23SStefan Behrens if (IS_ERR(sa)) 4269c11d2c23SStefan Behrens return PTR_ERR(sa); 4270c11d2c23SStefan Behrens 4271b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4272b27f7c0cSDavid Sterba kfree(sa); 4273b27f7c0cSDavid Sterba return -EPERM; 4274b27f7c0cSDavid Sterba } 4275b27f7c0cSDavid Sterba 42762ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4277c11d2c23SStefan Behrens 4278c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4279c11d2c23SStefan Behrens ret = -EFAULT; 4280c11d2c23SStefan Behrens 4281c11d2c23SStefan Behrens kfree(sa); 4282c11d2c23SStefan Behrens return ret; 4283c11d2c23SStefan Behrens } 4284c11d2c23SStefan Behrens 42852ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 42862ff7e61eSJeff Mahoney void __user *arg) 42873f6bcfbdSStefan Behrens { 42883f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 42893f6bcfbdSStefan Behrens int ret; 42903f6bcfbdSStefan Behrens 42913f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 42923f6bcfbdSStefan Behrens return -EPERM; 42933f6bcfbdSStefan Behrens 42943f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 42953f6bcfbdSStefan Behrens if (IS_ERR(p)) 42963f6bcfbdSStefan Behrens return PTR_ERR(p); 42973f6bcfbdSStefan Behrens 42983f6bcfbdSStefan Behrens switch (p->cmd) { 42993f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4300bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4301adfa97cbSIlya Dryomov ret = -EROFS; 4302adfa97cbSIlya Dryomov goto out; 4303adfa97cbSIlya Dryomov } 4304171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4305e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 43063f6bcfbdSStefan Behrens } else { 43072ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4308171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 43093f6bcfbdSStefan Behrens } 43103f6bcfbdSStefan Behrens break; 43113f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 43120b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 43133f6bcfbdSStefan Behrens ret = 0; 43143f6bcfbdSStefan Behrens break; 43153f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 431617d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 431797282031SAnand Jain ret = 0; 43183f6bcfbdSStefan Behrens break; 43193f6bcfbdSStefan Behrens default: 43203f6bcfbdSStefan Behrens ret = -EINVAL; 43213f6bcfbdSStefan Behrens break; 43223f6bcfbdSStefan Behrens } 43233f6bcfbdSStefan Behrens 43243f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 43253f6bcfbdSStefan Behrens ret = -EFAULT; 4326adfa97cbSIlya Dryomov out: 43273f6bcfbdSStefan Behrens kfree(p); 43283f6bcfbdSStefan Behrens return ret; 43293f6bcfbdSStefan Behrens } 43303f6bcfbdSStefan Behrens 4331d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4332d7728c96SJan Schmidt { 4333d7728c96SJan Schmidt int ret = 0; 4334d7728c96SJan Schmidt int i; 4335740c3d22SChris Mason u64 rel_ptr; 4336d7728c96SJan Schmidt int size; 4337806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4338d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4339d7728c96SJan Schmidt struct btrfs_path *path; 4340d7728c96SJan Schmidt 434182b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4342d7728c96SJan Schmidt return -EPERM; 4343d7728c96SJan Schmidt 4344d7728c96SJan Schmidt path = btrfs_alloc_path(); 4345d7728c96SJan Schmidt if (!path) { 4346d7728c96SJan Schmidt ret = -ENOMEM; 4347d7728c96SJan Schmidt goto out; 4348d7728c96SJan Schmidt } 4349d7728c96SJan Schmidt 4350d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4351d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4352d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4353d7728c96SJan Schmidt ipa = NULL; 4354d7728c96SJan Schmidt goto out; 4355d7728c96SJan Schmidt } 4356d7728c96SJan Schmidt 4357d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4358d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4359d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4360d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4361d7728c96SJan Schmidt ipath = NULL; 4362d7728c96SJan Schmidt goto out; 4363d7728c96SJan Schmidt } 4364d7728c96SJan Schmidt 4365d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4366d7728c96SJan Schmidt if (ret < 0) 4367d7728c96SJan Schmidt goto out; 4368d7728c96SJan Schmidt 4369d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4370745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4371745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4372740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4373d7728c96SJan Schmidt } 4374d7728c96SJan Schmidt 4375718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4376718dc5faSOmar Sandoval ipath->fspath, size); 4377d7728c96SJan Schmidt if (ret) { 4378d7728c96SJan Schmidt ret = -EFAULT; 4379d7728c96SJan Schmidt goto out; 4380d7728c96SJan Schmidt } 4381d7728c96SJan Schmidt 4382d7728c96SJan Schmidt out: 4383d7728c96SJan Schmidt btrfs_free_path(path); 4384d7728c96SJan Schmidt free_ipath(ipath); 4385d7728c96SJan Schmidt kfree(ipa); 4386d7728c96SJan Schmidt 4387d7728c96SJan Schmidt return ret; 4388d7728c96SJan Schmidt } 4389d7728c96SJan Schmidt 4390d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4391d7728c96SJan Schmidt { 4392d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4393d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4394d7728c96SJan Schmidt 4395d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4396d7728c96SJan Schmidt inodes->bytes_left -= c; 4397d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4398d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4399d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4400d7728c96SJan Schmidt inodes->elem_cnt += 3; 4401d7728c96SJan Schmidt } else { 4402d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4403d7728c96SJan Schmidt inodes->bytes_left = 0; 4404d7728c96SJan Schmidt inodes->elem_missed += 3; 4405d7728c96SJan Schmidt } 4406d7728c96SJan Schmidt 4407d7728c96SJan Schmidt return 0; 4408d7728c96SJan Schmidt } 4409d7728c96SJan Schmidt 44102ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4411d24a67b2SZygo Blaxell void __user *arg, int version) 4412d7728c96SJan Schmidt { 4413d7728c96SJan Schmidt int ret = 0; 4414d7728c96SJan Schmidt int size; 4415d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4416d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4417d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4418d24a67b2SZygo Blaxell bool ignore_offset; 4419d7728c96SJan Schmidt 4420d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4421d7728c96SJan Schmidt return -EPERM; 4422d7728c96SJan Schmidt 4423d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 44247b9ea627SShailendra Verma if (IS_ERR(loi)) 44257b9ea627SShailendra Verma return PTR_ERR(loi); 4426d7728c96SJan Schmidt 4427d24a67b2SZygo Blaxell if (version == 1) { 4428d24a67b2SZygo Blaxell ignore_offset = false; 4429b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4430d24a67b2SZygo Blaxell } else { 4431d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4432d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4433d24a67b2SZygo Blaxell ret = -EINVAL; 4434d24a67b2SZygo Blaxell goto out_loi; 4435d24a67b2SZygo Blaxell } 4436d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4437d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4438d24a67b2SZygo Blaxell ret = -EINVAL; 4439d24a67b2SZygo Blaxell goto out_loi; 4440d24a67b2SZygo Blaxell } 