1f46b5a66SChristoph Hellwig /* 2f46b5a66SChristoph Hellwig * Copyright (C) 2007 Oracle. All rights reserved. 3f46b5a66SChristoph Hellwig * 4f46b5a66SChristoph Hellwig * This program is free software; you can redistribute it and/or 5f46b5a66SChristoph Hellwig * modify it under the terms of the GNU General Public 6f46b5a66SChristoph Hellwig * License v2 as published by the Free Software Foundation. 7f46b5a66SChristoph Hellwig * 8f46b5a66SChristoph Hellwig * This program is distributed in the hope that it will be useful, 9f46b5a66SChristoph Hellwig * but WITHOUT ANY WARRANTY; without even the implied warranty of 10f46b5a66SChristoph Hellwig * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11f46b5a66SChristoph Hellwig * General Public License for more details. 12f46b5a66SChristoph Hellwig * 13f46b5a66SChristoph Hellwig * You should have received a copy of the GNU General Public 14f46b5a66SChristoph Hellwig * License along with this program; if not, write to the 15f46b5a66SChristoph Hellwig * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 16f46b5a66SChristoph Hellwig * Boston, MA 021110-1307, USA. 17f46b5a66SChristoph Hellwig */ 18f46b5a66SChristoph Hellwig 19f46b5a66SChristoph Hellwig #include <linux/kernel.h> 20f46b5a66SChristoph Hellwig #include <linux/bio.h> 21f46b5a66SChristoph Hellwig #include <linux/buffer_head.h> 22f46b5a66SChristoph Hellwig #include <linux/file.h> 23f46b5a66SChristoph Hellwig #include <linux/fs.h> 24cb8e7090SChristoph Hellwig #include <linux/fsnotify.h> 25f46b5a66SChristoph Hellwig #include <linux/pagemap.h> 26f46b5a66SChristoph Hellwig #include <linux/highmem.h> 27f46b5a66SChristoph Hellwig #include <linux/time.h> 28f46b5a66SChristoph Hellwig #include <linux/init.h> 29f46b5a66SChristoph Hellwig #include <linux/string.h> 30f46b5a66SChristoph Hellwig #include <linux/backing-dev.h> 31cb8e7090SChristoph Hellwig #include <linux/mount.h> 32f46b5a66SChristoph Hellwig #include <linux/mpage.h> 33cb8e7090SChristoph Hellwig #include <linux/namei.h> 34f46b5a66SChristoph Hellwig #include <linux/swap.h> 35f46b5a66SChristoph Hellwig #include <linux/writeback.h> 36f46b5a66SChristoph Hellwig #include <linux/compat.h> 37f46b5a66SChristoph Hellwig #include <linux/bit_spinlock.h> 38cb8e7090SChristoph Hellwig #include <linux/security.h> 39f46b5a66SChristoph Hellwig #include <linux/xattr.h> 407ea394f1SYan Zheng #include <linux/vmalloc.h> 415a0e3ad6STejun Heo #include <linux/slab.h> 42f7039b1dSLi Dongyang #include <linux/blkdev.h> 438ea05e3aSAlexander Block #include <linux/uuid.h> 4455e301fdSFilipe Brandenburger #include <linux/btrfs.h> 45416161dbSMark Fasheh #include <linux/uaccess.h> 46f46b5a66SChristoph Hellwig #include "ctree.h" 47f46b5a66SChristoph Hellwig #include "disk-io.h" 48f46b5a66SChristoph Hellwig #include "transaction.h" 49f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 50f46b5a66SChristoph Hellwig #include "print-tree.h" 51f46b5a66SChristoph Hellwig #include "volumes.h" 52925baeddSChris Mason #include "locking.h" 53581bb050SLi Zefan #include "inode-map.h" 54d7728c96SJan Schmidt #include "backref.h" 55606686eeSJosef Bacik #include "rcu-string.h" 5631db9f7cSAlexander Block #include "send.h" 573f6bcfbdSStefan Behrens #include "dev-replace.h" 5863541927SFilipe David Borba Manana #include "props.h" 593b02a68aSJeff Mahoney #include "sysfs.h" 60fcebe456SJosef Bacik #include "qgroup.h" 611ec9a1aeSFilipe Manana #include "tree-log.h" 62ebb8765bSAnand Jain #include "compression.h" 63f46b5a66SChristoph Hellwig 64abccd00fSHugo Mills #ifdef CONFIG_64BIT 65abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 66abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 67abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 68abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 69abccd00fSHugo Mills */ 70abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 71abccd00fSHugo Mills __u64 sec; 72abccd00fSHugo Mills __u32 nsec; 73abccd00fSHugo Mills } __attribute__ ((__packed__)); 74abccd00fSHugo Mills 75abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 76abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 77abccd00fSHugo Mills __u64 stransid; /* in */ 78abccd00fSHugo Mills __u64 rtransid; /* out */ 79abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 80abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 81abccd00fSHugo Mills __u64 flags; /* in */ 82abccd00fSHugo Mills __u64 reserved[16]; /* in */ 83abccd00fSHugo Mills } __attribute__ ((__packed__)); 84abccd00fSHugo Mills 85abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 86abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 87abccd00fSHugo Mills #endif 88abccd00fSHugo Mills 89abccd00fSHugo Mills 90416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 911c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 921c919a5eSMark Fasheh int no_time_update); 93416161dbSMark Fasheh 946cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 956cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 966cbff00fSChristoph Hellwig { 976cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 986cbff00fSChristoph Hellwig return flags; 996cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 1006cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1016cbff00fSChristoph Hellwig else 1026cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1036cbff00fSChristoph Hellwig } 104f46b5a66SChristoph Hellwig 1056cbff00fSChristoph Hellwig /* 1066cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 1076cbff00fSChristoph Hellwig */ 1086cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 1096cbff00fSChristoph Hellwig { 1106cbff00fSChristoph Hellwig unsigned int iflags = 0; 1116cbff00fSChristoph Hellwig 1126cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1136cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1156cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1176cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1196cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1216cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1226cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1236cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 124d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 125d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 126d0092bddSLi Zefan 12713f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 128d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 12913f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13013f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1316cbff00fSChristoph Hellwig 1326cbff00fSChristoph Hellwig return iflags; 1336cbff00fSChristoph Hellwig } 1346cbff00fSChristoph Hellwig 1356cbff00fSChristoph Hellwig /* 1366cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1376cbff00fSChristoph Hellwig */ 1386cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1396cbff00fSChristoph Hellwig { 1406cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1413cc79392SFilipe Manana unsigned int new_fl = 0; 1426cbff00fSChristoph Hellwig 1436cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1443cc79392SFilipe Manana new_fl |= S_SYNC; 1456cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1463cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1476cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1483cc79392SFilipe Manana new_fl |= S_APPEND; 1496cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1503cc79392SFilipe Manana new_fl |= S_NOATIME; 1516cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1523cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1533cc79392SFilipe Manana 1543cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1553cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1563cc79392SFilipe Manana new_fl); 1576cbff00fSChristoph Hellwig } 1586cbff00fSChristoph Hellwig 1596cbff00fSChristoph Hellwig /* 1606cbff00fSChristoph Hellwig * Inherit flags from the parent inode. 1616cbff00fSChristoph Hellwig * 162e27425d6SJosef Bacik * Currently only the compression flags and the cow flags are inherited. 1636cbff00fSChristoph Hellwig */ 1646cbff00fSChristoph Hellwig void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 1656cbff00fSChristoph Hellwig { 1660b4dcea5SChris Mason unsigned int flags; 1670b4dcea5SChris Mason 1680b4dcea5SChris Mason if (!dir) 1690b4dcea5SChris Mason return; 1700b4dcea5SChris Mason 1710b4dcea5SChris Mason flags = BTRFS_I(dir)->flags; 1726cbff00fSChristoph Hellwig 173e27425d6SJosef Bacik if (flags & BTRFS_INODE_NOCOMPRESS) { 174e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 175e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 176e27425d6SJosef Bacik } else if (flags & BTRFS_INODE_COMPRESS) { 177e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 178e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 179e27425d6SJosef Bacik } 1806cbff00fSChristoph Hellwig 181213490b3SLiu Bo if (flags & BTRFS_INODE_NODATACOW) { 182e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 183213490b3SLiu Bo if (S_ISREG(inode->i_mode)) 184213490b3SLiu Bo BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; 185213490b3SLiu Bo } 186e27425d6SJosef Bacik 1876cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1886cbff00fSChristoph Hellwig } 1896cbff00fSChristoph Hellwig 1906cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1916cbff00fSChristoph Hellwig { 192496ad9aaSAl Viro struct btrfs_inode *ip = BTRFS_I(file_inode(file)); 1936cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1946cbff00fSChristoph Hellwig 1956cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1966cbff00fSChristoph Hellwig return -EFAULT; 1976cbff00fSChristoph Hellwig return 0; 1986cbff00fSChristoph Hellwig } 1996cbff00fSChristoph Hellwig 20075e7cb7fSLiu Bo static int check_flags(unsigned int flags) 20175e7cb7fSLiu Bo { 20275e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 20375e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 20475e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 205e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 206e1e8fb6aSLi Zefan FS_NOCOW_FL)) 20775e7cb7fSLiu Bo return -EOPNOTSUPP; 20875e7cb7fSLiu Bo 20975e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 21075e7cb7fSLiu Bo return -EINVAL; 21175e7cb7fSLiu Bo 21275e7cb7fSLiu Bo return 0; 21375e7cb7fSLiu Bo } 21475e7cb7fSLiu Bo 2156cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 2166cbff00fSChristoph Hellwig { 217496ad9aaSAl Viro struct inode *inode = file_inode(file); 2180b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2196cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 2206cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 2216cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 2226cbff00fSChristoph Hellwig unsigned int flags, oldflags; 2236cbff00fSChristoph Hellwig int ret; 224f062abf0SLi Zefan u64 ip_oldflags; 225f062abf0SLi Zefan unsigned int i_oldflags; 2267e97b8daSDavid Sterba umode_t mode; 2276cbff00fSChristoph Hellwig 228bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 229bd60ea0fSDavid Sterba return -EPERM; 230bd60ea0fSDavid Sterba 231b83cc969SLi Zefan if (btrfs_root_readonly(root)) 232b83cc969SLi Zefan return -EROFS; 233b83cc969SLi Zefan 2346cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 2356cbff00fSChristoph Hellwig return -EFAULT; 2366cbff00fSChristoph Hellwig 23775e7cb7fSLiu Bo ret = check_flags(flags); 23875e7cb7fSLiu Bo if (ret) 23975e7cb7fSLiu Bo return ret; 2406cbff00fSChristoph Hellwig 241e7848683SJan Kara ret = mnt_want_write_file(file); 242e7848683SJan Kara if (ret) 243e7848683SJan Kara return ret; 244e7848683SJan Kara 2455955102cSAl Viro inode_lock(inode); 2466cbff00fSChristoph Hellwig 247f062abf0SLi Zefan ip_oldflags = ip->flags; 248f062abf0SLi Zefan i_oldflags = inode->i_flags; 2497e97b8daSDavid Sterba mode = inode->i_mode; 250f062abf0SLi Zefan 2516cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 2526cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 2536cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2546cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2556cbff00fSChristoph Hellwig ret = -EPERM; 2566cbff00fSChristoph Hellwig goto out_unlock; 2576cbff00fSChristoph Hellwig } 2586cbff00fSChristoph Hellwig } 2596cbff00fSChristoph Hellwig 2606cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2616cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2626cbff00fSChristoph Hellwig else 2636cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2646cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2656cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2666cbff00fSChristoph Hellwig else 2676cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2686cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2696cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2706cbff00fSChristoph Hellwig else 2716cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2726cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2736cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2746cbff00fSChristoph Hellwig else 2756cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2766cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2776cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2786cbff00fSChristoph Hellwig else 2796cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2806cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2816cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2826cbff00fSChristoph Hellwig else 2836cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 2847e97b8daSDavid Sterba if (flags & FS_NOCOW_FL) { 2857e97b8daSDavid Sterba if (S_ISREG(mode)) { 2867e97b8daSDavid Sterba /* 2877e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2887e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2897e97b8daSDavid Sterba * status of the file and will not set it. 2907e97b8daSDavid Sterba */ 2917e97b8daSDavid Sterba if (inode->i_size == 0) 2927e97b8daSDavid Sterba ip->flags |= BTRFS_INODE_NODATACOW 2937e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM; 2947e97b8daSDavid Sterba } else { 295e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 2967e97b8daSDavid Sterba } 2977e97b8daSDavid Sterba } else { 2987e97b8daSDavid Sterba /* 29901327610SNicholas D Steeves * Revert back under same assumptions as above 3007e97b8daSDavid Sterba */ 3017e97b8daSDavid Sterba if (S_ISREG(mode)) { 3027e97b8daSDavid Sterba if (inode->i_size == 0) 3037e97b8daSDavid Sterba ip->flags &= ~(BTRFS_INODE_NODATACOW 3047e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM); 3057e97b8daSDavid Sterba } else { 306e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 3077e97b8daSDavid Sterba } 3087e97b8daSDavid Sterba } 3096cbff00fSChristoph Hellwig 31075e7cb7fSLiu Bo /* 31175e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 31275e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 31375e7cb7fSLiu Bo * things smaller. 31475e7cb7fSLiu Bo */ 31575e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 31675e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 31775e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 31863541927SFilipe David Borba Manana 31963541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 32063541927SFilipe David Borba Manana if (ret && ret != -ENODATA) 32163541927SFilipe David Borba Manana goto out_drop; 32275e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 32363541927SFilipe David Borba Manana const char *comp; 32463541927SFilipe David Borba Manana 32575e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 32675e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 32763541927SFilipe David Borba Manana 3280b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 32963541927SFilipe David Borba Manana comp = "lzo"; 33063541927SFilipe David Borba Manana else 33163541927SFilipe David Borba Manana comp = "zlib"; 33263541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", 33363541927SFilipe David Borba Manana comp, strlen(comp), 0); 33463541927SFilipe David Borba Manana if (ret) 33563541927SFilipe David Borba Manana goto out_drop; 33663541927SFilipe David Borba Manana 337ebcb904dSLi Zefan } else { 33878a017a2SFilipe Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 33978a017a2SFilipe Manana if (ret && ret != -ENODATA) 34078a017a2SFilipe Manana goto out_drop; 341ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 34275e7cb7fSLiu Bo } 3436cbff00fSChristoph Hellwig 3444da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 345f062abf0SLi Zefan if (IS_ERR(trans)) { 346f062abf0SLi Zefan ret = PTR_ERR(trans); 347f062abf0SLi Zefan goto out_drop; 348f062abf0SLi Zefan } 3496cbff00fSChristoph Hellwig 350306424ccSLi Zefan btrfs_update_iflags(inode); 3510c4d2d95SJosef Bacik inode_inc_iversion(inode); 352c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3536cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3546cbff00fSChristoph Hellwig 3556cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 356f062abf0SLi Zefan out_drop: 357f062abf0SLi Zefan if (ret) { 358f062abf0SLi Zefan ip->flags = ip_oldflags; 359f062abf0SLi Zefan inode->i_flags = i_oldflags; 360f062abf0SLi Zefan } 3616cbff00fSChristoph Hellwig 3626cbff00fSChristoph Hellwig out_unlock: 3635955102cSAl Viro inode_unlock(inode); 364e7848683SJan Kara mnt_drop_write_file(file); 3652d4e6f6aSliubo return ret; 3666cbff00fSChristoph Hellwig } 3676cbff00fSChristoph Hellwig 3686cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 3696cbff00fSChristoph Hellwig { 370496ad9aaSAl Viro struct inode *inode = file_inode(file); 3716cbff00fSChristoph Hellwig 3726cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 3736cbff00fSChristoph Hellwig } 374f46b5a66SChristoph Hellwig 375f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 376f7039b1dSLi Dongyang { 3770b246afaSJeff Mahoney struct inode *inode = file_inode(file); 3780b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 379f7039b1dSLi Dongyang struct btrfs_device *device; 380f7039b1dSLi Dongyang struct request_queue *q; 381f7039b1dSLi Dongyang struct fstrim_range range; 382f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 383f7039b1dSLi Dongyang u64 num_devices = 0; 384815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 385f7039b1dSLi Dongyang int ret; 386f7039b1dSLi Dongyang 387f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 388f7039b1dSLi Dongyang return -EPERM; 389f7039b1dSLi Dongyang 3901f78160cSXiao Guangrong rcu_read_lock(); 3911f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3921f78160cSXiao Guangrong dev_list) { 393f7039b1dSLi Dongyang if (!device->bdev) 394f7039b1dSLi Dongyang continue; 395f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 396f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 397f7039b1dSLi Dongyang num_devices++; 398f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 399f7039b1dSLi Dongyang minlen); 400f7039b1dSLi Dongyang } 401f7039b1dSLi Dongyang } 4021f78160cSXiao Guangrong rcu_read_unlock(); 403f4c697e6SLukas Czerner 404f7039b1dSLi Dongyang if (!num_devices) 405f7039b1dSLi Dongyang return -EOPNOTSUPP; 406f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 407f7039b1dSLi Dongyang return -EFAULT; 408e515c18bSLukas Czerner if (range.start > total_bytes || 409e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 410f4c697e6SLukas Czerner return -EINVAL; 411f7039b1dSLi Dongyang 412f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 413f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 414815745cfSAl Viro ret = btrfs_trim_fs(fs_info->tree_root, &range); 415f7039b1dSLi Dongyang if (ret < 0) 416f7039b1dSLi Dongyang return ret; 417f7039b1dSLi Dongyang 418f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 419f7039b1dSLi Dongyang return -EFAULT; 420f7039b1dSLi Dongyang 421f7039b1dSLi Dongyang return 0; 422f7039b1dSLi Dongyang } 423f7039b1dSLi Dongyang 424dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 425dd5f9615SStefan Behrens { 42646e0f66aSChris Mason int i; 42746e0f66aSChris Mason 42846e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 42946e0f66aSChris Mason if (uuid[i]) 43046e0f66aSChris Mason return 0; 43146e0f66aSChris Mason } 43246e0f66aSChris Mason return 1; 433dd5f9615SStefan Behrens } 434dd5f9615SStefan Behrens 435d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 436cb8e7090SChristoph Hellwig struct dentry *dentry, 43772fd032eSSage Weil char *name, int namelen, 4386f72c7e2SArne Jansen u64 *async_transid, 4398696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 440f46b5a66SChristoph Hellwig { 4410b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 442f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 443f46b5a66SChristoph Hellwig struct btrfs_key key; 44449a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 445f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 446f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 447d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 44876dda93cSYan, Zheng struct btrfs_root *new_root; 449d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 450c2050a45SDeepa Dinamani struct timespec cur_time = current_time(dir); 4515662344bSTsutomu Itoh struct inode *inode; 452f46b5a66SChristoph Hellwig int ret; 453f46b5a66SChristoph Hellwig int err; 454f46b5a66SChristoph Hellwig u64 objectid; 455f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 4563de4586cSChris Mason u64 index = 0; 457d5c12070SMiao Xie u64 qgroup_reserved; 4588ea05e3aSAlexander Block uuid_le new_uuid; 459f46b5a66SChristoph Hellwig 46049a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 46149a3c4d9SDavid Sterba if (!root_item) 46249a3c4d9SDavid Sterba return -ENOMEM; 46349a3c4d9SDavid Sterba 4640b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 4652fbe8c8aSAl Viro if (ret) 46649a3c4d9SDavid Sterba goto fail_free; 4676a912213SJosef Bacik 468e09fe2d2SQu Wenruo /* 469e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 47001327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 471e09fe2d2SQu Wenruo */ 47249a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 47349a3c4d9SDavid Sterba ret = -ENOSPC; 47449a3c4d9SDavid Sterba goto fail_free; 47549a3c4d9SDavid Sterba } 476e09fe2d2SQu Wenruo 477d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 4789ed74f2dSJosef Bacik /* 479d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 480d5c12070SMiao Xie * of create_snapshot(). 4819ed74f2dSJosef Bacik */ 482d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 483dd5f9615SStefan Behrens 8, &qgroup_reserved, false); 484d5c12070SMiao Xie if (ret) 48549a3c4d9SDavid Sterba goto fail_free; 486f46b5a66SChristoph Hellwig 487d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 488d5c12070SMiao Xie if (IS_ERR(trans)) { 489d5c12070SMiao Xie ret = PTR_ERR(trans); 490de6e8200SLiu Bo btrfs_subvolume_release_metadata(root, &block_rsv, 491de6e8200SLiu Bo qgroup_reserved); 49249a3c4d9SDavid Sterba goto fail_free; 493d5c12070SMiao Xie } 494d5c12070SMiao Xie trans->block_rsv = &block_rsv; 495d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 496f46b5a66SChristoph Hellwig 4970b246afaSJeff Mahoney ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); 4986f72c7e2SArne Jansen if (ret) 4996f72c7e2SArne Jansen goto fail; 5006f72c7e2SArne Jansen 5014d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 5028e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 5038e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 5048e8a1e31SJosef Bacik goto fail; 5058e8a1e31SJosef Bacik } 506f46b5a66SChristoph Hellwig 507b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 508f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 509f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 5105d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 511f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 512f46b5a66SChristoph Hellwig 5130b246afaSJeff Mahoney write_extent_buffer_fsid(leaf, fs_info->fsid); 5140b246afaSJeff Mahoney write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 515f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 516f46b5a66SChristoph Hellwig 51749a3c4d9SDavid Sterba inode_item = &root_item->inode; 5183cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 5193cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 5203cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 521da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 5220b246afaSJeff Mahoney fs_info->nodesize); 5233cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 524f46b5a66SChristoph Hellwig 52549a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 52649a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 5273cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 52808fe4db1SLi Zefan 52949a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 53049a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 53149a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 53249a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 53349a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 53449a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 535f46b5a66SChristoph Hellwig 53649a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 53749a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 5388ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 53949a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 54049a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 54149a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 54249a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 54349a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 54449a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 545f46b5a66SChristoph Hellwig 546925baeddSChris Mason btrfs_tree_unlock(leaf); 547f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 548f46b5a66SChristoph Hellwig leaf = NULL; 549f46b5a66SChristoph Hellwig 55049a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 551f46b5a66SChristoph Hellwig 552f46b5a66SChristoph Hellwig key.objectid = objectid; 5535d4f98a2SYan Zheng key.offset = 0; 554962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 5550b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 55649a3c4d9SDavid Sterba root_item); 557f46b5a66SChristoph Hellwig if (ret) 558f46b5a66SChristoph Hellwig goto fail; 559f46b5a66SChristoph Hellwig 56076dda93cSYan, Zheng key.offset = (u64)-1; 5610b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 56279787eaaSJeff Mahoney if (IS_ERR(new_root)) { 56379787eaaSJeff Mahoney ret = PTR_ERR(new_root); 56466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 56579787eaaSJeff Mahoney goto fail; 56679787eaaSJeff Mahoney } 56776dda93cSYan, Zheng 56876dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 56976dda93cSYan, Zheng 57063541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 571ce598979SMark Fasheh if (ret) { 572ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 57366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 574ce598979SMark Fasheh goto fail; 575ce598979SMark Fasheh } 576ce598979SMark Fasheh 577f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 578f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 579f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 580f32e48e9SChandan Rajendra 581f46b5a66SChristoph Hellwig /* 582f46b5a66SChristoph Hellwig * insert the directory item 583f46b5a66SChristoph Hellwig */ 5843de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 58579787eaaSJeff Mahoney if (ret) { 58666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 58779787eaaSJeff Mahoney goto fail; 58879787eaaSJeff Mahoney } 5893de4586cSChris Mason 5903de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 59116cdcec7SMiao Xie name, namelen, dir, &key, 5923de4586cSChris Mason BTRFS_FT_DIR, index); 59379787eaaSJeff Mahoney if (ret) { 59466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 595f46b5a66SChristoph Hellwig goto fail; 59679787eaaSJeff Mahoney } 5970660b5afSChris Mason 59852c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 59952c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 60052c26179SYan Zheng BUG_ON(ret); 60152c26179SYan Zheng 6020b246afaSJeff Mahoney ret = btrfs_add_root_ref(trans, fs_info, 6034df27c4dSYan, Zheng objectid, root->root_key.objectid, 60433345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 6050660b5afSChris Mason BUG_ON(ret); 6060660b5afSChris Mason 6070b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid, 6086bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 609dd5f9615SStefan Behrens if (ret) 61066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 611dd5f9615SStefan Behrens 612f46b5a66SChristoph Hellwig fail: 61349a3c4d9SDavid Sterba kfree(root_item); 614d5c12070SMiao Xie trans->block_rsv = NULL; 615d5c12070SMiao Xie trans->bytes_reserved = 0; 616de6e8200SLiu Bo btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 617de6e8200SLiu Bo 61872fd032eSSage Weil if (async_transid) { 61972fd032eSSage Weil *async_transid = trans->transid; 62072fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 62100d71c9cSMiao Xie if (err) 62200d71c9cSMiao Xie err = btrfs_commit_transaction(trans, root); 62372fd032eSSage Weil } else { 62476dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 62572fd032eSSage Weil } 626f46b5a66SChristoph Hellwig if (err && !ret) 627f46b5a66SChristoph Hellwig ret = err; 6281a65e24bSChris Mason 6295662344bSTsutomu Itoh if (!ret) { 6305662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 631de6e8200SLiu Bo if (IS_ERR(inode)) 632de6e8200SLiu Bo return PTR_ERR(inode); 6335662344bSTsutomu Itoh d_instantiate(dentry, inode); 6345662344bSTsutomu Itoh } 635f46b5a66SChristoph Hellwig return ret; 63649a3c4d9SDavid Sterba 63749a3c4d9SDavid Sterba fail_free: 63849a3c4d9SDavid Sterba kfree(root_item); 63949a3c4d9SDavid Sterba return ret; 640f46b5a66SChristoph Hellwig } 641f46b5a66SChristoph Hellwig 6429ea24bbeSFilipe Manana static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) 6438257b2dcSMiao Xie { 6448257b2dcSMiao Xie s64 writers; 6458257b2dcSMiao Xie DEFINE_WAIT(wait); 6468257b2dcSMiao Xie 6478257b2dcSMiao Xie do { 6488257b2dcSMiao Xie prepare_to_wait(&root->subv_writers->wait, &wait, 6498257b2dcSMiao Xie TASK_UNINTERRUPTIBLE); 6508257b2dcSMiao Xie 6518257b2dcSMiao Xie writers = percpu_counter_sum(&root->subv_writers->counter); 6528257b2dcSMiao Xie if (writers) 6538257b2dcSMiao Xie schedule(); 6548257b2dcSMiao Xie 6558257b2dcSMiao Xie finish_wait(&root->subv_writers->wait, &wait); 6568257b2dcSMiao Xie } while (writers); 6578257b2dcSMiao Xie } 6588257b2dcSMiao Xie 659e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 660e9662f70SMiao Xie struct dentry *dentry, char *name, int namelen, 661e9662f70SMiao Xie u64 *async_transid, bool readonly, 662e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 663f46b5a66SChristoph Hellwig { 6640b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 6652e4bfab9SYan, Zheng struct inode *inode; 666f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 667f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 6682e4bfab9SYan, Zheng int ret; 669f46b5a66SChristoph Hellwig 67027cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 671f46b5a66SChristoph Hellwig return -EINVAL; 672f46b5a66SChristoph Hellwig 673a1ee7362SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 674a1ee7362SDavid Sterba if (!pending_snapshot) 675a1ee7362SDavid Sterba return -ENOMEM; 676a1ee7362SDavid Sterba 677b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 678b0c0ea63SDavid Sterba GFP_NOFS); 6798546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 6808546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 681b0c0ea63SDavid Sterba ret = -ENOMEM; 682b0c0ea63SDavid Sterba goto free_pending; 683b0c0ea63SDavid Sterba } 684b0c0ea63SDavid Sterba 6858257b2dcSMiao Xie atomic_inc(&root->will_be_snapshoted); 6864e857c58SPeter Zijlstra smp_mb__after_atomic(); 6879ea24bbeSFilipe Manana btrfs_wait_for_no_snapshoting_writes(root); 6888257b2dcSMiao Xie 6896a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 6906a03843dSMiao Xie if (ret) 691a1ee7362SDavid Sterba goto dec_and_free; 6926a03843dSMiao Xie 693578def7cSFilipe Manana btrfs_wait_ordered_extents(root, -1, 0, (u64)-1); 6946a03843dSMiao Xie 69566d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 69666d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 697d5c12070SMiao Xie /* 698d5c12070SMiao Xie * 1 - parent dir inode 699d5c12070SMiao Xie * 2 - dir entries 700d5c12070SMiao Xie * 1 - root item 701d5c12070SMiao Xie * 2 - root ref/backref 702d5c12070SMiao Xie * 1 - root of snapshot 703dd5f9615SStefan Behrens * 1 - UUID item 704d5c12070SMiao Xie */ 705d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 706dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 707ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 708ee3441b4SJeff Mahoney false); 709d5c12070SMiao Xie if (ret) 710a1ee7362SDavid Sterba goto dec_and_free; 711d5c12070SMiao Xie 712a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 713a22285a6SYan, Zheng pending_snapshot->root = root; 714b83cc969SLi Zefan pending_snapshot->readonly = readonly; 715e9662f70SMiao Xie pending_snapshot->dir = dir; 7168696c533SMiao Xie pending_snapshot->inherit = inherit; 717a22285a6SYan, Zheng 718d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 719a22285a6SYan, Zheng if (IS_ERR(trans)) { 720a22285a6SYan, Zheng ret = PTR_ERR(trans); 721a22285a6SYan, Zheng goto fail; 722a22285a6SYan, Zheng } 723a22285a6SYan, Zheng 7240b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 725a22285a6SYan, Zheng list_add(&pending_snapshot->list, 726a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 7270b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 72872fd032eSSage Weil if (async_transid) { 72972fd032eSSage Weil *async_transid = trans->transid; 73072fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 7310b246afaSJeff Mahoney fs_info->extent_root, 1); 73200d71c9cSMiao Xie if (ret) 73300d71c9cSMiao Xie ret = btrfs_commit_transaction(trans, root); 73472fd032eSSage Weil } else { 73572fd032eSSage Weil ret = btrfs_commit_transaction(trans, 7360b246afaSJeff Mahoney fs_info->extent_root); 73772fd032eSSage Weil } 738aec8030aSMiao Xie if (ret) 739c37b2b62SJosef Bacik goto fail; 740a22285a6SYan, Zheng 741a22285a6SYan, Zheng ret = pending_snapshot->error; 742f46b5a66SChristoph Hellwig if (ret) 7432e4bfab9SYan, Zheng goto fail; 744f46b5a66SChristoph Hellwig 745d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 746d3797308SChris Mason if (ret) 747d3797308SChris Mason goto fail; 748d3797308SChris Mason 7492b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7502e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7512e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7522e4bfab9SYan, Zheng goto fail; 7532e4bfab9SYan, Zheng } 7545662344bSTsutomu Itoh 7552e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7562e4bfab9SYan, Zheng ret = 0; 7572e4bfab9SYan, Zheng fail: 758d5c12070SMiao Xie btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, 759d5c12070SMiao Xie &pending_snapshot->block_rsv, 760d5c12070SMiao Xie pending_snapshot->qgroup_reserved); 761a1ee7362SDavid Sterba dec_and_free: 7629ea24bbeSFilipe Manana if (atomic_dec_and_test(&root->will_be_snapshoted)) 7639ea24bbeSFilipe Manana wake_up_atomic_t(&root->will_be_snapshoted); 764b0c0ea63SDavid Sterba free_pending: 765b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 7668546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 767a1ee7362SDavid Sterba kfree(pending_snapshot); 768a1ee7362SDavid Sterba 769f46b5a66SChristoph Hellwig return ret; 770f46b5a66SChristoph Hellwig } 771f46b5a66SChristoph Hellwig 7724260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7734260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7744260f7c7SSage Weil * whether the type of victim is right. 