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/statfs.h> 37f46b5a66SChristoph Hellwig #include <linux/compat.h> 38f46b5a66SChristoph Hellwig #include <linux/bit_spinlock.h> 39cb8e7090SChristoph Hellwig #include <linux/security.h> 40f46b5a66SChristoph Hellwig #include <linux/xattr.h> 417ea394f1SYan Zheng #include <linux/vmalloc.h> 425a0e3ad6STejun Heo #include <linux/slab.h> 43f7039b1dSLi Dongyang #include <linux/blkdev.h> 448ea05e3aSAlexander Block #include <linux/uuid.h> 4555e301fdSFilipe Brandenburger #include <linux/btrfs.h> 46416161dbSMark Fasheh #include <linux/uaccess.h> 47f46b5a66SChristoph Hellwig #include "ctree.h" 48f46b5a66SChristoph Hellwig #include "disk-io.h" 49f46b5a66SChristoph Hellwig #include "transaction.h" 50f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 51f46b5a66SChristoph Hellwig #include "print-tree.h" 52f46b5a66SChristoph Hellwig #include "volumes.h" 53925baeddSChris Mason #include "locking.h" 54581bb050SLi Zefan #include "inode-map.h" 55d7728c96SJan Schmidt #include "backref.h" 56606686eeSJosef Bacik #include "rcu-string.h" 5731db9f7cSAlexander Block #include "send.h" 583f6bcfbdSStefan Behrens #include "dev-replace.h" 5963541927SFilipe David Borba Manana #include "props.h" 603b02a68aSJeff Mahoney #include "sysfs.h" 61fcebe456SJosef Bacik #include "qgroup.h" 621ec9a1aeSFilipe Manana #include "tree-log.h" 63ebb8765bSAnand Jain #include "compression.h" 64f46b5a66SChristoph Hellwig 65abccd00fSHugo Mills #ifdef CONFIG_64BIT 66abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 67abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 68abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 69abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 70abccd00fSHugo Mills */ 71abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 72abccd00fSHugo Mills __u64 sec; 73abccd00fSHugo Mills __u32 nsec; 74abccd00fSHugo Mills } __attribute__ ((__packed__)); 75abccd00fSHugo Mills 76abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 77abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 78abccd00fSHugo Mills __u64 stransid; /* in */ 79abccd00fSHugo Mills __u64 rtransid; /* out */ 80abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 81abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 82abccd00fSHugo Mills __u64 flags; /* in */ 83abccd00fSHugo Mills __u64 reserved[16]; /* in */ 84abccd00fSHugo Mills } __attribute__ ((__packed__)); 85abccd00fSHugo Mills 86abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 87abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 88abccd00fSHugo Mills #endif 89abccd00fSHugo Mills 90abccd00fSHugo Mills 91416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 921c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 931c919a5eSMark Fasheh int no_time_update); 94416161dbSMark Fasheh 956cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 966cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 976cbff00fSChristoph Hellwig { 986cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 996cbff00fSChristoph Hellwig return flags; 1006cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 1016cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1026cbff00fSChristoph Hellwig else 1036cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1046cbff00fSChristoph Hellwig } 105f46b5a66SChristoph Hellwig 1066cbff00fSChristoph Hellwig /* 1076cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 1086cbff00fSChristoph Hellwig */ 1096cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 1106cbff00fSChristoph Hellwig { 1116cbff00fSChristoph Hellwig unsigned int iflags = 0; 1126cbff00fSChristoph Hellwig 1136cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1146cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1156cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1166cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1176cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1186cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1196cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1206cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1216cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1226cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1236cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1246cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 125d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 126d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 127d0092bddSLi Zefan 12813f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 129d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 13013f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13113f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1326cbff00fSChristoph Hellwig 1336cbff00fSChristoph Hellwig return iflags; 1346cbff00fSChristoph Hellwig } 1356cbff00fSChristoph Hellwig 1366cbff00fSChristoph Hellwig /* 1376cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1386cbff00fSChristoph Hellwig */ 1396cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1406cbff00fSChristoph Hellwig { 1416cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1423cc79392SFilipe Manana unsigned int new_fl = 0; 1436cbff00fSChristoph Hellwig 1446cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1453cc79392SFilipe Manana new_fl |= S_SYNC; 1466cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1473cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1486cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1493cc79392SFilipe Manana new_fl |= S_APPEND; 1506cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1513cc79392SFilipe Manana new_fl |= S_NOATIME; 1526cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1533cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1543cc79392SFilipe Manana 1553cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1563cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1573cc79392SFilipe Manana new_fl); 1586cbff00fSChristoph Hellwig } 1596cbff00fSChristoph Hellwig 1606cbff00fSChristoph Hellwig /* 1616cbff00fSChristoph Hellwig * Inherit flags from the parent inode. 1626cbff00fSChristoph Hellwig * 163e27425d6SJosef Bacik * Currently only the compression flags and the cow flags are inherited. 1646cbff00fSChristoph Hellwig */ 1656cbff00fSChristoph Hellwig void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 1666cbff00fSChristoph Hellwig { 1670b4dcea5SChris Mason unsigned int flags; 1680b4dcea5SChris Mason 1690b4dcea5SChris Mason if (!dir) 1700b4dcea5SChris Mason return; 1710b4dcea5SChris Mason 1720b4dcea5SChris Mason flags = BTRFS_I(dir)->flags; 1736cbff00fSChristoph Hellwig 174e27425d6SJosef Bacik if (flags & BTRFS_INODE_NOCOMPRESS) { 175e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 176e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 177e27425d6SJosef Bacik } else if (flags & BTRFS_INODE_COMPRESS) { 178e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 179e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 180e27425d6SJosef Bacik } 1816cbff00fSChristoph Hellwig 182213490b3SLiu Bo if (flags & BTRFS_INODE_NODATACOW) { 183e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 184213490b3SLiu Bo if (S_ISREG(inode->i_mode)) 185213490b3SLiu Bo BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; 186213490b3SLiu Bo } 187e27425d6SJosef Bacik 1886cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1896cbff00fSChristoph Hellwig } 1906cbff00fSChristoph Hellwig 1916cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1926cbff00fSChristoph Hellwig { 193496ad9aaSAl Viro struct btrfs_inode *ip = BTRFS_I(file_inode(file)); 1946cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1956cbff00fSChristoph Hellwig 1966cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1976cbff00fSChristoph Hellwig return -EFAULT; 1986cbff00fSChristoph Hellwig return 0; 1996cbff00fSChristoph Hellwig } 2006cbff00fSChristoph Hellwig 20175e7cb7fSLiu Bo static int check_flags(unsigned int flags) 20275e7cb7fSLiu Bo { 20375e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 20475e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 20575e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 206e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 207e1e8fb6aSLi Zefan FS_NOCOW_FL)) 20875e7cb7fSLiu Bo return -EOPNOTSUPP; 20975e7cb7fSLiu Bo 21075e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 21175e7cb7fSLiu Bo return -EINVAL; 21275e7cb7fSLiu Bo 21375e7cb7fSLiu Bo return 0; 21475e7cb7fSLiu Bo } 21575e7cb7fSLiu Bo 2166cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 2176cbff00fSChristoph Hellwig { 218496ad9aaSAl Viro struct inode *inode = file_inode(file); 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 32863541927SFilipe David Borba Manana if (root->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); 35204b285f3SDeepa Dinamani inode->i_ctime = current_fs_time(inode->i_sb); 3536cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3546cbff00fSChristoph Hellwig 3556cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 356f062abf0SLi Zefan out_drop: 357f062abf0SLi Zefan if (ret) { 358f062abf0SLi Zefan ip->flags = ip_oldflags; 359f062abf0SLi Zefan inode->i_flags = i_oldflags; 360f062abf0SLi Zefan } 3616cbff00fSChristoph Hellwig 3626cbff00fSChristoph Hellwig out_unlock: 3635955102cSAl Viro inode_unlock(inode); 364e7848683SJan Kara mnt_drop_write_file(file); 3652d4e6f6aSliubo return ret; 3666cbff00fSChristoph Hellwig } 3676cbff00fSChristoph Hellwig 3686cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 3696cbff00fSChristoph Hellwig { 370496ad9aaSAl Viro struct inode *inode = file_inode(file); 3716cbff00fSChristoph Hellwig 3726cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 3736cbff00fSChristoph Hellwig } 374f46b5a66SChristoph Hellwig 375f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 376f7039b1dSLi Dongyang { 37754563d41SAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 378f7039b1dSLi Dongyang struct btrfs_device *device; 379f7039b1dSLi Dongyang