4441d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4442b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4443d24a67b2SZygo Blaxell } 4444d24a67b2SZygo Blaxell 4445d7728c96SJan Schmidt path = btrfs_alloc_path(); 4446d7728c96SJan Schmidt if (!path) { 4447d7728c96SJan Schmidt ret = -ENOMEM; 4448d7728c96SJan Schmidt goto out; 4449d7728c96SJan Schmidt } 4450d7728c96SJan Schmidt 4451d7728c96SJan Schmidt inodes = init_data_container(size); 4452d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4453d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4454d7728c96SJan Schmidt inodes = NULL; 4455d7728c96SJan Schmidt goto out; 4456d7728c96SJan Schmidt } 4457d7728c96SJan Schmidt 44582ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4459d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4460df031f07SLiu Bo if (ret == -EINVAL) 4461d7728c96SJan Schmidt ret = -ENOENT; 4462d7728c96SJan Schmidt if (ret < 0) 4463d7728c96SJan Schmidt goto out; 4464d7728c96SJan Schmidt 4465718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4466718dc5faSOmar Sandoval size); 4467d7728c96SJan Schmidt if (ret) 4468d7728c96SJan Schmidt ret = -EFAULT; 4469d7728c96SJan Schmidt 4470d7728c96SJan Schmidt out: 4471d7728c96SJan Schmidt btrfs_free_path(path); 4472f54de068SDavid Sterba kvfree(inodes); 4473d24a67b2SZygo Blaxell out_loi: 4474d7728c96SJan Schmidt kfree(loi); 4475d7728c96SJan Schmidt 4476d7728c96SJan Schmidt return ret; 4477d7728c96SJan Schmidt } 4478d7728c96SJan Schmidt 4479008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4480c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4481c9e9f97bSIlya Dryomov { 4482c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4483c9e9f97bSIlya Dryomov 4484c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4485c9e9f97bSIlya Dryomov 44863009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4487837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4488837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4489837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4490a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4491a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4492837d5b6eSIlya Dryomov 4493c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4494c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4495c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 449619a39dceSIlya Dryomov 449719a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 449819a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 449919a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4500c9e9f97bSIlya Dryomov } 4501c9e9f97bSIlya Dryomov 45029ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4503c9e9f97bSIlya Dryomov { 4504496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4505c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4506c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4507c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4508ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4509c9e9f97bSIlya Dryomov int ret; 4510c9e9f97bSIlya Dryomov 4511c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4512c9e9f97bSIlya Dryomov return -EPERM; 4513c9e9f97bSIlya Dryomov 4514e54bfa31SLiu Bo ret = mnt_want_write_file(file); 45159ba1f6e4SLiu Bo if (ret) 45169ba1f6e4SLiu Bo return ret; 45179ba1f6e4SLiu Bo 4518ed0fb78fSIlya Dryomov again: 4519171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4520c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4521ed0fb78fSIlya Dryomov need_unlock = true; 4522ed0fb78fSIlya Dryomov goto locked; 4523ed0fb78fSIlya Dryomov } 4524ed0fb78fSIlya Dryomov 4525ed0fb78fSIlya Dryomov /* 452601327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4527ed0fb78fSIlya Dryomov * (1) some other op is running 4528ed0fb78fSIlya Dryomov * (2) balance is running 4529ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4530ed0fb78fSIlya Dryomov */ 4531ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4532ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4533ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 45343009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4535ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4536dccdb07bSDavid Sterba /* 4537dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4538dccdb07bSDavid Sterba * we'll reexamine the status again. 4539dccdb07bSDavid Sterba */ 4540ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4541ed0fb78fSIlya Dryomov 4542ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 45433009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4544ed0fb78fSIlya Dryomov /* this is (3) */ 4545ed0fb78fSIlya Dryomov need_unlock = false; 4546ed0fb78fSIlya Dryomov goto locked; 4547ed0fb78fSIlya Dryomov } 4548ed0fb78fSIlya Dryomov 4549ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4550ed0fb78fSIlya Dryomov goto again; 4551ed0fb78fSIlya Dryomov } else { 4552ed0fb78fSIlya Dryomov /* this is (2) */ 4553ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4554ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4555ed0fb78fSIlya Dryomov goto out; 4556ed0fb78fSIlya Dryomov } 4557ed0fb78fSIlya Dryomov } else { 4558ed0fb78fSIlya Dryomov /* this is (1) */ 4559ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4560e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4561ed0fb78fSIlya Dryomov goto out; 4562ed0fb78fSIlya Dryomov } 4563ed0fb78fSIlya Dryomov 4564ed0fb78fSIlya Dryomov locked: 4565171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4566c9e9f97bSIlya Dryomov 4567c9e9f97bSIlya Dryomov if (arg) { 4568c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4569c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4570c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4571ed0fb78fSIlya Dryomov goto out_unlock; 4572c9e9f97bSIlya Dryomov } 4573de322263SIlya Dryomov 4574de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4575de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4576de322263SIlya Dryomov ret = -ENOTCONN; 4577de322263SIlya Dryomov goto out_bargs; 4578de322263SIlya Dryomov } 4579de322263SIlya Dryomov 4580de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4581de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4582de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4583de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4584de322263SIlya Dryomov 4585de322263SIlya Dryomov goto do_balance; 4586de322263SIlya Dryomov } 4587c9e9f97bSIlya Dryomov } else { 4588c9e9f97bSIlya Dryomov bargs = NULL; 4589c9e9f97bSIlya Dryomov } 4590c9e9f97bSIlya Dryomov 4591ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4592837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4593837d5b6eSIlya Dryomov goto out_bargs; 4594837d5b6eSIlya Dryomov } 4595837d5b6eSIlya Dryomov 45968d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4597c9e9f97bSIlya Dryomov if (!bctl) { 4598c9e9f97bSIlya Dryomov ret = -ENOMEM; 4599c9e9f97bSIlya Dryomov goto out_bargs; 4600c9e9f97bSIlya Dryomov } 4601c9e9f97bSIlya Dryomov 4602c9e9f97bSIlya Dryomov if (arg) { 4603c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4604c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4605c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4606c9e9f97bSIlya Dryomov 4607c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4608f43ffb60SIlya Dryomov } else { 4609f43ffb60SIlya Dryomov /* balance everything - no