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 /* 1086cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 1096cbff00fSChristoph Hellwig */ 1106cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 1116cbff00fSChristoph Hellwig { 1126cbff00fSChristoph Hellwig unsigned int iflags = 0; 1136cbff00fSChristoph Hellwig 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1156cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1176cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1196cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1216cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1226cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1236cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1246cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1256cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 126d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 127d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 128d0092bddSLi Zefan 12913f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 130d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 13113f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13213f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1336cbff00fSChristoph Hellwig 1346cbff00fSChristoph Hellwig return iflags; 1356cbff00fSChristoph Hellwig } 1366cbff00fSChristoph Hellwig 1376cbff00fSChristoph Hellwig /* 1386cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1396cbff00fSChristoph Hellwig */ 1407b6a221eSDavid Sterba void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) 1416cbff00fSChristoph Hellwig { 1426cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1433cc79392SFilipe Manana unsigned int new_fl = 0; 1446cbff00fSChristoph Hellwig 1456cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1463cc79392SFilipe Manana new_fl |= S_SYNC; 1476cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1483cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1496cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1503cc79392SFilipe Manana new_fl |= S_APPEND; 1516cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1523cc79392SFilipe Manana new_fl |= S_NOATIME; 1536cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1543cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1553cc79392SFilipe Manana 1563cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1573cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1583cc79392SFilipe Manana new_fl); 1596cbff00fSChristoph Hellwig } 1606cbff00fSChristoph Hellwig 1616cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1626cbff00fSChristoph Hellwig { 163496ad9aaSAl Viro struct btrfs_inode *ip = BTRFS_I(file_inode(file)); 1646cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1656cbff00fSChristoph Hellwig 1666cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1676cbff00fSChristoph Hellwig return -EFAULT; 1686cbff00fSChristoph Hellwig return 0; 1696cbff00fSChristoph Hellwig } 1706cbff00fSChristoph Hellwig 17175e7cb7fSLiu Bo static int check_flags(unsigned int flags) 17275e7cb7fSLiu Bo { 17375e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17475e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17575e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 176e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 177e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17875e7cb7fSLiu Bo return -EOPNOTSUPP; 17975e7cb7fSLiu Bo 18075e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 18175e7cb7fSLiu Bo return -EINVAL; 18275e7cb7fSLiu Bo 18375e7cb7fSLiu Bo return 0; 18475e7cb7fSLiu Bo } 18575e7cb7fSLiu Bo 1866cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1876cbff00fSChristoph Hellwig { 188496ad9aaSAl Viro struct inode *inode = file_inode(file); 1890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1906cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1916cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 1926cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1936cbff00fSChristoph Hellwig unsigned int flags, oldflags; 1946cbff00fSChristoph Hellwig int ret; 195f062abf0SLi Zefan u64 ip_oldflags; 196f062abf0SLi Zefan unsigned int i_oldflags; 1977e97b8daSDavid Sterba umode_t mode; 1986cbff00fSChristoph Hellwig 199bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 200bd60ea0fSDavid Sterba return -EPERM; 201bd60ea0fSDavid Sterba 202b83cc969SLi Zefan if (btrfs_root_readonly(root)) 203b83cc969SLi Zefan return -EROFS; 204b83cc969SLi Zefan 2056cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 2066cbff00fSChristoph Hellwig return -EFAULT; 2076cbff00fSChristoph Hellwig 20875e7cb7fSLiu Bo ret = check_flags(flags); 20975e7cb7fSLiu Bo if (ret) 21075e7cb7fSLiu Bo return ret; 2116cbff00fSChristoph Hellwig 212e7848683SJan Kara ret = mnt_want_write_file(file); 213e7848683SJan Kara if (ret) 214e7848683SJan Kara return ret; 215e7848683SJan Kara 2165955102cSAl Viro inode_lock(inode); 2176cbff00fSChristoph Hellwig 218f062abf0SLi Zefan ip_oldflags = ip->flags; 219f062abf0SLi Zefan i_oldflags = inode->i_flags; 2207e97b8daSDavid Sterba mode = inode->i_mode; 221f062abf0SLi Zefan 2221905a0f7SDavid Sterba flags = btrfs_mask_fsflags_for_type(inode, flags); 2236cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 2246cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2256cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2266cbff00fSChristoph Hellwig ret = -EPERM; 2276cbff00fSChristoph Hellwig goto out_unlock; 2286cbff00fSChristoph Hellwig } 2296cbff00fSChristoph Hellwig } 2306cbff00fSChristoph Hellwig 2316cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2326cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2336cbff00fSChristoph Hellwig else 2346cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2356cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2366cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2376cbff00fSChristoph Hellwig else 2386cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2396cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2406cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2416cbff00fSChristoph Hellwig else 2426cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2436cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2446cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2456cbff00fSChristoph Hellwig else 2466cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2476cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2486cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2496cbff00fSChristoph Hellwig else 2506cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2516cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2526cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2536cbff00fSChristoph Hellwig else 2546cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 2557e97b8daSDavid Sterba if (flags & FS_NOCOW_FL) { 2567e97b8daSDavid Sterba if (S_ISREG(mode)) { 2577e97b8daSDavid Sterba /* 2587e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2597e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2607e97b8daSDavid Sterba * status of the file and will not set it. 2617e97b8daSDavid Sterba */ 2627e97b8daSDavid Sterba if (inode->i_size == 0) 2637e97b8daSDavid Sterba ip->flags |= BTRFS_INODE_NODATACOW 2647e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM; 2657e97b8daSDavid Sterba } else { 266e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 2677e97b8daSDavid Sterba } 2687e97b8daSDavid Sterba } else { 2697e97b8daSDavid Sterba /* 27001327610SNicholas D Steeves * Revert back under same assumptions as above 2717e97b8daSDavid Sterba */ 2727e97b8daSDavid Sterba if (S_ISREG(mode)) { 2737e97b8daSDavid Sterba if (inode->i_size == 0) 2747e97b8daSDavid Sterba ip->flags &= ~(BTRFS_INODE_NODATACOW 2757e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM); 2767e97b8daSDavid Sterba } else { 277e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 2787e97b8daSDavid Sterba } 2797e97b8daSDavid Sterba } 2806cbff00fSChristoph Hellwig 28175e7cb7fSLiu Bo /* 28275e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 28375e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 28475e7cb7fSLiu Bo * things smaller. 28575e7cb7fSLiu Bo */ 28675e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 28775e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 28875e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 28963541927SFilipe David Borba Manana 29063541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 29163541927SFilipe David Borba Manana if (ret && ret != -ENODATA) 29263541927SFilipe David Borba Manana goto out_drop; 29375e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 29463541927SFilipe David Borba Manana const char *comp; 29563541927SFilipe David Borba Manana 29675e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 29775e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 29863541927SFilipe David Borba Manana 29993370509SDavid Sterba comp = btrfs_compress_type2str(fs_info->compress_type); 30093370509SDavid Sterba if (!comp || comp[0] == 0) 30193370509SDavid Sterba comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 30293370509SDavid Sterba 30363541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", 30463541927SFilipe David Borba Manana comp, strlen(comp), 0); 30563541927SFilipe David Borba Manana if (ret) 30663541927SFilipe David Borba Manana goto out_drop; 30763541927SFilipe David Borba Manana 308ebcb904dSLi Zefan } else { 30978a017a2SFilipe Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 31078a017a2SFilipe Manana if (ret && ret != -ENODATA) 31178a017a2SFilipe Manana goto out_drop; 312ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 31375e7cb7fSLiu Bo } 3146cbff00fSChristoph Hellwig 3154da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 316f062abf0SLi Zefan if (IS_ERR(trans)) { 317f062abf0SLi Zefan ret = PTR_ERR(trans); 318f062abf0SLi Zefan goto out_drop; 319f062abf0SLi Zefan } 3206cbff00fSChristoph Hellwig 3217b6a221eSDavid Sterba btrfs_sync_inode_flags_to_i_flags(inode); 3220c4d2d95SJosef Bacik inode_inc_iversion(inode); 323c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3246cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3256cbff00fSChristoph Hellwig 3263a45bb20SJeff Mahoney btrfs_end_transaction(trans); 327f062abf0SLi Zefan out_drop: 328f062abf0SLi Zefan if (ret) { 329f062abf0SLi Zefan ip->flags = ip_oldflags; 330f062abf0SLi Zefan inode->i_flags = i_oldflags; 331f062abf0SLi Zefan } 3326cbff00fSChristoph Hellwig 3336cbff00fSChristoph Hellwig out_unlock: 3345955102cSAl Viro inode_unlock(inode); 335e7848683SJan Kara mnt_drop_write_file(file); 3362d4e6f6aSliubo return ret; 3376cbff00fSChristoph Hellwig } 3386cbff00fSChristoph Hellwig 3396cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 3406cbff00fSChristoph Hellwig { 341496ad9aaSAl Viro struct inode *inode = file_inode(file); 3426cbff00fSChristoph Hellwig 3436cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 3446cbff00fSChristoph Hellwig } 345f46b5a66SChristoph Hellwig 346f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 347f7039b1dSLi Dongyang { 3480b246afaSJeff Mahoney struct inode *inode = file_inode(file); 3490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 350f7039b1dSLi Dongyang struct btrfs_device *device; 351f7039b1dSLi Dongyang struct request_queue *q; 352f7039b1dSLi Dongyang struct fstrim_range range; 353f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 354f7039b1dSLi Dongyang u64 num_devices = 0; 355815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 356f7039b1dSLi Dongyang int ret; 357f7039b1dSLi Dongyang 358f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 359f7039b1dSLi Dongyang return -EPERM; 360f7039b1dSLi Dongyang 3611f78160cSXiao Guangrong rcu_read_lock(); 3621f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3631f78160cSXiao Guangrong dev_list) { 364f7039b1dSLi Dongyang if (!device->bdev) 365f7039b1dSLi Dongyang continue; 366f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 367f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 368f7039b1dSLi Dongyang num_devices++; 36950d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 370f7039b1dSLi Dongyang minlen); 371f7039b1dSLi Dongyang } 372f7039b1dSLi Dongyang } 3731f78160cSXiao Guangrong rcu_read_unlock(); 374f4c697e6SLukas Czerner 375f7039b1dSLi Dongyang if (!num_devices) 376f7039b1dSLi Dongyang return -EOPNOTSUPP; 377f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 378f7039b1dSLi Dongyang return -EFAULT; 379e515c18bSLukas Czerner if (range.start > total_bytes || 380e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 381f4c697e6SLukas Czerner return -EINVAL; 382f7039b1dSLi Dongyang 383f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 384f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 3852ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 386f7039b1dSLi Dongyang if (ret < 0) 387f7039b1dSLi Dongyang return ret; 388f7039b1dSLi Dongyang 389f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 390f7039b1dSLi Dongyang return -EFAULT; 391f7039b1dSLi Dongyang 392f7039b1dSLi Dongyang return 0; 393f7039b1dSLi Dongyang } 394f7039b1dSLi Dongyang 395dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 396dd5f9615SStefan Behrens { 39746e0f66aSChris Mason int i; 39846e0f66aSChris Mason 39946e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 40046e0f66aSChris Mason if (uuid[i]) 40146e0f66aSChris Mason return 0; 40246e0f66aSChris Mason } 40346e0f66aSChris Mason return 1; 404dd5f9615SStefan Behrens } 405dd5f9615SStefan Behrens 406d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 407cb8e7090SChristoph Hellwig struct dentry *dentry, 40852f75f4fSDavid Sterba const char *name, int namelen, 4096f72c7e2SArne Jansen u64 *async_transid, 4108696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 411f46b5a66SChristoph Hellwig { 4120b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 413f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 414f46b5a66SChristoph Hellwig struct btrfs_key key; 41549a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 416f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 417f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 418d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 41976dda93cSYan, Zheng struct btrfs_root *new_root; 420d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 421c2050a45SDeepa Dinamani struct timespec cur_time = current_time(dir); 4225662344bSTsutomu Itoh struct inode *inode; 423f46b5a66SChristoph Hellwig int ret; 424f46b5a66SChristoph Hellwig int err; 425f46b5a66SChristoph Hellwig u64 objectid; 426f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 4273de4586cSChris Mason u64 index = 0; 428d5c12070SMiao Xie u64 qgroup_reserved; 4298ea05e3aSAlexander Block uuid_le new_uuid; 430f46b5a66SChristoph Hellwig 43149a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 43249a3c4d9SDavid Sterba if (!root_item) 43349a3c4d9SDavid Sterba return -ENOMEM; 43449a3c4d9SDavid Sterba 4350b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 4362fbe8c8aSAl Viro if (ret) 43749a3c4d9SDavid Sterba goto fail_free; 4386a912213SJosef Bacik 439e09fe2d2SQu Wenruo /* 440e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 44101327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 442e09fe2d2SQu Wenruo */ 44349a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 44449a3c4d9SDavid Sterba ret = -ENOSPC; 44549a3c4d9SDavid Sterba goto fail_free; 44649a3c4d9SDavid Sterba } 447e09fe2d2SQu Wenruo 448d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 4499ed74f2dSJosef Bacik /* 450d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 451d5c12070SMiao Xie * of create_snapshot(). 4529ed74f2dSJosef Bacik */ 453d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 454dd5f9615SStefan Behrens 8, &qgroup_reserved, false); 455d5c12070SMiao Xie if (ret) 45649a3c4d9SDavid Sterba goto fail_free; 457f46b5a66SChristoph Hellwig 458d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 459d5c12070SMiao Xie if (IS_ERR(trans)) { 460d5c12070SMiao Xie ret = PTR_ERR(trans); 4617775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 46249a3c4d9SDavid Sterba goto fail_free; 463d5c12070SMiao Xie } 464d5c12070SMiao Xie trans->block_rsv = &block_rsv; 465d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 466f46b5a66SChristoph Hellwig 4670b246afaSJeff Mahoney ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); 4686f72c7e2SArne Jansen if (ret) 4696f72c7e2SArne Jansen goto fail; 4706f72c7e2SArne Jansen 4714d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 4728e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 4738e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 4748e8a1e31SJosef Bacik goto fail; 4758e8a1e31SJosef Bacik } 476f46b5a66SChristoph Hellwig 477b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 478f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 479f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 4805d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 481f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 482f46b5a66SChristoph Hellwig 4830b246afaSJeff Mahoney write_extent_buffer_fsid(leaf, fs_info->fsid); 4840b246afaSJeff Mahoney write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 485f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 486f46b5a66SChristoph Hellwig 48749a3c4d9SDavid Sterba inode_item = &root_item->inode; 4883cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 4893cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 4903cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 491da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 4920b246afaSJeff Mahoney fs_info->nodesize); 4933cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 494f46b5a66SChristoph Hellwig 49549a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 49649a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 4973cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 49808fe4db1SLi Zefan 49949a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 50049a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 50149a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 50249a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 50349a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 50449a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 505f46b5a66SChristoph Hellwig 50649a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 50749a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 5088ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 50949a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 51049a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 51149a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 51249a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 51349a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 51449a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 515f46b5a66SChristoph Hellwig 516925baeddSChris Mason btrfs_tree_unlock(leaf); 517f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 518f46b5a66SChristoph Hellwig leaf = NULL; 519f46b5a66SChristoph Hellwig 52049a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 521f46b5a66SChristoph Hellwig 522f46b5a66SChristoph Hellwig key.objectid = objectid; 5235d4f98a2SYan Zheng key.offset = 0; 524962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 5250b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 52649a3c4d9SDavid Sterba root_item); 527f46b5a66SChristoph Hellwig if (ret) 528f46b5a66SChristoph Hellwig goto fail; 529f46b5a66SChristoph Hellwig 53076dda93cSYan, Zheng key.offset = (u64)-1; 5310b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 53279787eaaSJeff Mahoney if (IS_ERR(new_root)) { 53379787eaaSJeff Mahoney ret = PTR_ERR(new_root); 53466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 53579787eaaSJeff Mahoney goto fail; 53679787eaaSJeff Mahoney } 53776dda93cSYan, Zheng 53876dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 53976dda93cSYan, Zheng 54063541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 541ce598979SMark Fasheh if (ret) { 542ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 54366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 544ce598979SMark Fasheh goto fail; 545ce598979SMark Fasheh } 546ce598979SMark Fasheh 547f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 548f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 549f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 550f32e48e9SChandan Rajendra 551f46b5a66SChristoph Hellwig /* 552f46b5a66SChristoph Hellwig * insert the directory item 553f46b5a66SChristoph Hellwig */ 554877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 55579787eaaSJeff Mahoney if (ret) { 55666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 55779787eaaSJeff Mahoney goto fail; 55879787eaaSJeff Mahoney } 5593de4586cSChris Mason 5603de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 5618e7611cfSNikolay Borisov name, namelen, BTRFS_I(dir), &key, 5623de4586cSChris Mason BTRFS_FT_DIR, index); 56379787eaaSJeff Mahoney if (ret) { 56466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 565f46b5a66SChristoph Hellwig goto fail; 56679787eaaSJeff Mahoney } 5670660b5afSChris Mason 5686ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 56952c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 57052c26179SYan Zheng BUG_ON(ret); 57152c26179SYan Zheng 5720b246afaSJeff Mahoney ret = btrfs_add_root_ref(trans, fs_info, 5734df27c4dSYan, Zheng objectid, root->root_key.objectid, 5744a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 5750660b5afSChris Mason BUG_ON(ret); 5760660b5afSChris Mason 5770b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid, 5786bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 579dd5f9615SStefan Behrens if (ret) 58066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 581dd5f9615SStefan Behrens 582f46b5a66SChristoph Hellwig fail: 58349a3c4d9SDavid Sterba kfree(root_item); 584d5c12070SMiao Xie trans->block_rsv = NULL; 585d5c12070SMiao Xie trans->bytes_reserved = 0; 5867775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 587de6e8200SLiu Bo 58872fd032eSSage Weil if (async_transid) { 58972fd032eSSage Weil *async_transid = trans->transid; 5903a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 59100d71c9cSMiao Xie if (err) 5923a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 59372fd032eSSage Weil } else { 5943a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 59572fd032eSSage Weil } 596f46b5a66SChristoph Hellwig if (err && !ret) 597f46b5a66SChristoph Hellwig ret = err; 5981a65e24bSChris Mason 5995662344bSTsutomu Itoh if (!ret) { 6005662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 601de6e8200SLiu Bo if (IS_ERR(inode)) 602de6e8200SLiu Bo return PTR_ERR(inode); 6035662344bSTsutomu Itoh d_instantiate(dentry, inode); 6045662344bSTsutomu Itoh } 605f46b5a66SChristoph Hellwig return ret; 60649a3c4d9SDavid Sterba 60749a3c4d9SDavid Sterba fail_free: 60849a3c4d9SDavid Sterba kfree(root_item); 60949a3c4d9SDavid Sterba return ret; 610f46b5a66SChristoph Hellwig } 611f46b5a66SChristoph Hellwig 612e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 61361d7e4cbSDavid Sterba struct dentry *dentry, 614e9662f70SMiao Xie u64 *async_transid, bool readonly, 615e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 616f46b5a66SChristoph Hellwig { 6170b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 6182e4bfab9SYan, Zheng struct inode *inode; 619f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 620f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 6212e4bfab9SYan, Zheng int ret; 622f46b5a66SChristoph Hellwig 62327cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 624f46b5a66SChristoph Hellwig return -EINVAL; 625f46b5a66SChristoph Hellwig 62623269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 627a1ee7362SDavid Sterba if (!pending_snapshot) 628a1ee7362SDavid Sterba return -ENOMEM; 629a1ee7362SDavid Sterba 630b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 63123269bf5SDavid Sterba GFP_KERNEL); 6328546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 6338546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 634b0c0ea63SDavid Sterba ret = -ENOMEM; 635b0c0ea63SDavid Sterba goto free_pending; 636b0c0ea63SDavid Sterba } 637b0c0ea63SDavid Sterba 638ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 6394e857c58SPeter Zijlstra smp_mb__after_atomic(); 64045bac0f3SLiu Bo /* wait for no snapshot writes */ 64145bac0f3SLiu Bo wait_event(root->subv_writers->wait, 64245bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 6438257b2dcSMiao Xie 64476f32e24SNikolay Borisov ret = btrfs_start_delalloc_inodes(root); 6456a03843dSMiao Xie if (ret) 646a1ee7362SDavid Sterba goto dec_and_free; 6476a03843dSMiao Xie 6486374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 6496a03843dSMiao Xie 65066d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 65166d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 652d5c12070SMiao Xie /* 653d5c12070SMiao Xie * 1 - parent dir inode 654d5c12070SMiao Xie * 2 - dir entries 655d5c12070SMiao Xie * 1 - root item 656d5c12070SMiao Xie * 2 - root ref/backref 657d5c12070SMiao Xie * 1 - root of snapshot 658dd5f9615SStefan Behrens * 1 - UUID item 659d5c12070SMiao Xie */ 660d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 661dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 662ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 663ee3441b4SJeff Mahoney false); 664d5c12070SMiao Xie if (ret) 665a1ee7362SDavid Sterba goto dec_and_free; 666d5c12070SMiao Xie 667a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 668a22285a6SYan, Zheng pending_snapshot->root = root; 669b83cc969SLi Zefan pending_snapshot->readonly = readonly; 670e9662f70SMiao Xie pending_snapshot->dir = dir; 6718696c533SMiao Xie pending_snapshot->inherit = inherit; 672a22285a6SYan, Zheng 673d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 674a22285a6SYan, Zheng if (IS_ERR(trans)) { 675a22285a6SYan, Zheng ret = PTR_ERR(trans); 676a22285a6SYan, Zheng goto fail; 677a22285a6SYan, Zheng } 678a22285a6SYan, Zheng 6790b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 680a22285a6SYan, Zheng list_add(&pending_snapshot->list, 681a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 6820b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 68372fd032eSSage Weil if (async_transid) { 68472fd032eSSage Weil *async_transid = trans->transid; 6853a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 68600d71c9cSMiao Xie if (ret) 6873a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 68872fd032eSSage Weil } else { 6893a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 69072fd032eSSage Weil } 691aec8030aSMiao Xie if (ret) 692c37b2b62SJosef Bacik goto fail; 693a22285a6SYan, Zheng 694a22285a6SYan, Zheng ret = pending_snapshot->error; 695f46b5a66SChristoph Hellwig if (ret) 6962e4bfab9SYan, Zheng goto fail; 697f46b5a66SChristoph Hellwig 698d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 699d3797308SChris Mason if (ret) 700d3797308SChris Mason goto fail; 701d3797308SChris Mason 7022b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7032e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7042e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7052e4bfab9SYan, Zheng goto fail; 7062e4bfab9SYan, Zheng } 7075662344bSTsutomu Itoh 7082e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7092e4bfab9SYan, Zheng ret = 0; 7102e4bfab9SYan, Zheng fail: 7117775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 712a1ee7362SDavid Sterba dec_and_free: 713ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 7144625956aSPeter Zijlstra wake_up_var(&root->will_be_snapshotted); 715b0c0ea63SDavid Sterba free_pending: 716b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 7178546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 718a1ee7362SDavid Sterba kfree(pending_snapshot); 719a1ee7362SDavid Sterba 720f46b5a66SChristoph Hellwig return ret; 721f46b5a66SChristoph Hellwig } 722f46b5a66SChristoph Hellwig 7234260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7244260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7254260f7c7SSage Weil * whether the type of victim is right. 7264260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7274260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7284260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7294260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7304260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7314260f7c7SSage Weil * a. be owner of dir, or 7324260f7c7SSage Weil * b. be owner of victim, or 7334260f7c7SSage Weil * c. have CAP_FOWNER capability 73401327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 7354260f7c7SSage Weil * links pointing to it. 7364260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7374260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7384260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7394260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 7404260f7c7SSage Weil * nfs_async_unlink(). 