struct request_queue *q; 380f7039b1dSLi Dongyang struct fstrim_range range; 381f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 382f7039b1dSLi Dongyang u64 num_devices = 0; 383815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 384f7039b1dSLi Dongyang int ret; 385f7039b1dSLi Dongyang 386f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 387f7039b1dSLi Dongyang return -EPERM; 388f7039b1dSLi Dongyang 3891f78160cSXiao Guangrong rcu_read_lock(); 3901f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3911f78160cSXiao Guangrong dev_list) { 392f7039b1dSLi Dongyang if (!device->bdev) 393f7039b1dSLi Dongyang continue; 394f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 395f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 396f7039b1dSLi Dongyang num_devices++; 397f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 398f7039b1dSLi Dongyang minlen); 399f7039b1dSLi Dongyang } 400f7039b1dSLi Dongyang } 4011f78160cSXiao Guangrong rcu_read_unlock(); 402f4c697e6SLukas Czerner 403f7039b1dSLi Dongyang if (!num_devices) 404f7039b1dSLi Dongyang return -EOPNOTSUPP; 405f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 406f7039b1dSLi Dongyang return -EFAULT; 407e515c18bSLukas Czerner if (range.start > total_bytes || 408e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 409f4c697e6SLukas Czerner return -EINVAL; 410f7039b1dSLi Dongyang 411f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 412f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 413815745cfSAl Viro ret = btrfs_trim_fs(fs_info->tree_root, &range); 414f7039b1dSLi Dongyang if (ret < 0) 415f7039b1dSLi Dongyang return ret; 416f7039b1dSLi Dongyang 417f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 418f7039b1dSLi Dongyang return -EFAULT; 419f7039b1dSLi Dongyang 420f7039b1dSLi Dongyang return 0; 421f7039b1dSLi Dongyang } 422f7039b1dSLi Dongyang 423dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 424dd5f9615SStefan Behrens { 42546e0f66aSChris Mason int i; 42646e0f66aSChris Mason 42746e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 42846e0f66aSChris Mason if (uuid[i]) 42946e0f66aSChris Mason return 0; 43046e0f66aSChris Mason } 43146e0f66aSChris Mason return 1; 432dd5f9615SStefan Behrens } 433dd5f9615SStefan Behrens 434d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 435cb8e7090SChristoph Hellwig struct dentry *dentry, 43672fd032eSSage Weil char *name, int namelen, 4376f72c7e2SArne Jansen u64 *async_transid, 4388696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 439f46b5a66SChristoph Hellwig { 440f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 441f46b5a66SChristoph Hellwig struct btrfs_key key; 44249a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 443f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 444f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 445d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 44676dda93cSYan, Zheng struct btrfs_root *new_root; 447d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 44804b285f3SDeepa Dinamani struct timespec cur_time = current_fs_time(dir->i_sb); 4495662344bSTsutomu Itoh struct inode *inode; 450f46b5a66SChristoph Hellwig int ret; 451f46b5a66SChristoph Hellwig int err; 452f46b5a66SChristoph Hellwig u64 objectid; 453f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 4543de4586cSChris Mason u64 index = 0; 455d5c12070SMiao Xie u64 qgroup_reserved; 4568ea05e3aSAlexander Block uuid_le new_uuid; 457f46b5a66SChristoph Hellwig 45849a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 45949a3c4d9SDavid Sterba if (!root_item) 46049a3c4d9SDavid Sterba return -ENOMEM; 46149a3c4d9SDavid Sterba 462581bb050SLi Zefan ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 4632fbe8c8aSAl Viro if (ret) 46449a3c4d9SDavid Sterba goto fail_free; 4656a912213SJosef Bacik 466e09fe2d2SQu Wenruo /* 467e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 46801327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 469e09fe2d2SQu Wenruo */ 47049a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 47149a3c4d9SDavid Sterba ret = -ENOSPC; 47249a3c4d9SDavid Sterba goto fail_free; 47349a3c4d9SDavid Sterba } 474e09fe2d2SQu Wenruo 475d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 4769ed74f2dSJosef Bacik /* 477d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 478d5c12070SMiao Xie * of create_snapshot(). 4799ed74f2dSJosef Bacik */ 480d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 481dd5f9615SStefan Behrens 8, &qgroup_reserved, false); 482d5c12070SMiao Xie if (ret) 48349a3c4d9SDavid Sterba goto fail_free; 484f46b5a66SChristoph Hellwig 485d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 486d5c12070SMiao Xie if (IS_ERR(trans)) { 487d5c12070SMiao Xie ret = PTR_ERR(trans); 488de6e8200SLiu Bo btrfs_subvolume_release_metadata(root, &block_rsv, 489de6e8200SLiu Bo qgroup_reserved); 49049a3c4d9SDavid Sterba goto fail_free; 491d5c12070SMiao Xie } 492d5c12070SMiao Xie trans->block_rsv = &block_rsv; 493d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 494f46b5a66SChristoph Hellwig 4958696c533SMiao Xie ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); 4966f72c7e2SArne Jansen if (ret) 4976f72c7e2SArne Jansen goto fail; 4986f72c7e2SArne Jansen 4994d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 5008e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 5018e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 5028e8a1e31SJosef Bacik goto fail; 5038e8a1e31SJosef Bacik } 504f46b5a66SChristoph Hellwig 5055d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 506f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 507f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 5085d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 509f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 510f46b5a66SChristoph Hellwig 5110a4e5586SRoss Kirk write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(), 512f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 5135d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 514b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(leaf), 5155d4f98a2SYan Zheng BTRFS_UUID_SIZE); 516f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 517f46b5a66SChristoph Hellwig 51849a3c4d9SDavid Sterba inode_item = &root_item->inode; 5193cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 5203cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 5213cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 522707e8a07SDavid Sterba btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); 5233cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 524f46b5a66SChristoph Hellwig 52549a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 52649a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 5273cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 52808fe4db1SLi Zefan 52949a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 53049a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 53149a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 53249a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 53349a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 53449a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 535f46b5a66SChristoph Hellwig 53649a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 53749a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 5388ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 53949a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 54049a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 54149a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 54249a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 54349a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 54449a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 545f46b5a66SChristoph Hellwig 546925baeddSChris Mason btrfs_tree_unlock(leaf); 547f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 548f46b5a66SChristoph Hellwig leaf = NULL; 549f46b5a66SChristoph Hellwig 55049a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 551f46b5a66SChristoph Hellwig 552f46b5a66SChristoph Hellwig key.objectid = objectid; 5535d4f98a2SYan Zheng key.offset = 0; 554962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 555f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 55649a3c4d9SDavid Sterba root_item); 557f46b5a66SChristoph Hellwig if (ret) 558f46b5a66SChristoph Hellwig goto fail; 559f46b5a66SChristoph Hellwig 56076dda93cSYan, Zheng key.offset = (u64)-1; 56176dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 56279787eaaSJeff Mahoney if (IS_ERR(new_root)) { 56379787eaaSJeff Mahoney ret = PTR_ERR(new_root); 5646d13f549SDavid Sterba btrfs_abort_transaction(trans, root, ret); 56579787eaaSJeff Mahoney goto fail; 56679787eaaSJeff Mahoney } 56776dda93cSYan, Zheng 56876dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 56976dda93cSYan, Zheng 57063541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 571ce598979SMark Fasheh if (ret) { 572ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 57379787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 574ce598979SMark Fasheh goto fail; 575ce598979SMark Fasheh } 576ce598979SMark Fasheh 577f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 578f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 579f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 580f32e48e9SChandan Rajendra 581f46b5a66SChristoph Hellwig /* 582f46b5a66SChristoph Hellwig * insert the directory item 583f46b5a66SChristoph Hellwig */ 5843de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 58579787eaaSJeff Mahoney if (ret) { 58679787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 58779787eaaSJeff Mahoney goto fail; 58879787eaaSJeff Mahoney } 5893de4586cSChris Mason 5903de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 59116cdcec7SMiao Xie name, namelen, dir, &key, 5923de4586cSChris Mason BTRFS_FT_DIR, index); 59379787eaaSJeff Mahoney if (ret) { 59479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 595f46b5a66SChristoph