filters */ 4610f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4611c9e9f97bSIlya Dryomov } 4612c9e9f97bSIlya Dryomov 46138eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 46148eb93459SDavid Sterba ret = -EINVAL; 46150f89abf5SChristian Engelmayer goto out_bctl; 46168eb93459SDavid Sterba } 46178eb93459SDavid Sterba 4618de322263SIlya Dryomov do_balance: 4619c9e9f97bSIlya Dryomov /* 4620149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4621149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4622149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4623149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4624c9e9f97bSIlya Dryomov */ 4625ed0fb78fSIlya Dryomov need_unlock = false; 4626ed0fb78fSIlya Dryomov 46276fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 46280f89abf5SChristian Engelmayer bctl = NULL; 4629ed0fb78fSIlya Dryomov 4630c9e9f97bSIlya Dryomov if (arg) { 4631c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4632c9e9f97bSIlya Dryomov ret = -EFAULT; 4633c9e9f97bSIlya Dryomov } 4634c9e9f97bSIlya Dryomov 46350f89abf5SChristian Engelmayer out_bctl: 46360f89abf5SChristian Engelmayer kfree(bctl); 4637c9e9f97bSIlya Dryomov out_bargs: 4638c9e9f97bSIlya Dryomov kfree(bargs); 4639ed0fb78fSIlya Dryomov out_unlock: 4640c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4641ed0fb78fSIlya Dryomov if (need_unlock) 4642171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4643ed0fb78fSIlya Dryomov out: 4644e54bfa31SLiu Bo mnt_drop_write_file(file); 4645c9e9f97bSIlya Dryomov return ret; 4646c9e9f97bSIlya Dryomov } 4647c9e9f97bSIlya Dryomov 46482ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4649837d5b6eSIlya Dryomov { 4650837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4651837d5b6eSIlya Dryomov return -EPERM; 4652837d5b6eSIlya Dryomov 4653837d5b6eSIlya Dryomov switch (cmd) { 4654837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 46550b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4656a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 46570b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4658837d5b6eSIlya Dryomov } 4659837d5b6eSIlya Dryomov 4660837d5b6eSIlya Dryomov return -EINVAL; 4661837d5b6eSIlya Dryomov } 4662837d5b6eSIlya Dryomov 46632ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 466419a39dceSIlya Dryomov void __user *arg) 466519a39dceSIlya Dryomov { 466619a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 466719a39dceSIlya Dryomov int ret = 0; 466819a39dceSIlya Dryomov 466919a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 467019a39dceSIlya Dryomov return -EPERM; 467119a39dceSIlya Dryomov 467219a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 467319a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 467419a39dceSIlya Dryomov ret = -ENOTCONN; 467519a39dceSIlya Dryomov goto out; 467619a39dceSIlya Dryomov } 467719a39dceSIlya Dryomov 46788d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 467919a39dceSIlya Dryomov if (!bargs) { 468019a39dceSIlya Dryomov ret = -ENOMEM; 468119a39dceSIlya Dryomov goto out; 468219a39dceSIlya Dryomov } 468319a39dceSIlya Dryomov 4684008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 468519a39dceSIlya Dryomov 468619a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 468719a39dceSIlya Dryomov ret = -EFAULT; 468819a39dceSIlya Dryomov 468919a39dceSIlya Dryomov kfree(bargs); 469019a39dceSIlya Dryomov out: 469119a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 469219a39dceSIlya Dryomov return ret; 469319a39dceSIlya Dryomov } 469419a39dceSIlya Dryomov 4695905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 46965d13a37bSArne Jansen { 46970b246afaSJeff Mahoney struct inode *inode = file_inode(file); 46980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 46995d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 47005d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 47015d13a37bSArne Jansen int ret; 47025d13a37bSArne Jansen int err; 47035d13a37bSArne Jansen 47045d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47055d13a37bSArne Jansen return -EPERM; 47065d13a37bSArne Jansen 4707905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4708905b0ddaSMiao Xie if (ret) 4709905b0ddaSMiao Xie return ret; 47105d13a37bSArne Jansen 47115d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4712905b0ddaSMiao Xie if (IS_ERR(sa)) { 4713905b0ddaSMiao Xie ret = PTR_ERR(sa); 4714905b0ddaSMiao Xie goto drop_write; 4715905b0ddaSMiao Xie } 47165d13a37bSArne Jansen 47170b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 47180b246afaSJeff Mahoney trans = btrfs_start_transaction(fs_info->tree_root, 2); 47195d13a37bSArne Jansen if (IS_ERR(trans)) { 47205d13a37bSArne Jansen ret = PTR_ERR(trans); 47215d13a37bSArne Jansen goto out; 47225d13a37bSArne Jansen } 47235d13a37bSArne Jansen 47245d13a37bSArne Jansen switch (sa->cmd) { 47255d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 47260b246afaSJeff Mahoney ret = btrfs_quota_enable(trans, fs_info); 47275d13a37bSArne Jansen break; 47285d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 47290b246afaSJeff Mahoney ret = btrfs_quota_disable(trans, fs_info); 47305d13a37bSArne Jansen break; 47315d13a37bSArne Jansen default: 47325d13a37bSArne Jansen ret = -EINVAL; 47335d13a37bSArne Jansen break; 47345d13a37bSArne Jansen } 47355d13a37bSArne Jansen 47363a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 47375d13a37bSArne Jansen if (err && !ret) 47385d13a37bSArne Jansen ret = err; 47395d13a37bSArne Jansen out: 47405d13a37bSArne Jansen kfree(sa); 47410b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4742905b0ddaSMiao Xie drop_write: 4743905b0ddaSMiao Xie mnt_drop_write_file(file); 47445d13a37bSArne Jansen return ret; 47455d13a37bSArne Jansen } 47465d13a37bSArne Jansen 4747905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 47485d13a37bSArne Jansen { 47490b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47500b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47510b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 47525d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 47535d13a37bSArne Jansen struct btrfs_trans_handle *trans; 47545d13a37bSArne Jansen int ret; 47555d13a37bSArne Jansen int err; 47565d13a37bSArne Jansen 47575d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47585d13a37bSArne Jansen return -EPERM; 47595d13a37bSArne Jansen 4760905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4761905b0ddaSMiao Xie if (ret) 4762905b0ddaSMiao Xie return ret; 47635d13a37bSArne Jansen 47645d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4765905b0ddaSMiao Xie if (IS_ERR(sa)) { 4766905b0ddaSMiao Xie ret = PTR_ERR(sa); 4767905b0ddaSMiao Xie goto drop_write; 4768905b0ddaSMiao Xie } 47695d13a37bSArne Jansen 47705d13a37bSArne Jansen trans = btrfs_join_transaction(root); 47715d13a37bSArne Jansen if (IS_ERR(trans)) { 47725d13a37bSArne Jansen ret = PTR_ERR(trans); 47735d13a37bSArne Jansen goto out; 47745d13a37bSArne Jansen } 47755d13a37bSArne Jansen 47765d13a37bSArne Jansen if (sa->assign) { 47770b246afaSJeff Mahoney ret = btrfs_add_qgroup_relation(trans, fs_info, 47785d13a37bSArne Jansen sa->src, sa->dst); 47795d13a37bSArne Jansen } else { 47800b246afaSJeff Mahoney ret = btrfs_del_qgroup_relation(trans, fs_info, 47815d13a37bSArne Jansen sa->src, sa->dst); 47825d13a37bSArne Jansen } 47835d13a37bSArne Jansen 4784e082f563SQu