7754260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7764260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7774260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7784260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7794260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7804260f7c7SSage Weil * a. be owner of dir, or 7814260f7c7SSage Weil * b. be owner of victim, or 7824260f7c7SSage Weil * c. have CAP_FOWNER capability 78301327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 7844260f7c7SSage Weil * links pointing to it. 7854260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7864260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7874260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7884260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 7894260f7c7SSage Weil * nfs_async_unlink(). 7904260f7c7SSage Weil */ 7914260f7c7SSage Weil 7924260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 7934260f7c7SSage Weil { 7944260f7c7SSage Weil int error; 7954260f7c7SSage Weil 7962b0143b5SDavid Howells if (d_really_is_negative(victim)) 7974260f7c7SSage Weil return -ENOENT; 7984260f7c7SSage Weil 7992b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8004fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8014260f7c7SSage Weil 8024260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 8034260f7c7SSage Weil if (error) 8044260f7c7SSage Weil return error; 8054260f7c7SSage Weil if (IS_APPEND(dir)) 8064260f7c7SSage Weil return -EPERM; 8072b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 8082b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 8094260f7c7SSage Weil return -EPERM; 8104260f7c7SSage Weil if (isdir) { 811e36cb0b8SDavid Howells if (!d_is_dir(victim)) 8124260f7c7SSage Weil return -ENOTDIR; 8134260f7c7SSage Weil if (IS_ROOT(victim)) 8144260f7c7SSage Weil return -EBUSY; 815e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 8164260f7c7SSage Weil return -EISDIR; 8174260f7c7SSage Weil if (IS_DEADDIR(dir)) 8184260f7c7SSage Weil return -ENOENT; 8194260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 8204260f7c7SSage Weil return -EBUSY; 8214260f7c7SSage Weil return 0; 8224260f7c7SSage Weil } 8234260f7c7SSage Weil 824cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 825cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 826cb8e7090SChristoph Hellwig { 8272b0143b5SDavid Howells if (d_really_is_positive(child)) 828cb8e7090SChristoph Hellwig return -EEXIST; 829cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 830cb8e7090SChristoph Hellwig return -ENOENT; 831cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 832cb8e7090SChristoph Hellwig } 833cb8e7090SChristoph Hellwig 834cb8e7090SChristoph Hellwig /* 835cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 836cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 837cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 838cb8e7090SChristoph Hellwig */ 83976dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 84076dda93cSYan, Zheng char *name, int namelen, 84172fd032eSSage Weil struct btrfs_root *snap_src, 8426f72c7e2SArne Jansen u64 *async_transid, bool readonly, 8438696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 844cb8e7090SChristoph Hellwig { 8452b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 8460b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 847cb8e7090SChristoph Hellwig struct dentry *dentry; 848cb8e7090SChristoph Hellwig int error; 849cb8e7090SChristoph Hellwig 85000235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 85100235411SAl Viro if (error == -EINTR) 85200235411SAl Viro return error; 853cb8e7090SChristoph Hellwig 854cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 855cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 856cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 857cb8e7090SChristoph Hellwig goto out_unlock; 858cb8e7090SChristoph Hellwig 85976dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 860cb8e7090SChristoph Hellwig if (error) 861a874a63eSLiu Bo goto out_dput; 862cb8e7090SChristoph Hellwig 8639c52057cSChris Mason /* 8649c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8659c52057cSChris Mason * check for them now when we can safely fail 8669c52057cSChris Mason */ 8679c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8689c52057cSChris Mason dir->i_ino, name, 8699c52057cSChris Mason namelen); 8709c52057cSChris Mason if (error) 8719c52057cSChris Mason goto out_dput; 8729c52057cSChris Mason 8730b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 87476dda93cSYan, Zheng 87576dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 87676dda93cSYan, Zheng goto out_up_read; 87776dda93cSYan, Zheng 8783de4586cSChris Mason if (snap_src) { 879e9662f70SMiao Xie error = create_snapshot(snap_src, dir, dentry, name, namelen, 8806f72c7e2SArne Jansen async_transid, readonly, inherit); 8813de4586cSChris Mason } else { 882d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 883d5c12070SMiao Xie async_transid, inherit); 8843de4586cSChris Mason } 88576dda93cSYan, Zheng if (!error) 88676dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 88776dda93cSYan, Zheng out_up_read: 8880b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 889cb8e7090SChristoph Hellwig out_dput: 890cb8e7090SChristoph Hellwig dput(dentry); 891cb8e7090SChristoph Hellwig out_unlock: 8925955102cSAl Viro inode_unlock(dir); 893cb8e7090SChristoph Hellwig return error; 894cb8e7090SChristoph Hellwig } 895cb8e7090SChristoph Hellwig 8964cb5300bSChris Mason /* 8974cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 8984cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 8994cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 9004cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 9014cb5300bSChris Mason * part of the file 9024cb5300bSChris Mason */ 903aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 9044cb5300bSChris Mason { 9054cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9064cb5300bSChris Mason struct extent_map *em = NULL; 9074cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 9084cb5300bSChris Mason u64 end; 9094cb5300bSChris Mason 9104cb5300bSChris Mason read_lock(&em_tree->lock); 91109cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 9124cb5300bSChris Mason read_unlock(&em_tree->lock); 9134cb5300bSChris Mason 9144cb5300bSChris Mason if (em) { 9154cb5300bSChris Mason end = extent_map_end(em); 9164cb5300bSChris Mason free_extent_map(em); 9174cb5300bSChris Mason if (end - offset > thresh) 9184cb5300bSChris Mason return 0; 9194cb5300bSChris Mason } 9204cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 9214cb5300bSChris Mason thresh /= 2; 9224cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 9234cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 9244cb5300bSChris Mason if (end >= thresh) 9254cb5300bSChris Mason return 0; 9264cb5300bSChris Mason return 1; 9274cb5300bSChris Mason } 9284cb5300bSChris Mason 9294cb5300bSChris Mason /* 9304cb5300bSChris Mason * helper function to walk through a file and find extents 9314cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 9324cb5300bSChris Mason * 9334cb5300bSChris Mason * This is used by the defragging code to find new and small 9344cb5300bSChris Mason * extents 9354cb5300bSChris Mason */ 9364cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 9374cb5300bSChris Mason struct inode *inode, u64 newer_than, 938aab110abSDavid Sterba u64 *off, u32 thresh) 9394cb5300bSChris Mason { 9404cb5300bSChris Mason struct btrfs_path *path; 9414cb5300bSChris Mason struct btrfs_key min_key; 9424cb5300bSChris Mason struct extent_buffer *leaf; 9434cb5300bSChris Mason struct btrfs_file_extent_item *extent; 9444cb5300bSChris Mason int type; 9454cb5300bSChris Mason int ret; 946a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 9474cb5300bSChris Mason 9484cb5300bSChris Mason path = btrfs_alloc_path(); 9494cb5300bSChris Mason if (!path) 9504cb5300bSChris Mason return -ENOMEM; 9514cb5300bSChris Mason 952a4689d2bSDavid Sterba min_key.objectid = ino; 9534cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9544cb5300bSChris Mason min_key.offset = *off; 9554cb5300bSChris Mason 9564cb5300bSChris Mason while (1) { 9576174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9584cb5300bSChris Mason if (ret != 0) 9594cb5300bSChris Mason goto none; 960f094c9bdSFilipe Manana process_slot: 961a4689d2bSDavid Sterba if (min_key.objectid != ino) 9624cb5300bSChris Mason goto none; 9634cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9644cb5300bSChris Mason goto none; 9654cb5300bSChris Mason 9664cb5300bSChris Mason leaf = path->nodes[0]; 9674cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9684cb5300bSChris Mason struct btrfs_file_extent_item); 9694cb5300bSChris Mason 9704cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9714cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9724cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9734cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9744cb5300bSChris Mason *off = min_key.offset; 9754cb5300bSChris Mason btrfs_free_path(path); 9764cb5300bSChris Mason return 0; 9774cb5300bSChris Mason } 9784cb5300bSChris Mason 979f094c9bdSFilipe Manana path->slots[0]++; 980f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 981f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 982f094c9bdSFilipe Manana goto process_slot; 983f094c9bdSFilipe Manana } 984f094c9bdSFilipe Manana 9854cb5300bSChris Mason if (min_key.offset == (u64)-1) 9864cb5300bSChris Mason goto none; 9874cb5300bSChris Mason 9884cb5300bSChris Mason min_key.offset++; 9894cb5300bSChris Mason btrfs_release_path(path); 9904cb5300bSChris Mason } 9914cb5300bSChris Mason none: 9924cb5300bSChris Mason btrfs_free_path(path); 9934cb5300bSChris Mason return -ENOENT; 9944cb5300bSChris Mason } 9954cb5300bSChris Mason 9966c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 99717ce6ef8SLiu Bo { 99817ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 999940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10006c282eb4SLi Zefan struct extent_map *em; 100109cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1002940100a4SChris Mason 1003940100a4SChris Mason /* 1004940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1005940100a4SChris Mason * the full extent lock 1006940100a4SChris Mason */ 1007940100a4SChris Mason read_lock(&em_tree->lock); 1008940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1009940100a4SChris Mason read_unlock(&em_tree->lock); 1010940100a4SChris Mason 1011940100a4SChris Mason if (!em) { 1012308d9800SFilipe Manana struct extent_state *cached = NULL; 1013308d9800SFilipe Manana u64 end = start + len - 1; 1014308d9800SFilipe Manana 1015940100a4SChris Mason /* get the big lock and read metadata off disk */ 1016ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1017940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 1018308d9800SFilipe Manana unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 1019940100a4SChris Mason 10206cf8bfbfSDan Carpenter if (IS_ERR(em)) 10216c282eb4SLi Zefan return NULL; 1022940100a4SChris Mason } 1023940100a4SChris Mason 10246c282eb4SLi Zefan return em; 10256c282eb4SLi Zefan } 10266c282eb4SLi Zefan 10276c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 10286c282eb4SLi Zefan { 10296c282eb4SLi Zefan struct extent_map *next; 10306c282eb4SLi Zefan bool ret = true; 10316c282eb4SLi Zefan 10326c282eb4SLi Zefan /* this is the last extent */ 10336c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 10346c282eb4SLi Zefan return false; 10356c282eb4SLi Zefan 10366c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1037e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1038e9512d72SChris Mason ret = false; 1039e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1040ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 10416c282eb4SLi Zefan ret = false; 10426c282eb4SLi Zefan 10436c282eb4SLi Zefan free_extent_map(next); 10446c282eb4SLi Zefan return ret; 10456c282eb4SLi Zefan } 10466c282eb4SLi Zefan 1047aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1048a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1049a43a2111SAndrew Mahone int compress) 10506c282eb4SLi Zefan { 10516c282eb4SLi Zefan struct extent_map *em; 10526c282eb4SLi Zefan int ret = 1; 10536c282eb4SLi Zefan bool next_mergeable = true; 10544a3560c4SLiu Bo bool prev_mergeable = true; 10556c282eb4SLi Zefan 10566c282eb4SLi Zefan /* 10576c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10586c282eb4SLi Zefan * defragging it 10596c282eb4SLi Zefan */ 10606c282eb4SLi Zefan if (start < *defrag_end) 10616c282eb4SLi Zefan return 1; 10626c282eb4SLi Zefan 10636c282eb4SLi Zefan *skip = 0; 10646c282eb4SLi Zefan 10656c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10666c282eb4SLi Zefan if (!em) 10676c282eb4SLi Zefan return 0; 10686c282eb4SLi Zefan 1069940100a4SChris Mason /* this will cover holes, and inline extents */ 107017ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1071940100a4SChris Mason ret = 0; 107217ce6ef8SLiu Bo goto out; 107317ce6ef8SLiu Bo } 107417ce6ef8SLiu Bo 10754a3560c4SLiu Bo if (!*defrag_end) 10764a3560c4SLiu Bo prev_mergeable = false; 10774a3560c4SLiu Bo 10786c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1079940100a4SChris Mason /* 10806c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10816c282eb4SLi Zefan * real extent, don't bother defragging it 1082940100a4SChris Mason */ 1083a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10844a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1085940100a4SChris Mason ret = 0; 108617ce6ef8SLiu Bo out: 1087940100a4SChris Mason /* 1088940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1089940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1090940100a4SChris Mason * so that the next tiny extent will force a defrag. 1091940100a4SChris Mason * 1092940100a4SChris Mason * The end result of this is that tiny extents before a single big 1093940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1094940100a4SChris Mason */ 1095940100a4SChris Mason if (ret) { 1096940100a4SChris Mason *defrag_end = extent_map_end(em); 1097940100a4SChris Mason } else { 1098940100a4SChris Mason *last_len = 0; 1099940100a4SChris Mason *skip = extent_map_end(em); 1100940100a4SChris Mason *defrag_end = 0; 1101940100a4SChris Mason } 1102940100a4SChris Mason 1103940100a4SChris Mason free_extent_map(em); 1104940100a4SChris Mason return ret; 1105940100a4SChris Mason } 1106940100a4SChris Mason 11074cb5300bSChris Mason /* 11084cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 11094cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 11104cb5300bSChris Mason * to COW and defrag. 11114cb5300bSChris Mason * 11124cb5300bSChris Mason * It also makes sure the delalloc code has enough 11134cb5300bSChris Mason * dirty data to avoid making new small extents as part 11144cb5300bSChris Mason * of the defrag 11154cb5300bSChris Mason * 11164cb5300bSChris Mason * It's a good idea to start RA on this range 11174cb5300bSChris Mason * before calling this. 11184cb5300bSChris Mason */ 11194cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 11204cb5300bSChris Mason struct page **pages, 11214cb5300bSChris Mason unsigned long start_index, 1122c41570c9SJustin Maggard unsigned long num_pages) 1123f46b5a66SChristoph Hellwig { 11244cb5300bSChris Mason unsigned long file_end; 11254cb5300bSChris Mason u64 isize = i_size_read(inode); 1126f46b5a66SChristoph Hellwig u64 page_start; 1127f46b5a66SChristoph Hellwig u64 page_end; 11281f12bd06SLiu Bo u64 page_cnt; 11294cb5300bSChris Mason int ret; 11304cb5300bSChris Mason int i; 11314cb5300bSChris Mason int i_done; 11324cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 11334cb5300bSChris Mason struct extent_state *cached_state = NULL; 1134600a45e1SMiao Xie struct extent_io_tree *tree; 11353b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 11364cb5300bSChris Mason 113709cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 11381f12bd06SLiu Bo if (!isize || start_index > file_end) 11391f12bd06SLiu Bo return 0; 11401f12bd06SLiu Bo 11411f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 11424cb5300bSChris Mason 11434cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 114409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 114509cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 11464cb5300bSChris Mason if (ret) 11474cb5300bSChris Mason return ret; 11484cb5300bSChris Mason i_done = 0; 1149600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11504cb5300bSChris Mason 11514cb5300bSChris Mason /* step one, lock all the pages */ 11521f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11534cb5300bSChris Mason struct page *page; 1154600a45e1SMiao Xie again: 1155a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11563b16a4e3SJosef Bacik start_index + i, mask); 11574cb5300bSChris Mason if (!page) 11584cb5300bSChris Mason break; 11594cb5300bSChris Mason 1160600a45e1SMiao Xie page_start = page_offset(page); 116109cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1162600a45e1SMiao Xie while (1) { 1163308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1164ff13db41SDavid Sterba &cached_state); 1165600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1166600a45e1SMiao Xie page_start); 1167308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1168308d9800SFilipe Manana &cached_state, GFP_NOFS); 1169600a45e1SMiao Xie if (!ordered) 1170600a45e1SMiao Xie break; 1171600a45e1SMiao Xie 1172600a45e1SMiao Xie unlock_page(page); 1173600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1174600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1175600a45e1SMiao Xie lock_page(page); 11761f12bd06SLiu Bo /* 11771f12bd06SLiu Bo * we unlocked the page above, so we need check if 11781f12bd06SLiu Bo * it was released or not. 11791f12bd06SLiu Bo */ 11801f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11811f12bd06SLiu Bo unlock_page(page); 118209cbfeafSKirill A. Shutemov put_page(page); 11831f12bd06SLiu Bo goto again; 11841f12bd06SLiu Bo } 1185600a45e1SMiao Xie } 1186600a45e1SMiao Xie 11874cb5300bSChris Mason if (!PageUptodate(page)) { 11884cb5300bSChris Mason btrfs_readpage(NULL, page); 11894cb5300bSChris Mason lock_page(page); 11904cb5300bSChris Mason if (!PageUptodate(page)) { 11914cb5300bSChris Mason unlock_page(page); 119209cbfeafSKirill A. Shutemov put_page(page); 11934cb5300bSChris Mason ret = -EIO; 11944cb5300bSChris Mason break; 11954cb5300bSChris Mason } 11964cb5300bSChris Mason } 1197600a45e1SMiao Xie 1198600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1199600a45e1SMiao Xie unlock_page(page); 120009cbfeafSKirill A. Shutemov put_page(page); 1201600a45e1SMiao Xie goto again; 1202600a45e1SMiao Xie } 1203600a45e1SMiao Xie 12044cb5300bSChris Mason pages[i] = page; 12054cb5300bSChris Mason i_done++; 12064cb5300bSChris Mason } 12074cb5300bSChris Mason if (!i_done || ret) 12084cb5300bSChris Mason goto out; 12094cb5300bSChris Mason 12104cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 12114cb5300bSChris Mason goto out; 12124cb5300bSChris Mason 12134cb5300bSChris Mason /* 12144cb5300bSChris Mason * so now we have a nice long stream of locked 12154cb5300bSChris Mason * and up to date pages, lets wait on them 12164cb5300bSChris Mason */ 12174cb5300bSChris Mason for (i = 0; i < i_done; i++) 12184cb5300bSChris Mason wait_on_page_writeback(pages[i]); 12194cb5300bSChris Mason 12204cb5300bSChris Mason page_start = page_offset(pages[0]); 122109cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 12224cb5300bSChris Mason 12234cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1224ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 12254cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 12264cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 12279e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 12289e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12294cb5300bSChris Mason 12301f12bd06SLiu Bo if (i_done != page_cnt) { 12319e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 12329e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 12339e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 12344cb5300bSChris Mason btrfs_delalloc_release_space(inode, 123509cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 123609cbfeafSKirill A. Shutemov (page_cnt - i_done) << PAGE_SHIFT); 12374cb5300bSChris Mason } 12384cb5300bSChris Mason 12394cb5300bSChris Mason 12409e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1241018ed4f7SDavid Sterba &cached_state); 12424cb5300bSChris Mason 12434cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 12444cb5300bSChris Mason page_start, page_end - 1, &cached_state, 12454cb5300bSChris Mason GFP_NOFS); 12464cb5300bSChris Mason 12474cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12484cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12494cb5300bSChris Mason ClearPageChecked(pages[i]); 12504cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12514cb5300bSChris Mason set_page_dirty(pages[i]); 12524cb5300bSChris Mason unlock_page(pages[i]); 125309cbfeafSKirill A. Shutemov put_page(pages[i]); 12544cb5300bSChris Mason } 12554cb5300bSChris Mason return i_done; 12564cb5300bSChris Mason out: 12574cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12584cb5300bSChris Mason unlock_page(pages[i]); 125909cbfeafSKirill A. Shutemov put_page(pages[i]); 12604cb5300bSChris Mason } 12617cf5b976SQu Wenruo btrfs_delalloc_release_space(inode, 126209cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 126309cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12644cb5300bSChris Mason return ret; 12654cb5300bSChris Mason 12664cb5300bSChris Mason } 12674cb5300bSChris Mason 12684cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12694cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12704cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12714cb5300bSChris Mason { 12720b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 12734cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12744cb5300bSChris Mason struct file_ra_state *ra = NULL; 12754cb5300bSChris Mason unsigned long last_index; 1276151a31b2SLi Zefan u64 isize = i_size_read(inode); 1277940100a4SChris Mason u64 last_len = 0; 1278940100a4SChris Mason u64 skip = 0; 1279940100a4SChris Mason u64 defrag_end = 0; 12804cb5300bSChris Mason u64 newer_off = range->start; 1281f46b5a66SChristoph Hellwig unsigned long i; 1282008873eaSLi Zefan unsigned long ra_index = 0; 1283f46b5a66SChristoph Hellwig int ret; 12844cb5300bSChris Mason int defrag_count = 0; 12851a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1286aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 128709cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1288c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1289ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 12904cb5300bSChris Mason struct page **pages = NULL; 12914cb5300bSChris Mason 12920abd5b17SLiu Bo if (isize == 0) 12930abd5b17SLiu Bo return 0; 12940abd5b17SLiu Bo 12950abd5b17SLiu Bo if (range->start >= isize) 12960abd5b17SLiu Bo return -EINVAL; 12971a419d85SLi Zefan 12981a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 12991a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 13001a419d85SLi Zefan return -EINVAL; 13011a419d85SLi Zefan if (range->compress_type) 13021a419d85SLi Zefan compress_type = range->compress_type; 13031a419d85SLi Zefan } 1304f46b5a66SChristoph Hellwig 13050abd5b17SLiu Bo if (extent_thresh == 0) 1306ee22184bSByongho Lee extent_thresh = SZ_256K; 1307f46b5a66SChristoph Hellwig 13084cb5300bSChris Mason /* 13094cb5300bSChris Mason * if we were not given a file, allocate a readahead 13104cb5300bSChris Mason * context 13114cb5300bSChris Mason */ 13124cb5300bSChris Mason if (!file) { 13134cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 13144cb5300bSChris Mason if (!ra) 13154cb5300bSChris Mason return -ENOMEM; 13164cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 13174cb5300bSChris Mason } else { 13184cb5300bSChris Mason ra = &file->f_ra; 13194cb5300bSChris Mason } 13204cb5300bSChris Mason 1321d9b0d9baSDulshani Gunawardhana pages = kmalloc_array(max_cluster, sizeof(struct page *), 13224cb5300bSChris Mason GFP_NOFS); 13234cb5300bSChris Mason if (!pages) { 13244cb5300bSChris Mason ret = -ENOMEM; 13254cb5300bSChris Mason goto out_ra; 13264cb5300bSChris Mason } 13274cb5300bSChris Mason 13284cb5300bSChris Mason /* find the last page to defrag */ 13291e701a32SChris Mason if (range->start + range->len > range->start) { 1330151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 133109cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 13321e701a32SChris Mason } else { 133309cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 13341e701a32SChris Mason } 13351e701a32SChris Mason 13364cb5300bSChris Mason if (newer_than) { 13374cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1338ee22184bSByongho Lee &newer_off, SZ_64K); 13394cb5300bSChris Mason if (!ret) { 13404cb5300bSChris Mason range->start = newer_off; 13414cb5300bSChris Mason /* 13424cb5300bSChris Mason * we always align our defrag to help keep 13434cb5300bSChris Mason * the extents in the file evenly spaced 13444cb5300bSChris Mason */ 134509cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 13464cb5300bSChris Mason } else 13474cb5300bSChris Mason goto out_ra; 13484cb5300bSChris Mason } else { 134909cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 13504cb5300bSChris Mason } 13514cb5300bSChris Mason if (!max_to_defrag) 1352070034bdSchandan max_to_defrag = last_index - i + 1; 13534cb5300bSChris Mason 13542a0f7f57SLi Zefan /* 13552a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13562a0f7f57SLi Zefan * written sequentially. 