7414260f7c7SSage Weil */ 7424260f7c7SSage Weil 7434260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 7444260f7c7SSage Weil { 7454260f7c7SSage Weil int error; 7464260f7c7SSage Weil 7472b0143b5SDavid Howells if (d_really_is_negative(victim)) 7484260f7c7SSage Weil return -ENOENT; 7494260f7c7SSage Weil 7502b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 7514fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 7524260f7c7SSage Weil 7534260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 7544260f7c7SSage Weil if (error) 7554260f7c7SSage Weil return error; 7564260f7c7SSage Weil if (IS_APPEND(dir)) 7574260f7c7SSage Weil return -EPERM; 7582b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 7592b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 7604260f7c7SSage Weil return -EPERM; 7614260f7c7SSage Weil if (isdir) { 762e36cb0b8SDavid Howells if (!d_is_dir(victim)) 7634260f7c7SSage Weil return -ENOTDIR; 7644260f7c7SSage Weil if (IS_ROOT(victim)) 7654260f7c7SSage Weil return -EBUSY; 766e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 7674260f7c7SSage Weil return -EISDIR; 7684260f7c7SSage Weil if (IS_DEADDIR(dir)) 7694260f7c7SSage Weil return -ENOENT; 7704260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 7714260f7c7SSage Weil return -EBUSY; 7724260f7c7SSage Weil return 0; 7734260f7c7SSage Weil } 7744260f7c7SSage Weil 775cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 776cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 777cb8e7090SChristoph Hellwig { 7782b0143b5SDavid Howells if (d_really_is_positive(child)) 779cb8e7090SChristoph Hellwig return -EEXIST; 780cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 781cb8e7090SChristoph Hellwig return -ENOENT; 782cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 783cb8e7090SChristoph Hellwig } 784cb8e7090SChristoph Hellwig 785cb8e7090SChristoph Hellwig /* 786cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 787cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 788cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 789cb8e7090SChristoph Hellwig */ 79092872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 79152f75f4fSDavid Sterba const char *name, int namelen, 79272fd032eSSage Weil struct btrfs_root *snap_src, 7936f72c7e2SArne Jansen u64 *async_transid, bool readonly, 7948696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 795cb8e7090SChristoph Hellwig { 7962b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 7970b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 798cb8e7090SChristoph Hellwig struct dentry *dentry; 799cb8e7090SChristoph Hellwig int error; 800cb8e7090SChristoph Hellwig 80100235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 80200235411SAl Viro if (error == -EINTR) 80300235411SAl Viro return error; 804cb8e7090SChristoph Hellwig 805cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 806cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 807cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 808cb8e7090SChristoph Hellwig goto out_unlock; 809cb8e7090SChristoph Hellwig 81076dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 811cb8e7090SChristoph Hellwig if (error) 812a874a63eSLiu Bo goto out_dput; 813cb8e7090SChristoph Hellwig 8149c52057cSChris Mason /* 8159c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8169c52057cSChris Mason * check for them now when we can safely fail 8179c52057cSChris Mason */ 8189c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8199c52057cSChris Mason dir->i_ino, name, 8209c52057cSChris Mason namelen); 8219c52057cSChris Mason if (error) 8229c52057cSChris Mason goto out_dput; 8239c52057cSChris Mason 8240b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 82576dda93cSYan, Zheng 82676dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 82776dda93cSYan, Zheng goto out_up_read; 82876dda93cSYan, Zheng 8293de4586cSChris Mason if (snap_src) { 83061d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 8316f72c7e2SArne Jansen async_transid, readonly, inherit); 8323de4586cSChris Mason } else { 833d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 834d5c12070SMiao Xie async_transid, inherit); 8353de4586cSChris Mason } 83676dda93cSYan, Zheng if (!error) 83776dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 83876dda93cSYan, Zheng out_up_read: 8390b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 840cb8e7090SChristoph Hellwig out_dput: 841cb8e7090SChristoph Hellwig dput(dentry); 842cb8e7090SChristoph Hellwig out_unlock: 8435955102cSAl Viro inode_unlock(dir); 844cb8e7090SChristoph Hellwig return error; 845cb8e7090SChristoph Hellwig } 846cb8e7090SChristoph Hellwig 8474cb5300bSChris Mason /* 8484cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 8494cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 8504cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 8514cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 8524cb5300bSChris Mason * part of the file 8534cb5300bSChris Mason */ 854aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 8554cb5300bSChris Mason { 8564cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 8574cb5300bSChris Mason struct extent_map *em = NULL; 8584cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 8594cb5300bSChris Mason u64 end; 8604cb5300bSChris Mason 8614cb5300bSChris Mason read_lock(&em_tree->lock); 86209cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 8634cb5300bSChris Mason read_unlock(&em_tree->lock); 8644cb5300bSChris Mason 8654cb5300bSChris Mason if (em) { 8664cb5300bSChris Mason end = extent_map_end(em); 8674cb5300bSChris Mason free_extent_map(em); 8684cb5300bSChris Mason if (end - offset > thresh) 8694cb5300bSChris Mason return 0; 8704cb5300bSChris Mason } 8714cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 8724cb5300bSChris Mason thresh /= 2; 8734cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 8744cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 8754cb5300bSChris Mason if (end >= thresh) 8764cb5300bSChris Mason return 0; 8774cb5300bSChris Mason return 1; 8784cb5300bSChris Mason } 8794cb5300bSChris Mason 8804cb5300bSChris Mason /* 8814cb5300bSChris Mason * helper function to walk through a file and find extents 8824cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 8834cb5300bSChris Mason * 8844cb5300bSChris Mason * This is used by the defragging code to find new and small 8854cb5300bSChris Mason * extents 8864cb5300bSChris Mason */ 8874cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 8884cb5300bSChris Mason struct inode *inode, u64 newer_than, 889aab110abSDavid Sterba u64 *off, u32 thresh) 8904cb5300bSChris Mason { 8914cb5300bSChris Mason struct btrfs_path *path; 8924cb5300bSChris Mason struct btrfs_key min_key; 8934cb5300bSChris Mason struct extent_buffer *leaf; 8944cb5300bSChris Mason struct btrfs_file_extent_item *extent; 8954cb5300bSChris Mason int type; 8964cb5300bSChris Mason int ret; 8974a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 8984cb5300bSChris Mason 8994cb5300bSChris Mason path = btrfs_alloc_path(); 9004cb5300bSChris Mason if (!path) 9014cb5300bSChris Mason return -ENOMEM; 9024cb5300bSChris Mason 903a4689d2bSDavid Sterba min_key.objectid = ino; 9044cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9054cb5300bSChris Mason min_key.offset = *off; 9064cb5300bSChris Mason 9074cb5300bSChris Mason while (1) { 9086174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9094cb5300bSChris Mason if (ret != 0) 9104cb5300bSChris Mason goto none; 911f094c9bdSFilipe Manana process_slot: 912a4689d2bSDavid Sterba if (min_key.objectid != ino) 9134cb5300bSChris Mason goto none; 9144cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9154cb5300bSChris Mason goto none; 9164cb5300bSChris Mason 9174cb5300bSChris Mason leaf = path->nodes[0]; 9184cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9194cb5300bSChris Mason struct btrfs_file_extent_item); 9204cb5300bSChris Mason 9214cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9224cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9234cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9244cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9254cb5300bSChris Mason *off = min_key.offset; 9264cb5300bSChris Mason btrfs_free_path(path); 9274cb5300bSChris Mason return 0; 9284cb5300bSChris Mason } 9294cb5300bSChris Mason 930f094c9bdSFilipe Manana path->slots[0]++; 931f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 932f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 933f094c9bdSFilipe Manana goto process_slot; 934f094c9bdSFilipe Manana } 935f094c9bdSFilipe Manana 9364cb5300bSChris Mason if (min_key.offset == (u64)-1) 9374cb5300bSChris Mason goto none; 9384cb5300bSChris Mason 9394cb5300bSChris Mason min_key.offset++; 9404cb5300bSChris Mason btrfs_release_path(path); 9414cb5300bSChris Mason } 9424cb5300bSChris Mason none: 9434cb5300bSChris Mason btrfs_free_path(path); 9444cb5300bSChris Mason return -ENOENT; 9454cb5300bSChris Mason } 9464cb5300bSChris Mason 9476c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 94817ce6ef8SLiu Bo { 94917ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 950940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9516c282eb4SLi Zefan struct extent_map *em; 95209cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 953940100a4SChris Mason 954940100a4SChris Mason /* 955940100a4SChris Mason * hopefully we have this extent in the tree already, try without 956940100a4SChris Mason * the full extent lock 957940100a4SChris Mason */ 958940100a4SChris Mason read_lock(&em_tree->lock); 959940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 960940100a4SChris Mason read_unlock(&em_tree->lock); 961940100a4SChris Mason 962940100a4SChris Mason if (!em) { 963308d9800SFilipe Manana struct extent_state *cached = NULL; 964308d9800SFilipe Manana u64 end = start + len - 1; 965308d9800SFilipe Manana 966940100a4SChris Mason /* get the big lock and read metadata off disk */ 967ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 968fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 969e43bbe5eSDavid Sterba unlock_extent_cached(io_tree, start, end, &cached); 970940100a4SChris Mason 9716cf8bfbfSDan Carpenter if (IS_ERR(em)) 9726c282eb4SLi Zefan return NULL; 973940100a4SChris Mason } 974940100a4SChris Mason 9756c282eb4SLi Zefan return em; 9766c282eb4SLi Zefan } 9776c282eb4SLi Zefan 9786c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 9796c282eb4SLi Zefan { 9806c282eb4SLi Zefan struct extent_map *next; 9816c282eb4SLi Zefan bool ret = true; 9826c282eb4SLi Zefan 9836c282eb4SLi Zefan /* this is the last extent */ 9846c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 9856c282eb4SLi Zefan return false; 9866c282eb4SLi Zefan 9876c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 988e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 989e9512d72SChris Mason ret = false; 990e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 991ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 9926c282eb4SLi Zefan ret = false; 9936c282eb4SLi Zefan 9946c282eb4SLi Zefan free_extent_map(next); 9956c282eb4SLi Zefan return ret; 9966c282eb4SLi Zefan } 9976c282eb4SLi Zefan 998aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 999a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1000a43a2111SAndrew Mahone int compress) 10016c282eb4SLi Zefan { 10026c282eb4SLi Zefan struct extent_map *em; 10036c282eb4SLi Zefan int ret = 1; 10046c282eb4SLi Zefan bool next_mergeable = true; 10054a3560c4SLiu Bo bool prev_mergeable = true; 10066c282eb4SLi Zefan 10076c282eb4SLi Zefan /* 10086c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10096c282eb4SLi Zefan * defragging it 10106c282eb4SLi Zefan */ 10116c282eb4SLi Zefan if (start < *defrag_end) 10126c282eb4SLi Zefan return 1; 10136c282eb4SLi Zefan 10146c282eb4SLi Zefan *skip = 0; 10156c282eb4SLi Zefan 10166c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10176c282eb4SLi Zefan if (!em) 10186c282eb4SLi Zefan return 0; 10196c282eb4SLi Zefan 1020940100a4SChris Mason /* this will cover holes, and inline extents */ 102117ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1022940100a4SChris Mason ret = 0; 102317ce6ef8SLiu Bo goto out; 102417ce6ef8SLiu Bo } 102517ce6ef8SLiu Bo 10264a3560c4SLiu Bo if (!*defrag_end) 10274a3560c4SLiu Bo prev_mergeable = false; 10284a3560c4SLiu Bo 10296c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1030940100a4SChris Mason /* 10316c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10326c282eb4SLi Zefan * real extent, don't bother defragging it 1033940100a4SChris Mason */ 1034a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10354a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1036940100a4SChris Mason ret = 0; 103717ce6ef8SLiu Bo out: 1038940100a4SChris Mason /* 1039940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1040940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1041940100a4SChris Mason * so that the next tiny extent will force a defrag. 1042940100a4SChris Mason * 1043940100a4SChris Mason * The end result of this is that tiny extents before a single big 1044940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1045940100a4SChris Mason */ 1046940100a4SChris Mason if (ret) { 1047940100a4SChris Mason *defrag_end = extent_map_end(em); 1048940100a4SChris Mason } else { 1049940100a4SChris Mason *last_len = 0; 1050940100a4SChris Mason *skip = extent_map_end(em); 1051940100a4SChris Mason *defrag_end = 0; 1052940100a4SChris Mason } 1053940100a4SChris Mason 1054940100a4SChris Mason free_extent_map(em); 1055940100a4SChris Mason return ret; 1056940100a4SChris Mason } 1057940100a4SChris Mason 10584cb5300bSChris Mason /* 10594cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 10604cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 10614cb5300bSChris Mason * to COW and defrag. 10624cb5300bSChris Mason * 10634cb5300bSChris Mason * It also makes sure the delalloc code has enough 10644cb5300bSChris Mason * dirty data to avoid making new small extents as part 10654cb5300bSChris Mason * of the defrag 10664cb5300bSChris Mason * 10674cb5300bSChris Mason * It's a good idea to start RA on this range 10684cb5300bSChris Mason * before calling this. 10694cb5300bSChris Mason */ 10704cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 10714cb5300bSChris Mason struct page **pages, 10724cb5300bSChris Mason unsigned long start_index, 1073c41570c9SJustin Maggard unsigned long num_pages) 1074f46b5a66SChristoph Hellwig { 10754cb5300bSChris Mason unsigned long file_end; 10764cb5300bSChris Mason u64 isize = i_size_read(inode); 1077f46b5a66SChristoph Hellwig u64 page_start; 1078f46b5a66SChristoph Hellwig u64 page_end; 10791f12bd06SLiu Bo u64 page_cnt; 10804cb5300bSChris Mason int ret; 10814cb5300bSChris Mason int i; 10824cb5300bSChris Mason int i_done; 10834cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 10844cb5300bSChris Mason struct extent_state *cached_state = NULL; 1085600a45e1SMiao Xie struct extent_io_tree *tree; 1086364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 10873b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 10884cb5300bSChris Mason 108909cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 10901f12bd06SLiu Bo if (!isize || start_index > file_end) 10911f12bd06SLiu Bo return 0; 10921f12bd06SLiu Bo 10931f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 10944cb5300bSChris Mason 1095364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 109609cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 109709cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 10984cb5300bSChris Mason if (ret) 10994cb5300bSChris Mason return ret; 11004cb5300bSChris Mason i_done = 0; 1101600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11024cb5300bSChris Mason 11034cb5300bSChris Mason /* step one, lock all the pages */ 11041f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11054cb5300bSChris Mason struct page *page; 1106600a45e1SMiao Xie again: 1107a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11083b16a4e3SJosef Bacik start_index + i, mask); 11094cb5300bSChris Mason if (!page) 11104cb5300bSChris Mason break; 11114cb5300bSChris Mason 1112600a45e1SMiao Xie page_start = page_offset(page); 111309cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1114600a45e1SMiao Xie while (1) { 1115308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1116ff13db41SDavid Sterba &cached_state); 1117600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1118600a45e1SMiao Xie page_start); 1119308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1120e43bbe5eSDavid Sterba &cached_state); 1121600a45e1SMiao Xie if (!ordered) 1122600a45e1SMiao Xie break; 1123600a45e1SMiao Xie 1124600a45e1SMiao Xie unlock_page(page); 1125600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1126600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1127600a45e1SMiao Xie lock_page(page); 11281f12bd06SLiu Bo /* 11291f12bd06SLiu Bo * we unlocked the page above, so we need check if 11301f12bd06SLiu Bo * it was released or not. 11311f12bd06SLiu Bo */ 11321f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11331f12bd06SLiu Bo unlock_page(page); 113409cbfeafSKirill A. Shutemov put_page(page); 11351f12bd06SLiu Bo goto again; 11361f12bd06SLiu Bo } 1137600a45e1SMiao Xie } 1138600a45e1SMiao Xie 11394cb5300bSChris Mason if (!PageUptodate(page)) { 11404cb5300bSChris Mason btrfs_readpage(NULL, page); 11414cb5300bSChris Mason lock_page(page); 11424cb5300bSChris Mason if (!PageUptodate(page)) { 11434cb5300bSChris Mason unlock_page(page); 114409cbfeafSKirill A. Shutemov put_page(page); 11454cb5300bSChris Mason ret = -EIO; 11464cb5300bSChris Mason break; 11474cb5300bSChris Mason } 11484cb5300bSChris Mason } 1149600a45e1SMiao Xie 1150600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1151600a45e1SMiao Xie unlock_page(page); 115209cbfeafSKirill A. Shutemov put_page(page); 1153600a45e1SMiao Xie goto again; 1154600a45e1SMiao Xie } 1155600a45e1SMiao Xie 11564cb5300bSChris Mason pages[i] = page; 11574cb5300bSChris Mason i_done++; 11584cb5300bSChris Mason } 11594cb5300bSChris Mason if (!i_done || ret) 11604cb5300bSChris Mason goto out; 11614cb5300bSChris Mason 11621751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 11634cb5300bSChris Mason goto out; 11644cb5300bSChris Mason 11654cb5300bSChris Mason /* 11664cb5300bSChris Mason * so now we have a nice long stream of locked 11674cb5300bSChris Mason * and up to date pages, lets wait on them 11684cb5300bSChris Mason */ 11694cb5300bSChris Mason for (i = 0; i < i_done; i++) 11704cb5300bSChris Mason wait_on_page_writeback(pages[i]); 11714cb5300bSChris Mason 11724cb5300bSChris Mason page_start = page_offset(pages[0]); 117309cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 11744cb5300bSChris Mason 11754cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1176ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 11774cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 11784cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 11799e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1180ae0f1625SDavid Sterba &cached_state); 11814cb5300bSChris Mason 11821f12bd06SLiu Bo if (i_done != page_cnt) { 11839e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 11849e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 11859e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1186bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 118709cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 118843b18595SQu Wenruo (page_cnt - i_done) << PAGE_SHIFT, true); 11894cb5300bSChris Mason } 11904cb5300bSChris Mason 11914cb5300bSChris Mason 11929e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1193018ed4f7SDavid Sterba &cached_state); 11944cb5300bSChris Mason 11954cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1196e43bbe5eSDavid Sterba page_start, page_end - 1, &cached_state); 11974cb5300bSChris Mason 11984cb5300bSChris Mason for (i = 0; i < i_done; i++) { 11994cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12004cb5300bSChris Mason ClearPageChecked(pages[i]); 12014cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12024cb5300bSChris Mason set_page_dirty(pages[i]); 12034cb5300bSChris Mason unlock_page(pages[i]); 120409cbfeafSKirill A. Shutemov put_page(pages[i]); 12054cb5300bSChris Mason } 120643b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 120743b18595SQu Wenruo false); 1208364ecf36SQu Wenruo extent_changeset_free(data_reserved); 12094cb5300bSChris Mason return i_done; 12104cb5300bSChris Mason out: 12114cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12124cb5300bSChris Mason unlock_page(pages[i]); 121309cbfeafSKirill A. Shutemov put_page(pages[i]); 12144cb5300bSChris Mason } 1215bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 121609cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 121743b18595SQu Wenruo page_cnt << PAGE_SHIFT, true); 121843b18595SQu Wenruo btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, 121943b18595SQu Wenruo true); 1220364ecf36SQu Wenruo extent_changeset_free(data_reserved); 12214cb5300bSChris Mason return ret; 12224cb5300bSChris Mason 12234cb5300bSChris Mason } 12244cb5300bSChris Mason 12254cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12264cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12274cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12284cb5300bSChris Mason { 12290b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 12304cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12314cb5300bSChris Mason struct file_ra_state *ra = NULL; 12324cb5300bSChris Mason unsigned long last_index; 1233151a31b2SLi Zefan u64 isize = i_size_read(inode); 1234940100a4SChris Mason u64 last_len = 0; 1235940100a4SChris Mason u64 skip = 0; 1236940100a4SChris Mason u64 defrag_end = 0; 12374cb5300bSChris Mason u64 newer_off = range->start; 1238f46b5a66SChristoph Hellwig unsigned long i; 1239008873eaSLi Zefan unsigned long ra_index = 0; 1240f46b5a66SChristoph Hellwig int ret; 12414cb5300bSChris Mason int defrag_count = 0; 12421a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1243aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 124409cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1245c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1246ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 12474cb5300bSChris Mason struct page **pages = NULL; 12481e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 12494cb5300bSChris Mason 12500abd5b17SLiu Bo if (isize == 0) 12510abd5b17SLiu Bo return 0; 12520abd5b17SLiu Bo 12530abd5b17SLiu Bo if (range->start >= isize) 12540abd5b17SLiu Bo return -EINVAL; 12551a419d85SLi Zefan 12561e2ef46dSDavid Sterba if (do_compress) { 12571a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 12581a419d85SLi Zefan return -EINVAL; 12591a419d85SLi Zefan if (range->compress_type) 12601a419d85SLi Zefan compress_type = range->compress_type; 12611a419d85SLi Zefan } 1262f46b5a66SChristoph Hellwig 12630abd5b17SLiu Bo if (extent_thresh == 0) 1264ee22184bSByongho Lee extent_thresh = SZ_256K; 1265f46b5a66SChristoph Hellwig 12664cb5300bSChris Mason /* 12670a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 12680a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 12690a52d108SDavid Sterba * we don't error out. 12704cb5300bSChris Mason */ 12714cb5300bSChris Mason if (!file) { 127263e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 12730a52d108SDavid Sterba if (ra) 12744cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 12754cb5300bSChris Mason } else { 12764cb5300bSChris Mason ra = &file->f_ra; 12774cb5300bSChris Mason } 12784cb5300bSChris Mason 127963e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 12804cb5300bSChris Mason if (!pages) { 12814cb5300bSChris Mason ret = -ENOMEM; 12824cb5300bSChris Mason goto out_ra; 12834cb5300bSChris Mason } 12844cb5300bSChris Mason 12854cb5300bSChris Mason /* find the last page to defrag */ 12861e701a32SChris Mason if (range->start + range->len > range->start) { 1287151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 128809cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 12891e701a32SChris Mason } else { 129009cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 12911e701a32SChris Mason } 12921e701a32SChris Mason 12934cb5300bSChris Mason if (newer_than) { 12944cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1295ee22184bSByongho Lee &newer_off, SZ_64K); 12964cb5300bSChris Mason if (!ret) { 12974cb5300bSChris Mason range->start = newer_off; 12984cb5300bSChris Mason /* 12994cb5300bSChris Mason * we always align our defrag to help keep 13004cb5300bSChris Mason * the extents in the file evenly spaced 13014cb5300bSChris Mason */ 130209cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 13034cb5300bSChris Mason } else 13044cb5300bSChris Mason goto out_ra; 13054cb5300bSChris Mason } else { 130609cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 13074cb5300bSChris Mason } 13084cb5300bSChris Mason if (!max_to_defrag) 1309070034bdSchandan max_to_defrag = last_index - i + 1; 13104cb5300bSChris Mason 13112a0f7f57SLi Zefan /* 13122a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13132a0f7f57SLi Zefan * written sequentially. 13142a0f7f57SLi Zefan */ 13152a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13162a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13172a0f7f57SLi Zefan 1318f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 131909cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 13204cb5300bSChris Mason /* 13214cb5300bSChris Mason * make sure we stop running if someone unmounts 13224cb5300bSChris Mason * the FS 13234cb5300bSChris Mason */ 13241751e8a6SLinus Torvalds if (!