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 3553a45bb20SJeff Mahoney btrfs_end_transaction(trans); 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++; 39850d0446eSSeraphime Kirkovski minlen = min_t(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); 4142ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &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, 43752f75f4fSDavid Sterba const 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); 4907775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 49149a3c4d9SDavid Sterba goto fail_free; 492d5c12070SMiao Xie } 493d5c12070SMiao Xie trans->block_rsv = &block_rsv; 494d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 495f46b5a66SChristoph Hellwig 4960b246afaSJeff Mahoney ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); 4976f72c7e2SArne Jansen if (ret) 4986f72c7e2SArne Jansen goto fail; 4996f72c7e2SArne Jansen 5004d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 5018e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 5028e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 5038e8a1e31SJosef Bacik goto fail; 5048e8a1e31SJosef Bacik } 505f46b5a66SChristoph Hellwig 506b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 507f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 508f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 5095d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 510f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 511f46b5a66SChristoph Hellwig 5120b246afaSJeff Mahoney write_extent_buffer_fsid(leaf, fs_info->fsid); 5130b246afaSJeff Mahoney write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 514f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 515f46b5a66SChristoph Hellwig 51649a3c4d9SDavid Sterba inode_item = &root_item->inode; 5173cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 5183cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 5193cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 520da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 5210b246afaSJeff Mahoney fs_info->nodesize); 5223cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 523f46b5a66SChristoph Hellwig 52449a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 52549a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 5263cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 52708fe4db1SLi Zefan 52849a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 52949a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 53049a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 53149a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 53249a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 53349a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 534f46b5a66SChristoph Hellwig 53549a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 53649a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 5378ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 53849a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 53949a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 54049a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 54149a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 54249a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 54349a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 544f46b5a66SChristoph Hellwig 545925baeddSChris Mason btrfs_tree_unlock(leaf); 546f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 547f46b5a66SChristoph Hellwig leaf = NULL; 548f46b5a66SChristoph Hellwig 54949a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 550f46b5a66SChristoph Hellwig 551f46b5a66SChristoph Hellwig key.objectid = objectid; 5525d4f98a2SYan Zheng key.offset = 0; 553962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 5540b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 55549a3c4d9SDavid Sterba root_item); 556f46b5a66SChristoph Hellwig if (ret) 557f46b5a66SChristoph Hellwig goto fail; 558f46b5a66SChristoph Hellwig 55976dda93cSYan, Zheng key.offset = (u64)-1; 5600b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 56179787eaaSJeff Mahoney if (IS_ERR(new_root)) { 56279787eaaSJeff Mahoney ret = PTR_ERR(new_root); 56366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 56479787eaaSJeff Mahoney goto fail; 56579787eaaSJeff Mahoney } 56676dda93cSYan, Zheng 56776dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 56876dda93cSYan, Zheng 56963541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 570ce598979SMark Fasheh if (ret) { 571ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 57266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 573ce598979SMark Fasheh goto fail; 574ce598979SMark Fasheh } 575ce598979SMark Fasheh 576f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 577f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 578f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 579f32e48e9SChandan Rajendra 580f46b5a66SChristoph Hellwig /* 581f46b5a66SChristoph Hellwig * insert the directory item 582f46b5a66SChristoph Hellwig */ 583877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 58479787eaaSJeff Mahoney if (ret) { 58566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 58679787eaaSJeff Mahoney goto fail; 58779787eaaSJeff Mahoney } 5883de4586cSChris Mason 5893de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 5908e7611cfSNikolay Borisov name, namelen, BTRFS_I(dir), &key, 5913de4586cSChris Mason BTRFS_FT_DIR, index); 59279787eaaSJeff Mahoney if (ret) { 59366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 594f46b5a66SChristoph Hellwig goto fail; 59579787eaaSJeff Mahoney } 5960660b5afSChris Mason 5976ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 59852c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 59952c26179SYan Zheng BUG_ON(ret); 60052c26179SYan Zheng 6010b246afaSJeff Mahoney ret = btrfs_add_root_ref(trans, fs_info, 6024df27c4dSYan, Zheng objectid, root->root_key.objectid, 6034a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 6040660b5afSChris Mason BUG_ON(ret); 6050660b5afSChris Mason 6060b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid, 6076bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 608dd5f9615SStefan Behrens if (ret) 60966642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 610dd5f9615SStefan Behrens 611f46b5a66SChristoph Hellwig fail: 61249a3c4d9SDavid Sterba kfree(root_item); 613d5c12070SMiao Xie trans->block_rsv = NULL; 614d5c12070SMiao Xie trans->bytes_reserved = 0; 6157775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 616de6e8200SLiu Bo 61772fd032eSSage Weil if (async_transid) { 61872fd032eSSage Weil *async_transid = trans->transid; 6193a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 62000d71c9cSMiao Xie if (err) 6213a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 62272fd032eSSage Weil } else { 6233a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 62472fd032eSSage Weil } 625f46b5a66SChristoph Hellwig if (err && !ret) 626f46b5a66SChristoph Hellwig ret = err; 6271a65e24bSChris Mason 6285662344bSTsutomu Itoh if (!ret) { 6295662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 630de6e8200SLiu Bo if (IS_ERR(inode)) 631de6e8200SLiu Bo return PTR_ERR(inode); 6325662344bSTsutomu Itoh d_instantiate(dentry, inode); 6335662344bSTsutomu Itoh } 634f46b5a66SChristoph Hellwig return ret; 63549a3c4d9SDavid Sterba 63649a3c4d9SDavid Sterba fail_free: 63749a3c4d9SDavid Sterba kfree(root_item); 63849a3c4d9SDavid Sterba return ret; 639f46b5a66SChristoph Hellwig } 640f46b5a66SChristoph Hellwig 6419ea24bbeSFilipe Manana static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) 6428257b2dcSMiao Xie { 6438257b2dcSMiao Xie s64 writers; 6448257b2dcSMiao Xie DEFINE_WAIT(wait); 6458257b2dcSMiao Xie 6468257b2dcSMiao Xie do { 6478257b2dcSMiao Xie prepare_to_wait(&root->subv_writers->wait, &wait, 6488257b2dcSMiao Xie TASK_UNINTERRUPTIBLE); 6498257b2dcSMiao Xie 6508257b2dcSMiao Xie writers = percpu_counter_sum(&root->subv_writers->counter); 6518257b2dcSMiao Xie if (writers) 6528257b2dcSMiao Xie schedule(); 6538257b2dcSMiao Xie 6548257b2dcSMiao Xie finish_wait(&root->subv_writers->wait, &wait); 6558257b2dcSMiao Xie } while (writers); 6568257b2dcSMiao Xie } 6578257b2dcSMiao Xie 658e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 65961d7e4cbSDavid Sterba struct dentry *dentry, 660e9662f70SMiao Xie u64 *async_transid, bool readonly, 661e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 662f46b5a66SChristoph Hellwig { 6630b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 6642e4bfab9SYan, Zheng struct inode *inode; 665f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 666f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 6672e4bfab9SYan, Zheng int ret; 668f46b5a66SChristoph Hellwig 66927cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 670f46b5a66SChristoph Hellwig return -EINVAL; 671f46b5a66SChristoph Hellwig 67223269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 673a1ee7362SDavid Sterba if (!pending_snapshot) 674a1ee7362SDavid Sterba return -ENOMEM; 675a1ee7362SDavid Sterba 676b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 67723269bf5SDavid Sterba GFP_KERNEL); 6788546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 6798546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 680b0c0ea63SDavid Sterba ret = -ENOMEM; 681b0c0ea63SDavid Sterba goto free_pending; 682b0c0ea63SDavid Sterba } 683b0c0ea63SDavid Sterba 6848257b2dcSMiao Xie atomic_inc(&root->will_be_snapshoted); 6854e857c58SPeter Zijlstra smp_mb__after_atomic(); 6869ea24bbeSFilipe Manana btrfs_wait_for_no_snapshoting_writes(root); 6878257b2dcSMiao Xie 6886a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 6896a03843dSMiao Xie if (ret) 690a1ee7362SDavid Sterba goto dec_and_free; 6916a03843dSMiao Xie 692578def7cSFilipe Manana btrfs_wait_ordered_extents(root, -1, 0, (u64)-1); 6936a03843dSMiao Xie 69466d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 69566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 696d5c12070SMiao Xie /* 697d5c12070SMiao Xie * 1 - parent dir inode 698d5c12070SMiao Xie * 2 - dir entries 699d5c12070SMiao Xie * 1 - root item 700d5c12070SMiao Xie * 2 - root ref/backref 701d5c12070SMiao Xie * 1 - root of snapshot 702dd5f9615SStefan Behrens * 1 - UUID item 703d5c12070SMiao Xie */ 704d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 705dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 706ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 707ee3441b4SJeff Mahoney false); 708d5c12070SMiao Xie if (ret) 709a1ee7362SDavid Sterba goto dec_and_free; 710d5c12070SMiao Xie 711a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 712a22285a6SYan, Zheng pending_snapshot->root = root; 713b83cc969SLi Zefan pending_snapshot->readonly = readonly; 714e9662f70SMiao Xie pending_snapshot->dir = dir; 7158696c533SMiao Xie pending_snapshot->inherit = inherit; 716a22285a6SYan, Zheng 717d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 718a22285a6SYan, Zheng if (IS_ERR(trans)) { 719a22285a6SYan, Zheng ret = PTR_ERR(trans); 720a22285a6SYan, Zheng goto fail; 721a22285a6SYan, Zheng } 722a22285a6SYan, Zheng 7230b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 724a22285a6SYan, Zheng list_add(&pending_snapshot->list, 725a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 7260b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 72772fd032eSSage Weil if (async_transid) { 72872fd032eSSage Weil *async_transid = trans->transid; 7293a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 73000d71c9cSMiao Xie if (ret) 7313a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 73272fd032eSSage Weil } else { 7333a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 73472fd032eSSage Weil } 735aec8030aSMiao Xie if (ret) 736c37b2b62SJosef Bacik goto fail; 737a22285a6SYan, Zheng 738a22285a6SYan, Zheng ret = pending_snapshot->error; 739f46b5a66SChristoph Hellwig if (ret) 7402e4bfab9SYan, Zheng goto fail; 741f46b5a66SChristoph Hellwig 742d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 743d3797308SChris Mason if (ret) 744d3797308SChris Mason goto fail; 745d3797308SChris Mason 7462b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7472e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7482e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7492e4bfab9SYan, Zheng goto fail; 7502e4bfab9SYan, Zheng } 7515662344bSTsutomu Itoh 7522e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7532e4bfab9SYan, Zheng ret = 0; 7542e4bfab9SYan, Zheng fail: 7557775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 756a1ee7362SDavid Sterba dec_and_free: 7579ea24bbeSFilipe Manana if (atomic_dec_and_test(&root->will_be_snapshoted)) 7589ea24bbeSFilipe Manana wake_up_atomic_t(&root->will_be_snapshoted); 759b0c0ea63SDavid Sterba free_pending: 760b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 7618546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 762a1ee7362SDavid Sterba kfree(pending_snapshot); 763a1ee7362SDavid Sterba 764f46b5a66SChristoph Hellwig return ret; 765f46b5a66SChristoph Hellwig } 766f46b5a66SChristoph Hellwig 7674260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7684260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7694260f7c7SSage Weil * whether the type of victim is right. 7704260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7714260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7724260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7734260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7744260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7754260f7c7SSage Weil * a. be owner of dir, or 7764260f7c7SSage Weil * b. be owner of victim, or 7774260f7c7SSage Weil * c. have CAP_FOWNER capability 77801327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 7794260f7c7SSage Weil * links pointing to it. 7804260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7814260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7824260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7834260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 7844260f7c7SSage Weil * nfs_async_unlink(). 7854260f7c7SSage Weil */ 7864260f7c7SSage Weil 7874260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 7884260f7c7SSage Weil { 7894260f7c7SSage Weil int error; 7904260f7c7SSage Weil 7912b0143b5SDavid Howells if (d_really_is_negative(victim)) 7924260f7c7SSage Weil return -ENOENT; 7934260f7c7SSage Weil 7942b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 7954fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 7964260f7c7SSage Weil 7974260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 7984260f7c7SSage Weil if (error) 7994260f7c7SSage Weil return error; 8004260f7c7SSage Weil if (IS_APPEND(dir)) 8014260f7c7SSage Weil return -EPERM; 8022b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 8032b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 8044260f7c7SSage Weil return -EPERM; 8054260f7c7SSage Weil if (isdir) { 806e36cb0b8SDavid Howells if (!d_is_dir(victim)) 8074260f7c7SSage Weil return -ENOTDIR; 8084260f7c7SSage Weil if (IS_ROOT(victim)) 8094260f7c7SSage Weil return -EBUSY; 810e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 8114260f7c7SSage Weil return -EISDIR; 8124260f7c7SSage Weil if (IS_DEADDIR(dir)) 8134260f7c7SSage Weil return -ENOENT; 8144260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 8154260f7c7SSage Weil return -EBUSY; 8164260f7c7SSage Weil return 0; 8174260f7c7SSage Weil } 8184260f7c7SSage Weil 819cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 820cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 821cb8e7090SChristoph Hellwig { 8222b0143b5SDavid Howells if (d_really_is_positive(child)) 823cb8e7090SChristoph Hellwig return -EEXIST; 824cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 825cb8e7090SChristoph Hellwig return -ENOENT; 826cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 827cb8e7090SChristoph Hellwig } 828cb8e7090SChristoph Hellwig 829cb8e7090SChristoph Hellwig /* 830cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 831cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 832cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 833cb8e7090SChristoph Hellwig */ 83492872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 83552f75f4fSDavid Sterba const char *name, int namelen, 83672fd032eSSage Weil struct btrfs_root *snap_src, 8376f72c7e2SArne Jansen u64 *async_transid, bool readonly, 8388696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 839cb8e7090SChristoph Hellwig { 8402b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 8410b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 842cb8e7090SChristoph Hellwig struct dentry *dentry; 843cb8e7090SChristoph Hellwig int error; 844cb8e7090SChristoph Hellwig 84500235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 84600235411SAl Viro if (error == -EINTR) 84700235411SAl Viro return error; 848cb8e7090SChristoph Hellwig 849cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 850cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 851cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 852cb8e7090SChristoph Hellwig goto out_unlock; 853cb8e7090SChristoph Hellwig 85476dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 855cb8e7090SChristoph Hellwig if (error) 856a874a63eSLiu Bo goto out_dput; 857cb8e7090SChristoph Hellwig 8589c52057cSChris Mason /* 8599c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8609c52057cSChris Mason * check for them now when we can safely fail 8619c52057cSChris Mason */ 8629c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8639c52057cSChris Mason dir->i_ino, name, 8649c52057cSChris Mason namelen); 8659c52057cSChris Mason if (error) 8669c52057cSChris Mason goto out_dput; 8679c52057cSChris Mason 8680b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 86976dda93cSYan, Zheng 87076dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 87176dda93cSYan, Zheng goto out_up_read; 87276dda93cSYan, Zheng 8733de4586cSChris Mason if (snap_src) { 87461d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 8756f72c7e2SArne Jansen async_transid, readonly, inherit); 8763de4586cSChris Mason } else { 877d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 878d5c12070SMiao Xie async_transid, inherit); 8793de4586cSChris Mason } 88076dda93cSYan, Zheng if (!error) 88176dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 88276dda93cSYan, Zheng out_up_read: 8830b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 884cb8e7090SChristoph Hellwig out_dput: 885cb8e7090SChristoph Hellwig dput(dentry); 886cb8e7090SChristoph Hellwig out_unlock: 8875955102cSAl Viro inode_unlock(dir); 888cb8e7090SChristoph Hellwig return error; 889cb8e7090SChristoph Hellwig } 890cb8e7090SChristoph Hellwig 8914cb5300bSChris Mason /* 8924cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 8934cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 8944cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 8954cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 8964cb5300bSChris Mason * part of the file 8974cb5300bSChris Mason */ 898aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 8994cb5300bSChris Mason { 9004cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9014cb5300bSChris Mason struct extent_map *em = NULL; 9024cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 9034cb5300bSChris Mason u64 end; 9044cb5300bSChris Mason 9054cb5300bSChris Mason read_lock(&em_tree->lock); 90609cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 9074cb5300bSChris Mason read_unlock(&em_tree->lock); 9084cb5300bSChris Mason 9094cb5300bSChris Mason if (em) { 9104cb5300bSChris Mason end = extent_map_end(em); 9114cb5300bSChris Mason free_extent_map(em); 9124cb5300bSChris Mason if (end - offset > thresh) 9134cb5300bSChris Mason return 0; 9144cb5300bSChris Mason } 9154cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 9164cb5300bSChris Mason thresh /= 2; 9174cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 9184cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 9194cb5300bSChris Mason if (end >= thresh) 9204cb5300bSChris Mason return 0; 9214cb5300bSChris Mason return 1; 9224cb5300bSChris Mason } 9234cb5300bSChris Mason 9244cb5300bSChris Mason /* 9254cb5300bSChris Mason * helper function to walk through a file and find extents 9264cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 9274cb5300bSChris Mason * 9284cb5300bSChris Mason * This is used by the defragging code to find new and small 9294cb5300bSChris Mason * extents 9304cb5300bSChris Mason */ 9314cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 9324cb5300bSChris Mason struct inode *inode, u64 newer_than, 933aab110abSDavid Sterba u64 *off, u32 thresh) 9344cb5300bSChris Mason { 9354cb5300bSChris Mason struct btrfs_path *path; 9364cb5300bSChris Mason struct btrfs_key min_key; 9374cb5300bSChris Mason struct extent_buffer *leaf; 9384cb5300bSChris Mason struct btrfs_file_extent_item *extent; 9394cb5300bSChris Mason int type; 9404cb5300bSChris Mason int ret; 9414a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 9424cb5300bSChris Mason 9434cb5300bSChris Mason path = btrfs_alloc_path(); 9444cb5300bSChris Mason if (!path) 9454cb5300bSChris Mason return -ENOMEM; 9464cb5300bSChris Mason 947a4689d2bSDavid Sterba min_key.objectid = ino; 9484cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9494cb5300bSChris Mason min_key.offset = *off; 9504cb5300bSChris Mason 9514cb5300bSChris Mason while (1) { 9526174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9534cb5300bSChris Mason if (ret != 0) 9544cb5300bSChris Mason goto none; 955f094c9bdSFilipe Manana process_slot: 956a4689d2bSDavid Sterba if (min_key.objectid != ino) 9574cb5300bSChris Mason goto none; 9584cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9594cb5300bSChris Mason goto none; 9604cb5300bSChris Mason 9614cb5300bSChris Mason leaf = path->nodes[0]; 9624cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9634cb5300bSChris Mason struct btrfs_file_extent_item); 9644cb5300bSChris Mason 9654cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9664cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9674cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9684cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9694cb5300bSChris Mason *off = min_key.offset; 9704cb5300bSChris Mason btrfs_free_path(path); 9714cb5300bSChris Mason return 0; 9724cb5300bSChris Mason } 9734cb5300bSChris Mason 974f094c9bdSFilipe Manana path->slots[0]++; 975f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 976f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 977f094c9bdSFilipe Manana goto process_slot; 978f094c9bdSFilipe Manana } 979f094c9bdSFilipe Manana 9804cb5300bSChris Mason if (min_key.offset == (u64)-1) 9814cb5300bSChris Mason goto none; 9824cb5300bSChris Mason 9834cb5300bSChris Mason min_key.offset++; 9844cb5300bSChris Mason btrfs_release_path(path); 9854cb5300bSChris Mason } 9864cb5300bSChris Mason none: 9874cb5300bSChris Mason btrfs_free_path(path); 9884cb5300bSChris Mason return -ENOENT; 9894cb5300bSChris Mason } 9904cb5300bSChris Mason 9916c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 99217ce6ef8SLiu Bo { 99317ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 994940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9956c282eb4SLi Zefan struct extent_map *em; 99609cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 997940100a4SChris Mason 998940100a4SChris Mason /* 999940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1000940100a4SChris Mason * the full extent lock 1001940100a4SChris Mason */ 1002940100a4SChris Mason read_lock(&em_tree->lock); 1003940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1004940100a4SChris Mason read_unlock(&em_tree->lock); 1005940100a4SChris Mason 1006940100a4SChris Mason if (!em) { 1007308d9800SFilipe Manana struct extent_state *cached = NULL; 1008308d9800SFilipe Manana u64 end = start + len - 1; 1009308d9800SFilipe Manana 1010940100a4SChris Mason /* get the big lock and read metadata off disk */ 1011ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1012fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 1013308d9800SFilipe Manana unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 1014940100a4SChris Mason 10156cf8bfbfSDan Carpenter if (IS_ERR(em)) 10166c282eb4SLi Zefan return NULL; 1017940100a4SChris Mason } 1018940100a4SChris Mason 10196c282eb4SLi Zefan return em; 10206c282eb4SLi Zefan } 10216c282eb4SLi Zefan 10226c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 10236c282eb4SLi Zefan { 10246c282eb4SLi Zefan struct extent_map *next; 10256c282eb4SLi Zefan bool ret = true; 10266c282eb4SLi Zefan 10276c282eb4SLi Zefan /* this is the last extent */ 10286c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 10296c282eb4SLi Zefan return false; 10306c282eb4SLi Zefan 10316c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1032e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1033e9512d72SChris Mason ret = false; 1034e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1035ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 10366c282eb4SLi Zefan ret = false; 10376c282eb4SLi Zefan 10386c282eb4SLi Zefan free_extent_map(next); 10396c282eb4SLi Zefan return ret; 10406c282eb4SLi Zefan } 10416c282eb4SLi Zefan 1042aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1043a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1044a43a2111SAndrew Mahone int compress) 10456c282eb4SLi Zefan { 10466c282eb4SLi Zefan struct extent_map *em; 10476c282eb4SLi Zefan int ret = 1; 10486c282eb4SLi Zefan bool next_mergeable = true; 10494a3560c4SLiu Bo bool prev_mergeable = true; 10506c282eb4SLi Zefan 10516c282eb4SLi Zefan /* 10526c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10536c282eb4SLi Zefan * defragging it 10546c282eb4SLi Zefan */ 10556c282eb4SLi Zefan if (start < *defrag_end) 10566c282eb4SLi Zefan return 1; 10576c282eb4SLi Zefan 10586c282eb4SLi Zefan *skip = 0; 10596c282eb4SLi Zefan 10606c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10616c282eb4SLi Zefan if (!em) 10626c282eb4SLi Zefan return 0; 10636c282eb4SLi Zefan 1064940100a4SChris Mason /* this will cover holes, and inline extents */ 106517ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1066940100a4SChris Mason ret = 0; 106717ce6ef8SLiu Bo goto out; 106817ce6ef8SLiu Bo } 106917ce6ef8SLiu Bo 10704a3560c4SLiu Bo if (!*defrag_end) 10714a3560c4SLiu Bo prev_mergeable = false; 10724a3560c4SLiu Bo 10736c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1074940100a4SChris Mason /* 10756c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10766c282eb4SLi Zefan * real extent, don't bother defragging it 1077940100a4SChris Mason */ 1078a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10794a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1080940100a4SChris Mason ret = 0; 108117ce6ef8SLiu Bo out: 1082940100a4SChris Mason /* 1083940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1084940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1085940100a4SChris Mason * so that the next tiny extent will force a defrag. 1086940100a4SChris Mason * 1087940100a4SChris Mason * The end result of this is that tiny extents before a single big 1088940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1089940100a4SChris Mason */ 1090940100a4SChris Mason if (ret) { 1091940100a4SChris Mason *defrag_end = extent_map_end(em); 1092940100a4SChris Mason } else { 1093940100a4SChris Mason *last_len = 0; 1094940100a4SChris Mason *skip = extent_map_end(em); 1095940100a4SChris Mason *defrag_end = 0; 1096940100a4SChris Mason } 1097940100a4SChris Mason 1098940100a4SChris Mason free_extent_map(em); 1099940100a4SChris Mason return ret; 1100940100a4SChris Mason } 1101940100a4SChris Mason 11024cb5300bSChris Mason /* 11034cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 11044cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 11054cb5300bSChris Mason * to COW and defrag. 