Wenruo /* update qgroup status and info */ 47850b246afaSJeff Mahoney err = btrfs_run_qgroups(trans, fs_info); 4786e082f563SQu Wenruo if (err < 0) 47870b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4788ad8403dfSAnand Jain "failed to update qgroup status and info"); 47893a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 47905d13a37bSArne Jansen if (err && !ret) 47915d13a37bSArne Jansen ret = err; 47925d13a37bSArne Jansen 47935d13a37bSArne Jansen out: 47945d13a37bSArne Jansen kfree(sa); 4795905b0ddaSMiao Xie drop_write: 4796905b0ddaSMiao Xie mnt_drop_write_file(file); 47975d13a37bSArne Jansen return ret; 47985d13a37bSArne Jansen } 47995d13a37bSArne Jansen 4800905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 48015d13a37bSArne Jansen { 48020b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48030b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48040b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48055d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 48065d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48075d13a37bSArne Jansen int ret; 48085d13a37bSArne Jansen int err; 48095d13a37bSArne Jansen 48105d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48115d13a37bSArne Jansen return -EPERM; 48125d13a37bSArne Jansen 4813905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4814905b0ddaSMiao Xie if (ret) 4815905b0ddaSMiao Xie return ret; 48165d13a37bSArne Jansen 48175d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4818905b0ddaSMiao Xie if (IS_ERR(sa)) { 4819905b0ddaSMiao Xie ret = PTR_ERR(sa); 4820905b0ddaSMiao Xie goto drop_write; 4821905b0ddaSMiao Xie } 48225d13a37bSArne Jansen 4823d86e56cfSMiao Xie if (!sa->qgroupid) { 4824d86e56cfSMiao Xie ret = -EINVAL; 4825d86e56cfSMiao Xie goto out; 4826d86e56cfSMiao Xie } 4827d86e56cfSMiao Xie 48285d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48295d13a37bSArne Jansen if (IS_ERR(trans)) { 48305d13a37bSArne Jansen ret = PTR_ERR(trans); 48315d13a37bSArne Jansen goto out; 48325d13a37bSArne Jansen } 48335d13a37bSArne Jansen 48345d13a37bSArne Jansen if (sa->create) { 48350b246afaSJeff Mahoney ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); 48365d13a37bSArne Jansen } else { 48370b246afaSJeff Mahoney ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); 48385d13a37bSArne Jansen } 48395d13a37bSArne Jansen 48403a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48415d13a37bSArne Jansen if (err && !ret) 48425d13a37bSArne Jansen ret = err; 48435d13a37bSArne Jansen 48445d13a37bSArne Jansen out: 48455d13a37bSArne Jansen kfree(sa); 4846905b0ddaSMiao Xie drop_write: 4847905b0ddaSMiao Xie mnt_drop_write_file(file); 48485d13a37bSArne Jansen return ret; 48495d13a37bSArne Jansen } 48505d13a37bSArne Jansen 4851905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 48525d13a37bSArne Jansen { 48530b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48550b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48565d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 48575d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48585d13a37bSArne Jansen int ret; 48595d13a37bSArne Jansen int err; 48605d13a37bSArne Jansen u64 qgroupid; 48615d13a37bSArne Jansen 48625d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48635d13a37bSArne Jansen return -EPERM; 48645d13a37bSArne Jansen 4865905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4866905b0ddaSMiao Xie if (ret) 4867905b0ddaSMiao Xie return ret; 48685d13a37bSArne Jansen 48695d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4870905b0ddaSMiao Xie if (IS_ERR(sa)) { 4871905b0ddaSMiao Xie ret = PTR_ERR(sa); 4872905b0ddaSMiao Xie goto drop_write; 4873905b0ddaSMiao Xie } 48745d13a37bSArne Jansen 48755d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48765d13a37bSArne Jansen if (IS_ERR(trans)) { 48775d13a37bSArne Jansen ret = PTR_ERR(trans); 48785d13a37bSArne Jansen goto out; 48795d13a37bSArne Jansen } 48805d13a37bSArne Jansen 48815d13a37bSArne Jansen qgroupid = sa->qgroupid; 48825d13a37bSArne Jansen if (!qgroupid) { 48835d13a37bSArne Jansen /* take the current subvol as qgroup */ 48845d13a37bSArne Jansen qgroupid = root->root_key.objectid; 48855d13a37bSArne Jansen } 48865d13a37bSArne Jansen 48870b246afaSJeff Mahoney ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); 48885d13a37bSArne Jansen 48893a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48905d13a37bSArne Jansen if (err && !ret) 48915d13a37bSArne Jansen ret = err; 48925d13a37bSArne Jansen 48935d13a37bSArne Jansen out: 48945d13a37bSArne Jansen kfree(sa); 4895905b0ddaSMiao Xie drop_write: 4896905b0ddaSMiao Xie mnt_drop_write_file(file); 48975d13a37bSArne Jansen return ret; 48985d13a37bSArne Jansen } 48995d13a37bSArne Jansen 49002f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 49012f232036SJan Schmidt { 49020b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49030b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49042f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49052f232036SJan Schmidt int ret; 49062f232036SJan Schmidt 49072f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49082f232036SJan Schmidt return -EPERM; 49092f232036SJan Schmidt 49102f232036SJan Schmidt ret = mnt_want_write_file(file); 49112f232036SJan Schmidt if (ret) 49122f232036SJan Schmidt return ret; 49132f232036SJan Schmidt 49142f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 49152f232036SJan Schmidt if (IS_ERR(qsa)) { 49162f232036SJan Schmidt ret = PTR_ERR(qsa); 49172f232036SJan Schmidt goto drop_write; 49182f232036SJan Schmidt } 49192f232036SJan Schmidt 49202f232036SJan Schmidt if (qsa->flags) { 49212f232036SJan Schmidt ret = -EINVAL; 49222f232036SJan Schmidt goto out; 49232f232036SJan Schmidt } 49242f232036SJan Schmidt 49250b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 49262f232036SJan Schmidt 49272f232036SJan Schmidt out: 49282f232036SJan Schmidt kfree(qsa); 49292f232036SJan Schmidt drop_write: 49302f232036SJan Schmidt mnt_drop_write_file(file); 49312f232036SJan Schmidt return ret; 49322f232036SJan Schmidt } 49332f232036SJan Schmidt 49342f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 49352f232036SJan Schmidt { 49360b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49370b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49382f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49392f232036SJan Schmidt int ret = 0; 49402f232036SJan Schmidt 49412f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49422f232036SJan Schmidt return -EPERM; 49432f232036SJan Schmidt 49448d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 49452f232036SJan Schmidt if (!qsa) 49462f232036SJan Schmidt return -ENOMEM; 49472f232036SJan Schmidt 49480b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 49492f232036SJan Schmidt qsa->flags = 1; 49500b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 49512f232036SJan Schmidt } 49522f232036SJan Schmidt 49532f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 49542f232036SJan Schmidt ret = -EFAULT; 49552f232036SJan Schmidt 49562f232036SJan Schmidt kfree(qsa); 49572f232036SJan Schmidt return ret; 49582f232036SJan Schmidt } 49592f232036SJan Schmidt 496057254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 496157254b6eSJan Schmidt { 49620b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49630b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 