Hellwig goto fail; 59679787eaaSJeff Mahoney } 5970660b5afSChris Mason 59852c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 59952c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 60052c26179SYan Zheng BUG_ON(ret); 60152c26179SYan Zheng 6020660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 6034df27c4dSYan, Zheng objectid, root->root_key.objectid, 60433345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 6050660b5afSChris Mason BUG_ON(ret); 6060660b5afSChris Mason 607dd5f9615SStefan Behrens ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 60849a3c4d9SDavid Sterba root_item->uuid, BTRFS_UUID_KEY_SUBVOL, 609dd5f9615SStefan Behrens objectid); 610dd5f9615SStefan Behrens if (ret) 611dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 612dd5f9615SStefan Behrens 613f46b5a66SChristoph Hellwig fail: 61449a3c4d9SDavid Sterba kfree(root_item); 615d5c12070SMiao Xie trans->block_rsv = NULL; 616d5c12070SMiao Xie trans->bytes_reserved = 0; 617de6e8200SLiu Bo btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 618de6e8200SLiu Bo 61972fd032eSSage Weil if (async_transid) { 62072fd032eSSage Weil *async_transid = trans->transid; 62172fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 62200d71c9cSMiao Xie if (err) 62300d71c9cSMiao Xie err = btrfs_commit_transaction(trans, root); 62472fd032eSSage Weil } else { 62576dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 62672fd032eSSage Weil } 627f46b5a66SChristoph Hellwig if (err && !ret) 628f46b5a66SChristoph Hellwig ret = err; 6291a65e24bSChris Mason 6305662344bSTsutomu Itoh if (!ret) { 6315662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 632de6e8200SLiu Bo if (IS_ERR(inode)) 633de6e8200SLiu Bo return PTR_ERR(inode); 6345662344bSTsutomu Itoh d_instantiate(dentry, inode); 6355662344bSTsutomu Itoh } 636f46b5a66SChristoph Hellwig return ret; 63749a3c4d9SDavid Sterba 63849a3c4d9SDavid Sterba fail_free: 63949a3c4d9SDavid Sterba kfree(root_item); 64049a3c4d9SDavid Sterba return ret; 641f46b5a66SChristoph Hellwig } 642f46b5a66SChristoph Hellwig 6439ea24bbeSFilipe Manana static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) 6448257b2dcSMiao Xie { 6458257b2dcSMiao Xie s64 writers; 6468257b2dcSMiao Xie DEFINE_WAIT(wait); 6478257b2dcSMiao Xie 6488257b2dcSMiao Xie do { 6498257b2dcSMiao Xie prepare_to_wait(&root->subv_writers->wait, &wait, 6508257b2dcSMiao Xie TASK_UNINTERRUPTIBLE); 6518257b2dcSMiao Xie 6528257b2dcSMiao Xie writers = percpu_counter_sum(&root->subv_writers->counter); 6538257b2dcSMiao Xie if (writers) 6548257b2dcSMiao Xie schedule(); 6558257b2dcSMiao Xie 6568257b2dcSMiao Xie finish_wait(&root->subv_writers->wait, &wait); 6578257b2dcSMiao Xie } while (writers); 6588257b2dcSMiao Xie } 6598257b2dcSMiao Xie 660e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 661e9662f70SMiao Xie struct dentry *dentry, char *name, int namelen, 662e9662f70SMiao Xie u64 *async_transid, bool readonly, 663e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 664f46b5a66SChristoph Hellwig { 6652e4bfab9SYan, Zheng struct inode *inode; 666f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 667f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 6682e4bfab9SYan, Zheng int ret; 669f46b5a66SChristoph Hellwig 67027cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 671f46b5a66SChristoph Hellwig return -EINVAL; 672f46b5a66SChristoph Hellwig 673a1ee7362SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 674a1ee7362SDavid Sterba if (!pending_snapshot) 675a1ee7362SDavid Sterba return -ENOMEM; 676a1ee7362SDavid Sterba 677b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 678b0c0ea63SDavid Sterba GFP_NOFS); 6798546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 6808546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 681b0c0ea63SDavid Sterba ret = -ENOMEM; 682b0c0ea63SDavid Sterba goto free_pending; 683b0c0ea63SDavid Sterba } 684b0c0ea63SDavid Sterba 6858257b2dcSMiao Xie atomic_inc(&root->will_be_snapshoted); 6864e857c58SPeter Zijlstra smp_mb__after_atomic(); 6879ea24bbeSFilipe Manana btrfs_wait_for_no_snapshoting_writes(root); 6888257b2dcSMiao Xie 6896a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 6906a03843dSMiao Xie if (ret) 691a1ee7362SDavid Sterba goto dec_and_free; 6926a03843dSMiao Xie 693578def7cSFilipe Manana btrfs_wait_ordered_extents(root, -1, 0, (u64)-1); 6946a03843dSMiao Xie 69566d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 69666d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 697d5c12070SMiao Xie /* 698d5c12070SMiao Xie * 1 - parent dir inode 699d5c12070SMiao Xie * 2 - dir entries 700d5c12070SMiao Xie * 1 - root item 701d5c12070SMiao Xie * 2 - root ref/backref 702d5c12070SMiao Xie * 1 - root of snapshot 703dd5f9615SStefan Behrens * 1 - UUID item 704d5c12070SMiao Xie */ 705d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 706dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 707ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 708ee3441b4SJeff Mahoney false); 709d5c12070SMiao Xie if (ret) 710a1ee7362SDavid Sterba goto dec_and_free; 711d5c12070SMiao Xie 712a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 713a22285a6SYan, Zheng pending_snapshot->root = root; 714b83cc969SLi Zefan pending_snapshot->readonly = readonly; 715e9662f70SMiao Xie pending_snapshot->dir = dir; 7168696c533SMiao Xie pending_snapshot->inherit = inherit; 717a22285a6SYan, Zheng 718d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 719a22285a6SYan, Zheng if (IS_ERR(trans)) { 720a22285a6SYan, Zheng ret = PTR_ERR(trans); 721a22285a6SYan, Zheng goto fail; 722a22285a6SYan, Zheng } 723a22285a6SYan, Zheng 7248351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 725a22285a6SYan, Zheng list_add(&pending_snapshot->list, 726a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 7278351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 72872fd032eSSage Weil if (async_transid) { 72972fd032eSSage Weil *async_transid = trans->transid; 73072fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 73172fd032eSSage Weil root->fs_info->extent_root, 1); 73200d71c9cSMiao Xie if (ret) 73300d71c9cSMiao Xie ret = btrfs_commit_transaction(trans, root); 73472fd032eSSage Weil } else { 73572fd032eSSage Weil ret = btrfs_commit_transaction(trans, 73672fd032eSSage Weil root->fs_info->extent_root); 73772fd032eSSage Weil } 738aec8030aSMiao Xie if (ret) 739c37b2b62SJosef Bacik goto fail; 740a22285a6SYan, Zheng 741a22285a6SYan, Zheng ret = pending_snapshot->error; 742f46b5a66SChristoph Hellwig if (ret) 7432e4bfab9SYan, Zheng goto fail; 744f46b5a66SChristoph Hellwig 745d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 746d3797308SChris Mason if (ret) 747d3797308SChris Mason goto fail; 748d3797308SChris Mason 7492b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7502e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7512e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7522e4bfab9SYan, Zheng goto fail; 7532e4bfab9SYan, Zheng } 7545662344bSTsutomu Itoh 7552e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7562e4bfab9SYan, Zheng ret = 0; 7572e4bfab9SYan, Zheng fail: 758d5c12070SMiao Xie btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, 759d5c12070SMiao Xie &pending_snapshot->block_rsv, 760d5c12070SMiao Xie pending_snapshot->qgroup_reserved); 761a1ee7362SDavid Sterba dec_and_free: 7629ea24bbeSFilipe Manana if (atomic_dec_and_test(&root->will_be_snapshoted)) 7639ea24bbeSFilipe Manana wake_up_atomic_t(&root->will_be_snapshoted); 764b0c0ea63SDavid Sterba free_pending: 765b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 7668546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 767a1ee7362SDavid Sterba kfree(pending_snapshot); 768a1ee7362SDavid Sterba 769f46b5a66SChristoph Hellwig return ret; 770f46b5a66SChristoph Hellwig } 771f46b5a66SChristoph Hellwig 7724260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7734260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7744260f7c7SSage Weil * whether the type of victim is right. 7754260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7764260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7774260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7784260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7794260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7804260f7c7SSage Weil * a. be owner of dir, or 7814260f7c7SSage Weil * b. be owner of victim, or 7824260f7c7SSage Weil * c. have CAP_FOWNER capability 78301327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 7844260f7c7SSage Weil * links pointing to it. 7854260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7864260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7874260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7884260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 7894260f7c7SSage Weil * nfs_async_unlink(). 