11064cb5300bSChris Mason * 11074cb5300bSChris Mason * It also makes sure the delalloc code has enough 11084cb5300bSChris Mason * dirty data to avoid making new small extents as part 11094cb5300bSChris Mason * of the defrag 11104cb5300bSChris Mason * 11114cb5300bSChris Mason * It's a good idea to start RA on this range 11124cb5300bSChris Mason * before calling this. 11134cb5300bSChris Mason */ 11144cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 11154cb5300bSChris Mason struct page **pages, 11164cb5300bSChris Mason unsigned long start_index, 1117c41570c9SJustin Maggard unsigned long num_pages) 1118f46b5a66SChristoph Hellwig { 11194cb5300bSChris Mason unsigned long file_end; 11204cb5300bSChris Mason u64 isize = i_size_read(inode); 1121f46b5a66SChristoph Hellwig u64 page_start; 1122f46b5a66SChristoph Hellwig u64 page_end; 11231f12bd06SLiu Bo u64 page_cnt; 11244cb5300bSChris Mason int ret; 11254cb5300bSChris Mason int i; 11264cb5300bSChris Mason int i_done; 11274cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 11284cb5300bSChris Mason struct extent_state *cached_state = NULL; 1129600a45e1SMiao Xie struct extent_io_tree *tree; 11303b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 11314cb5300bSChris Mason 113209cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 11331f12bd06SLiu Bo if (!isize || start_index > file_end) 11341f12bd06SLiu Bo return 0; 11351f12bd06SLiu Bo 11361f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 11374cb5300bSChris Mason 11384cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 113909cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 114009cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 11414cb5300bSChris Mason if (ret) 11424cb5300bSChris Mason return ret; 11434cb5300bSChris Mason i_done = 0; 1144600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11454cb5300bSChris Mason 11464cb5300bSChris Mason /* step one, lock all the pages */ 11471f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11484cb5300bSChris Mason struct page *page; 1149600a45e1SMiao Xie again: 1150a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11513b16a4e3SJosef Bacik start_index + i, mask); 11524cb5300bSChris Mason if (!page) 11534cb5300bSChris Mason break; 11544cb5300bSChris Mason 1155600a45e1SMiao Xie page_start = page_offset(page); 115609cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1157600a45e1SMiao Xie while (1) { 1158308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1159ff13db41SDavid Sterba &cached_state); 1160600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1161600a45e1SMiao Xie page_start); 1162308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1163308d9800SFilipe Manana &cached_state, GFP_NOFS); 1164600a45e1SMiao Xie if (!ordered) 1165600a45e1SMiao Xie break; 1166600a45e1SMiao Xie 1167600a45e1SMiao Xie unlock_page(page); 1168600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1169600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1170600a45e1SMiao Xie lock_page(page); 11711f12bd06SLiu Bo /* 11721f12bd06SLiu Bo * we unlocked the page above, so we need check if 11731f12bd06SLiu Bo * it was released or not. 11741f12bd06SLiu Bo */ 11751f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11761f12bd06SLiu Bo unlock_page(page); 117709cbfeafSKirill A. Shutemov put_page(page); 11781f12bd06SLiu Bo goto again; 11791f12bd06SLiu Bo } 1180600a45e1SMiao Xie } 1181600a45e1SMiao Xie 11824cb5300bSChris Mason if (!PageUptodate(page)) { 11834cb5300bSChris Mason btrfs_readpage(NULL, page); 11844cb5300bSChris Mason lock_page(page); 11854cb5300bSChris Mason if (!PageUptodate(page)) { 11864cb5300bSChris Mason unlock_page(page); 118709cbfeafSKirill A. Shutemov put_page(page); 11884cb5300bSChris Mason ret = -EIO; 11894cb5300bSChris Mason break; 11904cb5300bSChris Mason } 11914cb5300bSChris Mason } 1192600a45e1SMiao Xie 1193600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1194600a45e1SMiao Xie unlock_page(page); 119509cbfeafSKirill A. Shutemov put_page(page); 1196600a45e1SMiao Xie goto again; 1197600a45e1SMiao Xie } 1198600a45e1SMiao Xie 11994cb5300bSChris Mason pages[i] = page; 12004cb5300bSChris Mason i_done++; 12014cb5300bSChris Mason } 12024cb5300bSChris Mason if (!i_done || ret) 12034cb5300bSChris Mason goto out; 12044cb5300bSChris Mason 12054cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 12064cb5300bSChris Mason goto out; 12074cb5300bSChris Mason 12084cb5300bSChris Mason /* 12094cb5300bSChris Mason * so now we have a nice long stream of locked 12104cb5300bSChris Mason * and up to date pages, lets wait on them 12114cb5300bSChris Mason */ 12124cb5300bSChris Mason for (i = 0; i < i_done; i++) 12134cb5300bSChris Mason wait_on_page_writeback(pages[i]); 12144cb5300bSChris Mason 12154cb5300bSChris Mason page_start = page_offset(pages[0]); 121609cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 12174cb5300bSChris Mason 12184cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1219ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 12204cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 12214cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 12229e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 12239e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12244cb5300bSChris Mason 12251f12bd06SLiu Bo if (i_done != page_cnt) { 12269e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 12279e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 12289e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 12294cb5300bSChris Mason btrfs_delalloc_release_space(inode, 123009cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 123109cbfeafSKirill A. Shutemov (page_cnt - i_done) << PAGE_SHIFT); 12324cb5300bSChris Mason } 12334cb5300bSChris Mason 12344cb5300bSChris Mason 12359e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1236018ed4f7SDavid Sterba &cached_state); 12374cb5300bSChris Mason 12384cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 12394cb5300bSChris Mason page_start, page_end - 1, &cached_state, 12404cb5300bSChris Mason GFP_NOFS); 12414cb5300bSChris Mason 12424cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12434cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12444cb5300bSChris Mason ClearPageChecked(pages[i]); 12454cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12464cb5300bSChris Mason set_page_dirty(pages[i]); 12474cb5300bSChris Mason unlock_page(pages[i]); 124809cbfeafSKirill A. Shutemov put_page(pages[i]); 12494cb5300bSChris Mason } 12504cb5300bSChris Mason return i_done; 12514cb5300bSChris Mason out: 12524cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12534cb5300bSChris Mason unlock_page(pages[i]); 125409cbfeafSKirill A. Shutemov put_page(pages[i]); 12554cb5300bSChris Mason } 12567cf5b976SQu Wenruo btrfs_delalloc_release_space(inode, 125709cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 125809cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12594cb5300bSChris Mason return ret; 12604cb5300bSChris Mason 12614cb5300bSChris Mason } 12624cb5300bSChris Mason 12634cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12644cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12654cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12664cb5300bSChris Mason { 12670b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 12684cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12694cb5300bSChris Mason struct file_ra_state *ra = NULL; 12704cb5300bSChris Mason unsigned long last_index; 1271151a31b2SLi Zefan u64 isize = i_size_read(inode); 1272940100a4SChris Mason u64 last_len = 0; 1273940100a4SChris Mason u64 skip = 0; 1274940100a4SChris Mason u64 defrag_end = 0; 12754cb5300bSChris Mason u64 newer_off = range->start; 1276f46b5a66SChristoph Hellwig unsigned long i; 1277008873eaSLi Zefan unsigned long ra_index = 0; 1278f46b5a66SChristoph Hellwig int ret; 12794cb5300bSChris Mason int defrag_count = 0; 12801a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1281aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 128209cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1283c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1284ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 12854cb5300bSChris Mason struct page **pages = NULL; 12864cb5300bSChris Mason 12870abd5b17SLiu Bo if (isize == 0) 12880abd5b17SLiu Bo return 0; 12890abd5b17SLiu Bo 12900abd5b17SLiu Bo if (range->start >= isize) 12910abd5b17SLiu Bo return -EINVAL; 12921a419d85SLi Zefan 12931a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 12941a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 12951a419d85SLi Zefan return -EINVAL; 12961a419d85SLi Zefan if (range->compress_type) 12971a419d85SLi Zefan compress_type = range->compress_type; 12981a419d85SLi Zefan } 1299f46b5a66SChristoph Hellwig 13000abd5b17SLiu Bo if (extent_thresh == 0) 1301ee22184bSByongho Lee extent_thresh = SZ_256K; 1302f46b5a66SChristoph Hellwig 13034cb5300bSChris Mason /* 13044cb5300bSChris Mason * if we were not given a file, allocate a readahead 13054cb5300bSChris Mason * context 13064cb5300bSChris Mason */ 13074cb5300bSChris Mason if (!file) { 13084cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 13094cb5300bSChris Mason if (!ra) 13104cb5300bSChris Mason return -ENOMEM; 13114cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 13124cb5300bSChris Mason } else { 13134cb5300bSChris Mason ra = &file->f_ra; 13144cb5300bSChris Mason } 13154cb5300bSChris Mason 1316d9b0d9baSDulshani Gunawardhana pages = kmalloc_array(max_cluster, sizeof(struct page *), 13174cb5300bSChris Mason GFP_NOFS); 13184cb5300bSChris Mason if (!pages) { 13194cb5300bSChris Mason ret = -ENOMEM; 13204cb5300bSChris Mason goto out_ra; 13214cb5300bSChris Mason } 13224cb5300bSChris Mason 13234cb5300bSChris Mason /* find the last page to defrag */ 13241e701a32SChris Mason if (range->start + range->len > range->start) { 1325151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 132609cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 13271e701a32SChris Mason } else { 132809cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 13291e701a32SChris Mason } 13301e701a32SChris Mason 13314cb5300bSChris Mason if (newer_than) { 13324cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1333ee22184bSByongho Lee &newer_off, SZ_64K); 13344cb5300bSChris Mason if (!ret) { 13354cb5300bSChris Mason range->start = newer_off; 13364cb5300bSChris Mason /* 13374cb5300bSChris Mason * we always align our defrag to help keep 13384cb5300bSChris Mason * the extents in the file evenly spaced 13394cb5300bSChris Mason */ 134009cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 13414cb5300bSChris Mason } else 13424cb5300bSChris Mason goto out_ra; 13434cb5300bSChris Mason } else { 134409cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 13454cb5300bSChris Mason } 13464cb5300bSChris Mason if (!max_to_defrag) 1347070034bdSchandan max_to_defrag = last_index - i + 1; 13484cb5300bSChris Mason 13492a0f7f57SLi Zefan /* 13502a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13512a0f7f57SLi Zefan * written sequentially. 13522a0f7f57SLi Zefan */ 13532a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13542a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13552a0f7f57SLi Zefan 1356f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 135709cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 13584cb5300bSChris Mason /* 13594cb5300bSChris Mason * make sure we stop running if someone unmounts 13604cb5300bSChris Mason * the FS 13614cb5300bSChris Mason */ 13624cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 13634cb5300bSChris Mason break; 13644cb5300bSChris Mason 13650b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 13660b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1367210549ebSDavid Sterba ret = -EAGAIN; 1368210549ebSDavid Sterba break; 1369210549ebSDavid Sterba } 1370210549ebSDavid Sterba 137109cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 13726c282eb4SLi Zefan extent_thresh, &last_len, &skip, 1373a43a2111SAndrew Mahone &defrag_end, range->flags & 1374a43a2111SAndrew Mahone BTRFS_DEFRAG_RANGE_COMPRESS)) { 1375940100a4SChris Mason unsigned long next; 1376940100a4SChris Mason /* 1377940100a4SChris Mason * the should_defrag function tells us how much to skip 1378940100a4SChris Mason * bump our counter by the suggested amount 1379940100a4SChris Mason */ 138009cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1381940100a4SChris Mason i = max(i + 1, next); 1382940100a4SChris Mason continue; 1383940100a4SChris Mason } 1384008873eaSLi Zefan 1385008873eaSLi Zefan if (!newer_than) { 138609cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 138709cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1388008873eaSLi Zefan cluster = min(cluster, max_cluster); 1389008873eaSLi Zefan } else { 1390008873eaSLi Zefan cluster = max_cluster; 1391008873eaSLi Zefan } 1392008873eaSLi Zefan 1393008873eaSLi Zefan if (i + cluster > ra_index) { 1394008873eaSLi Zefan ra_index = max(i, ra_index); 1395008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1396008873eaSLi Zefan cluster); 1397e4826a5bSchandan ra_index += cluster; 1398008873eaSLi Zefan } 13994cb5300bSChris Mason 14005955102cSAl Viro inode_lock(inode); 1401633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 1402633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = compress_type; 1403008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1404ecb8bea8SLiu Bo if (ret < 0) { 14055955102cSAl Viro inode_unlock(inode); 14064cb5300bSChris Mason goto out_ra; 1407ecb8bea8SLiu Bo } 14084cb5300bSChris Mason 14094cb5300bSChris Mason defrag_count += ret; 1410d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 14115955102cSAl Viro inode_unlock(inode); 14124cb5300bSChris Mason 14134cb5300bSChris Mason if (newer_than) { 14144cb5300bSChris Mason if (newer_off == (u64)-1) 14154cb5300bSChris Mason break; 14164cb5300bSChris Mason 1417e1f041e1SLiu Bo if (ret > 0) 1418e1f041e1SLiu Bo i += ret; 1419e1f041e1SLiu Bo 14204cb5300bSChris Mason newer_off = max(newer_off + 1, 142109cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 14224cb5300bSChris Mason 1423ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1424ee22184bSByongho Lee &newer_off, SZ_64K); 14254cb5300bSChris Mason if (!ret) { 14264cb5300bSChris Mason range->start = newer_off; 142709cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14284cb5300bSChris Mason } else { 14294cb5300bSChris Mason break; 1430940100a4SChris Mason } 14314cb5300bSChris Mason } else { 1432008873eaSLi Zefan if (ret > 0) { 1433cbcc8326SLi Zefan i += ret; 143409cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1435008873eaSLi Zefan } else { 1436940100a4SChris Mason i++; 1437008873eaSLi Zefan last_len = 0; 1438008873eaSLi Zefan } 1439f46b5a66SChristoph Hellwig } 14404cb5300bSChris Mason } 1441f46b5a66SChristoph Hellwig 1442dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14431e701a32SChris Mason filemap_flush(inode->i_mapping); 1444dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1445dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1446dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1447dec8ef90SFilipe Manana } 14481e701a32SChris Mason 14491e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 14501e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 14511e701a32SChris Mason * we have to make sure the IO is actually started and that 14521e701a32SChris Mason * ordered extents get created before we return 14531e701a32SChris Mason */ 14540b246afaSJeff Mahoney atomic_inc(&fs_info->async_submit_draining); 14550b246afaSJeff Mahoney while (atomic_read(&fs_info->nr_async_submits) || 14560b246afaSJeff Mahoney atomic_read(&fs_info->async_delalloc_pages)) { 14570b246afaSJeff Mahoney wait_event(fs_info->async_submit_wait, 14580b246afaSJeff Mahoney (atomic_read(&fs_info->nr_async_submits) == 0 && 14590b246afaSJeff Mahoney atomic_read(&fs_info->async_delalloc_pages) == 0)); 14601e701a32SChris Mason } 14610b246afaSJeff Mahoney atomic_dec(&fs_info->async_submit_draining); 14621e701a32SChris Mason } 14631e701a32SChris Mason 14641a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14650b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 14661a419d85SLi Zefan } 14671a419d85SLi Zefan 146860ccf82fSDiego Calleja ret = defrag_count; 1469940100a4SChris Mason 14704cb5300bSChris Mason out_ra: 1471633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 14725955102cSAl Viro inode_lock(inode); 1473633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 14745955102cSAl Viro inode_unlock(inode); 1475633085c7SFilipe David Borba Manana } 14764cb5300bSChris Mason if (!file) 14774cb5300bSChris Mason kfree(ra); 14784cb5300bSChris Mason kfree(pages); 1479940100a4SChris Mason return ret; 1480f46b5a66SChristoph Hellwig } 1481f46b5a66SChristoph Hellwig 1482198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 148376dda93cSYan, Zheng void __user *arg) 1484f46b5a66SChristoph Hellwig { 14850b246afaSJeff Mahoney struct inode *inode = file_inode(file); 14860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1487f46b5a66SChristoph Hellwig u64 new_size; 1488f46b5a66SChristoph Hellwig u64 old_size; 1489f46b5a66SChristoph Hellwig u64 devid = 1; 14900b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1491f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1492f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1493f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1494f46b5a66SChristoph Hellwig char *sizestr; 14959a40f122SGui Hecheng char *retptr; 1496f46b5a66SChristoph Hellwig char *devstr = NULL; 1497f46b5a66SChristoph Hellwig int ret = 0; 1498f46b5a66SChristoph Hellwig int mod = 0; 1499f46b5a66SChristoph Hellwig 1500e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1501e441d54dSChris Mason return -EPERM; 1502e441d54dSChris Mason 1503198605a8SMiao Xie ret = mnt_want_write_file(file); 1504198605a8SMiao Xie if (ret) 1505198605a8SMiao Xie return ret; 1506198605a8SMiao Xie 15070b246afaSJeff Mahoney if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 150897547676SMiao Xie mnt_drop_write_file(file); 1509e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1510c9e9f97bSIlya Dryomov } 1511c9e9f97bSIlya Dryomov 15120b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 1513dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1514c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1515c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1516c9e9f97bSIlya Dryomov goto out; 1517c9e9f97bSIlya Dryomov } 15185516e595SMark Fasheh 15195516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1520f46b5a66SChristoph Hellwig 1521f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1522f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1523f46b5a66SChristoph Hellwig if (devstr) { 1524f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1525f46b5a66SChristoph Hellwig *devstr = '\0'; 1526f46b5a66SChristoph Hellwig devstr = vol_args->name; 152758dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 152858dfae63SZhangZhen if (ret) 152958dfae63SZhangZhen goto out_free; 1530dfd79829SMiao Xie if (!devid) { 1531dfd79829SMiao Xie ret = -EINVAL; 1532dfd79829SMiao Xie goto out_free; 1533dfd79829SMiao Xie } 15340b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1535f46b5a66SChristoph Hellwig } 1536dba60f3fSMiao Xie 15370b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1538f46b5a66SChristoph Hellwig if (!device) { 15390b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1540c1c9ff7cSGeert Uytterhoeven devid); 1541dfd79829SMiao Xie ret = -ENODEV; 1542c9e9f97bSIlya Dryomov goto out_free; 1543f46b5a66SChristoph Hellwig } 1544dba60f3fSMiao Xie 1545dba60f3fSMiao Xie if (!device->writeable) { 15460b246afaSJeff Mahoney btrfs_info(fs_info, 1547efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1548c1c9ff7cSGeert Uytterhoeven devid); 1549dfd79829SMiao Xie ret = -EPERM; 15504e42ae1bSLiu Bo goto out_free; 15514e42ae1bSLiu Bo } 15524e42ae1bSLiu Bo 1553f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1554f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1555f46b5a66SChristoph Hellwig else { 1556f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1557f46b5a66SChristoph Hellwig mod = -1; 1558f46b5a66SChristoph Hellwig sizestr++; 1559f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1560f46b5a66SChristoph Hellwig mod = 1; 1561f46b5a66SChristoph Hellwig sizestr++; 1562f46b5a66SChristoph Hellwig } 15639a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15649a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1565f46b5a66SChristoph Hellwig ret = -EINVAL; 1566c9e9f97bSIlya Dryomov goto out_free; 1567f46b5a66SChristoph Hellwig } 1568f46b5a66SChristoph Hellwig } 1569f46b5a66SChristoph Hellwig 157063a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1571dfd79829SMiao Xie ret = -EPERM; 157263a212abSStefan Behrens goto out_free; 157363a212abSStefan Behrens } 157463a212abSStefan Behrens 15757cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1576f46b5a66SChristoph Hellwig 1577f46b5a66SChristoph Hellwig if (mod < 0) { 1578f46b5a66SChristoph Hellwig if (new_size > old_size) { 1579f46b5a66SChristoph Hellwig ret = -EINVAL; 1580c9e9f97bSIlya Dryomov goto out_free; 1581f46b5a66SChristoph Hellwig } 1582f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1583f46b5a66SChristoph Hellwig } else if (mod > 0) { 1584eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1585902c68a4SGui Hecheng ret = -ERANGE; 1586eb8052e0SWenliang Fan goto out_free; 1587eb8052e0SWenliang Fan } 1588f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1589f46b5a66SChristoph Hellwig } 1590f46b5a66SChristoph Hellwig 1591ee22184bSByongho Lee if (new_size < SZ_256M) { 1592f46b5a66SChristoph Hellwig ret = -EINVAL; 1593c9e9f97bSIlya Dryomov goto out_free; 1594f46b5a66SChristoph Hellwig } 1595f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1596f46b5a66SChristoph Hellwig ret = -EFBIG; 1597c9e9f97bSIlya Dryomov goto out_free; 1598f46b5a66SChristoph Hellwig } 1599f46b5a66SChristoph Hellwig 16000b246afaSJeff Mahoney new_size = div_u64(new_size, fs_info->sectorsize); 16010b246afaSJeff Mahoney new_size *= fs_info->sectorsize; 1602f46b5a66SChristoph Hellwig 16030b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1604c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1605f46b5a66SChristoph Hellwig 1606f46b5a66SChristoph Hellwig if (new_size > old_size) { 1607a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 160898d5dc13STsutomu Itoh if (IS_ERR(trans)) { 160998d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1610c9e9f97bSIlya Dryomov goto out_free; 161198d5dc13STsutomu Itoh } 1612f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 16133a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1614ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1615f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 16160253f40eSjeff.liu } /* equal, nothing need to do */ 1617f46b5a66SChristoph Hellwig 1618c9e9f97bSIlya Dryomov out_free: 1619f46b5a66SChristoph Hellwig kfree(vol_args); 1620c9e9f97bSIlya Dryomov out: 16210b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 16220b246afaSJeff Mahoney atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 162318f39c41SIlya Dryomov mnt_drop_write_file(file); 1624f46b5a66SChristoph Hellwig return ret; 1625f46b5a66SChristoph Hellwig } 1626f46b5a66SChristoph Hellwig 162772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 162852f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 16296f72c7e2SArne Jansen u64 *transid, bool readonly, 16308696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1631f46b5a66SChristoph Hellwig { 1632f46b5a66SChristoph Hellwig int namelen; 16333de4586cSChris Mason int ret = 0; 1634f46b5a66SChristoph Hellwig 1635325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1636325c50e3SJeff Mahoney return -ENOTDIR; 1637325c50e3SJeff Mahoney 1638a874a63eSLiu Bo ret = mnt_want_write_file(file); 1639a874a63eSLiu Bo if (ret) 1640a874a63eSLiu Bo goto out; 1641a874a63eSLiu Bo 164272fd032eSSage Weil namelen = strlen(name); 164372fd032eSSage Weil if (strchr(name, '/')) { 1644f46b5a66SChristoph Hellwig ret = -EINVAL; 1645a874a63eSLiu Bo goto out_drop_write; 1646f46b5a66SChristoph Hellwig } 1647f46b5a66SChristoph Hellwig 164816780cabSChris Mason if (name[0] == '.' && 164916780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 165016780cabSChris Mason ret = -EEXIST; 1651a874a63eSLiu Bo goto out_drop_write; 165216780cabSChris Mason } 165316780cabSChris Mason 16543de4586cSChris Mason if (subvol) { 165572fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16566f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1657cb8e7090SChristoph Hellwig } else { 16582903ff01SAl Viro struct fd src = fdget(fd); 16593de4586cSChris Mason struct inode *src_inode; 16602903ff01SAl Viro if (!src.file) { 16613de4586cSChris Mason ret = -EINVAL; 1662a874a63eSLiu Bo goto out_drop_write; 16633de4586cSChris Mason } 16643de4586cSChris Mason 1665496ad9aaSAl Viro src_inode = file_inode(src.file); 1666496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1667c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1668efe120a0SFrank Holton "Snapshot src from another FS"); 166923ad5b17SKusanagi Kouichi ret = -EXDEV; 1670d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1671d0242061SDavid Sterba /* 1672d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1673d0242061SDavid Sterba * are limited to own subvolumes only 1674d0242061SDavid Sterba */ 1675d0242061SDavid Sterba ret = -EPERM; 1676ecd18815SAl Viro } else { 167772fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 167872fd032eSSage Weil BTRFS_I(src_inode)->root, 16796f72c7e2SArne Jansen transid, readonly, inherit); 1680ecd18815SAl Viro } 16812903ff01SAl Viro fdput(src); 1682cb8e7090SChristoph Hellwig } 1683a874a63eSLiu Bo out_drop_write: 1684a874a63eSLiu Bo mnt_drop_write_file(file); 1685f46b5a66SChristoph Hellwig out: 168672fd032eSSage Weil return ret; 168772fd032eSSage Weil } 168872fd032eSSage Weil 168972fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1690fa0d2b9bSLi Zefan void __user *arg, int subvol) 169172fd032eSSage Weil { 1692fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 169372fd032eSSage Weil int ret; 169472fd032eSSage Weil 1695325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1696325c50e3SJeff Mahoney return -ENOTDIR; 1697325c50e3SJeff Mahoney 1698fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1699fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1700fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1701fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1702fa0d2b9bSLi Zefan 1703fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1704b83cc969SLi Zefan vol_args->fd, subvol, 17056f72c7e2SArne Jansen NULL, false, NULL); 1706fa0d2b9bSLi Zefan 1707fa0d2b9bSLi Zefan kfree(vol_args); 1708fa0d2b9bSLi Zefan return ret; 1709fa0d2b9bSLi Zefan } 1710fa0d2b9bSLi Zefan 1711fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1712fa0d2b9bSLi Zefan void __user *arg, int subvol) 1713fa0d2b9bSLi Zefan { 1714fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1715fa0d2b9bSLi Zefan int ret; 1716fdfb1e4fSLi Zefan u64 transid = 0; 1717fdfb1e4fSLi Zefan u64 *ptr = NULL; 1718b83cc969SLi Zefan bool readonly = false; 17196f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 172072fd032eSSage Weil 1721325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1722325c50e3SJeff Mahoney return -ENOTDIR; 1723325c50e3SJeff Mahoney 1724fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1725fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1726fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1727fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1728fdfb1e4fSLi Zefan 1729b83cc969SLi Zefan if (vol_args->flags & 17306f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 17316f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1732b83cc969SLi Zefan ret = -EOPNOTSUPP; 1733c47ca32dSDan Carpenter goto free_args; 1734fdfb1e4fSLi Zefan } 1735fdfb1e4fSLi Zefan 1736fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1737fdfb1e4fSLi Zefan ptr = &transid; 1738b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1739b83cc969SLi Zefan readonly = true; 17406f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 174109cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 17426f72c7e2SArne Jansen ret = -EINVAL; 1743c47ca32dSDan Carpenter goto free_args; 17446f72c7e2SArne Jansen } 17456f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 17466f72c7e2SArne Jansen if (IS_ERR(inherit)) { 17476f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1748c47ca32dSDan Carpenter goto free_args; 17496f72c7e2SArne Jansen } 17506f72c7e2SArne Jansen } 175175eaa0e2SSage Weil 1752fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 17536f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17548696c533SMiao Xie readonly, inherit); 1755c47ca32dSDan Carpenter if (ret) 1756c47ca32dSDan Carpenter goto free_inherit; 175775eaa0e2SSage Weil 1758c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1759fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1760c47ca32dSDan Carpenter transid), 1761c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 176275eaa0e2SSage Weil ret = -EFAULT; 1763c47ca32dSDan Carpenter 1764c47ca32dSDan Carpenter free_inherit: 17656f72c7e2SArne Jansen kfree(inherit); 1766c47ca32dSDan Carpenter free_args: 1767c47ca32dSDan Carpenter kfree(vol_args); 1768f46b5a66SChristoph Hellwig return ret; 1769f46b5a66SChristoph Hellwig } 1770f46b5a66SChristoph Hellwig 17710caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17720caa102dSLi Zefan void __user *arg) 17730caa102dSLi Zefan { 1774496ad9aaSAl Viro struct inode *inode = file_inode(file); 17750b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17760caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17770caa102dSLi Zefan int ret = 0; 17780caa102dSLi Zefan u64 flags = 0; 17790caa102dSLi Zefan 