(inode->i_sb->s_flags & SB_ACTIVE)) 13254cb5300bSChris Mason break; 13264cb5300bSChris Mason 13270b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 13280b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1329210549ebSDavid Sterba ret = -EAGAIN; 1330210549ebSDavid Sterba break; 1331210549ebSDavid Sterba } 1332210549ebSDavid Sterba 133309cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 13346c282eb4SLi Zefan extent_thresh, &last_len, &skip, 13351e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1336940100a4SChris Mason unsigned long next; 1337940100a4SChris Mason /* 1338940100a4SChris Mason * the should_defrag function tells us how much to skip 1339940100a4SChris Mason * bump our counter by the suggested amount 1340940100a4SChris Mason */ 134109cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1342940100a4SChris Mason i = max(i + 1, next); 1343940100a4SChris Mason continue; 1344940100a4SChris Mason } 1345008873eaSLi Zefan 1346008873eaSLi Zefan if (!newer_than) { 134709cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 134809cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1349008873eaSLi Zefan cluster = min(cluster, max_cluster); 1350008873eaSLi Zefan } else { 1351008873eaSLi Zefan cluster = max_cluster; 1352008873eaSLi Zefan } 1353008873eaSLi Zefan 1354008873eaSLi Zefan if (i + cluster > ra_index) { 1355008873eaSLi Zefan ra_index = max(i, ra_index); 13560a52d108SDavid Sterba if (ra) 1357d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1358d3c0bab5SDavid Sterba file, ra_index, cluster); 1359e4826a5bSchandan ra_index += cluster; 1360008873eaSLi Zefan } 13614cb5300bSChris Mason 13625955102cSAl Viro inode_lock(inode); 13631e2ef46dSDavid Sterba if (do_compress) 1364eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1365008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1366ecb8bea8SLiu Bo if (ret < 0) { 13675955102cSAl Viro inode_unlock(inode); 13684cb5300bSChris Mason goto out_ra; 1369ecb8bea8SLiu Bo } 13704cb5300bSChris Mason 13714cb5300bSChris Mason defrag_count += ret; 1372d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 13735955102cSAl Viro inode_unlock(inode); 13744cb5300bSChris Mason 13754cb5300bSChris Mason if (newer_than) { 13764cb5300bSChris Mason if (newer_off == (u64)-1) 13774cb5300bSChris Mason break; 13784cb5300bSChris Mason 1379e1f041e1SLiu Bo if (ret > 0) 1380e1f041e1SLiu Bo i += ret; 1381e1f041e1SLiu Bo 13824cb5300bSChris Mason newer_off = max(newer_off + 1, 138309cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 13844cb5300bSChris Mason 1385ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1386ee22184bSByongho Lee &newer_off, SZ_64K); 13874cb5300bSChris Mason if (!ret) { 13884cb5300bSChris Mason range->start = newer_off; 138909cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 13904cb5300bSChris Mason } else { 13914cb5300bSChris Mason break; 1392940100a4SChris Mason } 13934cb5300bSChris Mason } else { 1394008873eaSLi Zefan if (ret > 0) { 1395cbcc8326SLi Zefan i += ret; 139609cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1397008873eaSLi Zefan } else { 1398940100a4SChris Mason i++; 1399008873eaSLi Zefan last_len = 0; 1400008873eaSLi Zefan } 1401f46b5a66SChristoph Hellwig } 14024cb5300bSChris Mason } 1403f46b5a66SChristoph Hellwig 1404dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14051e701a32SChris Mason filemap_flush(inode->i_mapping); 1406dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1407dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1408dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1409dec8ef90SFilipe Manana } 14101e701a32SChris Mason 14111a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14120b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 14135c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 14145c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 14151a419d85SLi Zefan } 14161a419d85SLi Zefan 141760ccf82fSDiego Calleja ret = defrag_count; 1418940100a4SChris Mason 14194cb5300bSChris Mason out_ra: 14201e2ef46dSDavid Sterba if (do_compress) { 14215955102cSAl Viro inode_lock(inode); 1422eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 14235955102cSAl Viro inode_unlock(inode); 1424633085c7SFilipe David Borba Manana } 14254cb5300bSChris Mason if (!file) 14264cb5300bSChris Mason kfree(ra); 14274cb5300bSChris Mason kfree(pages); 1428940100a4SChris Mason return ret; 1429f46b5a66SChristoph Hellwig } 1430f46b5a66SChristoph Hellwig 1431198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 143276dda93cSYan, Zheng void __user *arg) 1433f46b5a66SChristoph Hellwig { 14340b246afaSJeff Mahoney struct inode *inode = file_inode(file); 14350b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1436f46b5a66SChristoph Hellwig u64 new_size; 1437f46b5a66SChristoph Hellwig u64 old_size; 1438f46b5a66SChristoph Hellwig u64 devid = 1; 14390b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1440f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1441f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1442f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1443f46b5a66SChristoph Hellwig char *sizestr; 14449a40f122SGui Hecheng char *retptr; 1445f46b5a66SChristoph Hellwig char *devstr = NULL; 1446f46b5a66SChristoph Hellwig int ret = 0; 1447f46b5a66SChristoph Hellwig int mod = 0; 1448f46b5a66SChristoph Hellwig 1449e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1450e441d54dSChris Mason return -EPERM; 1451e441d54dSChris Mason 1452198605a8SMiao Xie ret = mnt_want_write_file(file); 1453198605a8SMiao Xie if (ret) 1454198605a8SMiao Xie return ret; 1455198605a8SMiao Xie 1456171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 145797547676SMiao Xie mnt_drop_write_file(file); 1458e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1459c9e9f97bSIlya Dryomov } 1460c9e9f97bSIlya Dryomov 1461dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1462c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1463c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1464c9e9f97bSIlya Dryomov goto out; 1465c9e9f97bSIlya Dryomov } 14665516e595SMark Fasheh 14675516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1468f46b5a66SChristoph Hellwig 1469f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1470f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1471f46b5a66SChristoph Hellwig if (devstr) { 1472f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1473f46b5a66SChristoph Hellwig *devstr = '\0'; 1474f46b5a66SChristoph Hellwig devstr = vol_args->name; 147558dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 147658dfae63SZhangZhen if (ret) 147758dfae63SZhangZhen goto out_free; 1478dfd79829SMiao Xie if (!devid) { 1479dfd79829SMiao Xie ret = -EINVAL; 1480dfd79829SMiao Xie goto out_free; 1481dfd79829SMiao Xie } 14820b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1483f46b5a66SChristoph Hellwig } 1484dba60f3fSMiao Xie 14850b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1486f46b5a66SChristoph Hellwig if (!device) { 14870b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1488c1c9ff7cSGeert Uytterhoeven devid); 1489dfd79829SMiao Xie ret = -ENODEV; 1490c9e9f97bSIlya Dryomov goto out_free; 1491f46b5a66SChristoph Hellwig } 1492dba60f3fSMiao Xie 1493ebbede42SAnand Jain if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 14940b246afaSJeff Mahoney btrfs_info(fs_info, 1495efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1496c1c9ff7cSGeert Uytterhoeven devid); 1497dfd79829SMiao Xie ret = -EPERM; 14984e42ae1bSLiu Bo goto out_free; 14994e42ae1bSLiu Bo } 15004e42ae1bSLiu Bo 1501f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1502f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1503f46b5a66SChristoph Hellwig else { 1504f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1505f46b5a66SChristoph Hellwig mod = -1; 1506f46b5a66SChristoph Hellwig sizestr++; 1507f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1508f46b5a66SChristoph Hellwig mod = 1; 1509f46b5a66SChristoph Hellwig sizestr++; 1510f46b5a66SChristoph Hellwig } 15119a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15129a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1513f46b5a66SChristoph Hellwig ret = -EINVAL; 1514c9e9f97bSIlya Dryomov goto out_free; 1515f46b5a66SChristoph Hellwig } 1516f46b5a66SChristoph Hellwig } 1517f46b5a66SChristoph Hellwig 1518401e29c1SAnand Jain if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1519dfd79829SMiao Xie ret = -EPERM; 152063a212abSStefan Behrens goto out_free; 152163a212abSStefan Behrens } 152263a212abSStefan Behrens 15237cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1524f46b5a66SChristoph Hellwig 1525f46b5a66SChristoph Hellwig if (mod < 0) { 1526f46b5a66SChristoph Hellwig if (new_size > old_size) { 1527f46b5a66SChristoph Hellwig ret = -EINVAL; 1528c9e9f97bSIlya Dryomov goto out_free; 1529f46b5a66SChristoph Hellwig } 1530f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1531f46b5a66SChristoph Hellwig } else if (mod > 0) { 1532eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1533902c68a4SGui Hecheng ret = -ERANGE; 1534eb8052e0SWenliang Fan goto out_free; 1535eb8052e0SWenliang Fan } 1536f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1537f46b5a66SChristoph Hellwig } 1538f46b5a66SChristoph Hellwig 1539ee22184bSByongho Lee if (new_size < SZ_256M) { 1540f46b5a66SChristoph Hellwig ret = -EINVAL; 1541c9e9f97bSIlya Dryomov goto out_free; 1542f46b5a66SChristoph Hellwig } 1543f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1544f46b5a66SChristoph Hellwig ret = -EFBIG; 1545c9e9f97bSIlya Dryomov goto out_free; 1546f46b5a66SChristoph Hellwig } 1547f46b5a66SChristoph Hellwig 154847f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1549f46b5a66SChristoph Hellwig 15500b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1551c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1552f46b5a66SChristoph Hellwig 1553f46b5a66SChristoph Hellwig if (new_size > old_size) { 1554a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 155598d5dc13STsutomu Itoh if (IS_ERR(trans)) { 155698d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1557c9e9f97bSIlya Dryomov goto out_free; 155898d5dc13STsutomu Itoh } 1559f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 15603a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1561ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1562f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 15630253f40eSjeff.liu } /* equal, nothing need to do */ 1564f46b5a66SChristoph Hellwig 1565c9e9f97bSIlya Dryomov out_free: 1566f46b5a66SChristoph Hellwig kfree(vol_args); 1567c9e9f97bSIlya Dryomov out: 1568171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 156918f39c41SIlya Dryomov mnt_drop_write_file(file); 1570f46b5a66SChristoph Hellwig return ret; 1571f46b5a66SChristoph Hellwig } 1572f46b5a66SChristoph Hellwig 157372fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 157452f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 15756f72c7e2SArne Jansen u64 *transid, bool readonly, 15768696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1577f46b5a66SChristoph Hellwig { 1578f46b5a66SChristoph Hellwig int namelen; 15793de4586cSChris Mason int ret = 0; 1580f46b5a66SChristoph Hellwig 1581325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1582325c50e3SJeff Mahoney return -ENOTDIR; 1583325c50e3SJeff Mahoney 1584a874a63eSLiu Bo ret = mnt_want_write_file(file); 1585a874a63eSLiu Bo if (ret) 1586a874a63eSLiu Bo goto out; 1587a874a63eSLiu Bo 158872fd032eSSage Weil namelen = strlen(name); 158972fd032eSSage Weil if (strchr(name, '/')) { 1590f46b5a66SChristoph Hellwig ret = -EINVAL; 1591a874a63eSLiu Bo goto out_drop_write; 1592f46b5a66SChristoph Hellwig } 1593f46b5a66SChristoph Hellwig 159416780cabSChris Mason if (name[0] == '.' && 159516780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 159616780cabSChris Mason ret = -EEXIST; 1597a874a63eSLiu Bo goto out_drop_write; 159816780cabSChris Mason } 159916780cabSChris Mason 16003de4586cSChris Mason if (subvol) { 160172fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16026f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1603cb8e7090SChristoph Hellwig } else { 16042903ff01SAl Viro struct fd src = fdget(fd); 16053de4586cSChris Mason struct inode *src_inode; 16062903ff01SAl Viro if (!src.file) { 16073de4586cSChris Mason ret = -EINVAL; 1608a874a63eSLiu Bo goto out_drop_write; 16093de4586cSChris Mason } 16103de4586cSChris Mason 1611496ad9aaSAl Viro src_inode = file_inode(src.file); 1612496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1613c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1614efe120a0SFrank Holton "Snapshot src from another FS"); 161523ad5b17SKusanagi Kouichi ret = -EXDEV; 1616d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1617d0242061SDavid Sterba /* 1618d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1619d0242061SDavid Sterba * are limited to own subvolumes only 1620d0242061SDavid Sterba */ 1621d0242061SDavid Sterba ret = -EPERM; 1622ecd18815SAl Viro } else { 162372fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 162472fd032eSSage Weil BTRFS_I(src_inode)->root, 16256f72c7e2SArne Jansen transid, readonly, inherit); 1626ecd18815SAl Viro } 16272903ff01SAl Viro fdput(src); 1628cb8e7090SChristoph Hellwig } 1629a874a63eSLiu Bo out_drop_write: 1630a874a63eSLiu Bo mnt_drop_write_file(file); 1631f46b5a66SChristoph Hellwig out: 163272fd032eSSage Weil return ret; 163372fd032eSSage Weil } 163472fd032eSSage Weil 163572fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1636fa0d2b9bSLi Zefan void __user *arg, int subvol) 163772fd032eSSage Weil { 1638fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 163972fd032eSSage Weil int ret; 164072fd032eSSage Weil 1641325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1642325c50e3SJeff Mahoney return -ENOTDIR; 1643325c50e3SJeff Mahoney 1644fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1645fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1646fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1647fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1648fa0d2b9bSLi Zefan 1649fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1650b83cc969SLi Zefan vol_args->fd, subvol, 16516f72c7e2SArne Jansen NULL, false, NULL); 1652fa0d2b9bSLi Zefan 1653fa0d2b9bSLi Zefan kfree(vol_args); 1654fa0d2b9bSLi Zefan return ret; 1655fa0d2b9bSLi Zefan } 1656fa0d2b9bSLi Zefan 1657fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1658fa0d2b9bSLi Zefan void __user *arg, int subvol) 1659fa0d2b9bSLi Zefan { 1660fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1661fa0d2b9bSLi Zefan int ret; 1662fdfb1e4fSLi Zefan u64 transid = 0; 1663fdfb1e4fSLi Zefan u64 *ptr = NULL; 1664b83cc969SLi Zefan bool readonly = false; 16656f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 166672fd032eSSage Weil 1667325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1668325c50e3SJeff Mahoney return -ENOTDIR; 1669325c50e3SJeff Mahoney 1670fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1671fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1672fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1673fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1674fdfb1e4fSLi Zefan 1675b83cc969SLi Zefan if (vol_args->flags & 16766f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 16776f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1678b83cc969SLi Zefan ret = -EOPNOTSUPP; 1679c47ca32dSDan Carpenter goto free_args; 1680fdfb1e4fSLi Zefan } 1681fdfb1e4fSLi Zefan 1682fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1683fdfb1e4fSLi Zefan ptr = &transid; 1684b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1685b83cc969SLi Zefan readonly = true; 16866f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 168709cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 16886f72c7e2SArne Jansen ret = -EINVAL; 1689c47ca32dSDan Carpenter goto free_args; 16906f72c7e2SArne Jansen } 16916f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 16926f72c7e2SArne Jansen if (IS_ERR(inherit)) { 16936f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1694c47ca32dSDan Carpenter goto free_args; 16956f72c7e2SArne Jansen } 16966f72c7e2SArne Jansen } 169775eaa0e2SSage Weil 1698fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 16996f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17008696c533SMiao Xie readonly, inherit); 1701c47ca32dSDan Carpenter if (ret) 1702c47ca32dSDan Carpenter goto free_inherit; 170375eaa0e2SSage Weil 1704c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1705fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1706c47ca32dSDan Carpenter transid), 1707c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 170875eaa0e2SSage Weil ret = -EFAULT; 1709c47ca32dSDan Carpenter 1710c47ca32dSDan Carpenter free_inherit: 17116f72c7e2SArne Jansen kfree(inherit); 1712c47ca32dSDan Carpenter free_args: 1713c47ca32dSDan Carpenter kfree(vol_args); 1714f46b5a66SChristoph Hellwig return ret; 1715f46b5a66SChristoph Hellwig } 1716f46b5a66SChristoph Hellwig 17170caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17180caa102dSLi Zefan void __user *arg) 17190caa102dSLi Zefan { 1720496ad9aaSAl Viro struct inode *inode = file_inode(file); 17210b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17220caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17230caa102dSLi Zefan int ret = 0; 17240caa102dSLi Zefan u64 flags = 0; 17250caa102dSLi Zefan 17264a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 17270caa102dSLi Zefan return -EINVAL; 17280caa102dSLi Zefan 17290b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 17300caa102dSLi Zefan if (btrfs_root_readonly(root)) 17310caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17320b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 17330caa102dSLi Zefan 17340caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 17350caa102dSLi Zefan ret = -EFAULT; 17360caa102dSLi Zefan 17370caa102dSLi Zefan return ret; 17380caa102dSLi Zefan } 17390caa102dSLi Zefan 17400caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 17410caa102dSLi Zefan void __user *arg) 17420caa102dSLi Zefan { 1743496ad9aaSAl Viro struct inode *inode = file_inode(file); 17440b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17450caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17460caa102dSLi Zefan struct btrfs_trans_handle *trans; 17470caa102dSLi Zefan u64 root_flags; 17480caa102dSLi Zefan u64 flags; 17490caa102dSLi Zefan int ret = 0; 17500caa102dSLi Zefan 1751bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1752bd60ea0fSDavid Sterba return -EPERM; 1753bd60ea0fSDavid Sterba 1754b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1755b9ca0664SLiu Bo if (ret) 1756b9ca0664SLiu Bo goto out; 17570caa102dSLi Zefan 17584a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1759b9ca0664SLiu Bo ret = -EINVAL; 1760b9ca0664SLiu Bo goto out_drop_write; 1761b9ca0664SLiu Bo } 17620caa102dSLi Zefan 1763b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1764b9ca0664SLiu Bo ret = -EFAULT; 1765b9ca0664SLiu Bo goto out_drop_write; 1766b9ca0664SLiu Bo } 17670caa102dSLi Zefan 1768b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1769b9ca0664SLiu Bo ret = -EINVAL; 1770b9ca0664SLiu Bo goto out_drop_write; 1771b9ca0664SLiu Bo } 17720caa102dSLi Zefan 1773b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1774b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1775b9ca0664SLiu Bo goto out_drop_write; 1776b9ca0664SLiu Bo } 17770caa102dSLi Zefan 17780b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 17790caa102dSLi Zefan 17800caa102dSLi Zefan /* nothing to do */ 17810caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1782b9ca0664SLiu Bo goto out_drop_sem; 17830caa102dSLi Zefan 17840caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 17852c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 17860caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 17870caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 17882c686537SDavid Sterba } else { 17892c686537SDavid Sterba /* 17902c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 17912c686537SDavid Sterba * send 17922c686537SDavid Sterba */ 17932c686537SDavid Sterba spin_lock(&root->root_item_lock); 17942c686537SDavid Sterba if (root->send_in_progress == 0) { 17950caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 17960caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 17972c686537SDavid Sterba spin_unlock(&root->root_item_lock); 17982c686537SDavid Sterba } else { 17992c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18000b246afaSJeff Mahoney btrfs_warn(fs_info, 18012c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18022c686537SDavid Sterba root->root_key.objectid); 18032c686537SDavid Sterba ret = -EPERM; 18042c686537SDavid Sterba goto out_drop_sem; 18052c686537SDavid Sterba } 18062c686537SDavid Sterba } 18070caa102dSLi Zefan 18080caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18090caa102dSLi Zefan if (IS_ERR(trans)) { 18100caa102dSLi Zefan ret = PTR_ERR(trans); 18110caa102dSLi Zefan goto out_reset; 18120caa102dSLi Zefan } 18130caa102dSLi Zefan 18140b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 18150caa102dSLi Zefan &root->root_key, &root->root_item); 18169417ebc8SNikolay Borisov if (ret < 0) { 18179417ebc8SNikolay Borisov btrfs_end_transaction(trans); 18189417ebc8SNikolay Borisov goto out_reset; 18199417ebc8SNikolay Borisov } 18200caa102dSLi Zefan 18219417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 18229417ebc8SNikolay Borisov 18230caa102dSLi Zefan out_reset: 18240caa102dSLi Zefan if (ret) 18250caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1826b9ca0664SLiu Bo out_drop_sem: 18270b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1828b9ca0664SLiu Bo out_drop_write: 1829b9ca0664SLiu Bo mnt_drop_write_file(file); 1830b9ca0664SLiu Bo out: 18310caa102dSLi Zefan return ret; 18320caa102dSLi Zefan } 18330caa102dSLi Zefan 1834ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1835ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1836ac8e9819SChris Mason { 1837abc6e134SChris Mason struct btrfs_key test; 1838abc6e134SChris Mason int ret; 1839abc6e134SChris Mason 1840abc6e134SChris Mason test.objectid = sk->min_objectid; 1841abc6e134SChris Mason test.type = sk->min_type; 1842abc6e134SChris Mason test.offset = sk->min_offset; 1843abc6e134SChris Mason 1844abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1845abc6e134SChris Mason if (ret < 0) 1846ac8e9819SChris Mason return 0; 1847abc6e134SChris Mason 1848abc6e134SChris Mason test.objectid = sk->max_objectid; 1849abc6e134SChris Mason test.type = sk->max_type; 1850abc6e134SChris Mason test.offset = sk->max_offset; 1851abc6e134SChris Mason 1852abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1853abc6e134SChris Mason if (ret > 0) 1854ac8e9819SChris Mason return 0; 1855ac8e9819SChris Mason return 1; 1856ac8e9819SChris Mason } 1857ac8e9819SChris Mason 1858df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 1859ac8e9819SChris Mason struct btrfs_key *key, 1860ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 18619b6e817dSGerhard Heift size_t *buf_size, 1862ba346b35SGerhard Heift char __user *ubuf, 1863ac8e9819SChris Mason unsigned long *sk_offset, 1864ac8e9819SChris Mason int *num_found) 1865ac8e9819SChris Mason { 1866ac8e9819SChris Mason u64 found_transid; 1867ac8e9819SChris Mason struct extent_buffer *leaf; 1868ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1869dd81d459SNaohiro Aota struct btrfs_key test; 1870ac8e9819SChris Mason unsigned long item_off; 1871ac8e9819SChris Mason unsigned long item_len; 1872ac8e9819SChris Mason int nritems; 1873ac8e9819SChris Mason int i; 1874ac8e9819SChris Mason int slot; 1875ac8e9819SChris Mason int ret = 0; 1876ac8e9819SChris Mason 1877ac8e9819SChris Mason leaf = path->nodes[0]; 1878ac8e9819SChris Mason slot = path->slots[0]; 1879ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1880ac8e9819SChris Mason 1881ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1882ac8e9819SChris Mason i = nritems; 1883ac8e9819SChris Mason goto advance_key; 1884ac8e9819SChris Mason } 1885ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1886ac8e9819SChris Mason 1887ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1888ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1889ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1890ac8e9819SChris Mason 189103b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 189203b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 189303b71c6cSGabriel de Perthuis continue; 189403b71c6cSGabriel de Perthuis 18959b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 18968f5f6178SGerhard Heift if (*num_found) { 1897ac8e9819SChris Mason ret = 1; 18988f5f6178SGerhard Heift goto out; 18998f5f6178SGerhard Heift } 19008f5f6178SGerhard Heift 19018f5f6178SGerhard Heift /* 19028f5f6178SGerhard Heift * return one empty item back for v1, which does not 19038f5f6178SGerhard Heift * handle -EOVERFLOW 19048f5f6178SGerhard Heift */ 19058f5f6178SGerhard Heift 19069b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 1907ac8e9819SChris Mason item_len = 0; 19088f5f6178SGerhard Heift ret = -EOVERFLOW; 19098f5f6178SGerhard Heift } 1910ac8e9819SChris Mason 19119b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 1912ac8e9819SChris Mason ret = 1; 191325c9bc2eSGerhard Heift goto out; 1914ac8e9819SChris Mason } 1915ac8e9819SChris Mason 1916ac8e9819SChris Mason sh.objectid = key->objectid; 1917ac8e9819SChris Mason sh.offset = key->offset; 1918ac8e9819SChris Mason sh.type = key->type; 1919ac8e9819SChris Mason sh.len = item_len; 1920ac8e9819SChris Mason sh.transid = found_transid; 1921ac8e9819SChris Mason 1922ac8e9819SChris Mason /* copy search result header */ 1923ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 1924ba346b35SGerhard Heift ret = -EFAULT; 1925ba346b35SGerhard Heift goto out; 1926ba346b35SGerhard Heift } 1927ba346b35SGerhard Heift 1928ac8e9819SChris Mason *sk_offset += sizeof(sh); 1929ac8e9819SChris Mason 1930ac8e9819SChris Mason if (item_len) { 1931ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 1932ac8e9819SChris Mason /* copy the item */ 1933ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 1934ba346b35SGerhard Heift item_off, item_len)) { 1935ba346b35SGerhard Heift ret = -EFAULT; 1936ba346b35SGerhard Heift goto out; 1937ba346b35SGerhard Heift } 1938ba346b35SGerhard Heift 1939ac8e9819SChris Mason *sk_offset += item_len; 1940ac8e9819SChris Mason } 1941e2156867SHugo Mills (*num_found)++; 1942ac8e9819SChris Mason 19438f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 19448f5f6178SGerhard Heift goto out; 19458f5f6178SGerhard Heift 194625c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 194725c9bc2eSGerhard Heift ret = 1; 194825c9bc2eSGerhard Heift goto out; 194925c9bc2eSGerhard Heift } 1950ac8e9819SChris Mason } 1951ac8e9819SChris Mason advance_key: 1952ac8e9819SChris Mason ret = 0; 1953dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 1954dd81d459SNaohiro Aota test.type = sk->max_type; 1955dd81d459SNaohiro Aota test.offset = sk->max_offset; 1956dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 1957dd81d459SNaohiro Aota ret = 1; 1958dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 1959abc6e134SChris Mason key->offset++; 1960dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 1961abc6e134SChris Mason key->offset = 0; 1962abc6e134SChris Mason key->type++; 1963dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 1964abc6e134SChris Mason key->offset = 0; 1965abc6e134SChris Mason key->type = 0; 1966abc6e134SChris Mason key->objectid++; 1967abc6e134SChris Mason } else 1968abc6e134SChris Mason ret = 1; 196925c9bc2eSGerhard Heift out: 1970ba346b35SGerhard Heift /* 1971ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 1972ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 1973ba346b35SGerhard Heift * * all items were found 1974ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 1975ba346b35SGerhard Heift * leaf 1976ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 1977ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 1978ba346b35SGerhard Heift */ 1979ac8e9819SChris Mason return ret; 1980ac8e9819SChris Mason } 1981ac8e9819SChris Mason 1982ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 198312544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 19849b6e817dSGerhard Heift size_t *buf_size, 1985ba346b35SGerhard Heift char __user *ubuf) 1986ac8e9819SChris Mason { 19870b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 1988ac8e9819SChris Mason struct btrfs_root *root; 1989ac8e9819SChris Mason struct btrfs_key key; 1990ac8e9819SChris Mason struct btrfs_path *path; 1991ac8e9819SChris Mason int ret; 1992ac8e9819SChris Mason int num_found = 0; 1993ac8e9819SChris Mason unsigned long sk_offset = 0; 1994ac8e9819SChris Mason 19959b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 19969b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 199712544442SGerhard Heift return -EOVERFLOW; 19989b6e817dSGerhard Heift } 199912544442SGerhard Heift 2000ac8e9819SChris Mason path = btrfs_alloc_path(); 2001ac8e9819SChris Mason if (!path) 2002ac8e9819SChris Mason return -ENOMEM; 2003ac8e9819SChris Mason 2004ac8e9819SChris Mason if (sk->tree_id == 0) { 2005ac8e9819SChris Mason /* search the root of the inode that was passed */ 2006ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2007ac8e9819SChris Mason } else { 2008ac8e9819SChris Mason key.objectid = sk->tree_id; 2009ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2010ac8e9819SChris Mason key.offset = (u64)-1; 2011ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2012ac8e9819SChris Mason if (IS_ERR(root)) { 2013ac8e9819SChris Mason btrfs_free_path(path); 2014ac8e9819SChris Mason return -ENOENT; 2015ac8e9819SChris Mason } 2016ac8e9819SChris Mason } 2017ac8e9819SChris Mason 2018ac8e9819SChris Mason key.objectid = sk->min_objectid; 2019ac8e9819SChris Mason key.type = sk->min_type; 2020ac8e9819SChris Mason key.offset = sk->min_offset; 2021ac8e9819SChris Mason 2022ac8e9819SChris Mason while (1) { 20236174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2024ac8e9819SChris Mason if (ret != 0) { 2025ac8e9819SChris Mason if (ret > 0) 2026ac8e9819SChris Mason ret = 0; 2027ac8e9819SChris Mason goto err; 2028ac8e9819SChris Mason } 2029df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2030ac8e9819SChris Mason &sk_offset, &num_found); 2031b3b4aa74SDavid Sterba btrfs_release_path(path); 203225c9bc2eSGerhard Heift if (ret) 2033ac8e9819SChris Mason break; 2034ac8e9819SChris Mason 2035ac8e9819SChris Mason } 20368f5f6178SGerhard Heift if (ret > 0) 2037ac8e9819SChris Mason ret = 0; 2038ac8e9819SChris Mason err: 2039ac8e9819SChris Mason sk->nr_items = num_found; 2040ac8e9819SChris Mason btrfs_free_path(path); 2041ac8e9819SChris Mason return ret; 2042ac8e9819SChris Mason } 2043ac8e9819SChris Mason 2044ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2045ac8e9819SChris Mason void __user *argp) 2046ac8e9819SChris Mason { 2047ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2048ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2049ac8e9819SChris Mason struct inode *inode; 2050ac8e9819SChris Mason int ret; 20519b6e817dSGerhard Heift size_t buf_size; 2052ac8e9819SChris Mason 2053ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2054ac8e9819SChris Mason return -EPERM; 2055ac8e9819SChris Mason 2056ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2057ac8e9819SChris Mason 2058ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2059ba346b35SGerhard Heift return -EFAULT; 2060ba346b35SGerhard Heift 2061ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2062ac8e9819SChris Mason 2063496ad9aaSAl Viro inode = file_inode(file); 2064ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 20658f5f6178SGerhard Heift 20668f5f6178SGerhard Heift /* 20678f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 20688f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 20698f5f6178SGerhard Heift */ 20708f5f6178SGerhard Heift if (ret == -EOVERFLOW) 20718f5f6178SGerhard Heift ret = 0; 20728f5f6178SGerhard Heift 2073ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2074ac8e9819SChris Mason ret = -EFAULT; 2075ac8e9819SChris Mason return ret; 2076ac8e9819SChris Mason } 2077ac8e9819SChris Mason 2078cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2079cc68a8a5SGerhard Heift void __user *argp) 2080cc68a8a5SGerhard Heift { 2081cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2082cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2083cc68a8a5SGerhard Heift struct inode *inode; 2084cc68a8a5SGerhard Heift int ret; 2085cc68a8a5SGerhard Heift size_t buf_size; 2086ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2087cc68a8a5SGerhard Heift 2088cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2089cc68a8a5SGerhard Heift return -EPERM; 2090cc68a8a5SGerhard Heift 2091cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2092cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2093cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2094cc68a8a5SGerhard Heift return -EFAULT; 2095cc68a8a5SGerhard Heift 2096cc68a8a5SGerhard Heift buf_size = args.buf_size; 2097cc68a8a5SGerhard Heift 2098cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2099cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2100cc68a8a5SGerhard Heift buf_size = buf_limit; 2101cc68a8a5SGerhard Heift 2102cc68a8a5SGerhard Heift inode = file_inode(file); 2103cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2104718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2105cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2106cc68a8a5SGerhard Heift ret = -EFAULT; 2107cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2108cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2109cc68a8a5SGerhard Heift ret = -EFAULT; 2110cc68a8a5SGerhard Heift 211176dda93cSYan, Zheng return ret; 211276dda93cSYan, Zheng } 211376dda93cSYan, Zheng 211498d377a0STARUISI Hiroaki /* 