13572a0f7f57SLi Zefan */ 13582a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13592a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13602a0f7f57SLi Zefan 1361f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 136209cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 13634cb5300bSChris Mason /* 13644cb5300bSChris Mason * make sure we stop running if someone unmounts 13654cb5300bSChris Mason * the FS 13664cb5300bSChris Mason */ 13674cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 13684cb5300bSChris Mason break; 13694cb5300bSChris Mason 13700b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 13710b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1372210549ebSDavid Sterba ret = -EAGAIN; 1373210549ebSDavid Sterba break; 1374210549ebSDavid Sterba } 1375210549ebSDavid Sterba 137609cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 13776c282eb4SLi Zefan extent_thresh, &last_len, &skip, 1378a43a2111SAndrew Mahone &defrag_end, range->flags & 1379a43a2111SAndrew Mahone BTRFS_DEFRAG_RANGE_COMPRESS)) { 1380940100a4SChris Mason unsigned long next; 1381940100a4SChris Mason /* 1382940100a4SChris Mason * the should_defrag function tells us how much to skip 1383940100a4SChris Mason * bump our counter by the suggested amount 1384940100a4SChris Mason */ 138509cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1386940100a4SChris Mason i = max(i + 1, next); 1387940100a4SChris Mason continue; 1388940100a4SChris Mason } 1389008873eaSLi Zefan 1390008873eaSLi Zefan if (!newer_than) { 139109cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 139209cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1393008873eaSLi Zefan cluster = min(cluster, max_cluster); 1394008873eaSLi Zefan } else { 1395008873eaSLi Zefan cluster = max_cluster; 1396008873eaSLi Zefan } 1397008873eaSLi Zefan 1398008873eaSLi Zefan if (i + cluster > ra_index) { 1399008873eaSLi Zefan ra_index = max(i, ra_index); 1400008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1401008873eaSLi Zefan cluster); 1402e4826a5bSchandan ra_index += cluster; 1403008873eaSLi Zefan } 14044cb5300bSChris Mason 14055955102cSAl Viro inode_lock(inode); 1406633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 1407633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = compress_type; 1408008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1409ecb8bea8SLiu Bo if (ret < 0) { 14105955102cSAl Viro inode_unlock(inode); 14114cb5300bSChris Mason goto out_ra; 1412ecb8bea8SLiu Bo } 14134cb5300bSChris Mason 14144cb5300bSChris Mason defrag_count += ret; 1415d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 14165955102cSAl Viro inode_unlock(inode); 14174cb5300bSChris Mason 14184cb5300bSChris Mason if (newer_than) { 14194cb5300bSChris Mason if (newer_off == (u64)-1) 14204cb5300bSChris Mason break; 14214cb5300bSChris Mason 1422e1f041e1SLiu Bo if (ret > 0) 1423e1f041e1SLiu Bo i += ret; 1424e1f041e1SLiu Bo 14254cb5300bSChris Mason newer_off = max(newer_off + 1, 142609cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 14274cb5300bSChris Mason 1428ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1429ee22184bSByongho Lee &newer_off, SZ_64K); 14304cb5300bSChris Mason if (!ret) { 14314cb5300bSChris Mason range->start = newer_off; 143209cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14334cb5300bSChris Mason } else { 14344cb5300bSChris Mason break; 1435940100a4SChris Mason } 14364cb5300bSChris Mason } else { 1437008873eaSLi Zefan if (ret > 0) { 1438cbcc8326SLi Zefan i += ret; 143909cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1440008873eaSLi Zefan } else { 1441940100a4SChris Mason i++; 1442008873eaSLi Zefan last_len = 0; 1443008873eaSLi Zefan } 1444f46b5a66SChristoph Hellwig } 14454cb5300bSChris Mason } 1446f46b5a66SChristoph Hellwig 1447dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14481e701a32SChris Mason filemap_flush(inode->i_mapping); 1449dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1450dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1451dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1452dec8ef90SFilipe Manana } 14531e701a32SChris Mason 14541e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 14551e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 14561e701a32SChris Mason * we have to make sure the IO is actually started and that 14571e701a32SChris Mason * ordered extents get created before we return 14581e701a32SChris Mason */ 14590b246afaSJeff Mahoney atomic_inc(&fs_info->async_submit_draining); 14600b246afaSJeff Mahoney while (atomic_read(&fs_info->nr_async_submits) || 14610b246afaSJeff Mahoney atomic_read(&fs_info->async_delalloc_pages)) { 14620b246afaSJeff Mahoney wait_event(fs_info->async_submit_wait, 14630b246afaSJeff Mahoney (atomic_read(&fs_info->nr_async_submits) == 0 && 14640b246afaSJeff Mahoney atomic_read(&fs_info->async_delalloc_pages) == 0)); 14651e701a32SChris Mason } 14660b246afaSJeff Mahoney atomic_dec(&fs_info->async_submit_draining); 14671e701a32SChris Mason } 14681e701a32SChris Mason 14691a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14700b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 14711a419d85SLi Zefan } 14721a419d85SLi Zefan 147360ccf82fSDiego Calleja ret = defrag_count; 1474940100a4SChris Mason 14754cb5300bSChris Mason out_ra: 1476633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 14775955102cSAl Viro inode_lock(inode); 1478633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 14795955102cSAl Viro inode_unlock(inode); 1480633085c7SFilipe David Borba Manana } 14814cb5300bSChris Mason if (!file) 14824cb5300bSChris Mason kfree(ra); 14834cb5300bSChris Mason kfree(pages); 1484940100a4SChris Mason return ret; 1485f46b5a66SChristoph Hellwig } 1486f46b5a66SChristoph Hellwig 1487198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 148876dda93cSYan, Zheng void __user *arg) 1489f46b5a66SChristoph Hellwig { 14900b246afaSJeff Mahoney struct inode *inode = file_inode(file); 14910b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1492f46b5a66SChristoph Hellwig u64 new_size; 1493f46b5a66SChristoph Hellwig u64 old_size; 1494f46b5a66SChristoph Hellwig u64 devid = 1; 14950b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1496f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1497f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1498f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1499f46b5a66SChristoph Hellwig char *sizestr; 15009a40f122SGui Hecheng char *retptr; 1501f46b5a66SChristoph Hellwig char *devstr = NULL; 1502f46b5a66SChristoph Hellwig int ret = 0; 1503f46b5a66SChristoph Hellwig int mod = 0; 1504f46b5a66SChristoph Hellwig 1505e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1506e441d54dSChris Mason return -EPERM; 1507e441d54dSChris Mason 1508198605a8SMiao Xie ret = mnt_want_write_file(file); 1509198605a8SMiao Xie if (ret) 1510198605a8SMiao Xie return ret; 1511198605a8SMiao Xie 15120b246afaSJeff Mahoney if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 151397547676SMiao Xie mnt_drop_write_file(file); 1514e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1515c9e9f97bSIlya Dryomov } 1516c9e9f97bSIlya Dryomov 15170b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 1518dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1519c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1520c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1521c9e9f97bSIlya Dryomov goto out; 1522c9e9f97bSIlya Dryomov } 15235516e595SMark Fasheh 15245516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1525f46b5a66SChristoph Hellwig 1526f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1527f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1528f46b5a66SChristoph Hellwig if (devstr) { 1529f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1530f46b5a66SChristoph Hellwig *devstr = '\0'; 1531f46b5a66SChristoph Hellwig devstr = vol_args->name; 153258dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 153358dfae63SZhangZhen if (ret) 153458dfae63SZhangZhen goto out_free; 1535dfd79829SMiao Xie if (!devid) { 1536dfd79829SMiao Xie ret = -EINVAL; 1537dfd79829SMiao Xie goto out_free; 1538dfd79829SMiao Xie } 15390b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1540f46b5a66SChristoph Hellwig } 1541dba60f3fSMiao Xie 15420b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1543f46b5a66SChristoph Hellwig if (!device) { 15440b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1545c1c9ff7cSGeert Uytterhoeven devid); 1546dfd79829SMiao Xie ret = -ENODEV; 1547c9e9f97bSIlya Dryomov goto out_free; 1548f46b5a66SChristoph Hellwig } 1549dba60f3fSMiao Xie 1550dba60f3fSMiao Xie if (!device->writeable) { 15510b246afaSJeff Mahoney btrfs_info(fs_info, 1552efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1553c1c9ff7cSGeert Uytterhoeven devid); 1554dfd79829SMiao Xie ret = -EPERM; 15554e42ae1bSLiu Bo goto out_free; 15564e42ae1bSLiu Bo } 15574e42ae1bSLiu Bo 1558f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1559f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1560f46b5a66SChristoph Hellwig else { 1561f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1562f46b5a66SChristoph Hellwig mod = -1; 1563f46b5a66SChristoph Hellwig sizestr++; 1564f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1565f46b5a66SChristoph Hellwig mod = 1; 1566f46b5a66SChristoph Hellwig sizestr++; 1567f46b5a66SChristoph Hellwig } 15689a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15699a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1570f46b5a66SChristoph Hellwig ret = -EINVAL; 1571c9e9f97bSIlya Dryomov goto out_free; 1572f46b5a66SChristoph Hellwig } 1573f46b5a66SChristoph Hellwig } 1574f46b5a66SChristoph Hellwig 157563a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1576dfd79829SMiao Xie ret = -EPERM; 157763a212abSStefan Behrens goto out_free; 157863a212abSStefan Behrens } 157963a212abSStefan Behrens 15807cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1581f46b5a66SChristoph Hellwig 1582f46b5a66SChristoph Hellwig if (mod < 0) { 1583f46b5a66SChristoph Hellwig if (new_size > old_size) { 1584f46b5a66SChristoph Hellwig ret = -EINVAL; 1585c9e9f97bSIlya Dryomov goto out_free; 1586f46b5a66SChristoph Hellwig } 1587f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1588f46b5a66SChristoph Hellwig } else if (mod > 0) { 1589eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1590902c68a4SGui Hecheng ret = -ERANGE; 1591eb8052e0SWenliang Fan goto out_free; 1592eb8052e0SWenliang Fan } 1593f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1594f46b5a66SChristoph Hellwig } 1595f46b5a66SChristoph Hellwig 1596ee22184bSByongho Lee if (new_size < SZ_256M) { 1597f46b5a66SChristoph Hellwig ret = -EINVAL; 1598c9e9f97bSIlya Dryomov goto out_free; 1599f46b5a66SChristoph Hellwig } 1600f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1601f46b5a66SChristoph Hellwig ret = -EFBIG; 1602c9e9f97bSIlya Dryomov goto out_free; 1603f46b5a66SChristoph Hellwig } 1604f46b5a66SChristoph Hellwig 16050b246afaSJeff Mahoney new_size = div_u64(new_size, fs_info->sectorsize); 16060b246afaSJeff Mahoney new_size *= fs_info->sectorsize; 1607f46b5a66SChristoph Hellwig 16080b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1609c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1610f46b5a66SChristoph Hellwig 1611f46b5a66SChristoph Hellwig if (new_size > old_size) { 1612a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 161398d5dc13STsutomu Itoh if (IS_ERR(trans)) { 161498d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1615c9e9f97bSIlya Dryomov goto out_free; 161698d5dc13STsutomu Itoh } 1617f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1618f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1619ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1620f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 16210253f40eSjeff.liu } /* equal, nothing need to do */ 1622f46b5a66SChristoph Hellwig 1623c9e9f97bSIlya Dryomov out_free: 1624f46b5a66SChristoph Hellwig kfree(vol_args); 1625c9e9f97bSIlya Dryomov out: 16260b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 16270b246afaSJeff Mahoney atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 162818f39c41SIlya Dryomov mnt_drop_write_file(file); 1629f46b5a66SChristoph Hellwig return ret; 1630f46b5a66SChristoph Hellwig } 1631f46b5a66SChristoph Hellwig 163272fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 16336f72c7e2SArne Jansen char *name, unsigned long fd, int subvol, 16346f72c7e2SArne Jansen u64 *transid, bool readonly, 16358696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1636f46b5a66SChristoph Hellwig { 1637f46b5a66SChristoph Hellwig int namelen; 16383de4586cSChris Mason int ret = 0; 1639f46b5a66SChristoph Hellwig 1640325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1641325c50e3SJeff Mahoney return -ENOTDIR; 1642325c50e3SJeff Mahoney 1643a874a63eSLiu Bo ret = mnt_want_write_file(file); 1644a874a63eSLiu Bo if (ret) 1645a874a63eSLiu Bo goto out; 1646a874a63eSLiu Bo 164772fd032eSSage Weil namelen = strlen(name); 164872fd032eSSage Weil if (strchr(name, '/')) { 1649f46b5a66SChristoph Hellwig ret = -EINVAL; 1650a874a63eSLiu Bo goto out_drop_write; 1651f46b5a66SChristoph Hellwig } 1652f46b5a66SChristoph Hellwig 165316780cabSChris Mason if (name[0] == '.' && 165416780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 165516780cabSChris Mason ret = -EEXIST; 1656a874a63eSLiu Bo goto out_drop_write; 165716780cabSChris Mason } 165816780cabSChris Mason 16593de4586cSChris Mason if (subvol) { 166072fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16616f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1662cb8e7090SChristoph Hellwig } else { 16632903ff01SAl Viro struct fd src = fdget(fd); 16643de4586cSChris Mason struct inode *src_inode; 16652903ff01SAl Viro if (!src.file) { 16663de4586cSChris Mason ret = -EINVAL; 1667a874a63eSLiu Bo goto out_drop_write; 16683de4586cSChris Mason } 16693de4586cSChris Mason 1670496ad9aaSAl Viro src_inode = file_inode(src.file); 1671496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1672c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1673efe120a0SFrank Holton "Snapshot src from another FS"); 167423ad5b17SKusanagi Kouichi ret = -EXDEV; 1675d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1676d0242061SDavid Sterba /* 1677d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1678d0242061SDavid Sterba * are limited to own subvolumes only 1679d0242061SDavid Sterba */ 1680d0242061SDavid Sterba ret = -EPERM; 1681ecd18815SAl Viro } else { 168272fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 168372fd032eSSage Weil BTRFS_I(src_inode)->root, 16846f72c7e2SArne Jansen transid, readonly, inherit); 1685ecd18815SAl Viro } 16862903ff01SAl Viro fdput(src); 1687cb8e7090SChristoph Hellwig } 1688a874a63eSLiu Bo out_drop_write: 1689a874a63eSLiu Bo mnt_drop_write_file(file); 1690f46b5a66SChristoph Hellwig out: 169172fd032eSSage Weil return ret; 169272fd032eSSage Weil } 169372fd032eSSage Weil 169472fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1695fa0d2b9bSLi Zefan void __user *arg, int subvol) 169672fd032eSSage Weil { 1697fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 169872fd032eSSage Weil int ret; 169972fd032eSSage Weil 1700325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1701325c50e3SJeff Mahoney return -ENOTDIR; 1702325c50e3SJeff Mahoney 1703fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1704fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1705fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1706fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1707fa0d2b9bSLi Zefan 1708fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1709b83cc969SLi Zefan vol_args->fd, subvol, 17106f72c7e2SArne Jansen NULL, false, NULL); 1711fa0d2b9bSLi Zefan 1712fa0d2b9bSLi Zefan kfree(vol_args); 1713fa0d2b9bSLi Zefan return ret; 1714fa0d2b9bSLi Zefan } 1715fa0d2b9bSLi Zefan 1716fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1717fa0d2b9bSLi Zefan void __user *arg, int subvol) 1718fa0d2b9bSLi Zefan { 1719fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1720fa0d2b9bSLi Zefan int ret; 1721fdfb1e4fSLi Zefan u64 transid = 0; 1722fdfb1e4fSLi Zefan u64 *ptr = NULL; 1723b83cc969SLi Zefan bool readonly = false; 17246f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 172572fd032eSSage Weil 1726325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1727325c50e3SJeff Mahoney return -ENOTDIR; 1728325c50e3SJeff Mahoney 1729fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1730fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1731fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1732fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1733fdfb1e4fSLi Zefan 1734b83cc969SLi Zefan if (vol_args->flags & 17356f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 17366f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1737b83cc969SLi Zefan ret = -EOPNOTSUPP; 1738c47ca32dSDan Carpenter goto free_args; 1739fdfb1e4fSLi Zefan } 1740fdfb1e4fSLi Zefan 1741fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1742fdfb1e4fSLi Zefan ptr = &transid; 1743b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1744b83cc969SLi Zefan readonly = true; 17456f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 174609cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 17476f72c7e2SArne Jansen ret = -EINVAL; 1748c47ca32dSDan Carpenter goto free_args; 17496f72c7e2SArne Jansen } 17506f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 17516f72c7e2SArne Jansen if (IS_ERR(inherit)) { 17526f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1753c47ca32dSDan Carpenter goto free_args; 17546f72c7e2SArne Jansen } 17556f72c7e2SArne Jansen } 175675eaa0e2SSage Weil 1757fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 17586f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17598696c533SMiao Xie readonly, inherit); 1760c47ca32dSDan Carpenter if (ret) 1761c47ca32dSDan Carpenter goto free_inherit; 176275eaa0e2SSage Weil 1763c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1764fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1765c47ca32dSDan Carpenter transid), 1766c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 176775eaa0e2SSage Weil ret = -EFAULT; 1768c47ca32dSDan Carpenter 1769c47ca32dSDan Carpenter free_inherit: 17706f72c7e2SArne Jansen kfree(inherit); 1771c47ca32dSDan Carpenter free_args: 1772c47ca32dSDan Carpenter kfree(vol_args); 1773f46b5a66SChristoph Hellwig return ret; 1774f46b5a66SChristoph Hellwig } 1775f46b5a66SChristoph Hellwig 17760caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17770caa102dSLi Zefan void __user *arg) 17780caa102dSLi Zefan { 1779496ad9aaSAl Viro struct inode *inode = file_inode(file); 17800b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17810caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17820caa102dSLi Zefan int ret = 0; 17830caa102dSLi Zefan u64 flags = 0; 17840caa102dSLi Zefan 178533345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 17860caa102dSLi Zefan return -EINVAL; 17870caa102dSLi Zefan 17880b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 17890caa102dSLi Zefan if (btrfs_root_readonly(root)) 17900caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17910b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 17920caa102dSLi Zefan 17930caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 17940caa102dSLi Zefan ret = -EFAULT; 17950caa102dSLi Zefan 17960caa102dSLi Zefan return ret; 17970caa102dSLi Zefan } 17980caa102dSLi Zefan 17990caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 18000caa102dSLi Zefan void __user *arg) 18010caa102dSLi Zefan { 1802496ad9aaSAl Viro struct inode *inode = file_inode(file); 18030b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 18040caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18050caa102dSLi Zefan struct btrfs_trans_handle *trans; 18060caa102dSLi Zefan u64 root_flags; 18070caa102dSLi Zefan u64 flags; 18080caa102dSLi Zefan int ret = 0; 18090caa102dSLi Zefan 1810bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1811bd60ea0fSDavid Sterba return -EPERM; 1812bd60ea0fSDavid Sterba 1813b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1814b9ca0664SLiu Bo if (ret) 1815b9ca0664SLiu Bo goto out; 18160caa102dSLi Zefan 1817b9ca0664SLiu Bo if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 1818b9ca0664SLiu Bo ret = -EINVAL; 1819b9ca0664SLiu Bo goto out_drop_write; 1820b9ca0664SLiu Bo } 18210caa102dSLi Zefan 1822b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1823b9ca0664SLiu Bo ret = -EFAULT; 1824b9ca0664SLiu Bo goto out_drop_write; 1825b9ca0664SLiu Bo } 18260caa102dSLi Zefan 1827b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1828b9ca0664SLiu Bo ret = -EINVAL; 1829b9ca0664SLiu Bo goto out_drop_write; 1830b9ca0664SLiu Bo } 18310caa102dSLi Zefan 1832b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1833b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1834b9ca0664SLiu Bo goto out_drop_write; 1835b9ca0664SLiu Bo } 18360caa102dSLi Zefan 18370b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 18380caa102dSLi Zefan 18390caa102dSLi Zefan /* nothing to do */ 18400caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1841b9ca0664SLiu Bo goto out_drop_sem; 18420caa102dSLi Zefan 18430caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 18442c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 18450caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18460caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 18472c686537SDavid Sterba } else { 18482c686537SDavid Sterba /* 18492c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 18502c686537SDavid Sterba * send 18512c686537SDavid Sterba */ 18522c686537SDavid Sterba spin_lock(&root->root_item_lock); 18532c686537SDavid Sterba if (root->send_in_progress == 0) { 18540caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18550caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 18562c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18572c686537SDavid Sterba } else { 18582c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18590b246afaSJeff Mahoney btrfs_warn(fs_info, 18602c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18612c686537SDavid Sterba root->root_key.objectid); 18622c686537SDavid Sterba ret = -EPERM; 18632c686537SDavid Sterba goto out_drop_sem; 18642c686537SDavid Sterba } 18652c686537SDavid Sterba } 18660caa102dSLi Zefan 18670caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18680caa102dSLi Zefan if (IS_ERR(trans)) { 18690caa102dSLi Zefan ret = PTR_ERR(trans); 18700caa102dSLi Zefan goto out_reset; 18710caa102dSLi Zefan } 18720caa102dSLi Zefan 18730b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 18740caa102dSLi Zefan &root->root_key, &root->root_item); 18750caa102dSLi Zefan 18760caa102dSLi Zefan btrfs_commit_transaction(trans, root); 18770caa102dSLi Zefan out_reset: 18780caa102dSLi Zefan if (ret) 18790caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1880b9ca0664SLiu Bo out_drop_sem: 18810b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1882b9ca0664SLiu Bo out_drop_write: 1883b9ca0664SLiu Bo mnt_drop_write_file(file); 1884b9ca0664SLiu Bo out: 18850caa102dSLi Zefan return ret; 18860caa102dSLi Zefan } 18870caa102dSLi Zefan 188876dda93cSYan, Zheng /* 188976dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 189076dda93cSYan, Zheng */ 189176dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 189276dda93cSYan, Zheng { 18930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 189476dda93cSYan, Zheng struct btrfs_path *path; 1895175a2b87SJosef Bacik struct btrfs_dir_item *di; 189676dda93cSYan, Zheng struct btrfs_key key; 1897175a2b87SJosef Bacik u64 dir_id; 189876dda93cSYan, Zheng int ret; 189976dda93cSYan, Zheng 190076dda93cSYan, Zheng path = btrfs_alloc_path(); 190176dda93cSYan, Zheng if (!path) 190276dda93cSYan, Zheng return -ENOMEM; 190376dda93cSYan, Zheng 1904175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 19050b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 19060b246afaSJeff Mahoney di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path, 1907175a2b87SJosef Bacik dir_id, "default", 7, 0); 1908175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1909175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1910175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 191172de6b53SGuangyu Sun ret = -EPERM; 19120b246afaSJeff Mahoney btrfs_err(fs_info, 19135d163e0eSJeff Mahoney "deleting default subvolume %llu is not allowed", 19145d163e0eSJeff Mahoney key.objectid); 1915175a2b87SJosef Bacik goto out; 1916175a2b87SJosef Bacik } 1917175a2b87SJosef Bacik btrfs_release_path(path); 1918175a2b87SJosef Bacik } 1919175a2b87SJosef Bacik 192076dda93cSYan, Zheng key.objectid = root->root_key.objectid; 192176dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 192276dda93cSYan, Zheng key.offset = (u64)-1; 192376dda93cSYan, Zheng 19240b246afaSJeff Mahoney ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 192576dda93cSYan, Zheng if (ret < 0) 192676dda93cSYan, Zheng goto out; 192776dda93cSYan, Zheng BUG_ON(ret == 0); 192876dda93cSYan, Zheng 192976dda93cSYan, Zheng ret = 0; 193076dda93cSYan, Zheng if (path->slots[0] > 0) { 193176dda93cSYan, Zheng path->slots[0]--; 193276dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 193376dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 193476dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 193576dda93cSYan, Zheng ret = -ENOTEMPTY; 193676dda93cSYan, Zheng } 193776dda93cSYan, Zheng out: 193876dda93cSYan, Zheng btrfs_free_path(path); 193976dda93cSYan, Zheng return ret; 194076dda93cSYan, Zheng } 194176dda93cSYan, Zheng 1942ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1943ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1944ac8e9819SChris Mason { 1945abc6e134SChris Mason struct btrfs_key test; 1946abc6e134SChris Mason int ret; 1947abc6e134SChris Mason 1948abc6e134SChris Mason test.objectid = sk->min_objectid; 1949abc6e134SChris Mason test.type = sk->min_type; 1950abc6e134SChris Mason test.offset = sk->min_offset; 1951abc6e134SChris Mason 1952abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1953abc6e134SChris Mason if (ret < 0) 1954ac8e9819SChris Mason return 0; 1955abc6e134SChris Mason 1956abc6e134SChris Mason test.objectid = sk->max_objectid; 1957abc6e134SChris Mason test.type = sk->max_type; 1958abc6e134SChris Mason test.offset = sk->max_offset; 1959abc6e134SChris Mason 1960abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1961abc6e134SChris Mason if (ret > 0) 1962ac8e9819SChris Mason return 0; 1963ac8e9819SChris Mason return 1; 1964ac8e9819SChris Mason } 1965ac8e9819SChris Mason 1966df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 1967ac8e9819SChris Mason struct btrfs_key *key, 1968ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19699b6e817dSGerhard Heift size_t *buf_size, 1970ba346b35SGerhard Heift char __user *ubuf, 1971ac8e9819SChris Mason unsigned long *sk_offset, 1972ac8e9819SChris Mason int *num_found) 1973ac8e9819SChris Mason { 1974ac8e9819SChris Mason u64 found_transid; 1975ac8e9819SChris Mason struct extent_buffer *leaf; 1976ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1977dd81d459SNaohiro Aota struct btrfs_key test; 1978ac8e9819SChris Mason unsigned long item_off; 1979ac8e9819SChris Mason unsigned long item_len; 1980ac8e9819SChris Mason int nritems; 1981ac8e9819SChris Mason int i; 1982ac8e9819SChris Mason int slot; 1983ac8e9819SChris Mason int ret = 0; 1984ac8e9819SChris Mason 1985ac8e9819SChris Mason leaf = path->nodes[0]; 1986ac8e9819SChris Mason slot = path->slots[0]; 1987ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1988ac8e9819SChris Mason 1989ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1990ac8e9819SChris Mason i = nritems; 1991ac8e9819SChris Mason goto advance_key; 1992ac8e9819SChris Mason } 1993ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1994ac8e9819SChris Mason 1995ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1996ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1997ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1998ac8e9819SChris Mason 199903b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 200003b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 200103b71c6cSGabriel de Perthuis continue; 200203b71c6cSGabriel de Perthuis 20039b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 20048f5f6178SGerhard Heift if (*num_found) { 2005ac8e9819SChris Mason ret = 1; 20068f5f6178SGerhard Heift goto out; 20078f5f6178SGerhard Heift } 20088f5f6178SGerhard Heift 20098f5f6178SGerhard Heift /* 20108f5f6178SGerhard Heift * return one empty item back for v1, which does not 20118f5f6178SGerhard Heift * handle -EOVERFLOW 20128f5f6178SGerhard Heift */ 20138f5f6178SGerhard Heift 20149b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2015ac8e9819SChris Mason item_len = 0; 20168f5f6178SGerhard Heift ret = -EOVERFLOW; 20178f5f6178SGerhard Heift } 2018ac8e9819SChris Mason 20199b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2020ac8e9819SChris Mason ret = 1; 202125c9bc2eSGerhard Heift goto out; 2022ac8e9819SChris Mason } 2023ac8e9819SChris Mason 2024ac8e9819SChris Mason sh.objectid = key->objectid; 2025ac8e9819SChris Mason sh.offset = key->offset; 2026ac8e9819SChris Mason sh.type = key->type; 2027ac8e9819SChris Mason sh.len = item_len; 2028ac8e9819SChris Mason sh.transid = found_transid; 2029ac8e9819SChris Mason 2030ac8e9819SChris Mason /* copy search result header */ 2031ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2032ba346b35SGerhard Heift ret = -EFAULT; 2033ba346b35SGerhard Heift goto out; 2034ba346b35SGerhard Heift } 2035ba346b35SGerhard Heift 2036ac8e9819SChris Mason *sk_offset += sizeof(sh); 2037ac8e9819SChris Mason 2038ac8e9819SChris Mason if (item_len) { 2039ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2040ac8e9819SChris Mason /* copy the item */ 2041ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2042ba346b35SGerhard Heift item_off, item_len)) { 2043ba346b35SGerhard Heift ret = -EFAULT; 2044ba346b35SGerhard Heift goto out; 2045ba346b35SGerhard Heift } 2046ba346b35SGerhard Heift 2047ac8e9819SChris Mason *sk_offset += item_len; 2048ac8e9819SChris Mason } 2049e2156867SHugo Mills (*num_found)++; 2050ac8e9819SChris Mason 20518f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20528f5f6178SGerhard Heift goto out; 20538f5f6178SGerhard Heift 205425c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 205525c9bc2eSGerhard Heift ret = 1; 205625c9bc2eSGerhard Heift goto out; 205725c9bc2eSGerhard Heift } 2058ac8e9819SChris Mason } 2059ac8e9819SChris Mason advance_key: 2060ac8e9819SChris Mason ret = 0; 2061dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2062dd81d459SNaohiro Aota test.type = sk->max_type; 2063dd81d459SNaohiro Aota test.offset = sk->max_offset; 2064dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2065dd81d459SNaohiro Aota ret = 1; 2066dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2067abc6e134SChris Mason key->offset++; 2068dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2069abc6e134SChris Mason key->offset = 0; 2070abc6e134SChris Mason key->type++; 2071dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2072abc6e134SChris Mason key->offset = 0; 2073abc6e134SChris Mason key->type = 0; 2074abc6e134SChris Mason key->objectid++; 2075abc6e134SChris Mason } else 2076abc6e134SChris Mason ret = 1; 207725c9bc2eSGerhard Heift out: 2078ba346b35SGerhard Heift /* 2079ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2080ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2081ba346b35SGerhard Heift * * all items were found 2082ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2083ba346b35SGerhard Heift * leaf 2084ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2085ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2086ba346b35SGerhard Heift */ 2087ac8e9819SChris Mason return ret; 2088ac8e9819SChris Mason } 2089ac8e9819SChris Mason 2090ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 209112544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 20929b6e817dSGerhard Heift size_t *buf_size, 2093ba346b35SGerhard Heift char __user *ubuf) 2094ac8e9819SChris Mason { 20950b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2096ac8e9819SChris Mason struct btrfs_root *root; 2097ac8e9819SChris Mason struct btrfs_key key; 2098ac8e9819SChris Mason struct btrfs_path *path; 2099ac8e9819SChris Mason int ret; 2100ac8e9819SChris Mason int num_found = 0; 2101ac8e9819SChris Mason unsigned long sk_offset = 0; 2102ac8e9819SChris Mason 21039b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 21049b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 210512544442SGerhard Heift return -EOVERFLOW; 21069b6e817dSGerhard Heift } 210712544442SGerhard Heift 2108ac8e9819SChris Mason path = btrfs_alloc_path(); 2109ac8e9819SChris Mason if (!path) 2110ac8e9819SChris Mason return -ENOMEM; 2111ac8e9819SChris Mason 2112ac8e9819SChris Mason if (sk->tree_id == 0) { 2113ac8e9819SChris Mason /* search the root of the inode that was passed */ 2114ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2115ac8e9819SChris Mason } else { 2116ac8e9819SChris Mason key.objectid = sk->tree_id; 2117ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2118ac8e9819SChris Mason key.offset = (u64)-1; 2119ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2120ac8e9819SChris Mason if (IS_ERR(root)) { 2121ac8e9819SChris Mason btrfs_free_path(path); 2122ac8e9819SChris Mason return -ENOENT; 2123ac8e9819SChris Mason } 2124ac8e9819SChris Mason } 2125ac8e9819SChris Mason 2126ac8e9819SChris Mason key.objectid = sk->min_objectid; 2127ac8e9819SChris Mason key.type = sk->min_type; 2128ac8e9819SChris Mason key.offset = sk->min_offset; 2129ac8e9819SChris Mason 2130ac8e9819SChris Mason while (1) { 21316174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2132ac8e9819SChris Mason if (ret != 0) { 2133ac8e9819SChris Mason if (ret > 0) 2134ac8e9819SChris Mason ret = 0; 2135ac8e9819SChris Mason goto err; 2136ac8e9819SChris Mason } 2137df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2138ac8e9819SChris Mason &sk_offset, &num_found); 2139b3b4aa74SDavid Sterba btrfs_release_path(path); 214025c9bc2eSGerhard Heift if (ret) 2141ac8e9819SChris Mason break; 2142ac8e9819SChris Mason 2143ac8e9819SChris Mason } 21448f5f6178SGerhard Heift if (ret > 0) 2145ac8e9819SChris Mason ret = 0; 2146ac8e9819SChris Mason err: 2147ac8e9819SChris Mason sk->nr_items = num_found; 2148ac8e9819SChris Mason btrfs_free_path(path); 2149ac8e9819SChris Mason return ret; 