17804a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 17810caa102dSLi Zefan return -EINVAL; 17820caa102dSLi Zefan 17830b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 17840caa102dSLi Zefan if (btrfs_root_readonly(root)) 17850caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17860b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 17870caa102dSLi Zefan 17880caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 17890caa102dSLi Zefan ret = -EFAULT; 17900caa102dSLi Zefan 17910caa102dSLi Zefan return ret; 17920caa102dSLi Zefan } 17930caa102dSLi Zefan 17940caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 17950caa102dSLi Zefan void __user *arg) 17960caa102dSLi Zefan { 1797496ad9aaSAl Viro struct inode *inode = file_inode(file); 17980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17990caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 18000caa102dSLi Zefan struct btrfs_trans_handle *trans; 18010caa102dSLi Zefan u64 root_flags; 18020caa102dSLi Zefan u64 flags; 18030caa102dSLi Zefan int ret = 0; 18040caa102dSLi Zefan 1805bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1806bd60ea0fSDavid Sterba return -EPERM; 1807bd60ea0fSDavid Sterba 1808b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1809b9ca0664SLiu Bo if (ret) 1810b9ca0664SLiu Bo goto out; 18110caa102dSLi Zefan 18124a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1813b9ca0664SLiu Bo ret = -EINVAL; 1814b9ca0664SLiu Bo goto out_drop_write; 1815b9ca0664SLiu Bo } 18160caa102dSLi Zefan 1817b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1818b9ca0664SLiu Bo ret = -EFAULT; 1819b9ca0664SLiu Bo goto out_drop_write; 1820b9ca0664SLiu Bo } 18210caa102dSLi Zefan 1822b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1823b9ca0664SLiu Bo ret = -EINVAL; 1824b9ca0664SLiu Bo goto out_drop_write; 1825b9ca0664SLiu Bo } 18260caa102dSLi Zefan 1827b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1828b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1829b9ca0664SLiu Bo goto out_drop_write; 1830b9ca0664SLiu Bo } 18310caa102dSLi Zefan 18320b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 18330caa102dSLi Zefan 18340caa102dSLi Zefan /* nothing to do */ 18350caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1836b9ca0664SLiu Bo goto out_drop_sem; 18370caa102dSLi Zefan 18380caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 18392c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 18400caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18410caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 18422c686537SDavid Sterba } else { 18432c686537SDavid Sterba /* 18442c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 18452c686537SDavid Sterba * send 18462c686537SDavid Sterba */ 18472c686537SDavid Sterba spin_lock(&root->root_item_lock); 18482c686537SDavid Sterba if (root->send_in_progress == 0) { 18490caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18500caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 18512c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18522c686537SDavid Sterba } else { 18532c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18540b246afaSJeff Mahoney btrfs_warn(fs_info, 18552c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18562c686537SDavid Sterba root->root_key.objectid); 18572c686537SDavid Sterba ret = -EPERM; 18582c686537SDavid Sterba goto out_drop_sem; 18592c686537SDavid Sterba } 18602c686537SDavid Sterba } 18610caa102dSLi Zefan 18620caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18630caa102dSLi Zefan if (IS_ERR(trans)) { 18640caa102dSLi Zefan ret = PTR_ERR(trans); 18650caa102dSLi Zefan goto out_reset; 18660caa102dSLi Zefan } 18670caa102dSLi Zefan 18680b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 18690caa102dSLi Zefan &root->root_key, &root->root_item); 18700caa102dSLi Zefan 18713a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 18720caa102dSLi Zefan out_reset: 18730caa102dSLi Zefan if (ret) 18740caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1875b9ca0664SLiu Bo out_drop_sem: 18760b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1877b9ca0664SLiu Bo out_drop_write: 1878b9ca0664SLiu Bo mnt_drop_write_file(file); 1879b9ca0664SLiu Bo out: 18800caa102dSLi Zefan return ret; 18810caa102dSLi Zefan } 18820caa102dSLi Zefan 188376dda93cSYan, Zheng /* 188476dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 188576dda93cSYan, Zheng */ 188676dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 188776dda93cSYan, Zheng { 18880b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 188976dda93cSYan, Zheng struct btrfs_path *path; 1890175a2b87SJosef Bacik struct btrfs_dir_item *di; 189176dda93cSYan, Zheng struct btrfs_key key; 1892175a2b87SJosef Bacik u64 dir_id; 189376dda93cSYan, Zheng int ret; 189476dda93cSYan, Zheng 189576dda93cSYan, Zheng path = btrfs_alloc_path(); 189676dda93cSYan, Zheng if (!path) 189776dda93cSYan, Zheng return -ENOMEM; 189876dda93cSYan, Zheng 1899175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 19000b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 19010b246afaSJeff Mahoney di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path, 1902175a2b87SJosef Bacik dir_id, "default", 7, 0); 1903175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1904175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1905175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 190672de6b53SGuangyu Sun ret = -EPERM; 19070b246afaSJeff Mahoney btrfs_err(fs_info, 19085d163e0eSJeff Mahoney "deleting default subvolume %llu is not allowed", 19095d163e0eSJeff Mahoney key.objectid); 1910175a2b87SJosef Bacik goto out; 1911175a2b87SJosef Bacik } 1912175a2b87SJosef Bacik btrfs_release_path(path); 1913175a2b87SJosef Bacik } 1914175a2b87SJosef Bacik 191576dda93cSYan, Zheng key.objectid = root->root_key.objectid; 191676dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 191776dda93cSYan, Zheng key.offset = (u64)-1; 191876dda93cSYan, Zheng 19190b246afaSJeff Mahoney ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 192076dda93cSYan, Zheng if (ret < 0) 192176dda93cSYan, Zheng goto out; 192276dda93cSYan, Zheng BUG_ON(ret == 0); 192376dda93cSYan, Zheng 192476dda93cSYan, Zheng ret = 0; 192576dda93cSYan, Zheng if (path->slots[0] > 0) { 192676dda93cSYan, Zheng path->slots[0]--; 192776dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 192876dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 192976dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 193076dda93cSYan, Zheng ret = -ENOTEMPTY; 193176dda93cSYan, Zheng } 193276dda93cSYan, Zheng out: 193376dda93cSYan, Zheng btrfs_free_path(path); 193476dda93cSYan, Zheng return ret; 193576dda93cSYan, Zheng } 193676dda93cSYan, Zheng 1937ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1938ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1939ac8e9819SChris Mason { 1940abc6e134SChris Mason struct btrfs_key test; 1941abc6e134SChris Mason int ret; 1942abc6e134SChris Mason 1943abc6e134SChris Mason test.objectid = sk->min_objectid; 1944abc6e134SChris Mason test.type = sk->min_type; 1945abc6e134SChris Mason test.offset = sk->min_offset; 1946abc6e134SChris Mason 1947abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1948abc6e134SChris Mason if (ret < 0) 1949ac8e9819SChris Mason return 0; 1950abc6e134SChris Mason 1951abc6e134SChris Mason test.objectid = sk->max_objectid; 1952abc6e134SChris Mason test.type = sk->max_type; 1953abc6e134SChris Mason test.offset = sk->max_offset; 1954abc6e134SChris Mason 1955abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1956abc6e134SChris Mason if (ret > 0) 1957ac8e9819SChris Mason return 0; 1958ac8e9819SChris Mason return 1; 1959ac8e9819SChris Mason } 1960ac8e9819SChris Mason 1961df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 1962ac8e9819SChris Mason struct btrfs_key *key, 1963ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19649b6e817dSGerhard Heift size_t *buf_size, 1965ba346b35SGerhard Heift char __user *ubuf, 1966ac8e9819SChris Mason unsigned long *sk_offset, 1967ac8e9819SChris Mason int *num_found) 1968ac8e9819SChris Mason { 1969ac8e9819SChris Mason u64 found_transid; 1970ac8e9819SChris Mason struct extent_buffer *leaf; 1971ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1972dd81d459SNaohiro Aota struct btrfs_key test; 1973ac8e9819SChris Mason unsigned long item_off; 1974ac8e9819SChris Mason unsigned long item_len; 1975ac8e9819SChris Mason int nritems; 1976ac8e9819SChris Mason int i; 1977ac8e9819SChris Mason int slot; 1978ac8e9819SChris Mason int ret = 0; 1979ac8e9819SChris Mason 1980ac8e9819SChris Mason leaf = path->nodes[0]; 1981ac8e9819SChris Mason slot = path->slots[0]; 1982ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1983ac8e9819SChris Mason 1984ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1985ac8e9819SChris Mason i = nritems; 1986ac8e9819SChris Mason goto advance_key; 1987ac8e9819SChris Mason } 1988ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1989ac8e9819SChris Mason 1990ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1991ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1992ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1993ac8e9819SChris Mason 199403b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 199503b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 199603b71c6cSGabriel de Perthuis continue; 199703b71c6cSGabriel de Perthuis 19989b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 19998f5f6178SGerhard Heift if (*num_found) { 2000ac8e9819SChris Mason ret = 1; 20018f5f6178SGerhard Heift goto out; 20028f5f6178SGerhard Heift } 20038f5f6178SGerhard Heift 20048f5f6178SGerhard Heift /* 20058f5f6178SGerhard Heift * return one empty item back for v1, which does not 20068f5f6178SGerhard Heift * handle -EOVERFLOW 20078f5f6178SGerhard Heift */ 20088f5f6178SGerhard Heift 20099b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2010ac8e9819SChris Mason item_len = 0; 20118f5f6178SGerhard Heift ret = -EOVERFLOW; 20128f5f6178SGerhard Heift } 2013ac8e9819SChris Mason 20149b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2015ac8e9819SChris Mason ret = 1; 201625c9bc2eSGerhard Heift goto out; 2017ac8e9819SChris Mason } 2018ac8e9819SChris Mason 2019ac8e9819SChris Mason sh.objectid = key->objectid; 2020ac8e9819SChris Mason sh.offset = key->offset; 2021ac8e9819SChris Mason sh.type = key->type; 2022ac8e9819SChris Mason sh.len = item_len; 2023ac8e9819SChris Mason sh.transid = found_transid; 2024ac8e9819SChris Mason 2025ac8e9819SChris Mason /* copy search result header */ 2026ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2027ba346b35SGerhard Heift ret = -EFAULT; 2028ba346b35SGerhard Heift goto out; 2029ba346b35SGerhard Heift } 2030ba346b35SGerhard Heift 2031ac8e9819SChris Mason *sk_offset += sizeof(sh); 2032ac8e9819SChris Mason 2033ac8e9819SChris Mason if (item_len) { 2034ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2035ac8e9819SChris Mason /* copy the item */ 2036ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2037ba346b35SGerhard Heift item_off, item_len)) { 2038ba346b35SGerhard Heift ret = -EFAULT; 2039ba346b35SGerhard Heift goto out; 2040ba346b35SGerhard Heift } 2041ba346b35SGerhard Heift 2042ac8e9819SChris Mason *sk_offset += item_len; 2043ac8e9819SChris Mason } 2044e2156867SHugo Mills (*num_found)++; 2045ac8e9819SChris Mason 20468f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20478f5f6178SGerhard Heift goto out; 20488f5f6178SGerhard Heift 204925c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 205025c9bc2eSGerhard Heift ret = 1; 205125c9bc2eSGerhard Heift goto out; 205225c9bc2eSGerhard Heift } 2053ac8e9819SChris Mason } 2054ac8e9819SChris Mason advance_key: 2055ac8e9819SChris Mason ret = 0; 2056dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2057dd81d459SNaohiro Aota test.type = sk->max_type; 2058dd81d459SNaohiro Aota test.offset = sk->max_offset; 2059dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2060dd81d459SNaohiro Aota ret = 1; 2061dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2062abc6e134SChris Mason key->offset++; 2063dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2064abc6e134SChris Mason key->offset = 0; 2065abc6e134SChris Mason key->type++; 2066dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2067abc6e134SChris Mason key->offset = 0; 2068abc6e134SChris Mason key->type = 0; 2069abc6e134SChris Mason key->objectid++; 2070abc6e134SChris Mason } else 2071abc6e134SChris Mason ret = 1; 207225c9bc2eSGerhard Heift out: 2073ba346b35SGerhard Heift /* 2074ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2075ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2076ba346b35SGerhard Heift * * all items were found 2077ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2078ba346b35SGerhard Heift * leaf 2079ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2080ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2081ba346b35SGerhard Heift */ 2082ac8e9819SChris Mason return ret; 2083ac8e9819SChris Mason } 2084ac8e9819SChris Mason 2085ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 208612544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 20879b6e817dSGerhard Heift size_t *buf_size, 2088ba346b35SGerhard Heift char __user *ubuf) 2089ac8e9819SChris Mason { 20900b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2091ac8e9819SChris Mason struct btrfs_root *root; 2092ac8e9819SChris Mason struct btrfs_key key; 2093ac8e9819SChris Mason struct btrfs_path *path; 2094ac8e9819SChris Mason int ret; 2095ac8e9819SChris Mason int num_found = 0; 2096ac8e9819SChris Mason unsigned long sk_offset = 0; 2097ac8e9819SChris Mason 20989b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 20999b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 210012544442SGerhard Heift return -EOVERFLOW; 21019b6e817dSGerhard Heift } 210212544442SGerhard Heift 2103ac8e9819SChris Mason path = btrfs_alloc_path(); 2104ac8e9819SChris Mason if (!path) 2105ac8e9819SChris Mason return -ENOMEM; 2106ac8e9819SChris Mason 2107ac8e9819SChris Mason if (sk->tree_id == 0) { 2108ac8e9819SChris Mason /* search the root of the inode that was passed */ 2109ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2110ac8e9819SChris Mason } else { 2111ac8e9819SChris Mason key.objectid = sk->tree_id; 2112ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2113ac8e9819SChris Mason key.offset = (u64)-1; 2114ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2115ac8e9819SChris Mason if (IS_ERR(root)) { 2116ac8e9819SChris Mason btrfs_free_path(path); 2117ac8e9819SChris Mason return -ENOENT; 2118ac8e9819SChris Mason } 2119ac8e9819SChris Mason } 2120ac8e9819SChris Mason 2121ac8e9819SChris Mason key.objectid = sk->min_objectid; 2122ac8e9819SChris Mason key.type = sk->min_type; 2123ac8e9819SChris Mason key.offset = sk->min_offset; 2124ac8e9819SChris Mason 2125ac8e9819SChris Mason while (1) { 21266174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2127ac8e9819SChris Mason if (ret != 0) { 2128ac8e9819SChris Mason if (ret > 0) 2129ac8e9819SChris Mason ret = 0; 2130ac8e9819SChris Mason goto err; 2131ac8e9819SChris Mason } 2132df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2133ac8e9819SChris Mason &sk_offset, &num_found); 2134b3b4aa74SDavid Sterba btrfs_release_path(path); 213525c9bc2eSGerhard Heift if (ret) 2136ac8e9819SChris Mason break; 2137ac8e9819SChris Mason 2138ac8e9819SChris Mason } 21398f5f6178SGerhard Heift if (ret > 0) 2140ac8e9819SChris Mason ret = 0; 2141ac8e9819SChris Mason err: 2142ac8e9819SChris Mason sk->nr_items = num_found; 2143ac8e9819SChris Mason btrfs_free_path(path); 2144ac8e9819SChris Mason return ret; 2145ac8e9819SChris Mason } 2146ac8e9819SChris Mason 2147ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2148ac8e9819SChris Mason void __user *argp) 2149ac8e9819SChris Mason { 2150ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2151ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2152ac8e9819SChris Mason struct inode *inode; 2153ac8e9819SChris Mason int ret; 21549b6e817dSGerhard Heift size_t buf_size; 2155ac8e9819SChris Mason 2156ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2157ac8e9819SChris Mason return -EPERM; 2158ac8e9819SChris Mason 2159ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2160ac8e9819SChris Mason 2161ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2162ba346b35SGerhard Heift return -EFAULT; 2163ba346b35SGerhard Heift 2164ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2165ac8e9819SChris Mason 2166496ad9aaSAl Viro inode = file_inode(file); 2167ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21688f5f6178SGerhard Heift 21698f5f6178SGerhard Heift /* 21708f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21718f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21728f5f6178SGerhard Heift */ 21738f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21748f5f6178SGerhard Heift ret = 0; 21758f5f6178SGerhard Heift 2176ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2177ac8e9819SChris Mason ret = -EFAULT; 2178ac8e9819SChris Mason return ret; 2179ac8e9819SChris Mason } 2180ac8e9819SChris Mason 2181cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2182cc68a8a5SGerhard Heift void __user *argp) 2183cc68a8a5SGerhard Heift { 2184cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2185cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2186cc68a8a5SGerhard Heift struct inode *inode; 2187cc68a8a5SGerhard Heift int ret; 2188cc68a8a5SGerhard Heift size_t buf_size; 2189ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2190cc68a8a5SGerhard Heift 2191cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2192cc68a8a5SGerhard Heift return -EPERM; 2193cc68a8a5SGerhard Heift 2194cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2195cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2196cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2197cc68a8a5SGerhard Heift return -EFAULT; 2198cc68a8a5SGerhard Heift 2199cc68a8a5SGerhard Heift buf_size = args.buf_size; 2200cc68a8a5SGerhard Heift 2201cc68a8a5SGerhard Heift if (buf_size < sizeof(struct btrfs_ioctl_search_header)) 2202cc68a8a5SGerhard Heift return -EOVERFLOW; 2203cc68a8a5SGerhard Heift 2204cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2205cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2206cc68a8a5SGerhard Heift buf_size = buf_limit; 2207cc68a8a5SGerhard Heift 2208cc68a8a5SGerhard Heift inode = file_inode(file); 2209cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2210cc68a8a5SGerhard Heift (char *)(&uarg->buf[0])); 2211cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2212cc68a8a5SGerhard Heift ret = -EFAULT; 2213cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2214cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2215cc68a8a5SGerhard Heift ret = -EFAULT; 2216cc68a8a5SGerhard Heift 221776dda93cSYan, Zheng return ret; 221876dda93cSYan, Zheng } 221976dda93cSYan, Zheng 222098d377a0STARUISI Hiroaki /* 2221ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2222ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 222398d377a0STARUISI Hiroaki */ 222498d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 222598d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 222698d377a0STARUISI Hiroaki { 222798d377a0STARUISI Hiroaki struct btrfs_root *root; 222898d377a0STARUISI Hiroaki struct btrfs_key key; 2229ac8e9819SChris Mason char *ptr; 223098d377a0STARUISI Hiroaki int ret = -1; 223198d377a0STARUISI Hiroaki int slot; 223298d377a0STARUISI Hiroaki int len; 223398d377a0STARUISI Hiroaki int total_len = 0; 223498d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 223598d377a0STARUISI Hiroaki struct extent_buffer *l; 223698d377a0STARUISI Hiroaki struct btrfs_path *path; 223798d377a0STARUISI Hiroaki 223898d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 223998d377a0STARUISI Hiroaki name[0]='\0'; 224098d377a0STARUISI Hiroaki return 0; 224198d377a0STARUISI Hiroaki } 224298d377a0STARUISI Hiroaki 224398d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 224498d377a0STARUISI Hiroaki if (!path) 224598d377a0STARUISI Hiroaki return -ENOMEM; 224698d377a0STARUISI Hiroaki 2247ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 224898d377a0STARUISI Hiroaki 224998d377a0STARUISI Hiroaki key.objectid = tree_id; 225098d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 225198d377a0STARUISI Hiroaki key.offset = (u64)-1; 225298d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 225398d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2254f14d104dSDavid Sterba btrfs_err(info, "could not find root %llu", tree_id); 22558ad6fcabSChris Mason ret = -ENOENT; 22568ad6fcabSChris Mason goto out; 225798d377a0STARUISI Hiroaki } 225898d377a0STARUISI Hiroaki 225998d377a0STARUISI Hiroaki key.objectid = dirid; 226098d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22618ad6fcabSChris Mason key.offset = (u64)-1; 226298d377a0STARUISI Hiroaki 226398d377a0STARUISI Hiroaki while (1) { 226498d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 226598d377a0STARUISI Hiroaki if (ret < 0) 226698d377a0STARUISI Hiroaki goto out; 226718674c6cSFilipe David Borba Manana else if (ret > 0) { 226818674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 226918674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 227018674c6cSFilipe David Borba Manana if (ret < 0) 227118674c6cSFilipe David Borba Manana goto out; 227218674c6cSFilipe David Borba Manana else if (ret > 0) { 2273ac8e9819SChris Mason ret = -ENOENT; 227498d377a0STARUISI Hiroaki goto out; 2275ac8e9819SChris Mason } 227618674c6cSFilipe David Borba Manana } 227718674c6cSFilipe David Borba Manana 227818674c6cSFilipe David Borba Manana l = path->nodes[0]; 227918674c6cSFilipe David Borba Manana slot = path->slots[0]; 228018674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 228198d377a0STARUISI Hiroaki 228298d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 228398d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 228498d377a0STARUISI Hiroaki ptr -= len + 1; 228598d377a0STARUISI Hiroaki total_len += len + 1; 2286a696cf35SFilipe David Borba Manana if (ptr < name) { 2287a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 228898d377a0STARUISI Hiroaki goto out; 2289a696cf35SFilipe David Borba Manana } 229098d377a0STARUISI Hiroaki 229198d377a0STARUISI Hiroaki *(ptr + len) = '/'; 229298d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 229398d377a0STARUISI Hiroaki 229498d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 229598d377a0STARUISI Hiroaki break; 229698d377a0STARUISI Hiroaki 2297b3b4aa74SDavid Sterba btrfs_release_path(path); 229898d377a0STARUISI Hiroaki key.objectid = key.offset; 22998ad6fcabSChris Mason key.offset = (u64)-1; 230098d377a0STARUISI Hiroaki dirid = key.objectid; 230198d377a0STARUISI Hiroaki } 230277906a50SLi Zefan memmove(name, ptr, total_len); 230398d377a0STARUISI Hiroaki name[total_len] = '\0'; 230498d377a0STARUISI Hiroaki ret = 0; 230598d377a0STARUISI Hiroaki out: 230698d377a0STARUISI Hiroaki btrfs_free_path(path); 2307ac8e9819SChris Mason return ret; 2308ac8e9819SChris Mason } 2309ac8e9819SChris Mason 2310ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2311ac8e9819SChris Mason void __user *argp) 2312ac8e9819SChris Mason { 2313ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2314ac8e9819SChris Mason struct inode *inode; 231501b810b8SDavid Sterba int ret = 0; 2316ac8e9819SChris Mason 23172354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 23182354d08fSJulia Lawall if (IS_ERR(args)) 23192354d08fSJulia Lawall return PTR_ERR(args); 2320c2b96929SDan Carpenter 2321496ad9aaSAl Viro inode = file_inode(file); 2322ac8e9819SChris Mason 232301b810b8SDavid Sterba /* 232401b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 232501b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 232601b810b8SDavid Sterba */ 23271b53ac4dSChris Mason if (args->treeid == 0) 23281b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 23291b53ac4dSChris Mason 233001b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 233101b810b8SDavid Sterba args->name[0] = 0; 233201b810b8SDavid Sterba goto out; 233301b810b8SDavid Sterba } 233401b810b8SDavid Sterba 233501b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 233601b810b8SDavid Sterba ret = -EPERM; 233701b810b8SDavid Sterba goto out; 233801b810b8SDavid Sterba } 233901b810b8SDavid Sterba 2340ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2341ac8e9819SChris Mason args->treeid, args->objectid, 2342ac8e9819SChris Mason args->name); 2343ac8e9819SChris Mason 234401b810b8SDavid Sterba out: 2345ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2346ac8e9819SChris Mason ret = -EFAULT; 2347ac8e9819SChris Mason 2348ac8e9819SChris Mason kfree(args); 234998d377a0STARUISI Hiroaki return ret; 235098d377a0STARUISI Hiroaki } 235198d377a0STARUISI Hiroaki 235276dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 235376dda93cSYan, Zheng void __user *arg) 235476dda93cSYan, Zheng { 235554563d41SAl Viro struct dentry *parent = file->f_path.dentry; 23560b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 235776dda93cSYan, Zheng struct dentry *dentry; 23582b0143b5SDavid Howells struct inode *dir = d_inode(parent); 235976dda93cSYan, Zheng struct inode *inode; 236076dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 236176dda93cSYan, Zheng struct btrfs_root *dest = NULL; 236276dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 236376dda93cSYan, Zheng struct btrfs_trans_handle *trans; 2364c58aaad2SMiao Xie struct btrfs_block_rsv block_rsv; 2365521e0546SDavid Sterba u64 root_flags; 2366c58aaad2SMiao Xie u64 qgroup_reserved; 236776dda93cSYan, Zheng int namelen; 236876dda93cSYan, Zheng int ret; 236976dda93cSYan, Zheng int err = 0; 237076dda93cSYan, Zheng 2371325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2372325c50e3SJeff Mahoney return -ENOTDIR; 2373325c50e3SJeff Mahoney 237476dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 237576dda93cSYan, Zheng if (IS_ERR(vol_args)) 237676dda93cSYan, Zheng return PTR_ERR(vol_args); 237776dda93cSYan, Zheng 237876dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 237976dda93cSYan, Zheng namelen = strlen(vol_args->name); 238076dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 238176dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 238276dda93cSYan, Zheng err = -EINVAL; 238376dda93cSYan, Zheng goto out; 238476dda93cSYan, Zheng } 238576dda93cSYan, Zheng 2386a561be71SAl Viro err = mnt_want_write_file(file); 238776dda93cSYan, Zheng if (err) 238876dda93cSYan, Zheng goto out; 238976dda93cSYan, Zheng 2390521e0546SDavid Sterba 239100235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 239200235411SAl Viro if (err == -EINTR) 239300235411SAl Viro goto out_drop_write; 239476dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 239576dda93cSYan, Zheng if (IS_ERR(dentry)) { 239676dda93cSYan, Zheng err = PTR_ERR(dentry); 239776dda93cSYan, Zheng goto out_unlock_dir; 239876dda93cSYan, Zheng } 239976dda93cSYan, Zheng 24002b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 240176dda93cSYan, Zheng err = -ENOENT; 240276dda93cSYan, Zheng goto out_dput; 240376dda93cSYan, Zheng } 240476dda93cSYan, Zheng 24052b0143b5SDavid Howells inode = d_inode(dentry); 24064260f7c7SSage Weil dest = BTRFS_I(inode)->root; 24074260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 24084260f7c7SSage Weil /* 24094260f7c7SSage Weil * Regular user. Only allow this with a special mount 24104260f7c7SSage Weil * option, when the user has write+exec access to the 24114260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 24124260f7c7SSage Weil * allowed. 24134260f7c7SSage Weil * 24144260f7c7SSage Weil * Note that this is _not_ check that the subvol is 24154260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 24164260f7c7SSage Weil * otherwise be able to delete. 24174260f7c7SSage Weil * 24184260f7c7SSage Weil * Users who want to delete empty subvols should try 24194260f7c7SSage Weil * rmdir(2). 24204260f7c7SSage Weil */ 24214260f7c7SSage Weil err = -EPERM; 24220b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 24234260f7c7SSage Weil goto out_dput; 24244260f7c7SSage Weil 24254260f7c7SSage Weil /* 24264260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 24274260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 24284260f7c7SSage Weil * must be called on the dentry referencing the root 24294260f7c7SSage Weil * of the subvol, not a random directory contained 24304260f7c7SSage Weil * within it. 24314260f7c7SSage Weil */ 24324260f7c7SSage Weil err = -EINVAL; 24334260f7c7SSage Weil if (root == dest) 24344260f7c7SSage Weil goto out_dput; 24354260f7c7SSage Weil 24364260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 24374260f7c7SSage Weil if (err) 24384260f7c7SSage Weil goto out_dput; 24395c39da5bSMiao Xie } 24404260f7c7SSage Weil 24415c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24424260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24434260f7c7SSage Weil if (err) 24444260f7c7SSage Weil goto out_dput; 24454260f7c7SSage Weil 24464a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 244776dda93cSYan, Zheng err = -EINVAL; 244876dda93cSYan, Zheng goto out_dput; 244976dda93cSYan, Zheng } 245076dda93cSYan, Zheng 24515955102cSAl Viro inode_lock(inode); 2452521e0546SDavid Sterba 2453521e0546SDavid Sterba /* 2454521e0546SDavid Sterba * Don't allow to delete a subvolume with send in progress. This is 2455521e0546SDavid Sterba * inside the i_mutex so the error handling that has to drop the bit 2456521e0546SDavid Sterba * again is not run concurrently. 