496457254b6eSJan Schmidt 496557254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 496657254b6eSJan Schmidt return -EPERM; 496757254b6eSJan Schmidt 49680b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 496957254b6eSJan Schmidt } 497057254b6eSJan Schmidt 4971abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 4972abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 49738ea05e3aSAlexander Block { 4974496ad9aaSAl Viro struct inode *inode = file_inode(file); 49750b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49768ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 49778ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 49788ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 4979c2050a45SDeepa Dinamani struct timespec ct = current_time(inode); 49808ea05e3aSAlexander Block int ret = 0; 4981dd5f9615SStefan Behrens int received_uuid_changed; 49828ea05e3aSAlexander Block 4983bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 4984bd60ea0fSDavid Sterba return -EPERM; 4985bd60ea0fSDavid Sterba 49868ea05e3aSAlexander Block ret = mnt_want_write_file(file); 49878ea05e3aSAlexander Block if (ret < 0) 49888ea05e3aSAlexander Block return ret; 49898ea05e3aSAlexander Block 49900b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 49918ea05e3aSAlexander Block 49924a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 49938ea05e3aSAlexander Block ret = -EINVAL; 49948ea05e3aSAlexander Block goto out; 49958ea05e3aSAlexander Block } 49968ea05e3aSAlexander Block 49978ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 49988ea05e3aSAlexander Block ret = -EROFS; 49998ea05e3aSAlexander Block goto out; 50008ea05e3aSAlexander Block } 50018ea05e3aSAlexander Block 5002dd5f9615SStefan Behrens /* 5003dd5f9615SStefan Behrens * 1 - root item 5004dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5005dd5f9615SStefan Behrens */ 5006dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 50078ea05e3aSAlexander Block if (IS_ERR(trans)) { 50088ea05e3aSAlexander Block ret = PTR_ERR(trans); 50098ea05e3aSAlexander Block trans = NULL; 50108ea05e3aSAlexander Block goto out; 50118ea05e3aSAlexander Block } 50128ea05e3aSAlexander Block 50138ea05e3aSAlexander Block sa->rtransid = trans->transid; 50148ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 50158ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 50168ea05e3aSAlexander Block 5017dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5018dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5019dd5f9615SStefan Behrens if (received_uuid_changed && 5020d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 5021d1957791SLu Fengqi ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 5022dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5023dd5f9615SStefan Behrens root->root_key.objectid); 5024d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 5025d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 5026d87ff758SNikolay Borisov btrfs_end_transaction(trans); 5027d87ff758SNikolay Borisov goto out; 5028d87ff758SNikolay Borisov } 5029d87ff758SNikolay Borisov } 50308ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 50318ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 50328ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 50333cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 50343cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 50353cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 50363cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 50378ea05e3aSAlexander Block 50380b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 50398ea05e3aSAlexander Block &root->root_key, &root->root_item); 50408ea05e3aSAlexander Block if (ret < 0) { 50413a45bb20SJeff Mahoney btrfs_end_transaction(trans); 50428ea05e3aSAlexander Block goto out; 5043dd5f9615SStefan Behrens } 5044dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5045cdb345a8SLu Fengqi ret = btrfs_uuid_tree_add(trans, sa->uuid, 5046dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5047dd5f9615SStefan Behrens root->root_key.objectid); 5048dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 504966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5050efd38150SNikolay Borisov btrfs_end_transaction(trans); 5051dd5f9615SStefan Behrens goto out; 5052dd5f9615SStefan Behrens } 5053dd5f9615SStefan Behrens } 50543a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5055abccd00fSHugo Mills out: 50560b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5057abccd00fSHugo Mills mnt_drop_write_file(file); 5058abccd00fSHugo Mills return ret; 5059abccd00fSHugo Mills } 5060abccd00fSHugo Mills 5061abccd00fSHugo Mills #ifdef CONFIG_64BIT 5062abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5063abccd00fSHugo Mills void __user *arg) 5064abccd00fSHugo Mills { 5065abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5066abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5067abccd00fSHugo Mills int ret = 0; 5068abccd00fSHugo Mills 5069abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 50707b9ea627SShailendra Verma if (IS_ERR(args32)) 50717b9ea627SShailendra Verma return PTR_ERR(args32); 5072abccd00fSHugo Mills 50738d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 507484dbeb87SDan Carpenter if (!args64) { 507584dbeb87SDan Carpenter ret = -ENOMEM; 5076abccd00fSHugo Mills goto out; 5077abccd00fSHugo Mills } 5078abccd00fSHugo Mills 5079abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5080abccd00fSHugo Mills args64->stransid = args32->stransid; 5081abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5082abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5083abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5084abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5085abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5086abccd00fSHugo Mills args64->flags = args32->flags; 5087abccd00fSHugo Mills 5088abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5089abccd00fSHugo Mills if (ret) 5090abccd00fSHugo Mills goto out; 5091abccd00fSHugo Mills 5092abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5093abccd00fSHugo Mills args32->stransid = args64->stransid; 5094abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5095abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5096abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5097abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5098abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5099abccd00fSHugo Mills args32->flags = args64->flags; 5100abccd00fSHugo Mills 5101abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5102abccd00fSHugo Mills if (ret) 5103abccd00fSHugo Mills ret = -EFAULT; 5104abccd00fSHugo Mills 5105abccd00fSHugo Mills out: 5106abccd00fSHugo Mills kfree(args32); 5107abccd00fSHugo Mills kfree(args64); 5108abccd00fSHugo Mills return ret; 5109abccd00fSHugo Mills } 5110abccd00fSHugo Mills #endif 5111abccd00fSHugo Mills 5112abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5113abccd00fSHugo Mills void __user *arg) 5114abccd00fSHugo Mills { 5115abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5116abccd00fSHugo Mills int ret = 0; 5117abccd00fSHugo Mills 5118abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 51197b9ea627SShailendra Verma if (IS_ERR(sa)) 51207b9ea627SShailendra Verma return PTR_ERR(sa); 5121abccd00fSHugo Mills 