7904260f7c7SSage Weil */ 7914260f7c7SSage Weil 7924260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 7934260f7c7SSage Weil { 7944260f7c7SSage Weil int error; 7954260f7c7SSage Weil 7962b0143b5SDavid Howells if (d_really_is_negative(victim)) 7974260f7c7SSage Weil return -ENOENT; 7984260f7c7SSage Weil 7992b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 8004fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8014260f7c7SSage Weil 8024260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 8034260f7c7SSage Weil if (error) 8044260f7c7SSage Weil return error; 8054260f7c7SSage Weil if (IS_APPEND(dir)) 8064260f7c7SSage Weil return -EPERM; 8072b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 8082b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 8094260f7c7SSage Weil return -EPERM; 8104260f7c7SSage Weil if (isdir) { 811e36cb0b8SDavid Howells if (!d_is_dir(victim)) 8124260f7c7SSage Weil return -ENOTDIR; 8134260f7c7SSage Weil if (IS_ROOT(victim)) 8144260f7c7SSage Weil return -EBUSY; 815e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 8164260f7c7SSage Weil return -EISDIR; 8174260f7c7SSage Weil if (IS_DEADDIR(dir)) 8184260f7c7SSage Weil return -ENOENT; 8194260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 8204260f7c7SSage Weil return -EBUSY; 8214260f7c7SSage Weil return 0; 8224260f7c7SSage Weil } 8234260f7c7SSage Weil 824cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 825cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 826cb8e7090SChristoph Hellwig { 8272b0143b5SDavid Howells if (d_really_is_positive(child)) 828cb8e7090SChristoph Hellwig return -EEXIST; 829cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 830cb8e7090SChristoph Hellwig return -ENOENT; 831cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 832cb8e7090SChristoph Hellwig } 833cb8e7090SChristoph Hellwig 834cb8e7090SChristoph Hellwig /* 835cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 836cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 837cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 838cb8e7090SChristoph Hellwig */ 83976dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 84076dda93cSYan, Zheng char *name, int namelen, 84172fd032eSSage Weil struct btrfs_root *snap_src, 8426f72c7e2SArne Jansen u64 *async_transid, bool readonly, 8438696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 844cb8e7090SChristoph Hellwig { 8452b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 846cb8e7090SChristoph Hellwig struct dentry *dentry; 847cb8e7090SChristoph Hellwig int error; 848cb8e7090SChristoph Hellwig 84900235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 85000235411SAl Viro if (error == -EINTR) 85100235411SAl Viro return error; 852cb8e7090SChristoph Hellwig 853cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 854cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 855cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 856cb8e7090SChristoph Hellwig goto out_unlock; 857cb8e7090SChristoph Hellwig 85876dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 859cb8e7090SChristoph Hellwig if (error) 860a874a63eSLiu Bo goto out_dput; 861cb8e7090SChristoph Hellwig 8629c52057cSChris Mason /* 8639c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8649c52057cSChris Mason * check for them now when we can safely fail 8659c52057cSChris Mason */ 8669c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8679c52057cSChris Mason dir->i_ino, name, 8689c52057cSChris Mason namelen); 8699c52057cSChris Mason if (error) 8709c52057cSChris Mason goto out_dput; 8719c52057cSChris Mason 87276dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 87376dda93cSYan, Zheng 87476dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 87576dda93cSYan, Zheng goto out_up_read; 87676dda93cSYan, Zheng 8773de4586cSChris Mason if (snap_src) { 878e9662f70SMiao Xie error = create_snapshot(snap_src, dir, dentry, name, namelen, 8796f72c7e2SArne Jansen async_transid, readonly, inherit); 8803de4586cSChris Mason } else { 881d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 882d5c12070SMiao Xie async_transid, inherit); 8833de4586cSChris Mason } 88476dda93cSYan, Zheng if (!error) 88576dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 88676dda93cSYan, Zheng out_up_read: 88776dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 888cb8e7090SChristoph Hellwig out_dput: 889cb8e7090SChristoph Hellwig dput(dentry); 890cb8e7090SChristoph Hellwig out_unlock: 8915955102cSAl Viro inode_unlock(dir); 892cb8e7090SChristoph Hellwig return error; 893cb8e7090SChristoph Hellwig } 894cb8e7090SChristoph Hellwig 8954cb5300bSChris Mason /* 8964cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 8974cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 8984cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 8994cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 9004cb5300bSChris Mason * part of the file 9014cb5300bSChris Mason */ 902aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 9034cb5300bSChris Mason { 9044cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9054cb5300bSChris Mason struct extent_map *em = NULL; 9064cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 9074cb5300bSChris Mason u64 end; 9084cb5300bSChris Mason 9094cb5300bSChris Mason read_lock(&em_tree->lock); 91009cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 9114cb5300bSChris Mason read_unlock(&em_tree->lock); 9124cb5300bSChris Mason 9134cb5300bSChris Mason if (em) { 9144cb5300bSChris Mason end = extent_map_end(em); 9154cb5300bSChris Mason free_extent_map(em); 9164cb5300bSChris Mason if (end - offset > thresh) 9174cb5300bSChris Mason return 0; 9184cb5300bSChris Mason } 9194cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 9204cb5300bSChris Mason thresh /= 2; 9214cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 9224cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 9234cb5300bSChris Mason if (end >= thresh) 9244cb5300bSChris Mason return 0; 9254cb5300bSChris Mason return 1; 9264cb5300bSChris Mason } 9274cb5300bSChris Mason 9284cb5300bSChris Mason /* 9294cb5300bSChris Mason * helper function to walk through a file and find extents 9304cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 9314cb5300bSChris Mason * 9324cb5300bSChris Mason * This is used by the defragging code to find new and small 9334cb5300bSChris Mason * extents 9344cb5300bSChris Mason */ 9354cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 9364cb5300bSChris Mason struct inode *inode, u64 newer_than, 937aab110abSDavid Sterba u64 *off, u32 thresh) 9384cb5300bSChris Mason { 9394cb5300bSChris Mason struct btrfs_path *path; 9404cb5300bSChris Mason struct btrfs_key min_key; 9414cb5300bSChris Mason struct extent_buffer *leaf; 9424cb5300bSChris Mason struct btrfs_file_extent_item *extent; 9434cb5300bSChris Mason int type; 9444cb5300bSChris Mason int ret; 945a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 9464cb5300bSChris Mason 9474cb5300bSChris Mason path = btrfs_alloc_path(); 9484cb5300bSChris Mason if (!path) 9494cb5300bSChris Mason return -ENOMEM; 9504cb5300bSChris Mason 951a4689d2bSDavid Sterba min_key.objectid = ino; 9524cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9534cb5300bSChris Mason min_key.offset = *off; 9544cb5300bSChris Mason 9554cb5300bSChris Mason while (1) { 9566174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9574cb5300bSChris Mason if (ret != 0) 9584cb5300bSChris Mason goto none; 959f094c9bdSFilipe Manana process_slot: 960a4689d2bSDavid Sterba if (min_key.objectid != ino) 9614cb5300bSChris Mason goto none; 9624cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9634cb5300bSChris Mason goto none; 9644cb5300bSChris Mason 9654cb5300bSChris Mason leaf = path->nodes[0]; 9664cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9674cb5300bSChris Mason struct btrfs_file_extent_item); 9684cb5300bSChris Mason 9694cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9704cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9714cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9724cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9734cb5300bSChris Mason *off = min_key.offset; 9744cb5300bSChris Mason btrfs_free_path(path); 9754cb5300bSChris Mason return 0; 9764cb5300bSChris Mason } 9774cb5300bSChris Mason 978f094c9bdSFilipe Manana path->slots[0]++; 979f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 980f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 981f094c9bdSFilipe Manana goto process_slot; 982f094c9bdSFilipe Manana } 983f094c9bdSFilipe Manana 9844cb5300bSChris Mason if (min_key.offset == (u64)-1) 9854cb5300bSChris Mason goto none; 9864cb5300bSChris Mason 9874cb5300bSChris Mason min_key.offset++; 9884cb5300bSChris Mason btrfs_release_path(path); 9894cb5300bSChris Mason } 9904cb5300bSChris Mason none: 9914cb5300bSChris Mason btrfs_free_path(path); 9924cb5300bSChris Mason return -ENOENT; 9934cb5300bSChris Mason } 9944cb5300bSChris Mason 9956c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 99617ce6ef8SLiu Bo { 99717ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 998940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9996c282eb4SLi Zefan struct extent_map *em; 100009cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 1001940100a4SChris Mason 1002940100a4SChris Mason /* 1003940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1004940100a4SChris Mason * the full extent lock 1005940100a4SChris Mason */ 1006940100a4SChris Mason read_lock(&em_tree->lock); 1007940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1008940100a4SChris Mason read_unlock(&em_tree->lock); 1009940100a4SChris Mason 1010940100a4SChris Mason if (!em) { 1011308d9800SFilipe Manana struct extent_state *cached = NULL; 1012308d9800SFilipe Manana u64 end = start + len - 1; 1013308d9800SFilipe Manana 1014940100a4SChris Mason /* get the big lock and read metadata off disk */ 1015ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 1016940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 1017308d9800SFilipe Manana unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 1018940100a4SChris Mason 10196cf8bfbfSDan Carpenter if (IS_ERR(em)) 10206c282eb4SLi Zefan return NULL; 1021940100a4SChris Mason } 1022940100a4SChris Mason 10236c282eb4SLi Zefan return em; 10246c282eb4SLi Zefan } 10256c282eb4SLi Zefan 10266c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 10276c282eb4SLi Zefan { 10286c282eb4SLi Zefan struct extent_map *next; 10296c282eb4SLi Zefan bool ret = true; 10306c282eb4SLi Zefan 10316c282eb4SLi Zefan /* this is the last extent */ 10326c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 10336c282eb4SLi Zefan return false; 10346c282eb4SLi Zefan 10356c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1036e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1037e9512d72SChris Mason ret = false; 1038e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1039ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 10406c282eb4SLi Zefan ret = false; 10416c282eb4SLi Zefan 10426c282eb4SLi Zefan free_extent_map(next); 10436c282eb4SLi Zefan return ret; 10446c282eb4SLi Zefan } 10456c282eb4SLi Zefan 1046aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1047a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1048a43a2111SAndrew Mahone int compress) 10496c282eb4SLi Zefan { 10506c282eb4SLi Zefan struct extent_map *em; 10516c282eb4SLi Zefan int ret = 1; 10526c282eb4SLi Zefan bool next_mergeable = true; 10534a3560c4SLiu Bo bool prev_mergeable = true; 10546c282eb4SLi Zefan 10556c282eb4SLi Zefan /* 10566c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10576c282eb4SLi Zefan * defragging it 10586c282eb4SLi Zefan */ 10596c282eb4SLi Zefan if (start < *defrag_end) 10606c282eb4SLi Zefan return 1; 10616c282eb4SLi Zefan 10626c282eb4SLi Zefan *skip = 0; 10636c282eb4SLi Zefan 10646c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10656c282eb4SLi Zefan if (!em) 10666c282eb4SLi Zefan return 0; 10676c282eb4SLi Zefan 1068940100a4SChris Mason /* this will cover holes, and inline extents */ 106917ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1070940100a4SChris Mason ret = 0; 107117ce6ef8SLiu Bo goto out; 107217ce6ef8SLiu Bo } 107317ce6ef8SLiu Bo 10744a3560c4SLiu Bo if (!