2457521e0546SDavid Sterba */ 2458521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2459c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2460c55bfa67SFilipe Manana if (dest->send_in_progress == 0) { 2461c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2462521e0546SDavid Sterba root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2463521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2464521e0546SDavid Sterba } else { 2465521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 24660b246afaSJeff Mahoney btrfs_warn(fs_info, 2467521e0546SDavid Sterba "Attempt to delete subvolume %llu during send", 2468c55bfa67SFilipe Manana dest->root_key.objectid); 2469521e0546SDavid Sterba err = -EPERM; 2470909e26dcSOmar Sandoval goto out_unlock_inode; 2471521e0546SDavid Sterba } 2472521e0546SDavid Sterba 24730b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 247476dda93cSYan, Zheng 247576dda93cSYan, Zheng err = may_destroy_subvol(dest); 247676dda93cSYan, Zheng if (err) 247776dda93cSYan, Zheng goto out_up_write; 247876dda93cSYan, Zheng 2479c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2480c58aaad2SMiao Xie /* 2481c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2482c58aaad2SMiao Xie * ref/backref. 2483c58aaad2SMiao Xie */ 2484c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2485ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2486c58aaad2SMiao Xie if (err) 2487c58aaad2SMiao Xie goto out_up_write; 2488c58aaad2SMiao Xie 2489a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2490a22285a6SYan, Zheng if (IS_ERR(trans)) { 2491a22285a6SYan, Zheng err = PTR_ERR(trans); 2492c58aaad2SMiao Xie goto out_release; 2493a22285a6SYan, Zheng } 2494c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2495c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2496a22285a6SYan, Zheng 249743663557SNikolay Borisov btrfs_record_snapshot_destroy(trans, BTRFS_I(dir)); 24982be63d5cSFilipe Manana 249976dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 250076dda93cSYan, Zheng dest->root_key.objectid, 250176dda93cSYan, Zheng dentry->d_name.name, 250276dda93cSYan, Zheng dentry->d_name.len); 250379787eaaSJeff Mahoney if (ret) { 250479787eaaSJeff Mahoney err = ret; 250566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 250679787eaaSJeff Mahoney goto out_end_trans; 250779787eaaSJeff Mahoney } 250876dda93cSYan, Zheng 250976dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 251076dda93cSYan, Zheng 251176dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 251276dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 251376dda93cSYan, Zheng dest->root_item.drop_level = 0; 251476dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 251576dda93cSYan, Zheng 251627cdeb70SMiao Xie if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 251776dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 25180b246afaSJeff Mahoney fs_info->tree_root, 251976dda93cSYan, Zheng dest->root_key.objectid); 252079787eaaSJeff Mahoney if (ret) { 252166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 252279787eaaSJeff Mahoney err = ret; 252379787eaaSJeff Mahoney goto out_end_trans; 2524d68fc57bSYan, Zheng } 252579787eaaSJeff Mahoney } 2526dd5f9615SStefan Behrens 25270b246afaSJeff Mahoney ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid, 25286bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, 2529dd5f9615SStefan Behrens dest->root_key.objectid); 2530dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 253166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 2532dd5f9615SStefan Behrens err = ret; 2533dd5f9615SStefan Behrens goto out_end_trans; 2534dd5f9615SStefan Behrens } 2535dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 25360b246afaSJeff Mahoney ret = btrfs_uuid_tree_rem(trans, fs_info, 2537dd5f9615SStefan Behrens dest->root_item.received_uuid, 2538dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2539dd5f9615SStefan Behrens dest->root_key.objectid); 2540dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 254166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 2542dd5f9615SStefan Behrens err = ret; 2543dd5f9615SStefan Behrens goto out_end_trans; 2544dd5f9615SStefan Behrens } 2545dd5f9615SStefan Behrens } 2546dd5f9615SStefan Behrens 254779787eaaSJeff Mahoney out_end_trans: 2548c58aaad2SMiao Xie trans->block_rsv = NULL; 2549c58aaad2SMiao Xie trans->bytes_reserved = 0; 25503a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 255179787eaaSJeff Mahoney if (ret && !err) 255279787eaaSJeff Mahoney err = ret; 255376dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2554c58aaad2SMiao Xie out_release: 25557775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 255676dda93cSYan, Zheng out_up_write: 25570b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 2558521e0546SDavid Sterba if (err) { 2559521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2560c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2561c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2562521e0546SDavid Sterba root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2563521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2564521e0546SDavid Sterba } 2565909e26dcSOmar Sandoval out_unlock_inode: 25665955102cSAl Viro inode_unlock(inode); 256776dda93cSYan, Zheng if (!err) { 256864ad6c48SOmar Sandoval d_invalidate(dentry); 256976dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 257076dda93cSYan, Zheng d_delete(dentry); 257161155aa0SDavid Sterba ASSERT(dest->send_in_progress == 0); 2572fa6ac876SLiu Bo 2573fa6ac876SLiu Bo /* the last ref */ 257457cdc8dbSDavid Sterba if (dest->ino_cache_inode) { 257557cdc8dbSDavid Sterba iput(dest->ino_cache_inode); 257657cdc8dbSDavid Sterba dest->ino_cache_inode = NULL; 2577fa6ac876SLiu Bo } 257876dda93cSYan, Zheng } 257976dda93cSYan, Zheng out_dput: 258076dda93cSYan, Zheng dput(dentry); 258176dda93cSYan, Zheng out_unlock_dir: 25825955102cSAl Viro inode_unlock(dir); 258300235411SAl Viro out_drop_write: 25842a79f17eSAl Viro mnt_drop_write_file(file); 258576dda93cSYan, Zheng out: 258676dda93cSYan, Zheng kfree(vol_args); 258776dda93cSYan, Zheng return err; 258876dda93cSYan, Zheng } 258976dda93cSYan, Zheng 25901e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2591f46b5a66SChristoph Hellwig { 2592496ad9aaSAl Viro struct inode *inode = file_inode(file); 2593f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25941e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2595c146afadSYan Zheng int ret; 2596c146afadSYan Zheng 259725122d15SIlya Dryomov ret = mnt_want_write_file(file); 259825122d15SIlya Dryomov if (ret) 259925122d15SIlya Dryomov return ret; 2600b83cc969SLi Zefan 260125122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 260225122d15SIlya Dryomov ret = -EROFS; 260325122d15SIlya Dryomov goto out; 26045ac00addSStefan Behrens } 2605f46b5a66SChristoph Hellwig 2606f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2607f46b5a66SChristoph Hellwig case S_IFDIR: 2608e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2609e441d54dSChris Mason ret = -EPERM; 2610e441d54dSChris Mason goto out; 2611e441d54dSChris Mason } 2612de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2613f46b5a66SChristoph Hellwig break; 2614f46b5a66SChristoph Hellwig case S_IFREG: 2615e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2616e441d54dSChris Mason ret = -EINVAL; 2617e441d54dSChris Mason goto out; 2618e441d54dSChris Mason } 26191e701a32SChris Mason 26201e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 26211e701a32SChris Mason if (!range) { 26221e701a32SChris Mason ret = -ENOMEM; 26231e701a32SChris Mason goto out; 26241e701a32SChris Mason } 26251e701a32SChris Mason 26261e701a32SChris Mason if (argp) { 26271e701a32SChris Mason if (copy_from_user(range, argp, 26281e701a32SChris Mason sizeof(*range))) { 26291e701a32SChris Mason ret = -EFAULT; 26301e701a32SChris Mason kfree(range); 2631683be16eSDan Carpenter goto out; 26321e701a32SChris Mason } 26331e701a32SChris Mason /* compression requires us to start the IO */ 26341e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 26351e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 26361e701a32SChris Mason range->extent_thresh = (u32)-1; 26371e701a32SChris Mason } 26381e701a32SChris Mason } else { 26391e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 26401e701a32SChris Mason range->len = (u64)-1; 26411e701a32SChris Mason } 2642496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 26434cb5300bSChris Mason range, 0, 0); 26444cb5300bSChris Mason if (ret > 0) 26454cb5300bSChris Mason ret = 0; 26461e701a32SChris Mason kfree(range); 2647f46b5a66SChristoph Hellwig break; 26488929ecfaSYan, Zheng default: 26498929ecfaSYan, Zheng ret = -EINVAL; 2650f46b5a66SChristoph Hellwig } 2651e441d54dSChris Mason out: 265225122d15SIlya Dryomov mnt_drop_write_file(file); 2653e441d54dSChris Mason return ret; 2654f46b5a66SChristoph Hellwig } 2655f46b5a66SChristoph Hellwig 26562ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2657f46b5a66SChristoph Hellwig { 2658f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2659f46b5a66SChristoph Hellwig int ret; 2660f46b5a66SChristoph Hellwig 2661e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2662e441d54dSChris Mason return -EPERM; 2663e441d54dSChris Mason 26640b246afaSJeff Mahoney if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) 2665e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2666c9e9f97bSIlya Dryomov 26670b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 2668dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2669c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2670c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2671c9e9f97bSIlya Dryomov goto out; 2672c9e9f97bSIlya Dryomov } 2673f46b5a66SChristoph Hellwig 26745516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 26752ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 2676f46b5a66SChristoph Hellwig 267743d20761SAnand Jain if (!ret) 26780b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 267943d20761SAnand Jain 2680f46b5a66SChristoph Hellwig kfree(vol_args); 2681c9e9f97bSIlya Dryomov out: 26820b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 26830b246afaSJeff Mahoney atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 2684f46b5a66SChristoph Hellwig return ret; 2685f46b5a66SChristoph Hellwig } 2686f46b5a66SChristoph Hellwig 26876b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2688f46b5a66SChristoph Hellwig { 26890b246afaSJeff Mahoney struct inode *inode = file_inode(file); 26900b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 26916b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 2692f46b5a66SChristoph Hellwig int ret; 2693f46b5a66SChristoph Hellwig 2694e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2695e441d54dSChris Mason return -EPERM; 2696e441d54dSChris Mason 2697da24927bSMiao Xie ret = mnt_want_write_file(file); 2698da24927bSMiao Xie if (ret) 2699da24927bSMiao Xie return ret; 2700c146afadSYan Zheng 2701dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2702c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2703c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2704c47ca32dSDan Carpenter goto err_drop; 2705c9e9f97bSIlya Dryomov } 2706f46b5a66SChristoph Hellwig 27076b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 2708735654eaSDavid Sterba if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) 2709735654eaSDavid Sterba return -EOPNOTSUPP; 2710f46b5a66SChristoph Hellwig 27110b246afaSJeff Mahoney if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 2712183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2713183860f6SAnand Jain goto out; 2714183860f6SAnand Jain } 2715183860f6SAnand Jain 27160b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 2717735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 27182ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 27196b526ed7SAnand Jain } else { 27206b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 27212ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 27226b526ed7SAnand Jain } 27230b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 27240b246afaSJeff Mahoney atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 2725183860f6SAnand Jain 27266b526ed7SAnand Jain if (!ret) { 2727735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 27280b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 27296b526ed7SAnand Jain vol_args->devid); 27306b526ed7SAnand Jain else 27310b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 27326b526ed7SAnand Jain vol_args->name); 27336b526ed7SAnand Jain } 2734183860f6SAnand Jain out: 2735183860f6SAnand Jain kfree(vol_args); 2736c47ca32dSDan Carpenter err_drop: 27374ac20c70SIlya Dryomov mnt_drop_write_file(file); 2738f46b5a66SChristoph Hellwig return ret; 2739f46b5a66SChristoph Hellwig } 2740f46b5a66SChristoph Hellwig 2741f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2742f46b5a66SChristoph Hellwig { 27430b246afaSJeff Mahoney struct inode *inode = file_inode(file); 27440b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2745f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2746f46b5a66SChristoph Hellwig int ret; 2747f46b5a66SChristoph Hellwig 2748f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2749f46b5a66SChristoph Hellwig return -EPERM; 2750f46b5a66SChristoph Hellwig 2751f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 2752f46b5a66SChristoph Hellwig if (ret) 2753f46b5a66SChristoph Hellwig return ret; 2754f46b5a66SChristoph Hellwig 27550b246afaSJeff Mahoney if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 2756f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 275758d7bbf8SDavid Sterba goto out_drop_write; 275858d7bbf8SDavid Sterba } 275958d7bbf8SDavid Sterba 276058d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 276158d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 276258d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 2763f46b5a66SChristoph Hellwig goto out; 2764f46b5a66SChristoph Hellwig } 2765f46b5a66SChristoph Hellwig 276658d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 27670b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 27682ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 27690b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 2770f46b5a66SChristoph Hellwig 2771f46b5a66SChristoph Hellwig if (!ret) 27720b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 2773f46b5a66SChristoph Hellwig kfree(vol_args); 277458d7bbf8SDavid Sterba out: 27750b246afaSJeff Mahoney atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 277658d7bbf8SDavid Sterba out_drop_write: 2777f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 277858d7bbf8SDavid Sterba 2779f46b5a66SChristoph Hellwig return ret; 2780f46b5a66SChristoph Hellwig } 2781f46b5a66SChristoph Hellwig 27822ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 27832ff7e61eSJeff Mahoney void __user *arg) 2784475f6387SJan Schmidt { 2785027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2786475f6387SJan Schmidt struct btrfs_device *device; 27870b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2788027ed2f0SLi Zefan int ret = 0; 2789475f6387SJan Schmidt 2790027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2791027ed2f0SLi Zefan if (!fi_args) 2792027ed2f0SLi Zefan return -ENOMEM; 2793027ed2f0SLi Zefan 2794f7171750SFilipe David Borba Manana mutex_lock(&fs_devices->device_list_mutex); 2795027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 27960b246afaSJeff Mahoney memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 2797475f6387SJan Schmidt 2798d7641a49SByongho Lee list_for_each_entry(device, &fs_devices->devices, dev_list) { 2799027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2800027ed2f0SLi Zefan fi_args->max_id = device->devid; 2801475f6387SJan Schmidt } 2802475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2803475f6387SJan Schmidt 28040b246afaSJeff Mahoney fi_args->nodesize = fs_info->super_copy->nodesize; 28050b246afaSJeff Mahoney fi_args->sectorsize = fs_info->super_copy->sectorsize; 28060b246afaSJeff Mahoney fi_args->clone_alignment = fs_info->super_copy->sectorsize; 280780a773fbSDavid Sterba 2808027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2809027ed2f0SLi Zefan ret = -EFAULT; 2810475f6387SJan Schmidt 2811027ed2f0SLi Zefan kfree(fi_args); 2812027ed2f0SLi Zefan return ret; 2813475f6387SJan Schmidt } 2814475f6387SJan Schmidt 28152ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 28162ff7e61eSJeff Mahoney void __user *arg) 2817475f6387SJan Schmidt { 2818475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2819475f6387SJan Schmidt struct btrfs_device *dev; 28200b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2821475f6387SJan Schmidt int ret = 0; 2822475f6387SJan Schmidt char *s_uuid = NULL; 2823475f6387SJan Schmidt 2824475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2825475f6387SJan Schmidt if (IS_ERR(di_args)) 2826475f6387SJan Schmidt return PTR_ERR(di_args); 2827475f6387SJan Schmidt 2828dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2829475f6387SJan Schmidt s_uuid = di_args->uuid; 2830475f6387SJan Schmidt 2831475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 28320b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 2833475f6387SJan Schmidt 2834475f6387SJan Schmidt if (!dev) { 2835475f6387SJan Schmidt ret = -ENODEV; 2836475f6387SJan Schmidt goto out; 2837475f6387SJan Schmidt } 2838475f6387SJan Schmidt 2839475f6387SJan Schmidt di_args->devid = dev->devid; 28407cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 28417cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2842475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2843a27202fbSJim Meyering if (dev->name) { 2844606686eeSJosef Bacik struct rcu_string *name; 2845606686eeSJosef Bacik 2846606686eeSJosef Bacik rcu_read_lock(); 2847606686eeSJosef Bacik name = rcu_dereference(dev->name); 2848606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2849606686eeSJosef Bacik rcu_read_unlock(); 2850a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2851a27202fbSJim Meyering } else { 285299ba55adSStefan Behrens di_args->path[0] = '\0'; 2853a27202fbSJim Meyering } 2854475f6387SJan Schmidt 2855475f6387SJan Schmidt out: 285655793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2857475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2858475f6387SJan Schmidt ret = -EFAULT; 2859475f6387SJan Schmidt 2860475f6387SJan Schmidt kfree(di_args); 2861475f6387SJan Schmidt return ret; 2862475f6387SJan Schmidt } 2863475f6387SJan Schmidt 2864f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2865416161dbSMark Fasheh { 2866416161dbSMark Fasheh struct page *page; 2867416161dbSMark Fasheh 2868416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2869416161dbSMark Fasheh if (!page) 287031314002SFilipe Manana return ERR_PTR(-ENOMEM); 2871416161dbSMark Fasheh 2872416161dbSMark Fasheh if (!PageUptodate(page)) { 287331314002SFilipe Manana int ret; 287431314002SFilipe Manana 287531314002SFilipe Manana ret = btrfs_readpage(NULL, page); 287631314002SFilipe Manana if (ret) 287731314002SFilipe Manana return ERR_PTR(ret); 2878416161dbSMark Fasheh lock_page(page); 2879416161dbSMark Fasheh if (!PageUptodate(page)) { 2880416161dbSMark Fasheh unlock_page(page); 288109cbfeafSKirill A. Shutemov put_page(page); 288231314002SFilipe Manana return ERR_PTR(-EIO); 2883416161dbSMark Fasheh } 288431314002SFilipe Manana if (page->mapping != inode->i_mapping) { 2885416161dbSMark Fasheh unlock_page(page); 288609cbfeafSKirill A. Shutemov put_page(page); 288731314002SFilipe Manana return ERR_PTR(-EAGAIN); 2888416161dbSMark Fasheh } 2889416161dbSMark Fasheh } 2890416161dbSMark Fasheh 2891416161dbSMark Fasheh return page; 2892416161dbSMark Fasheh } 2893416161dbSMark Fasheh 2894f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2895f4414602SMark Fasheh int num_pages, u64 off) 2896f4414602SMark Fasheh { 2897f4414602SMark Fasheh int i; 289809cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 2899f4414602SMark Fasheh 2900f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 290131314002SFilipe Manana again: 2902f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 290331314002SFilipe Manana if (IS_ERR(pages[i])) { 290431314002SFilipe Manana int err = PTR_ERR(pages[i]); 290531314002SFilipe Manana 290631314002SFilipe Manana if (err == -EAGAIN) 290731314002SFilipe Manana goto again; 290831314002SFilipe Manana pages[i] = NULL; 290931314002SFilipe Manana return err; 291031314002SFilipe Manana } 2911f4414602SMark Fasheh } 2912f4414602SMark Fasheh return 0; 2913f4414602SMark Fasheh } 2914f4414602SMark Fasheh 2915e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 2916e0bd70c6SFilipe Manana bool retry_range_locking) 291777fe20dcSMark Fasheh { 2918e0bd70c6SFilipe Manana /* 2919e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 2920e0bd70c6SFilipe Manana * another, and lock file content. 2921e0bd70c6SFilipe Manana * The locking order is: 2922e0bd70c6SFilipe Manana * 2923e0bd70c6SFilipe Manana * 1) pages 2924e0bd70c6SFilipe Manana * 2) range in the inode's io tree 2925e0bd70c6SFilipe Manana */ 292677fe20dcSMark Fasheh while (1) { 292777fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 292877fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 292977fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 293077fe20dcSMark Fasheh off + len - 1); 2931ff5df9b8SFilipe Manana if ((!ordered || 2932ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2933ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 293477fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2935ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2936ff5df9b8SFilipe Manana if (ordered) 2937ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 293877fe20dcSMark Fasheh break; 2939ff5df9b8SFilipe Manana } 294077fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 294177fe20dcSMark Fasheh if (ordered) 294277fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 2943e0bd70c6SFilipe Manana if (!retry_range_locking) 2944e0bd70c6SFilipe Manana return -EAGAIN; 294577fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 294677fe20dcSMark Fasheh } 2947e0bd70c6SFilipe Manana return 0; 294877fe20dcSMark Fasheh } 294977fe20dcSMark Fasheh 2950f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2951416161dbSMark Fasheh { 29525955102cSAl Viro inode_unlock(inode1); 29535955102cSAl Viro inode_unlock(inode2); 2954416161dbSMark Fasheh } 2955416161dbSMark Fasheh 2956f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2957f4414602SMark Fasheh { 2958f4414602SMark Fasheh if (inode1 < inode2) 2959f4414602SMark Fasheh swap(inode1, inode2); 2960f4414602SMark Fasheh 29615955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 29625955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 2963f4414602SMark Fasheh } 2964f4414602SMark Fasheh 2965f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2966f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2967f4414602SMark Fasheh { 2968f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2969f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2970f4414602SMark Fasheh } 2971f4414602SMark Fasheh 2972e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2973e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 2974e0bd70c6SFilipe Manana bool retry_range_locking) 2975416161dbSMark Fasheh { 2976e0bd70c6SFilipe Manana int ret; 2977e0bd70c6SFilipe Manana 2978416161dbSMark Fasheh if (inode1 < inode2) { 2979416161dbSMark Fasheh swap(inode1, inode2); 2980416161dbSMark Fasheh swap(loff1, loff2); 2981416161dbSMark Fasheh } 2982e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 2983e0bd70c6SFilipe Manana if (ret) 2984e0bd70c6SFilipe Manana return ret; 2985e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 2986e0bd70c6SFilipe Manana if (ret) 2987e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 2988e0bd70c6SFilipe Manana loff1 + len - 1); 2989e0bd70c6SFilipe Manana return ret; 2990416161dbSMark Fasheh } 2991f4414602SMark Fasheh 2992f4414602SMark Fasheh struct cmp_pages { 2993f4414602SMark Fasheh int num_pages; 2994f4414602SMark Fasheh struct page **src_pages; 2995f4414602SMark Fasheh struct page **dst_pages; 2996f4414602SMark Fasheh }; 2997f4414602SMark Fasheh 2998f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 2999f4414602SMark Fasheh { 3000f4414602SMark Fasheh int i; 3001f4414602SMark Fasheh struct page *pg; 3002f4414602SMark Fasheh 3003f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 3004f4414602SMark Fasheh pg = cmp->src_pages[i]; 3005e0bd70c6SFilipe Manana if (pg) { 3006e0bd70c6SFilipe Manana unlock_page(pg); 300709cbfeafSKirill A. Shutemov put_page(pg); 3008e0bd70c6SFilipe Manana } 3009f4414602SMark Fasheh pg = cmp->dst_pages[i]; 3010e0bd70c6SFilipe Manana if (pg) { 3011e0bd70c6SFilipe Manana unlock_page(pg); 301209cbfeafSKirill A. Shutemov put_page(pg); 3013f4414602SMark Fasheh } 3014e0bd70c6SFilipe Manana } 3015f4414602SMark Fasheh kfree(cmp->src_pages); 3016f4414602SMark Fasheh kfree(cmp->dst_pages); 3017f4414602SMark Fasheh } 3018f4414602SMark Fasheh 3019f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 3020f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 3021f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 3022f4414602SMark Fasheh { 3023f4414602SMark Fasheh int ret; 302409cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 3025f4414602SMark Fasheh struct page **src_pgarr, **dst_pgarr; 3026f4414602SMark Fasheh 3027f4414602SMark Fasheh /* 3028f4414602SMark Fasheh * We must gather up all the pages before we initiate our 3029f4414602SMark Fasheh * extent locking. We use an array for the page pointers. Size 3030f4414602SMark Fasheh * of the array is bounded by len, which is in turn bounded by 3031f4414602SMark Fasheh * BTRFS_MAX_DEDUPE_LEN. 3032f4414602SMark Fasheh */ 303366722f7cSDavid Sterba src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 303466722f7cSDavid Sterba dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 3035f4414602SMark Fasheh if (!src_pgarr || !dst_pgarr) { 3036f4414602SMark Fasheh kfree(src_pgarr); 3037f4414602SMark Fasheh kfree(dst_pgarr); 3038f4414602SMark Fasheh return -ENOMEM; 3039f4414602SMark Fasheh } 3040f4414602SMark Fasheh cmp->num_pages = num_pages; 3041f4414602SMark Fasheh cmp->src_pages = src_pgarr; 3042f4414602SMark Fasheh cmp->dst_pages = dst_pgarr; 3043f4414602SMark Fasheh 3044b1517622SFilipe Manana /* 3045b1517622SFilipe Manana * If deduping ranges in the same inode, locking rules make it mandatory 3046b1517622SFilipe Manana * to always lock pages in ascending order to avoid deadlocks with 3047b1517622SFilipe Manana * concurrent tasks (such as starting writeback/delalloc). 