5122abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5123abccd00fSHugo Mills 5124abccd00fSHugo Mills if (ret) 5125abccd00fSHugo Mills goto out; 5126abccd00fSHugo Mills 51278ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 51288ea05e3aSAlexander Block if (ret) 51298ea05e3aSAlexander Block ret = -EFAULT; 51308ea05e3aSAlexander Block 51318ea05e3aSAlexander Block out: 51328ea05e3aSAlexander Block kfree(sa); 51338ea05e3aSAlexander Block return ret; 51348ea05e3aSAlexander Block } 51358ea05e3aSAlexander Block 5136867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5137867ab667Sjeff.liu { 51380b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5140a1b83ac5SAnand Jain size_t len; 5141867ab667Sjeff.liu int ret; 5142a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5143a1b83ac5SAnand Jain 51440b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 51450b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 51460b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5147a1b83ac5SAnand Jain 5148a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5149867ab667Sjeff.liu 5150867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 51510b246afaSJeff Mahoney btrfs_warn(fs_info, 51520b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 51530b246afaSJeff Mahoney --len); 5154867ab667Sjeff.liu } 5155867ab667Sjeff.liu 5156867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5157867ab667Sjeff.liu 5158867ab667Sjeff.liu return ret ? -EFAULT : 0; 5159867ab667Sjeff.liu } 5160867ab667Sjeff.liu 5161a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5162a8bfd4abSjeff.liu { 51630b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51640b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51650b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51660b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5167a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5168a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5169a8bfd4abSjeff.liu int ret; 5170a8bfd4abSjeff.liu 5171a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5172a8bfd4abSjeff.liu return -EPERM; 5173a8bfd4abSjeff.liu 5174a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5175a8bfd4abSjeff.liu return -EFAULT; 5176a8bfd4abSjeff.liu 5177a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 51780b246afaSJeff Mahoney btrfs_err(fs_info, 51795d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5180a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5181a8bfd4abSjeff.liu return -EINVAL; 5182a8bfd4abSjeff.liu } 5183a8bfd4abSjeff.liu 5184a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5185a8bfd4abSjeff.liu if (ret) 5186a8bfd4abSjeff.liu return ret; 5187a8bfd4abSjeff.liu 5188a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5189a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5190a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5191a8bfd4abSjeff.liu goto out_unlock; 5192a8bfd4abSjeff.liu } 5193a8bfd4abSjeff.liu 51940b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5195a8bfd4abSjeff.liu strcpy(super_block->label, label); 51960b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 51973a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5198a8bfd4abSjeff.liu 5199a8bfd4abSjeff.liu out_unlock: 5200a8bfd4abSjeff.liu mnt_drop_write_file(file); 5201a8bfd4abSjeff.liu return ret; 5202a8bfd4abSjeff.liu } 5203a8bfd4abSjeff.liu 52042eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 52052eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 52062eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 52072eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 52082eaa055fSJeff Mahoney 5209d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 52102eaa055fSJeff Mahoney { 52114d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 52122eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 52132eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 52142eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 52152eaa055fSJeff Mahoney }; 52162eaa055fSJeff Mahoney 52172eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52182eaa055fSJeff Mahoney return -EFAULT; 52192eaa055fSJeff Mahoney 52202eaa055fSJeff Mahoney return 0; 52212eaa055fSJeff Mahoney } 52222eaa055fSJeff Mahoney 52232eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 52242eaa055fSJeff Mahoney { 52250b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52260b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52270b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 52282eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 52292eaa055fSJeff Mahoney 52302eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 52312eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 52322eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 52332eaa055fSJeff Mahoney 52342eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52352eaa055fSJeff Mahoney return -EFAULT; 52362eaa055fSJeff Mahoney 52372eaa055fSJeff Mahoney return 0; 52382eaa055fSJeff Mahoney } 52392eaa055fSJeff Mahoney 52402ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 52413b02a68aSJeff Mahoney enum btrfs_feature_set set, 52422eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 52432eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 52442eaa055fSJeff Mahoney { 52453b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 52463b02a68aSJeff Mahoney char *names; 52472eaa055fSJeff Mahoney u64 disallowed, unsupported; 52482eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 52492eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 52502eaa055fSJeff Mahoney 52512eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 52522eaa055fSJeff Mahoney if (unsupported) { 52533b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 52543b02a68aSJeff Mahoney if (names) { 52550b246afaSJeff Mahoney btrfs_warn(fs_info, 52563b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 52573b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 52583b02a68aSJeff Mahoney kfree(names); 52593b02a68aSJeff Mahoney } else 52600b246afaSJeff Mahoney btrfs_warn(fs_info, 52612eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 52622eaa055fSJeff Mahoney type, unsupported); 52632eaa055fSJeff Mahoney return -EOPNOTSUPP; 52642eaa055fSJeff Mahoney } 52652eaa055fSJeff Mahoney 52662eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 52672eaa055fSJeff Mahoney if (disallowed) { 52683b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 52693b02a68aSJeff Mahoney if (names) { 52700b246afaSJeff Mahoney btrfs_warn(fs_info, 52713b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 52723b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 52733b02a68aSJeff Mahoney kfree(names); 52743b02a68aSJeff Mahoney } else 52750b246afaSJeff Mahoney btrfs_warn(fs_info, 52762eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 52772eaa055fSJeff Mahoney type, disallowed); 52782eaa055fSJeff Mahoney return -EPERM; 52792eaa055fSJeff Mahoney } 52802eaa055fSJeff Mahoney 52812eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 52822eaa055fSJeff Mahoney if (disallowed) { 52833b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 52843b02a68aSJeff Mahoney if (names) { 52850b246afaSJeff Mahoney btrfs_warn(fs_info, 52863b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 52873b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 52883b02a68aSJeff Mahoney kfree(names); 52893b02a68aSJeff Mahoney } else 