*defrag_end) 10754a3560c4SLiu Bo prev_mergeable = false; 10764a3560c4SLiu Bo 10776c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1078940100a4SChris Mason /* 10796c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10806c282eb4SLi Zefan * real extent, don't bother defragging it 1081940100a4SChris Mason */ 1082a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10834a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1084940100a4SChris Mason ret = 0; 108517ce6ef8SLiu Bo out: 1086940100a4SChris Mason /* 1087940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1088940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1089940100a4SChris Mason * so that the next tiny extent will force a defrag. 1090940100a4SChris Mason * 1091940100a4SChris Mason * The end result of this is that tiny extents before a single big 1092940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1093940100a4SChris Mason */ 1094940100a4SChris Mason if (ret) { 1095940100a4SChris Mason *defrag_end = extent_map_end(em); 1096940100a4SChris Mason } else { 1097940100a4SChris Mason *last_len = 0; 1098940100a4SChris Mason *skip = extent_map_end(em); 1099940100a4SChris Mason *defrag_end = 0; 1100940100a4SChris Mason } 1101940100a4SChris Mason 1102940100a4SChris Mason free_extent_map(em); 1103940100a4SChris Mason return ret; 1104940100a4SChris Mason } 1105940100a4SChris Mason 11064cb5300bSChris Mason /* 11074cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 11084cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 11094cb5300bSChris Mason * to COW and defrag. 11104cb5300bSChris Mason * 11114cb5300bSChris Mason * It also makes sure the delalloc code has enough 11124cb5300bSChris Mason * dirty data to avoid making new small extents as part 11134cb5300bSChris Mason * of the defrag 11144cb5300bSChris Mason * 11154cb5300bSChris Mason * It's a good idea to start RA on this range 11164cb5300bSChris Mason * before calling this. 11174cb5300bSChris Mason */ 11184cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 11194cb5300bSChris Mason struct page **pages, 11204cb5300bSChris Mason unsigned long start_index, 1121c41570c9SJustin Maggard unsigned long num_pages) 1122f46b5a66SChristoph Hellwig { 11234cb5300bSChris Mason unsigned long file_end; 11244cb5300bSChris Mason u64 isize = i_size_read(inode); 1125f46b5a66SChristoph Hellwig u64 page_start; 1126f46b5a66SChristoph Hellwig u64 page_end; 11271f12bd06SLiu Bo u64 page_cnt; 11284cb5300bSChris Mason int ret; 11294cb5300bSChris Mason int i; 11304cb5300bSChris Mason int i_done; 11314cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 11324cb5300bSChris Mason struct extent_state *cached_state = NULL; 1133600a45e1SMiao Xie struct extent_io_tree *tree; 11343b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 11354cb5300bSChris Mason 113609cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 11371f12bd06SLiu Bo if (!isize || start_index > file_end) 11381f12bd06SLiu Bo return 0; 11391f12bd06SLiu Bo 11401f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 11414cb5300bSChris Mason 11424cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 114309cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 114409cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 11454cb5300bSChris Mason if (ret) 11464cb5300bSChris Mason return ret; 11474cb5300bSChris Mason i_done = 0; 1148600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11494cb5300bSChris Mason 11504cb5300bSChris Mason /* step one, lock all the pages */ 11511f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11524cb5300bSChris Mason struct page *page; 1153600a45e1SMiao Xie again: 1154a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11553b16a4e3SJosef Bacik start_index + i, mask); 11564cb5300bSChris Mason if (!page) 11574cb5300bSChris Mason break; 11584cb5300bSChris Mason 1159600a45e1SMiao Xie page_start = page_offset(page); 116009cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1161600a45e1SMiao Xie while (1) { 1162308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1163ff13db41SDavid Sterba &cached_state); 1164600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1165600a45e1SMiao Xie page_start); 1166308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1167308d9800SFilipe Manana &cached_state, GFP_NOFS); 1168600a45e1SMiao Xie if (!ordered) 1169600a45e1SMiao Xie break; 1170600a45e1SMiao Xie 1171600a45e1SMiao Xie unlock_page(page); 1172600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1173600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1174600a45e1SMiao Xie lock_page(page); 11751f12bd06SLiu Bo /* 11761f12bd06SLiu Bo * we unlocked the page above, so we need check if 11771f12bd06SLiu Bo * it was released or not. 11781f12bd06SLiu Bo */ 11791f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11801f12bd06SLiu Bo unlock_page(page); 118109cbfeafSKirill A. Shutemov put_page(page); 11821f12bd06SLiu Bo goto again; 11831f12bd06SLiu Bo } 1184600a45e1SMiao Xie } 1185600a45e1SMiao Xie 11864cb5300bSChris Mason if (!PageUptodate(page)) { 11874cb5300bSChris Mason btrfs_readpage(NULL, page); 11884cb5300bSChris Mason lock_page(page); 11894cb5300bSChris Mason if (!PageUptodate(page)) { 11904cb5300bSChris Mason unlock_page(page); 119109cbfeafSKirill A. Shutemov put_page(page); 11924cb5300bSChris Mason ret = -EIO; 11934cb5300bSChris Mason break; 11944cb5300bSChris Mason } 11954cb5300bSChris Mason } 1196600a45e1SMiao Xie 1197600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1198600a45e1SMiao Xie unlock_page(page); 119909cbfeafSKirill A. Shutemov put_page(page); 1200600a45e1SMiao Xie goto again; 1201600a45e1SMiao Xie } 1202600a45e1SMiao Xie 12034cb5300bSChris Mason pages[i] = page; 12044cb5300bSChris Mason i_done++; 12054cb5300bSChris Mason } 12064cb5300bSChris Mason if (!i_done || ret) 12074cb5300bSChris Mason goto out; 12084cb5300bSChris Mason 12094cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 12104cb5300bSChris Mason goto out; 12114cb5300bSChris Mason 12124cb5300bSChris Mason /* 12134cb5300bSChris Mason * so now we have a nice long stream of locked 12144cb5300bSChris Mason * and up to date pages, lets wait on them 12154cb5300bSChris Mason */ 12164cb5300bSChris Mason for (i = 0; i < i_done; i++) 12174cb5300bSChris Mason wait_on_page_writeback(pages[i]); 12184cb5300bSChris Mason 12194cb5300bSChris Mason page_start = page_offset(pages[0]); 122009cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 12214cb5300bSChris Mason 12224cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1223ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 12244cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 12254cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 12269e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 12279e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12284cb5300bSChris Mason 12291f12bd06SLiu Bo if (i_done != page_cnt) { 12309e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 12319e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 12329e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 12334cb5300bSChris Mason btrfs_delalloc_release_space(inode, 123409cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 123509cbfeafSKirill A. Shutemov (page_cnt - i_done) << PAGE_SHIFT); 12364cb5300bSChris Mason } 12374cb5300bSChris Mason 12384cb5300bSChris Mason 12399e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1240018ed4f7SDavid Sterba &cached_state); 12414cb5300bSChris Mason 12424cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 12434cb5300bSChris Mason page_start, page_end - 1, &cached_state, 12444cb5300bSChris Mason GFP_NOFS); 12454cb5300bSChris Mason 12464cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12474cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12484cb5300bSChris Mason ClearPageChecked(pages[i]); 12494cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12504cb5300bSChris Mason set_page_dirty(pages[i]); 12514cb5300bSChris Mason unlock_page(pages[i]); 125209cbfeafSKirill A. Shutemov put_page(pages[i]); 12534cb5300bSChris Mason } 12544cb5300bSChris Mason return i_done; 12554cb5300bSChris Mason out: 12564cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12574cb5300bSChris Mason unlock_page(pages[i]); 125809cbfeafSKirill A. Shutemov put_page(pages[i]); 12594cb5300bSChris Mason } 12607cf5b976SQu Wenruo btrfs_delalloc_release_space(inode, 126109cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 126209cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 12634cb5300bSChris Mason return ret; 12644cb5300bSChris Mason 12654cb5300bSChris Mason } 12664cb5300bSChris Mason 12674cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12684cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12694cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12704cb5300bSChris Mason { 12714cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12724cb5300bSChris