3048b1517622SFilipe Manana */ 3049b1517622SFilipe Manana if (src == dst && dst_loff < loff) { 3050b1517622SFilipe Manana swap(src_pgarr, dst_pgarr); 3051b1517622SFilipe Manana swap(loff, dst_loff); 3052b1517622SFilipe Manana } 3053b1517622SFilipe Manana 3054b1517622SFilipe Manana ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff); 3055f4414602SMark Fasheh if (ret) 3056f4414602SMark Fasheh goto out; 3057f4414602SMark Fasheh 3058b1517622SFilipe Manana ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff); 3059f4414602SMark Fasheh 3060f4414602SMark Fasheh out: 3061f4414602SMark Fasheh if (ret) 3062f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 3063f4414602SMark Fasheh return 0; 3064416161dbSMark Fasheh } 3065416161dbSMark Fasheh 30661a287cfeSDavid Sterba static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp) 3067416161dbSMark Fasheh { 3068416161dbSMark Fasheh int ret = 0; 3069f4414602SMark Fasheh int i; 3070416161dbSMark Fasheh struct page *src_page, *dst_page; 307109cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 3072416161dbSMark Fasheh void *addr, *dst_addr; 3073416161dbSMark Fasheh 3074f4414602SMark Fasheh i = 0; 3075416161dbSMark Fasheh while (len) { 307609cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 3077416161dbSMark Fasheh cmp_len = len; 3078416161dbSMark Fasheh 3079f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 3080f4414602SMark Fasheh 3081f4414602SMark Fasheh src_page = cmp->src_pages[i]; 3082f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 3083e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 3084e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 3085f4414602SMark Fasheh 3086416161dbSMark Fasheh addr = kmap_atomic(src_page); 3087416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 3088416161dbSMark Fasheh 3089416161dbSMark Fasheh flush_dcache_page(src_page); 3090416161dbSMark Fasheh flush_dcache_page(dst_page); 3091416161dbSMark Fasheh 3092416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 30932b3909f8SDarrick J. Wong ret = -EBADE; 3094416161dbSMark Fasheh 3095416161dbSMark Fasheh kunmap_atomic(addr); 3096416161dbSMark Fasheh kunmap_atomic(dst_addr); 3097416161dbSMark Fasheh 3098416161dbSMark Fasheh if (ret) 3099416161dbSMark Fasheh break; 3100416161dbSMark Fasheh 3101416161dbSMark Fasheh len -= cmp_len; 3102f4414602SMark Fasheh i++; 3103416161dbSMark Fasheh } 3104416161dbSMark Fasheh 3105416161dbSMark Fasheh return ret; 3106416161dbSMark Fasheh } 3107416161dbSMark Fasheh 3108e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3109e1d227a4SMark Fasheh u64 olen) 3110416161dbSMark Fasheh { 3111e1d227a4SMark Fasheh u64 len = *plen; 3112416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3113416161dbSMark Fasheh 3114e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 3115416161dbSMark Fasheh return -EINVAL; 3116e1d227a4SMark Fasheh 3117e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 3118e1d227a4SMark Fasheh if (off + len == inode->i_size) 3119e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 3120e1d227a4SMark Fasheh 3121416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 3122416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3123416161dbSMark Fasheh return -EINVAL; 3124416161dbSMark Fasheh 3125416161dbSMark Fasheh return 0; 3126416161dbSMark Fasheh } 3127416161dbSMark Fasheh 3128e1d227a4SMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 3129416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 3130416161dbSMark Fasheh { 3131416161dbSMark Fasheh int ret; 3132e1d227a4SMark Fasheh u64 len = olen; 3133f4414602SMark Fasheh struct cmp_pages cmp; 3134fc4badd9SOmar Sandoval bool same_inode = (src == dst); 31350efa9f48SMark Fasheh u64 same_lock_start = 0; 31360efa9f48SMark Fasheh u64 same_lock_len = 0; 3137416161dbSMark Fasheh 3138113e8283SFilipe Manana if (len == 0) 3139113e8283SFilipe Manana return 0; 3140113e8283SFilipe Manana 3141fc4badd9SOmar Sandoval if (same_inode) 31425955102cSAl Viro inode_lock(src); 3143fc4badd9SOmar Sandoval else 3144fc4badd9SOmar Sandoval btrfs_double_inode_lock(src, dst); 31450efa9f48SMark Fasheh 31460efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 31470efa9f48SMark Fasheh if (ret) 31480efa9f48SMark Fasheh goto out_unlock; 3149fc4badd9SOmar Sandoval 3150fc4badd9SOmar Sandoval ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3151f4dfe687SFilipe Manana if (ret) 3152f4dfe687SFilipe Manana goto out_unlock; 31530efa9f48SMark Fasheh 3154fc4badd9SOmar Sandoval if (same_inode) { 31550efa9f48SMark Fasheh /* 31560efa9f48SMark Fasheh * Single inode case wants the same checks, except we 31570efa9f48SMark Fasheh * don't want our length pushed out past i_size as 31580efa9f48SMark Fasheh * comparing that data range makes no sense. 31590efa9f48SMark Fasheh * 31600efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 31610efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 31620efa9f48SMark Fasheh * reject the request. 31630efa9f48SMark Fasheh * 31640efa9f48SMark Fasheh * This effectively means we require aligned extents 31650efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 31660efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 31670efa9f48SMark Fasheh * i_size. 31680efa9f48SMark Fasheh */ 31690efa9f48SMark Fasheh if (len != olen) { 31700efa9f48SMark Fasheh ret = -EINVAL; 31710efa9f48SMark Fasheh goto out_unlock; 31720efa9f48SMark Fasheh } 31730efa9f48SMark Fasheh 31740efa9f48SMark Fasheh /* Check for overlapping ranges */ 31750efa9f48SMark Fasheh if (dst_loff + len > loff && dst_loff < loff + len) { 31760efa9f48SMark Fasheh ret = -EINVAL; 31770efa9f48SMark Fasheh goto out_unlock; 31780efa9f48SMark Fasheh } 31790efa9f48SMark Fasheh 31800efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 31810efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 31820efa9f48SMark Fasheh } 3183416161dbSMark Fasheh 3184416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 3185416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3186416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 3187416161dbSMark Fasheh ret = -EINVAL; 3188416161dbSMark Fasheh goto out_unlock; 3189416161dbSMark Fasheh } 3190416161dbSMark Fasheh 3191e0bd70c6SFilipe Manana again: 3192f4414602SMark Fasheh ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 3193f4414602SMark Fasheh if (ret) 3194f4414602SMark Fasheh goto out_unlock; 3195f4414602SMark Fasheh 31960efa9f48SMark Fasheh if (same_inode) 3197e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3198e0bd70c6SFilipe Manana false); 31990efa9f48SMark Fasheh else 3200e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3201e0bd70c6SFilipe Manana false); 3202e0bd70c6SFilipe Manana /* 3203e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3204e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3205e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3206e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3207e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3208e0bd70c6SFilipe Manana * range locking). 3209e0bd70c6SFilipe Manana */ 3210e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3211e0bd70c6SFilipe Manana /* 3212e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3213e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3214e0bd70c6SFilipe Manana */ 3215e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3216e0bd70c6SFilipe Manana if (same_inode) { 3217e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3218e0bd70c6SFilipe Manana same_lock_len); 3219e0bd70c6SFilipe Manana } else { 3220e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3221e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3222e0bd70c6SFilipe Manana } 3223e0bd70c6SFilipe Manana goto again; 3224e0bd70c6SFilipe Manana } 3225e0bd70c6SFilipe Manana ASSERT(ret == 0); 3226e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3227e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3228e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3229e0bd70c6SFilipe Manana return ret; 3230e0bd70c6SFilipe Manana } 3231f4414602SMark Fasheh 3232207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 32331a287cfeSDavid Sterba ret = btrfs_cmp_data(olen, &cmp); 3234416161dbSMark Fasheh if (ret == 0) 32351c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3236416161dbSMark Fasheh 32370efa9f48SMark Fasheh if (same_inode) 32380efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 32390efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 32400efa9f48SMark Fasheh else 3241f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3242f4414602SMark Fasheh 3243f4414602SMark Fasheh btrfs_cmp_data_free(&cmp); 3244416161dbSMark Fasheh out_unlock: 32450efa9f48SMark Fasheh if (same_inode) 32465955102cSAl Viro inode_unlock(src); 32470efa9f48SMark Fasheh else 3248f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3249416161dbSMark Fasheh 3250416161dbSMark Fasheh return ret; 3251416161dbSMark Fasheh } 3252416161dbSMark Fasheh 3253ee22184bSByongho Lee #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3254416161dbSMark Fasheh 32552b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 32562b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3257416161dbSMark Fasheh { 32582b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 32592b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3260416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 32612b3909f8SDarrick J. Wong ssize_t res; 3262416161dbSMark Fasheh 32632b3909f8SDarrick J. Wong if (olen > BTRFS_MAX_DEDUPE_LEN) 32642b3909f8SDarrick J. Wong olen = BTRFS_MAX_DEDUPE_LEN; 3265416161dbSMark Fasheh 326609cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3267416161dbSMark Fasheh /* 3268416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3269416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3270416161dbSMark Fasheh * this situation without an update. 3271416161dbSMark Fasheh */ 32722b3909f8SDarrick J. Wong return -EINVAL; 3273416161dbSMark Fasheh } 3274416161dbSMark Fasheh 32752b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 32762b3909f8SDarrick J. Wong if (res) 32772b3909f8SDarrick J. Wong return res; 32782b3909f8SDarrick J. Wong return olen; 3279416161dbSMark Fasheh } 3280416161dbSMark Fasheh 3281f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3282f82a9901SFilipe Manana struct inode *inode, 3283f82a9901SFilipe Manana u64 endoff, 3284f82a9901SFilipe Manana const u64 destoff, 32851c919a5eSMark Fasheh const u64 olen, 32861c919a5eSMark Fasheh int no_time_update) 3287f82a9901SFilipe Manana { 3288f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3289f82a9901SFilipe Manana int ret; 3290f82a9901SFilipe Manana 3291f82a9901SFilipe Manana inode_inc_iversion(inode); 32921c919a5eSMark Fasheh if (!no_time_update) 3293c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3294f82a9901SFilipe Manana /* 3295f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3296f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3297f82a9901SFilipe Manana */ 3298f82a9901SFilipe Manana if (endoff > destoff + olen) 3299f82a9901SFilipe Manana endoff = destoff + olen; 3300f82a9901SFilipe Manana if (endoff > inode->i_size) 33016ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3302f82a9901SFilipe Manana 3303f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3304f82a9901SFilipe Manana if (ret) { 330566642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 33063a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3307f82a9901SFilipe Manana goto out; 3308f82a9901SFilipe Manana } 33093a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3310f82a9901SFilipe Manana out: 3311f82a9901SFilipe Manana return ret; 3312f82a9901SFilipe Manana } 3313f82a9901SFilipe Manana 3314a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 33157ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 33167ffbb598SFilipe Manana const struct btrfs_path *path, 33177ffbb598SFilipe Manana const u64 hole_offset, 33187ffbb598SFilipe Manana const u64 hole_len) 33197ffbb598SFilipe Manana { 3320a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 33217ffbb598SFilipe Manana struct extent_map *em; 33227ffbb598SFilipe Manana int ret; 33237ffbb598SFilipe Manana 33247ffbb598SFilipe Manana em = alloc_extent_map(); 33257ffbb598SFilipe Manana if (!em) { 3326a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33277ffbb598SFilipe Manana return; 33287ffbb598SFilipe Manana } 33297ffbb598SFilipe Manana 333014f59796SFilipe Manana if (path) { 333114f59796SFilipe Manana struct btrfs_file_extent_item *fi; 333214f59796SFilipe Manana 333314f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 333414f59796SFilipe Manana struct btrfs_file_extent_item); 3335a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 33367ffbb598SFilipe Manana em->generation = -1; 33377ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 33387ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33397ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3340a2f392e4SNikolay Borisov &inode->runtime_flags); 33417ffbb598SFilipe Manana } else { 33427ffbb598SFilipe Manana em->start = hole_offset; 33437ffbb598SFilipe Manana em->len = hole_len; 33447ffbb598SFilipe Manana em->ram_bytes = em->len; 33457ffbb598SFilipe Manana em->orig_start = hole_offset; 33467ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33477ffbb598SFilipe Manana em->block_len = 0; 33487ffbb598SFilipe Manana em->orig_block_len = 0; 33497ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33507ffbb598SFilipe Manana em->generation = trans->transid; 33517ffbb598SFilipe Manana } 33527ffbb598SFilipe Manana 33537ffbb598SFilipe Manana while (1) { 33547ffbb598SFilipe Manana write_lock(&em_tree->lock); 33557ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33567ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33577ffbb598SFilipe Manana if (ret != -EEXIST) { 33587ffbb598SFilipe Manana free_extent_map(em); 33597ffbb598SFilipe Manana break; 33607ffbb598SFilipe Manana } 3361a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 33627ffbb598SFilipe Manana em->start + em->len - 1, 0); 33637ffbb598SFilipe Manana } 33647ffbb598SFilipe Manana 3365ee39b432SDavid Sterba if (ret) 3366a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33677ffbb598SFilipe Manana } 33687ffbb598SFilipe Manana 33698039d87dSFilipe Manana /* 33708039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33718039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33728039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33738039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33748039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33758039d87dSFilipe Manana * 33768039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33778039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33788039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33798039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33808039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33818039d87dSFilipe Manana * i_mutex, which we are holding here). 33828039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33838039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33848039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33858039d87dSFilipe Manana * 33868039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33878039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33888039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33898039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33908039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33918039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33928039d87dSFilipe Manana * the former. 33938039d87dSFilipe Manana */ 33944a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 33958039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33968039d87dSFilipe Manana struct btrfs_path *path, 33978039d87dSFilipe Manana struct btrfs_key *new_key, 33988039d87dSFilipe Manana const u64 drop_start, 33998039d87dSFilipe Manana const u64 datal, 34008039d87dSFilipe Manana const u64 skip, 34018039d87dSFilipe Manana const u64 size, 34028039d87dSFilipe Manana char *inline_data) 34038039d87dSFilipe Manana { 34040b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 34058039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 34068039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 34070b246afaSJeff Mahoney fs_info->sectorsize); 34088039d87dSFilipe Manana int ret; 34098039d87dSFilipe Manana struct btrfs_key key; 34108039d87dSFilipe Manana 34118039d87dSFilipe Manana if (new_key->offset > 0) 34128039d87dSFilipe Manana return -EOPNOTSUPP; 34138039d87dSFilipe Manana 34144a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 34158039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 34168039d87dSFilipe Manana key.offset = 0; 34178039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 34188039d87dSFilipe Manana if (ret < 0) { 34198039d87dSFilipe Manana return ret; 34208039d87dSFilipe Manana } else if (ret > 0) { 34218039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 34228039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 34238039d87dSFilipe Manana if (ret < 0) 34248039d87dSFilipe Manana return ret; 34258039d87dSFilipe Manana else if (ret > 0) 34268039d87dSFilipe Manana goto copy_inline_extent; 34278039d87dSFilipe Manana } 34288039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 34294a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34308039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 34318039d87dSFilipe Manana ASSERT(key.offset > 0); 34328039d87dSFilipe Manana return -EOPNOTSUPP; 34338039d87dSFilipe Manana } 34348039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34358039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34368039d87dSFilipe Manana u64 ext_len; 34378039d87dSFilipe Manana 34388039d87dSFilipe Manana /* 34398039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34408039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34418039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34428039d87dSFilipe Manana */ 34438039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34448039d87dSFilipe Manana struct btrfs_file_extent_item); 34458039d87dSFilipe Manana /* 34468039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34478039d87dSFilipe Manana * following it. 34488039d87dSFilipe Manana */ 34498039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34508039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34518039d87dSFilipe Manana goto copy_inline_extent; 34528039d87dSFilipe Manana 34538039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34548039d87dSFilipe Manana if (ext_len > aligned_end) 34558039d87dSFilipe Manana return -EOPNOTSUPP; 34568039d87dSFilipe Manana 34578039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34588039d87dSFilipe Manana if (ret < 0) { 34598039d87dSFilipe Manana return ret; 34608039d87dSFilipe Manana } else if (ret == 0) { 34618039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34628039d87dSFilipe Manana path->slots[0]); 34634a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34648039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34658039d87dSFilipe Manana return -EOPNOTSUPP; 34668039d87dSFilipe Manana } 34678039d87dSFilipe Manana } 34688039d87dSFilipe Manana 34698039d87dSFilipe Manana copy_inline_extent: 34708039d87dSFilipe Manana /* 34718039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34728039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34738039d87dSFilipe Manana */ 34748039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34758039d87dSFilipe Manana /* 34768039d87dSFilipe Manana * If the destination inode has an inline extent... 34778039d87dSFilipe Manana * This would require copying the data from the source inline 34788039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34798039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34808039d87dSFilipe Manana * or just one of them, which would require decompressing and 34818039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34828039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34838039d87dSFilipe Manana * inline extent). 34848039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34858039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34868039d87dSFilipe Manana */ 34878039d87dSFilipe Manana return -EOPNOTSUPP; 34888039d87dSFilipe Manana } 34898039d87dSFilipe Manana 34908039d87dSFilipe Manana btrfs_release_path(path); 34918039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34928039d87dSFilipe Manana if (ret) 34938039d87dSFilipe Manana return ret; 34948039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34958039d87dSFilipe Manana if (ret) 34968039d87dSFilipe Manana return ret; 34978039d87dSFilipe Manana 34988039d87dSFilipe Manana if (skip) { 34998039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 35008039d87dSFilipe Manana 35018039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 35028039d87dSFilipe Manana } 35038039d87dSFilipe Manana 35048039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 35058039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 35068039d87dSFilipe Manana path->slots[0]), 35078039d87dSFilipe Manana size); 35088039d87dSFilipe Manana inode_add_bytes(dst, datal); 35098039d87dSFilipe Manana 35108039d87dSFilipe Manana return 0; 35118039d87dSFilipe Manana } 35128039d87dSFilipe Manana 351332b7c687SMark Fasheh /** 351432b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 351532b7c687SMark Fasheh * 351632b7c687SMark Fasheh * @src: Inode to clone from 351732b7c687SMark Fasheh * @inode: Inode to clone to 351832b7c687SMark Fasheh * @off: Offset within source to start clone from 351932b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 35201c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 352132b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 35221c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 352332b7c687SMark Fasheh */ 352432b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3525f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 35261c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3527f46b5a66SChristoph Hellwig { 35280b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3529f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 353032b7c687SMark Fasheh struct btrfs_path *path = NULL; 3531f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 353232b7c687SMark Fasheh struct btrfs_trans_handle *trans; 353332b7c687SMark Fasheh char *buf = NULL; 3534ae01a0abSYan Zheng struct btrfs_key key; 3535f46b5a66SChristoph Hellwig u32 nritems; 3536f46b5a66SChristoph Hellwig int slot; 3537ae01a0abSYan Zheng int ret; 3538f82a9901SFilipe Manana const u64 len = olen_aligned; 3539f82a9901SFilipe Manana u64 last_dest_end = destoff; 3540ae01a0abSYan Zheng 3541ae01a0abSYan Zheng ret = -ENOMEM; 35420b246afaSJeff Mahoney buf = kmalloc(fs_info->nodesize, GFP_KERNEL | __GFP_NOWARN); 354315351955SDavid Sterba if (!buf) { 35440b246afaSJeff Mahoney buf = vmalloc(fs_info->nodesize); 3545ae01a0abSYan Zheng if (!buf) 354632b7c687SMark Fasheh return ret; 354715351955SDavid Sterba } 3548ae01a0abSYan Zheng 3549ae01a0abSYan Zheng path = btrfs_alloc_path(); 3550ae01a0abSYan Zheng if (!path) { 355115351955SDavid Sterba kvfree(buf); 355232b7c687SMark Fasheh return ret; 3553ae01a0abSYan Zheng } 355432b7c687SMark Fasheh 3555e4058b54SDavid Sterba path->reada = READA_FORWARD; 3556c5c9cd4dSSage Weil /* clone data */ 35574a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3558ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35592c463823SFilipe Manana key.offset = off; 3560f46b5a66SChristoph Hellwig 3561f46b5a66SChristoph Hellwig while (1) { 3562de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3563df858e76SFilipe Manana 3564f46b5a66SChristoph Hellwig /* 3565f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3566f46b5a66SChristoph Hellwig * tree. 3567f46b5a66SChristoph Hellwig */ 3568e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3569362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3570362a20c5SDavid Sterba 0, 0); 3571f46b5a66SChristoph Hellwig if (ret < 0) 3572f46b5a66SChristoph Hellwig goto out; 35732c463823SFilipe Manana /* 35742c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35752c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35762c463823SFilipe Manana * it might overlap our target range, therefore process it. 35772c463823SFilipe Manana */ 35782c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35792c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35802c463823SFilipe Manana path->slots[0] - 1); 35812c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35822c463823SFilipe Manana path->slots[0]--; 35832c463823SFilipe Manana } 3584f46b5a66SChristoph Hellwig 3585ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3586e4355f34SFilipe David Borba Manana process_slot: 3587ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3588362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3589f46b5a66SChristoph Hellwig if (ret < 0) 3590f46b5a66SChristoph Hellwig goto out; 3591f46b5a66SChristoph Hellwig if (ret > 0) 3592f46b5a66SChristoph Hellwig break; 3593ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3594f46b5a66SChristoph Hellwig } 3595f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3596f46b5a66SChristoph Hellwig slot = path->slots[0]; 3597f46b5a66SChristoph Hellwig 3598ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3599962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 36004a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3601f46b5a66SChristoph Hellwig break; 3602f46b5a66SChristoph Hellwig 3603962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3604c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3605c5c9cd4dSSage Weil int type; 360631840ae1SZheng Yan u32 size; 360731840ae1SZheng Yan struct btrfs_key new_key; 3608c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3609c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3610c5c9cd4dSSage Weil u8 comp; 3611f82a9901SFilipe Manana u64 drop_start; 361231840ae1SZheng Yan 3613c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3614c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3615c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3616c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3617c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3618c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3619d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3620d397712bSChris Mason extent); 3621d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3622d397712bSChris Mason extent); 3623c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3624d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3625d397712bSChris Mason extent); 3626c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3627c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3628c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3629c5c9cd4dSSage Weil extent); 3630c5c9cd4dSSage Weil } 363131840ae1SZheng Yan 36322c463823SFilipe Manana /* 36332c463823SFilipe Manana * The first search might have left us at an extent 36342c463823SFilipe Manana * item that ends before our target range's start, can 36352c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 36362c463823SFilipe Manana */ 36372c463823SFilipe Manana if (key.offset + datal <= off) { 3638e4355f34SFilipe David Borba Manana path->slots[0]++; 3639e4355f34SFilipe David Borba Manana goto process_slot; 36402c463823SFilipe Manana } else if (key.offset >= off + len) { 36412c463823SFilipe Manana break; 3642e4355f34SFilipe David Borba Manana } 3643df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3644e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3645e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3646e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3647e4355f34SFilipe David Borba Manana size); 3648e4355f34SFilipe David Borba Manana 3649e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3650e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3651c5c9cd4dSSage Weil 365231840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36534a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36544d728ec7SLi Zefan if (off <= key.offset) 3655c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36564d728ec7SLi Zefan else 36574d728ec7SLi Zefan new_key.offset = destoff; 3658c5c9cd4dSSage Weil 3659b6f3409bSSage Weil /* 3660f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3661f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3662f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3663f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3664f82a9901SFilipe Manana * partially overlaps it). 