2115ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2116ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 211798d377a0STARUISI Hiroaki */ 211898d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 211998d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 212098d377a0STARUISI Hiroaki { 212198d377a0STARUISI Hiroaki struct btrfs_root *root; 212298d377a0STARUISI Hiroaki struct btrfs_key key; 2123ac8e9819SChris Mason char *ptr; 212498d377a0STARUISI Hiroaki int ret = -1; 212598d377a0STARUISI Hiroaki int slot; 212698d377a0STARUISI Hiroaki int len; 212798d377a0STARUISI Hiroaki int total_len = 0; 212898d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 212998d377a0STARUISI Hiroaki struct extent_buffer *l; 213098d377a0STARUISI Hiroaki struct btrfs_path *path; 213198d377a0STARUISI Hiroaki 213298d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 213398d377a0STARUISI Hiroaki name[0]='\0'; 213498d377a0STARUISI Hiroaki return 0; 213598d377a0STARUISI Hiroaki } 213698d377a0STARUISI Hiroaki 213798d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 213898d377a0STARUISI Hiroaki if (!path) 213998d377a0STARUISI Hiroaki return -ENOMEM; 214098d377a0STARUISI Hiroaki 2141c8bcbfbdSNikolay Borisov ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 214298d377a0STARUISI Hiroaki 214398d377a0STARUISI Hiroaki key.objectid = tree_id; 214498d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 214598d377a0STARUISI Hiroaki key.offset = (u64)-1; 214698d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 214798d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2148f14d104dSDavid Sterba btrfs_err(info, "could not find root %llu", tree_id); 21498ad6fcabSChris Mason ret = -ENOENT; 21508ad6fcabSChris Mason goto out; 215198d377a0STARUISI Hiroaki } 215298d377a0STARUISI Hiroaki 215398d377a0STARUISI Hiroaki key.objectid = dirid; 215498d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 21558ad6fcabSChris Mason key.offset = (u64)-1; 215698d377a0STARUISI Hiroaki 215798d377a0STARUISI Hiroaki while (1) { 215898d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 215998d377a0STARUISI Hiroaki if (ret < 0) 216098d377a0STARUISI Hiroaki goto out; 216118674c6cSFilipe David Borba Manana else if (ret > 0) { 216218674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 216318674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 216418674c6cSFilipe David Borba Manana if (ret < 0) 216518674c6cSFilipe David Borba Manana goto out; 216618674c6cSFilipe David Borba Manana else if (ret > 0) { 2167ac8e9819SChris Mason ret = -ENOENT; 216898d377a0STARUISI Hiroaki goto out; 2169ac8e9819SChris Mason } 217018674c6cSFilipe David Borba Manana } 217118674c6cSFilipe David Borba Manana 217218674c6cSFilipe David Borba Manana l = path->nodes[0]; 217318674c6cSFilipe David Borba Manana slot = path->slots[0]; 217418674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 217598d377a0STARUISI Hiroaki 217698d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 217798d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 217898d377a0STARUISI Hiroaki ptr -= len + 1; 217998d377a0STARUISI Hiroaki total_len += len + 1; 2180a696cf35SFilipe David Borba Manana if (ptr < name) { 2181a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 218298d377a0STARUISI Hiroaki goto out; 2183a696cf35SFilipe David Borba Manana } 218498d377a0STARUISI Hiroaki 218598d377a0STARUISI Hiroaki *(ptr + len) = '/'; 218698d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 218798d377a0STARUISI Hiroaki 218898d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 218998d377a0STARUISI Hiroaki break; 219098d377a0STARUISI Hiroaki 2191b3b4aa74SDavid Sterba btrfs_release_path(path); 219298d377a0STARUISI Hiroaki key.objectid = key.offset; 21938ad6fcabSChris Mason key.offset = (u64)-1; 219498d377a0STARUISI Hiroaki dirid = key.objectid; 219598d377a0STARUISI Hiroaki } 219677906a50SLi Zefan memmove(name, ptr, total_len); 219798d377a0STARUISI Hiroaki name[total_len] = '\0'; 219898d377a0STARUISI Hiroaki ret = 0; 219998d377a0STARUISI Hiroaki out: 220098d377a0STARUISI Hiroaki btrfs_free_path(path); 2201ac8e9819SChris Mason return ret; 2202ac8e9819SChris Mason } 2203ac8e9819SChris Mason 2204ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2205ac8e9819SChris Mason void __user *argp) 2206ac8e9819SChris Mason { 2207ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2208ac8e9819SChris Mason struct inode *inode; 220901b810b8SDavid Sterba int ret = 0; 2210ac8e9819SChris Mason 22112354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 22122354d08fSJulia Lawall if (IS_ERR(args)) 22132354d08fSJulia Lawall return PTR_ERR(args); 2214c2b96929SDan Carpenter 2215496ad9aaSAl Viro inode = file_inode(file); 2216ac8e9819SChris Mason 221701b810b8SDavid Sterba /* 221801b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 221901b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 222001b810b8SDavid Sterba */ 22211b53ac4dSChris Mason if (args->treeid == 0) 22221b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 22231b53ac4dSChris Mason 222401b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 222501b810b8SDavid Sterba args->name[0] = 0; 222601b810b8SDavid Sterba goto out; 222701b810b8SDavid Sterba } 222801b810b8SDavid Sterba 222901b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 223001b810b8SDavid Sterba ret = -EPERM; 223101b810b8SDavid Sterba goto out; 223201b810b8SDavid Sterba } 223301b810b8SDavid Sterba 2234ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2235ac8e9819SChris Mason args->treeid, args->objectid, 2236ac8e9819SChris Mason args->name); 2237ac8e9819SChris Mason 223801b810b8SDavid Sterba out: 2239ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2240ac8e9819SChris Mason ret = -EFAULT; 2241ac8e9819SChris Mason 2242ac8e9819SChris Mason kfree(args); 224398d377a0STARUISI Hiroaki return ret; 224498d377a0STARUISI Hiroaki } 224598d377a0STARUISI Hiroaki 224676dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 224776dda93cSYan, Zheng void __user *arg) 224876dda93cSYan, Zheng { 224954563d41SAl Viro struct dentry *parent = file->f_path.dentry; 22500b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 225176dda93cSYan, Zheng struct dentry *dentry; 22522b0143b5SDavid Howells struct inode *dir = d_inode(parent); 225376dda93cSYan, Zheng struct inode *inode; 225476dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 225576dda93cSYan, Zheng struct btrfs_root *dest = NULL; 225676dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 225776dda93cSYan, Zheng int namelen; 225876dda93cSYan, Zheng int err = 0; 225976dda93cSYan, Zheng 2260325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2261325c50e3SJeff Mahoney return -ENOTDIR; 2262325c50e3SJeff Mahoney 226376dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 226476dda93cSYan, Zheng if (IS_ERR(vol_args)) 226576dda93cSYan, Zheng return PTR_ERR(vol_args); 226676dda93cSYan, Zheng 226776dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 226876dda93cSYan, Zheng namelen = strlen(vol_args->name); 226976dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 227076dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 227176dda93cSYan, Zheng err = -EINVAL; 227276dda93cSYan, Zheng goto out; 227376dda93cSYan, Zheng } 227476dda93cSYan, Zheng 2275a561be71SAl Viro err = mnt_want_write_file(file); 227676dda93cSYan, Zheng if (err) 227776dda93cSYan, Zheng goto out; 227876dda93cSYan, Zheng 2279521e0546SDavid Sterba 228000235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 228100235411SAl Viro if (err == -EINTR) 228200235411SAl Viro goto out_drop_write; 228376dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 228476dda93cSYan, Zheng if (IS_ERR(dentry)) { 228576dda93cSYan, Zheng err = PTR_ERR(dentry); 228676dda93cSYan, Zheng goto out_unlock_dir; 228776dda93cSYan, Zheng } 228876dda93cSYan, Zheng 22892b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 229076dda93cSYan, Zheng err = -ENOENT; 229176dda93cSYan, Zheng goto out_dput; 229276dda93cSYan, Zheng } 229376dda93cSYan, Zheng 22942b0143b5SDavid Howells inode = d_inode(dentry); 22954260f7c7SSage Weil dest = BTRFS_I(inode)->root; 22964260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 22974260f7c7SSage Weil /* 22984260f7c7SSage Weil * Regular user. Only allow this with a special mount 22994260f7c7SSage Weil * option, when the user has write+exec access to the 23004260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 23014260f7c7SSage Weil * allowed. 23024260f7c7SSage Weil * 23034260f7c7SSage Weil * Note that this is _not_ check that the subvol is 23044260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 23054260f7c7SSage Weil * otherwise be able to delete. 23064260f7c7SSage Weil * 23074260f7c7SSage Weil * Users who want to delete empty subvols should try 23084260f7c7SSage Weil * rmdir(2). 23094260f7c7SSage Weil */ 23104260f7c7SSage Weil err = -EPERM; 23110b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 23124260f7c7SSage Weil goto out_dput; 23134260f7c7SSage Weil 23144260f7c7SSage Weil /* 23154260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 23164260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 23174260f7c7SSage Weil * must be called on the dentry referencing the root 23184260f7c7SSage Weil * of the subvol, not a random directory contained 23194260f7c7SSage Weil * within it. 23204260f7c7SSage Weil */ 23214260f7c7SSage Weil err = -EINVAL; 23224260f7c7SSage Weil if (root == dest) 23234260f7c7SSage Weil goto out_dput; 23244260f7c7SSage Weil 23254260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 23264260f7c7SSage Weil if (err) 23274260f7c7SSage Weil goto out_dput; 23285c39da5bSMiao Xie } 23294260f7c7SSage Weil 23305c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 23314260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 23324260f7c7SSage Weil if (err) 23334260f7c7SSage Weil goto out_dput; 23344260f7c7SSage Weil 23354a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 233676dda93cSYan, Zheng err = -EINVAL; 233776dda93cSYan, Zheng goto out_dput; 233876dda93cSYan, Zheng } 233976dda93cSYan, Zheng 23405955102cSAl Viro inode_lock(inode); 2341f60a2364SMisono Tomohiro err = btrfs_delete_subvolume(dir, dentry); 23425955102cSAl Viro inode_unlock(inode); 2343f60a2364SMisono Tomohiro if (!err) 234476dda93cSYan, Zheng d_delete(dentry); 2345fa6ac876SLiu Bo 234676dda93cSYan, Zheng out_dput: 234776dda93cSYan, Zheng dput(dentry); 234876dda93cSYan, Zheng out_unlock_dir: 23495955102cSAl Viro inode_unlock(dir); 235000235411SAl Viro out_drop_write: 23512a79f17eSAl Viro mnt_drop_write_file(file); 235276dda93cSYan, Zheng out: 235376dda93cSYan, Zheng kfree(vol_args); 235476dda93cSYan, Zheng return err; 235576dda93cSYan, Zheng } 235676dda93cSYan, Zheng 23571e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2358f46b5a66SChristoph Hellwig { 2359496ad9aaSAl Viro struct inode *inode = file_inode(file); 2360f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 23611e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2362c146afadSYan Zheng int ret; 2363c146afadSYan Zheng 236425122d15SIlya Dryomov ret = mnt_want_write_file(file); 236525122d15SIlya Dryomov if (ret) 236625122d15SIlya Dryomov return ret; 2367b83cc969SLi Zefan 236825122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 236925122d15SIlya Dryomov ret = -EROFS; 237025122d15SIlya Dryomov goto out; 23715ac00addSStefan Behrens } 2372f46b5a66SChristoph Hellwig 2373f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2374f46b5a66SChristoph Hellwig case S_IFDIR: 2375e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2376e441d54dSChris Mason ret = -EPERM; 2377e441d54dSChris Mason goto out; 2378e441d54dSChris Mason } 2379de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2380f46b5a66SChristoph Hellwig break; 2381f46b5a66SChristoph Hellwig case S_IFREG: 2382e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2383e441d54dSChris Mason ret = -EINVAL; 2384e441d54dSChris Mason goto out; 2385e441d54dSChris Mason } 23861e701a32SChris Mason 23871e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 23881e701a32SChris Mason if (!range) { 23891e701a32SChris Mason ret = -ENOMEM; 23901e701a32SChris Mason goto out; 23911e701a32SChris Mason } 23921e701a32SChris Mason 23931e701a32SChris Mason if (argp) { 23941e701a32SChris Mason if (copy_from_user(range, argp, 23951e701a32SChris Mason sizeof(*range))) { 23961e701a32SChris Mason ret = -EFAULT; 23971e701a32SChris Mason kfree(range); 2398683be16eSDan Carpenter goto out; 23991e701a32SChris Mason } 24001e701a32SChris Mason /* compression requires us to start the IO */ 24011e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 24021e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 24031e701a32SChris Mason range->extent_thresh = (u32)-1; 24041e701a32SChris Mason } 24051e701a32SChris Mason } else { 24061e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 24071e701a32SChris Mason range->len = (u64)-1; 24081e701a32SChris Mason } 2409496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 24107c829b72SAnand Jain range, BTRFS_OLDEST_GENERATION, 0); 24114cb5300bSChris Mason if (ret > 0) 24124cb5300bSChris Mason ret = 0; 24131e701a32SChris Mason kfree(range); 2414f46b5a66SChristoph Hellwig break; 24158929ecfaSYan, Zheng default: 24168929ecfaSYan, Zheng ret = -EINVAL; 2417f46b5a66SChristoph Hellwig } 2418e441d54dSChris Mason out: 241925122d15SIlya Dryomov mnt_drop_write_file(file); 2420e441d54dSChris Mason return ret; 2421f46b5a66SChristoph Hellwig } 2422f46b5a66SChristoph Hellwig 24232ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2424f46b5a66SChristoph Hellwig { 2425f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2426f46b5a66SChristoph Hellwig int ret; 2427f46b5a66SChristoph Hellwig 2428e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2429e441d54dSChris Mason return -EPERM; 2430e441d54dSChris Mason 2431171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 2432e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2433c9e9f97bSIlya Dryomov 2434dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2435c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2436c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2437c9e9f97bSIlya Dryomov goto out; 2438c9e9f97bSIlya Dryomov } 2439f46b5a66SChristoph Hellwig 24405516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 24412ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 2442f46b5a66SChristoph Hellwig 244343d20761SAnand Jain if (!ret) 24440b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 244543d20761SAnand Jain 2446f46b5a66SChristoph Hellwig kfree(vol_args); 2447c9e9f97bSIlya Dryomov out: 2448171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2449f46b5a66SChristoph Hellwig return ret; 2450f46b5a66SChristoph Hellwig } 2451f46b5a66SChristoph Hellwig 24526b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2453f46b5a66SChristoph Hellwig { 24540b246afaSJeff Mahoney struct inode *inode = file_inode(file); 24550b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 24566b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 2457f46b5a66SChristoph Hellwig int ret; 2458f46b5a66SChristoph Hellwig 2459e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2460e441d54dSChris Mason return -EPERM; 2461e441d54dSChris Mason 2462da24927bSMiao Xie ret = mnt_want_write_file(file); 2463da24927bSMiao Xie if (ret) 2464da24927bSMiao Xie return ret; 2465c146afadSYan Zheng 2466dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2467c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2468c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2469c47ca32dSDan Carpenter goto err_drop; 2470c9e9f97bSIlya Dryomov } 2471f46b5a66SChristoph Hellwig 24726b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 2473735654eaSDavid Sterba if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) 2474735654eaSDavid Sterba return -EOPNOTSUPP; 2475f46b5a66SChristoph Hellwig 2476171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 2477183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2478183860f6SAnand Jain goto out; 2479183860f6SAnand Jain } 2480183860f6SAnand Jain 2481735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 24822ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 24836b526ed7SAnand Jain } else { 24846b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 24852ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 24866b526ed7SAnand Jain } 2487171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2488183860f6SAnand Jain 24896b526ed7SAnand Jain if (!ret) { 2490735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 24910b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 24926b526ed7SAnand Jain vol_args->devid); 24936b526ed7SAnand Jain else 24940b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 24956b526ed7SAnand Jain vol_args->name); 24966b526ed7SAnand Jain } 2497183860f6SAnand Jain out: 2498183860f6SAnand Jain kfree(vol_args); 2499c47ca32dSDan Carpenter err_drop: 25004ac20c70SIlya Dryomov mnt_drop_write_file(file); 2501f46b5a66SChristoph Hellwig return ret; 2502f46b5a66SChristoph Hellwig } 2503f46b5a66SChristoph Hellwig 2504f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2505f46b5a66SChristoph Hellwig { 25060b246afaSJeff Mahoney struct inode *inode = file_inode(file); 25070b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2508f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2509f46b5a66SChristoph Hellwig int ret; 2510f46b5a66SChristoph Hellwig 2511f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2512f46b5a66SChristoph Hellwig return -EPERM; 2513f46b5a66SChristoph Hellwig 2514f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 2515f46b5a66SChristoph Hellwig if (ret) 2516f46b5a66SChristoph Hellwig return ret; 2517f46b5a66SChristoph Hellwig 2518171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 2519f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 252058d7bbf8SDavid Sterba goto out_drop_write; 252158d7bbf8SDavid Sterba } 252258d7bbf8SDavid Sterba 252358d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 252458d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 252558d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 2526f46b5a66SChristoph Hellwig goto out; 2527f46b5a66SChristoph Hellwig } 2528f46b5a66SChristoph Hellwig 252958d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 25302ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 2531f46b5a66SChristoph Hellwig 2532f46b5a66SChristoph Hellwig if (!ret) 25330b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 2534f46b5a66SChristoph Hellwig kfree(vol_args); 253558d7bbf8SDavid Sterba out: 2536171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 253758d7bbf8SDavid Sterba out_drop_write: 2538f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 253958d7bbf8SDavid Sterba 2540f46b5a66SChristoph Hellwig return ret; 2541f46b5a66SChristoph Hellwig } 2542f46b5a66SChristoph Hellwig 25432ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 25442ff7e61eSJeff Mahoney void __user *arg) 2545475f6387SJan Schmidt { 2546027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2547475f6387SJan Schmidt struct btrfs_device *device; 25480b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2549027ed2f0SLi Zefan int ret = 0; 2550475f6387SJan Schmidt 2551027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2552027ed2f0SLi Zefan if (!fi_args) 2553027ed2f0SLi Zefan return -ENOMEM; 2554027ed2f0SLi Zefan 2555d03262c7SDavid Sterba rcu_read_lock(); 2556027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2557475f6387SJan Schmidt 2558d03262c7SDavid Sterba list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 2559027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2560027ed2f0SLi Zefan fi_args->max_id = device->devid; 2561475f6387SJan Schmidt } 2562d03262c7SDavid Sterba rcu_read_unlock(); 2563475f6387SJan Schmidt 2564d03262c7SDavid Sterba memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 2565bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 2566bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 2567bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 256880a773fbSDavid Sterba 2569027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2570027ed2f0SLi Zefan ret = -EFAULT; 2571475f6387SJan Schmidt 2572027ed2f0SLi Zefan kfree(fi_args); 2573027ed2f0SLi Zefan return ret; 2574475f6387SJan Schmidt } 2575475f6387SJan Schmidt 25762ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 25772ff7e61eSJeff Mahoney void __user *arg) 2578475f6387SJan Schmidt { 2579475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2580475f6387SJan Schmidt struct btrfs_device *dev; 2581475f6387SJan Schmidt int ret = 0; 2582475f6387SJan Schmidt char *s_uuid = NULL; 2583475f6387SJan Schmidt 2584475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2585475f6387SJan Schmidt if (IS_ERR(di_args)) 2586475f6387SJan Schmidt return PTR_ERR(di_args); 2587475f6387SJan Schmidt 2588dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2589475f6387SJan Schmidt s_uuid = di_args->uuid; 2590475f6387SJan Schmidt 2591c5593ca3SDavid Sterba rcu_read_lock(); 25920b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 2593475f6387SJan Schmidt 2594475f6387SJan Schmidt if (!dev) { 2595475f6387SJan Schmidt ret = -ENODEV; 2596475f6387SJan Schmidt goto out; 2597475f6387SJan Schmidt } 2598475f6387SJan Schmidt 2599475f6387SJan Schmidt di_args->devid = dev->devid; 26007cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 26017cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2602475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2603a27202fbSJim Meyering if (dev->name) { 2604606686eeSJosef Bacik struct rcu_string *name; 2605606686eeSJosef Bacik 2606606686eeSJosef Bacik name = rcu_dereference(dev->name); 26076670d4c2SXiongfeng Wang strncpy(di_args->path, name->str, sizeof(di_args->path) - 1); 2608a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2609a27202fbSJim Meyering } else { 261099ba55adSStefan Behrens di_args->path[0] = '\0'; 2611a27202fbSJim Meyering } 2612475f6387SJan Schmidt 2613475f6387SJan Schmidt out: 2614c5593ca3SDavid Sterba rcu_read_unlock(); 2615475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2616475f6387SJan Schmidt ret = -EFAULT; 2617475f6387SJan Schmidt 2618475f6387SJan Schmidt kfree(di_args); 2619475f6387SJan Schmidt return ret; 2620475f6387SJan Schmidt } 2621475f6387SJan Schmidt 2622f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2623416161dbSMark Fasheh { 2624416161dbSMark Fasheh struct page *page; 2625416161dbSMark Fasheh 2626416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2627416161dbSMark Fasheh if (!page) 262831314002SFilipe Manana return ERR_PTR(-ENOMEM); 2629416161dbSMark Fasheh 2630416161dbSMark Fasheh if (!PageUptodate(page)) { 263131314002SFilipe Manana int ret; 263231314002SFilipe Manana 263331314002SFilipe Manana ret = btrfs_readpage(NULL, page); 263431314002SFilipe Manana if (ret) 263531314002SFilipe Manana return ERR_PTR(ret); 2636416161dbSMark Fasheh lock_page(page); 2637416161dbSMark Fasheh if (!PageUptodate(page)) { 2638416161dbSMark Fasheh unlock_page(page); 263909cbfeafSKirill A. Shutemov put_page(page); 264031314002SFilipe Manana return ERR_PTR(-EIO); 2641416161dbSMark Fasheh } 264231314002SFilipe Manana if (page->mapping != inode->i_mapping) { 2643416161dbSMark Fasheh unlock_page(page); 264409cbfeafSKirill A. Shutemov put_page(page); 264531314002SFilipe Manana return ERR_PTR(-EAGAIN); 2646416161dbSMark Fasheh } 2647416161dbSMark Fasheh } 2648416161dbSMark Fasheh 2649416161dbSMark Fasheh return page; 2650416161dbSMark Fasheh } 2651416161dbSMark Fasheh 2652f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2653f4414602SMark Fasheh int num_pages, u64 off) 2654f4414602SMark Fasheh { 2655f4414602SMark Fasheh int i; 265609cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 2657f4414602SMark Fasheh 2658f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 265931314002SFilipe Manana again: 2660f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 266131314002SFilipe Manana if (IS_ERR(pages[i])) { 266231314002SFilipe Manana int err = PTR_ERR(pages[i]); 266331314002SFilipe Manana 266431314002SFilipe Manana if (err == -EAGAIN) 266531314002SFilipe Manana goto again; 266631314002SFilipe Manana pages[i] = NULL; 266731314002SFilipe Manana return err; 266831314002SFilipe Manana } 2669f4414602SMark Fasheh } 2670f4414602SMark Fasheh return 0; 2671f4414602SMark Fasheh } 2672f4414602SMark Fasheh 2673e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 2674e0bd70c6SFilipe Manana bool retry_range_locking) 267577fe20dcSMark Fasheh { 2676e0bd70c6SFilipe Manana /* 2677e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 2678e0bd70c6SFilipe Manana * another, and lock file content. 2679e0bd70c6SFilipe Manana * The locking order is: 2680e0bd70c6SFilipe Manana * 2681e0bd70c6SFilipe Manana * 1) pages 2682e0bd70c6SFilipe Manana * 2) range in the inode's io tree 2683e0bd70c6SFilipe Manana */ 268477fe20dcSMark Fasheh while (1) { 268577fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 268677fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 268777fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 268877fe20dcSMark Fasheh off + len - 1); 2689ff5df9b8SFilipe Manana if ((!ordered || 2690ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2691ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 269277fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2693ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2694ff5df9b8SFilipe Manana if (ordered) 2695ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 269677fe20dcSMark Fasheh break; 2697ff5df9b8SFilipe Manana } 269877fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 269977fe20dcSMark Fasheh if (ordered) 270077fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 2701e0bd70c6SFilipe Manana if (!retry_range_locking) 2702e0bd70c6SFilipe Manana return -EAGAIN; 270377fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 270477fe20dcSMark Fasheh } 2705e0bd70c6SFilipe Manana return 0; 270677fe20dcSMark Fasheh } 270777fe20dcSMark Fasheh 2708f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2709416161dbSMark Fasheh { 27105955102cSAl Viro inode_unlock(inode1); 27115955102cSAl Viro inode_unlock(inode2); 2712416161dbSMark Fasheh } 2713416161dbSMark Fasheh 2714f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2715f4414602SMark Fasheh { 2716f4414602SMark Fasheh if (inode1 < inode2) 2717f4414602SMark Fasheh swap(inode1, inode2); 2718f4414602SMark Fasheh 27195955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 27205955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 2721f4414602SMark Fasheh } 2722f4414602SMark Fasheh 2723f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2724f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2725f4414602SMark Fasheh { 2726f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2727f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2728f4414602SMark Fasheh } 2729f4414602SMark Fasheh 2730e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2731e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 2732e0bd70c6SFilipe Manana bool retry_range_locking) 2733416161dbSMark Fasheh { 2734e0bd70c6SFilipe Manana int ret; 2735e0bd70c6SFilipe Manana 2736416161dbSMark Fasheh if (inode1 < inode2) { 2737416161dbSMark Fasheh swap(inode1, inode2); 2738416161dbSMark Fasheh swap(loff1, loff2); 2739416161dbSMark Fasheh } 2740e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 2741e0bd70c6SFilipe Manana if (ret) 2742e0bd70c6SFilipe Manana return ret; 2743e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 2744e0bd70c6SFilipe Manana if (ret) 2745e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 2746e0bd70c6SFilipe Manana loff1 + len - 1); 2747e0bd70c6SFilipe Manana return ret; 2748416161dbSMark Fasheh } 2749f4414602SMark Fasheh 2750f4414602SMark Fasheh struct cmp_pages { 2751f4414602SMark Fasheh int num_pages; 2752f4414602SMark Fasheh struct page **src_pages; 2753f4414602SMark Fasheh struct page **dst_pages; 2754f4414602SMark Fasheh }; 2755f4414602SMark Fasheh 2756f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 2757f4414602SMark Fasheh { 2758f4414602SMark Fasheh int i; 2759f4414602SMark Fasheh struct page *pg; 2760f4414602SMark Fasheh 2761f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 2762f4414602SMark Fasheh pg = cmp->src_pages[i]; 2763e0bd70c6SFilipe Manana if (pg) { 2764e0bd70c6SFilipe Manana unlock_page(pg); 276509cbfeafSKirill A. Shutemov put_page(pg); 2766e0bd70c6SFilipe Manana } 2767f4414602SMark Fasheh pg = cmp->dst_pages[i]; 2768e0bd70c6SFilipe Manana if (pg) { 2769e0bd70c6SFilipe Manana unlock_page(pg); 277009cbfeafSKirill A. Shutemov put_page(pg); 2771f4414602SMark Fasheh } 2772e0bd70c6SFilipe Manana } 2773f4414602SMark Fasheh kfree(cmp->src_pages); 2774f4414602SMark Fasheh kfree(cmp->dst_pages); 2775f4414602SMark Fasheh } 2776f4414602SMark Fasheh 2777f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 2778f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 2779f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 2780f4414602SMark Fasheh { 2781f4414602SMark Fasheh int ret; 278209cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 2783f4414602SMark Fasheh struct page **src_pgarr, **dst_pgarr; 2784f4414602SMark Fasheh 2785f4414602SMark Fasheh /* 2786f4414602SMark Fasheh * We must gather up all the pages before we initiate our 2787f4414602SMark Fasheh * extent locking. We use an array for the page pointers. Size 2788f4414602SMark Fasheh * of the array is bounded by len, which is in turn bounded by 2789f4414602SMark Fasheh * BTRFS_MAX_DEDUPE_LEN. 2790f4414602SMark Fasheh */ 279166722f7cSDavid Sterba src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 279266722f7cSDavid Sterba dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 2793f4414602SMark Fasheh if (!src_pgarr || !dst_pgarr) { 2794f4414602SMark Fasheh kfree(src_pgarr); 2795f4414602SMark Fasheh kfree(dst_pgarr); 2796f4414602SMark Fasheh return -ENOMEM; 2797f4414602SMark Fasheh } 2798f4414602SMark Fasheh cmp->num_pages = num_pages; 2799f4414602SMark Fasheh cmp->src_pages = src_pgarr; 2800f4414602SMark Fasheh cmp->dst_pages = dst_pgarr; 2801f4414602SMark Fasheh 2802b1517622SFilipe Manana /* 2803b1517622SFilipe Manana * If deduping ranges in the same inode, locking rules make it mandatory 2804b1517622SFilipe Manana * to always lock pages in ascending order to avoid deadlocks with 2805b1517622SFilipe Manana * concurrent tasks (such as starting writeback/delalloc). 