Mason struct file_ra_state *ra = NULL; 12734cb5300bSChris Mason unsigned long last_index; 1274151a31b2SLi Zefan u64 isize = i_size_read(inode); 1275940100a4SChris Mason u64 last_len = 0; 1276940100a4SChris Mason u64 skip = 0; 1277940100a4SChris Mason u64 defrag_end = 0; 12784cb5300bSChris Mason u64 newer_off = range->start; 1279f46b5a66SChristoph Hellwig unsigned long i; 1280008873eaSLi Zefan unsigned long ra_index = 0; 1281f46b5a66SChristoph Hellwig int ret; 12824cb5300bSChris Mason int defrag_count = 0; 12831a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1284aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 128509cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1286c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1287ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 12884cb5300bSChris Mason struct page **pages = NULL; 12894cb5300bSChris Mason 12900abd5b17SLiu Bo if (isize == 0) 12910abd5b17SLiu Bo return 0; 12920abd5b17SLiu Bo 12930abd5b17SLiu Bo if (range->start >= isize) 12940abd5b17SLiu Bo return -EINVAL; 12951a419d85SLi Zefan 12961a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 12971a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 12981a419d85SLi Zefan return -EINVAL; 12991a419d85SLi Zefan if (range->compress_type) 13001a419d85SLi Zefan compress_type = range->compress_type; 13011a419d85SLi Zefan } 1302f46b5a66SChristoph Hellwig 13030abd5b17SLiu Bo if (extent_thresh == 0) 1304ee22184bSByongho Lee extent_thresh = SZ_256K; 1305f46b5a66SChristoph Hellwig 13064cb5300bSChris Mason /* 13074cb5300bSChris Mason * if we were not given a file, allocate a readahead 13084cb5300bSChris Mason * context 13094cb5300bSChris Mason */ 13104cb5300bSChris Mason if (!file) { 13114cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 13124cb5300bSChris Mason if (!ra) 13134cb5300bSChris Mason return -ENOMEM; 13144cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 13154cb5300bSChris Mason } else { 13164cb5300bSChris Mason ra = &file->f_ra; 13174cb5300bSChris Mason } 13184cb5300bSChris Mason 1319d9b0d9baSDulshani Gunawardhana pages = kmalloc_array(max_cluster, sizeof(struct page *), 13204cb5300bSChris Mason GFP_NOFS); 13214cb5300bSChris Mason if (!pages) { 13224cb5300bSChris Mason ret = -ENOMEM; 13234cb5300bSChris Mason goto out_ra; 13244cb5300bSChris Mason } 13254cb5300bSChris Mason 13264cb5300bSChris Mason /* find the last page to defrag */ 13271e701a32SChris Mason if (range->start + range->len > range->start) { 1328151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 132909cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 13301e701a32SChris Mason } else { 133109cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 13321e701a32SChris Mason } 13331e701a32SChris Mason 13344cb5300bSChris Mason if (newer_than) { 13354cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1336ee22184bSByongho Lee &newer_off, SZ_64K); 13374cb5300bSChris Mason if (!ret) { 13384cb5300bSChris Mason range->start = newer_off; 13394cb5300bSChris Mason /* 13404cb5300bSChris Mason * we always align our defrag to help keep 13414cb5300bSChris Mason * the extents in the file evenly spaced 13424cb5300bSChris Mason */ 134309cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 13444cb5300bSChris Mason } else 13454cb5300bSChris Mason goto out_ra; 13464cb5300bSChris Mason } else { 134709cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 13484cb5300bSChris Mason } 13494cb5300bSChris Mason if (!max_to_defrag) 1350070034bdSchandan max_to_defrag = last_index - i + 1; 13514cb5300bSChris Mason 13522a0f7f57SLi Zefan /* 13532a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13542a0f7f57SLi Zefan * written sequentially. 13552a0f7f57SLi Zefan */ 13562a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13572a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13582a0f7f57SLi Zefan 1359f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 136009cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 13614cb5300bSChris Mason /* 13624cb5300bSChris Mason * make sure we stop running if someone unmounts 13634cb5300bSChris Mason * the FS 13644cb5300bSChris Mason */ 13654cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 13664cb5300bSChris Mason break; 13674cb5300bSChris Mason 1368210549ebSDavid Sterba if (btrfs_defrag_cancelled(root->fs_info)) { 1369f14d104dSDavid Sterba btrfs_debug(root->fs_info, "defrag_file cancelled"); 1370210549ebSDavid Sterba ret = -EAGAIN; 1371210549ebSDavid Sterba break; 1372210549ebSDavid Sterba } 1373210549ebSDavid Sterba 137409cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 13756c282eb4SLi Zefan extent_thresh, &last_len, &skip, 1376a43a2111SAndrew Mahone &defrag_end, range->flags & 1377a43a2111SAndrew Mahone BTRFS_DEFRAG_RANGE_COMPRESS)) { 1378940100a4SChris Mason unsigned long next; 1379940100a4SChris Mason /* 1380940100a4SChris Mason * the should_defrag function tells us how much to skip 1381940100a4SChris Mason * bump our counter by the suggested amount 1382940100a4SChris Mason */ 138309cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1384940100a4SChris Mason i = max(i + 1, next); 1385940100a4SChris Mason continue; 1386940100a4SChris Mason } 1387008873eaSLi Zefan 1388008873eaSLi Zefan if (!newer_than) { 138909cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 139009cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1391008873eaSLi Zefan cluster = min(cluster, max_cluster); 1392008873eaSLi Zefan } else { 1393008873eaSLi Zefan cluster = max_cluster; 1394008873eaSLi Zefan } 1395008873eaSLi Zefan 1396008873eaSLi Zefan if (i + cluster > ra_index) { 1397008873eaSLi Zefan ra_index = max(i, ra_index); 1398008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1399008873eaSLi Zefan cluster); 1400e4826a5bSchandan ra_index += cluster; 1401008873eaSLi Zefan } 14024cb5300bSChris Mason 14035955102cSAl Viro inode_lock(inode); 1404633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 1405633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = compress_type; 1406008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1407ecb8bea8SLiu Bo if (ret < 0) { 14085955102cSAl Viro inode_unlock(inode); 14094cb5300bSChris Mason goto out_ra; 1410ecb8bea8SLiu Bo } 14114cb5300bSChris Mason 14124cb5300bSChris Mason defrag_count += ret; 1413d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 14145955102cSAl Viro inode_unlock(inode); 14154cb5300bSChris Mason 14164cb5300bSChris Mason if (newer_than) { 14174cb5300bSChris Mason if (newer_off == (u64)-1) 14184cb5300bSChris Mason break; 14194cb5300bSChris Mason 1420e1f041e1SLiu Bo if (ret > 0) 1421e1f041e1SLiu Bo i += ret; 1422e1f041e1SLiu Bo 14234cb5300bSChris Mason newer_off = max(newer_off + 1, 142409cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 14254cb5300bSChris Mason 1426ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1427ee22184bSByongho Lee &newer_off, SZ_64K); 14284cb5300bSChris Mason if (!ret) { 14294cb5300bSChris Mason range->start = newer_off; 143009cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14314cb5300bSChris Mason } else { 14324cb5300bSChris Mason break; 1433940100a4SChris Mason } 14344cb5300bSChris Mason } else { 1435008873eaSLi Zefan if (ret > 0) { 1436cbcc8326SLi Zefan i += ret; 143709cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1438008873eaSLi Zefan } else { 1439940100a4SChris Mason i++; 1440008873eaSLi Zefan last_len = 0; 1441008873eaSLi Zefan } 1442f46b5a66SChristoph Hellwig } 14434cb5300bSChris Mason } 1444f46b5a66SChristoph Hellwig 1445dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14461e701a32SChris Mason filemap_flush(inode->i_mapping); 1447dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1448dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1449dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1450dec8ef90SFilipe Manana } 14511e701a32SChris Mason 14521e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 14531e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 14541e701a32SChris Mason * we have to make sure the IO is actually started and that 14551e701a32SChris Mason * ordered extents get created before we return 14561e701a32SChris Mason */ 14571e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 14581e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 14591e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 14601e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 14611e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 14621e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 14631e701a32SChris Mason } 14641e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 14651e701a32SChris Mason } 14661e701a32SChris Mason 14671a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14682b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); 14691a419d85SLi Zefan } 14701a419d85SLi Zefan 147160ccf82fSDiego Calleja ret = defrag_count; 1472940100a4SChris Mason 14734cb5300bSChris Mason out_ra: 1474633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 14755955102cSAl Viro inode_lock(inode); 1476633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 14775955102cSAl Viro inode_unlock(inode); 1478633085c7SFilipe David Borba Manana } 14794cb5300bSChris Mason if (!file) 14804cb5300bSChris Mason kfree(ra); 14814cb5300bSChris Mason kfree(pages); 1482940100a4SChris Mason return ret; 1483f46b5a66SChristoph Hellwig } 1484f46b5a66SChristoph Hellwig 1485198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 148676dda93cSYan, Zheng void __user *arg) 1487f46b5a66SChristoph Hellwig { 1488f46b5a66SChristoph Hellwig u64 new_size; 1489f46b5a66SChristoph Hellwig u64 old_size; 1490f46b5a66SChristoph Hellwig u64 devid = 1; 1491496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 1492f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1493f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1494f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1495f46b5a66SChristoph Hellwig char *sizestr; 14969a40f122SGui Hecheng