2150ac8e9819SChris Mason } 2151ac8e9819SChris Mason 2152ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2153ac8e9819SChris Mason void __user *argp) 2154ac8e9819SChris Mason { 2155ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2156ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2157ac8e9819SChris Mason struct inode *inode; 2158ac8e9819SChris Mason int ret; 21599b6e817dSGerhard Heift size_t buf_size; 2160ac8e9819SChris Mason 2161ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2162ac8e9819SChris Mason return -EPERM; 2163ac8e9819SChris Mason 2164ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2165ac8e9819SChris Mason 2166ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2167ba346b35SGerhard Heift return -EFAULT; 2168ba346b35SGerhard Heift 2169ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2170ac8e9819SChris Mason 2171496ad9aaSAl Viro inode = file_inode(file); 2172ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21738f5f6178SGerhard Heift 21748f5f6178SGerhard Heift /* 21758f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21768f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21778f5f6178SGerhard Heift */ 21788f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21798f5f6178SGerhard Heift ret = 0; 21808f5f6178SGerhard Heift 2181ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2182ac8e9819SChris Mason ret = -EFAULT; 2183ac8e9819SChris Mason return ret; 2184ac8e9819SChris Mason } 2185ac8e9819SChris Mason 2186cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2187cc68a8a5SGerhard Heift void __user *argp) 2188cc68a8a5SGerhard Heift { 2189cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2190cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2191cc68a8a5SGerhard Heift struct inode *inode; 2192cc68a8a5SGerhard Heift int ret; 2193cc68a8a5SGerhard Heift size_t buf_size; 2194ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2195cc68a8a5SGerhard Heift 2196cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2197cc68a8a5SGerhard Heift return -EPERM; 2198cc68a8a5SGerhard Heift 2199cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2200cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2201cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2202cc68a8a5SGerhard Heift return -EFAULT; 2203cc68a8a5SGerhard Heift 2204cc68a8a5SGerhard Heift buf_size = args.buf_size; 2205cc68a8a5SGerhard Heift 2206cc68a8a5SGerhard Heift if (buf_size < sizeof(struct btrfs_ioctl_search_header)) 2207cc68a8a5SGerhard Heift return -EOVERFLOW; 2208cc68a8a5SGerhard Heift 2209cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2210cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2211cc68a8a5SGerhard Heift buf_size = buf_limit; 2212cc68a8a5SGerhard Heift 2213cc68a8a5SGerhard Heift inode = file_inode(file); 2214cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2215cc68a8a5SGerhard Heift (char *)(&uarg->buf[0])); 2216cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2217cc68a8a5SGerhard Heift ret = -EFAULT; 2218cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2219cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2220cc68a8a5SGerhard Heift ret = -EFAULT; 2221cc68a8a5SGerhard Heift 222276dda93cSYan, Zheng return ret; 222376dda93cSYan, Zheng } 222476dda93cSYan, Zheng 222598d377a0STARUISI Hiroaki /* 2226ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2227ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 222898d377a0STARUISI Hiroaki */ 222998d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 223098d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 223198d377a0STARUISI Hiroaki { 223298d377a0STARUISI Hiroaki struct btrfs_root *root; 223398d377a0STARUISI Hiroaki struct btrfs_key key; 2234ac8e9819SChris Mason char *ptr; 223598d377a0STARUISI Hiroaki int ret = -1; 223698d377a0STARUISI Hiroaki int slot; 223798d377a0STARUISI Hiroaki int len; 223898d377a0STARUISI Hiroaki int total_len = 0; 223998d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 224098d377a0STARUISI Hiroaki struct extent_buffer *l; 224198d377a0STARUISI Hiroaki struct btrfs_path *path; 224298d377a0STARUISI Hiroaki 224398d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 224498d377a0STARUISI Hiroaki name[0]='\0'; 224598d377a0STARUISI Hiroaki return 0; 224698d377a0STARUISI Hiroaki } 224798d377a0STARUISI Hiroaki 224898d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 224998d377a0STARUISI Hiroaki if (!path) 225098d377a0STARUISI Hiroaki return -ENOMEM; 225198d377a0STARUISI Hiroaki 2252ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 225398d377a0STARUISI Hiroaki 225498d377a0STARUISI Hiroaki key.objectid = tree_id; 225598d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 225698d377a0STARUISI Hiroaki key.offset = (u64)-1; 225798d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 225898d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2259f14d104dSDavid Sterba btrfs_err(info, "could not find root %llu", tree_id); 22608ad6fcabSChris Mason ret = -ENOENT; 22618ad6fcabSChris Mason goto out; 226298d377a0STARUISI Hiroaki } 226398d377a0STARUISI Hiroaki 226498d377a0STARUISI Hiroaki key.objectid = dirid; 226598d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22668ad6fcabSChris Mason key.offset = (u64)-1; 226798d377a0STARUISI Hiroaki 226898d377a0STARUISI Hiroaki while (1) { 226998d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 227098d377a0STARUISI Hiroaki if (ret < 0) 227198d377a0STARUISI Hiroaki goto out; 227218674c6cSFilipe David Borba Manana else if (ret > 0) { 227318674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 227418674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 227518674c6cSFilipe David Borba Manana if (ret < 0) 227618674c6cSFilipe David Borba Manana goto out; 227718674c6cSFilipe David Borba Manana else if (ret > 0) { 2278ac8e9819SChris Mason ret = -ENOENT; 227998d377a0STARUISI Hiroaki goto out; 2280ac8e9819SChris Mason } 228118674c6cSFilipe David Borba Manana } 228218674c6cSFilipe David Borba Manana 228318674c6cSFilipe David Borba Manana l = path->nodes[0]; 228418674c6cSFilipe David Borba Manana slot = path->slots[0]; 228518674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 228698d377a0STARUISI Hiroaki 228798d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 228898d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 228998d377a0STARUISI Hiroaki ptr -= len + 1; 229098d377a0STARUISI Hiroaki total_len += len + 1; 2291a696cf35SFilipe David Borba Manana if (ptr < name) { 2292a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 229398d377a0STARUISI Hiroaki goto out; 2294a696cf35SFilipe David Borba Manana } 229598d377a0STARUISI Hiroaki 229698d377a0STARUISI Hiroaki *(ptr + len) = '/'; 229798d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 229898d377a0STARUISI Hiroaki 229998d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 230098d377a0STARUISI Hiroaki break; 230198d377a0STARUISI Hiroaki 2302b3b4aa74SDavid Sterba btrfs_release_path(path); 230398d377a0STARUISI Hiroaki key.objectid = key.offset; 23048ad6fcabSChris Mason key.offset = (u64)-1; 230598d377a0STARUISI Hiroaki dirid = key.objectid; 230698d377a0STARUISI Hiroaki } 230777906a50SLi Zefan memmove(name, ptr, total_len); 230898d377a0STARUISI Hiroaki name[total_len] = '\0'; 230998d377a0STARUISI Hiroaki ret = 0; 231098d377a0STARUISI Hiroaki out: 231198d377a0STARUISI Hiroaki btrfs_free_path(path); 2312ac8e9819SChris Mason return ret; 2313ac8e9819SChris Mason } 2314ac8e9819SChris Mason 2315ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2316ac8e9819SChris Mason void __user *argp) 2317ac8e9819SChris Mason { 2318ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2319ac8e9819SChris Mason struct inode *inode; 232001b810b8SDavid Sterba int ret = 0; 2321ac8e9819SChris Mason 23222354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 23232354d08fSJulia Lawall if (IS_ERR(args)) 23242354d08fSJulia Lawall return PTR_ERR(args); 2325c2b96929SDan Carpenter 2326496ad9aaSAl Viro inode = file_inode(file); 2327ac8e9819SChris Mason 232801b810b8SDavid Sterba /* 232901b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 233001b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 233101b810b8SDavid Sterba */ 23321b53ac4dSChris Mason if (args->treeid == 0) 23331b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 23341b53ac4dSChris Mason 233501b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 233601b810b8SDavid Sterba args->name[0] = 0; 233701b810b8SDavid Sterba goto out; 233801b810b8SDavid Sterba } 233901b810b8SDavid Sterba 234001b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 234101b810b8SDavid Sterba ret = -EPERM; 234201b810b8SDavid Sterba goto out; 234301b810b8SDavid Sterba } 234401b810b8SDavid Sterba 2345ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2346ac8e9819SChris Mason args->treeid, args->objectid, 2347ac8e9819SChris Mason args->name); 2348ac8e9819SChris Mason 234901b810b8SDavid Sterba out: 2350ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2351ac8e9819SChris Mason ret = -EFAULT; 2352ac8e9819SChris Mason 2353ac8e9819SChris Mason kfree(args); 235498d377a0STARUISI Hiroaki return ret; 235598d377a0STARUISI Hiroaki } 235698d377a0STARUISI Hiroaki 235776dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 235876dda93cSYan, Zheng void __user *arg) 235976dda93cSYan, Zheng { 236054563d41SAl Viro struct dentry *parent = file->f_path.dentry; 23610b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 236276dda93cSYan, Zheng struct dentry *dentry; 23632b0143b5SDavid Howells struct inode *dir = d_inode(parent); 236476dda93cSYan, Zheng struct inode *inode; 236576dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 236676dda93cSYan, Zheng struct btrfs_root *dest = NULL; 236776dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 236876dda93cSYan, Zheng struct btrfs_trans_handle *trans; 2369c58aaad2SMiao Xie struct btrfs_block_rsv block_rsv; 2370521e0546SDavid Sterba u64 root_flags; 2371c58aaad2SMiao Xie u64 qgroup_reserved; 237276dda93cSYan, Zheng int namelen; 237376dda93cSYan, Zheng int ret; 237476dda93cSYan, Zheng int err = 0; 237576dda93cSYan, Zheng 2376325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2377325c50e3SJeff Mahoney return -ENOTDIR; 2378325c50e3SJeff Mahoney 237976dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 238076dda93cSYan, Zheng if (IS_ERR(vol_args)) 238176dda93cSYan, Zheng return PTR_ERR(vol_args); 238276dda93cSYan, Zheng 238376dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 238476dda93cSYan, Zheng namelen = strlen(vol_args->name); 238576dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 238676dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 238776dda93cSYan, Zheng err = -EINVAL; 238876dda93cSYan, Zheng goto out; 238976dda93cSYan, Zheng } 239076dda93cSYan, Zheng 2391a561be71SAl Viro err = mnt_want_write_file(file); 239276dda93cSYan, Zheng if (err) 239376dda93cSYan, Zheng goto out; 239476dda93cSYan, Zheng 2395521e0546SDavid Sterba 239600235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 239700235411SAl Viro if (err == -EINTR) 239800235411SAl Viro goto out_drop_write; 239976dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 240076dda93cSYan, Zheng if (IS_ERR(dentry)) { 240176dda93cSYan, Zheng err = PTR_ERR(dentry); 240276dda93cSYan, Zheng goto out_unlock_dir; 240376dda93cSYan, Zheng } 240476dda93cSYan, Zheng 24052b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 240676dda93cSYan, Zheng err = -ENOENT; 240776dda93cSYan, Zheng goto out_dput; 240876dda93cSYan, Zheng } 240976dda93cSYan, Zheng 24102b0143b5SDavid Howells inode = d_inode(dentry); 24114260f7c7SSage Weil dest = BTRFS_I(inode)->root; 24124260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 24134260f7c7SSage Weil /* 24144260f7c7SSage Weil * Regular user. Only allow this with a special mount 24154260f7c7SSage Weil * option, when the user has write+exec access to the 24164260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 24174260f7c7SSage Weil * allowed. 24184260f7c7SSage Weil * 24194260f7c7SSage Weil * Note that this is _not_ check that the subvol is 24204260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 24214260f7c7SSage Weil * otherwise be able to delete. 24224260f7c7SSage Weil * 24234260f7c7SSage Weil * Users who want to delete empty subvols should try 24244260f7c7SSage Weil * rmdir(2). 24254260f7c7SSage Weil */ 24264260f7c7SSage Weil err = -EPERM; 24270b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 24284260f7c7SSage Weil goto out_dput; 24294260f7c7SSage Weil 24304260f7c7SSage Weil /* 24314260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 24324260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 24334260f7c7SSage Weil * must be called on the dentry referencing the root 24344260f7c7SSage Weil * of the subvol, not a random directory contained 24354260f7c7SSage Weil * within it. 24364260f7c7SSage Weil */ 24374260f7c7SSage Weil err = -EINVAL; 24384260f7c7SSage Weil if (root == dest) 24394260f7c7SSage Weil goto out_dput; 24404260f7c7SSage Weil 24414260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 24424260f7c7SSage Weil if (err) 24434260f7c7SSage Weil goto out_dput; 24445c39da5bSMiao Xie } 24454260f7c7SSage Weil 24465c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24474260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24484260f7c7SSage Weil if (err) 24494260f7c7SSage Weil goto out_dput; 24504260f7c7SSage Weil 245133345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 245276dda93cSYan, Zheng err = -EINVAL; 245376dda93cSYan, Zheng goto out_dput; 245476dda93cSYan, Zheng } 245576dda93cSYan, Zheng 24565955102cSAl Viro inode_lock(inode); 2457521e0546SDavid Sterba 2458521e0546SDavid Sterba /* 2459521e0546SDavid Sterba * Don't allow to delete a subvolume with send in progress. This is 2460521e0546SDavid Sterba * inside the i_mutex so the error handling that has to drop the bit 2461521e0546SDavid Sterba * again is not run concurrently. 2462521e0546SDavid Sterba */ 2463521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2464c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2465c55bfa67SFilipe Manana if (dest->send_in_progress == 0) { 2466c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2467521e0546SDavid Sterba root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2468521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2469521e0546SDavid Sterba } else { 2470521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 24710b246afaSJeff Mahoney btrfs_warn(fs_info, 2472521e0546SDavid Sterba "Attempt to delete subvolume %llu during send", 2473c55bfa67SFilipe Manana dest->root_key.objectid); 2474521e0546SDavid Sterba err = -EPERM; 2475909e26dcSOmar Sandoval goto out_unlock_inode; 2476521e0546SDavid Sterba } 2477521e0546SDavid Sterba 24780b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 247976dda93cSYan, Zheng 248076dda93cSYan, Zheng err = may_destroy_subvol(dest); 248176dda93cSYan, Zheng if (err) 248276dda93cSYan, Zheng goto out_up_write; 248376dda93cSYan, Zheng 2484c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2485c58aaad2SMiao Xie /* 2486c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2487c58aaad2SMiao Xie * ref/backref. 2488c58aaad2SMiao Xie */ 2489c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2490ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2491c58aaad2SMiao Xie if (err) 2492c58aaad2SMiao Xie goto out_up_write; 2493c58aaad2SMiao Xie 2494a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2495a22285a6SYan, Zheng if (IS_ERR(trans)) { 2496a22285a6SYan, Zheng err = PTR_ERR(trans); 2497c58aaad2SMiao Xie goto out_release; 2498a22285a6SYan, Zheng } 2499c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2500c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2501a22285a6SYan, Zheng 25022be63d5cSFilipe Manana btrfs_record_snapshot_destroy(trans, dir); 25032be63d5cSFilipe Manana 250476dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 250576dda93cSYan, Zheng dest->root_key.objectid, 250676dda93cSYan, Zheng dentry->d_name.name, 250776dda93cSYan, Zheng dentry->d_name.len); 250879787eaaSJeff Mahoney if (ret) { 250979787eaaSJeff Mahoney err = ret; 251066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 251179787eaaSJeff Mahoney goto out_end_trans; 251279787eaaSJeff Mahoney } 251376dda93cSYan, Zheng 251476dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 251576dda93cSYan, Zheng 251676dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 251776dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 251876dda93cSYan, Zheng dest->root_item.drop_level = 0; 251976dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 252076dda93cSYan, Zheng 252127cdeb70SMiao Xie if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 252276dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 25230b246afaSJeff Mahoney fs_info->tree_root, 252476dda93cSYan, Zheng dest->root_key.objectid); 252579787eaaSJeff Mahoney if (ret) { 252666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 252779787eaaSJeff Mahoney err = ret; 252879787eaaSJeff Mahoney goto out_end_trans; 2529d68fc57bSYan, Zheng } 253079787eaaSJeff Mahoney } 2531dd5f9615SStefan Behrens 25320b246afaSJeff Mahoney ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid, 25336bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, 2534dd5f9615SStefan Behrens dest->root_key.objectid); 2535dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 253666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 2537dd5f9615SStefan Behrens err = ret; 2538dd5f9615SStefan Behrens goto out_end_trans; 2539dd5f9615SStefan Behrens } 2540dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 25410b246afaSJeff Mahoney ret = btrfs_uuid_tree_rem(trans, fs_info, 2542dd5f9615SStefan Behrens dest->root_item.received_uuid, 2543dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2544dd5f9615SStefan Behrens dest->root_key.objectid); 2545dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 254666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 2547dd5f9615SStefan Behrens err = ret; 2548dd5f9615SStefan Behrens goto out_end_trans; 2549dd5f9615SStefan Behrens } 2550dd5f9615SStefan Behrens } 2551dd5f9615SStefan Behrens 255279787eaaSJeff Mahoney out_end_trans: 2553c58aaad2SMiao Xie trans->block_rsv = NULL; 2554c58aaad2SMiao Xie trans->bytes_reserved = 0; 2555531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 255679787eaaSJeff Mahoney if (ret && !err) 255779787eaaSJeff Mahoney err = ret; 255876dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2559c58aaad2SMiao Xie out_release: 2560c58aaad2SMiao Xie btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 256176dda93cSYan, Zheng out_up_write: 25620b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 2563521e0546SDavid Sterba if (err) { 2564521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2565c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2566c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2567521e0546SDavid Sterba root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2568521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2569521e0546SDavid Sterba } 2570909e26dcSOmar Sandoval out_unlock_inode: 25715955102cSAl Viro inode_unlock(inode); 257276dda93cSYan, Zheng if (!err) { 257364ad6c48SOmar Sandoval d_invalidate(dentry); 257476dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 257576dda93cSYan, Zheng d_delete(dentry); 257661155aa0SDavid Sterba ASSERT(dest->send_in_progress == 0); 2577fa6ac876SLiu Bo 2578fa6ac876SLiu Bo /* the last ref */ 257957cdc8dbSDavid Sterba if (dest->ino_cache_inode) { 258057cdc8dbSDavid Sterba iput(dest->ino_cache_inode); 258157cdc8dbSDavid Sterba dest->ino_cache_inode = NULL; 2582fa6ac876SLiu Bo } 258376dda93cSYan, Zheng } 258476dda93cSYan, Zheng out_dput: 258576dda93cSYan, Zheng dput(dentry); 258676dda93cSYan, Zheng out_unlock_dir: 25875955102cSAl Viro inode_unlock(dir); 258800235411SAl Viro out_drop_write: 25892a79f17eSAl Viro mnt_drop_write_file(file); 259076dda93cSYan, Zheng out: 259176dda93cSYan, Zheng kfree(vol_args); 259276dda93cSYan, Zheng return err; 259376dda93cSYan, Zheng } 259476dda93cSYan, Zheng 25951e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2596f46b5a66SChristoph Hellwig { 2597496ad9aaSAl Viro struct inode *inode = file_inode(file); 2598f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25991e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2600c146afadSYan Zheng int ret; 2601c146afadSYan Zheng 260225122d15SIlya Dryomov ret = mnt_want_write_file(file); 260325122d15SIlya Dryomov if (ret) 260425122d15SIlya Dryomov return ret; 2605b83cc969SLi Zefan 260625122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 260725122d15SIlya Dryomov ret = -EROFS; 260825122d15SIlya Dryomov goto out; 26095ac00addSStefan Behrens } 2610f46b5a66SChristoph Hellwig 2611f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2612f46b5a66SChristoph Hellwig case S_IFDIR: 2613e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2614e441d54dSChris Mason ret = -EPERM; 2615e441d54dSChris Mason goto out; 2616e441d54dSChris Mason } 2617de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 26188929ecfaSYan, Zheng if (ret) 26198929ecfaSYan, Zheng goto out; 2620de78b51aSEric Sandeen ret = btrfs_defrag_root(root->fs_info->extent_root); 2621f46b5a66SChristoph Hellwig break; 2622f46b5a66SChristoph Hellwig case S_IFREG: 2623e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2624e441d54dSChris Mason ret = -EINVAL; 2625e441d54dSChris Mason goto out; 2626e441d54dSChris Mason } 26271e701a32SChris Mason 26281e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 26291e701a32SChris Mason if (!range) { 26301e701a32SChris Mason ret = -ENOMEM; 26311e701a32SChris Mason goto out; 26321e701a32SChris Mason } 26331e701a32SChris Mason 26341e701a32SChris Mason if (argp) { 26351e701a32SChris Mason if (copy_from_user(range, argp, 26361e701a32SChris Mason sizeof(*range))) { 26371e701a32SChris Mason ret = -EFAULT; 26381e701a32SChris Mason kfree(range); 2639683be16eSDan Carpenter goto out; 26401e701a32SChris Mason } 26411e701a32SChris Mason /* compression requires us to start the IO */ 26421e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 26431e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 26441e701a32SChris Mason range->extent_thresh = (u32)-1; 26451e701a32SChris Mason } 26461e701a32SChris Mason } else { 26471e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 26481e701a32SChris Mason range->len = (u64)-1; 26491e701a32SChris Mason } 2650496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 26514cb5300bSChris Mason range, 0, 0); 26524cb5300bSChris Mason if (ret > 0) 26534cb5300bSChris Mason ret = 0; 26541e701a32SChris Mason kfree(range); 2655f46b5a66SChristoph Hellwig break; 26568929ecfaSYan, Zheng default: 26578929ecfaSYan, Zheng ret = -EINVAL; 2658f46b5a66SChristoph Hellwig } 2659e441d54dSChris Mason out: 266025122d15SIlya Dryomov mnt_drop_write_file(file); 2661e441d54dSChris Mason return ret; 2662f46b5a66SChristoph Hellwig } 2663f46b5a66SChristoph Hellwig 2664b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2665f46b5a66SChristoph Hellwig { 26660b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 2667f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2668f46b5a66SChristoph Hellwig int ret; 2669f46b5a66SChristoph Hellwig 2670e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2671e441d54dSChris Mason return -EPERM; 2672e441d54dSChris Mason 26730b246afaSJeff Mahoney if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) 2674e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2675c9e9f97bSIlya Dryomov 26760b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 2677dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2678c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2679c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2680c9e9f97bSIlya Dryomov goto out; 2681c9e9f97bSIlya Dryomov } 2682f46b5a66SChristoph Hellwig 26835516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 26845112febbSJeff Mahoney ret = btrfs_init_new_device(root->fs_info, vol_args->name); 2685f46b5a66SChristoph Hellwig 268643d20761SAnand Jain if (!ret) 26870b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 268843d20761SAnand Jain 2689f46b5a66SChristoph Hellwig kfree(vol_args); 2690c9e9f97bSIlya Dryomov out: 26910b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 26920b246afaSJeff Mahoney atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 2693f46b5a66SChristoph Hellwig return ret; 2694f46b5a66SChristoph Hellwig } 2695f46b5a66SChristoph Hellwig 26966b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2697f46b5a66SChristoph Hellwig { 26980b246afaSJeff Mahoney struct inode *inode = file_inode(file); 26990b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 27000b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 27016b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 2702f46b5a66SChristoph Hellwig int ret; 2703f46b5a66SChristoph Hellwig 2704e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2705e441d54dSChris Mason return -EPERM; 2706e441d54dSChris Mason 2707da24927bSMiao Xie ret = mnt_want_write_file(file); 2708da24927bSMiao Xie if (ret) 2709da24927bSMiao Xie return ret; 2710c146afadSYan Zheng 2711dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2712c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2713c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2714c47ca32dSDan Carpenter goto err_drop; 2715c9e9f97bSIlya Dryomov } 2716f46b5a66SChristoph Hellwig 27176b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 2718735654eaSDavid Sterba if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) 2719735654eaSDavid Sterba return -EOPNOTSUPP; 2720f46b5a66SChristoph Hellwig 27210b246afaSJeff Mahoney if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 2722183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2723183860f6SAnand Jain goto out; 2724183860f6SAnand Jain } 2725183860f6SAnand Jain 27260b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 2727735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 27286b526ed7SAnand Jain ret = btrfs_rm_device(root, NULL, vol_args->devid); 27296b526ed7SAnand Jain } else { 27306b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 27316b526ed7SAnand Jain ret = btrfs_rm_device(root, vol_args->name, 0); 27326b526ed7SAnand Jain } 27330b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 27340b246afaSJeff Mahoney atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 2735183860f6SAnand Jain 27366b526ed7SAnand Jain if (!ret) { 2737735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 27380b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 27396b526ed7SAnand Jain vol_args->devid); 27406b526ed7SAnand Jain else 27410b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 27426b526ed7SAnand Jain vol_args->name); 27436b526ed7SAnand Jain } 2744183860f6SAnand Jain out: 2745183860f6SAnand Jain kfree(vol_args); 2746c47ca32dSDan Carpenter err_drop: 27474ac20c70SIlya Dryomov mnt_drop_write_file(file); 2748f46b5a66SChristoph Hellwig return ret; 2749f46b5a66SChristoph Hellwig } 2750f46b5a66SChristoph Hellwig 2751f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2752f46b5a66SChristoph Hellwig { 27530b246afaSJeff Mahoney struct inode *inode = file_inode(file); 27540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 27550b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 2756f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2757f46b5a66SChristoph Hellwig int ret; 2758f46b5a66SChristoph Hellwig 2759f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2760f46b5a66SChristoph Hellwig return -EPERM; 2761f46b5a66SChristoph Hellwig 2762f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 2763f46b5a66SChristoph Hellwig if (ret) 2764f46b5a66SChristoph Hellwig return ret; 2765f46b5a66SChristoph Hellwig 27660b246afaSJeff Mahoney if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 2767f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 276858d7bbf8SDavid Sterba goto out_drop_write; 276958d7bbf8SDavid Sterba } 277058d7bbf8SDavid Sterba 277158d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 277258d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 277358d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 2774f46b5a66SChristoph Hellwig goto out; 2775f46b5a66SChristoph Hellwig } 2776f46b5a66SChristoph Hellwig 277758d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 27780b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 27796b526ed7SAnand Jain ret = btrfs_rm_device(root, vol_args->name, 0); 27800b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 2781f46b5a66SChristoph Hellwig 2782f46b5a66SChristoph Hellwig if (!ret) 27830b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 2784f46b5a66SChristoph Hellwig kfree(vol_args); 278558d7bbf8SDavid Sterba out: 27860b246afaSJeff Mahoney atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 278758d7bbf8SDavid Sterba out_drop_write: 2788f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 278958d7bbf8SDavid Sterba 2790f46b5a66SChristoph Hellwig return ret; 2791f46b5a66SChristoph Hellwig } 2792f46b5a66SChristoph Hellwig 2793475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2794475f6387SJan Schmidt { 27950b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 2796027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2797475f6387SJan Schmidt struct btrfs_device *device; 27980b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2799027ed2f0SLi Zefan int ret = 0; 2800475f6387SJan Schmidt 2801027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2802027ed2f0SLi Zefan if (!fi_args) 2803027ed2f0SLi Zefan return -ENOMEM; 2804027ed2f0SLi Zefan 2805f7171750SFilipe David Borba Manana mutex_lock(&fs_devices->device_list_mutex); 2806027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 28070b246afaSJeff Mahoney memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 2808475f6387SJan Schmidt 2809d7641a49SByongho Lee list_for_each_entry(device, &fs_devices->devices, dev_list) { 2810027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2811027ed2f0SLi Zefan fi_args->max_id = device->devid; 2812475f6387SJan Schmidt } 2813475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2814475f6387SJan Schmidt 28150b246afaSJeff Mahoney fi_args->nodesize = fs_info->super_copy->nodesize; 28160b246afaSJeff Mahoney fi_args->sectorsize = fs_info->super_copy->sectorsize; 28170b246afaSJeff Mahoney fi_args->clone_alignment = fs_info->super_copy->sectorsize; 281880a773fbSDavid Sterba 2819027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2820027ed2f0SLi Zefan ret = -EFAULT; 2821475f6387SJan Schmidt 2822027ed2f0SLi Zefan kfree(fi_args); 2823027ed2f0SLi Zefan return ret; 2824475f6387SJan Schmidt } 2825475f6387SJan Schmidt 2826475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2827475f6387SJan Schmidt { 28280b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 2829475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2830475f6387SJan Schmidt struct btrfs_device *dev; 28310b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2832475f6387SJan Schmidt int ret = 0; 2833475f6387SJan Schmidt char *s_uuid = NULL; 2834475f6387SJan Schmidt 2835475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2836475f6387SJan Schmidt if (IS_ERR(di_args)) 2837475f6387SJan Schmidt return PTR_ERR(di_args); 2838475f6387SJan Schmidt 2839dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2840475f6387SJan Schmidt s_uuid = di_args->uuid; 2841475f6387SJan Schmidt 2842475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 28430b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 2844475f6387SJan Schmidt 2845475f6387SJan Schmidt if (!dev) { 2846475f6387SJan Schmidt ret = -ENODEV; 2847475f6387SJan Schmidt goto out; 2848475f6387SJan Schmidt } 2849475f6387SJan Schmidt 2850475f6387SJan Schmidt di_args->devid = dev->devid; 28517cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 28527cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2853475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2854a27202fbSJim Meyering if (dev->name) { 2855606686eeSJosef Bacik struct rcu_string *name; 2856606686eeSJosef Bacik 2857606686eeSJosef Bacik rcu_read_lock(); 2858606686eeSJosef Bacik name = rcu_dereference(dev->name); 2859606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2860606686eeSJosef Bacik rcu_read_unlock(); 2861a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2862a27202fbSJim Meyering } else { 286399ba55adSStefan Behrens di_args->path[0] = '\0'; 2864a27202fbSJim Meyering } 2865475f6387SJan Schmidt 2866475f6387SJan Schmidt out: 286755793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2868475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2869475f6387SJan Schmidt ret = -EFAULT; 2870475f6387SJan Schmidt 2871475f6387SJan Schmidt kfree(di_args); 2872475f6387SJan Schmidt return ret; 2873475f6387SJan Schmidt } 2874475f6387SJan Schmidt 2875f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2876416161dbSMark Fasheh { 2877416161dbSMark Fasheh struct page *page; 2878416161dbSMark Fasheh 2879416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2880416161dbSMark Fasheh if (!page) 288131314002SFilipe Manana return ERR_PTR(-ENOMEM); 2882416161dbSMark Fasheh 2883416161dbSMark Fasheh if (!PageUptodate(page)) { 288431314002SFilipe Manana int ret; 288531314002SFilipe Manana 288631314002SFilipe Manana ret = btrfs_readpage(NULL, page); 288731314002SFilipe Manana if (ret) 288831314002SFilipe Manana return ERR_PTR(ret); 2889416161dbSMark Fasheh lock_page(page); 2890416161dbSMark Fasheh if (!PageUptodate(page)) { 2891416161dbSMark Fasheh unlock_page(page); 289209cbfeafSKirill A. Shutemov put_page(page); 289331314002SFilipe Manana return ERR_PTR(-EIO); 2894416161dbSMark Fasheh } 289531314002SFilipe Manana if (page->mapping != inode->i_mapping) { 2896416161dbSMark Fasheh unlock_page(page); 289709cbfeafSKirill A. Shutemov put_page(page); 289831314002SFilipe Manana return ERR_PTR(-EAGAIN); 2899416161dbSMark Fasheh } 2900416161dbSMark Fasheh } 2901416161dbSMark Fasheh 2902416161dbSMark Fasheh return page; 2903416161dbSMark Fasheh } 2904416161dbSMark Fasheh 2905f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2906f4414602SMark Fasheh int num_pages, u64 off) 2907f4414602SMark Fasheh { 2908f4414602SMark Fasheh int i; 290909cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 2910f4414602SMark Fasheh 2911f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 291231314002SFilipe Manana again: 2913f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 291431314002SFilipe Manana if (IS_ERR(pages[i])) { 291531314002SFilipe Manana int err = PTR_ERR(pages[i]); 291631314002SFilipe Manana 291731314002SFilipe Manana if (err == -EAGAIN) 291831314002SFilipe Manana goto again; 291931314002SFilipe Manana pages[i] = NULL; 292031314002SFilipe Manana return err; 292131314002SFilipe Manana } 2922f4414602SMark Fasheh } 2923f4414602SMark Fasheh return 0; 2924f4414602SMark Fasheh } 2925f4414602SMark Fasheh 2926e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 2927e0bd70c6SFilipe Manana bool retry_range_locking) 292877fe20dcSMark Fasheh { 2929e0bd70c6SFilipe Manana /* 2930e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 2931e0bd70c6SFilipe Manana * another, and lock file content. 