2806b1517622SFilipe Manana */ 2807b1517622SFilipe Manana if (src == dst && dst_loff < loff) { 2808b1517622SFilipe Manana swap(src_pgarr, dst_pgarr); 2809b1517622SFilipe Manana swap(loff, dst_loff); 2810b1517622SFilipe Manana } 2811b1517622SFilipe Manana 2812b1517622SFilipe Manana ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff); 2813f4414602SMark Fasheh if (ret) 2814f4414602SMark Fasheh goto out; 2815f4414602SMark Fasheh 2816b1517622SFilipe Manana ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff); 2817f4414602SMark Fasheh 2818f4414602SMark Fasheh out: 2819f4414602SMark Fasheh if (ret) 2820f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 282178ad4ce0SNaohiro Aota return ret; 2822416161dbSMark Fasheh } 2823416161dbSMark Fasheh 28241a287cfeSDavid Sterba static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp) 2825416161dbSMark Fasheh { 2826416161dbSMark Fasheh int ret = 0; 2827f4414602SMark Fasheh int i; 2828416161dbSMark Fasheh struct page *src_page, *dst_page; 282909cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 2830416161dbSMark Fasheh void *addr, *dst_addr; 2831416161dbSMark Fasheh 2832f4414602SMark Fasheh i = 0; 2833416161dbSMark Fasheh while (len) { 283409cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 2835416161dbSMark Fasheh cmp_len = len; 2836416161dbSMark Fasheh 2837f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 2838f4414602SMark Fasheh 2839f4414602SMark Fasheh src_page = cmp->src_pages[i]; 2840f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 2841e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 2842e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 2843f4414602SMark Fasheh 2844416161dbSMark Fasheh addr = kmap_atomic(src_page); 2845416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 2846416161dbSMark Fasheh 2847416161dbSMark Fasheh flush_dcache_page(src_page); 2848416161dbSMark Fasheh flush_dcache_page(dst_page); 2849416161dbSMark Fasheh 2850416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 28512b3909f8SDarrick J. Wong ret = -EBADE; 2852416161dbSMark Fasheh 2853416161dbSMark Fasheh kunmap_atomic(addr); 2854416161dbSMark Fasheh kunmap_atomic(dst_addr); 2855416161dbSMark Fasheh 2856416161dbSMark Fasheh if (ret) 2857416161dbSMark Fasheh break; 2858416161dbSMark Fasheh 2859416161dbSMark Fasheh len -= cmp_len; 2860f4414602SMark Fasheh i++; 2861416161dbSMark Fasheh } 2862416161dbSMark Fasheh 2863416161dbSMark Fasheh return ret; 2864416161dbSMark Fasheh } 2865416161dbSMark Fasheh 2866e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 2867e1d227a4SMark Fasheh u64 olen) 2868416161dbSMark Fasheh { 2869e1d227a4SMark Fasheh u64 len = *plen; 2870416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 2871416161dbSMark Fasheh 2872e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 2873416161dbSMark Fasheh return -EINVAL; 2874e1d227a4SMark Fasheh 2875e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 2876e1d227a4SMark Fasheh if (off + len == inode->i_size) 2877e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 2878e1d227a4SMark Fasheh 2879416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 2880416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 2881416161dbSMark Fasheh return -EINVAL; 2882416161dbSMark Fasheh 2883416161dbSMark Fasheh return 0; 2884416161dbSMark Fasheh } 2885416161dbSMark Fasheh 2886e1d227a4SMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 2887416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 2888416161dbSMark Fasheh { 2889416161dbSMark Fasheh int ret; 2890e1d227a4SMark Fasheh u64 len = olen; 2891f4414602SMark Fasheh struct cmp_pages cmp; 2892fc4badd9SOmar Sandoval bool same_inode = (src == dst); 28930efa9f48SMark Fasheh u64 same_lock_start = 0; 28940efa9f48SMark Fasheh u64 same_lock_len = 0; 2895416161dbSMark Fasheh 2896113e8283SFilipe Manana if (len == 0) 2897113e8283SFilipe Manana return 0; 2898113e8283SFilipe Manana 2899fc4badd9SOmar Sandoval if (same_inode) 29005955102cSAl Viro inode_lock(src); 2901fc4badd9SOmar Sandoval else 2902fc4badd9SOmar Sandoval btrfs_double_inode_lock(src, dst); 29030efa9f48SMark Fasheh 29040efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 29050efa9f48SMark Fasheh if (ret) 29060efa9f48SMark Fasheh goto out_unlock; 2907fc4badd9SOmar Sandoval 2908fc4badd9SOmar Sandoval ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 2909f4dfe687SFilipe Manana if (ret) 2910f4dfe687SFilipe Manana goto out_unlock; 29110efa9f48SMark Fasheh 2912fc4badd9SOmar Sandoval if (same_inode) { 29130efa9f48SMark Fasheh /* 29140efa9f48SMark Fasheh * Single inode case wants the same checks, except we 29150efa9f48SMark Fasheh * don't want our length pushed out past i_size as 29160efa9f48SMark Fasheh * comparing that data range makes no sense. 29170efa9f48SMark Fasheh * 29180efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 29190efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 29200efa9f48SMark Fasheh * reject the request. 29210efa9f48SMark Fasheh * 29220efa9f48SMark Fasheh * This effectively means we require aligned extents 29230efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 29240efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 29250efa9f48SMark Fasheh * i_size. 29260efa9f48SMark Fasheh */ 29270efa9f48SMark Fasheh if (len != olen) { 29280efa9f48SMark Fasheh ret = -EINVAL; 29290efa9f48SMark Fasheh goto out_unlock; 29300efa9f48SMark Fasheh } 29310efa9f48SMark Fasheh 29320efa9f48SMark Fasheh /* Check for overlapping ranges */ 29330efa9f48SMark Fasheh if (dst_loff + len > loff && dst_loff < loff + len) { 29340efa9f48SMark Fasheh ret = -EINVAL; 29350efa9f48SMark Fasheh goto out_unlock; 29360efa9f48SMark Fasheh } 29370efa9f48SMark Fasheh 29380efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 29390efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 29400efa9f48SMark Fasheh } 2941416161dbSMark Fasheh 2942416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 2943416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 2944416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 2945416161dbSMark Fasheh ret = -EINVAL; 2946416161dbSMark Fasheh goto out_unlock; 2947416161dbSMark Fasheh } 2948416161dbSMark Fasheh 2949e0bd70c6SFilipe Manana again: 2950f4414602SMark Fasheh ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 2951f4414602SMark Fasheh if (ret) 2952f4414602SMark Fasheh goto out_unlock; 2953f4414602SMark Fasheh 29540efa9f48SMark Fasheh if (same_inode) 2955e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 2956e0bd70c6SFilipe Manana false); 29570efa9f48SMark Fasheh else 2958e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 2959e0bd70c6SFilipe Manana false); 2960e0bd70c6SFilipe Manana /* 2961e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 2962e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 2963e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 2964e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 2965e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 2966e0bd70c6SFilipe Manana * range locking). 2967e0bd70c6SFilipe Manana */ 2968e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 2969e0bd70c6SFilipe Manana /* 2970e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 2971e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 2972e0bd70c6SFilipe Manana */ 2973e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 2974e0bd70c6SFilipe Manana if (same_inode) { 2975e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 2976e0bd70c6SFilipe Manana same_lock_len); 2977e0bd70c6SFilipe Manana } else { 2978e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 2979e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 2980e0bd70c6SFilipe Manana } 2981e0bd70c6SFilipe Manana goto again; 2982e0bd70c6SFilipe Manana } 2983e0bd70c6SFilipe Manana ASSERT(ret == 0); 2984e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 2985e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 2986e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 2987e0bd70c6SFilipe Manana return ret; 2988e0bd70c6SFilipe Manana } 2989f4414602SMark Fasheh 2990207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 29911a287cfeSDavid Sterba ret = btrfs_cmp_data(olen, &cmp); 2992416161dbSMark Fasheh if (ret == 0) 29931c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 2994416161dbSMark Fasheh 29950efa9f48SMark Fasheh if (same_inode) 29960efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 29970efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 29980efa9f48SMark Fasheh else 2999f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3000f4414602SMark Fasheh 3001f4414602SMark Fasheh btrfs_cmp_data_free(&cmp); 3002416161dbSMark Fasheh out_unlock: 30030efa9f48SMark Fasheh if (same_inode) 30045955102cSAl Viro inode_unlock(src); 30050efa9f48SMark Fasheh else 3006f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3007416161dbSMark Fasheh 3008416161dbSMark Fasheh return ret; 3009416161dbSMark Fasheh } 3010416161dbSMark Fasheh 3011ee22184bSByongho Lee #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3012416161dbSMark Fasheh 30132b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 30142b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3015416161dbSMark Fasheh { 30162b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 30172b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3018416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 30192b3909f8SDarrick J. Wong ssize_t res; 3020416161dbSMark Fasheh 30212b3909f8SDarrick J. Wong if (olen > BTRFS_MAX_DEDUPE_LEN) 30222b3909f8SDarrick J. Wong olen = BTRFS_MAX_DEDUPE_LEN; 3023416161dbSMark Fasheh 302409cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3025416161dbSMark Fasheh /* 3026416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3027416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3028416161dbSMark Fasheh * this situation without an update. 3029416161dbSMark Fasheh */ 30302b3909f8SDarrick J. Wong return -EINVAL; 3031416161dbSMark Fasheh } 3032416161dbSMark Fasheh 30332b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 30342b3909f8SDarrick J. Wong if (res) 30352b3909f8SDarrick J. Wong return res; 30362b3909f8SDarrick J. Wong return olen; 3037416161dbSMark Fasheh } 3038416161dbSMark Fasheh 3039f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3040f82a9901SFilipe Manana struct inode *inode, 3041f82a9901SFilipe Manana u64 endoff, 3042f82a9901SFilipe Manana const u64 destoff, 30431c919a5eSMark Fasheh const u64 olen, 30441c919a5eSMark Fasheh int no_time_update) 3045f82a9901SFilipe Manana { 3046f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3047f82a9901SFilipe Manana int ret; 3048f82a9901SFilipe Manana 3049f82a9901SFilipe Manana inode_inc_iversion(inode); 30501c919a5eSMark Fasheh if (!no_time_update) 3051c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3052f82a9901SFilipe Manana /* 3053f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3054f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3055f82a9901SFilipe Manana */ 3056f82a9901SFilipe Manana if (endoff > destoff + olen) 3057f82a9901SFilipe Manana endoff = destoff + olen; 3058f82a9901SFilipe Manana if (endoff > inode->i_size) 30596ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3060f82a9901SFilipe Manana 3061f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3062f82a9901SFilipe Manana if (ret) { 306366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 30643a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3065f82a9901SFilipe Manana goto out; 3066f82a9901SFilipe Manana } 30673a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3068f82a9901SFilipe Manana out: 3069f82a9901SFilipe Manana return ret; 3070f82a9901SFilipe Manana } 3071f82a9901SFilipe Manana 3072a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 30737ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 30747ffbb598SFilipe Manana const struct btrfs_path *path, 30757ffbb598SFilipe Manana const u64 hole_offset, 30767ffbb598SFilipe Manana const u64 hole_len) 30777ffbb598SFilipe Manana { 3078a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 30797ffbb598SFilipe Manana struct extent_map *em; 30807ffbb598SFilipe Manana int ret; 30817ffbb598SFilipe Manana 30827ffbb598SFilipe Manana em = alloc_extent_map(); 30837ffbb598SFilipe Manana if (!em) { 3084a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 30857ffbb598SFilipe Manana return; 30867ffbb598SFilipe Manana } 30877ffbb598SFilipe Manana 308814f59796SFilipe Manana if (path) { 308914f59796SFilipe Manana struct btrfs_file_extent_item *fi; 309014f59796SFilipe Manana 309114f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 309214f59796SFilipe Manana struct btrfs_file_extent_item); 3093a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 30947ffbb598SFilipe Manana em->generation = -1; 30957ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 30967ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 30977ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3098a2f392e4SNikolay Borisov &inode->runtime_flags); 30997ffbb598SFilipe Manana } else { 31007ffbb598SFilipe Manana em->start = hole_offset; 31017ffbb598SFilipe Manana em->len = hole_len; 31027ffbb598SFilipe Manana em->ram_bytes = em->len; 31037ffbb598SFilipe Manana em->orig_start = hole_offset; 31047ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 31057ffbb598SFilipe Manana em->block_len = 0; 31067ffbb598SFilipe Manana em->orig_block_len = 0; 31077ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 31087ffbb598SFilipe Manana em->generation = trans->transid; 31097ffbb598SFilipe Manana } 31107ffbb598SFilipe Manana 31117ffbb598SFilipe Manana while (1) { 31127ffbb598SFilipe Manana write_lock(&em_tree->lock); 31137ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 31147ffbb598SFilipe Manana write_unlock(&em_tree->lock); 31157ffbb598SFilipe Manana if (ret != -EEXIST) { 31167ffbb598SFilipe Manana free_extent_map(em); 31177ffbb598SFilipe Manana break; 31187ffbb598SFilipe Manana } 3119a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 31207ffbb598SFilipe Manana em->start + em->len - 1, 0); 31217ffbb598SFilipe Manana } 31227ffbb598SFilipe Manana 3123ee39b432SDavid Sterba if (ret) 3124a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 31257ffbb598SFilipe Manana } 31267ffbb598SFilipe Manana 31278039d87dSFilipe Manana /* 31288039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 31298039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 31308039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 31318039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 31328039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 31338039d87dSFilipe Manana * 31348039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 31358039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 31368039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 31378039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 31388039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 31398039d87dSFilipe Manana * i_mutex, which we are holding here). 31408039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 31418039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 31428039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 31438039d87dSFilipe Manana * 31448039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 31458039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 31468039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 31478039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 31488039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 31498039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 31508039d87dSFilipe Manana * the former. 31518039d87dSFilipe Manana */ 31524a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 31538039d87dSFilipe Manana struct btrfs_trans_handle *trans, 31548039d87dSFilipe Manana struct btrfs_path *path, 31558039d87dSFilipe Manana struct btrfs_key *new_key, 31568039d87dSFilipe Manana const u64 drop_start, 31578039d87dSFilipe Manana const u64 datal, 31588039d87dSFilipe Manana const u64 skip, 31598039d87dSFilipe Manana const u64 size, 31608039d87dSFilipe Manana char *inline_data) 31618039d87dSFilipe Manana { 31620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 31638039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 31648039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 31650b246afaSJeff Mahoney fs_info->sectorsize); 31668039d87dSFilipe Manana int ret; 31678039d87dSFilipe Manana struct btrfs_key key; 31688039d87dSFilipe Manana 31698039d87dSFilipe Manana if (new_key->offset > 0) 31708039d87dSFilipe Manana return -EOPNOTSUPP; 31718039d87dSFilipe Manana 31724a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 31738039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 31748039d87dSFilipe Manana key.offset = 0; 31758039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 31768039d87dSFilipe Manana if (ret < 0) { 31778039d87dSFilipe Manana return ret; 31788039d87dSFilipe Manana } else if (ret > 0) { 31798039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 31808039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 31818039d87dSFilipe Manana if (ret < 0) 31828039d87dSFilipe Manana return ret; 31838039d87dSFilipe Manana else if (ret > 0) 31848039d87dSFilipe Manana goto copy_inline_extent; 31858039d87dSFilipe Manana } 31868039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 31874a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 31888039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 31898039d87dSFilipe Manana ASSERT(key.offset > 0); 31908039d87dSFilipe Manana return -EOPNOTSUPP; 31918039d87dSFilipe Manana } 31928039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 31938039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 31948039d87dSFilipe Manana u64 ext_len; 31958039d87dSFilipe Manana 31968039d87dSFilipe Manana /* 31978039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 31988039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 31998039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 32008039d87dSFilipe Manana */ 32018039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 32028039d87dSFilipe Manana struct btrfs_file_extent_item); 32038039d87dSFilipe Manana /* 32048039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 32058039d87dSFilipe Manana * following it. 32068039d87dSFilipe Manana */ 32078039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 32088039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 32098039d87dSFilipe Manana goto copy_inline_extent; 32108039d87dSFilipe Manana 32118039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 32128039d87dSFilipe Manana if (ext_len > aligned_end) 32138039d87dSFilipe Manana return -EOPNOTSUPP; 32148039d87dSFilipe Manana 32158039d87dSFilipe Manana ret = btrfs_next_item(root, path); 32168039d87dSFilipe Manana if (ret < 0) { 32178039d87dSFilipe Manana return ret; 32188039d87dSFilipe Manana } else if (ret == 0) { 32198039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 32208039d87dSFilipe Manana path->slots[0]); 32214a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 32228039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 32238039d87dSFilipe Manana return -EOPNOTSUPP; 32248039d87dSFilipe Manana } 32258039d87dSFilipe Manana } 32268039d87dSFilipe Manana 32278039d87dSFilipe Manana copy_inline_extent: 32288039d87dSFilipe Manana /* 32298039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 32308039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 32318039d87dSFilipe Manana */ 32328039d87dSFilipe Manana if (i_size_read(dst) > datal) { 32338039d87dSFilipe Manana /* 32348039d87dSFilipe Manana * If the destination inode has an inline extent... 32358039d87dSFilipe Manana * This would require copying the data from the source inline 32368039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 32378039d87dSFilipe Manana * But this is really complex, both extents can be compressed 32388039d87dSFilipe Manana * or just one of them, which would require decompressing and 32398039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 32408039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 32418039d87dSFilipe Manana * inline extent). 32428039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 32438039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 32448039d87dSFilipe Manana */ 32458039d87dSFilipe Manana return -EOPNOTSUPP; 32468039d87dSFilipe Manana } 32478039d87dSFilipe Manana 32488039d87dSFilipe Manana btrfs_release_path(path); 32498039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 32508039d87dSFilipe Manana if (ret) 32518039d87dSFilipe Manana return ret; 32528039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 32538039d87dSFilipe Manana if (ret) 32548039d87dSFilipe Manana return ret; 32558039d87dSFilipe Manana 32568039d87dSFilipe Manana if (skip) { 32578039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 32588039d87dSFilipe Manana 32598039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 32608039d87dSFilipe Manana } 32618039d87dSFilipe Manana 32628039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 32638039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 32648039d87dSFilipe Manana path->slots[0]), 32658039d87dSFilipe Manana size); 32668039d87dSFilipe Manana inode_add_bytes(dst, datal); 32678039d87dSFilipe Manana 32688039d87dSFilipe Manana return 0; 32698039d87dSFilipe Manana } 32708039d87dSFilipe Manana 327132b7c687SMark Fasheh /** 327232b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 327332b7c687SMark Fasheh * 327432b7c687SMark Fasheh * @src: Inode to clone from 327532b7c687SMark Fasheh * @inode: Inode to clone to 327632b7c687SMark Fasheh * @off: Offset within source to start clone from 327732b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 32781c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 327932b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 32801c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 328132b7c687SMark Fasheh */ 328232b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3283f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 32841c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3285f46b5a66SChristoph Hellwig { 32860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3287f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 328832b7c687SMark Fasheh struct btrfs_path *path = NULL; 3289f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 329032b7c687SMark Fasheh struct btrfs_trans_handle *trans; 329132b7c687SMark Fasheh char *buf = NULL; 3292ae01a0abSYan Zheng struct btrfs_key key; 3293f46b5a66SChristoph Hellwig u32 nritems; 3294f46b5a66SChristoph Hellwig int slot; 3295ae01a0abSYan Zheng int ret; 3296f82a9901SFilipe Manana const u64 len = olen_aligned; 3297f82a9901SFilipe Manana u64 last_dest_end = destoff; 3298ae01a0abSYan Zheng 3299ae01a0abSYan Zheng ret = -ENOMEM; 3300752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3301ae01a0abSYan Zheng if (!buf) 330232b7c687SMark Fasheh return ret; 3303ae01a0abSYan Zheng 3304ae01a0abSYan Zheng path = btrfs_alloc_path(); 3305ae01a0abSYan Zheng if (!path) { 330615351955SDavid Sterba kvfree(buf); 330732b7c687SMark Fasheh return ret; 3308ae01a0abSYan Zheng } 330932b7c687SMark Fasheh 3310e4058b54SDavid Sterba path->reada = READA_FORWARD; 3311c5c9cd4dSSage Weil /* clone data */ 33124a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3313ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 33142c463823SFilipe Manana key.offset = off; 3315f46b5a66SChristoph Hellwig 3316f46b5a66SChristoph Hellwig while (1) { 3317de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3318df858e76SFilipe Manana 3319f46b5a66SChristoph Hellwig /* 3320f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3321f46b5a66SChristoph Hellwig * tree. 3322f46b5a66SChristoph Hellwig */ 3323e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3324362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3325362a20c5SDavid Sterba 0, 0); 3326f46b5a66SChristoph Hellwig if (ret < 0) 3327f46b5a66SChristoph Hellwig goto out; 33282c463823SFilipe Manana /* 33292c463823SFilipe Manana * First search, if no extent item that starts at offset off was 33302c463823SFilipe Manana * found but the previous item is an extent item, it's possible 33312c463823SFilipe Manana * it might overlap our target range, therefore process it. 33322c463823SFilipe Manana */ 33332c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 33342c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 33352c463823SFilipe Manana path->slots[0] - 1); 33362c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 33372c463823SFilipe Manana path->slots[0]--; 33382c463823SFilipe Manana } 3339f46b5a66SChristoph Hellwig 3340ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3341e4355f34SFilipe David Borba Manana process_slot: 3342ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3343362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3344f46b5a66SChristoph Hellwig if (ret < 0) 3345f46b5a66SChristoph Hellwig goto out; 3346f46b5a66SChristoph Hellwig if (ret > 0) 3347f46b5a66SChristoph Hellwig break; 3348ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3349f46b5a66SChristoph Hellwig } 3350f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3351f46b5a66SChristoph Hellwig slot = path->slots[0]; 3352f46b5a66SChristoph Hellwig 3353ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3354962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 33554a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3356f46b5a66SChristoph Hellwig break; 3357f46b5a66SChristoph Hellwig 3358962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3359c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3360c5c9cd4dSSage Weil int type; 336131840ae1SZheng Yan u32 size; 336231840ae1SZheng Yan struct btrfs_key new_key; 3363c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3364c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3365c5c9cd4dSSage Weil u8 comp; 3366f82a9901SFilipe Manana u64 drop_start; 336731840ae1SZheng Yan 3368c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3369c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3370c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3371c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3372c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3373c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3374d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3375d397712bSChris Mason extent); 3376d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3377d397712bSChris Mason extent); 3378c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3379d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3380d397712bSChris Mason extent); 3381c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3382c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3383c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3384c5c9cd4dSSage Weil extent); 3385c5c9cd4dSSage Weil } 338631840ae1SZheng Yan 33872c463823SFilipe Manana /* 33882c463823SFilipe Manana * The first search might have left us at an extent 33892c463823SFilipe Manana * item that ends before our target range's start, can 33902c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 33912c463823SFilipe Manana */ 33922c463823SFilipe Manana if (key.offset + datal <= off) { 3393e4355f34SFilipe David Borba Manana path->slots[0]++; 3394e4355f34SFilipe David Borba Manana goto process_slot; 33952c463823SFilipe Manana } else if (key.offset >= off + len) { 33962c463823SFilipe Manana break; 3397e4355f34SFilipe David Borba Manana } 3398df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3399e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3400e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3401e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3402e4355f34SFilipe David Borba Manana size); 3403e4355f34SFilipe David Borba Manana 3404e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3405e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3406c5c9cd4dSSage Weil 340731840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 34084a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 34094d728ec7SLi Zefan if (off <= key.offset) 3410c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 34114d728ec7SLi Zefan else 34124d728ec7SLi Zefan new_key.offset = destoff; 3413c5c9cd4dSSage Weil 3414b6f3409bSSage Weil /* 3415f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3416f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3417f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3418f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3419f82a9901SFilipe Manana * partially overlaps it). 