3665f82a9901SFilipe Manana */ 3666f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3667f82a9901SFilipe Manana drop_start = last_dest_end; 3668f82a9901SFilipe Manana else 3669f82a9901SFilipe Manana drop_start = new_key.offset; 3670f82a9901SFilipe Manana 3671f82a9901SFilipe Manana /* 3672b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3673b6f3409bSSage Weil * 1 - add new extent 3674b6f3409bSSage Weil * 1 - inode update 3675b6f3409bSSage Weil */ 3676b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3677a22285a6SYan, Zheng if (IS_ERR(trans)) { 3678a22285a6SYan, Zheng ret = PTR_ERR(trans); 3679a22285a6SYan, Zheng goto out; 3680a22285a6SYan, Zheng } 3681a22285a6SYan, Zheng 3682c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3683c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3684d72c0842SLi Zefan /* 3685d72c0842SLi Zefan * a | --- range to clone ---| b 3686d72c0842SLi Zefan * | ------------- extent ------------- | 3687d72c0842SLi Zefan */ 3688d72c0842SLi Zefan 368993915584SAntonio Ospite /* subtract range b */ 3690d72c0842SLi Zefan if (key.offset + datal > off + len) 3691d72c0842SLi Zefan datal = off + len - key.offset; 3692d72c0842SLi Zefan 369393915584SAntonio Ospite /* subtract range a */ 3694a22285a6SYan, Zheng if (off > key.offset) { 3695a22285a6SYan, Zheng datao += off - key.offset; 3696a22285a6SYan, Zheng datal -= off - key.offset; 3697a22285a6SYan, Zheng } 3698a22285a6SYan, Zheng 36995dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3700f82a9901SFilipe Manana drop_start, 3701a22285a6SYan, Zheng new_key.offset + datal, 37022671485dSJosef Bacik 1); 370379787eaaSJeff Mahoney if (ret) { 37043f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 370500fdf13aSLiu Bo btrfs_abort_transaction(trans, 370666642832SJeff Mahoney ret); 37073a45bb20SJeff Mahoney btrfs_end_transaction(trans); 370879787eaaSJeff Mahoney goto out; 370979787eaaSJeff Mahoney } 3710a22285a6SYan, Zheng 371131840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 371231840ae1SZheng Yan &new_key, size); 371379787eaaSJeff Mahoney if (ret) { 371466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 37153a45bb20SJeff Mahoney btrfs_end_transaction(trans); 371679787eaaSJeff Mahoney goto out; 371779787eaaSJeff Mahoney } 371831840ae1SZheng Yan 371931840ae1SZheng Yan leaf = path->nodes[0]; 372031840ae1SZheng Yan slot = path->slots[0]; 372131840ae1SZheng Yan write_extent_buffer(leaf, buf, 372231840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 372331840ae1SZheng Yan size); 3724ae01a0abSYan Zheng 3725f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3726f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3727c5c9cd4dSSage Weil 3728c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3729c5c9cd4dSSage Weil if (!disko) 3730c5c9cd4dSSage Weil datao = 0; 3731c5c9cd4dSSage Weil 3732c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3733c5c9cd4dSSage Weil datao); 3734c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3735c5c9cd4dSSage Weil datal); 3736fcebe456SJosef Bacik 3737c5c9cd4dSSage Weil if (disko) { 3738c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 37392ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 37402ff7e61eSJeff Mahoney fs_info, 37415d4f98a2SYan Zheng disko, diskl, 0, 3742f46b5a66SChristoph Hellwig root->root_key.objectid, 37434a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 3744b06c4bf5SFilipe Manana new_key.offset - datao); 374579787eaaSJeff Mahoney if (ret) { 374679787eaaSJeff Mahoney btrfs_abort_transaction(trans, 374779787eaaSJeff Mahoney ret); 37483a45bb20SJeff Mahoney btrfs_end_transaction(trans); 374979787eaaSJeff Mahoney goto out; 375079787eaaSJeff Mahoney 375179787eaaSJeff Mahoney } 3752f46b5a66SChristoph Hellwig } 3753c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3754c5c9cd4dSSage Weil u64 skip = 0; 3755c5c9cd4dSSage Weil u64 trim = 0; 3756ed958762SFilipe Manana 3757c5c9cd4dSSage Weil if (off > key.offset) { 3758c5c9cd4dSSage Weil skip = off - key.offset; 3759c5c9cd4dSSage Weil new_key.offset += skip; 376031840ae1SZheng Yan } 3761d397712bSChris Mason 3762c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3763c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3764d397712bSChris Mason 3765c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3766c5c9cd4dSSage Weil ret = -EINVAL; 37673a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3768c5c9cd4dSSage Weil goto out; 376931840ae1SZheng Yan } 3770c5c9cd4dSSage Weil size -= skip + trim; 3771c5c9cd4dSSage Weil datal -= skip + trim; 3772a22285a6SYan, Zheng 37734a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 37748039d87dSFilipe Manana trans, path, 37758039d87dSFilipe Manana &new_key, 3776f82a9901SFilipe Manana drop_start, 37778039d87dSFilipe Manana datal, 37788039d87dSFilipe Manana skip, size, buf); 377979787eaaSJeff Mahoney if (ret) { 37803f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37813a29bc09SChris Mason btrfs_abort_transaction(trans, 378279787eaaSJeff Mahoney ret); 37833a45bb20SJeff Mahoney btrfs_end_transaction(trans); 378479787eaaSJeff Mahoney goto out; 378579787eaaSJeff Mahoney } 3786c5c9cd4dSSage Weil leaf = path->nodes[0]; 3787c5c9cd4dSSage Weil slot = path->slots[0]; 3788c5c9cd4dSSage Weil } 3789c5c9cd4dSSage Weil 37907ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 37917ffbb598SFilipe Manana if (drop_start < new_key.offset) 3792a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 379314f59796SFilipe Manana NULL, drop_start, 37947ffbb598SFilipe Manana new_key.offset - drop_start); 37957ffbb598SFilipe Manana 3796a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 3797a2f392e4SNikolay Borisov path, 0, 0); 37987ffbb598SFilipe Manana 3799c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3800b3b4aa74SDavid Sterba btrfs_release_path(path); 3801c5c9cd4dSSage Weil 380262e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 38030b246afaSJeff Mahoney fs_info->sectorsize); 3804f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3805f82a9901SFilipe Manana last_dest_end, 38061c919a5eSMark Fasheh destoff, olen, 38071c919a5eSMark Fasheh no_time_update); 3808f82a9901SFilipe Manana if (ret) 380979787eaaSJeff Mahoney goto out; 38102c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 38112c463823SFilipe Manana break; 3812a22285a6SYan, Zheng } 3813b3b4aa74SDavid Sterba btrfs_release_path(path); 3814df858e76SFilipe Manana key.offset = next_key_min_offset; 381569ae5e44SWang Xiaoguang 381669ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 381769ae5e44SWang Xiaoguang ret = -EINTR; 381869ae5e44SWang Xiaoguang goto out; 381969ae5e44SWang Xiaoguang } 3820ae01a0abSYan Zheng } 3821f46b5a66SChristoph Hellwig ret = 0; 382232b7c687SMark Fasheh 3823f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3824f82a9901SFilipe Manana /* 3825f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3826f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3827f82a9901SFilipe Manana */ 3828b3b4aa74SDavid Sterba btrfs_release_path(path); 3829f82a9901SFilipe Manana 3830f82a9901SFilipe Manana /* 3831f82a9901SFilipe Manana * 1 - remove extent(s) 3832f82a9901SFilipe Manana * 1 - inode update 3833f82a9901SFilipe Manana */ 3834f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3835f82a9901SFilipe Manana if (IS_ERR(trans)) { 3836f82a9901SFilipe Manana ret = PTR_ERR(trans); 3837f82a9901SFilipe Manana goto out; 3838f82a9901SFilipe Manana } 3839f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3840f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3841f82a9901SFilipe Manana if (ret) { 3842f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 384366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 38443a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3845f82a9901SFilipe Manana goto out; 3846f82a9901SFilipe Manana } 3847a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 3848a2f392e4SNikolay Borisov last_dest_end, 384914f59796SFilipe Manana destoff + len - last_dest_end); 3850f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38511c919a5eSMark Fasheh destoff, olen, no_time_update); 3852f82a9901SFilipe Manana } 3853f82a9901SFilipe Manana 3854f46b5a66SChristoph Hellwig out: 385532b7c687SMark Fasheh btrfs_free_path(path); 385615351955SDavid Sterba kvfree(buf); 385732b7c687SMark Fasheh return ret; 385832b7c687SMark Fasheh } 385932b7c687SMark Fasheh 38603db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 386132b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 386232b7c687SMark Fasheh { 386354563d41SAl Viro struct inode *inode = file_inode(file); 38643db11b2eSZach Brown struct inode *src = file_inode(file_src); 38650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 386632b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 386732b7c687SMark Fasheh int ret; 386832b7c687SMark Fasheh u64 len = olen; 38690b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 38703db11b2eSZach Brown int same_inode = src == inode; 387132b7c687SMark Fasheh 387232b7c687SMark Fasheh /* 387332b7c687SMark Fasheh * TODO: 387432b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 387532b7c687SMark Fasheh * decompress into destination's address_space (the file offset 387632b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 387732b7c687SMark Fasheh * otherwise reinsert) a subrange. 387800fdf13aSLiu Bo * 387900fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 388000fdf13aSLiu Bo * be either compressed or non-compressed. 388132b7c687SMark Fasheh */ 388232b7c687SMark Fasheh 388332b7c687SMark Fasheh if (btrfs_root_readonly(root)) 388432b7c687SMark Fasheh return -EROFS; 388532b7c687SMark Fasheh 38863db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 38873db11b2eSZach Brown src->i_sb != inode->i_sb) 38883db11b2eSZach Brown return -EXDEV; 388932b7c687SMark Fasheh 389032b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 389132b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 389232b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 38933db11b2eSZach Brown return -EINVAL; 389432b7c687SMark Fasheh 389532b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 38963db11b2eSZach Brown return -EISDIR; 389732b7c687SMark Fasheh 389832b7c687SMark Fasheh if (!same_inode) { 3899293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 390032b7c687SMark Fasheh } else { 39015955102cSAl Viro inode_lock(src); 390232b7c687SMark Fasheh } 390332b7c687SMark Fasheh 390432b7c687SMark Fasheh /* determine range to clone */ 390532b7c687SMark Fasheh ret = -EINVAL; 390632b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 390732b7c687SMark Fasheh goto out_unlock; 390832b7c687SMark Fasheh if (len == 0) 390932b7c687SMark Fasheh olen = len = src->i_size - off; 391032b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 391132b7c687SMark Fasheh if (off + len == src->i_size) 391232b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 391332b7c687SMark Fasheh 3914ccccf3d6SFilipe Manana if (len == 0) { 3915ccccf3d6SFilipe Manana ret = 0; 3916ccccf3d6SFilipe Manana goto out_unlock; 3917ccccf3d6SFilipe Manana } 3918ccccf3d6SFilipe Manana 391932b7c687SMark Fasheh /* verify the end result is block aligned */ 392032b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 392132b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 392232b7c687SMark Fasheh goto out_unlock; 392332b7c687SMark Fasheh 392432b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 392532b7c687SMark Fasheh if (same_inode) { 392632b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 392732b7c687SMark Fasheh goto out_unlock; 392832b7c687SMark Fasheh } 392932b7c687SMark Fasheh 393032b7c687SMark Fasheh if (destoff > inode->i_size) { 393132b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 393232b7c687SMark Fasheh if (ret) 393332b7c687SMark Fasheh goto out_unlock; 393432b7c687SMark Fasheh } 393532b7c687SMark Fasheh 3936c125b8bfSFilipe Manana /* 3937c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3938c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3939c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3940c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3941c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3942c125b8bfSFilipe Manana */ 3943c125b8bfSFilipe Manana if (same_inode) { 3944c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3945c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 394632b7c687SMark Fasheh 3947e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 3948c125b8bfSFilipe Manana } else { 3949e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 3950e0bd70c6SFilipe Manana true); 3951e0bd70c6SFilipe Manana } 3952e0bd70c6SFilipe Manana ASSERT(ret == 0); 3953e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3954e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3955e0bd70c6SFilipe Manana goto out_unlock; 3956c125b8bfSFilipe Manana } 395732b7c687SMark Fasheh 39581c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 395932b7c687SMark Fasheh 3960c125b8bfSFilipe Manana if (same_inode) { 3961c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3962c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3963c125b8bfSFilipe Manana 3964c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3965c125b8bfSFilipe Manana } else { 3966293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 3967c125b8bfSFilipe Manana } 3968c125b8bfSFilipe Manana /* 3969c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3970c125b8bfSFilipe Manana * data immediately and not the previous data. 3971c125b8bfSFilipe Manana */ 397265bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 397309cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 397409cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 3975f46b5a66SChristoph Hellwig out_unlock: 3976293a8489SMark Fasheh if (!same_inode) 3977293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 3978293a8489SMark Fasheh else 39795955102cSAl Viro inode_unlock(src); 39803db11b2eSZach Brown return ret; 39813db11b2eSZach Brown } 39823db11b2eSZach Brown 398304b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 398404b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 39853db11b2eSZach Brown { 398604b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 3987c5c9cd4dSSage Weil } 3988c5c9cd4dSSage Weil 3989f46b5a66SChristoph Hellwig /* 3990f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3991f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3992f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3993f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3994f46b5a66SChristoph Hellwig */ 3995b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 3996f46b5a66SChristoph Hellwig { 3997496ad9aaSAl Viro struct inode *inode = file_inode(file); 39980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3999f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4000f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 40011ab86aedSSage Weil int ret; 4002f46b5a66SChristoph Hellwig 40031ab86aedSSage Weil ret = -EPERM; 4004df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 4005f46b5a66SChristoph Hellwig goto out; 40061ab86aedSSage Weil 40071ab86aedSSage Weil ret = -EINPROGRESS; 40081ab86aedSSage Weil if (file->private_data) 40091ab86aedSSage Weil goto out; 40109ca9ee09SSage Weil 4011b83cc969SLi Zefan ret = -EROFS; 4012b83cc969SLi Zefan if (btrfs_root_readonly(root)) 4013b83cc969SLi Zefan goto out; 4014b83cc969SLi Zefan 4015a561be71SAl Viro ret = mnt_want_write_file(file); 4016c146afadSYan Zheng if (ret) 4017c146afadSYan Zheng goto out; 4018c146afadSYan Zheng 40190b246afaSJeff Mahoney atomic_inc(&fs_info->open_ioctl_trans); 40209ca9ee09SSage Weil 4021f46b5a66SChristoph Hellwig ret = -ENOMEM; 40227a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 4023abd30bb0STsutomu Itoh if (IS_ERR(trans)) 40241ab86aedSSage Weil goto out_drop; 40251ab86aedSSage Weil 40261ab86aedSSage Weil file->private_data = trans; 40271ab86aedSSage Weil return 0; 40281ab86aedSSage Weil 40291ab86aedSSage Weil out_drop: 40300b246afaSJeff Mahoney atomic_dec(&fs_info->open_ioctl_trans); 40312a79f17eSAl Viro mnt_drop_write_file(file); 4032f46b5a66SChristoph Hellwig out: 4033f46b5a66SChristoph Hellwig return ret; 4034f46b5a66SChristoph Hellwig } 4035f46b5a66SChristoph Hellwig 40366ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40376ef5ed0dSJosef Bacik { 4038496ad9aaSAl Viro struct inode *inode = file_inode(file); 40390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 40406ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40416ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40426ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40436ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40446ef5ed0dSJosef Bacik struct btrfs_path *path; 40456ef5ed0dSJosef Bacik struct btrfs_key location; 40466ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40476ef5ed0dSJosef Bacik u64 objectid = 0; 40486ef5ed0dSJosef Bacik u64 dir_id; 40493c04ce01SMiao Xie int ret; 40506ef5ed0dSJosef Bacik 40516ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40526ef5ed0dSJosef Bacik return -EPERM; 40536ef5ed0dSJosef Bacik 40543c04ce01SMiao Xie ret = mnt_want_write_file(file); 40553c04ce01SMiao Xie if (ret) 40563c04ce01SMiao Xie return ret; 40573c04ce01SMiao Xie 40583c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40593c04ce01SMiao Xie ret = -EFAULT; 40603c04ce01SMiao Xie goto out; 40613c04ce01SMiao Xie } 40626ef5ed0dSJosef Bacik 40636ef5ed0dSJosef Bacik if (!objectid) 40641cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40656ef5ed0dSJosef Bacik 40666ef5ed0dSJosef Bacik location.objectid = objectid; 40676ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40686ef5ed0dSJosef Bacik location.offset = (u64)-1; 40696ef5ed0dSJosef Bacik 40700b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 40713c04ce01SMiao Xie if (IS_ERR(new_root)) { 40723c04ce01SMiao Xie ret = PTR_ERR(new_root); 40733c04ce01SMiao Xie goto out; 40743c04ce01SMiao Xie } 40756ef5ed0dSJosef Bacik 40766ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40773c04ce01SMiao Xie if (!path) { 40783c04ce01SMiao Xie ret = -ENOMEM; 40793c04ce01SMiao Xie goto out; 40803c04ce01SMiao Xie } 40816ef5ed0dSJosef Bacik path->leave_spinning = 1; 40826ef5ed0dSJosef Bacik 40836ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 408498d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40856ef5ed0dSJosef Bacik btrfs_free_path(path); 40863c04ce01SMiao Xie ret = PTR_ERR(trans); 40873c04ce01SMiao Xie goto out; 40886ef5ed0dSJosef Bacik } 40896ef5ed0dSJosef Bacik 40900b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 40910b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 40926ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4093cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40946ef5ed0dSJosef Bacik btrfs_free_path(path); 40953a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40960b246afaSJeff Mahoney btrfs_err(fs_info, 40975d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 40983c04ce01SMiao Xie ret = -ENOENT; 40993c04ce01SMiao Xie goto out; 41006ef5ed0dSJosef Bacik } 41016ef5ed0dSJosef Bacik 41026ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 41036ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 41046ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 41056ef5ed0dSJosef Bacik btrfs_free_path(path); 41066ef5ed0dSJosef Bacik 41070b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 41083a45bb20SJeff Mahoney btrfs_end_transaction(trans); 41093c04ce01SMiao Xie out: 41103c04ce01SMiao Xie mnt_drop_write_file(file); 41113c04ce01SMiao Xie return ret; 41126ef5ed0dSJosef Bacik } 41136ef5ed0dSJosef Bacik 41145af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 4115bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4116bf5fc093SJosef Bacik { 4117bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4118bf5fc093SJosef Bacik 4119bf5fc093SJosef Bacik space->total_bytes = 0; 4120bf5fc093SJosef Bacik space->used_bytes = 0; 4121bf5fc093SJosef Bacik space->flags = 0; 4122bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4123bf5fc093SJosef Bacik space->flags = block_group->flags; 4124bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4125bf5fc093SJosef Bacik space->used_bytes += 4126bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4127bf5fc093SJosef Bacik } 4128bf5fc093SJosef Bacik } 4129bf5fc093SJosef Bacik 41302ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 41312ff7e61eSJeff Mahoney void __user *arg) 41321406e432SJosef Bacik { 41331406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41341406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41351406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41367fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 413713f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41381406e432SJosef Bacik struct btrfs_space_info *info; 4139bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 4140bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4141bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4142bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 4143bf5fc093SJosef Bacik int num_types = 4; 41447fde62bfSChris Mason int alloc_size; 41451406e432SJosef Bacik int ret = 0; 414651788b1bSDan Rosenberg u64 slot_count = 0; 4147bf5fc093SJosef Bacik int i, c; 41481406e432SJosef Bacik 41491406e432SJosef Bacik if (copy_from_user(&space_args, 41501406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41511406e432SJosef Bacik sizeof(space_args))) 41521406e432SJosef Bacik return -EFAULT; 41531406e432SJosef Bacik 4154bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4155bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4156bf5fc093SJosef Bacik 4157bf5fc093SJosef Bacik info = NULL; 41587fde62bfSChris Mason rcu_read_lock(); 41590b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4160bf5fc093SJosef Bacik list) { 4161bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4162bf5fc093SJosef Bacik info = tmp; 4163bf5fc093SJosef Bacik break; 4164bf5fc093SJosef Bacik } 4165bf5fc093SJosef Bacik } 41667fde62bfSChris Mason rcu_read_unlock(); 41671406e432SJosef Bacik 4168bf5fc093SJosef Bacik if (!info) 4169bf5fc093SJosef Bacik continue; 4170bf5fc093SJosef Bacik 4171bf5fc093SJosef Bacik down_read(&info->groups_sem); 4172bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4173bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4174bf5fc093SJosef Bacik slot_count++; 4175bf5fc093SJosef Bacik } 4176bf5fc093SJosef Bacik up_read(&info->groups_sem); 4177bf5fc093SJosef Bacik } 4178bf5fc093SJosef Bacik 417936523e95SDavid Sterba /* 418036523e95SDavid Sterba * Global block reserve, exported as a space_info 418136523e95SDavid Sterba */ 418236523e95SDavid Sterba slot_count++; 418336523e95SDavid Sterba 41847fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 41857fde62bfSChris Mason if (space_args.space_slots == 0) { 41867fde62bfSChris Mason space_args.total_spaces = slot_count; 41877fde62bfSChris Mason goto out; 41887fde62bfSChris Mason } 4189bf5fc093SJosef Bacik 419051788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4191bf5fc093SJosef Bacik 41927fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4193bf5fc093SJosef Bacik 41947fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41957fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41967fde62bfSChris Mason */ 419709cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 41987fde62bfSChris Mason return -ENOMEM; 41997fde62bfSChris Mason 42007fde62bfSChris Mason space_args.total_spaces = 0; 42018d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 42027fde62bfSChris Mason if (!dest) 42037fde62bfSChris Mason return -ENOMEM; 42047fde62bfSChris Mason dest_orig = dest; 42057fde62bfSChris Mason 42067fde62bfSChris Mason /* now we have a buffer to copy into */ 4207bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4208bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4209bf5fc093SJosef Bacik 421051788b1bSDan Rosenberg if (!slot_count) 421151788b1bSDan Rosenberg break; 421251788b1bSDan Rosenberg 4213bf5fc093SJosef Bacik info = NULL; 42141406e432SJosef Bacik rcu_read_lock(); 42150b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4216bf5fc093SJosef Bacik list) { 4217bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4218bf5fc093SJosef Bacik info = tmp; 42197fde62bfSChris Mason break; 4220bf5fc093SJosef Bacik } 4221bf5fc093SJosef Bacik } 4222bf5fc093SJosef Bacik rcu_read_unlock(); 42237fde62bfSChris Mason 4224bf5fc093SJosef Bacik if (!info) 4225bf5fc093SJosef Bacik continue; 4226bf5fc093SJosef Bacik down_read(&info->groups_sem); 4227bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4228bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 42295af3e8ccSStefan Behrens btrfs_get_block_group_info( 42305af3e8ccSStefan Behrens &info->block_groups[c], &space); 42317fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42321406e432SJosef Bacik dest++; 42331406e432SJosef Bacik space_args.total_spaces++; 423451788b1bSDan Rosenberg slot_count--; 42351406e432SJosef Bacik } 423651788b1bSDan Rosenberg if (!slot_count) 423751788b1bSDan Rosenberg break; 4238bf5fc093SJosef Bacik } 4239bf5fc093SJosef Bacik up_read(&info->groups_sem); 4240bf5fc093SJosef Bacik } 42411406e432SJosef Bacik 424236523e95SDavid Sterba /* 424336523e95SDavid Sterba * Add global block reserve 424436523e95SDavid Sterba */ 424536523e95SDavid Sterba if (slot_count) { 42460b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 424736523e95SDavid Sterba 424836523e95SDavid Sterba spin_lock(&block_rsv->lock); 424936523e95SDavid Sterba space.total_bytes = block_rsv->size; 425036523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 425136523e95SDavid Sterba spin_unlock(&block_rsv->lock); 425236523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 425336523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 425436523e95SDavid Sterba space_args.total_spaces++; 425536523e95SDavid Sterba } 425636523e95SDavid Sterba 42572eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42587fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42597fde62bfSChris Mason 42607fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42617fde62bfSChris Mason ret = -EFAULT; 42627fde62bfSChris Mason 42637fde62bfSChris Mason kfree(dest_orig); 42647fde62bfSChris Mason out: 42657fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42661406e432SJosef Bacik ret = -EFAULT; 42671406e432SJosef Bacik 42681406e432SJosef Bacik return ret; 42691406e432SJosef Bacik } 42701406e432SJosef Bacik 4271f46b5a66SChristoph Hellwig /* 4272f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4273f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4274f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4275f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4276f46b5a66SChristoph Hellwig */ 4277f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 4278f46b5a66SChristoph Hellwig { 4279496ad9aaSAl Viro struct inode *inode = file_inode(file); 4280f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4281f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4282f46b5a66SChristoph Hellwig 4283f46b5a66SChristoph Hellwig trans = file->private_data; 42841ab86aedSSage Weil if (!trans) 42851ab86aedSSage Weil return -EINVAL; 4286b214107eSChristoph Hellwig file->private_data = NULL; 42879ca9ee09SSage Weil 42883a45bb20SJeff Mahoney btrfs_end_transaction(trans); 42891ab86aedSSage Weil 4290a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 42919ca9ee09SSage Weil 42922a79f17eSAl Viro mnt_drop_write_file(file); 42931ab86aedSSage Weil return 0; 4294f46b5a66SChristoph Hellwig } 4295f46b5a66SChristoph Hellwig 42969a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42979a8c28beSMiao Xie void __user *argp) 429846204592SSage Weil { 429946204592SSage Weil struct btrfs_trans_handle *trans; 430046204592SSage Weil u64 transid; 4301db5b493aSTsutomu Itoh int ret; 430246204592SSage Weil 4303d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4304ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4305ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 430698d5dc13STsutomu Itoh return PTR_ERR(trans); 4307ff7c1d33SMiao Xie 4308ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4309ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4310ff7c1d33SMiao Xie goto out; 4311ff7c1d33SMiao Xie } 431246204592SSage Weil transid = trans->transid; 43133a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 43148b2b2d3cSTsutomu Itoh if (ret) { 43153a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4316db5b493aSTsutomu Itoh return ret; 43178b2b2d3cSTsutomu Itoh } 4318ff7c1d33SMiao Xie out: 431946204592SSage Weil if (argp) 432046204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 432146204592SSage Weil return -EFAULT; 432246204592SSage Weil return 0; 432346204592SSage Weil } 432446204592SSage Weil 43252ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 43269a8c28beSMiao Xie void __user *argp) 432746204592SSage Weil { 432846204592SSage Weil u64 transid; 432946204592SSage Weil 433046204592SSage Weil if (argp) { 433146204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 433246204592SSage Weil return -EFAULT; 433346204592SSage Weil } else { 433446204592SSage Weil transid = 0; /* current trans */ 433546204592SSage Weil } 43362ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 433746204592SSage Weil } 433846204592SSage Weil 4339b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4340475f6387SJan Schmidt { 43410b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4342475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4343b8e95489SMiao Xie int ret; 4344475f6387SJan Schmidt 4345475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4346475f6387SJan Schmidt return -EPERM; 4347475f6387SJan Schmidt 4348475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4349475f6387SJan Schmidt if (IS_ERR(sa)) 4350475f6387SJan Schmidt return PTR_ERR(sa); 4351475f6387SJan Schmidt 4352b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4353b8e95489SMiao Xie ret = mnt_want_write_file(file); 4354b8e95489SMiao Xie if (ret) 4355b8e95489SMiao Xie goto out; 4356b8e95489SMiao Xie } 4357b8e95489SMiao Xie 43580b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 435963a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 436063a212abSStefan Behrens 0); 4361475f6387SJan Schmidt 4362475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4363475f6387SJan Schmidt ret = -EFAULT; 4364475f6387SJan Schmidt 4365b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4366b8e95489SMiao Xie mnt_drop_write_file(file); 4367b8e95489SMiao Xie out: 4368475f6387SJan Schmidt kfree(sa); 4369475f6387SJan Schmidt return ret; 4370475f6387SJan Schmidt } 4371475f6387SJan Schmidt 43722ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4373475f6387SJan Schmidt { 4374475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4375475f6387SJan Schmidt return -EPERM; 4376475f6387SJan Schmidt 43772ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4378475f6387SJan