2932e0bd70c6SFilipe Manana * The locking order is: 2933e0bd70c6SFilipe Manana * 2934e0bd70c6SFilipe Manana * 1) pages 2935e0bd70c6SFilipe Manana * 2) range in the inode's io tree 2936e0bd70c6SFilipe Manana */ 293777fe20dcSMark Fasheh while (1) { 293877fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 293977fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 294077fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 294177fe20dcSMark Fasheh off + len - 1); 2942ff5df9b8SFilipe Manana if ((!ordered || 2943ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2944ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 294577fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2946ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2947ff5df9b8SFilipe Manana if (ordered) 2948ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 294977fe20dcSMark Fasheh break; 2950ff5df9b8SFilipe Manana } 295177fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 295277fe20dcSMark Fasheh if (ordered) 295377fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 2954e0bd70c6SFilipe Manana if (!retry_range_locking) 2955e0bd70c6SFilipe Manana return -EAGAIN; 295677fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 295777fe20dcSMark Fasheh } 2958e0bd70c6SFilipe Manana return 0; 295977fe20dcSMark Fasheh } 296077fe20dcSMark Fasheh 2961f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2962416161dbSMark Fasheh { 29635955102cSAl Viro inode_unlock(inode1); 29645955102cSAl Viro inode_unlock(inode2); 2965416161dbSMark Fasheh } 2966416161dbSMark Fasheh 2967f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2968f4414602SMark Fasheh { 2969f4414602SMark Fasheh if (inode1 < inode2) 2970f4414602SMark Fasheh swap(inode1, inode2); 2971f4414602SMark Fasheh 29725955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 29735955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 2974f4414602SMark Fasheh } 2975f4414602SMark Fasheh 2976f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2977f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2978f4414602SMark Fasheh { 2979f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2980f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2981f4414602SMark Fasheh } 2982f4414602SMark Fasheh 2983e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2984e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 2985e0bd70c6SFilipe Manana bool retry_range_locking) 2986416161dbSMark Fasheh { 2987e0bd70c6SFilipe Manana int ret; 2988e0bd70c6SFilipe Manana 2989416161dbSMark Fasheh if (inode1 < inode2) { 2990416161dbSMark Fasheh swap(inode1, inode2); 2991416161dbSMark Fasheh swap(loff1, loff2); 2992416161dbSMark Fasheh } 2993e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 2994e0bd70c6SFilipe Manana if (ret) 2995e0bd70c6SFilipe Manana return ret; 2996e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 2997e0bd70c6SFilipe Manana if (ret) 2998e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 2999e0bd70c6SFilipe Manana loff1 + len - 1); 3000e0bd70c6SFilipe Manana return ret; 3001416161dbSMark Fasheh } 3002f4414602SMark Fasheh 3003f4414602SMark Fasheh struct cmp_pages { 3004f4414602SMark Fasheh int num_pages; 3005f4414602SMark Fasheh struct page **src_pages; 3006f4414602SMark Fasheh struct page **dst_pages; 3007f4414602SMark Fasheh }; 3008f4414602SMark Fasheh 3009f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 3010f4414602SMark Fasheh { 3011f4414602SMark Fasheh int i; 3012f4414602SMark Fasheh struct page *pg; 3013f4414602SMark Fasheh 3014f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 3015f4414602SMark Fasheh pg = cmp->src_pages[i]; 3016e0bd70c6SFilipe Manana if (pg) { 3017e0bd70c6SFilipe Manana unlock_page(pg); 301809cbfeafSKirill A. Shutemov put_page(pg); 3019e0bd70c6SFilipe Manana } 3020f4414602SMark Fasheh pg = cmp->dst_pages[i]; 3021e0bd70c6SFilipe Manana if (pg) { 3022e0bd70c6SFilipe Manana unlock_page(pg); 302309cbfeafSKirill A. Shutemov put_page(pg); 3024f4414602SMark Fasheh } 3025e0bd70c6SFilipe Manana } 3026f4414602SMark Fasheh kfree(cmp->src_pages); 3027f4414602SMark Fasheh kfree(cmp->dst_pages); 3028f4414602SMark Fasheh } 3029f4414602SMark Fasheh 3030f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 3031f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 3032f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 3033f4414602SMark Fasheh { 3034f4414602SMark Fasheh int ret; 303509cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 3036f4414602SMark Fasheh struct page **src_pgarr, **dst_pgarr; 3037f4414602SMark Fasheh 3038f4414602SMark Fasheh /* 3039f4414602SMark Fasheh * We must gather up all the pages before we initiate our 3040f4414602SMark Fasheh * extent locking. We use an array for the page pointers. Size 3041f4414602SMark Fasheh * of the array is bounded by len, which is in turn bounded by 3042f4414602SMark Fasheh * BTRFS_MAX_DEDUPE_LEN. 3043f4414602SMark Fasheh */ 304466722f7cSDavid Sterba src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 304566722f7cSDavid Sterba dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 3046f4414602SMark Fasheh if (!src_pgarr || !dst_pgarr) { 3047f4414602SMark Fasheh kfree(src_pgarr); 3048f4414602SMark Fasheh kfree(dst_pgarr); 3049f4414602SMark Fasheh return -ENOMEM; 3050f4414602SMark Fasheh } 3051f4414602SMark Fasheh cmp->num_pages = num_pages; 3052f4414602SMark Fasheh cmp->src_pages = src_pgarr; 3053f4414602SMark Fasheh cmp->dst_pages = dst_pgarr; 3054f4414602SMark Fasheh 3055f4414602SMark Fasheh ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff); 3056f4414602SMark Fasheh if (ret) 3057f4414602SMark Fasheh goto out; 3058f4414602SMark Fasheh 3059f4414602SMark Fasheh ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff); 3060f4414602SMark Fasheh 3061f4414602SMark Fasheh out: 3062f4414602SMark Fasheh if (ret) 3063f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 3064f4414602SMark Fasheh return 0; 3065416161dbSMark Fasheh } 3066416161dbSMark Fasheh 3067416161dbSMark Fasheh static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, 3068f4414602SMark Fasheh u64 dst_loff, u64 len, struct cmp_pages *cmp) 3069416161dbSMark Fasheh { 3070416161dbSMark Fasheh int ret = 0; 3071f4414602SMark Fasheh int i; 3072416161dbSMark Fasheh struct page *src_page, *dst_page; 307309cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 3074416161dbSMark Fasheh void *addr, *dst_addr; 3075416161dbSMark Fasheh 3076f4414602SMark Fasheh i = 0; 3077416161dbSMark Fasheh while (len) { 307809cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 3079416161dbSMark Fasheh cmp_len = len; 3080416161dbSMark Fasheh 3081f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 3082f4414602SMark Fasheh 3083f4414602SMark Fasheh src_page = cmp->src_pages[i]; 3084f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 3085e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 3086e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 3087f4414602SMark Fasheh 3088416161dbSMark Fasheh addr = kmap_atomic(src_page); 3089416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 3090416161dbSMark Fasheh 3091416161dbSMark Fasheh flush_dcache_page(src_page); 3092416161dbSMark Fasheh flush_dcache_page(dst_page); 3093416161dbSMark Fasheh 3094416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 30952b3909f8SDarrick J. Wong ret = -EBADE; 3096416161dbSMark Fasheh 3097416161dbSMark Fasheh kunmap_atomic(addr); 3098416161dbSMark Fasheh kunmap_atomic(dst_addr); 3099416161dbSMark Fasheh 3100416161dbSMark Fasheh if (ret) 3101416161dbSMark Fasheh break; 3102416161dbSMark Fasheh 3103416161dbSMark Fasheh len -= cmp_len; 3104f4414602SMark Fasheh i++; 3105416161dbSMark Fasheh } 3106416161dbSMark Fasheh 3107416161dbSMark Fasheh return ret; 3108416161dbSMark Fasheh } 3109416161dbSMark Fasheh 3110e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3111e1d227a4SMark Fasheh u64 olen) 3112416161dbSMark Fasheh { 3113e1d227a4SMark Fasheh u64 len = *plen; 3114416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3115416161dbSMark Fasheh 3116e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 3117416161dbSMark Fasheh return -EINVAL; 3118e1d227a4SMark Fasheh 3119e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 3120e1d227a4SMark Fasheh if (off + len == inode->i_size) 3121e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 3122e1d227a4SMark Fasheh 3123416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 3124416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3125416161dbSMark Fasheh return -EINVAL; 3126416161dbSMark Fasheh 3127416161dbSMark Fasheh return 0; 3128416161dbSMark Fasheh } 3129416161dbSMark Fasheh 3130e1d227a4SMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 3131416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 3132416161dbSMark Fasheh { 3133416161dbSMark Fasheh int ret; 3134e1d227a4SMark Fasheh u64 len = olen; 3135f4414602SMark Fasheh struct cmp_pages cmp; 31360efa9f48SMark Fasheh int same_inode = 0; 31370efa9f48SMark Fasheh u64 same_lock_start = 0; 31380efa9f48SMark Fasheh u64 same_lock_len = 0; 3139416161dbSMark Fasheh 3140416161dbSMark Fasheh if (src == dst) 31410efa9f48SMark Fasheh same_inode = 1; 3142416161dbSMark Fasheh 3143113e8283SFilipe Manana if (len == 0) 3144113e8283SFilipe Manana return 0; 3145113e8283SFilipe Manana 31460efa9f48SMark Fasheh if (same_inode) { 31475955102cSAl Viro inode_lock(src); 31480efa9f48SMark Fasheh 31490efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 31500efa9f48SMark Fasheh if (ret) 31510efa9f48SMark Fasheh goto out_unlock; 3152f4dfe687SFilipe Manana ret = extent_same_check_offsets(src, dst_loff, &len, olen); 3153f4dfe687SFilipe Manana if (ret) 3154f4dfe687SFilipe Manana goto out_unlock; 31550efa9f48SMark Fasheh 31560efa9f48SMark Fasheh /* 31570efa9f48SMark Fasheh * Single inode case wants the same checks, except we 31580efa9f48SMark Fasheh * don't want our length pushed out past i_size as 31590efa9f48SMark Fasheh * comparing that data range makes no sense. 31600efa9f48SMark Fasheh * 31610efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 31620efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 31630efa9f48SMark Fasheh * reject the request. 31640efa9f48SMark Fasheh * 31650efa9f48SMark Fasheh * This effectively means we require aligned extents 31660efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 31670efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 31680efa9f48SMark Fasheh * i_size. 31690efa9f48SMark Fasheh */ 31700efa9f48SMark Fasheh if (len != olen) { 31710efa9f48SMark Fasheh ret = -EINVAL; 31720efa9f48SMark Fasheh goto out_unlock; 31730efa9f48SMark Fasheh } 31740efa9f48SMark Fasheh 31750efa9f48SMark Fasheh /* Check for overlapping ranges */ 31760efa9f48SMark Fasheh if (dst_loff + len > loff && dst_loff < loff + len) { 31770efa9f48SMark Fasheh ret = -EINVAL; 31780efa9f48SMark Fasheh goto out_unlock; 31790efa9f48SMark Fasheh } 31800efa9f48SMark Fasheh 31810efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 31820efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 31830efa9f48SMark Fasheh } else { 3184f4414602SMark Fasheh btrfs_double_inode_lock(src, dst); 3185416161dbSMark Fasheh 3186e1d227a4SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 3187416161dbSMark Fasheh if (ret) 3188416161dbSMark Fasheh goto out_unlock; 3189416161dbSMark Fasheh 3190e1d227a4SMark Fasheh ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3191416161dbSMark Fasheh if (ret) 3192416161dbSMark Fasheh goto out_unlock; 31930efa9f48SMark Fasheh } 3194416161dbSMark Fasheh 3195416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 3196416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3197416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 3198416161dbSMark Fasheh ret = -EINVAL; 3199416161dbSMark Fasheh goto out_unlock; 3200416161dbSMark Fasheh } 3201416161dbSMark Fasheh 3202e0bd70c6SFilipe Manana again: 3203f4414602SMark Fasheh ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 3204f4414602SMark Fasheh if (ret) 3205f4414602SMark Fasheh goto out_unlock; 3206f4414602SMark Fasheh 32070efa9f48SMark Fasheh if (same_inode) 3208e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3209e0bd70c6SFilipe Manana false); 32100efa9f48SMark Fasheh else 3211e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3212e0bd70c6SFilipe Manana false); 3213e0bd70c6SFilipe Manana /* 3214e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3215e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3216e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3217e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3218e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3219e0bd70c6SFilipe Manana * range locking). 3220e0bd70c6SFilipe Manana */ 3221e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3222e0bd70c6SFilipe Manana /* 3223e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3224e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3225e0bd70c6SFilipe Manana */ 3226e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3227e0bd70c6SFilipe Manana if (same_inode) { 3228e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3229e0bd70c6SFilipe Manana same_lock_len); 3230e0bd70c6SFilipe Manana } else { 3231e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3232e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3233e0bd70c6SFilipe Manana } 3234e0bd70c6SFilipe Manana goto again; 3235e0bd70c6SFilipe Manana } 3236e0bd70c6SFilipe Manana ASSERT(ret == 0); 3237e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3238e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3239e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3240e0bd70c6SFilipe Manana return ret; 3241e0bd70c6SFilipe Manana } 3242f4414602SMark Fasheh 3243207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 3244f4414602SMark Fasheh ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); 3245416161dbSMark Fasheh if (ret == 0) 32461c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3247416161dbSMark Fasheh 32480efa9f48SMark Fasheh if (same_inode) 32490efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 32500efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 32510efa9f48SMark Fasheh else 3252f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3253f4414602SMark Fasheh 3254f4414602SMark Fasheh btrfs_cmp_data_free(&cmp); 3255416161dbSMark Fasheh out_unlock: 32560efa9f48SMark Fasheh if (same_inode) 32575955102cSAl Viro inode_unlock(src); 32580efa9f48SMark Fasheh else 3259f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3260416161dbSMark Fasheh 3261416161dbSMark Fasheh return ret; 3262416161dbSMark Fasheh } 3263416161dbSMark Fasheh 3264ee22184bSByongho Lee #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3265416161dbSMark Fasheh 32662b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 32672b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3268416161dbSMark Fasheh { 32692b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 32702b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3271416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 32722b3909f8SDarrick J. Wong ssize_t res; 3273416161dbSMark Fasheh 32742b3909f8SDarrick J. Wong if (olen > BTRFS_MAX_DEDUPE_LEN) 32752b3909f8SDarrick J. Wong olen = BTRFS_MAX_DEDUPE_LEN; 3276416161dbSMark Fasheh 327709cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3278416161dbSMark Fasheh /* 3279416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3280416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3281416161dbSMark Fasheh * this situation without an update. 3282416161dbSMark Fasheh */ 32832b3909f8SDarrick J. Wong return -EINVAL; 3284416161dbSMark Fasheh } 3285416161dbSMark Fasheh 32862b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 32872b3909f8SDarrick J. Wong if (res) 32882b3909f8SDarrick J. Wong return res; 32892b3909f8SDarrick J. Wong return olen; 3290416161dbSMark Fasheh } 3291416161dbSMark Fasheh 3292f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3293f82a9901SFilipe Manana struct inode *inode, 3294f82a9901SFilipe Manana u64 endoff, 3295f82a9901SFilipe Manana const u64 destoff, 32961c919a5eSMark Fasheh const u64 olen, 32971c919a5eSMark Fasheh int no_time_update) 3298f82a9901SFilipe Manana { 3299f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3300f82a9901SFilipe Manana int ret; 3301f82a9901SFilipe Manana 3302f82a9901SFilipe Manana inode_inc_iversion(inode); 33031c919a5eSMark Fasheh if (!no_time_update) 3304c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3305f82a9901SFilipe Manana /* 3306f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3307f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3308f82a9901SFilipe Manana */ 3309f82a9901SFilipe Manana if (endoff > destoff + olen) 3310f82a9901SFilipe Manana endoff = destoff + olen; 3311f82a9901SFilipe Manana if (endoff > inode->i_size) 3312f82a9901SFilipe Manana btrfs_i_size_write(inode, endoff); 3313f82a9901SFilipe Manana 3314f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3315f82a9901SFilipe Manana if (ret) { 331666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 3317f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3318f82a9901SFilipe Manana goto out; 3319f82a9901SFilipe Manana } 3320f82a9901SFilipe Manana ret = btrfs_end_transaction(trans, root); 3321f82a9901SFilipe Manana out: 3322f82a9901SFilipe Manana return ret; 3323f82a9901SFilipe Manana } 3324f82a9901SFilipe Manana 33257ffbb598SFilipe Manana static void clone_update_extent_map(struct inode *inode, 33267ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 33277ffbb598SFilipe Manana const struct btrfs_path *path, 33287ffbb598SFilipe Manana const u64 hole_offset, 33297ffbb598SFilipe Manana const u64 hole_len) 33307ffbb598SFilipe Manana { 33317ffbb598SFilipe Manana struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 33327ffbb598SFilipe Manana struct extent_map *em; 33337ffbb598SFilipe Manana int ret; 33347ffbb598SFilipe Manana 33357ffbb598SFilipe Manana em = alloc_extent_map(); 33367ffbb598SFilipe Manana if (!em) { 33377ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 33387ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 33397ffbb598SFilipe Manana return; 33407ffbb598SFilipe Manana } 33417ffbb598SFilipe Manana 334214f59796SFilipe Manana if (path) { 334314f59796SFilipe Manana struct btrfs_file_extent_item *fi; 334414f59796SFilipe Manana 334514f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 334614f59796SFilipe Manana struct btrfs_file_extent_item); 33477ffbb598SFilipe Manana btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 33487ffbb598SFilipe Manana em->generation = -1; 33497ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 33507ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33517ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 33527ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 33537ffbb598SFilipe Manana } else { 33547ffbb598SFilipe Manana em->start = hole_offset; 33557ffbb598SFilipe Manana em->len = hole_len; 33567ffbb598SFilipe Manana em->ram_bytes = em->len; 33577ffbb598SFilipe Manana em->orig_start = hole_offset; 33587ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33597ffbb598SFilipe Manana em->block_len = 0; 33607ffbb598SFilipe Manana em->orig_block_len = 0; 33617ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33627ffbb598SFilipe Manana em->generation = trans->transid; 33637ffbb598SFilipe Manana } 33647ffbb598SFilipe Manana 33657ffbb598SFilipe Manana while (1) { 33667ffbb598SFilipe Manana write_lock(&em_tree->lock); 33677ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33687ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33697ffbb598SFilipe Manana if (ret != -EEXIST) { 33707ffbb598SFilipe Manana free_extent_map(em); 33717ffbb598SFilipe Manana break; 33727ffbb598SFilipe Manana } 33737ffbb598SFilipe Manana btrfs_drop_extent_cache(inode, em->start, 33747ffbb598SFilipe Manana em->start + em->len - 1, 0); 33757ffbb598SFilipe Manana } 33767ffbb598SFilipe Manana 3377ee39b432SDavid Sterba if (ret) 33787ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 33797ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 33807ffbb598SFilipe Manana } 33817ffbb598SFilipe Manana 33828039d87dSFilipe Manana /* 33838039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33848039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33858039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33868039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33878039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33888039d87dSFilipe Manana * 33898039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33908039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33918039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33928039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33938039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33948039d87dSFilipe Manana * i_mutex, which we are holding here). 33958039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33968039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33978039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33988039d87dSFilipe Manana * 33998039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 34008039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 34018039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 34028039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 34038039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 34048039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 34058039d87dSFilipe Manana * the former. 34068039d87dSFilipe Manana */ 34078039d87dSFilipe Manana static int clone_copy_inline_extent(struct inode *src, 34088039d87dSFilipe Manana struct inode *dst, 34098039d87dSFilipe Manana struct btrfs_trans_handle *trans, 34108039d87dSFilipe Manana struct btrfs_path *path, 34118039d87dSFilipe Manana struct btrfs_key *new_key, 34128039d87dSFilipe Manana const u64 drop_start, 34138039d87dSFilipe Manana const u64 datal, 34148039d87dSFilipe Manana const u64 skip, 34158039d87dSFilipe Manana const u64 size, 34168039d87dSFilipe Manana char *inline_data) 34178039d87dSFilipe Manana { 34180b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 34198039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 34208039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 34210b246afaSJeff Mahoney fs_info->sectorsize); 34228039d87dSFilipe Manana int ret; 34238039d87dSFilipe Manana struct btrfs_key key; 34248039d87dSFilipe Manana 34258039d87dSFilipe Manana if (new_key->offset > 0) 34268039d87dSFilipe Manana return -EOPNOTSUPP; 34278039d87dSFilipe Manana 34288039d87dSFilipe Manana key.objectid = btrfs_ino(dst); 34298039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 34308039d87dSFilipe Manana key.offset = 0; 34318039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 34328039d87dSFilipe Manana if (ret < 0) { 34338039d87dSFilipe Manana return ret; 34348039d87dSFilipe Manana } else if (ret > 0) { 34358039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 34368039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 34378039d87dSFilipe Manana if (ret < 0) 34388039d87dSFilipe Manana return ret; 34398039d87dSFilipe Manana else if (ret > 0) 34408039d87dSFilipe Manana goto copy_inline_extent; 34418039d87dSFilipe Manana } 34428039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 34438039d87dSFilipe Manana if (key.objectid == btrfs_ino(dst) && 34448039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 34458039d87dSFilipe Manana ASSERT(key.offset > 0); 34468039d87dSFilipe Manana return -EOPNOTSUPP; 34478039d87dSFilipe Manana } 34488039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34498039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34508039d87dSFilipe Manana u64 ext_len; 34518039d87dSFilipe Manana 34528039d87dSFilipe Manana /* 34538039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34548039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34558039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34568039d87dSFilipe Manana */ 34578039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34588039d87dSFilipe Manana struct btrfs_file_extent_item); 34598039d87dSFilipe Manana /* 34608039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34618039d87dSFilipe Manana * following it. 34628039d87dSFilipe Manana */ 34638039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34648039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34658039d87dSFilipe Manana goto copy_inline_extent; 34668039d87dSFilipe Manana 34678039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34688039d87dSFilipe Manana if (ext_len > aligned_end) 34698039d87dSFilipe Manana return -EOPNOTSUPP; 34708039d87dSFilipe Manana 34718039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34728039d87dSFilipe Manana if (ret < 0) { 34738039d87dSFilipe Manana return ret; 34748039d87dSFilipe Manana } else if (ret == 0) { 34758039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34768039d87dSFilipe Manana path->slots[0]); 34778039d87dSFilipe Manana if (key.objectid == btrfs_ino(dst) && 34788039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34798039d87dSFilipe Manana return -EOPNOTSUPP; 34808039d87dSFilipe Manana } 34818039d87dSFilipe Manana } 34828039d87dSFilipe Manana 34838039d87dSFilipe Manana copy_inline_extent: 34848039d87dSFilipe Manana /* 34858039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34868039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34878039d87dSFilipe Manana */ 34888039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34898039d87dSFilipe Manana /* 34908039d87dSFilipe Manana * If the destination inode has an inline extent... 34918039d87dSFilipe Manana * This would require copying the data from the source inline 34928039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34938039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34948039d87dSFilipe Manana * or just one of them, which would require decompressing and 34958039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34968039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34978039d87dSFilipe Manana * inline extent). 