3420f82a9901SFilipe Manana */ 3421f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3422f82a9901SFilipe Manana drop_start = last_dest_end; 3423f82a9901SFilipe Manana else 3424f82a9901SFilipe Manana drop_start = new_key.offset; 3425f82a9901SFilipe Manana 3426f82a9901SFilipe Manana /* 3427b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3428b6f3409bSSage Weil * 1 - add new extent 3429b6f3409bSSage Weil * 1 - inode update 3430b6f3409bSSage Weil */ 3431b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3432a22285a6SYan, Zheng if (IS_ERR(trans)) { 3433a22285a6SYan, Zheng ret = PTR_ERR(trans); 3434a22285a6SYan, Zheng goto out; 3435a22285a6SYan, Zheng } 3436a22285a6SYan, Zheng 3437c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3438c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3439d72c0842SLi Zefan /* 3440d72c0842SLi Zefan * a | --- range to clone ---| b 3441d72c0842SLi Zefan * | ------------- extent ------------- | 3442d72c0842SLi Zefan */ 3443d72c0842SLi Zefan 344493915584SAntonio Ospite /* subtract range b */ 3445d72c0842SLi Zefan if (key.offset + datal > off + len) 3446d72c0842SLi Zefan datal = off + len - key.offset; 3447d72c0842SLi Zefan 344893915584SAntonio Ospite /* subtract range a */ 3449a22285a6SYan, Zheng if (off > key.offset) { 3450a22285a6SYan, Zheng datao += off - key.offset; 3451a22285a6SYan, Zheng datal -= off - key.offset; 3452a22285a6SYan, Zheng } 3453a22285a6SYan, Zheng 34545dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3455f82a9901SFilipe Manana drop_start, 3456a22285a6SYan, Zheng new_key.offset + datal, 34572671485dSJosef Bacik 1); 345879787eaaSJeff Mahoney if (ret) { 34593f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 346000fdf13aSLiu Bo btrfs_abort_transaction(trans, 346166642832SJeff Mahoney ret); 34623a45bb20SJeff Mahoney btrfs_end_transaction(trans); 346379787eaaSJeff Mahoney goto out; 346479787eaaSJeff Mahoney } 3465a22285a6SYan, Zheng 346631840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 346731840ae1SZheng Yan &new_key, size); 346879787eaaSJeff Mahoney if (ret) { 346966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 34703a45bb20SJeff Mahoney btrfs_end_transaction(trans); 347179787eaaSJeff Mahoney goto out; 347279787eaaSJeff Mahoney } 347331840ae1SZheng Yan 347431840ae1SZheng Yan leaf = path->nodes[0]; 347531840ae1SZheng Yan slot = path->slots[0]; 347631840ae1SZheng Yan write_extent_buffer(leaf, buf, 347731840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 347831840ae1SZheng Yan size); 3479ae01a0abSYan Zheng 3480f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3481f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3482c5c9cd4dSSage Weil 3483c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3484c5c9cd4dSSage Weil if (!disko) 3485c5c9cd4dSSage Weil datao = 0; 3486c5c9cd4dSSage Weil 3487c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3488c5c9cd4dSSage Weil datao); 3489c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3490c5c9cd4dSSage Weil datal); 3491fcebe456SJosef Bacik 3492c5c9cd4dSSage Weil if (disko) { 3493c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 34942ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 349584f7d8e6SJosef Bacik root, 34965d4f98a2SYan Zheng disko, diskl, 0, 3497f46b5a66SChristoph Hellwig root->root_key.objectid, 34984a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 3499b06c4bf5SFilipe Manana new_key.offset - datao); 350079787eaaSJeff Mahoney if (ret) { 350179787eaaSJeff Mahoney btrfs_abort_transaction(trans, 350279787eaaSJeff Mahoney ret); 35033a45bb20SJeff Mahoney btrfs_end_transaction(trans); 350479787eaaSJeff Mahoney goto out; 350579787eaaSJeff Mahoney 350679787eaaSJeff Mahoney } 3507f46b5a66SChristoph Hellwig } 3508c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3509c5c9cd4dSSage Weil u64 skip = 0; 3510c5c9cd4dSSage Weil u64 trim = 0; 3511ed958762SFilipe Manana 3512c5c9cd4dSSage Weil if (off > key.offset) { 3513c5c9cd4dSSage Weil skip = off - key.offset; 3514c5c9cd4dSSage Weil new_key.offset += skip; 351531840ae1SZheng Yan } 3516d397712bSChris Mason 3517c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3518c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3519d397712bSChris Mason 3520c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3521c5c9cd4dSSage Weil ret = -EINVAL; 35223a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3523c5c9cd4dSSage Weil goto out; 352431840ae1SZheng Yan } 3525c5c9cd4dSSage Weil size -= skip + trim; 3526c5c9cd4dSSage Weil datal -= skip + trim; 3527a22285a6SYan, Zheng 35284a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 35298039d87dSFilipe Manana trans, path, 35308039d87dSFilipe Manana &new_key, 3531f82a9901SFilipe Manana drop_start, 35328039d87dSFilipe Manana datal, 35338039d87dSFilipe Manana skip, size, buf); 353479787eaaSJeff Mahoney if (ret) { 35353f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 35363a29bc09SChris Mason btrfs_abort_transaction(trans, 353779787eaaSJeff Mahoney ret); 35383a45bb20SJeff Mahoney btrfs_end_transaction(trans); 353979787eaaSJeff Mahoney goto out; 354079787eaaSJeff Mahoney } 3541c5c9cd4dSSage Weil leaf = path->nodes[0]; 3542c5c9cd4dSSage Weil slot = path->slots[0]; 3543c5c9cd4dSSage Weil } 3544c5c9cd4dSSage Weil 35457ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 35467ffbb598SFilipe Manana if (drop_start < new_key.offset) 3547a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 354814f59796SFilipe Manana NULL, drop_start, 35497ffbb598SFilipe Manana new_key.offset - drop_start); 35507ffbb598SFilipe Manana 3551a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 3552a2f392e4SNikolay Borisov path, 0, 0); 35537ffbb598SFilipe Manana 3554c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3555b3b4aa74SDavid Sterba btrfs_release_path(path); 3556c5c9cd4dSSage Weil 355762e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 35580b246afaSJeff Mahoney fs_info->sectorsize); 3559f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3560f82a9901SFilipe Manana last_dest_end, 35611c919a5eSMark Fasheh destoff, olen, 35621c919a5eSMark Fasheh no_time_update); 3563f82a9901SFilipe Manana if (ret) 356479787eaaSJeff Mahoney goto out; 35652c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 35662c463823SFilipe Manana break; 3567a22285a6SYan, Zheng } 3568b3b4aa74SDavid Sterba btrfs_release_path(path); 3569df858e76SFilipe Manana key.offset = next_key_min_offset; 357069ae5e44SWang Xiaoguang 357169ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 357269ae5e44SWang Xiaoguang ret = -EINTR; 357369ae5e44SWang Xiaoguang goto out; 357469ae5e44SWang Xiaoguang } 3575ae01a0abSYan Zheng } 3576f46b5a66SChristoph Hellwig ret = 0; 357732b7c687SMark Fasheh 3578f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3579f82a9901SFilipe Manana /* 3580f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3581f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3582f82a9901SFilipe Manana */ 3583b3b4aa74SDavid Sterba btrfs_release_path(path); 3584f82a9901SFilipe Manana 3585f82a9901SFilipe Manana /* 3586f82a9901SFilipe Manana * 1 - remove extent(s) 3587f82a9901SFilipe Manana * 1 - inode update 3588f82a9901SFilipe Manana */ 3589f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3590f82a9901SFilipe Manana if (IS_ERR(trans)) { 3591f82a9901SFilipe Manana ret = PTR_ERR(trans); 3592f82a9901SFilipe Manana goto out; 3593f82a9901SFilipe Manana } 3594f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3595f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3596f82a9901SFilipe Manana if (ret) { 3597f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 359866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 35993a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3600f82a9901SFilipe Manana goto out; 3601f82a9901SFilipe Manana } 3602a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 3603a2f392e4SNikolay Borisov last_dest_end, 360414f59796SFilipe Manana destoff + len - last_dest_end); 3605f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 36061c919a5eSMark Fasheh destoff, olen, no_time_update); 3607f82a9901SFilipe Manana } 3608f82a9901SFilipe Manana 3609f46b5a66SChristoph Hellwig out: 361032b7c687SMark Fasheh btrfs_free_path(path); 361115351955SDavid Sterba kvfree(buf); 361232b7c687SMark Fasheh return ret; 361332b7c687SMark Fasheh } 361432b7c687SMark Fasheh 36153db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 361632b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 361732b7c687SMark Fasheh { 361854563d41SAl Viro struct inode *inode = file_inode(file); 36193db11b2eSZach Brown struct inode *src = file_inode(file_src); 36200b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 362132b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 362232b7c687SMark Fasheh int ret; 362332b7c687SMark Fasheh u64 len = olen; 36240b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 36253db11b2eSZach Brown int same_inode = src == inode; 362632b7c687SMark Fasheh 362732b7c687SMark Fasheh /* 362832b7c687SMark Fasheh * TODO: 362932b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 363032b7c687SMark Fasheh * decompress into destination's address_space (the file offset 363132b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 363232b7c687SMark Fasheh * otherwise reinsert) a subrange. 363300fdf13aSLiu Bo * 363400fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 363500fdf13aSLiu Bo * be either compressed or non-compressed. 363632b7c687SMark Fasheh */ 363732b7c687SMark Fasheh 363832b7c687SMark Fasheh if (btrfs_root_readonly(root)) 363932b7c687SMark Fasheh return -EROFS; 364032b7c687SMark Fasheh 36413db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 36423db11b2eSZach Brown src->i_sb != inode->i_sb) 36433db11b2eSZach Brown return -EXDEV; 364432b7c687SMark Fasheh 364532b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 364632b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 364732b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 36483db11b2eSZach Brown return -EINVAL; 364932b7c687SMark Fasheh 365032b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 36513db11b2eSZach Brown return -EISDIR; 365232b7c687SMark Fasheh 365332b7c687SMark Fasheh if (!same_inode) { 3654293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 365532b7c687SMark Fasheh } else { 36565955102cSAl Viro inode_lock(src); 365732b7c687SMark Fasheh } 365832b7c687SMark Fasheh 365932b7c687SMark Fasheh /* determine range to clone */ 366032b7c687SMark Fasheh ret = -EINVAL; 366132b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 366232b7c687SMark Fasheh goto out_unlock; 366332b7c687SMark Fasheh if (len == 0) 366432b7c687SMark Fasheh olen = len = src->i_size - off; 366532b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 366632b7c687SMark Fasheh if (off + len == src->i_size) 366732b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 366832b7c687SMark Fasheh 3669ccccf3d6SFilipe Manana if (len == 0) { 3670ccccf3d6SFilipe Manana ret = 0; 3671ccccf3d6SFilipe Manana goto out_unlock; 3672ccccf3d6SFilipe Manana } 3673ccccf3d6SFilipe Manana 367432b7c687SMark Fasheh /* verify the end result is block aligned */ 367532b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 367632b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 367732b7c687SMark Fasheh goto out_unlock; 367832b7c687SMark Fasheh 367932b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 368032b7c687SMark Fasheh if (same_inode) { 368132b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 368232b7c687SMark Fasheh goto out_unlock; 368332b7c687SMark Fasheh } 368432b7c687SMark Fasheh 368532b7c687SMark Fasheh if (destoff > inode->i_size) { 368632b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 368732b7c687SMark Fasheh if (ret) 368832b7c687SMark Fasheh goto out_unlock; 368932b7c687SMark Fasheh } 369032b7c687SMark Fasheh 3691c125b8bfSFilipe Manana /* 3692c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3693c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3694c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3695c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3696c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3697c125b8bfSFilipe Manana */ 3698c125b8bfSFilipe Manana if (same_inode) { 3699c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3700c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 370132b7c687SMark Fasheh 3702e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 3703c125b8bfSFilipe Manana } else { 3704e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 3705e0bd70c6SFilipe Manana true); 3706e0bd70c6SFilipe Manana } 3707e0bd70c6SFilipe Manana ASSERT(ret == 0); 3708e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3709e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3710e0bd70c6SFilipe Manana goto out_unlock; 3711c125b8bfSFilipe Manana } 371232b7c687SMark Fasheh 37131c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 371432b7c687SMark Fasheh 3715c125b8bfSFilipe Manana if (same_inode) { 3716c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3717c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3718c125b8bfSFilipe Manana 3719c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3720c125b8bfSFilipe Manana } else { 3721293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 3722c125b8bfSFilipe Manana } 3723c125b8bfSFilipe Manana /* 3724c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3725c125b8bfSFilipe Manana * data immediately and not the previous data. 3726c125b8bfSFilipe Manana */ 372765bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 372809cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 372909cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 3730f46b5a66SChristoph Hellwig out_unlock: 3731293a8489SMark Fasheh if (!same_inode) 3732293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 3733293a8489SMark Fasheh else 37345955102cSAl Viro inode_unlock(src); 37353db11b2eSZach Brown return ret; 37363db11b2eSZach Brown } 37373db11b2eSZach Brown 373804b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 373904b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 37403db11b2eSZach Brown { 374104b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 3742c5c9cd4dSSage Weil } 3743c5c9cd4dSSage Weil 37446ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 37456ef5ed0dSJosef Bacik { 3746496ad9aaSAl Viro struct inode *inode = file_inode(file); 37470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 37486ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 37496ef5ed0dSJosef Bacik struct btrfs_root *new_root; 37506ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 37516ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 37526ef5ed0dSJosef Bacik struct btrfs_path *path; 37536ef5ed0dSJosef Bacik struct btrfs_key location; 37546ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 37556ef5ed0dSJosef Bacik u64 objectid = 0; 37566ef5ed0dSJosef Bacik u64 dir_id; 37573c04ce01SMiao Xie int ret; 37586ef5ed0dSJosef Bacik 37596ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 37606ef5ed0dSJosef Bacik return -EPERM; 37616ef5ed0dSJosef Bacik 37623c04ce01SMiao Xie ret = mnt_want_write_file(file); 37633c04ce01SMiao Xie if (ret) 37643c04ce01SMiao Xie return ret; 37653c04ce01SMiao Xie 37663c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 37673c04ce01SMiao Xie ret = -EFAULT; 37683c04ce01SMiao Xie goto out; 37693c04ce01SMiao Xie } 37706ef5ed0dSJosef Bacik 37716ef5ed0dSJosef Bacik if (!objectid) 37721cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 37736ef5ed0dSJosef Bacik 37746ef5ed0dSJosef Bacik location.objectid = objectid; 37756ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 37766ef5ed0dSJosef Bacik location.offset = (u64)-1; 37776ef5ed0dSJosef Bacik 37780b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 37793c04ce01SMiao Xie if (IS_ERR(new_root)) { 37803c04ce01SMiao Xie ret = PTR_ERR(new_root); 37813c04ce01SMiao Xie goto out; 37823c04ce01SMiao Xie } 37836d6d2829Ssatoru takeuchi if (!is_fstree(new_root->objectid)) { 37846d6d2829Ssatoru takeuchi ret = -ENOENT; 37856d6d2829Ssatoru takeuchi goto out; 37866d6d2829Ssatoru takeuchi } 37876ef5ed0dSJosef Bacik 37886ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 37893c04ce01SMiao Xie if (!path) { 37903c04ce01SMiao Xie ret = -ENOMEM; 37913c04ce01SMiao Xie goto out; 37923c04ce01SMiao Xie } 37936ef5ed0dSJosef Bacik path->leave_spinning = 1; 37946ef5ed0dSJosef Bacik 37956ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 379698d5dc13STsutomu Itoh if (IS_ERR(trans)) { 37976ef5ed0dSJosef Bacik btrfs_free_path(path); 37983c04ce01SMiao Xie ret = PTR_ERR(trans); 37993c04ce01SMiao Xie goto out; 38006ef5ed0dSJosef Bacik } 38016ef5ed0dSJosef Bacik 38020b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 38030b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 38046ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 3805cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 38066ef5ed0dSJosef Bacik btrfs_free_path(path); 38073a45bb20SJeff Mahoney btrfs_end_transaction(trans); 38080b246afaSJeff Mahoney btrfs_err(fs_info, 38095d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 38103c04ce01SMiao Xie ret = -ENOENT; 38113c04ce01SMiao Xie goto out; 38126ef5ed0dSJosef Bacik } 38136ef5ed0dSJosef Bacik 38146ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 38156ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 38166ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 38176ef5ed0dSJosef Bacik btrfs_free_path(path); 38186ef5ed0dSJosef Bacik 38190b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 38203a45bb20SJeff Mahoney btrfs_end_transaction(trans); 38213c04ce01SMiao Xie out: 38223c04ce01SMiao Xie mnt_drop_write_file(file); 38233c04ce01SMiao Xie return ret; 38246ef5ed0dSJosef Bacik } 38256ef5ed0dSJosef Bacik 3826c065f5b1SSu Yue static void get_block_group_info(struct list_head *groups_list, 3827bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 3828bf5fc093SJosef Bacik { 3829bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 3830bf5fc093SJosef Bacik 3831bf5fc093SJosef Bacik space->total_bytes = 0; 3832bf5fc093SJosef Bacik space->used_bytes = 0; 3833bf5fc093SJosef Bacik space->flags = 0; 3834bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 3835bf5fc093SJosef Bacik space->flags = block_group->flags; 3836bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 3837bf5fc093SJosef Bacik space->used_bytes += 3838bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 3839bf5fc093SJosef Bacik } 3840bf5fc093SJosef Bacik } 3841bf5fc093SJosef Bacik 38422ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 38432ff7e61eSJeff Mahoney void __user *arg) 38441406e432SJosef Bacik { 38451406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 38461406e432SJosef Bacik struct btrfs_ioctl_space_info space; 38471406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 38487fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 384913f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 38501406e432SJosef Bacik struct btrfs_space_info *info; 3851315d8e98SColin Ian King static const u64 types[] = { 3852315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 3853bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 3854bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 3855315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 3856315d8e98SColin Ian King }; 3857bf5fc093SJosef Bacik int num_types = 4; 38587fde62bfSChris Mason int alloc_size; 38591406e432SJosef Bacik int ret = 0; 386051788b1bSDan Rosenberg u64 slot_count = 0; 3861bf5fc093SJosef Bacik int i, c; 38621406e432SJosef Bacik 38631406e432SJosef Bacik if (copy_from_user(&space_args, 38641406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 38651406e432SJosef Bacik sizeof(space_args))) 38661406e432SJosef Bacik return -EFAULT; 38671406e432SJosef Bacik 3868bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3869bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3870bf5fc093SJosef Bacik 3871bf5fc093SJosef Bacik info = NULL; 38727fde62bfSChris Mason rcu_read_lock(); 38730b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 3874bf5fc093SJosef Bacik list) { 3875bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3876bf5fc093SJosef Bacik info = tmp; 3877bf5fc093SJosef Bacik break; 3878bf5fc093SJosef Bacik } 3879bf5fc093SJosef Bacik } 38807fde62bfSChris Mason rcu_read_unlock(); 38811406e432SJosef Bacik 3882bf5fc093SJosef Bacik if (!info) 3883bf5fc093SJosef Bacik continue; 3884bf5fc093SJosef Bacik 3885bf5fc093SJosef Bacik down_read(&info->groups_sem); 3886bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3887bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 3888bf5fc093SJosef Bacik slot_count++; 3889bf5fc093SJosef Bacik } 3890bf5fc093SJosef Bacik up_read(&info->groups_sem); 3891bf5fc093SJosef Bacik } 3892bf5fc093SJosef Bacik 389336523e95SDavid Sterba /* 389436523e95SDavid Sterba * Global block reserve, exported as a space_info 389536523e95SDavid Sterba */ 389636523e95SDavid Sterba slot_count++; 389736523e95SDavid Sterba 38987fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 38997fde62bfSChris Mason if (space_args.space_slots == 0) { 39007fde62bfSChris Mason space_args.total_spaces = slot_count; 39017fde62bfSChris Mason goto out; 39027fde62bfSChris Mason } 3903bf5fc093SJosef Bacik 390451788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 3905bf5fc093SJosef Bacik 39067fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 3907bf5fc093SJosef Bacik 39087fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 39097fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 39107fde62bfSChris Mason */ 391109cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 39127fde62bfSChris Mason return -ENOMEM; 39137fde62bfSChris Mason 39147fde62bfSChris Mason space_args.total_spaces = 0; 39158d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 39167fde62bfSChris Mason if (!dest) 39177fde62bfSChris Mason return -ENOMEM; 39187fde62bfSChris Mason dest_orig = dest; 39197fde62bfSChris Mason 39207fde62bfSChris Mason /* now we have a buffer to copy into */ 3921bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3922bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3923bf5fc093SJosef Bacik 392451788b1bSDan Rosenberg if (!slot_count) 392551788b1bSDan Rosenberg break; 392651788b1bSDan Rosenberg 3927bf5fc093SJosef Bacik info = NULL; 39281406e432SJosef Bacik rcu_read_lock(); 39290b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 3930bf5fc093SJosef Bacik list) { 3931bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3932bf5fc093SJosef Bacik info = tmp; 39337fde62bfSChris Mason break; 3934bf5fc093SJosef Bacik } 3935bf5fc093SJosef Bacik } 3936bf5fc093SJosef Bacik rcu_read_unlock(); 39377fde62bfSChris Mason 3938bf5fc093SJosef Bacik if (!info) 3939bf5fc093SJosef Bacik continue; 3940bf5fc093SJosef Bacik down_read(&info->groups_sem); 3941bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3942bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 3943c065f5b1SSu Yue get_block_group_info(&info->block_groups[c], 3944c065f5b1SSu Yue &space); 39457fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 39461406e432SJosef Bacik dest++; 39471406e432SJosef Bacik space_args.total_spaces++; 394851788b1bSDan Rosenberg slot_count--; 39491406e432SJosef Bacik } 395051788b1bSDan Rosenberg if (!slot_count) 395151788b1bSDan Rosenberg break; 3952bf5fc093SJosef Bacik } 3953bf5fc093SJosef Bacik up_read(&info->groups_sem); 3954bf5fc093SJosef Bacik } 39551406e432SJosef Bacik 395636523e95SDavid Sterba /* 395736523e95SDavid Sterba * Add global block reserve 395836523e95SDavid Sterba */ 395936523e95SDavid Sterba if (slot_count) { 39600b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 396136523e95SDavid Sterba 396236523e95SDavid Sterba spin_lock(&block_rsv->lock); 396336523e95SDavid Sterba space.total_bytes = block_rsv->size; 396436523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 396536523e95SDavid Sterba spin_unlock(&block_rsv->lock); 396636523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 396736523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 396836523e95SDavid Sterba space_args.total_spaces++; 396936523e95SDavid Sterba } 397036523e95SDavid Sterba 39712eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 39727fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 39737fde62bfSChris Mason 39747fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 39757fde62bfSChris Mason ret = -EFAULT; 39767fde62bfSChris Mason 39777fde62bfSChris Mason kfree(dest_orig); 39787fde62bfSChris Mason out: 39797fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 39801406e432SJosef Bacik ret = -EFAULT; 39811406e432SJosef Bacik 39821406e432SJosef Bacik return ret; 39831406e432SJosef Bacik } 39841406e432SJosef Bacik 39859a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 39869a8c28beSMiao Xie void __user *argp) 398746204592SSage Weil { 398846204592SSage Weil struct btrfs_trans_handle *trans; 398946204592SSage Weil u64 transid; 3990db5b493aSTsutomu Itoh int ret; 399146204592SSage Weil 3992d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 3993ff7c1d33SMiao Xie if (IS_ERR(trans)) { 3994ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 399598d5dc13STsutomu Itoh return PTR_ERR(trans); 3996ff7c1d33SMiao Xie 3997ff7c1d33SMiao Xie /* No running transaction, don't bother */ 3998ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 3999ff7c1d33SMiao Xie goto out; 4000ff7c1d33SMiao Xie } 400146204592SSage Weil transid = trans->transid; 40023a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 40038b2b2d3cSTsutomu Itoh if (ret) { 40043a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4005db5b493aSTsutomu Itoh return ret; 40068b2b2d3cSTsutomu Itoh } 4007ff7c1d33SMiao Xie out: 400846204592SSage Weil if (argp) 400946204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 401046204592SSage Weil return -EFAULT; 401146204592SSage Weil return 0; 401246204592SSage Weil } 401346204592SSage Weil 40142ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 40159a8c28beSMiao Xie void __user *argp) 401646204592SSage Weil { 401746204592SSage Weil u64 transid; 401846204592SSage Weil 401946204592SSage Weil if (argp) { 402046204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 402146204592SSage Weil return -EFAULT; 402246204592SSage Weil } else { 402346204592SSage Weil transid = 0; /* current trans */ 402446204592SSage Weil } 40252ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 402646204592SSage Weil } 402746204592SSage Weil 4028b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4029475f6387SJan Schmidt { 40300b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4031475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4032b8e95489SMiao Xie int ret; 4033475f6387SJan Schmidt 4034475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4035475f6387SJan Schmidt return -EPERM; 4036475f6387SJan Schmidt 4037475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4038475f6387SJan Schmidt if (IS_ERR(sa)) 4039475f6387SJan Schmidt return PTR_ERR(sa); 4040475f6387SJan Schmidt 4041b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4042b8e95489SMiao Xie ret = mnt_want_write_file(file); 4043b8e95489SMiao Xie if (ret) 4044b8e95489SMiao Xie goto out; 4045b8e95489SMiao Xie } 4046b8e95489SMiao Xie 40470b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 404863a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 404963a212abSStefan Behrens 0); 4050475f6387SJan Schmidt 4051475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4052475f6387SJan Schmidt ret = -EFAULT; 4053475f6387SJan Schmidt 4054b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4055b8e95489SMiao Xie mnt_drop_write_file(file); 4056b8e95489SMiao Xie out: 4057475f6387SJan Schmidt kfree(sa); 4058475f6387SJan Schmidt return ret; 4059475f6387SJan Schmidt } 4060475f6387SJan Schmidt 40612ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4062475f6387SJan Schmidt { 4063475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4064475f6387SJan Schmidt return -EPERM; 4065475f6387SJan Schmidt 40662ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4067475f6387SJan Schmidt } 4068475f6387SJan Schmidt 40692ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4070475f6387SJan Schmidt void __user *arg) 4071475f6387SJan Schmidt { 4072475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4073475f6387SJan Schmidt int ret; 4074475f6387SJan Schmidt 4075475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4076475f6387SJan Schmidt return -EPERM; 4077475f6387SJan Schmidt 4078475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4079475f6387SJan Schmidt if (IS_ERR(sa)) 4080475f6387SJan Schmidt return PTR_ERR(sa); 4081475f6387SJan Schmidt 40822ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4083475f6387SJan Schmidt 4084475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4085475f6387SJan Schmidt ret = -EFAULT; 4086475f6387SJan Schmidt 4087475f6387SJan Schmidt kfree(sa); 4088475f6387SJan Schmidt return ret; 4089475f6387SJan Schmidt } 4090475f6387SJan Schmidt 40912ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4092b27f7c0cSDavid Sterba void __user *arg) 4093c11d2c23SStefan Behrens { 4094c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4095c11d2c23SStefan Behrens int ret; 4096c11d2c23SStefan Behrens 4097c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4098c11d2c23SStefan Behrens if (IS_ERR(sa)) 4099c11d2c23SStefan Behrens return PTR_ERR(sa); 4100c11d2c23SStefan Behrens 