52900b246afaSJeff Mahoney btrfs_warn(fs_info, 52912eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 52922eaa055fSJeff Mahoney type, disallowed); 52932eaa055fSJeff Mahoney return -EPERM; 52942eaa055fSJeff Mahoney } 52952eaa055fSJeff Mahoney 52962eaa055fSJeff Mahoney return 0; 52972eaa055fSJeff Mahoney } 52982eaa055fSJeff Mahoney 52992ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 53002ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 53012eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 53022eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 53032eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 53042eaa055fSJeff Mahoney 53052eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 53062eaa055fSJeff Mahoney { 53070b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53080b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53090b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 53100b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53112eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 53122eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 53132eaa055fSJeff Mahoney u64 newflags; 53142eaa055fSJeff Mahoney int ret; 53152eaa055fSJeff Mahoney 53162eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 53172eaa055fSJeff Mahoney return -EPERM; 53182eaa055fSJeff Mahoney 53192eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 53202eaa055fSJeff Mahoney return -EFAULT; 53212eaa055fSJeff Mahoney 53222eaa055fSJeff Mahoney /* Nothing to do */ 53232eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 53242eaa055fSJeff Mahoney !flags[0].incompat_flags) 53252eaa055fSJeff Mahoney return 0; 53262eaa055fSJeff Mahoney 53272ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 53282eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 53292eaa055fSJeff Mahoney if (ret) 53302eaa055fSJeff Mahoney return ret; 53312eaa055fSJeff Mahoney 53322ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 53332eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 53342eaa055fSJeff Mahoney if (ret) 53352eaa055fSJeff Mahoney return ret; 53362eaa055fSJeff Mahoney 53372ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 53382eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 53392eaa055fSJeff Mahoney if (ret) 53402eaa055fSJeff Mahoney return ret; 53412eaa055fSJeff Mahoney 53427ab19625SDavid Sterba ret = mnt_want_write_file(file); 53437ab19625SDavid Sterba if (ret) 53447ab19625SDavid Sterba return ret; 53457ab19625SDavid Sterba 53468051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 53477ab19625SDavid Sterba if (IS_ERR(trans)) { 53487ab19625SDavid Sterba ret = PTR_ERR(trans); 53497ab19625SDavid Sterba goto out_drop_write; 53507ab19625SDavid Sterba } 53512eaa055fSJeff Mahoney 53520b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 53532eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 53542eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 53552eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 53562eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 53572eaa055fSJeff Mahoney 53582eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 53592eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 53602eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 53612eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 53622eaa055fSJeff Mahoney 53632eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 53642eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 53652eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 53662eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 53670b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53682eaa055fSJeff Mahoney 53693a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 53707ab19625SDavid Sterba out_drop_write: 53717ab19625SDavid Sterba mnt_drop_write_file(file); 53727ab19625SDavid Sterba 53737ab19625SDavid Sterba return ret; 53742eaa055fSJeff Mahoney } 53752eaa055fSJeff Mahoney 53762351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 53772351f431SJosef Bacik { 53782351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 53792351f431SJosef Bacik int ret; 53802351f431SJosef Bacik 53812351f431SJosef Bacik if (compat) { 53822351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 53832351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 53842351f431SJosef Bacik 53852351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 53862351f431SJosef Bacik if (ret) 53872351f431SJosef Bacik return -EFAULT; 53882351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 53892351f431SJosef Bacik if (!arg) 53902351f431SJosef Bacik return -ENOMEM; 53912351f431SJosef Bacik arg->send_fd = args32.send_fd; 53922351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 53932351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 53942351f431SJosef Bacik arg->parent_root = args32.parent_root; 53952351f431SJosef Bacik arg->flags = args32.flags; 53962351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 53972351f431SJosef Bacik sizeof(args32.reserved)); 53982351f431SJosef Bacik #else 53992351f431SJosef Bacik return -ENOTTY; 54002351f431SJosef Bacik #endif 54012351f431SJosef Bacik } else { 54022351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 54032351f431SJosef Bacik if (IS_ERR(arg)) 54042351f431SJosef Bacik return PTR_ERR(arg); 54052351f431SJosef Bacik } 54062351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 54072351f431SJosef Bacik kfree(arg); 54082351f431SJosef Bacik return ret; 54092351f431SJosef Bacik } 54102351f431SJosef Bacik 5411f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5412f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5413f46b5a66SChristoph Hellwig { 54140b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54150b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54160b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54174bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5418f46b5a66SChristoph Hellwig 5419f46b5a66SChristoph Hellwig switch (cmd) { 54206cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 54216cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 54226cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 54236cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 54246cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 54256cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5426f7039b1dSLi Dongyang case FITRIM: 5427f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5428f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5429fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5430fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5431fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 54323de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5433fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 54346f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 54356f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 543676dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 543776dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 54380caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 54390caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 54400caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 54410caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 54426ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 