34988039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34998039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 35008039d87dSFilipe Manana */ 35018039d87dSFilipe Manana return -EOPNOTSUPP; 35028039d87dSFilipe Manana } 35038039d87dSFilipe Manana 35048039d87dSFilipe Manana btrfs_release_path(path); 35058039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 35068039d87dSFilipe Manana if (ret) 35078039d87dSFilipe Manana return ret; 35088039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 35098039d87dSFilipe Manana if (ret) 35108039d87dSFilipe Manana return ret; 35118039d87dSFilipe Manana 35128039d87dSFilipe Manana if (skip) { 35138039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 35148039d87dSFilipe Manana 35158039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 35168039d87dSFilipe Manana } 35178039d87dSFilipe Manana 35188039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 35198039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 35208039d87dSFilipe Manana path->slots[0]), 35218039d87dSFilipe Manana size); 35228039d87dSFilipe Manana inode_add_bytes(dst, datal); 35238039d87dSFilipe Manana 35248039d87dSFilipe Manana return 0; 35258039d87dSFilipe Manana } 35268039d87dSFilipe Manana 352732b7c687SMark Fasheh /** 352832b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 352932b7c687SMark Fasheh * 353032b7c687SMark Fasheh * @src: Inode to clone from 353132b7c687SMark Fasheh * @inode: Inode to clone to 353232b7c687SMark Fasheh * @off: Offset within source to start clone from 353332b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 35341c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 353532b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 35361c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 353732b7c687SMark Fasheh */ 353832b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3539f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 35401c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3541f46b5a66SChristoph Hellwig { 35420b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3543f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 354432b7c687SMark Fasheh struct btrfs_path *path = NULL; 3545f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 354632b7c687SMark Fasheh struct btrfs_trans_handle *trans; 354732b7c687SMark Fasheh char *buf = NULL; 3548ae01a0abSYan Zheng struct btrfs_key key; 3549f46b5a66SChristoph Hellwig u32 nritems; 3550f46b5a66SChristoph Hellwig int slot; 3551ae01a0abSYan Zheng int ret; 3552f82a9901SFilipe Manana const u64 len = olen_aligned; 3553f82a9901SFilipe Manana u64 last_dest_end = destoff; 3554ae01a0abSYan Zheng 3555ae01a0abSYan Zheng ret = -ENOMEM; 35560b246afaSJeff Mahoney buf = kmalloc(fs_info->nodesize, GFP_KERNEL | __GFP_NOWARN); 355715351955SDavid Sterba if (!buf) { 35580b246afaSJeff Mahoney buf = vmalloc(fs_info->nodesize); 3559ae01a0abSYan Zheng if (!buf) 356032b7c687SMark Fasheh return ret; 356115351955SDavid Sterba } 3562ae01a0abSYan Zheng 3563ae01a0abSYan Zheng path = btrfs_alloc_path(); 3564ae01a0abSYan Zheng if (!path) { 356515351955SDavid Sterba kvfree(buf); 356632b7c687SMark Fasheh return ret; 3567ae01a0abSYan Zheng } 356832b7c687SMark Fasheh 3569e4058b54SDavid Sterba path->reada = READA_FORWARD; 3570c5c9cd4dSSage Weil /* clone data */ 357133345d01SLi Zefan key.objectid = btrfs_ino(src); 3572ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35732c463823SFilipe Manana key.offset = off; 3574f46b5a66SChristoph Hellwig 3575f46b5a66SChristoph Hellwig while (1) { 3576de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3577df858e76SFilipe Manana 3578f46b5a66SChristoph Hellwig /* 3579f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3580f46b5a66SChristoph Hellwig * tree. 3581f46b5a66SChristoph Hellwig */ 3582e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3583362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3584362a20c5SDavid Sterba 0, 0); 3585f46b5a66SChristoph Hellwig if (ret < 0) 3586f46b5a66SChristoph Hellwig goto out; 35872c463823SFilipe Manana /* 35882c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35892c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35902c463823SFilipe Manana * it might overlap our target range, therefore process it. 35912c463823SFilipe Manana */ 35922c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35932c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35942c463823SFilipe Manana path->slots[0] - 1); 35952c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35962c463823SFilipe Manana path->slots[0]--; 35972c463823SFilipe Manana } 3598f46b5a66SChristoph Hellwig 3599ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3600e4355f34SFilipe David Borba Manana process_slot: 3601ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3602362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3603f46b5a66SChristoph Hellwig if (ret < 0) 3604f46b5a66SChristoph Hellwig goto out; 3605f46b5a66SChristoph Hellwig if (ret > 0) 3606f46b5a66SChristoph Hellwig break; 3607ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3608f46b5a66SChristoph Hellwig } 3609f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3610f46b5a66SChristoph Hellwig slot = path->slots[0]; 3611f46b5a66SChristoph Hellwig 3612ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3613962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 361433345d01SLi Zefan key.objectid != btrfs_ino(src)) 3615f46b5a66SChristoph Hellwig break; 3616f46b5a66SChristoph Hellwig 3617962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3618c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3619c5c9cd4dSSage Weil int type; 362031840ae1SZheng Yan u32 size; 362131840ae1SZheng Yan struct btrfs_key new_key; 3622c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3623c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3624c5c9cd4dSSage Weil u8 comp; 3625f82a9901SFilipe Manana u64 drop_start; 362631840ae1SZheng Yan 3627c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3628c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3629c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3630c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3631c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3632c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3633d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3634d397712bSChris Mason extent); 3635d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3636d397712bSChris Mason extent); 3637c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3638d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3639d397712bSChris Mason extent); 3640c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3641c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3642c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3643c5c9cd4dSSage Weil extent); 3644c5c9cd4dSSage Weil } 364531840ae1SZheng Yan 36462c463823SFilipe Manana /* 36472c463823SFilipe Manana * The first search might have left us at an extent 36482c463823SFilipe Manana * item that ends before our target range's start, can 36492c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 36502c463823SFilipe Manana */ 36512c463823SFilipe Manana if (key.offset + datal <= off) { 3652e4355f34SFilipe David Borba Manana path->slots[0]++; 3653e4355f34SFilipe David Borba Manana goto process_slot; 36542c463823SFilipe Manana } else if (key.offset >= off + len) { 36552c463823SFilipe Manana break; 3656e4355f34SFilipe David Borba Manana } 3657df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3658e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3659e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3660e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3661e4355f34SFilipe David Borba Manana size); 3662e4355f34SFilipe David Borba Manana 3663e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3664e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3665c5c9cd4dSSage Weil 366631840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 366733345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 36684d728ec7SLi Zefan if (off <= key.offset) 3669c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36704d728ec7SLi Zefan else 36714d728ec7SLi Zefan new_key.offset = destoff; 3672c5c9cd4dSSage Weil 3673b6f3409bSSage Weil /* 3674f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3675f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3676f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3677f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3678f82a9901SFilipe Manana * partially overlaps it). 3679f82a9901SFilipe Manana */ 3680f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3681f82a9901SFilipe Manana drop_start = last_dest_end; 3682f82a9901SFilipe Manana else 3683f82a9901SFilipe Manana drop_start = new_key.offset; 3684f82a9901SFilipe Manana 3685f82a9901SFilipe Manana /* 3686b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3687b6f3409bSSage Weil * 1 - add new extent 3688b6f3409bSSage Weil * 1 - inode update 3689b6f3409bSSage Weil */ 3690b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3691a22285a6SYan, Zheng if (IS_ERR(trans)) { 3692a22285a6SYan, Zheng ret = PTR_ERR(trans); 3693a22285a6SYan, Zheng goto out; 3694a22285a6SYan, Zheng } 3695a22285a6SYan, Zheng 3696c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3697c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3698d72c0842SLi Zefan /* 3699d72c0842SLi Zefan * a | --- range to clone ---| b 3700d72c0842SLi Zefan * | ------------- extent ------------- | 3701d72c0842SLi Zefan */ 3702d72c0842SLi Zefan 370393915584SAntonio Ospite /* subtract range b */ 3704d72c0842SLi Zefan if (key.offset + datal > off + len) 3705d72c0842SLi Zefan datal = off + len - key.offset; 3706d72c0842SLi Zefan 370793915584SAntonio Ospite /* subtract range a */ 3708a22285a6SYan, Zheng if (off > key.offset) { 3709a22285a6SYan, Zheng datao += off - key.offset; 3710a22285a6SYan, Zheng datal -= off - key.offset; 3711a22285a6SYan, Zheng } 3712a22285a6SYan, Zheng 37135dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3714f82a9901SFilipe Manana drop_start, 3715a22285a6SYan, Zheng new_key.offset + datal, 37162671485dSJosef Bacik 1); 371779787eaaSJeff Mahoney if (ret) { 37183f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 371900fdf13aSLiu Bo btrfs_abort_transaction(trans, 372066642832SJeff Mahoney ret); 372179787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 372279787eaaSJeff Mahoney goto out; 372379787eaaSJeff Mahoney } 3724a22285a6SYan, Zheng 372531840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 372631840ae1SZheng Yan &new_key, size); 372779787eaaSJeff Mahoney if (ret) { 372866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 372979787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 373079787eaaSJeff Mahoney goto out; 373179787eaaSJeff Mahoney } 373231840ae1SZheng Yan 373331840ae1SZheng Yan leaf = path->nodes[0]; 373431840ae1SZheng Yan slot = path->slots[0]; 373531840ae1SZheng Yan write_extent_buffer(leaf, buf, 373631840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 373731840ae1SZheng Yan size); 3738ae01a0abSYan Zheng 3739f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3740f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3741c5c9cd4dSSage Weil 3742c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3743c5c9cd4dSSage Weil if (!disko) 3744c5c9cd4dSSage Weil datao = 0; 3745c5c9cd4dSSage Weil 3746c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3747c5c9cd4dSSage Weil datao); 3748c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3749c5c9cd4dSSage Weil datal); 3750fcebe456SJosef Bacik 3751c5c9cd4dSSage Weil if (disko) { 3752c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 3753ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 37545d4f98a2SYan Zheng disko, diskl, 0, 3755f46b5a66SChristoph Hellwig root->root_key.objectid, 375633345d01SLi Zefan btrfs_ino(inode), 3757b06c4bf5SFilipe Manana new_key.offset - datao); 375879787eaaSJeff Mahoney if (ret) { 375979787eaaSJeff Mahoney btrfs_abort_transaction(trans, 376079787eaaSJeff Mahoney ret); 376179787eaaSJeff Mahoney btrfs_end_transaction(trans, 376279787eaaSJeff Mahoney root); 376379787eaaSJeff Mahoney goto out; 376479787eaaSJeff Mahoney 376579787eaaSJeff Mahoney } 3766f46b5a66SChristoph Hellwig } 3767c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3768c5c9cd4dSSage Weil u64 skip = 0; 3769c5c9cd4dSSage Weil u64 trim = 0; 3770ed958762SFilipe Manana 3771c5c9cd4dSSage Weil if (off > key.offset) { 3772c5c9cd4dSSage Weil skip = off - key.offset; 3773c5c9cd4dSSage Weil new_key.offset += skip; 377431840ae1SZheng Yan } 3775d397712bSChris Mason 3776c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3777c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3778d397712bSChris Mason 3779c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3780c5c9cd4dSSage Weil ret = -EINVAL; 3781a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 3782c5c9cd4dSSage Weil goto out; 378331840ae1SZheng Yan } 3784c5c9cd4dSSage Weil size -= skip + trim; 3785c5c9cd4dSSage Weil datal -= skip + trim; 3786a22285a6SYan, Zheng 37878039d87dSFilipe Manana ret = clone_copy_inline_extent(src, inode, 37888039d87dSFilipe Manana trans, path, 37898039d87dSFilipe Manana &new_key, 3790f82a9901SFilipe Manana drop_start, 37918039d87dSFilipe Manana datal, 37928039d87dSFilipe Manana skip, size, buf); 379379787eaaSJeff Mahoney if (ret) { 37943f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37953a29bc09SChris Mason btrfs_abort_transaction(trans, 379679787eaaSJeff Mahoney ret); 379779787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 379879787eaaSJeff Mahoney goto out; 379979787eaaSJeff Mahoney } 3800c5c9cd4dSSage Weil leaf = path->nodes[0]; 3801c5c9cd4dSSage Weil slot = path->slots[0]; 3802c5c9cd4dSSage Weil } 3803c5c9cd4dSSage Weil 38047ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 38057ffbb598SFilipe Manana if (drop_start < new_key.offset) 38067ffbb598SFilipe Manana clone_update_extent_map(inode, trans, 380714f59796SFilipe Manana NULL, drop_start, 38087ffbb598SFilipe Manana new_key.offset - drop_start); 38097ffbb598SFilipe Manana 381014f59796SFilipe Manana clone_update_extent_map(inode, trans, path, 0, 0); 38117ffbb598SFilipe Manana 3812c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3813b3b4aa74SDavid Sterba btrfs_release_path(path); 3814c5c9cd4dSSage Weil 381562e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 38160b246afaSJeff Mahoney fs_info->sectorsize); 3817f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3818f82a9901SFilipe Manana last_dest_end, 38191c919a5eSMark Fasheh destoff, olen, 38201c919a5eSMark Fasheh no_time_update); 3821f82a9901SFilipe Manana if (ret) 382279787eaaSJeff Mahoney goto out; 38232c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 38242c463823SFilipe Manana break; 3825a22285a6SYan, Zheng } 3826b3b4aa74SDavid Sterba btrfs_release_path(path); 3827df858e76SFilipe Manana key.offset = next_key_min_offset; 382869ae5e44SWang Xiaoguang 382969ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 383069ae5e44SWang Xiaoguang ret = -EINTR; 383169ae5e44SWang Xiaoguang goto out; 383269ae5e44SWang Xiaoguang } 3833ae01a0abSYan Zheng } 3834f46b5a66SChristoph Hellwig ret = 0; 383532b7c687SMark Fasheh 3836f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3837f82a9901SFilipe Manana /* 3838f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3839f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3840f82a9901SFilipe Manana */ 3841b3b4aa74SDavid Sterba btrfs_release_path(path); 3842f82a9901SFilipe Manana 3843f82a9901SFilipe Manana /* 3844f82a9901SFilipe Manana * 1 - remove extent(s) 3845f82a9901SFilipe Manana * 1 - inode update 3846f82a9901SFilipe Manana */ 3847f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3848f82a9901SFilipe Manana if (IS_ERR(trans)) { 3849f82a9901SFilipe Manana ret = PTR_ERR(trans); 3850f82a9901SFilipe Manana goto out; 3851f82a9901SFilipe Manana } 3852f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3853f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3854f82a9901SFilipe Manana if (ret) { 3855f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 385666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 3857f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3858f82a9901SFilipe Manana goto out; 3859f82a9901SFilipe Manana } 386014f59796SFilipe Manana clone_update_extent_map(inode, trans, NULL, last_dest_end, 386114f59796SFilipe Manana destoff + len - last_dest_end); 3862f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38631c919a5eSMark Fasheh destoff, olen, no_time_update); 3864f82a9901SFilipe Manana } 3865f82a9901SFilipe Manana 3866f46b5a66SChristoph Hellwig out: 386732b7c687SMark Fasheh btrfs_free_path(path); 386815351955SDavid Sterba kvfree(buf); 386932b7c687SMark Fasheh return ret; 387032b7c687SMark Fasheh } 387132b7c687SMark Fasheh 38723db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 387332b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 387432b7c687SMark Fasheh { 387554563d41SAl Viro struct inode *inode = file_inode(file); 38763db11b2eSZach Brown struct inode *src = file_inode(file_src); 38770b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 387832b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 387932b7c687SMark Fasheh int ret; 388032b7c687SMark Fasheh u64 len = olen; 38810b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 38823db11b2eSZach Brown int same_inode = src == inode; 388332b7c687SMark Fasheh 388432b7c687SMark Fasheh /* 388532b7c687SMark Fasheh * TODO: 388632b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 388732b7c687SMark Fasheh * decompress into destination's address_space (the file offset 388832b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 388932b7c687SMark Fasheh * otherwise reinsert) a subrange. 389000fdf13aSLiu Bo * 389100fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 389200fdf13aSLiu Bo * be either compressed or non-compressed. 389332b7c687SMark Fasheh */ 389432b7c687SMark Fasheh 389532b7c687SMark Fasheh if (btrfs_root_readonly(root)) 389632b7c687SMark Fasheh return -EROFS; 389732b7c687SMark Fasheh 38983db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 38993db11b2eSZach Brown src->i_sb != inode->i_sb) 39003db11b2eSZach Brown return -EXDEV; 390132b7c687SMark Fasheh 390232b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 390332b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 390432b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 39053db11b2eSZach Brown return -EINVAL; 390632b7c687SMark Fasheh 390732b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 39083db11b2eSZach Brown return -EISDIR; 390932b7c687SMark Fasheh 391032b7c687SMark Fasheh if (!same_inode) { 3911293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 391232b7c687SMark Fasheh } else { 39135955102cSAl Viro inode_lock(src); 391432b7c687SMark Fasheh } 391532b7c687SMark Fasheh 391632b7c687SMark Fasheh /* determine range to clone */ 391732b7c687SMark Fasheh ret = -EINVAL; 391832b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 391932b7c687SMark Fasheh goto out_unlock; 392032b7c687SMark Fasheh if (len == 0) 392132b7c687SMark Fasheh olen = len = src->i_size - off; 392232b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 392332b7c687SMark Fasheh if (off + len == src->i_size) 392432b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 392532b7c687SMark Fasheh 3926ccccf3d6SFilipe Manana if (len == 0) { 3927ccccf3d6SFilipe Manana ret = 0; 3928ccccf3d6SFilipe Manana goto out_unlock; 3929ccccf3d6SFilipe Manana } 3930ccccf3d6SFilipe Manana 393132b7c687SMark Fasheh /* verify the end result is block aligned */ 393232b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 393332b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 393432b7c687SMark Fasheh goto out_unlock; 393532b7c687SMark Fasheh 393632b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 393732b7c687SMark Fasheh if (same_inode) { 393832b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 393932b7c687SMark Fasheh goto out_unlock; 394032b7c687SMark Fasheh } 394132b7c687SMark Fasheh 394232b7c687SMark Fasheh if (destoff > inode->i_size) { 394332b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 394432b7c687SMark Fasheh if (ret) 394532b7c687SMark Fasheh goto out_unlock; 394632b7c687SMark Fasheh } 394732b7c687SMark Fasheh 3948c125b8bfSFilipe Manana /* 3949c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3950c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3951c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3952c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3953c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3954c125b8bfSFilipe Manana */ 3955c125b8bfSFilipe Manana if (same_inode) { 3956c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3957c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 395832b7c687SMark Fasheh 3959e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 3960c125b8bfSFilipe Manana } else { 3961e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 3962e0bd70c6SFilipe Manana true); 3963e0bd70c6SFilipe Manana } 3964e0bd70c6SFilipe Manana ASSERT(ret == 0); 3965e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3966e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3967e0bd70c6SFilipe Manana goto out_unlock; 3968c125b8bfSFilipe Manana } 396932b7c687SMark Fasheh 39701c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 397132b7c687SMark Fasheh 3972c125b8bfSFilipe Manana if (same_inode) { 3973c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3974c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3975c125b8bfSFilipe Manana 3976c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3977c125b8bfSFilipe Manana } else { 3978293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 3979c125b8bfSFilipe Manana } 3980c125b8bfSFilipe Manana /* 3981c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3982c125b8bfSFilipe Manana * data immediately and not the previous data. 3983c125b8bfSFilipe Manana */ 398465bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 398509cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 398609cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 3987f46b5a66SChristoph Hellwig out_unlock: 3988293a8489SMark Fasheh if (!same_inode) 3989293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 3990293a8489SMark Fasheh else 39915955102cSAl Viro inode_unlock(src); 39923db11b2eSZach Brown return ret; 39933db11b2eSZach Brown } 39943db11b2eSZach Brown 39953db11b2eSZach Brown ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in, 39963db11b2eSZach Brown struct file *file_out, loff_t pos_out, 39973db11b2eSZach Brown size_t len, unsigned int flags) 39983db11b2eSZach Brown { 39993db11b2eSZach Brown ssize_t ret; 40003db11b2eSZach Brown 40013db11b2eSZach Brown ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out); 40023db11b2eSZach Brown if (ret == 0) 40033db11b2eSZach Brown ret = len; 40043db11b2eSZach Brown return ret; 40053db11b2eSZach Brown } 40063db11b2eSZach Brown 400704b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 400804b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 40093db11b2eSZach Brown { 401004b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 4011c5c9cd4dSSage Weil } 4012c5c9cd4dSSage Weil 4013f46b5a66SChristoph Hellwig /* 4014f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4015f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4016f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4017f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4018f46b5a66SChristoph Hellwig */ 4019b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 4020f46b5a66SChristoph Hellwig { 4021496ad9aaSAl Viro struct inode *inode = file_inode(file); 40220b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 4023f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4024f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 40251ab86aedSSage Weil int ret; 4026f46b5a66SChristoph Hellwig 40271ab86aedSSage Weil ret = -EPERM; 4028df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 4029f46b5a66SChristoph Hellwig goto out; 40301ab86aedSSage Weil 40311ab86aedSSage Weil ret = -EINPROGRESS; 40321ab86aedSSage Weil if (file->private_data) 40331ab86aedSSage Weil goto out; 40349ca9ee09SSage Weil 4035b83cc969SLi Zefan ret = -EROFS; 4036b83cc969SLi Zefan if (btrfs_root_readonly(root)) 4037b83cc969SLi Zefan goto out; 4038b83cc969SLi Zefan 4039a561be71SAl Viro ret = mnt_want_write_file(file); 4040c146afadSYan Zheng if (ret) 4041c146afadSYan Zheng goto out; 4042c146afadSYan Zheng 40430b246afaSJeff Mahoney atomic_inc(&fs_info->open_ioctl_trans); 40449ca9ee09SSage Weil 4045f46b5a66SChristoph Hellwig ret = -ENOMEM; 40467a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 4047abd30bb0STsutomu Itoh if (IS_ERR(trans)) 40481ab86aedSSage Weil goto out_drop; 40491ab86aedSSage Weil 40501ab86aedSSage Weil file->private_data = trans; 40511ab86aedSSage Weil return 0; 40521ab86aedSSage Weil 40531ab86aedSSage Weil out_drop: 40540b246afaSJeff Mahoney atomic_dec(&fs_info->open_ioctl_trans); 40552a79f17eSAl Viro mnt_drop_write_file(file); 4056f46b5a66SChristoph Hellwig out: 4057f46b5a66SChristoph Hellwig return ret; 4058f46b5a66SChristoph Hellwig } 4059f46b5a66SChristoph Hellwig 40606ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40616ef5ed0dSJosef Bacik { 4062496ad9aaSAl Viro struct inode *inode = file_inode(file); 40630b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 40646ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40656ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40666ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40676ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40686ef5ed0dSJosef Bacik struct btrfs_path *path; 40696ef5ed0dSJosef Bacik struct btrfs_key location; 40706ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40716ef5ed0dSJosef Bacik u64 objectid = 0; 40726ef5ed0dSJosef Bacik u64 dir_id; 40733c04ce01SMiao Xie int ret; 40746ef5ed0dSJosef Bacik 40756ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40766ef5ed0dSJosef Bacik return -EPERM; 40776ef5ed0dSJosef Bacik 40783c04ce01SMiao Xie ret = mnt_want_write_file(file); 40793c04ce01SMiao Xie if (ret) 40803c04ce01SMiao Xie return ret; 40813c04ce01SMiao Xie 40823c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40833c04ce01SMiao Xie ret = -EFAULT; 40843c04ce01SMiao Xie goto out; 40853c04ce01SMiao Xie } 40866ef5ed0dSJosef Bacik 40876ef5ed0dSJosef Bacik if (!objectid) 40881cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40896ef5ed0dSJosef Bacik 40906ef5ed0dSJosef Bacik location.objectid = objectid; 40916ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40926ef5ed0dSJosef Bacik location.offset = (u64)-1; 40936ef5ed0dSJosef Bacik 40940b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 40953c04ce01SMiao Xie if (IS_ERR(new_root)) { 40963c04ce01SMiao Xie ret = PTR_ERR(new_root); 40973c04ce01SMiao Xie goto out; 40983c04ce01SMiao Xie } 40996ef5ed0dSJosef Bacik 41006ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 41013c04ce01SMiao Xie if (!path) { 41023c04ce01SMiao Xie ret = -ENOMEM; 41033c04ce01SMiao Xie goto out; 41043c04ce01SMiao Xie } 41056ef5ed0dSJosef Bacik path->leave_spinning = 1; 41066ef5ed0dSJosef Bacik 41076ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 410898d5dc13STsutomu Itoh if (IS_ERR(trans)) { 41096ef5ed0dSJosef Bacik btrfs_free_path(path); 41103c04ce01SMiao Xie ret = PTR_ERR(trans); 41113c04ce01SMiao Xie goto out; 41126ef5ed0dSJosef Bacik } 41136ef5ed0dSJosef Bacik 41140b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 41150b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 41166ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4117cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 41186ef5ed0dSJosef Bacik btrfs_free_path(path); 41196ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 41200b246afaSJeff Mahoney btrfs_err(fs_info, 41215d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 41223c04ce01SMiao Xie ret = -ENOENT; 41233c04ce01SMiao Xie goto out; 41246ef5ed0dSJosef Bacik } 41256ef5ed0dSJosef Bacik 41266ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 41276ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 41286ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 41296ef5ed0dSJosef Bacik btrfs_free_path(path); 41306ef5ed0dSJosef Bacik 41310b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 41326ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 41333c04ce01SMiao Xie out: 41343c04ce01SMiao Xie mnt_drop_write_file(file); 41353c04ce01SMiao Xie return ret; 41366ef5ed0dSJosef Bacik } 41376ef5ed0dSJosef Bacik 41385af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 4139bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4140bf5fc093SJosef Bacik { 4141bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4142bf5fc093SJosef Bacik 4143bf5fc093SJosef Bacik space->total_bytes = 0; 4144bf5fc093SJosef Bacik space->used_bytes = 0; 4145bf5fc093SJosef Bacik space->flags = 0; 4146bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4147bf5fc093SJosef Bacik space->flags = block_group->flags; 4148bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4149bf5fc093SJosef Bacik space->used_bytes += 4150bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4151bf5fc093SJosef Bacik } 4152bf5fc093SJosef Bacik } 4153bf5fc093SJosef Bacik 415448a3b636SEric Sandeen static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 41551406e432SJosef Bacik { 41560b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 41571406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41581406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41591406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41607fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 416113f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41621406e432SJosef Bacik struct btrfs_space_info *info; 4163bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 4164bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4165bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4166bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 4167bf5fc093SJosef Bacik int num_types = 4; 41687fde62bfSChris Mason int alloc_size; 41691406e432SJosef Bacik int ret = 0; 417051788b1bSDan Rosenberg u64 slot_count = 0; 4171bf5fc093SJosef Bacik int i, c; 41721406e432SJosef Bacik 41731406e432SJosef Bacik if (copy_from_user(&space_args, 41741406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41751406e432SJosef Bacik sizeof(space_args))) 41761406e432SJosef Bacik return -EFAULT; 41771406e432SJosef Bacik 4178bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4179bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4180bf5fc093SJosef Bacik 4181bf5fc093SJosef Bacik info = NULL; 41827fde62bfSChris Mason rcu_read_lock(); 41830b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4184bf5fc093SJosef Bacik list) { 4185bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4186bf5fc093SJosef Bacik info = tmp; 4187bf5fc093SJosef Bacik break; 4188bf5fc093SJosef Bacik } 4189bf5fc093SJosef Bacik } 41907fde62bfSChris Mason rcu_read_unlock(); 41911406e432SJosef Bacik 4192bf5fc093SJosef Bacik if (!info) 4193bf5fc093SJosef Bacik continue; 4194bf5fc093SJosef Bacik 4195bf5fc093SJosef Bacik down_read(&info->groups_sem); 4196bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4197bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4198bf5fc093SJosef Bacik slot_count++; 4199bf5fc093SJosef Bacik } 4200bf5fc093SJosef Bacik up_read(&info->groups_sem); 4201bf5fc093SJosef Bacik } 4202bf5fc093SJosef Bacik 420336523e95SDavid Sterba /* 420436523e95SDavid Sterba * Global block reserve, exported as a space_info 420536523e95SDavid Sterba */ 420636523e95SDavid Sterba slot_count++; 420736523e95SDavid Sterba 42087fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 42097fde62bfSChris Mason if (space_args.space_slots == 0) { 42107fde62bfSChris Mason space_args.total_spaces = slot_count; 42117fde62bfSChris Mason goto out; 42127fde62bfSChris Mason } 4213bf5fc093SJosef Bacik 421451788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4215bf5fc093SJosef Bacik 42167fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4217bf5fc093SJosef Bacik 42187fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 42197fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 42207fde62bfSChris Mason */ 422109cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 42227fde62bfSChris Mason return -ENOMEM; 42237fde62bfSChris Mason 42247fde62bfSChris Mason space_args.total_spaces = 0; 42258d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 42267fde62bfSChris Mason if (!dest) 42277fde62bfSChris Mason return -ENOMEM; 42287fde62bfSChris Mason dest_orig = dest; 42297fde62bfSChris Mason 42307fde62bfSChris Mason /* now we have a buffer to copy into */ 4231bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4232bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4233bf5fc093SJosef Bacik 423451788b1bSDan Rosenberg if (!slot_count) 423551788b1bSDan Rosenberg break; 423651788b1bSDan Rosenberg 4237bf5fc093SJosef Bacik info = NULL; 42381406e432SJosef Bacik rcu_read_lock(); 42390b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4240bf5fc093SJosef Bacik list) { 4241bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4242bf5fc093SJosef Bacik info = tmp; 42437fde62bfSChris Mason break; 4244bf5fc093SJosef Bacik } 4245bf5fc093SJosef Bacik } 4246bf5fc093SJosef Bacik rcu_read_unlock(); 42477fde62bfSChris Mason 4248bf5fc093SJosef Bacik if (!info) 4249bf5fc093SJosef Bacik continue; 4250bf5fc093SJosef Bacik down_read(&info->groups_sem); 4251bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4252bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 42535af3e8ccSStefan Behrens btrfs_get_block_group_info( 42545af3e8ccSStefan Behrens &info->block_groups[c], &space); 42557fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42561406e432SJosef Bacik dest++; 42571406e432SJosef Bacik space_args.total_spaces++; 425851788b1bSDan Rosenberg slot_count--; 42591406e432SJosef Bacik } 426051788b1bSDan Rosenberg if (!slot_count) 426151788b1bSDan Rosenberg break; 4262bf5fc093SJosef Bacik } 4263bf5fc093SJosef Bacik up_read(&info->groups_sem); 4264bf5fc093SJosef Bacik } 42651406e432SJosef Bacik 426636523e95SDavid Sterba /* 426736523e95SDavid Sterba * Add global block reserve 426836523e95SDavid Sterba */ 426936523e95SDavid Sterba if (slot_count) { 42700b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 427136523e95SDavid Sterba 427236523e95SDavid Sterba spin_lock(&block_rsv->lock); 427336523e95SDavid Sterba space.total_bytes = block_rsv->size; 427436523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 