char *retptr; 1497f46b5a66SChristoph Hellwig char *devstr = NULL; 1498f46b5a66SChristoph Hellwig int ret = 0; 1499f46b5a66SChristoph Hellwig int mod = 0; 1500f46b5a66SChristoph Hellwig 1501e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1502e441d54dSChris Mason return -EPERM; 1503e441d54dSChris Mason 1504198605a8SMiao Xie ret = mnt_want_write_file(file); 1505198605a8SMiao Xie if (ret) 1506198605a8SMiao Xie return ret; 1507198605a8SMiao Xie 15085ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 15095ac00addSStefan Behrens 1)) { 151097547676SMiao Xie mnt_drop_write_file(file); 1511e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1512c9e9f97bSIlya Dryomov } 1513c9e9f97bSIlya Dryomov 15145ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 1515dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1516c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1517c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1518c9e9f97bSIlya Dryomov goto out; 1519c9e9f97bSIlya Dryomov } 15205516e595SMark Fasheh 15215516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1522f46b5a66SChristoph Hellwig 1523f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1524f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1525f46b5a66SChristoph Hellwig if (devstr) { 1526f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1527f46b5a66SChristoph Hellwig *devstr = '\0'; 1528f46b5a66SChristoph Hellwig devstr = vol_args->name; 152958dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 153058dfae63SZhangZhen if (ret) 153158dfae63SZhangZhen goto out_free; 1532dfd79829SMiao Xie if (!devid) { 1533dfd79829SMiao Xie ret = -EINVAL; 1534dfd79829SMiao Xie goto out_free; 1535dfd79829SMiao Xie } 1536efe120a0SFrank Holton btrfs_info(root->fs_info, "resizing devid %llu", devid); 1537f46b5a66SChristoph Hellwig } 1538dba60f3fSMiao Xie 1539aa1b8cd4SStefan Behrens device = btrfs_find_device(root->fs_info, devid, NULL, NULL); 1540f46b5a66SChristoph Hellwig if (!device) { 1541efe120a0SFrank Holton btrfs_info(root->fs_info, "resizer unable to find device %llu", 1542c1c9ff7cSGeert Uytterhoeven devid); 1543dfd79829SMiao Xie ret = -ENODEV; 1544c9e9f97bSIlya Dryomov goto out_free; 1545f46b5a66SChristoph Hellwig } 1546dba60f3fSMiao Xie 1547dba60f3fSMiao Xie if (!device->writeable) { 1548efe120a0SFrank Holton btrfs_info(root->fs_info, 1549efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1550c1c9ff7cSGeert Uytterhoeven devid); 1551dfd79829SMiao Xie ret = -EPERM; 15524e42ae1bSLiu Bo goto out_free; 15534e42ae1bSLiu Bo } 15544e42ae1bSLiu Bo 1555f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1556f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1557f46b5a66SChristoph Hellwig else { 1558f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1559f46b5a66SChristoph Hellwig mod = -1; 1560f46b5a66SChristoph Hellwig sizestr++; 1561f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1562f46b5a66SChristoph Hellwig mod = 1; 1563f46b5a66SChristoph Hellwig sizestr++; 1564f46b5a66SChristoph Hellwig } 15659a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15669a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1567f46b5a66SChristoph Hellwig ret = -EINVAL; 1568c9e9f97bSIlya Dryomov goto out_free; 1569f46b5a66SChristoph Hellwig } 1570f46b5a66SChristoph Hellwig } 1571f46b5a66SChristoph Hellwig 157263a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1573dfd79829SMiao Xie ret = -EPERM; 157463a212abSStefan Behrens goto out_free; 157563a212abSStefan Behrens } 157663a212abSStefan Behrens 15777cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1578f46b5a66SChristoph Hellwig 1579f46b5a66SChristoph Hellwig if (mod < 0) { 1580f46b5a66SChristoph Hellwig if (new_size > old_size) { 1581f46b5a66SChristoph Hellwig ret = -EINVAL; 1582c9e9f97bSIlya Dryomov goto out_free; 1583f46b5a66SChristoph Hellwig } 1584f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1585f46b5a66SChristoph Hellwig } else if (mod > 0) { 1586eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1587902c68a4SGui Hecheng ret = -ERANGE; 1588eb8052e0SWenliang Fan goto out_free; 1589eb8052e0SWenliang Fan } 1590f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1591f46b5a66SChristoph Hellwig } 1592f46b5a66SChristoph Hellwig 1593ee22184bSByongho Lee if (new_size < SZ_256M) { 1594f46b5a66SChristoph Hellwig ret = -EINVAL; 1595c9e9f97bSIlya Dryomov goto out_free; 1596f46b5a66SChristoph Hellwig } 1597f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1598f46b5a66SChristoph Hellwig ret = -EFBIG; 1599c9e9f97bSIlya Dryomov goto out_free; 1600f46b5a66SChristoph Hellwig } 1601f46b5a66SChristoph Hellwig 1602b8b93addSDavid Sterba new_size = div_u64(new_size, root->sectorsize); 1603f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1604f46b5a66SChristoph Hellwig 1605ecaeb14bSDavid Sterba btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu", 1606c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1607f46b5a66SChristoph Hellwig 1608f46b5a66SChristoph Hellwig if (new_size > old_size) { 1609a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 161098d5dc13STsutomu Itoh if (IS_ERR(trans)) { 161198d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1612c9e9f97bSIlya Dryomov goto out_free; 161398d5dc13STsutomu Itoh } 1614f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1615f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1616ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1617f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 16180253f40eSjeff.liu } /* equal, nothing need to do */ 1619f46b5a66SChristoph Hellwig 1620c9e9f97bSIlya Dryomov out_free: 1621f46b5a66SChristoph Hellwig kfree(vol_args); 1622c9e9f97bSIlya Dryomov out: 1623c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 16245ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 162518f39c41SIlya Dryomov mnt_drop_write_file(file); 1626f46b5a66SChristoph Hellwig return ret; 1627f46b5a66SChristoph Hellwig } 1628f46b5a66SChristoph Hellwig 162972fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 16306f72c7e2SArne Jansen char *name, unsigned long fd, int subvol, 16316f72c7e2SArne Jansen u64 *transid, bool readonly, 16328696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1633f46b5a66SChristoph Hellwig { 1634f46b5a66SChristoph Hellwig int namelen; 16353de4586cSChris Mason int ret = 0; 1636f46b5a66SChristoph Hellwig 1637a874a63eSLiu Bo ret = mnt_want_write_file(file); 1638a874a63eSLiu Bo if (ret) 1639a874a63eSLiu Bo goto out; 1640a874a63eSLiu Bo 164172fd032eSSage Weil namelen = strlen(name); 164272fd032eSSage Weil if (strchr(name, '/')) { 1643f46b5a66SChristoph Hellwig ret = -EINVAL; 1644a874a63eSLiu Bo goto out_drop_write; 1645f46b5a66SChristoph Hellwig } 1646f46b5a66SChristoph Hellwig 164716780cabSChris Mason if (name[0] == '.' && 164816780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 164916780cabSChris Mason ret = -EEXIST; 1650a874a63eSLiu Bo goto out_drop_write; 165116780cabSChris Mason } 165216780cabSChris Mason 16533de4586cSChris Mason if (subvol) { 165472fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16556f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1656cb8e7090SChristoph Hellwig } else { 16572903ff01SAl Viro struct fd src = fdget(fd); 16583de4586cSChris Mason struct inode *src_inode; 16592903ff01SAl Viro if (!src.file) { 16603de4586cSChris Mason ret = -EINVAL; 1661a874a63eSLiu Bo goto out_drop_write; 16623de4586cSChris Mason } 16633de4586cSChris Mason 1664496ad9aaSAl Viro src_inode = file_inode(src.file); 1665496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1666c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1667efe120a0SFrank Holton "Snapshot src from another FS"); 166823ad5b17SKusanagi Kouichi ret = -EXDEV; 1669d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1670d0242061SDavid Sterba /* 1671d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1672d0242061SDavid Sterba * are limited to own subvolumes only 1673d0242061SDavid Sterba */ 1674d0242061SDavid Sterba ret = -EPERM; 1675ecd18815SAl Viro } else { 167672fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 167772fd032eSSage Weil BTRFS_I(src_inode)->root, 16786f72c7e2SArne Jansen transid, readonly, inherit); 1679ecd18815SAl Viro } 16802903ff01SAl Viro fdput(src); 1681cb8e7090SChristoph Hellwig } 1682a874a63eSLiu Bo out_drop_write: 1683a874a63eSLiu Bo mnt_drop_write_file(file); 1684f46b5a66SChristoph Hellwig out: 168572fd032eSSage Weil return ret; 168672fd032eSSage Weil } 168772fd032eSSage Weil 168872fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1689fa0d2b9bSLi Zefan void __user *arg, int subvol) 169072fd032eSSage Weil { 1691fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 169272fd032eSSage Weil int ret; 169372fd032eSSage Weil 1694fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1695fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1696fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1697fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1698fa0d2b9bSLi Zefan 1699fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1700b83cc969SLi Zefan vol_args->fd, subvol, 17016f72c7e2SArne Jansen NULL, false, NULL); 1702fa0d2b9bSLi Zefan 1703fa0d2b9bSLi Zefan kfree(vol_args); 1704fa0d2b9bSLi Zefan return ret; 1705fa0d2b9bSLi Zefan } 1706fa0d2b9bSLi Zefan 1707fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1708fa0d2b9bSLi Zefan void __user *arg, int subvol) 1709fa0d2b9bSLi Zefan { 1710fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1711fa0d2b9bSLi Zefan int ret; 1712fdfb1e4fSLi Zefan u64 transid = 0; 1713fdfb1e4fSLi Zefan u64 *ptr = NULL; 1714b83cc969SLi Zefan bool readonly = false; 17156f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 171672fd032eSSage Weil 1717fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1718fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1719fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1720fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1721fdfb1e4fSLi Zefan 1722b83cc969SLi Zefan