Schmidt } 4379475f6387SJan Schmidt 43802ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4381475f6387SJan Schmidt void __user *arg) 4382475f6387SJan Schmidt { 4383475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4384475f6387SJan Schmidt int ret; 4385475f6387SJan Schmidt 4386475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4387475f6387SJan Schmidt return -EPERM; 4388475f6387SJan Schmidt 4389475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4390475f6387SJan Schmidt if (IS_ERR(sa)) 4391475f6387SJan Schmidt return PTR_ERR(sa); 4392475f6387SJan Schmidt 43932ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4394475f6387SJan Schmidt 4395475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4396475f6387SJan Schmidt ret = -EFAULT; 4397475f6387SJan Schmidt 4398475f6387SJan Schmidt kfree(sa); 4399475f6387SJan Schmidt return ret; 4400475f6387SJan Schmidt } 4401475f6387SJan Schmidt 44022ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4403b27f7c0cSDavid Sterba void __user *arg) 4404c11d2c23SStefan Behrens { 4405c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4406c11d2c23SStefan Behrens int ret; 4407c11d2c23SStefan Behrens 4408c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4409c11d2c23SStefan Behrens if (IS_ERR(sa)) 4410c11d2c23SStefan Behrens return PTR_ERR(sa); 4411c11d2c23SStefan Behrens 4412b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4413b27f7c0cSDavid Sterba kfree(sa); 4414b27f7c0cSDavid Sterba return -EPERM; 4415b27f7c0cSDavid Sterba } 4416b27f7c0cSDavid Sterba 44172ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4418c11d2c23SStefan Behrens 4419c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4420c11d2c23SStefan Behrens ret = -EFAULT; 4421c11d2c23SStefan Behrens 4422c11d2c23SStefan Behrens kfree(sa); 4423c11d2c23SStefan Behrens return ret; 4424c11d2c23SStefan Behrens } 4425c11d2c23SStefan Behrens 44262ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 44272ff7e61eSJeff Mahoney void __user *arg) 44283f6bcfbdSStefan Behrens { 44293f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 44303f6bcfbdSStefan Behrens int ret; 44313f6bcfbdSStefan Behrens 44323f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44333f6bcfbdSStefan Behrens return -EPERM; 44343f6bcfbdSStefan Behrens 44353f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44363f6bcfbdSStefan Behrens if (IS_ERR(p)) 44373f6bcfbdSStefan Behrens return PTR_ERR(p); 44383f6bcfbdSStefan Behrens 44393f6bcfbdSStefan Behrens switch (p->cmd) { 44403f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 44410b246afaSJeff Mahoney if (fs_info->sb->s_flags & MS_RDONLY) { 4442adfa97cbSIlya Dryomov ret = -EROFS; 4443adfa97cbSIlya Dryomov goto out; 4444adfa97cbSIlya Dryomov } 44453f6bcfbdSStefan Behrens if (atomic_xchg( 44460b246afaSJeff Mahoney &fs_info->mutually_exclusive_operation_running, 1)) { 4447e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44483f6bcfbdSStefan Behrens } else { 44492ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 44503f6bcfbdSStefan Behrens atomic_set( 44510b246afaSJeff Mahoney &fs_info->mutually_exclusive_operation_running, 0); 44523f6bcfbdSStefan Behrens } 44533f6bcfbdSStefan Behrens break; 44543f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44550b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 44563f6bcfbdSStefan Behrens ret = 0; 44573f6bcfbdSStefan Behrens break; 44583f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 44590b246afaSJeff Mahoney ret = btrfs_dev_replace_cancel(fs_info, p); 44603f6bcfbdSStefan Behrens break; 44613f6bcfbdSStefan Behrens default: 44623f6bcfbdSStefan Behrens ret = -EINVAL; 44633f6bcfbdSStefan Behrens break; 44643f6bcfbdSStefan Behrens } 44653f6bcfbdSStefan Behrens 44663f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 44673f6bcfbdSStefan Behrens ret = -EFAULT; 4468adfa97cbSIlya Dryomov out: 44693f6bcfbdSStefan Behrens kfree(p); 44703f6bcfbdSStefan Behrens return ret; 44713f6bcfbdSStefan Behrens } 44723f6bcfbdSStefan Behrens 4473d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4474d7728c96SJan Schmidt { 4475d7728c96SJan Schmidt int ret = 0; 4476d7728c96SJan Schmidt int i; 4477740c3d22SChris Mason u64 rel_ptr; 4478d7728c96SJan Schmidt int size; 4479806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4480d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4481d7728c96SJan Schmidt struct btrfs_path *path; 4482d7728c96SJan Schmidt 448382b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4484d7728c96SJan Schmidt return -EPERM; 4485d7728c96SJan Schmidt 4486d7728c96SJan Schmidt path = btrfs_alloc_path(); 4487d7728c96SJan Schmidt if (!path) { 4488d7728c96SJan Schmidt ret = -ENOMEM; 4489d7728c96SJan Schmidt goto out; 4490d7728c96SJan Schmidt } 4491d7728c96SJan Schmidt 4492d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4493d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4494d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4495d7728c96SJan Schmidt ipa = NULL; 4496d7728c96SJan Schmidt goto out; 4497d7728c96SJan Schmidt } 4498d7728c96SJan Schmidt 4499d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4500d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4501d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4502d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4503d7728c96SJan Schmidt ipath = NULL; 4504d7728c96SJan Schmidt goto out; 4505d7728c96SJan Schmidt } 4506d7728c96SJan Schmidt 4507d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4508d7728c96SJan Schmidt if (ret < 0) 4509d7728c96SJan Schmidt goto out; 4510d7728c96SJan Schmidt 4511d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4512745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4513745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4514740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4515d7728c96SJan Schmidt } 4516d7728c96SJan Schmidt 4517745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 4518745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 4519d7728c96SJan Schmidt if (ret) { 4520d7728c96SJan Schmidt ret = -EFAULT; 4521d7728c96SJan Schmidt goto out; 4522d7728c96SJan Schmidt } 4523d7728c96SJan Schmidt 4524d7728c96SJan Schmidt out: 4525d7728c96SJan Schmidt btrfs_free_path(path); 4526d7728c96SJan Schmidt free_ipath(ipath); 4527d7728c96SJan Schmidt kfree(ipa); 4528d7728c96SJan Schmidt 4529d7728c96SJan Schmidt return ret; 4530d7728c96SJan Schmidt } 4531d7728c96SJan Schmidt 4532d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4533d7728c96SJan Schmidt { 4534d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4535d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4536d7728c96SJan Schmidt 4537d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4538d7728c96SJan Schmidt inodes->bytes_left -= c; 4539d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4540d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4541d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4542d7728c96SJan Schmidt inodes->elem_cnt += 3; 4543d7728c96SJan Schmidt } else { 4544d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4545d7728c96SJan Schmidt inodes->bytes_left = 0; 4546d7728c96SJan Schmidt inodes->elem_missed += 3; 4547d7728c96SJan Schmidt } 4548d7728c96SJan Schmidt 4549d7728c96SJan Schmidt return 0; 4550d7728c96SJan Schmidt } 4551d7728c96SJan Schmidt 45522ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4553d7728c96SJan Schmidt void __user *arg) 4554d7728c96SJan Schmidt { 4555d7728c96SJan Schmidt int ret = 0; 4556d7728c96SJan Schmidt int size; 4557d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4558d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4559d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4560d7728c96SJan Schmidt 4561d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4562d7728c96SJan Schmidt return -EPERM; 4563d7728c96SJan Schmidt 4564d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 45657b9ea627SShailendra Verma if (IS_ERR(loi)) 45667b9ea627SShailendra Verma return PTR_ERR(loi); 4567d7728c96SJan Schmidt 4568d7728c96SJan Schmidt path = btrfs_alloc_path(); 4569d7728c96SJan Schmidt if (!path) { 4570d7728c96SJan Schmidt ret = -ENOMEM; 4571d7728c96SJan Schmidt goto out; 4572d7728c96SJan Schmidt } 4573d7728c96SJan Schmidt 4574ee22184bSByongho Lee size = min_t(u32, loi->size, SZ_64K); 4575d7728c96SJan Schmidt inodes = init_data_container(size); 4576d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4577d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4578d7728c96SJan Schmidt inodes = NULL; 4579d7728c96SJan Schmidt goto out; 4580d7728c96SJan Schmidt } 4581d7728c96SJan Schmidt 45822ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4583df031f07SLiu Bo build_ino_list, inodes); 4584df031f07SLiu Bo if (ret == -EINVAL) 4585d7728c96SJan Schmidt ret = -ENOENT; 4586d7728c96SJan Schmidt if (ret < 0) 4587d7728c96SJan Schmidt goto out; 4588d7728c96SJan Schmidt 4589745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 4590745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 4591d7728c96SJan Schmidt if (ret) 4592d7728c96SJan Schmidt ret = -EFAULT; 4593d7728c96SJan Schmidt 4594d7728c96SJan Schmidt out: 4595d7728c96SJan Schmidt btrfs_free_path(path); 4596425d17a2SLiu Bo vfree(inodes); 4597d7728c96SJan Schmidt kfree(loi); 4598d7728c96SJan Schmidt 4599d7728c96SJan Schmidt return ret; 4600d7728c96SJan Schmidt } 4601d7728c96SJan Schmidt 460219a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4603c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4604c9e9f97bSIlya Dryomov { 4605c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4606c9e9f97bSIlya Dryomov 4607c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4608c9e9f97bSIlya Dryomov 4609837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 4610837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4611837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4612837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4613a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4614a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4615837d5b6eSIlya Dryomov 4616c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4617c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4618c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 461919a39dceSIlya Dryomov 462019a39dceSIlya Dryomov if (lock) { 462119a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 462219a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 462319a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 462419a39dceSIlya Dryomov } else { 462519a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 462619a39dceSIlya Dryomov } 4627c9e9f97bSIlya Dryomov } 4628c9e9f97bSIlya Dryomov 46299ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4630c9e9f97bSIlya Dryomov { 4631496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4632c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4633c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4634c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4635ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4636c9e9f97bSIlya Dryomov int ret; 4637c9e9f97bSIlya Dryomov 4638c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4639c9e9f97bSIlya Dryomov return -EPERM; 4640c9e9f97bSIlya Dryomov 4641e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46429ba1f6e4SLiu Bo if (ret) 46439ba1f6e4SLiu Bo return ret; 46449ba1f6e4SLiu Bo 4645ed0fb78fSIlya Dryomov again: 4646ed0fb78fSIlya Dryomov if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 4647c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 4648c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4649ed0fb78fSIlya Dryomov need_unlock = true; 4650ed0fb78fSIlya Dryomov goto locked; 4651ed0fb78fSIlya Dryomov } 4652ed0fb78fSIlya Dryomov 4653ed0fb78fSIlya Dryomov /* 465401327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4655ed0fb78fSIlya Dryomov * (1) some other op is running 4656ed0fb78fSIlya Dryomov * (2) balance is running 4657ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4658ed0fb78fSIlya Dryomov */ 4659ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4660ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4661ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 4662ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 4663ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4664ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 4665ed0fb78fSIlya Dryomov goto again; 4666ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4667ed0fb78fSIlya Dryomov 4668ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 4669ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 4670ed0fb78fSIlya Dryomov /* this is (3) */ 4671ed0fb78fSIlya Dryomov need_unlock = false; 4672ed0fb78fSIlya Dryomov goto locked; 4673ed0fb78fSIlya Dryomov } 4674ed0fb78fSIlya Dryomov 4675ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4676ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4677ed0fb78fSIlya Dryomov goto again; 4678ed0fb78fSIlya Dryomov } else { 4679ed0fb78fSIlya Dryomov /* this is (2) */ 4680ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4681ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4682ed0fb78fSIlya Dryomov goto out; 4683ed0fb78fSIlya Dryomov } 4684ed0fb78fSIlya Dryomov } else { 4685ed0fb78fSIlya Dryomov /* this is (1) */ 4686ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4687e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4688ed0fb78fSIlya Dryomov goto out; 4689ed0fb78fSIlya Dryomov } 4690ed0fb78fSIlya Dryomov 4691ed0fb78fSIlya Dryomov locked: 4692ed0fb78fSIlya Dryomov BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); 4693c9e9f97bSIlya Dryomov 4694c9e9f97bSIlya Dryomov if (arg) { 4695c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4696c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4697c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4698ed0fb78fSIlya Dryomov goto out_unlock; 4699c9e9f97bSIlya Dryomov } 4700de322263SIlya Dryomov 4701de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4702de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4703de322263SIlya Dryomov ret = -ENOTCONN; 4704de322263SIlya Dryomov goto out_bargs; 4705de322263SIlya Dryomov } 4706de322263SIlya Dryomov 4707de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4708de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4709de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4710de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4711de322263SIlya Dryomov 4712de322263SIlya Dryomov goto do_balance; 4713de322263SIlya Dryomov } 4714c9e9f97bSIlya Dryomov } else { 4715c9e9f97bSIlya Dryomov bargs = NULL; 4716c9e9f97bSIlya Dryomov } 4717c9e9f97bSIlya Dryomov 4718ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4719837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4720837d5b6eSIlya Dryomov goto out_bargs; 4721837d5b6eSIlya Dryomov } 4722837d5b6eSIlya Dryomov 47238d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4724c9e9f97bSIlya Dryomov if (!bctl) { 4725c9e9f97bSIlya Dryomov ret = -ENOMEM; 4726c9e9f97bSIlya Dryomov goto out_bargs; 4727c9e9f97bSIlya Dryomov } 4728c9e9f97bSIlya Dryomov 4729c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 4730c9e9f97bSIlya Dryomov if (arg) { 4731c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4732c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4733c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4734c9e9f97bSIlya Dryomov 4735c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4736f43ffb60SIlya Dryomov } else { 4737f43ffb60SIlya Dryomov /* balance everything - no filters */ 4738f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4739c9e9f97bSIlya Dryomov } 4740c9e9f97bSIlya Dryomov 47418eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47428eb93459SDavid Sterba ret = -EINVAL; 47430f89abf5SChristian Engelmayer goto out_bctl; 47448eb93459SDavid Sterba } 47458eb93459SDavid Sterba 4746de322263SIlya Dryomov do_balance: 4747c9e9f97bSIlya Dryomov /* 4748ed0fb78fSIlya Dryomov * Ownership of bctl and mutually_exclusive_operation_running 4749ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4750ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 4751ed0fb78fSIlya Dryomov * free_fs_info. mutually_exclusive_operation_running is 4752ed0fb78fSIlya Dryomov * cleared in __cancel_balance. 4753c9e9f97bSIlya Dryomov */ 4754ed0fb78fSIlya Dryomov need_unlock = false; 4755ed0fb78fSIlya Dryomov 4756ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 47570f89abf5SChristian Engelmayer bctl = NULL; 4758ed0fb78fSIlya Dryomov 4759c9e9f97bSIlya Dryomov if (arg) { 4760c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4761c9e9f97bSIlya Dryomov ret = -EFAULT; 4762c9e9f97bSIlya Dryomov } 4763c9e9f97bSIlya Dryomov 47640f89abf5SChristian Engelmayer out_bctl: 47650f89abf5SChristian Engelmayer kfree(bctl); 4766c9e9f97bSIlya Dryomov out_bargs: 4767c9e9f97bSIlya Dryomov kfree(bargs); 4768ed0fb78fSIlya Dryomov out_unlock: 4769c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4770c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4771ed0fb78fSIlya Dryomov if (need_unlock) 4772ed0fb78fSIlya Dryomov atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 4773ed0fb78fSIlya Dryomov out: 4774e54bfa31SLiu Bo mnt_drop_write_file(file); 4775c9e9f97bSIlya Dryomov return ret; 4776c9e9f97bSIlya Dryomov } 4777c9e9f97bSIlya Dryomov 47782ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4779837d5b6eSIlya Dryomov { 4780837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4781837d5b6eSIlya Dryomov return -EPERM; 4782837d5b6eSIlya Dryomov 4783837d5b6eSIlya Dryomov switch (cmd) { 4784837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 47850b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4786a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47870b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4788837d5b6eSIlya Dryomov } 4789837d5b6eSIlya Dryomov 4790837d5b6eSIlya Dryomov return -EINVAL; 4791837d5b6eSIlya Dryomov } 4792837d5b6eSIlya Dryomov 47932ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 479419a39dceSIlya Dryomov void __user *arg) 479519a39dceSIlya Dryomov { 479619a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 479719a39dceSIlya Dryomov int ret = 0; 479819a39dceSIlya Dryomov 479919a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 480019a39dceSIlya Dryomov return -EPERM; 480119a39dceSIlya Dryomov 480219a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 480319a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 480419a39dceSIlya Dryomov ret = -ENOTCONN; 480519a39dceSIlya Dryomov goto out; 480619a39dceSIlya Dryomov } 480719a39dceSIlya Dryomov 48088d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 480919a39dceSIlya Dryomov if (!bargs) { 481019a39dceSIlya Dryomov ret = -ENOMEM; 481119a39dceSIlya Dryomov goto out; 481219a39dceSIlya Dryomov } 481319a39dceSIlya Dryomov 481419a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 481519a39dceSIlya Dryomov 481619a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 481719a39dceSIlya Dryomov ret = -EFAULT; 481819a39dceSIlya Dryomov 481919a39dceSIlya Dryomov kfree(bargs); 482019a39dceSIlya Dryomov out: 482119a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 482219a39dceSIlya Dryomov return ret; 482319a39dceSIlya Dryomov } 482419a39dceSIlya Dryomov 4825905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 48265d13a37bSArne Jansen { 48270b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48280b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48295d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 48305d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 48315d13a37bSArne Jansen int ret; 48325d13a37bSArne Jansen int err; 48335d13a37bSArne Jansen 48345d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48355d13a37bSArne Jansen return -EPERM; 48365d13a37bSArne Jansen 4837905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4838905b0ddaSMiao Xie if (ret) 4839905b0ddaSMiao Xie return ret; 48405d13a37bSArne Jansen 48415d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4842905b0ddaSMiao Xie if (IS_ERR(sa)) { 4843905b0ddaSMiao Xie ret = PTR_ERR(sa); 4844905b0ddaSMiao Xie goto drop_write; 4845905b0ddaSMiao Xie } 48465d13a37bSArne Jansen 48470b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48480b246afaSJeff Mahoney trans = btrfs_start_transaction(fs_info->tree_root, 2); 48495d13a37bSArne Jansen if (IS_ERR(trans)) { 48505d13a37bSArne Jansen ret = PTR_ERR(trans); 48515d13a37bSArne Jansen goto out; 48525d13a37bSArne Jansen } 48535d13a37bSArne Jansen 48545d13a37bSArne Jansen switch (sa->cmd) { 48555d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 48560b246afaSJeff Mahoney ret = btrfs_quota_enable(trans, fs_info); 48575d13a37bSArne Jansen break; 48585d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 48590b246afaSJeff Mahoney ret = btrfs_quota_disable(trans, fs_info); 48605d13a37bSArne Jansen break; 48615d13a37bSArne Jansen default: 48625d13a37bSArne Jansen ret = -EINVAL; 48635d13a37bSArne Jansen break; 48645d13a37bSArne Jansen } 48655d13a37bSArne Jansen 48663a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 48675d13a37bSArne Jansen if (err && !ret) 48685d13a37bSArne Jansen ret = err; 48695d13a37bSArne Jansen out: 48705d13a37bSArne Jansen kfree(sa); 48710b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4872905b0ddaSMiao Xie drop_write: 4873905b0ddaSMiao Xie mnt_drop_write_file(file); 48745d13a37bSArne Jansen return ret; 48755d13a37bSArne Jansen } 48765d13a37bSArne Jansen 4877905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48785d13a37bSArne Jansen { 48790b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48800b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48810b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48825d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48835d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48845d13a37bSArne Jansen int ret; 48855d13a37bSArne Jansen int err; 48865d13a37bSArne Jansen 48875d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48885d13a37bSArne Jansen return -EPERM; 48895d13a37bSArne Jansen 4890905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4891905b0ddaSMiao Xie if (ret) 4892905b0ddaSMiao Xie return ret; 48935d13a37bSArne Jansen 48945d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4895905b0ddaSMiao Xie if (IS_ERR(sa)) { 4896905b0ddaSMiao Xie ret = PTR_ERR(sa); 4897905b0ddaSMiao Xie goto drop_write; 4898905b0ddaSMiao Xie } 48995d13a37bSArne Jansen 49005d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49015d13a37bSArne Jansen if (IS_ERR(trans)) { 49025d13a37bSArne Jansen ret = PTR_ERR(trans); 49035d13a37bSArne Jansen goto out; 49045d13a37bSArne Jansen } 49055d13a37bSArne Jansen 49065d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49075d13a37bSArne Jansen if (sa->assign) { 49080b246afaSJeff Mahoney ret = btrfs_add_qgroup_relation(trans, fs_info, 49095d13a37bSArne Jansen sa->src, sa->dst); 49105d13a37bSArne Jansen } else { 49110b246afaSJeff Mahoney ret = btrfs_del_qgroup_relation(trans, fs_info, 49125d13a37bSArne Jansen sa->src, sa->dst); 49135d13a37bSArne Jansen } 49145d13a37bSArne Jansen 4915e082f563SQu Wenruo /* update qgroup status and info */ 49160b246afaSJeff Mahoney err = btrfs_run_qgroups(trans, fs_info); 4917e082f563SQu Wenruo if (err < 0) 49180b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4919ad8403dfSAnand Jain "failed to update qgroup status and info"); 49203a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49215d13a37bSArne Jansen if (err && !ret) 49225d13a37bSArne Jansen ret = err; 49235d13a37bSArne Jansen 49245d13a37bSArne Jansen out: 49255d13a37bSArne Jansen kfree(sa); 4926905b0ddaSMiao Xie drop_write: 4927905b0ddaSMiao Xie mnt_drop_write_file(file); 49285d13a37bSArne Jansen return ret; 49295d13a37bSArne Jansen } 49305d13a37bSArne Jansen 4931905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 49325d13a37bSArne Jansen { 49330b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49340b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49350b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49365d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49375d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49385d13a37bSArne Jansen int ret; 49395d13a37bSArne Jansen int err; 49405d13a37bSArne Jansen 49415d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49425d13a37bSArne Jansen return -EPERM; 49435d13a37bSArne Jansen 4944905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4945905b0ddaSMiao Xie if (ret) 4946905b0ddaSMiao Xie return ret; 49475d13a37bSArne Jansen 49485d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4949905b0ddaSMiao Xie if (IS_ERR(sa)) { 4950905b0ddaSMiao Xie ret = PTR_ERR(sa); 4951905b0ddaSMiao Xie goto drop_write; 4952905b0ddaSMiao Xie } 49535d13a37bSArne Jansen 4954d86e56cfSMiao Xie if (!sa->qgroupid) { 4955d86e56cfSMiao Xie ret = -EINVAL; 4956d86e56cfSMiao Xie goto out; 4957d86e56cfSMiao Xie } 4958d86e56cfSMiao Xie 49595d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49605d13a37bSArne Jansen if (IS_ERR(trans)) { 49615d13a37bSArne Jansen ret = PTR_ERR(trans); 49625d13a37bSArne Jansen goto out; 49635d13a37bSArne Jansen } 49645d13a37bSArne Jansen 49655d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49665d13a37bSArne Jansen if (sa->create) { 49670b246afaSJeff Mahoney ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); 49685d13a37bSArne Jansen } else { 49690b246afaSJeff Mahoney ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); 49705d13a37bSArne Jansen } 49715d13a37bSArne Jansen 49723a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49735d13a37bSArne Jansen if (err && !ret) 49745d13a37bSArne Jansen ret = err; 49755d13a37bSArne Jansen 49765d13a37bSArne Jansen out: 49775d13a37bSArne Jansen kfree(sa); 4978905b0ddaSMiao Xie drop_write: 4979905b0ddaSMiao Xie mnt_drop_write_file(file); 49805d13a37bSArne Jansen return ret; 49815d13a37bSArne Jansen } 49825d13a37bSArne Jansen 4983905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49845d13a37bSArne Jansen { 49850b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49870b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49885d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49895d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49905d13a37bSArne Jansen int ret; 49915d13a37bSArne Jansen int err; 49925d13a37bSArne Jansen u64 qgroupid; 49935d13a37bSArne Jansen 49945d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49955d13a37bSArne Jansen return -EPERM; 49965d13a37bSArne Jansen 4997905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4998905b0ddaSMiao Xie if (ret) 4999905b0ddaSMiao Xie return ret; 50005d13a37bSArne Jansen 50015d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 5002905b0ddaSMiao Xie if (IS_ERR(sa)) { 5003905b0ddaSMiao Xie ret = PTR_ERR(sa); 5004905b0ddaSMiao Xie goto drop_write; 5005905b0ddaSMiao Xie } 50065d13a37bSArne Jansen 50075d13a37bSArne Jansen trans = btrfs_join_transaction(root); 50085d13a37bSArne Jansen if (IS_ERR(trans)) { 50095d13a37bSArne Jansen ret = PTR_ERR(trans); 50105d13a37bSArne Jansen goto out; 50115d13a37bSArne Jansen } 50125d13a37bSArne Jansen 50135d13a37bSArne Jansen qgroupid = sa->qgroupid; 50145d13a37bSArne Jansen if (!qgroupid) { 50155d13a37bSArne Jansen /* take the current subvol as qgroup */ 50165d13a37bSArne Jansen qgroupid = root->root_key.objectid; 50175d13a37bSArne Jansen } 50185d13a37bSArne Jansen 50195d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 50200b246afaSJeff Mahoney ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); 50215d13a37bSArne Jansen 50223a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 50235d13a37bSArne Jansen if (err && !ret) 50245d13a37bSArne Jansen ret = err; 50255d13a37bSArne Jansen 50265d13a37bSArne Jansen out: 50275d13a37bSArne Jansen kfree(sa); 5028905b0ddaSMiao Xie drop_write: 5029905b0ddaSMiao Xie mnt_drop_write_file(file); 50305d13a37bSArne Jansen return ret; 50315d13a37bSArne Jansen } 50325d13a37bSArne Jansen 50332f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 50342f232036SJan Schmidt { 50350b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50360b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50372f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50382f232036SJan Schmidt int ret; 50392f232036SJan Schmidt 50402f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50412f232036SJan Schmidt return -EPERM; 50422f232036SJan Schmidt 50432f232036SJan Schmidt ret = mnt_want_write_file(file); 50442f232036SJan Schmidt if (ret) 50452f232036SJan Schmidt return ret; 50462f232036SJan Schmidt 50472f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50482f232036SJan Schmidt if (IS_ERR(qsa)) { 50492f232036SJan Schmidt ret = PTR_ERR(qsa); 50502f232036SJan Schmidt goto drop_write; 50512f232036SJan Schmidt } 50522f232036SJan Schmidt 50532f232036SJan Schmidt if (qsa->flags) { 50542f232036SJan Schmidt ret = -EINVAL; 50552f232036SJan Schmidt