427536523e95SDavid Sterba spin_unlock(&block_rsv->lock); 427636523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 427736523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 427836523e95SDavid Sterba space_args.total_spaces++; 427936523e95SDavid Sterba } 428036523e95SDavid Sterba 42812eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42827fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42837fde62bfSChris Mason 42847fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42857fde62bfSChris Mason ret = -EFAULT; 42867fde62bfSChris Mason 42877fde62bfSChris Mason kfree(dest_orig); 42887fde62bfSChris Mason out: 42897fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42901406e432SJosef Bacik ret = -EFAULT; 42911406e432SJosef Bacik 42921406e432SJosef Bacik return ret; 42931406e432SJosef Bacik } 42941406e432SJosef Bacik 4295f46b5a66SChristoph Hellwig /* 4296f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4297f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4298f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4299f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4300f46b5a66SChristoph Hellwig */ 4301f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 4302f46b5a66SChristoph Hellwig { 4303496ad9aaSAl Viro struct inode *inode = file_inode(file); 4304f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4305f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4306f46b5a66SChristoph Hellwig 4307f46b5a66SChristoph Hellwig trans = file->private_data; 43081ab86aedSSage Weil if (!trans) 43091ab86aedSSage Weil return -EINVAL; 4310b214107eSChristoph Hellwig file->private_data = NULL; 43119ca9ee09SSage Weil 43121ab86aedSSage Weil btrfs_end_transaction(trans, root); 43131ab86aedSSage Weil 4314a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 43159ca9ee09SSage Weil 43162a79f17eSAl Viro mnt_drop_write_file(file); 43171ab86aedSSage Weil return 0; 4318f46b5a66SChristoph Hellwig } 4319f46b5a66SChristoph Hellwig 43209a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 43219a8c28beSMiao Xie void __user *argp) 432246204592SSage Weil { 432346204592SSage Weil struct btrfs_trans_handle *trans; 432446204592SSage Weil u64 transid; 4325db5b493aSTsutomu Itoh int ret; 432646204592SSage Weil 4327d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4328ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4329ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 433098d5dc13STsutomu Itoh return PTR_ERR(trans); 4331ff7c1d33SMiao Xie 4332ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4333ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4334ff7c1d33SMiao Xie goto out; 4335ff7c1d33SMiao Xie } 433646204592SSage Weil transid = trans->transid; 4337db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 43388b2b2d3cSTsutomu Itoh if (ret) { 43398b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 4340db5b493aSTsutomu Itoh return ret; 43418b2b2d3cSTsutomu Itoh } 4342ff7c1d33SMiao Xie out: 434346204592SSage Weil if (argp) 434446204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 434546204592SSage Weil return -EFAULT; 434646204592SSage Weil return 0; 434746204592SSage Weil } 434846204592SSage Weil 43499a8c28beSMiao Xie static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, 43509a8c28beSMiao Xie void __user *argp) 435146204592SSage Weil { 435246204592SSage Weil u64 transid; 435346204592SSage Weil 435446204592SSage Weil if (argp) { 435546204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 435646204592SSage Weil return -EFAULT; 435746204592SSage Weil } else { 435846204592SSage Weil transid = 0; /* current trans */ 435946204592SSage Weil } 436046204592SSage Weil return btrfs_wait_for_commit(root, transid); 436146204592SSage Weil } 436246204592SSage Weil 4363b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4364475f6387SJan Schmidt { 43650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4366475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4367b8e95489SMiao Xie int ret; 4368475f6387SJan Schmidt 4369475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4370475f6387SJan Schmidt return -EPERM; 4371475f6387SJan Schmidt 4372475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4373475f6387SJan Schmidt if (IS_ERR(sa)) 4374475f6387SJan Schmidt return PTR_ERR(sa); 4375475f6387SJan Schmidt 4376b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4377b8e95489SMiao Xie ret = mnt_want_write_file(file); 4378b8e95489SMiao Xie if (ret) 4379b8e95489SMiao Xie goto out; 4380b8e95489SMiao Xie } 4381b8e95489SMiao Xie 43820b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 438363a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 438463a212abSStefan Behrens 0); 4385475f6387SJan Schmidt 4386475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4387475f6387SJan Schmidt ret = -EFAULT; 4388475f6387SJan Schmidt 4389b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4390b8e95489SMiao Xie mnt_drop_write_file(file); 4391b8e95489SMiao Xie out: 4392475f6387SJan Schmidt kfree(sa); 4393475f6387SJan Schmidt return ret; 4394475f6387SJan Schmidt } 4395475f6387SJan Schmidt 4396475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 4397475f6387SJan Schmidt { 4398475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4399475f6387SJan Schmidt return -EPERM; 4400475f6387SJan Schmidt 4401aa1b8cd4SStefan Behrens return btrfs_scrub_cancel(root->fs_info); 4402475f6387SJan Schmidt } 4403475f6387SJan Schmidt 4404475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 4405475f6387SJan Schmidt void __user *arg) 4406475f6387SJan Schmidt { 4407475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4408475f6387SJan Schmidt int ret; 4409475f6387SJan Schmidt 4410475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4411475f6387SJan Schmidt return -EPERM; 4412475f6387SJan Schmidt 4413475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4414475f6387SJan Schmidt if (IS_ERR(sa)) 4415475f6387SJan Schmidt return PTR_ERR(sa); 4416475f6387SJan Schmidt 4417475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 4418475f6387SJan Schmidt 4419475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4420475f6387SJan Schmidt ret = -EFAULT; 4421475f6387SJan Schmidt 4422475f6387SJan Schmidt kfree(sa); 4423475f6387SJan Schmidt return ret; 4424475f6387SJan Schmidt } 4425475f6387SJan Schmidt 4426c11d2c23SStefan Behrens static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, 4427b27f7c0cSDavid Sterba void __user *arg) 4428c11d2c23SStefan Behrens { 4429c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4430c11d2c23SStefan Behrens int ret; 4431c11d2c23SStefan Behrens 4432c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4433c11d2c23SStefan Behrens if (IS_ERR(sa)) 4434c11d2c23SStefan Behrens return PTR_ERR(sa); 4435c11d2c23SStefan Behrens 4436b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4437b27f7c0cSDavid Sterba kfree(sa); 4438b27f7c0cSDavid Sterba return -EPERM; 4439b27f7c0cSDavid Sterba } 4440b27f7c0cSDavid Sterba 4441b27f7c0cSDavid Sterba ret = btrfs_get_dev_stats(root, sa); 4442c11d2c23SStefan Behrens 4443c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4444c11d2c23SStefan Behrens ret = -EFAULT; 4445c11d2c23SStefan Behrens 4446c11d2c23SStefan Behrens kfree(sa); 4447c11d2c23SStefan Behrens return ret; 4448c11d2c23SStefan Behrens } 4449c11d2c23SStefan Behrens 44503f6bcfbdSStefan Behrens static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) 44513f6bcfbdSStefan Behrens { 44520b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 44533f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 44543f6bcfbdSStefan Behrens int ret; 44553f6bcfbdSStefan Behrens 44563f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44573f6bcfbdSStefan Behrens return -EPERM; 44583f6bcfbdSStefan Behrens 44593f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44603f6bcfbdSStefan Behrens if (IS_ERR(p)) 44613f6bcfbdSStefan Behrens return PTR_ERR(p); 44623f6bcfbdSStefan Behrens 44633f6bcfbdSStefan Behrens switch (p->cmd) { 44643f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 44650b246afaSJeff Mahoney if (fs_info->sb->s_flags & MS_RDONLY) { 4466adfa97cbSIlya Dryomov ret = -EROFS; 4467adfa97cbSIlya Dryomov goto out; 4468adfa97cbSIlya Dryomov } 44693f6bcfbdSStefan Behrens if (atomic_xchg( 44700b246afaSJeff Mahoney &fs_info->mutually_exclusive_operation_running, 1)) { 4471e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44723f6bcfbdSStefan Behrens } else { 4473b5255456SAnand Jain ret = btrfs_dev_replace_by_ioctl(root, p); 44743f6bcfbdSStefan Behrens atomic_set( 44750b246afaSJeff Mahoney &fs_info->mutually_exclusive_operation_running, 0); 44763f6bcfbdSStefan Behrens } 44773f6bcfbdSStefan Behrens break; 44783f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44790b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 44803f6bcfbdSStefan Behrens ret = 0; 44813f6bcfbdSStefan Behrens break; 44823f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 44830b246afaSJeff Mahoney ret = btrfs_dev_replace_cancel(fs_info, p); 44843f6bcfbdSStefan Behrens break; 44853f6bcfbdSStefan Behrens default: 44863f6bcfbdSStefan Behrens ret = -EINVAL; 44873f6bcfbdSStefan Behrens break; 44883f6bcfbdSStefan Behrens } 44893f6bcfbdSStefan Behrens 44903f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 44913f6bcfbdSStefan Behrens ret = -EFAULT; 4492adfa97cbSIlya Dryomov out: 44933f6bcfbdSStefan Behrens kfree(p); 44943f6bcfbdSStefan Behrens return ret; 44953f6bcfbdSStefan Behrens } 44963f6bcfbdSStefan Behrens 4497d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4498d7728c96SJan Schmidt { 4499d7728c96SJan Schmidt int ret = 0; 4500d7728c96SJan Schmidt int i; 4501740c3d22SChris Mason u64 rel_ptr; 4502d7728c96SJan Schmidt int size; 4503806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4504d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4505d7728c96SJan Schmidt struct btrfs_path *path; 4506d7728c96SJan Schmidt 450782b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4508d7728c96SJan Schmidt return -EPERM; 4509d7728c96SJan Schmidt 4510d7728c96SJan Schmidt path = btrfs_alloc_path(); 4511d7728c96SJan Schmidt if (!path) { 4512d7728c96SJan Schmidt ret = -ENOMEM; 4513d7728c96SJan Schmidt goto out; 4514d7728c96SJan Schmidt } 4515d7728c96SJan Schmidt 4516d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4517d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4518d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4519d7728c96SJan Schmidt ipa = NULL; 4520d7728c96SJan Schmidt goto out; 4521d7728c96SJan Schmidt } 4522d7728c96SJan Schmidt 4523d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4524d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4525d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4526d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4527d7728c96SJan Schmidt ipath = NULL; 4528d7728c96SJan Schmidt goto out; 4529d7728c96SJan Schmidt } 4530d7728c96SJan Schmidt 4531d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4532d7728c96SJan Schmidt if (ret < 0) 4533d7728c96SJan Schmidt goto out; 4534d7728c96SJan Schmidt 4535d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4536745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4537745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4538740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4539d7728c96SJan Schmidt } 4540d7728c96SJan Schmidt 4541745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 4542745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 4543d7728c96SJan Schmidt if (ret) { 4544d7728c96SJan Schmidt ret = -EFAULT; 4545d7728c96SJan Schmidt goto out; 4546d7728c96SJan Schmidt } 4547d7728c96SJan Schmidt 4548d7728c96SJan Schmidt out: 4549d7728c96SJan Schmidt btrfs_free_path(path); 4550d7728c96SJan Schmidt free_ipath(ipath); 4551d7728c96SJan Schmidt kfree(ipa); 4552d7728c96SJan Schmidt 4553d7728c96SJan Schmidt return ret; 4554d7728c96SJan Schmidt } 4555d7728c96SJan Schmidt 4556d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4557d7728c96SJan Schmidt { 4558d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4559d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4560d7728c96SJan Schmidt 4561d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4562d7728c96SJan Schmidt inodes->bytes_left -= c; 4563d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4564d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4565d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4566d7728c96SJan Schmidt inodes->elem_cnt += 3; 4567d7728c96SJan Schmidt } else { 4568d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4569d7728c96SJan Schmidt inodes->bytes_left = 0; 4570d7728c96SJan Schmidt inodes->elem_missed += 3; 4571d7728c96SJan Schmidt } 4572d7728c96SJan Schmidt 4573d7728c96SJan Schmidt return 0; 4574d7728c96SJan Schmidt } 4575d7728c96SJan Schmidt 4576d7728c96SJan Schmidt static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 4577d7728c96SJan Schmidt void __user *arg) 4578d7728c96SJan Schmidt { 4579d7728c96SJan Schmidt int ret = 0; 4580d7728c96SJan Schmidt int size; 4581d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4582d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4583d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4584d7728c96SJan Schmidt 4585d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4586d7728c96SJan Schmidt return -EPERM; 4587d7728c96SJan Schmidt 4588d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 45897b9ea627SShailendra Verma if (IS_ERR(loi)) 45907b9ea627SShailendra Verma return PTR_ERR(loi); 4591d7728c96SJan Schmidt 4592d7728c96SJan Schmidt path = btrfs_alloc_path(); 4593d7728c96SJan Schmidt if (!path) { 4594d7728c96SJan Schmidt ret = -ENOMEM; 4595d7728c96SJan Schmidt goto out; 4596d7728c96SJan Schmidt } 4597d7728c96SJan Schmidt 4598ee22184bSByongho Lee size = min_t(u32, loi->size, SZ_64K); 4599d7728c96SJan Schmidt inodes = init_data_container(size); 4600d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4601d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4602d7728c96SJan Schmidt inodes = NULL; 4603d7728c96SJan Schmidt goto out; 4604d7728c96SJan Schmidt } 4605d7728c96SJan Schmidt 4606df031f07SLiu Bo ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, 4607df031f07SLiu Bo build_ino_list, inodes); 4608df031f07SLiu Bo if (ret == -EINVAL) 4609d7728c96SJan Schmidt ret = -ENOENT; 4610d7728c96SJan Schmidt if (ret < 0) 4611d7728c96SJan Schmidt goto out; 4612d7728c96SJan Schmidt 4613745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 4614745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 4615d7728c96SJan Schmidt if (ret) 4616d7728c96SJan Schmidt ret = -EFAULT; 4617d7728c96SJan Schmidt 4618d7728c96SJan Schmidt out: 4619d7728c96SJan Schmidt btrfs_free_path(path); 4620425d17a2SLiu Bo vfree(inodes); 4621d7728c96SJan Schmidt kfree(loi); 4622d7728c96SJan Schmidt 4623d7728c96SJan Schmidt return ret; 4624d7728c96SJan Schmidt } 4625d7728c96SJan Schmidt 462619a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4627c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4628c9e9f97bSIlya Dryomov { 4629c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4630c9e9f97bSIlya Dryomov 4631c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4632c9e9f97bSIlya Dryomov 4633837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 4634837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4635837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4636837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4637a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4638a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4639837d5b6eSIlya Dryomov 4640c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4641c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4642c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 464319a39dceSIlya Dryomov 464419a39dceSIlya Dryomov if (lock) { 464519a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 464619a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 464719a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 464819a39dceSIlya Dryomov } else { 464919a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 465019a39dceSIlya Dryomov } 4651c9e9f97bSIlya Dryomov } 4652c9e9f97bSIlya Dryomov 46539ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4654c9e9f97bSIlya Dryomov { 4655496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4656c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4657c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4658c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4659ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4660c9e9f97bSIlya Dryomov int ret; 4661c9e9f97bSIlya Dryomov 4662c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4663c9e9f97bSIlya Dryomov return -EPERM; 4664c9e9f97bSIlya Dryomov 4665e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46669ba1f6e4SLiu Bo if (ret) 46679ba1f6e4SLiu Bo return ret; 46689ba1f6e4SLiu Bo 4669ed0fb78fSIlya Dryomov again: 4670ed0fb78fSIlya Dryomov if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 4671c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 4672c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4673ed0fb78fSIlya Dryomov need_unlock = true; 4674ed0fb78fSIlya Dryomov goto locked; 4675ed0fb78fSIlya Dryomov } 4676ed0fb78fSIlya Dryomov 4677ed0fb78fSIlya Dryomov /* 467801327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4679ed0fb78fSIlya Dryomov * (1) some other op is running 4680ed0fb78fSIlya Dryomov * (2) balance is running 4681ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4682ed0fb78fSIlya Dryomov */ 4683ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4684ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4685ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 4686ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 4687ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4688ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 4689ed0fb78fSIlya Dryomov goto again; 4690ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4691ed0fb78fSIlya Dryomov 4692ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 4693ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 4694ed0fb78fSIlya Dryomov /* this is (3) */ 4695ed0fb78fSIlya Dryomov need_unlock = false; 4696ed0fb78fSIlya Dryomov goto locked; 4697ed0fb78fSIlya Dryomov } 4698ed0fb78fSIlya Dryomov 4699ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4700ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4701ed0fb78fSIlya Dryomov goto again; 4702ed0fb78fSIlya Dryomov } else { 4703ed0fb78fSIlya Dryomov /* this is (2) */ 4704ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4705ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4706ed0fb78fSIlya Dryomov goto out; 4707ed0fb78fSIlya Dryomov } 4708ed0fb78fSIlya Dryomov } else { 4709ed0fb78fSIlya Dryomov /* this is (1) */ 4710ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4711e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4712ed0fb78fSIlya Dryomov goto out; 4713ed0fb78fSIlya Dryomov } 4714ed0fb78fSIlya Dryomov 4715ed0fb78fSIlya Dryomov locked: 4716ed0fb78fSIlya Dryomov BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); 4717c9e9f97bSIlya Dryomov 4718c9e9f97bSIlya Dryomov if (arg) { 4719c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4720c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4721c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4722ed0fb78fSIlya Dryomov goto out_unlock; 4723c9e9f97bSIlya Dryomov } 4724de322263SIlya Dryomov 4725de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4726de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4727de322263SIlya Dryomov ret = -ENOTCONN; 4728de322263SIlya Dryomov goto out_bargs; 4729de322263SIlya Dryomov } 4730de322263SIlya Dryomov 4731de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4732de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4733de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4734de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4735de322263SIlya Dryomov 4736de322263SIlya Dryomov goto do_balance; 4737de322263SIlya Dryomov } 4738c9e9f97bSIlya Dryomov } else { 4739c9e9f97bSIlya Dryomov bargs = NULL; 4740c9e9f97bSIlya Dryomov } 4741c9e9f97bSIlya Dryomov 4742ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4743837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4744837d5b6eSIlya Dryomov goto out_bargs; 4745837d5b6eSIlya Dryomov } 4746837d5b6eSIlya Dryomov 47478d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4748c9e9f97bSIlya Dryomov if (!bctl) { 4749c9e9f97bSIlya Dryomov ret = -ENOMEM; 4750c9e9f97bSIlya Dryomov goto out_bargs; 4751c9e9f97bSIlya Dryomov } 4752c9e9f97bSIlya Dryomov 4753c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 4754c9e9f97bSIlya Dryomov if (arg) { 4755c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4756c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4757c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4758c9e9f97bSIlya Dryomov 4759c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4760f43ffb60SIlya Dryomov } else { 4761f43ffb60SIlya Dryomov /* balance everything - no filters */ 4762f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4763c9e9f97bSIlya Dryomov } 4764c9e9f97bSIlya Dryomov 47658eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47668eb93459SDavid Sterba ret = -EINVAL; 47670f89abf5SChristian Engelmayer goto out_bctl; 47688eb93459SDavid Sterba } 47698eb93459SDavid Sterba 4770de322263SIlya Dryomov do_balance: 4771c9e9f97bSIlya Dryomov /* 4772ed0fb78fSIlya Dryomov * Ownership of bctl and mutually_exclusive_operation_running 4773ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4774ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 4775ed0fb78fSIlya Dryomov * free_fs_info. mutually_exclusive_operation_running is 4776ed0fb78fSIlya Dryomov * cleared in __cancel_balance. 4777c9e9f97bSIlya Dryomov */ 4778ed0fb78fSIlya Dryomov need_unlock = false; 4779ed0fb78fSIlya Dryomov 4780ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 47810f89abf5SChristian Engelmayer bctl = NULL; 4782ed0fb78fSIlya Dryomov 4783c9e9f97bSIlya Dryomov if (arg) { 4784c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4785c9e9f97bSIlya Dryomov ret = -EFAULT; 4786c9e9f97bSIlya Dryomov } 4787c9e9f97bSIlya Dryomov 47880f89abf5SChristian Engelmayer out_bctl: 47890f89abf5SChristian Engelmayer kfree(bctl); 4790c9e9f97bSIlya Dryomov out_bargs: 4791c9e9f97bSIlya Dryomov kfree(bargs); 4792ed0fb78fSIlya Dryomov out_unlock: 4793c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4794c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4795ed0fb78fSIlya Dryomov if (need_unlock) 4796ed0fb78fSIlya Dryomov atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 4797ed0fb78fSIlya Dryomov out: 4798e54bfa31SLiu Bo mnt_drop_write_file(file); 4799c9e9f97bSIlya Dryomov return ret; 4800c9e9f97bSIlya Dryomov } 4801c9e9f97bSIlya Dryomov 4802837d5b6eSIlya Dryomov static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 4803837d5b6eSIlya Dryomov { 48040b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 48050b246afaSJeff Mahoney 4806837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4807837d5b6eSIlya Dryomov return -EPERM; 4808837d5b6eSIlya Dryomov 4809837d5b6eSIlya Dryomov switch (cmd) { 4810837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 48110b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4812a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 48130b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4814837d5b6eSIlya Dryomov } 4815837d5b6eSIlya Dryomov 4816837d5b6eSIlya Dryomov return -EINVAL; 4817837d5b6eSIlya Dryomov } 4818837d5b6eSIlya Dryomov 481919a39dceSIlya Dryomov static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 482019a39dceSIlya Dryomov void __user *arg) 482119a39dceSIlya Dryomov { 482219a39dceSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 482319a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 482419a39dceSIlya Dryomov int ret = 0; 482519a39dceSIlya Dryomov 482619a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 482719a39dceSIlya Dryomov return -EPERM; 482819a39dceSIlya Dryomov 482919a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 483019a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 483119a39dceSIlya Dryomov ret = -ENOTCONN; 483219a39dceSIlya Dryomov goto out; 483319a39dceSIlya Dryomov } 483419a39dceSIlya Dryomov 48358d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 483619a39dceSIlya Dryomov if (!bargs) { 483719a39dceSIlya Dryomov ret = -ENOMEM; 483819a39dceSIlya Dryomov goto out; 483919a39dceSIlya Dryomov } 484019a39dceSIlya Dryomov 484119a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 484219a39dceSIlya Dryomov 484319a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 484419a39dceSIlya Dryomov ret = -EFAULT; 484519a39dceSIlya Dryomov 484619a39dceSIlya Dryomov kfree(bargs); 484719a39dceSIlya Dryomov out: 484819a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 484919a39dceSIlya Dryomov return ret; 485019a39dceSIlya Dryomov } 485119a39dceSIlya Dryomov 4852905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 48535d13a37bSArne Jansen { 48540b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48550b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48565d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 48575d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 48585d13a37bSArne Jansen int ret; 48595d13a37bSArne Jansen int err; 48605d13a37bSArne Jansen 48615d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48625d13a37bSArne Jansen return -EPERM; 48635d13a37bSArne Jansen 4864905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4865905b0ddaSMiao Xie if (ret) 4866905b0ddaSMiao Xie return ret; 48675d13a37bSArne Jansen 48685d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4869905b0ddaSMiao Xie if (IS_ERR(sa)) { 4870905b0ddaSMiao Xie ret = PTR_ERR(sa); 4871905b0ddaSMiao Xie goto drop_write; 4872905b0ddaSMiao Xie } 48735d13a37bSArne Jansen 48740b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48750b246afaSJeff Mahoney trans = btrfs_start_transaction(fs_info->tree_root, 2); 48765d13a37bSArne Jansen if (IS_ERR(trans)) { 48775d13a37bSArne Jansen ret = PTR_ERR(trans); 48785d13a37bSArne Jansen goto out; 48795d13a37bSArne Jansen } 48805d13a37bSArne Jansen 48815d13a37bSArne Jansen switch (sa->cmd) { 48825d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 48830b246afaSJeff Mahoney ret = btrfs_quota_enable(trans, fs_info); 48845d13a37bSArne Jansen break; 48855d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 48860b246afaSJeff Mahoney ret = btrfs_quota_disable(trans, fs_info); 48875d13a37bSArne Jansen break; 48885d13a37bSArne Jansen default: 48895d13a37bSArne Jansen ret = -EINVAL; 48905d13a37bSArne Jansen break; 48915d13a37bSArne Jansen } 48925d13a37bSArne Jansen 48930b246afaSJeff Mahoney err = btrfs_commit_transaction(trans, fs_info->tree_root); 48945d13a37bSArne Jansen if (err && !ret) 48955d13a37bSArne Jansen ret = err; 48965d13a37bSArne Jansen out: 48975d13a37bSArne Jansen kfree(sa); 48980b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4899905b0ddaSMiao Xie drop_write: 4900905b0ddaSMiao Xie mnt_drop_write_file(file); 49015d13a37bSArne Jansen return ret; 49025d13a37bSArne Jansen } 49035d13a37bSArne Jansen 4904905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 49055d13a37bSArne Jansen { 49060b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49070b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49080b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49095d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 49105d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49115d13a37bSArne Jansen int ret; 49125d13a37bSArne Jansen int err; 49135d13a37bSArne Jansen 49145d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49155d13a37bSArne Jansen return -EPERM; 49165d13a37bSArne Jansen 4917905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4918905b0ddaSMiao Xie if (ret) 4919905b0ddaSMiao Xie return ret; 49205d13a37bSArne Jansen 49215d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4922905b0ddaSMiao Xie if (IS_ERR(sa)) { 4923905b0ddaSMiao Xie ret = PTR_ERR(sa); 4924905b0ddaSMiao Xie goto drop_write; 4925905b0ddaSMiao Xie } 49265d13a37bSArne Jansen 49275d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49285d13a37bSArne Jansen if (IS_ERR(trans)) { 49295d13a37bSArne Jansen ret = PTR_ERR(trans); 49305d13a37bSArne Jansen goto out; 49315d13a37bSArne Jansen } 49325d13a37bSArne Jansen 49335d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49345d13a37bSArne Jansen if (sa->assign) { 49350b246afaSJeff Mahoney ret = btrfs_add_qgroup_relation(trans, fs_info, 49365d13a37bSArne Jansen sa->src, sa->dst); 49375d13a37bSArne Jansen } else { 49380b246afaSJeff Mahoney ret = btrfs_del_qgroup_relation(trans, fs_info, 49395d13a37bSArne Jansen sa->src, sa->dst); 49405d13a37bSArne Jansen } 49415d13a37bSArne Jansen 4942e082f563SQu Wenruo /* update qgroup status and info */ 49430b246afaSJeff Mahoney err = btrfs_run_qgroups(trans, fs_info); 4944e082f563SQu Wenruo if (err < 0) 49450b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4946ad8403dfSAnand Jain "failed to update qgroup status and info"); 49475d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 49485d13a37bSArne Jansen if (err && !ret) 49495d13a37bSArne Jansen ret = err; 49505d13a37bSArne Jansen 49515d13a37bSArne Jansen out: 49525d13a37bSArne Jansen kfree(sa); 4953905b0ddaSMiao Xie drop_write: 4954905b0ddaSMiao Xie mnt_drop_write_file(file); 49555d13a37bSArne Jansen return ret; 49565d13a37bSArne Jansen } 49575d13a37bSArne Jansen 4958905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 49595d13a37bSArne Jansen { 49600b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49610b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49620b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49635d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49645d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49655d13a37bSArne Jansen int ret; 49665d13a37bSArne Jansen int err; 49675d13a37bSArne Jansen 49685d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49695d13a37bSArne Jansen return -EPERM; 49705d13a37bSArne Jansen 4971905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4972905b0ddaSMiao Xie if (ret) 4973905b0ddaSMiao Xie return ret; 49745d13a37bSArne Jansen 49755d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4976905b0ddaSMiao Xie if (IS_ERR(sa)) { 4977905b0ddaSMiao Xie ret = PTR_ERR(sa); 4978905b0ddaSMiao Xie goto drop_write; 4979905b0ddaSMiao Xie } 49805d13a37bSArne Jansen 4981d86e56cfSMiao Xie if (!sa->qgroupid) { 4982d86e56cfSMiao Xie ret = -EINVAL; 4983d86e56cfSMiao Xie goto out; 4984d86e56cfSMiao Xie } 4985d86e56cfSMiao Xie 49865d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49875d13a37bSArne Jansen if (IS_ERR(trans)) { 49885d13a37bSArne Jansen ret = PTR_ERR(trans); 49895d13a37bSArne Jansen goto out; 49905d13a37bSArne Jansen } 49915d13a37bSArne Jansen 49925d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49935d13a37bSArne Jansen if (sa->create) { 49940b246afaSJeff Mahoney ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); 49955d13a37bSArne Jansen } else { 49960b246afaSJeff Mahoney ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); 49975d13a37bSArne Jansen } 49985d13a37bSArne Jansen 49995d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 50005d13a37bSArne Jansen if (err && !ret) 50015d13a37bSArne Jansen ret = err; 50025d13a37bSArne Jansen 50035d13a37bSArne Jansen out: 50045d13a37bSArne Jansen kfree(sa); 5005905b0ddaSMiao Xie drop_write: 5006905b0ddaSMiao Xie mnt_drop_write_file(file); 50075d13a37bSArne Jansen return ret; 50085d13a37bSArne Jansen } 50095d13a37bSArne Jansen 5010905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 50115d13a37bSArne Jansen { 50120b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50130b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50140b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 50155d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 