4101b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4102b27f7c0cSDavid Sterba kfree(sa); 4103b27f7c0cSDavid Sterba return -EPERM; 4104b27f7c0cSDavid Sterba } 4105b27f7c0cSDavid Sterba 41062ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4107c11d2c23SStefan Behrens 4108c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4109c11d2c23SStefan Behrens ret = -EFAULT; 4110c11d2c23SStefan Behrens 4111c11d2c23SStefan Behrens kfree(sa); 4112c11d2c23SStefan Behrens return ret; 4113c11d2c23SStefan Behrens } 4114c11d2c23SStefan Behrens 41152ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 41162ff7e61eSJeff Mahoney void __user *arg) 41173f6bcfbdSStefan Behrens { 41183f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 41193f6bcfbdSStefan Behrens int ret; 41203f6bcfbdSStefan Behrens 41213f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 41223f6bcfbdSStefan Behrens return -EPERM; 41233f6bcfbdSStefan Behrens 41243f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 41253f6bcfbdSStefan Behrens if (IS_ERR(p)) 41263f6bcfbdSStefan Behrens return PTR_ERR(p); 41273f6bcfbdSStefan Behrens 41283f6bcfbdSStefan Behrens switch (p->cmd) { 41293f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4130bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4131adfa97cbSIlya Dryomov ret = -EROFS; 4132adfa97cbSIlya Dryomov goto out; 4133adfa97cbSIlya Dryomov } 4134171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4135e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 41363f6bcfbdSStefan Behrens } else { 41372ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4138171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 41393f6bcfbdSStefan Behrens } 41403f6bcfbdSStefan Behrens break; 41413f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 41420b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 41433f6bcfbdSStefan Behrens ret = 0; 41443f6bcfbdSStefan Behrens break; 41453f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 414617d202b9SAnand Jain p->result = btrfs_dev_replace_cancel(fs_info); 414797282031SAnand Jain ret = 0; 41483f6bcfbdSStefan Behrens break; 41493f6bcfbdSStefan Behrens default: 41503f6bcfbdSStefan Behrens ret = -EINVAL; 41513f6bcfbdSStefan Behrens break; 41523f6bcfbdSStefan Behrens } 41533f6bcfbdSStefan Behrens 41543f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 41553f6bcfbdSStefan Behrens ret = -EFAULT; 4156adfa97cbSIlya Dryomov out: 41573f6bcfbdSStefan Behrens kfree(p); 41583f6bcfbdSStefan Behrens return ret; 41593f6bcfbdSStefan Behrens } 41603f6bcfbdSStefan Behrens 4161d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4162d7728c96SJan Schmidt { 4163d7728c96SJan Schmidt int ret = 0; 4164d7728c96SJan Schmidt int i; 4165740c3d22SChris Mason u64 rel_ptr; 4166d7728c96SJan Schmidt int size; 4167806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4168d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4169d7728c96SJan Schmidt struct btrfs_path *path; 4170d7728c96SJan Schmidt 417182b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4172d7728c96SJan Schmidt return -EPERM; 4173d7728c96SJan Schmidt 4174d7728c96SJan Schmidt path = btrfs_alloc_path(); 4175d7728c96SJan Schmidt if (!path) { 4176d7728c96SJan Schmidt ret = -ENOMEM; 4177d7728c96SJan Schmidt goto out; 4178d7728c96SJan Schmidt } 4179d7728c96SJan Schmidt 4180d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4181d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4182d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4183d7728c96SJan Schmidt ipa = NULL; 4184d7728c96SJan Schmidt goto out; 4185d7728c96SJan Schmidt } 4186d7728c96SJan Schmidt 4187d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4188d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4189d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4190d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4191d7728c96SJan Schmidt ipath = NULL; 4192d7728c96SJan Schmidt goto out; 4193d7728c96SJan Schmidt } 4194d7728c96SJan Schmidt 4195d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4196d7728c96SJan Schmidt if (ret < 0) 4197d7728c96SJan Schmidt goto out; 4198d7728c96SJan Schmidt 4199d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4200745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4201745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4202740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4203d7728c96SJan Schmidt } 4204d7728c96SJan Schmidt 4205718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4206718dc5faSOmar Sandoval ipath->fspath, size); 4207d7728c96SJan Schmidt if (ret) { 4208d7728c96SJan Schmidt ret = -EFAULT; 4209d7728c96SJan Schmidt goto out; 4210d7728c96SJan Schmidt } 4211d7728c96SJan Schmidt 4212d7728c96SJan Schmidt out: 4213d7728c96SJan Schmidt btrfs_free_path(path); 4214d7728c96SJan Schmidt free_ipath(ipath); 4215d7728c96SJan Schmidt kfree(ipa); 4216d7728c96SJan Schmidt 4217d7728c96SJan Schmidt return ret; 4218d7728c96SJan Schmidt } 4219d7728c96SJan Schmidt 4220d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4221d7728c96SJan Schmidt { 4222d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4223d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4224d7728c96SJan Schmidt 4225d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4226d7728c96SJan Schmidt inodes->bytes_left -= c; 4227d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4228d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4229d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4230d7728c96SJan Schmidt inodes->elem_cnt += 3; 4231d7728c96SJan Schmidt } else { 4232d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4233d7728c96SJan Schmidt inodes->bytes_left = 0; 4234d7728c96SJan Schmidt inodes->elem_missed += 3; 4235d7728c96SJan Schmidt } 4236d7728c96SJan Schmidt 4237d7728c96SJan Schmidt return 0; 4238d7728c96SJan Schmidt } 4239d7728c96SJan Schmidt 42402ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4241d24a67b2SZygo Blaxell void __user *arg, int version) 4242d7728c96SJan Schmidt { 4243d7728c96SJan Schmidt int ret = 0; 4244d7728c96SJan Schmidt int size; 4245d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4246d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4247d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4248d24a67b2SZygo Blaxell bool ignore_offset; 4249d7728c96SJan Schmidt 4250d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4251d7728c96SJan Schmidt return -EPERM; 4252d7728c96SJan Schmidt 4253d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 42547b9ea627SShailendra Verma if (IS_ERR(loi)) 42557b9ea627SShailendra Verma return PTR_ERR(loi); 4256d7728c96SJan Schmidt 4257d24a67b2SZygo Blaxell if (version == 1) { 4258d24a67b2SZygo Blaxell ignore_offset = false; 4259b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_64K); 4260d24a67b2SZygo Blaxell } else { 4261d24a67b2SZygo Blaxell /* All reserved bits must be 0 for now */ 4262d24a67b2SZygo Blaxell if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { 4263d24a67b2SZygo Blaxell ret = -EINVAL; 4264d24a67b2SZygo Blaxell goto out_loi; 4265d24a67b2SZygo Blaxell } 4266d24a67b2SZygo Blaxell /* Only accept flags we have defined so far */ 4267d24a67b2SZygo Blaxell if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { 4268d24a67b2SZygo Blaxell ret = -EINVAL; 4269d24a67b2SZygo Blaxell goto out_loi; 4270d24a67b2SZygo Blaxell } 4271d24a67b2SZygo Blaxell ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 4272b115e3bcSZygo Blaxell size = min_t(u32, loi->size, SZ_16M); 4273d24a67b2SZygo Blaxell } 4274d24a67b2SZygo Blaxell 4275d7728c96SJan Schmidt path = btrfs_alloc_path(); 4276d7728c96SJan Schmidt if (!path) { 4277d7728c96SJan Schmidt ret = -ENOMEM; 4278d7728c96SJan Schmidt goto out; 4279d7728c96SJan Schmidt } 4280d7728c96SJan Schmidt 4281d7728c96SJan Schmidt inodes = init_data_container(size); 4282d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4283d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4284d7728c96SJan Schmidt inodes = NULL; 4285d7728c96SJan Schmidt goto out; 4286d7728c96SJan Schmidt } 4287d7728c96SJan Schmidt 42882ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4289d24a67b2SZygo Blaxell build_ino_list, inodes, ignore_offset); 4290df031f07SLiu Bo if (ret == -EINVAL) 4291d7728c96SJan Schmidt ret = -ENOENT; 4292d7728c96SJan Schmidt if (ret < 0) 4293d7728c96SJan Schmidt goto out; 4294d7728c96SJan Schmidt 4295718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4296718dc5faSOmar Sandoval size); 4297d7728c96SJan Schmidt if (ret) 4298d7728c96SJan Schmidt ret = -EFAULT; 4299d7728c96SJan Schmidt 4300d7728c96SJan Schmidt out: 4301d7728c96SJan Schmidt btrfs_free_path(path); 4302f54de068SDavid Sterba kvfree(inodes); 4303d24a67b2SZygo Blaxell out_loi: 4304d7728c96SJan Schmidt kfree(loi); 4305d7728c96SJan Schmidt 4306d7728c96SJan Schmidt return ret; 4307d7728c96SJan Schmidt } 4308d7728c96SJan Schmidt 4309008ef096SDavid Sterba void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 4310c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4311c9e9f97bSIlya Dryomov { 4312c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4313c9e9f97bSIlya Dryomov 4314c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4315c9e9f97bSIlya Dryomov 43163009a62fSDavid Sterba if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 4317837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4318837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4319837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4320a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4321a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4322837d5b6eSIlya Dryomov 4323c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4324c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4325c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 432619a39dceSIlya Dryomov 432719a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 432819a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 432919a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 4330c9e9f97bSIlya Dryomov } 4331c9e9f97bSIlya Dryomov 43329ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4333c9e9f97bSIlya Dryomov { 4334496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4335c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4336c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4337c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4338ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4339c9e9f97bSIlya Dryomov int ret; 4340c9e9f97bSIlya Dryomov 4341c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4342c9e9f97bSIlya Dryomov return -EPERM; 4343c9e9f97bSIlya Dryomov 4344e54bfa31SLiu Bo ret = mnt_want_write_file(file); 43459ba1f6e4SLiu Bo if (ret) 43469ba1f6e4SLiu Bo return ret; 43479ba1f6e4SLiu Bo 4348ed0fb78fSIlya Dryomov again: 4349171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4350c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4351ed0fb78fSIlya Dryomov need_unlock = true; 4352ed0fb78fSIlya Dryomov goto locked; 4353ed0fb78fSIlya Dryomov } 4354ed0fb78fSIlya Dryomov 4355ed0fb78fSIlya Dryomov /* 435601327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4357ed0fb78fSIlya Dryomov * (1) some other op is running 4358ed0fb78fSIlya Dryomov * (2) balance is running 4359ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4360ed0fb78fSIlya Dryomov */ 4361ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4362ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4363ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 43643009a62fSDavid Sterba if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4365ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4366dccdb07bSDavid Sterba /* 4367dccdb07bSDavid Sterba * Lock released to allow other waiters to continue, 4368dccdb07bSDavid Sterba * we'll reexamine the status again. 4369dccdb07bSDavid Sterba */ 4370ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4371ed0fb78fSIlya Dryomov 4372ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 43733009a62fSDavid Sterba !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 4374ed0fb78fSIlya Dryomov /* this is (3) */ 4375ed0fb78fSIlya Dryomov need_unlock = false; 4376ed0fb78fSIlya Dryomov goto locked; 4377ed0fb78fSIlya Dryomov } 4378ed0fb78fSIlya Dryomov 4379ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4380ed0fb78fSIlya Dryomov goto again; 4381ed0fb78fSIlya Dryomov } else { 4382ed0fb78fSIlya Dryomov /* this is (2) */ 4383ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4384ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4385ed0fb78fSIlya Dryomov goto out; 4386ed0fb78fSIlya Dryomov } 4387ed0fb78fSIlya Dryomov } else { 4388ed0fb78fSIlya Dryomov /* this is (1) */ 4389ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4390e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4391ed0fb78fSIlya Dryomov goto out; 4392ed0fb78fSIlya Dryomov } 4393ed0fb78fSIlya Dryomov 4394ed0fb78fSIlya Dryomov locked: 4395171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4396c9e9f97bSIlya Dryomov 4397c9e9f97bSIlya Dryomov if (arg) { 4398c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4399c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4400c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4401ed0fb78fSIlya Dryomov goto out_unlock; 4402c9e9f97bSIlya Dryomov } 4403de322263SIlya Dryomov 4404de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4405de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4406de322263SIlya Dryomov ret = -ENOTCONN; 4407de322263SIlya Dryomov goto out_bargs; 4408de322263SIlya Dryomov } 4409de322263SIlya Dryomov 4410de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4411de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4412de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4413de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4414de322263SIlya Dryomov 4415de322263SIlya Dryomov goto do_balance; 4416de322263SIlya Dryomov } 4417c9e9f97bSIlya Dryomov } else { 4418c9e9f97bSIlya Dryomov bargs = NULL; 4419c9e9f97bSIlya Dryomov } 4420c9e9f97bSIlya Dryomov 4421ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4422837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4423837d5b6eSIlya Dryomov goto out_bargs; 4424837d5b6eSIlya Dryomov } 4425837d5b6eSIlya Dryomov 44268d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4427c9e9f97bSIlya Dryomov if (!bctl) { 4428c9e9f97bSIlya Dryomov ret = -ENOMEM; 4429c9e9f97bSIlya Dryomov goto out_bargs; 4430c9e9f97bSIlya Dryomov } 4431c9e9f97bSIlya Dryomov 4432c9e9f97bSIlya Dryomov if (arg) { 4433c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4434c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4435c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4436c9e9f97bSIlya Dryomov 4437c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4438f43ffb60SIlya Dryomov } else { 4439f43ffb60SIlya Dryomov /* balance everything - no filters */ 4440f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4441c9e9f97bSIlya Dryomov } 4442c9e9f97bSIlya Dryomov 44438eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 44448eb93459SDavid Sterba ret = -EINVAL; 44450f89abf5SChristian Engelmayer goto out_bctl; 44468eb93459SDavid Sterba } 44478eb93459SDavid Sterba 4448de322263SIlya Dryomov do_balance: 4449c9e9f97bSIlya Dryomov /* 4450149196a2SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to 4451149196a2SDavid Sterba * btrfs_balance. bctl is freed in reset_balance_state, or, if 4452149196a2SDavid Sterba * restriper was paused all the way until unmount, in free_fs_info. 4453149196a2SDavid Sterba * The flag should be cleared after reset_balance_state. 4454c9e9f97bSIlya Dryomov */ 4455ed0fb78fSIlya Dryomov need_unlock = false; 4456ed0fb78fSIlya Dryomov 44576fcf6e2bSDavid Sterba ret = btrfs_balance(fs_info, bctl, bargs); 44580f89abf5SChristian Engelmayer bctl = NULL; 4459ed0fb78fSIlya Dryomov 4460c9e9f97bSIlya Dryomov if (arg) { 4461c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4462c9e9f97bSIlya Dryomov ret = -EFAULT; 4463c9e9f97bSIlya Dryomov } 4464c9e9f97bSIlya Dryomov 44650f89abf5SChristian Engelmayer out_bctl: 44660f89abf5SChristian Engelmayer kfree(bctl); 4467c9e9f97bSIlya Dryomov out_bargs: 4468c9e9f97bSIlya Dryomov kfree(bargs); 4469ed0fb78fSIlya Dryomov out_unlock: 4470c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4471ed0fb78fSIlya Dryomov if (need_unlock) 4472171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4473ed0fb78fSIlya Dryomov out: 4474e54bfa31SLiu Bo mnt_drop_write_file(file); 4475c9e9f97bSIlya Dryomov return ret; 4476c9e9f97bSIlya Dryomov } 4477c9e9f97bSIlya Dryomov 44782ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4479837d5b6eSIlya Dryomov { 4480837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4481837d5b6eSIlya Dryomov return -EPERM; 4482837d5b6eSIlya Dryomov 4483837d5b6eSIlya Dryomov switch (cmd) { 4484837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 44850b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4486a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 44870b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4488837d5b6eSIlya Dryomov } 4489837d5b6eSIlya Dryomov 4490837d5b6eSIlya Dryomov return -EINVAL; 4491837d5b6eSIlya Dryomov } 4492837d5b6eSIlya Dryomov 44932ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 449419a39dceSIlya Dryomov void __user *arg) 449519a39dceSIlya Dryomov { 449619a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 449719a39dceSIlya Dryomov int ret = 0; 449819a39dceSIlya Dryomov 449919a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 450019a39dceSIlya Dryomov return -EPERM; 450119a39dceSIlya Dryomov 450219a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 450319a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 450419a39dceSIlya Dryomov ret = -ENOTCONN; 450519a39dceSIlya Dryomov goto out; 450619a39dceSIlya Dryomov } 450719a39dceSIlya Dryomov 45088d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 450919a39dceSIlya Dryomov if (!bargs) { 451019a39dceSIlya Dryomov ret = -ENOMEM; 451119a39dceSIlya Dryomov goto out; 451219a39dceSIlya Dryomov } 451319a39dceSIlya Dryomov 4514008ef096SDavid Sterba btrfs_update_ioctl_balance_args(fs_info, bargs); 451519a39dceSIlya Dryomov 451619a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 451719a39dceSIlya Dryomov ret = -EFAULT; 451819a39dceSIlya Dryomov 451919a39dceSIlya Dryomov kfree(bargs); 452019a39dceSIlya Dryomov out: 452119a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 452219a39dceSIlya Dryomov return ret; 452319a39dceSIlya Dryomov } 452419a39dceSIlya Dryomov 4525905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 45265d13a37bSArne Jansen { 45270b246afaSJeff Mahoney struct inode *inode = file_inode(file); 45280b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 45295d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 45305d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 45315d13a37bSArne Jansen int ret; 45325d13a37bSArne Jansen int err; 45335d13a37bSArne Jansen 45345d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 45355d13a37bSArne Jansen return -EPERM; 45365d13a37bSArne Jansen 4537905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4538905b0ddaSMiao Xie if (ret) 4539905b0ddaSMiao Xie return ret; 45405d13a37bSArne Jansen 45415d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4542905b0ddaSMiao Xie if (IS_ERR(sa)) { 4543905b0ddaSMiao Xie ret = PTR_ERR(sa); 4544905b0ddaSMiao Xie goto drop_write; 4545905b0ddaSMiao Xie } 45465d13a37bSArne Jansen 45470b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 45480b246afaSJeff Mahoney trans = btrfs_start_transaction(fs_info->tree_root, 2); 45495d13a37bSArne Jansen if (IS_ERR(trans)) { 45505d13a37bSArne Jansen ret = PTR_ERR(trans); 45515d13a37bSArne Jansen goto out; 45525d13a37bSArne Jansen } 45535d13a37bSArne Jansen 45545d13a37bSArne Jansen switch (sa->cmd) { 45555d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 45560b246afaSJeff Mahoney ret = btrfs_quota_enable(trans, fs_info); 45575d13a37bSArne Jansen break; 45585d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 45590b246afaSJeff Mahoney ret = btrfs_quota_disable(trans, fs_info); 45605d13a37bSArne Jansen break; 45615d13a37bSArne Jansen default: 45625d13a37bSArne Jansen ret = -EINVAL; 45635d13a37bSArne Jansen break; 45645d13a37bSArne Jansen } 45655d13a37bSArne Jansen 45663a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 45675d13a37bSArne Jansen if (err && !ret) 45685d13a37bSArne Jansen ret = err; 45695d13a37bSArne Jansen out: 45705d13a37bSArne Jansen kfree(sa); 45710b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4572905b0ddaSMiao Xie drop_write: 4573905b0ddaSMiao Xie mnt_drop_write_file(file); 45745d13a37bSArne Jansen return ret; 45755d13a37bSArne Jansen } 45765d13a37bSArne Jansen 4577905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 45785d13a37bSArne Jansen { 45790b246afaSJeff Mahoney struct inode *inode = file_inode(file); 45800b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 45810b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 45825d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 45835d13a37bSArne Jansen struct btrfs_trans_handle *trans; 45845d13a37bSArne Jansen int ret; 45855d13a37bSArne Jansen int err; 45865d13a37bSArne Jansen 45875d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 45885d13a37bSArne Jansen return -EPERM; 45895d13a37bSArne Jansen 4590905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4591905b0ddaSMiao Xie if (ret) 4592905b0ddaSMiao Xie return ret; 45935d13a37bSArne Jansen 45945d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4595905b0ddaSMiao Xie if (IS_ERR(sa)) { 4596905b0ddaSMiao Xie ret = PTR_ERR(sa); 4597905b0ddaSMiao Xie goto drop_write; 4598905b0ddaSMiao Xie } 45995d13a37bSArne Jansen 46005d13a37bSArne Jansen trans = btrfs_join_transaction(root); 46015d13a37bSArne Jansen if (IS_ERR(trans)) { 46025d13a37bSArne Jansen ret = PTR_ERR(trans); 46035d13a37bSArne Jansen goto out; 46045d13a37bSArne Jansen } 46055d13a37bSArne Jansen 46065d13a37bSArne Jansen if (sa->assign) { 46070b246afaSJeff Mahoney ret = btrfs_add_qgroup_relation(trans, fs_info, 46085d13a37bSArne Jansen sa->src, sa->dst); 46095d13a37bSArne Jansen } else { 46100b246afaSJeff Mahoney ret = btrfs_del_qgroup_relation(trans, fs_info, 46115d13a37bSArne Jansen sa->src, sa->dst); 46125d13a37bSArne Jansen } 46135d13a37bSArne Jansen 4614e082f563SQu Wenruo /* update qgroup status and info */ 46150b246afaSJeff Mahoney err = btrfs_run_qgroups(trans, fs_info); 4616e082f563SQu Wenruo if (err < 0) 46170b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4618ad8403dfSAnand Jain "failed to update qgroup status and info"); 46193a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 46205d13a37bSArne Jansen if (err && !ret) 46215d13a37bSArne Jansen ret = err; 46225d13a37bSArne Jansen 46235d13a37bSArne Jansen out: 46245d13a37bSArne Jansen kfree(sa); 4625905b0ddaSMiao Xie drop_write: 4626905b0ddaSMiao Xie mnt_drop_write_file(file); 46275d13a37bSArne Jansen return ret; 46285d13a37bSArne Jansen } 46295d13a37bSArne Jansen 4630905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 46315d13a37bSArne Jansen { 46320b246afaSJeff Mahoney struct inode *inode = file_inode(file); 46330b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 46340b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 46355d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 46365d13a37bSArne Jansen struct btrfs_trans_handle *trans; 46375d13a37bSArne Jansen int ret; 46385d13a37bSArne Jansen int err; 46395d13a37bSArne Jansen 46405d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 46415d13a37bSArne Jansen return -EPERM; 46425d13a37bSArne Jansen 4643905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4644905b0ddaSMiao Xie if (ret) 4645905b0ddaSMiao Xie return ret; 46465d13a37bSArne Jansen 46475d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4648905b0ddaSMiao Xie if (IS_ERR(sa)) { 4649905b0ddaSMiao Xie ret = PTR_ERR(sa); 4650905b0ddaSMiao Xie goto drop_write; 4651905b0ddaSMiao Xie } 46525d13a37bSArne Jansen 4653d86e56cfSMiao Xie if (!sa->qgroupid) { 4654d86e56cfSMiao Xie ret = -EINVAL; 4655d86e56cfSMiao Xie goto out; 4656d86e56cfSMiao Xie } 4657d86e56cfSMiao Xie 46585d13a37bSArne Jansen trans = btrfs_join_transaction(root); 46595d13a37bSArne Jansen if (IS_ERR(trans)) { 46605d13a37bSArne Jansen ret = PTR_ERR(trans); 46615d13a37bSArne Jansen goto out; 46625d13a37bSArne Jansen } 46635d13a37bSArne Jansen 46645d13a37bSArne Jansen if (sa->create) { 46650b246afaSJeff Mahoney ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); 46665d13a37bSArne Jansen } else { 46670b246afaSJeff Mahoney ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); 46685d13a37bSArne Jansen } 46695d13a37bSArne Jansen 46703a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 46715d13a37bSArne Jansen if (err && !ret) 46725d13a37bSArne Jansen ret = err; 46735d13a37bSArne Jansen 46745d13a37bSArne Jansen out: 46755d13a37bSArne Jansen kfree(sa); 4676905b0ddaSMiao Xie drop_write: 4677905b0ddaSMiao Xie mnt_drop_write_file(file); 46785d13a37bSArne Jansen return ret; 46795d13a37bSArne Jansen } 46805d13a37bSArne Jansen 4681905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 46825d13a37bSArne Jansen { 46830b246afaSJeff Mahoney struct inode *inode = file_inode(file); 46840b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 46850b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 46865d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 46875d13a37bSArne Jansen struct btrfs_trans_handle *trans; 46885d13a37bSArne Jansen int ret; 46895d13a37bSArne Jansen int err; 46905d13a37bSArne Jansen u64 qgroupid; 46915d13a37bSArne Jansen 46925d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 46935d13a37bSArne Jansen return -EPERM; 46945d13a37bSArne Jansen 4695905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4696905b0ddaSMiao Xie if (ret) 4697905b0ddaSMiao Xie return ret; 46985d13a37bSArne Jansen 46995d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4700905b0ddaSMiao Xie if (IS_ERR(sa)) { 4701905b0ddaSMiao Xie ret = PTR_ERR(sa); 4702905b0ddaSMiao Xie goto drop_write; 4703905b0ddaSMiao Xie } 47045d13a37bSArne Jansen 47055d13a37bSArne Jansen trans = btrfs_join_transaction(root); 47065d13a37bSArne Jansen if (IS_ERR(trans)) { 47075d13a37bSArne Jansen ret = PTR_ERR(trans); 47085d13a37bSArne Jansen goto out; 47095d13a37bSArne Jansen } 47105d13a37bSArne Jansen 47115d13a37bSArne Jansen qgroupid = sa->qgroupid; 47125d13a37bSArne Jansen if (!qgroupid) { 47135d13a37bSArne Jansen /* take the current subvol as qgroup */ 47145d13a37bSArne Jansen qgroupid = root->root_key.objectid; 47155d13a37bSArne Jansen } 47165d13a37bSArne Jansen 47170b246afaSJeff Mahoney ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); 47185d13a37bSArne Jansen 47193a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 47205d13a37bSArne Jansen if (err && !ret) 47215d13a37bSArne Jansen ret = err; 47225d13a37bSArne Jansen 47235d13a37bSArne Jansen out: 47245d13a37bSArne Jansen kfree(sa); 4725905b0ddaSMiao Xie drop_write: 4726905b0ddaSMiao Xie mnt_drop_write_file(file); 47275d13a37bSArne Jansen return ret; 47285d13a37bSArne Jansen } 47295d13a37bSArne Jansen 47302f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 47312f232036SJan Schmidt { 47320b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47330b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47342f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 47352f232036SJan Schmidt int ret; 47362f232036SJan Schmidt 47372f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 47382f232036SJan Schmidt return -EPERM; 47392f232036SJan Schmidt 47402f232036SJan Schmidt ret = mnt_want_write_file(file); 47412f232036SJan Schmidt if (ret) 47422f232036SJan Schmidt return ret; 47432f232036SJan Schmidt 47442f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 47452f232036SJan Schmidt if (IS_ERR(qsa)) { 47462f232036SJan Schmidt ret = PTR_ERR(qsa); 47472f232036SJan Schmidt goto drop_write; 47482f232036SJan Schmidt } 47492f232036SJan Schmidt 47502f232036SJan Schmidt if (qsa->flags) { 47512f232036SJan Schmidt ret = -EINVAL; 47522f232036SJan Schmidt goto out; 47532f232036SJan Schmidt } 47542f232036SJan Schmidt 47550b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 47562f232036SJan Schmidt 47572f232036SJan Schmidt out: 47582f232036SJan Schmidt kfree(qsa); 47592f232036SJan Schmidt drop_write: 47602f232036SJan Schmidt mnt_drop_write_file(file); 47612f232036SJan Schmidt return ret; 47622f232036SJan Schmidt } 47632f232036SJan Schmidt 47642f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 47652f232036SJan Schmidt { 47660b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47670b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47682f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 47692f232036SJan Schmidt int ret = 0; 47702f232036SJan Schmidt 47712f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 47722f232036SJan Schmidt return -EPERM; 47732f232036SJan Schmidt 47748d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 47752f232036SJan Schmidt if (!qsa) 47762f232036SJan Schmidt return -ENOMEM; 47772f232036SJan Schmidt 47780b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 47792f232036SJan Schmidt qsa->flags = 1; 47800b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 47812f232036SJan Schmidt } 47822f232036SJan Schmidt 47832f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 47842f232036SJan Schmidt ret = -EFAULT; 47852f232036SJan Schmidt 47862f232036SJan Schmidt kfree(qsa); 47872f232036SJan Schmidt return ret; 47882f232036SJan Schmidt } 47892f232036SJan Schmidt 