goto out; 50562f232036SJan Schmidt } 50572f232036SJan Schmidt 50580b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 50592f232036SJan Schmidt 50602f232036SJan Schmidt out: 50612f232036SJan Schmidt kfree(qsa); 50622f232036SJan Schmidt drop_write: 50632f232036SJan Schmidt mnt_drop_write_file(file); 50642f232036SJan Schmidt return ret; 50652f232036SJan Schmidt } 50662f232036SJan Schmidt 50672f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50682f232036SJan Schmidt { 50690b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50712f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50722f232036SJan Schmidt int ret = 0; 50732f232036SJan Schmidt 50742f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50752f232036SJan Schmidt return -EPERM; 50762f232036SJan Schmidt 50778d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 50782f232036SJan Schmidt if (!qsa) 50792f232036SJan Schmidt return -ENOMEM; 50802f232036SJan Schmidt 50810b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50822f232036SJan Schmidt qsa->flags = 1; 50830b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 50842f232036SJan Schmidt } 50852f232036SJan Schmidt 50862f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50872f232036SJan Schmidt ret = -EFAULT; 50882f232036SJan Schmidt 50892f232036SJan Schmidt kfree(qsa); 50902f232036SJan Schmidt return ret; 50912f232036SJan Schmidt } 50922f232036SJan Schmidt 509357254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 509457254b6eSJan Schmidt { 50950b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50960b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 509757254b6eSJan Schmidt 509857254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 509957254b6eSJan Schmidt return -EPERM; 510057254b6eSJan Schmidt 51010b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 510257254b6eSJan Schmidt } 510357254b6eSJan Schmidt 5104abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5105abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 51068ea05e3aSAlexander Block { 5107496ad9aaSAl Viro struct inode *inode = file_inode(file); 51080b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 51098ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 51108ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 51118ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 5112c2050a45SDeepa Dinamani struct timespec ct = current_time(inode); 51138ea05e3aSAlexander Block int ret = 0; 5114dd5f9615SStefan Behrens int received_uuid_changed; 51158ea05e3aSAlexander Block 5116bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5117bd60ea0fSDavid Sterba return -EPERM; 5118bd60ea0fSDavid Sterba 51198ea05e3aSAlexander Block ret = mnt_want_write_file(file); 51208ea05e3aSAlexander Block if (ret < 0) 51218ea05e3aSAlexander Block return ret; 51228ea05e3aSAlexander Block 51230b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51248ea05e3aSAlexander Block 51254a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 51268ea05e3aSAlexander Block ret = -EINVAL; 51278ea05e3aSAlexander Block goto out; 51288ea05e3aSAlexander Block } 51298ea05e3aSAlexander Block 51308ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 51318ea05e3aSAlexander Block ret = -EROFS; 51328ea05e3aSAlexander Block goto out; 51338ea05e3aSAlexander Block } 51348ea05e3aSAlexander Block 5135dd5f9615SStefan Behrens /* 5136dd5f9615SStefan Behrens * 1 - root item 5137dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5138dd5f9615SStefan Behrens */ 5139dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51408ea05e3aSAlexander Block if (IS_ERR(trans)) { 51418ea05e3aSAlexander Block ret = PTR_ERR(trans); 51428ea05e3aSAlexander Block trans = NULL; 51438ea05e3aSAlexander Block goto out; 51448ea05e3aSAlexander Block } 51458ea05e3aSAlexander Block 51468ea05e3aSAlexander Block sa->rtransid = trans->transid; 51478ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51488ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51498ea05e3aSAlexander Block 5150dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5151dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5152dd5f9615SStefan Behrens if (received_uuid_changed && 5153dd5f9615SStefan Behrens !btrfs_is_empty_uuid(root_item->received_uuid)) 51540b246afaSJeff Mahoney btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid, 5155dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5156dd5f9615SStefan Behrens root->root_key.objectid); 51578ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51588ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51598ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51603cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51613cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51623cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51633cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51648ea05e3aSAlexander Block 51650b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 51668ea05e3aSAlexander Block &root->root_key, &root->root_item); 51678ea05e3aSAlexander Block if (ret < 0) { 51683a45bb20SJeff Mahoney btrfs_end_transaction(trans); 51698ea05e3aSAlexander Block goto out; 5170dd5f9615SStefan Behrens } 5171dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 51720b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid, 5173dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5174dd5f9615SStefan Behrens root->root_key.objectid); 5175dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 517666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5177dd5f9615SStefan Behrens goto out; 5178dd5f9615SStefan Behrens } 5179dd5f9615SStefan Behrens } 51803a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5181dd5f9615SStefan Behrens if (ret < 0) { 518266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 51838ea05e3aSAlexander Block goto out; 51848ea05e3aSAlexander Block } 51858ea05e3aSAlexander Block 5186abccd00fSHugo Mills out: 51870b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5188abccd00fSHugo Mills mnt_drop_write_file(file); 5189abccd00fSHugo Mills return ret; 5190abccd00fSHugo Mills } 5191abccd00fSHugo Mills 5192abccd00fSHugo Mills #ifdef CONFIG_64BIT 5193abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5194abccd00fSHugo Mills void __user *arg) 5195abccd00fSHugo Mills { 5196abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5197abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5198abccd00fSHugo Mills int ret = 0; 5199abccd00fSHugo Mills 5200abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 52017b9ea627SShailendra Verma if (IS_ERR(args32)) 52027b9ea627SShailendra Verma return PTR_ERR(args32); 5203abccd00fSHugo Mills 52048d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 520584dbeb87SDan Carpenter if (!args64) { 520684dbeb87SDan Carpenter ret = -ENOMEM; 5207abccd00fSHugo Mills goto out; 5208abccd00fSHugo Mills } 5209abccd00fSHugo Mills 5210abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5211abccd00fSHugo Mills args64->stransid = args32->stransid; 5212abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5213abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5214abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5215abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5216abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5217abccd00fSHugo Mills args64->flags = args32->flags; 5218abccd00fSHugo Mills 5219abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5220abccd00fSHugo Mills if (ret) 5221abccd00fSHugo Mills goto out; 5222abccd00fSHugo Mills 5223abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5224abccd00fSHugo Mills args32->stransid = args64->stransid; 5225abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5226abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5227abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5228abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5229abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5230abccd00fSHugo Mills args32->flags = args64->flags; 5231abccd00fSHugo Mills 5232abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5233abccd00fSHugo Mills if (ret) 5234abccd00fSHugo Mills ret = -EFAULT; 5235abccd00fSHugo Mills 5236abccd00fSHugo Mills out: 5237abccd00fSHugo Mills kfree(args32); 5238abccd00fSHugo Mills kfree(args64); 5239abccd00fSHugo Mills return ret; 5240abccd00fSHugo Mills } 5241abccd00fSHugo Mills #endif 5242abccd00fSHugo Mills 5243abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5244abccd00fSHugo Mills void __user *arg) 5245abccd00fSHugo Mills { 5246abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5247abccd00fSHugo Mills int ret = 0; 5248abccd00fSHugo Mills 5249abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 52507b9ea627SShailendra Verma if (IS_ERR(sa)) 52517b9ea627SShailendra Verma return PTR_ERR(sa); 5252abccd00fSHugo Mills 5253abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5254abccd00fSHugo Mills 5255abccd00fSHugo Mills if (ret) 5256abccd00fSHugo Mills goto out; 5257abccd00fSHugo Mills 52588ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52598ea05e3aSAlexander Block if (ret) 52608ea05e3aSAlexander Block ret = -EFAULT; 52618ea05e3aSAlexander Block 52628ea05e3aSAlexander Block out: 52638ea05e3aSAlexander Block kfree(sa); 52648ea05e3aSAlexander Block return ret; 52658ea05e3aSAlexander Block } 52668ea05e3aSAlexander Block 5267867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5268867ab667Sjeff.liu { 52690b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5271a1b83ac5SAnand Jain size_t len; 5272867ab667Sjeff.liu int ret; 5273a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5274a1b83ac5SAnand Jain 52750b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 52760b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 52770b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5278a1b83ac5SAnand Jain 5279a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5280867ab667Sjeff.liu 5281867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 52820b246afaSJeff Mahoney btrfs_warn(fs_info, 52830b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 52840b246afaSJeff Mahoney --len); 5285867ab667Sjeff.liu } 5286867ab667Sjeff.liu 5287867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5288867ab667Sjeff.liu 5289867ab667Sjeff.liu return ret ? -EFAULT : 0; 5290867ab667Sjeff.liu } 5291867ab667Sjeff.liu 5292a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5293a8bfd4abSjeff.liu { 52940b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52950b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52960b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52970b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5298a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5299a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5300a8bfd4abSjeff.liu int ret; 5301a8bfd4abSjeff.liu 5302a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5303a8bfd4abSjeff.liu return -EPERM; 5304a8bfd4abSjeff.liu 5305a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5306a8bfd4abSjeff.liu return -EFAULT; 5307a8bfd4abSjeff.liu 5308a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 53090b246afaSJeff Mahoney btrfs_err(fs_info, 53105d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5311a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5312a8bfd4abSjeff.liu return -EINVAL; 5313a8bfd4abSjeff.liu } 5314a8bfd4abSjeff.liu 5315a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5316a8bfd4abSjeff.liu if (ret) 5317a8bfd4abSjeff.liu return ret; 5318a8bfd4abSjeff.liu 5319a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5320a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5321a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5322a8bfd4abSjeff.liu goto out_unlock; 5323a8bfd4abSjeff.liu } 5324a8bfd4abSjeff.liu 53250b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5326a8bfd4abSjeff.liu strcpy(super_block->label, label); 53270b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53283a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5329a8bfd4abSjeff.liu 5330a8bfd4abSjeff.liu out_unlock: 5331a8bfd4abSjeff.liu mnt_drop_write_file(file); 5332a8bfd4abSjeff.liu return ret; 5333a8bfd4abSjeff.liu } 5334a8bfd4abSjeff.liu 53352eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 53362eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 53372eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53382eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53392eaa055fSJeff Mahoney 5340d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53412eaa055fSJeff Mahoney { 53424d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53432eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53442eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53452eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53462eaa055fSJeff Mahoney }; 53472eaa055fSJeff Mahoney 53482eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53492eaa055fSJeff Mahoney return -EFAULT; 53502eaa055fSJeff Mahoney 53512eaa055fSJeff Mahoney return 0; 53522eaa055fSJeff Mahoney } 53532eaa055fSJeff Mahoney 53542eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53552eaa055fSJeff Mahoney { 53560b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53570b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53580b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53592eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53602eaa055fSJeff Mahoney 53612eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53622eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53632eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53642eaa055fSJeff Mahoney 53652eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53662eaa055fSJeff Mahoney return -EFAULT; 53672eaa055fSJeff Mahoney 53682eaa055fSJeff Mahoney return 0; 53692eaa055fSJeff Mahoney } 53702eaa055fSJeff Mahoney 53712ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 53723b02a68aSJeff Mahoney enum btrfs_feature_set set, 53732eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53742eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53752eaa055fSJeff Mahoney { 53763b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 53773b02a68aSJeff Mahoney char *names; 53782eaa055fSJeff Mahoney u64 disallowed, unsupported; 53792eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53802eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53812eaa055fSJeff Mahoney 53822eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53832eaa055fSJeff Mahoney if (unsupported) { 53843b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53853b02a68aSJeff Mahoney if (names) { 53860b246afaSJeff Mahoney btrfs_warn(fs_info, 53873b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53883b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53893b02a68aSJeff Mahoney kfree(names); 53903b02a68aSJeff Mahoney } else 53910b246afaSJeff Mahoney btrfs_warn(fs_info, 53922eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53932eaa055fSJeff Mahoney type, unsupported); 53942eaa055fSJeff Mahoney return -EOPNOTSUPP; 53952eaa055fSJeff Mahoney } 53962eaa055fSJeff Mahoney 53972eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53982eaa055fSJeff Mahoney if (disallowed) { 53993b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54003b02a68aSJeff Mahoney if (names) { 54010b246afaSJeff Mahoney btrfs_warn(fs_info, 54023b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 54033b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54043b02a68aSJeff Mahoney kfree(names); 54053b02a68aSJeff Mahoney } else 54060b246afaSJeff Mahoney btrfs_warn(fs_info, 54072eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 54082eaa055fSJeff Mahoney type, disallowed); 54092eaa055fSJeff Mahoney return -EPERM; 54102eaa055fSJeff Mahoney } 54112eaa055fSJeff Mahoney 54122eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 54132eaa055fSJeff Mahoney if (disallowed) { 54143b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54153b02a68aSJeff Mahoney if (names) { 54160b246afaSJeff Mahoney btrfs_warn(fs_info, 54173b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 54183b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54193b02a68aSJeff Mahoney kfree(names); 54203b02a68aSJeff Mahoney } else 54210b246afaSJeff Mahoney btrfs_warn(fs_info, 54222eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 54232eaa055fSJeff Mahoney type, disallowed); 54242eaa055fSJeff Mahoney return -EPERM; 54252eaa055fSJeff Mahoney } 54262eaa055fSJeff Mahoney 54272eaa055fSJeff Mahoney return 0; 54282eaa055fSJeff Mahoney } 54292eaa055fSJeff Mahoney 54302ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 54312ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 54322eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 54332eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 54342eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 54352eaa055fSJeff Mahoney 54362eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 54372eaa055fSJeff Mahoney { 54380b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54400b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54410b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 54422eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54432eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54442eaa055fSJeff Mahoney u64 newflags; 54452eaa055fSJeff Mahoney int ret; 54462eaa055fSJeff Mahoney 54472eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54482eaa055fSJeff Mahoney return -EPERM; 54492eaa055fSJeff Mahoney 54502eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54512eaa055fSJeff Mahoney return -EFAULT; 54522eaa055fSJeff Mahoney 54532eaa055fSJeff Mahoney /* Nothing to do */ 54542eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54552eaa055fSJeff Mahoney !flags[0].incompat_flags) 54562eaa055fSJeff Mahoney return 0; 54572eaa055fSJeff Mahoney 54582ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 54592eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54602eaa055fSJeff Mahoney if (ret) 54612eaa055fSJeff Mahoney return ret; 54622eaa055fSJeff Mahoney 54632ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 54642eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54652eaa055fSJeff Mahoney if (ret) 54662eaa055fSJeff Mahoney return ret; 54672eaa055fSJeff Mahoney 54682ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 54692eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54702eaa055fSJeff Mahoney if (ret) 54712eaa055fSJeff Mahoney return ret; 54722eaa055fSJeff Mahoney 54737ab19625SDavid Sterba ret = mnt_want_write_file(file); 54747ab19625SDavid Sterba if (ret) 54757ab19625SDavid Sterba return ret; 54767ab19625SDavid Sterba 54778051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54787ab19625SDavid Sterba if (IS_ERR(trans)) { 54797ab19625SDavid Sterba ret = PTR_ERR(trans); 54807ab19625SDavid Sterba goto out_drop_write; 54817ab19625SDavid Sterba } 54822eaa055fSJeff Mahoney 54830b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 54842eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54852eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54862eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54872eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54882eaa055fSJeff Mahoney 54892eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54902eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54912eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54922eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54932eaa055fSJeff Mahoney 54942eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54952eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54962eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54972eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54980b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 54992eaa055fSJeff Mahoney 55003a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 55017ab19625SDavid Sterba out_drop_write: 55027ab19625SDavid Sterba mnt_drop_write_file(file); 55037ab19625SDavid Sterba 55047ab19625SDavid Sterba return ret; 55052eaa055fSJeff Mahoney } 55062eaa055fSJeff Mahoney 5507f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5508f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5509f46b5a66SChristoph Hellwig { 55100b246afaSJeff Mahoney struct inode *inode = file_inode(file); 55110b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 55120b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 55134bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5514f46b5a66SChristoph Hellwig 5515f46b5a66SChristoph Hellwig switch (cmd) { 55166cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 55176cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 55186cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 55196cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 55206cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 55216cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5522f7039b1dSLi Dongyang case FITRIM: 5523f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5524f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5525fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5526fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5527fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 55283de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5529fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 55306f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 55316f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 553276dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 553376dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 55340caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 55350caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 55360caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 55370caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55386ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55396ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5540f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55411e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55421e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55431e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5544f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5545198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5546f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55472ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5548f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5549da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55506b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55516b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5552475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 55532ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5554475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 55552ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5556f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55579ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5558f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 5559f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 5560f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 5561f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 5562ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5563ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5564cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5565cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5566ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5567ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5568d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5569d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5570d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 55712ff7e61eSJeff Mahoney return btrfs_ioctl_logical_to_ino(fs_info, argp); 55721406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55732ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 55749b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55759b199859SFilipe David Borba Manana int ret; 55769b199859SFilipe David Borba Manana 55770b246afaSJeff Mahoney ret = btrfs_start_delalloc_roots(fs_info, 0, -1); 55789b199859SFilipe David Borba Manana if (ret) 55799b199859SFilipe David Borba Manana return ret; 55800b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 55812fad4e83SDavid Sterba /* 55822fad4e83SDavid Sterba * The transaction thread may want to do more work, 558301327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 55842fad4e83SDavid Sterba * processing uncleaned subvols. 55852fad4e83SDavid Sterba */ 55860b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 55879b199859SFilipe David Borba Manana return ret; 55889b199859SFilipe David Borba Manana } 558946204592SSage Weil case BTRFS_IOC_START_SYNC: 55909a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 559146204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55922ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5593475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5594b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5595475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 55962ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5597475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 55982ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5599c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 56009ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5601837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 56022ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 560319a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 56042ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 56058ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 56068ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5607abccd00fSHugo Mills #ifdef CONFIG_64BIT 5608abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5609abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5610abccd00fSHugo Mills #endif 561131db9f7cSAlexander Block case BTRFS_IOC_SEND: 561231db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 5613c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 56142ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 56155d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5616905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 56175d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5618905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 56195d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5620905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 56215d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5622905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 56232f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 56242f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 56252f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 56262f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 562757254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 562857254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 56293f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 56302ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5631867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5632867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5633a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5634a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 56352eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5636d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 56372eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 56382eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 56392eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56402eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5641f46b5a66SChristoph Hellwig } 5642f46b5a66SChristoph Hellwig 5643f46b5a66SChristoph Hellwig return -ENOTTY; 5644f46b5a66SChristoph Hellwig } 56454c63c245SLuke Dashjr 56464c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56474c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56484c63c245SLuke Dashjr { 56492a362249SJeff Mahoney /* 56502a362249SJeff Mahoney * These all access 32-bit values anyway so no further 56512a362249SJeff Mahoney * handling is necessary. 56522a362249SJeff Mahoney */ 56534c63c245SLuke Dashjr switch (cmd) { 56544c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56554c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56564c63c245SLuke Dashjr break; 56574c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56584c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56594c63c245SLuke Dashjr break; 56604c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56614c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56624c63c245SLuke Dashjr break; 56634c63c245SLuke Dashjr } 56644c63c245SLuke Dashjr 56654c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56664c63c245SLuke Dashjr } 56674c63c245SLuke Dashjr #endif 5668