50165d13a37bSArne Jansen struct btrfs_trans_handle *trans; 50175d13a37bSArne Jansen int ret; 50185d13a37bSArne Jansen int err; 50195d13a37bSArne Jansen u64 qgroupid; 50205d13a37bSArne Jansen 50215d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 50225d13a37bSArne Jansen return -EPERM; 50235d13a37bSArne Jansen 5024905b0ddaSMiao Xie ret = mnt_want_write_file(file); 5025905b0ddaSMiao Xie if (ret) 5026905b0ddaSMiao Xie return ret; 50275d13a37bSArne Jansen 50285d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5029905b0ddaSMiao Xie if (IS_ERR(sa)) { 5030905b0ddaSMiao Xie ret = PTR_ERR(sa); 5031905b0ddaSMiao Xie goto drop_write; 5032905b0ddaSMiao Xie } 50335d13a37bSArne Jansen 50345d13a37bSArne Jansen trans = btrfs_join_transaction(root); 50355d13a37bSArne Jansen if (IS_ERR(trans)) { 50365d13a37bSArne Jansen ret = PTR_ERR(trans); 50375d13a37bSArne Jansen goto out; 50385d13a37bSArne Jansen } 50395d13a37bSArne Jansen 50405d13a37bSArne Jansen qgroupid = sa->qgroupid; 50415d13a37bSArne Jansen if (!qgroupid) { 50425d13a37bSArne Jansen /* take the current subvol as qgroup */ 50435d13a37bSArne Jansen qgroupid = root->root_key.objectid; 50445d13a37bSArne Jansen } 50455d13a37bSArne Jansen 50465d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 50470b246afaSJeff Mahoney ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); 50485d13a37bSArne Jansen 50495d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 50505d13a37bSArne Jansen if (err && !ret) 50515d13a37bSArne Jansen ret = err; 50525d13a37bSArne Jansen 50535d13a37bSArne Jansen out: 50545d13a37bSArne Jansen kfree(sa); 5055905b0ddaSMiao Xie drop_write: 5056905b0ddaSMiao Xie mnt_drop_write_file(file); 50575d13a37bSArne Jansen return ret; 50585d13a37bSArne Jansen } 50595d13a37bSArne Jansen 50602f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 50612f232036SJan Schmidt { 50620b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50630b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50642f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50652f232036SJan Schmidt int ret; 50662f232036SJan Schmidt 50672f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50682f232036SJan Schmidt return -EPERM; 50692f232036SJan Schmidt 50702f232036SJan Schmidt ret = mnt_want_write_file(file); 50712f232036SJan Schmidt if (ret) 50722f232036SJan Schmidt return ret; 50732f232036SJan Schmidt 50742f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50752f232036SJan Schmidt if (IS_ERR(qsa)) { 50762f232036SJan Schmidt ret = PTR_ERR(qsa); 50772f232036SJan Schmidt goto drop_write; 50782f232036SJan Schmidt } 50792f232036SJan Schmidt 50802f232036SJan Schmidt if (qsa->flags) { 50812f232036SJan Schmidt ret = -EINVAL; 50822f232036SJan Schmidt goto out; 50832f232036SJan Schmidt } 50842f232036SJan Schmidt 50850b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 50862f232036SJan Schmidt 50872f232036SJan Schmidt out: 50882f232036SJan Schmidt kfree(qsa); 50892f232036SJan Schmidt drop_write: 50902f232036SJan Schmidt mnt_drop_write_file(file); 50912f232036SJan Schmidt return ret; 50922f232036SJan Schmidt } 50932f232036SJan Schmidt 50942f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50952f232036SJan Schmidt { 50960b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50970b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50982f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50992f232036SJan Schmidt int ret = 0; 51002f232036SJan Schmidt 51012f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 51022f232036SJan Schmidt return -EPERM; 51032f232036SJan Schmidt 51048d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 51052f232036SJan Schmidt if (!qsa) 51062f232036SJan Schmidt return -ENOMEM; 51072f232036SJan Schmidt 51080b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 51092f232036SJan Schmidt qsa->flags = 1; 51100b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 51112f232036SJan Schmidt } 51122f232036SJan Schmidt 51132f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 51142f232036SJan Schmidt ret = -EFAULT; 51152f232036SJan Schmidt 51162f232036SJan Schmidt kfree(qsa); 51172f232036SJan Schmidt return ret; 51182f232036SJan Schmidt } 51192f232036SJan Schmidt 512057254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 512157254b6eSJan Schmidt { 51220b246afaSJeff Mahoney struct inode *inode = file_inode(file); 51230b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 512457254b6eSJan Schmidt 512557254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 512657254b6eSJan Schmidt return -EPERM; 512757254b6eSJan Schmidt 51280b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 512957254b6eSJan Schmidt } 513057254b6eSJan Schmidt 5131abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5132abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 51338ea05e3aSAlexander Block { 5134496ad9aaSAl Viro struct inode *inode = file_inode(file); 51350b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51368ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 51378ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 51388ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 5139c2050a45SDeepa Dinamani struct timespec ct = current_time(inode); 51408ea05e3aSAlexander Block int ret = 0; 5141dd5f9615SStefan Behrens int received_uuid_changed; 51428ea05e3aSAlexander Block 5143bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5144bd60ea0fSDavid Sterba return -EPERM; 5145bd60ea0fSDavid Sterba 51468ea05e3aSAlexander Block ret = mnt_want_write_file(file); 51478ea05e3aSAlexander Block if (ret < 0) 51488ea05e3aSAlexander Block return ret; 51498ea05e3aSAlexander Block 51500b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51518ea05e3aSAlexander Block 51528ea05e3aSAlexander Block if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 51538ea05e3aSAlexander Block ret = -EINVAL; 51548ea05e3aSAlexander Block goto out; 51558ea05e3aSAlexander Block } 51568ea05e3aSAlexander Block 51578ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 51588ea05e3aSAlexander Block ret = -EROFS; 51598ea05e3aSAlexander Block goto out; 51608ea05e3aSAlexander Block } 51618ea05e3aSAlexander Block 5162dd5f9615SStefan Behrens /* 5163dd5f9615SStefan Behrens * 1 - root item 5164dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5165dd5f9615SStefan Behrens */ 5166dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51678ea05e3aSAlexander Block if (IS_ERR(trans)) { 51688ea05e3aSAlexander Block ret = PTR_ERR(trans); 51698ea05e3aSAlexander Block trans = NULL; 51708ea05e3aSAlexander Block goto out; 51718ea05e3aSAlexander Block } 51728ea05e3aSAlexander Block 51738ea05e3aSAlexander Block sa->rtransid = trans->transid; 51748ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51758ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51768ea05e3aSAlexander Block 5177dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5178dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5179dd5f9615SStefan Behrens if (received_uuid_changed && 5180dd5f9615SStefan Behrens !btrfs_is_empty_uuid(root_item->received_uuid)) 51810b246afaSJeff Mahoney btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid, 5182dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5183dd5f9615SStefan Behrens root->root_key.objectid); 51848ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51858ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51868ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51873cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51883cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51893cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51903cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51918ea05e3aSAlexander Block 51920b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 51938ea05e3aSAlexander Block &root->root_key, &root->root_item); 51948ea05e3aSAlexander Block if (ret < 0) { 51958ea05e3aSAlexander Block btrfs_end_transaction(trans, root); 51968ea05e3aSAlexander Block goto out; 5197dd5f9615SStefan Behrens } 5198dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 51990b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid, 5200dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5201dd5f9615SStefan Behrens root->root_key.objectid); 5202dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 520366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5204dd5f9615SStefan Behrens goto out; 5205dd5f9615SStefan Behrens } 5206dd5f9615SStefan Behrens } 52078ea05e3aSAlexander Block ret = btrfs_commit_transaction(trans, root); 5208dd5f9615SStefan Behrens if (ret < 0) { 520966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 52108ea05e3aSAlexander Block goto out; 52118ea05e3aSAlexander Block } 52128ea05e3aSAlexander Block 5213abccd00fSHugo Mills out: 52140b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5215abccd00fSHugo Mills mnt_drop_write_file(file); 5216abccd00fSHugo Mills return ret; 5217abccd00fSHugo Mills } 5218abccd00fSHugo Mills 5219abccd00fSHugo Mills #ifdef CONFIG_64BIT 5220abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5221abccd00fSHugo Mills void __user *arg) 5222abccd00fSHugo Mills { 5223abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5224abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5225abccd00fSHugo Mills int ret = 0; 5226abccd00fSHugo Mills 5227abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 52287b9ea627SShailendra Verma if (IS_ERR(args32)) 52297b9ea627SShailendra Verma return PTR_ERR(args32); 5230abccd00fSHugo Mills 52318d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 523284dbeb87SDan Carpenter if (!args64) { 523384dbeb87SDan Carpenter ret = -ENOMEM; 5234abccd00fSHugo Mills goto out; 5235abccd00fSHugo Mills } 5236abccd00fSHugo Mills 5237abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5238abccd00fSHugo Mills args64->stransid = args32->stransid; 5239abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5240abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5241abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5242abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5243abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5244abccd00fSHugo Mills args64->flags = args32->flags; 5245abccd00fSHugo Mills 5246abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5247abccd00fSHugo Mills if (ret) 5248abccd00fSHugo Mills goto out; 5249abccd00fSHugo Mills 5250abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5251abccd00fSHugo Mills args32->stransid = args64->stransid; 5252abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5253abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5254abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5255abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5256abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5257abccd00fSHugo Mills args32->flags = args64->flags; 5258abccd00fSHugo Mills 5259abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5260abccd00fSHugo Mills if (ret) 5261abccd00fSHugo Mills ret = -EFAULT; 5262abccd00fSHugo Mills 5263abccd00fSHugo Mills out: 5264abccd00fSHugo Mills kfree(args32); 5265abccd00fSHugo Mills kfree(args64); 5266abccd00fSHugo Mills return ret; 5267abccd00fSHugo Mills } 5268abccd00fSHugo Mills #endif 5269abccd00fSHugo Mills 5270abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5271abccd00fSHugo Mills void __user *arg) 5272abccd00fSHugo Mills { 5273abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5274abccd00fSHugo Mills int ret = 0; 5275abccd00fSHugo Mills 5276abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 52777b9ea627SShailendra Verma if (IS_ERR(sa)) 52787b9ea627SShailendra Verma return PTR_ERR(sa); 5279abccd00fSHugo Mills 5280abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5281abccd00fSHugo Mills 5282abccd00fSHugo Mills if (ret) 5283abccd00fSHugo Mills goto out; 5284abccd00fSHugo Mills 52858ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52868ea05e3aSAlexander Block if (ret) 52878ea05e3aSAlexander Block ret = -EFAULT; 52888ea05e3aSAlexander Block 52898ea05e3aSAlexander Block out: 52908ea05e3aSAlexander Block kfree(sa); 52918ea05e3aSAlexander Block return ret; 52928ea05e3aSAlexander Block } 52938ea05e3aSAlexander Block 5294867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5295867ab667Sjeff.liu { 52960b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52970b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5298a1b83ac5SAnand Jain size_t len; 5299867ab667Sjeff.liu int ret; 5300a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5301a1b83ac5SAnand Jain 53020b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 53030b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 53040b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5305a1b83ac5SAnand Jain 5306a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5307867ab667Sjeff.liu 5308867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 53090b246afaSJeff Mahoney btrfs_warn(fs_info, 53100b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 53110b246afaSJeff Mahoney --len); 5312867ab667Sjeff.liu } 5313867ab667Sjeff.liu 5314867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5315867ab667Sjeff.liu 5316867ab667Sjeff.liu return ret ? -EFAULT : 0; 5317867ab667Sjeff.liu } 5318867ab667Sjeff.liu 5319a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5320a8bfd4abSjeff.liu { 53210b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53220b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53230b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 53240b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5325a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5326a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5327a8bfd4abSjeff.liu int ret; 5328a8bfd4abSjeff.liu 5329a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5330a8bfd4abSjeff.liu return -EPERM; 5331a8bfd4abSjeff.liu 5332a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5333a8bfd4abSjeff.liu return -EFAULT; 5334a8bfd4abSjeff.liu 5335a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 53360b246afaSJeff Mahoney btrfs_err(fs_info, 53375d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5338a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5339a8bfd4abSjeff.liu return -EINVAL; 5340a8bfd4abSjeff.liu } 5341a8bfd4abSjeff.liu 5342a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5343a8bfd4abSjeff.liu if (ret) 5344a8bfd4abSjeff.liu return ret; 5345a8bfd4abSjeff.liu 5346a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5347a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5348a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5349a8bfd4abSjeff.liu goto out_unlock; 5350a8bfd4abSjeff.liu } 5351a8bfd4abSjeff.liu 53520b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5353a8bfd4abSjeff.liu strcpy(super_block->label, label); 53540b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5355d0270acaSJeff Mahoney ret = btrfs_commit_transaction(trans, root); 5356a8bfd4abSjeff.liu 5357a8bfd4abSjeff.liu out_unlock: 5358a8bfd4abSjeff.liu mnt_drop_write_file(file); 5359a8bfd4abSjeff.liu return ret; 5360a8bfd4abSjeff.liu } 5361a8bfd4abSjeff.liu 53622eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 53632eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 53642eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53652eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53662eaa055fSJeff Mahoney 5367d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53682eaa055fSJeff Mahoney { 53694d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53702eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53712eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53722eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53732eaa055fSJeff Mahoney }; 53742eaa055fSJeff Mahoney 53752eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53762eaa055fSJeff Mahoney return -EFAULT; 53772eaa055fSJeff Mahoney 53782eaa055fSJeff Mahoney return 0; 53792eaa055fSJeff Mahoney } 53802eaa055fSJeff Mahoney 53812eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53822eaa055fSJeff Mahoney { 53830b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53840b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53850b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53862eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53872eaa055fSJeff Mahoney 53882eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53892eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53902eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53912eaa055fSJeff Mahoney 53922eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53932eaa055fSJeff Mahoney return -EFAULT; 53942eaa055fSJeff Mahoney 53952eaa055fSJeff Mahoney return 0; 53962eaa055fSJeff Mahoney } 53972eaa055fSJeff Mahoney 53983b02a68aSJeff Mahoney static int check_feature_bits(struct btrfs_root *root, 53993b02a68aSJeff Mahoney enum btrfs_feature_set set, 54002eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 54012eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 54022eaa055fSJeff Mahoney { 54030b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 54043b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 54053b02a68aSJeff Mahoney char *names; 54062eaa055fSJeff Mahoney u64 disallowed, unsupported; 54072eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 54082eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 54092eaa055fSJeff Mahoney 54102eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 54112eaa055fSJeff Mahoney if (unsupported) { 54123b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 54133b02a68aSJeff Mahoney if (names) { 54140b246afaSJeff Mahoney btrfs_warn(fs_info, 54153b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 54163b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54173b02a68aSJeff Mahoney kfree(names); 54183b02a68aSJeff Mahoney } else 54190b246afaSJeff Mahoney btrfs_warn(fs_info, 54202eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 54212eaa055fSJeff Mahoney type, unsupported); 54222eaa055fSJeff Mahoney return -EOPNOTSUPP; 54232eaa055fSJeff Mahoney } 54242eaa055fSJeff Mahoney 54252eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 54262eaa055fSJeff Mahoney if (disallowed) { 54273b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54283b02a68aSJeff Mahoney if (names) { 54290b246afaSJeff Mahoney btrfs_warn(fs_info, 54303b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 54313b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54323b02a68aSJeff Mahoney kfree(names); 54333b02a68aSJeff Mahoney } else 54340b246afaSJeff Mahoney btrfs_warn(fs_info, 54352eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 54362eaa055fSJeff Mahoney type, disallowed); 54372eaa055fSJeff Mahoney return -EPERM; 54382eaa055fSJeff Mahoney } 54392eaa055fSJeff Mahoney 54402eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 54412eaa055fSJeff Mahoney if (disallowed) { 54423b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54433b02a68aSJeff Mahoney if (names) { 54440b246afaSJeff Mahoney btrfs_warn(fs_info, 54453b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 54463b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54473b02a68aSJeff Mahoney kfree(names); 54483b02a68aSJeff Mahoney } else 54490b246afaSJeff Mahoney btrfs_warn(fs_info, 54502eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 54512eaa055fSJeff Mahoney type, disallowed); 54522eaa055fSJeff Mahoney return -EPERM; 54532eaa055fSJeff Mahoney } 54542eaa055fSJeff Mahoney 54552eaa055fSJeff Mahoney return 0; 54562eaa055fSJeff Mahoney } 54572eaa055fSJeff Mahoney 54582eaa055fSJeff Mahoney #define check_feature(root, change_mask, flags, mask_base) \ 54593b02a68aSJeff Mahoney check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ 54602eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 54612eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 54622eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 54632eaa055fSJeff Mahoney 54642eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 54652eaa055fSJeff Mahoney { 54660b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54670b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54680b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54690b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 54702eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54712eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54722eaa055fSJeff Mahoney u64 newflags; 54732eaa055fSJeff Mahoney int ret; 54742eaa055fSJeff Mahoney 54752eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54762eaa055fSJeff Mahoney return -EPERM; 54772eaa055fSJeff Mahoney 54782eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54792eaa055fSJeff Mahoney return -EFAULT; 54802eaa055fSJeff Mahoney 54812eaa055fSJeff Mahoney /* Nothing to do */ 54822eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54832eaa055fSJeff Mahoney !flags[0].incompat_flags) 54842eaa055fSJeff Mahoney return 0; 54852eaa055fSJeff Mahoney 54862eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_flags, 54872eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54882eaa055fSJeff Mahoney if (ret) 54892eaa055fSJeff Mahoney return ret; 54902eaa055fSJeff Mahoney 54912eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_ro_flags, 54922eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54932eaa055fSJeff Mahoney if (ret) 54942eaa055fSJeff Mahoney return ret; 54952eaa055fSJeff Mahoney 54962eaa055fSJeff Mahoney ret = check_feature(root, flags[0].incompat_flags, 54972eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54982eaa055fSJeff Mahoney if (ret) 54992eaa055fSJeff Mahoney return ret; 55002eaa055fSJeff Mahoney 55017ab19625SDavid Sterba ret = mnt_want_write_file(file); 55027ab19625SDavid Sterba if (ret) 55037ab19625SDavid Sterba return ret; 55047ab19625SDavid Sterba 55058051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 55067ab19625SDavid Sterba if (IS_ERR(trans)) { 55077ab19625SDavid Sterba ret = PTR_ERR(trans); 55087ab19625SDavid Sterba goto out_drop_write; 55097ab19625SDavid Sterba } 55102eaa055fSJeff Mahoney 55110b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 55122eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 55132eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 55142eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 55152eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 55162eaa055fSJeff Mahoney 55172eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 55182eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 55192eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 55202eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 55212eaa055fSJeff Mahoney 55222eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 55232eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 55242eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 55252eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 55260b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 55272eaa055fSJeff Mahoney 55287ab19625SDavid Sterba ret = btrfs_commit_transaction(trans, root); 55297ab19625SDavid Sterba out_drop_write: 55307ab19625SDavid Sterba mnt_drop_write_file(file); 55317ab19625SDavid Sterba 55327ab19625SDavid Sterba return ret; 55332eaa055fSJeff Mahoney } 55342eaa055fSJeff Mahoney 5535f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5536f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5537f46b5a66SChristoph Hellwig { 55380b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55400b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55414bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5542f46b5a66SChristoph Hellwig 5543f46b5a66SChristoph Hellwig switch (cmd) { 55446cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 55456cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 55466cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 55476cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 55486cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 55496cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5550f7039b1dSLi Dongyang case FITRIM: 5551f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5552f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5553fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5554fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5555fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 55563de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5557fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 55586f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 55596f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 556076dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 556176dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 55620caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 55630caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 55640caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 55650caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55666ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55676ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5568f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55691e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55701e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55711e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5572f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5573198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5574f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55754bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 5576f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5577da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55786b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55796b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5580475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 5581475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 5582475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 5583475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 5584f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55859ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5586f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 5587f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 5588f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 5589f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 5590ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5591ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5592cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5593cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5594ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5595ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5596d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5597d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5598d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5599d7728c96SJan Schmidt return btrfs_ioctl_logical_to_ino(root, argp); 56001406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 56011406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 56029b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 56039b199859SFilipe David Borba Manana int ret; 56049b199859SFilipe David Borba Manana 56050b246afaSJeff Mahoney ret = btrfs_start_delalloc_roots(fs_info, 0, -1); 56069b199859SFilipe David Borba Manana if (ret) 56079b199859SFilipe David Borba Manana return ret; 56080b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 56092fad4e83SDavid Sterba /* 56102fad4e83SDavid Sterba * The transaction thread may want to do more work, 561101327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 56122fad4e83SDavid Sterba * processing uncleaned subvols. 56132fad4e83SDavid Sterba */ 56140b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 56159b199859SFilipe David Borba Manana return ret; 56169b199859SFilipe David Borba Manana } 561746204592SSage Weil case BTRFS_IOC_START_SYNC: 56189a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 561946204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 56209a8c28beSMiao Xie return btrfs_ioctl_wait_sync(root, argp); 5621475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5622b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5623475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 5624475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 5625475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 5626475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 5627c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 56289ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5629837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 5630837d5b6eSIlya Dryomov return btrfs_ioctl_balance_ctl(root, arg); 563119a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 563219a39dceSIlya Dryomov return btrfs_ioctl_balance_progress(root, argp); 56338ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 56348ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5635abccd00fSHugo Mills #ifdef CONFIG_64BIT 5636abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5637abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5638abccd00fSHugo Mills #endif 563931db9f7cSAlexander Block case BTRFS_IOC_SEND: 564031db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 5641c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 5642b27f7c0cSDavid Sterba return btrfs_ioctl_get_dev_stats(root, argp); 56435d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5644905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 56455d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5646905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 56475d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5648905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 56495d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5650905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 56512f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 56522f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 56532f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 56542f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 565557254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 565657254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 56573f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 56583f6bcfbdSStefan Behrens return btrfs_ioctl_dev_replace(root, argp); 5659867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5660867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5661a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5662a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 56632eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5664d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 56652eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 56662eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 56672eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56682eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5669f46b5a66SChristoph Hellwig } 5670f46b5a66SChristoph Hellwig 5671f46b5a66SChristoph Hellwig return -ENOTTY; 5672f46b5a66SChristoph Hellwig } 56734c63c245SLuke Dashjr 56744c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56754c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56764c63c245SLuke Dashjr { 56774c63c245SLuke Dashjr switch (cmd) { 56784c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56794c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56804c63c245SLuke Dashjr break; 56814c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56824c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56834c63c245SLuke Dashjr break; 56844c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56854c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56864c63c245SLuke Dashjr break; 56874c63c245SLuke Dashjr default: 56884c63c245SLuke Dashjr return -ENOIOCTLCMD; 56894c63c245SLuke Dashjr } 56904c63c245SLuke Dashjr 56914c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56924c63c245SLuke Dashjr } 56934c63c245SLuke Dashjr #endif 5694