479057254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 479157254b6eSJan Schmidt { 47920b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 479457254b6eSJan Schmidt 479557254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 479657254b6eSJan Schmidt return -EPERM; 479757254b6eSJan Schmidt 47980b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 479957254b6eSJan Schmidt } 480057254b6eSJan Schmidt 4801abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 4802abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 48038ea05e3aSAlexander Block { 4804496ad9aaSAl Viro struct inode *inode = file_inode(file); 48050b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48068ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 48078ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 48088ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 4809c2050a45SDeepa Dinamani struct timespec ct = current_time(inode); 48108ea05e3aSAlexander Block int ret = 0; 4811dd5f9615SStefan Behrens int received_uuid_changed; 48128ea05e3aSAlexander Block 4813bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 4814bd60ea0fSDavid Sterba return -EPERM; 4815bd60ea0fSDavid Sterba 48168ea05e3aSAlexander Block ret = mnt_want_write_file(file); 48178ea05e3aSAlexander Block if (ret < 0) 48188ea05e3aSAlexander Block return ret; 48198ea05e3aSAlexander Block 48200b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48218ea05e3aSAlexander Block 48224a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 48238ea05e3aSAlexander Block ret = -EINVAL; 48248ea05e3aSAlexander Block goto out; 48258ea05e3aSAlexander Block } 48268ea05e3aSAlexander Block 48278ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 48288ea05e3aSAlexander Block ret = -EROFS; 48298ea05e3aSAlexander Block goto out; 48308ea05e3aSAlexander Block } 48318ea05e3aSAlexander Block 4832dd5f9615SStefan Behrens /* 4833dd5f9615SStefan Behrens * 1 - root item 4834dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 4835dd5f9615SStefan Behrens */ 4836dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 48378ea05e3aSAlexander Block if (IS_ERR(trans)) { 48388ea05e3aSAlexander Block ret = PTR_ERR(trans); 48398ea05e3aSAlexander Block trans = NULL; 48408ea05e3aSAlexander Block goto out; 48418ea05e3aSAlexander Block } 48428ea05e3aSAlexander Block 48438ea05e3aSAlexander Block sa->rtransid = trans->transid; 48448ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 48458ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 48468ea05e3aSAlexander Block 4847dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 4848dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 4849dd5f9615SStefan Behrens if (received_uuid_changed && 4850d87ff758SNikolay Borisov !btrfs_is_empty_uuid(root_item->received_uuid)) { 4851d87ff758SNikolay Borisov ret = btrfs_uuid_tree_rem(trans, fs_info, 4852d87ff758SNikolay Borisov root_item->received_uuid, 4853dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4854dd5f9615SStefan Behrens root->root_key.objectid); 4855d87ff758SNikolay Borisov if (ret && ret != -ENOENT) { 4856d87ff758SNikolay Borisov btrfs_abort_transaction(trans, ret); 4857d87ff758SNikolay Borisov btrfs_end_transaction(trans); 4858d87ff758SNikolay Borisov goto out; 4859d87ff758SNikolay Borisov } 4860d87ff758SNikolay Borisov } 48618ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 48628ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 48638ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 48643cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 48653cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 48663cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 48673cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 48688ea05e3aSAlexander Block 48690b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 48708ea05e3aSAlexander Block &root->root_key, &root->root_item); 48718ea05e3aSAlexander Block if (ret < 0) { 48723a45bb20SJeff Mahoney btrfs_end_transaction(trans); 48738ea05e3aSAlexander Block goto out; 4874dd5f9615SStefan Behrens } 4875dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 48760b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid, 4877dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4878dd5f9615SStefan Behrens root->root_key.objectid); 4879dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 488066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 4881efd38150SNikolay Borisov btrfs_end_transaction(trans); 4882dd5f9615SStefan Behrens goto out; 4883dd5f9615SStefan Behrens } 4884dd5f9615SStefan Behrens } 48853a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 4886abccd00fSHugo Mills out: 48870b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4888abccd00fSHugo Mills mnt_drop_write_file(file); 4889abccd00fSHugo Mills return ret; 4890abccd00fSHugo Mills } 4891abccd00fSHugo Mills 4892abccd00fSHugo Mills #ifdef CONFIG_64BIT 4893abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 4894abccd00fSHugo Mills void __user *arg) 4895abccd00fSHugo Mills { 4896abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 4897abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 4898abccd00fSHugo Mills int ret = 0; 4899abccd00fSHugo Mills 4900abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 49017b9ea627SShailendra Verma if (IS_ERR(args32)) 49027b9ea627SShailendra Verma return PTR_ERR(args32); 4903abccd00fSHugo Mills 49048d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 490584dbeb87SDan Carpenter if (!args64) { 490684dbeb87SDan Carpenter ret = -ENOMEM; 4907abccd00fSHugo Mills goto out; 4908abccd00fSHugo Mills } 4909abccd00fSHugo Mills 4910abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 4911abccd00fSHugo Mills args64->stransid = args32->stransid; 4912abccd00fSHugo Mills args64->rtransid = args32->rtransid; 4913abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 4914abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 4915abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 4916abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 4917abccd00fSHugo Mills args64->flags = args32->flags; 4918abccd00fSHugo Mills 4919abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 4920abccd00fSHugo Mills if (ret) 4921abccd00fSHugo Mills goto out; 4922abccd00fSHugo Mills 4923abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 4924abccd00fSHugo Mills args32->stransid = args64->stransid; 4925abccd00fSHugo Mills args32->rtransid = args64->rtransid; 4926abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 4927abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 4928abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 4929abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 4930abccd00fSHugo Mills args32->flags = args64->flags; 4931abccd00fSHugo Mills 4932abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 4933abccd00fSHugo Mills if (ret) 4934abccd00fSHugo Mills ret = -EFAULT; 4935abccd00fSHugo Mills 4936abccd00fSHugo Mills out: 4937abccd00fSHugo Mills kfree(args32); 4938abccd00fSHugo Mills kfree(args64); 4939abccd00fSHugo Mills return ret; 4940abccd00fSHugo Mills } 4941abccd00fSHugo Mills #endif 4942abccd00fSHugo Mills 4943abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 4944abccd00fSHugo Mills void __user *arg) 4945abccd00fSHugo Mills { 4946abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 4947abccd00fSHugo Mills int ret = 0; 4948abccd00fSHugo Mills 4949abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 49507b9ea627SShailendra Verma if (IS_ERR(sa)) 49517b9ea627SShailendra Verma return PTR_ERR(sa); 4952abccd00fSHugo Mills 4953abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 4954abccd00fSHugo Mills 4955abccd00fSHugo Mills if (ret) 4956abccd00fSHugo Mills goto out; 4957abccd00fSHugo Mills 49588ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 49598ea05e3aSAlexander Block if (ret) 49608ea05e3aSAlexander Block ret = -EFAULT; 49618ea05e3aSAlexander Block 49628ea05e3aSAlexander Block out: 49638ea05e3aSAlexander Block kfree(sa); 49648ea05e3aSAlexander Block return ret; 49658ea05e3aSAlexander Block } 49668ea05e3aSAlexander Block 4967867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 4968867ab667Sjeff.liu { 49690b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 4971a1b83ac5SAnand Jain size_t len; 4972867ab667Sjeff.liu int ret; 4973a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 4974a1b83ac5SAnand Jain 49750b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 49760b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 49770b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 4978a1b83ac5SAnand Jain 4979a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 4980867ab667Sjeff.liu 4981867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 49820b246afaSJeff Mahoney btrfs_warn(fs_info, 49830b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 49840b246afaSJeff Mahoney --len); 4985867ab667Sjeff.liu } 4986867ab667Sjeff.liu 4987867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 4988867ab667Sjeff.liu 4989867ab667Sjeff.liu return ret ? -EFAULT : 0; 4990867ab667Sjeff.liu } 4991867ab667Sjeff.liu 4992a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 4993a8bfd4abSjeff.liu { 49940b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49950b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49960b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49970b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 4998a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 4999a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5000a8bfd4abSjeff.liu int ret; 5001a8bfd4abSjeff.liu 5002a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5003a8bfd4abSjeff.liu return -EPERM; 5004a8bfd4abSjeff.liu 5005a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5006a8bfd4abSjeff.liu return -EFAULT; 5007a8bfd4abSjeff.liu 5008a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 50090b246afaSJeff Mahoney btrfs_err(fs_info, 50105d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5011a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5012a8bfd4abSjeff.liu return -EINVAL; 5013a8bfd4abSjeff.liu } 5014a8bfd4abSjeff.liu 5015a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5016a8bfd4abSjeff.liu if (ret) 5017a8bfd4abSjeff.liu return ret; 5018a8bfd4abSjeff.liu 5019a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5020a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5021a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5022a8bfd4abSjeff.liu goto out_unlock; 5023a8bfd4abSjeff.liu } 5024a8bfd4abSjeff.liu 50250b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5026a8bfd4abSjeff.liu strcpy(super_block->label, label); 50270b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 50283a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5029a8bfd4abSjeff.liu 5030a8bfd4abSjeff.liu out_unlock: 5031a8bfd4abSjeff.liu mnt_drop_write_file(file); 5032a8bfd4abSjeff.liu return ret; 5033a8bfd4abSjeff.liu } 5034a8bfd4abSjeff.liu 50352eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 50362eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 50372eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 50382eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 50392eaa055fSJeff Mahoney 5040d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 50412eaa055fSJeff Mahoney { 50424d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 50432eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 50442eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 50452eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 50462eaa055fSJeff Mahoney }; 50472eaa055fSJeff Mahoney 50482eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 50492eaa055fSJeff Mahoney return -EFAULT; 50502eaa055fSJeff Mahoney 50512eaa055fSJeff Mahoney return 0; 50522eaa055fSJeff Mahoney } 50532eaa055fSJeff Mahoney 50542eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 50552eaa055fSJeff Mahoney { 50560b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50570b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50580b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 50592eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 50602eaa055fSJeff Mahoney 50612eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 50622eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 50632eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 50642eaa055fSJeff Mahoney 50652eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 50662eaa055fSJeff Mahoney return -EFAULT; 50672eaa055fSJeff Mahoney 50682eaa055fSJeff Mahoney return 0; 50692eaa055fSJeff Mahoney } 50702eaa055fSJeff Mahoney 50712ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 50723b02a68aSJeff Mahoney enum btrfs_feature_set set, 50732eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 50742eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 50752eaa055fSJeff Mahoney { 50763b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 50773b02a68aSJeff Mahoney char *names; 50782eaa055fSJeff Mahoney u64 disallowed, unsupported; 50792eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 50802eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 50812eaa055fSJeff Mahoney 50822eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 50832eaa055fSJeff Mahoney if (unsupported) { 50843b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 50853b02a68aSJeff Mahoney if (names) { 50860b246afaSJeff Mahoney btrfs_warn(fs_info, 50873b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 50883b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 50893b02a68aSJeff Mahoney kfree(names); 50903b02a68aSJeff Mahoney } else 50910b246afaSJeff Mahoney btrfs_warn(fs_info, 50922eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 50932eaa055fSJeff Mahoney type, unsupported); 50942eaa055fSJeff Mahoney return -EOPNOTSUPP; 50952eaa055fSJeff Mahoney } 50962eaa055fSJeff Mahoney 50972eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 50982eaa055fSJeff Mahoney if (disallowed) { 50993b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 51003b02a68aSJeff Mahoney if (names) { 51010b246afaSJeff Mahoney btrfs_warn(fs_info, 51023b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 51033b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 51043b02a68aSJeff Mahoney kfree(names); 51053b02a68aSJeff Mahoney } else 51060b246afaSJeff Mahoney btrfs_warn(fs_info, 51072eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 51082eaa055fSJeff Mahoney type, disallowed); 51092eaa055fSJeff Mahoney return -EPERM; 51102eaa055fSJeff Mahoney } 51112eaa055fSJeff Mahoney 51122eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 51132eaa055fSJeff Mahoney if (disallowed) { 51143b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 51153b02a68aSJeff Mahoney if (names) { 51160b246afaSJeff Mahoney btrfs_warn(fs_info, 51173b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 51183b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 51193b02a68aSJeff Mahoney kfree(names); 51203b02a68aSJeff Mahoney } else 51210b246afaSJeff Mahoney btrfs_warn(fs_info, 51222eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 51232eaa055fSJeff Mahoney type, disallowed); 51242eaa055fSJeff Mahoney return -EPERM; 51252eaa055fSJeff Mahoney } 51262eaa055fSJeff Mahoney 51272eaa055fSJeff Mahoney return 0; 51282eaa055fSJeff Mahoney } 51292eaa055fSJeff Mahoney 51302ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 51312ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 51322eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 51332eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 51342eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 51352eaa055fSJeff Mahoney 51362eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 51372eaa055fSJeff Mahoney { 51380b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51400b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 51410b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 51422eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 51432eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 51442eaa055fSJeff Mahoney u64 newflags; 51452eaa055fSJeff Mahoney int ret; 51462eaa055fSJeff Mahoney 51472eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 51482eaa055fSJeff Mahoney return -EPERM; 51492eaa055fSJeff Mahoney 51502eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 51512eaa055fSJeff Mahoney return -EFAULT; 51522eaa055fSJeff Mahoney 51532eaa055fSJeff Mahoney /* Nothing to do */ 51542eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 51552eaa055fSJeff Mahoney !flags[0].incompat_flags) 51562eaa055fSJeff Mahoney return 0; 51572eaa055fSJeff Mahoney 51582ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 51592eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 51602eaa055fSJeff Mahoney if (ret) 51612eaa055fSJeff Mahoney return ret; 51622eaa055fSJeff Mahoney 51632ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 51642eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 51652eaa055fSJeff Mahoney if (ret) 51662eaa055fSJeff Mahoney return ret; 51672eaa055fSJeff Mahoney 51682ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 51692eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 51702eaa055fSJeff Mahoney if (ret) 51712eaa055fSJeff Mahoney return ret; 51722eaa055fSJeff Mahoney 51737ab19625SDavid Sterba ret = mnt_want_write_file(file); 51747ab19625SDavid Sterba if (ret) 51757ab19625SDavid Sterba return ret; 51767ab19625SDavid Sterba 51778051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 51787ab19625SDavid Sterba if (IS_ERR(trans)) { 51797ab19625SDavid Sterba ret = PTR_ERR(trans); 51807ab19625SDavid Sterba goto out_drop_write; 51817ab19625SDavid Sterba } 51822eaa055fSJeff Mahoney 51830b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 51842eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 51852eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 51862eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 51872eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 51882eaa055fSJeff Mahoney 51892eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 51902eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 51912eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 51922eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 51932eaa055fSJeff Mahoney 51942eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 51952eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 51962eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 51972eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 51980b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 51992eaa055fSJeff Mahoney 52003a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 52017ab19625SDavid Sterba out_drop_write: 52027ab19625SDavid Sterba mnt_drop_write_file(file); 52037ab19625SDavid Sterba 52047ab19625SDavid Sterba return ret; 52052eaa055fSJeff Mahoney } 52062eaa055fSJeff Mahoney 52072351f431SJosef Bacik static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) 52082351f431SJosef Bacik { 52092351f431SJosef Bacik struct btrfs_ioctl_send_args *arg; 52102351f431SJosef Bacik int ret; 52112351f431SJosef Bacik 52122351f431SJosef Bacik if (compat) { 52132351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 52142351f431SJosef Bacik struct btrfs_ioctl_send_args_32 args32; 52152351f431SJosef Bacik 52162351f431SJosef Bacik ret = copy_from_user(&args32, argp, sizeof(args32)); 52172351f431SJosef Bacik if (ret) 52182351f431SJosef Bacik return -EFAULT; 52192351f431SJosef Bacik arg = kzalloc(sizeof(*arg), GFP_KERNEL); 52202351f431SJosef Bacik if (!arg) 52212351f431SJosef Bacik return -ENOMEM; 52222351f431SJosef Bacik arg->send_fd = args32.send_fd; 52232351f431SJosef Bacik arg->clone_sources_count = args32.clone_sources_count; 52242351f431SJosef Bacik arg->clone_sources = compat_ptr(args32.clone_sources); 52252351f431SJosef Bacik arg->parent_root = args32.parent_root; 52262351f431SJosef Bacik arg->flags = args32.flags; 52272351f431SJosef Bacik memcpy(arg->reserved, args32.reserved, 52282351f431SJosef Bacik sizeof(args32.reserved)); 52292351f431SJosef Bacik #else 52302351f431SJosef Bacik return -ENOTTY; 52312351f431SJosef Bacik #endif 52322351f431SJosef Bacik } else { 52332351f431SJosef Bacik arg = memdup_user(argp, sizeof(*arg)); 52342351f431SJosef Bacik if (IS_ERR(arg)) 52352351f431SJosef Bacik return PTR_ERR(arg); 52362351f431SJosef Bacik } 52372351f431SJosef Bacik ret = btrfs_ioctl_send(file, arg); 52382351f431SJosef Bacik kfree(arg); 52392351f431SJosef Bacik return ret; 52402351f431SJosef Bacik } 52412351f431SJosef Bacik 5242f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5243f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5244f46b5a66SChristoph Hellwig { 52450b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52460b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52470b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52484bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5249f46b5a66SChristoph Hellwig 5250f46b5a66SChristoph Hellwig switch (cmd) { 52516cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 52526cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 52536cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 52546cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 52556cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 52566cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5257f7039b1dSLi Dongyang case FITRIM: 5258f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5259f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5260fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5261fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5262fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 52633de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5264fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 52656f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 52666f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 526776dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 526876dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 52690caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 52700caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 52710caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 52720caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 52736ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 52746ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5275f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 52761e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 52771e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 52781e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5279f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5280198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5281f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 52822ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5283f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5284da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 52856b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 52866b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5287475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 52882ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5289475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 52902ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5291f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 52929ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5293ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5294ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5295cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5296cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5297ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5298ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5299d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5300d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5301d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5302d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5303d24a67b2SZygo Blaxell case BTRFS_IOC_LOGICAL_INO_V2: 5304d24a67b2SZygo Blaxell return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 53051406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 53062ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 53079b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 53089b199859SFilipe David Borba Manana int ret; 53099b199859SFilipe David Borba Manana 531082b3e53bSNikolay Borisov ret = btrfs_start_delalloc_roots(fs_info, -1); 53119b199859SFilipe David Borba Manana if (ret) 53129b199859SFilipe David Borba Manana return ret; 53130b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 53142fad4e83SDavid Sterba /* 53152fad4e83SDavid Sterba * The transaction thread may want to do more work, 531601327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 53172fad4e83SDavid Sterba * processing uncleaned subvols. 53182fad4e83SDavid Sterba */ 53190b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 53209b199859SFilipe David Borba Manana return ret; 53219b199859SFilipe David Borba Manana } 532246204592SSage Weil case BTRFS_IOC_START_SYNC: 53239a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 532446204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 53252ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5326475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5327b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5328475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 53292ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5330475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 53312ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5332c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 53339ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5334837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 53352ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 533619a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 53372ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 53388ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 53398ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5340abccd00fSHugo Mills #ifdef CONFIG_64BIT 5341abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5342abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5343abccd00fSHugo Mills #endif 534431db9f7cSAlexander Block case BTRFS_IOC_SEND: 53452351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, false); 53462351f431SJosef Bacik #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 53472351f431SJosef Bacik case BTRFS_IOC_SEND_32: 53482351f431SJosef Bacik return _btrfs_ioctl_send(file, argp, true); 53492351f431SJosef Bacik #endif 5350c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 53512ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 53525d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5353905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 53545d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5355905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 53565d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5357905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 53585d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5359905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 53602f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 53612f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 53622f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 53632f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 536457254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 536557254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 53663f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 53672ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5368867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5369867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5370a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5371a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 53722eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5373d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 53742eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 53752eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 53762eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 53772eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5378f46b5a66SChristoph Hellwig } 5379f46b5a66SChristoph Hellwig 5380f46b5a66SChristoph Hellwig return -ENOTTY; 5381f46b5a66SChristoph Hellwig } 53824c63c245SLuke Dashjr 53834c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 53844c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 53854c63c245SLuke Dashjr { 53862a362249SJeff Mahoney /* 53872a362249SJeff Mahoney * These all access 32-bit values anyway so no further 53882a362249SJeff Mahoney * handling is necessary. 53892a362249SJeff Mahoney */ 53904c63c245SLuke Dashjr switch (cmd) { 53914c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 53924c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 53934c63c245SLuke Dashjr break; 53944c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 53954c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 53964c63c245SLuke Dashjr break; 53974c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 53984c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 53994c63c245SLuke Dashjr break; 54004c63c245SLuke Dashjr } 54014c63c245SLuke Dashjr 54024c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 54034c63c245SLuke Dashjr } 54044c63c245SLuke Dashjr #endif 5405