54436ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5444f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 54451e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 54461e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 54471e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5448f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5449198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5450f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 54512ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5452f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5453da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 54546b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 54556b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5456475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 54572ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5458475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 54592ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5460f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 54619ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5462ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5463ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5464cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5465cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5466ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5467ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5468d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5469d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5470d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5471d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5472d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5473d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 54741406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 54752ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 54769b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 54779b199859SFilipe David Borba Manana int ret; 54789b199859SFilipe David Borba Manana 547982b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 54809b199859SFilipe David Borba Manana if (ret) 54819b199859SFilipe David Borba Manana return ret; 54820b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 54832fad4e83SDavid Sterba /* 54842fad4e83SDavid Sterba * The transaction thread may want to do more work, 548501327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 54862fad4e83SDavid Sterba * processing uncleaned subvols. 54872fad4e83SDavid Sterba */ 54880b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 54899b199859SFilipe David Borba Manana return ret; 54909b199859SFilipe David Borba Manana } 549146204592SSage Weil case BTRFS_IOC_START_SYNC: 54929a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 549346204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 54942ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5495475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5496b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5497475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 54982ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5499475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 55002ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5501c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55029ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5503837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 55042ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 550519a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 55062ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 55078ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55088ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5509abccd00fSHugo Mills #ifdef CONFIG_64BIT 5510abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5511abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5512abccd00fSHugo Mills #endif 551331db9f7cSAlexander Block case BTRFS_IOC_SEND: 55142351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 55152351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 55162351f431SJosef Bacik case BTRFS_IOC_SEND_32: 55172351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 55182351f431SJosef Bacik #endif 5519c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 55202ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 55215d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5522905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55235d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5524905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55255d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5526905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55275d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5528905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 55292f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 55302f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 55312f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 55322f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 553357254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 553457254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 55353f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 55362ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5537867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5538867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5539a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5540a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 55412eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5542d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 55432eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 55442eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 55452eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 55462eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5547e4202ac9SDavid Sterba case FS_IOC_FSGETXATTR: 5548e4202ac9SDavid Sterba return btrfs_ioctl_fsgetxattr(file, argp); 5549025f2121SDavid Sterba case FS_IOC_FSSETXATTR: 5550025f2121SDavid Sterba return btrfs_ioctl_fssetxattr(file, argp); 5551f46b5a66SChristoph Hellwig } 5552f46b5a66SChristoph Hellwig 5553f46b5a66SChristoph Hellwig return -ENOTTY; 5554f46b5a66SChristoph Hellwig } 55554c63c245SLuke Dashjr 55564c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 55574c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 55584c63c245SLuke Dashjr { 55592a362249SJeff Mahoney /* 55602a362249SJeff Mahoney * These all access 32-bit values anyway so no further 55612a362249SJeff Mahoney * handling is necessary. 55622a362249SJeff Mahoney */ 55634c63c245SLuke Dashjr switch (cmd) { 55644c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 55654c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 55664c63c245SLuke Dashjr break; 55674c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 55684c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 55694c63c245SLuke Dashjr break; 55704c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 55714c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 55724c63c245SLuke Dashjr break; 55734c63c245SLuke Dashjr } 55744c63c245SLuke Dashjr 55754c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 55764c63c245SLuke Dashjr } 55774c63c245SLuke Dashjr #endif 5578