if (vol_args->flags & 17236f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 17246f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1725b83cc969SLi Zefan ret = -EOPNOTSUPP; 1726c47ca32dSDan Carpenter goto free_args; 1727fdfb1e4fSLi Zefan } 1728fdfb1e4fSLi Zefan 1729fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1730fdfb1e4fSLi Zefan ptr = &transid; 1731b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1732b83cc969SLi Zefan readonly = true; 17336f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 173409cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 17356f72c7e2SArne Jansen ret = -EINVAL; 1736c47ca32dSDan Carpenter goto free_args; 17376f72c7e2SArne Jansen } 17386f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 17396f72c7e2SArne Jansen if (IS_ERR(inherit)) { 17406f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1741c47ca32dSDan Carpenter goto free_args; 17426f72c7e2SArne Jansen } 17436f72c7e2SArne Jansen } 174475eaa0e2SSage Weil 1745fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 17466f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17478696c533SMiao Xie readonly, inherit); 1748c47ca32dSDan Carpenter if (ret) 1749c47ca32dSDan Carpenter goto free_inherit; 175075eaa0e2SSage Weil 1751c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1752fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1753c47ca32dSDan Carpenter transid), 1754c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 175575eaa0e2SSage Weil ret = -EFAULT; 1756c47ca32dSDan Carpenter 1757c47ca32dSDan Carpenter free_inherit: 17586f72c7e2SArne Jansen kfree(inherit); 1759c47ca32dSDan Carpenter free_args: 1760c47ca32dSDan Carpenter kfree(vol_args); 1761f46b5a66SChristoph Hellwig return ret; 1762f46b5a66SChristoph Hellwig } 1763f46b5a66SChristoph Hellwig 17640caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17650caa102dSLi Zefan void __user *arg) 17660caa102dSLi Zefan { 1767496ad9aaSAl Viro struct inode *inode = file_inode(file); 17680caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17690caa102dSLi Zefan int ret = 0; 17700caa102dSLi Zefan u64 flags = 0; 17710caa102dSLi Zefan 177233345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 17730caa102dSLi Zefan return -EINVAL; 17740caa102dSLi Zefan 17750caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 17760caa102dSLi Zefan if (btrfs_root_readonly(root)) 17770caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17780caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 17790caa102dSLi Zefan 17800caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 17810caa102dSLi Zefan ret = -EFAULT; 17820caa102dSLi Zefan 17830caa102dSLi Zefan return ret; 17840caa102dSLi Zefan } 17850caa102dSLi Zefan 17860caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 17870caa102dSLi Zefan void __user *arg) 17880caa102dSLi Zefan { 1789496ad9aaSAl Viro struct inode *inode = file_inode(file); 17900caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17910caa102dSLi Zefan struct btrfs_trans_handle *trans; 17920caa102dSLi Zefan u64 root_flags; 17930caa102dSLi Zefan u64 flags; 17940caa102dSLi Zefan int ret = 0; 17950caa102dSLi Zefan 1796bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1797bd60ea0fSDavid Sterba return -EPERM; 1798bd60ea0fSDavid Sterba 1799b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1800b9ca0664SLiu Bo if (ret) 1801b9ca0664SLiu Bo goto out; 18020caa102dSLi Zefan 1803b9ca0664SLiu Bo if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 1804b9ca0664SLiu Bo ret = -EINVAL; 1805b9ca0664SLiu Bo goto out_drop_write; 1806b9ca0664SLiu Bo } 18070caa102dSLi Zefan 1808b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1809b9ca0664SLiu Bo ret = -EFAULT; 1810b9ca0664SLiu Bo goto out_drop_write; 1811b9ca0664SLiu Bo } 18120caa102dSLi Zefan 1813b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1814b9ca0664SLiu Bo ret = -EINVAL; 1815b9ca0664SLiu Bo goto out_drop_write; 1816b9ca0664SLiu Bo } 18170caa102dSLi Zefan 1818b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1819b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1820b9ca0664SLiu Bo goto out_drop_write; 1821b9ca0664SLiu Bo } 18220caa102dSLi Zefan 18230caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 18240caa102dSLi Zefan 18250caa102dSLi Zefan /* nothing to do */ 18260caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1827b9ca0664SLiu Bo goto out_drop_sem; 18280caa102dSLi Zefan 18290caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 18302c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 18310caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18320caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 18332c686537SDavid Sterba } else { 18342c686537SDavid Sterba /* 18352c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 18362c686537SDavid Sterba * send 18372c686537SDavid Sterba */ 18382c686537SDavid Sterba spin_lock(&root->root_item_lock); 18392c686537SDavid Sterba if (root->send_in_progress == 0) { 18400caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18410caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 18422c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18432c686537SDavid Sterba } else { 18442c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18452c686537SDavid Sterba btrfs_warn(root->fs_info, 18462c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18472c686537SDavid Sterba root->root_key.objectid); 18482c686537SDavid Sterba ret = -EPERM; 18492c686537SDavid Sterba goto out_drop_sem; 18502c686537SDavid Sterba } 18512c686537SDavid Sterba } 18520caa102dSLi Zefan 18530caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18540caa102dSLi Zefan if (IS_ERR(trans)) { 18550caa102dSLi Zefan ret = PTR_ERR(trans); 18560caa102dSLi Zefan goto out_reset; 18570caa102dSLi Zefan } 18580caa102dSLi Zefan 1859b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 18600caa102dSLi Zefan &root->root_key, &root->root_item); 18610caa102dSLi Zefan 18620caa102dSLi Zefan btrfs_commit_transaction(trans, root); 18630caa102dSLi Zefan out_reset: 18640caa102dSLi Zefan if (ret) 18650caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1866b9ca0664SLiu Bo out_drop_sem: 18670caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 1868b9ca0664SLiu Bo out_drop_write: 1869b9ca0664SLiu Bo mnt_drop_write_file(file); 1870b9ca0664SLiu Bo out: 18710caa102dSLi Zefan return ret; 18720caa102dSLi Zefan } 18730caa102dSLi Zefan 187476dda93cSYan, Zheng /* 187576dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 187676dda93cSYan, Zheng */ 187776dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 187876dda93cSYan, Zheng { 187976dda93cSYan, Zheng struct btrfs_path *path; 1880175a2b87SJosef Bacik struct btrfs_dir_item *di; 188176dda93cSYan, Zheng struct btrfs_key key; 1882175a2b87SJosef Bacik u64 dir_id; 188376dda93cSYan, Zheng int ret; 188476dda93cSYan, Zheng 188576dda93cSYan, Zheng path = btrfs_alloc_path(); 188676dda93cSYan, Zheng if (!path) 188776dda93cSYan, Zheng return -ENOMEM; 188876dda93cSYan, Zheng 1889175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 1890175a2b87SJosef Bacik dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 1891175a2b87SJosef Bacik di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, 1892175a2b87SJosef Bacik dir_id, "default", 7, 0); 1893175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1894175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1895175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 189672de6b53SGuangyu Sun ret = -EPERM; 189772de6b53SGuangyu Sun btrfs_err(root->fs_info, "deleting default subvolume " 189872de6b53SGuangyu Sun "%llu is not allowed", key.objectid); 1899175a2b87SJosef Bacik goto out; 1900175a2b87SJosef Bacik } 1901175a2b87SJosef Bacik btrfs_release_path(path); 1902175a2b87SJosef Bacik } 1903175a2b87SJosef Bacik 190476dda93cSYan, Zheng key.objectid = root->root_key.objectid; 190576dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 190676dda93cSYan, Zheng key.offset = (u64)-1; 190776dda93cSYan, Zheng 190876dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 190976dda93cSYan, Zheng &key, path, 0, 0); 191076dda93cSYan, Zheng if (ret < 0) 191176dda93cSYan, Zheng goto out; 191276dda93cSYan, Zheng BUG_ON(ret == 0); 191376dda93cSYan, Zheng 191476dda93cSYan, Zheng ret = 0; 191576dda93cSYan, Zheng if (path->slots[0] > 0) { 191676dda93cSYan, Zheng path->slots[0]--; 191776dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 191876dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 191976dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 192076dda93cSYan, Zheng ret = -ENOTEMPTY; 192176dda93cSYan, Zheng } 192276dda93cSYan, Zheng out: 192376dda93cSYan, Zheng btrfs_free_path(path); 192476dda93cSYan, Zheng return ret; 192576dda93cSYan, Zheng } 192676dda93cSYan, Zheng 1927ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1928ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1929ac8e9819SChris Mason { 1930abc6e134SChris Mason struct btrfs_key test; 1931abc6e134SChris Mason int ret; 1932abc6e134SChris Mason 1933abc6e134SChris Mason test.objectid = sk->min_objectid; 1934abc6e134SChris Mason test.type = sk->min_type; 1935abc6e134SChris Mason test.offset = sk->min_offset; 1936abc6e134SChris Mason 1937abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1938abc6e134SChris Mason if (ret < 0) 1939ac8e9819SChris Mason return 0; 1940abc6e134SChris Mason 1941abc6e134SChris Mason test.objectid = sk->max_objectid; 1942abc6e134SChris Mason test.type = sk->max_type; 1943abc6e134SChris Mason test.offset = sk->max_offset; 1944abc6e134SChris Mason 1945abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1946abc6e134SChris Mason if (ret > 0) 1947ac8e9819SChris Mason return 0; 1948ac8e9819SChris Mason return 1; 1949ac8e9819SChris Mason } 1950ac8e9819SChris Mason 1951df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 1952ac8e9819SChris Mason struct btrfs_key *key, 1953ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19549b6e817dSGerhard Heift size_t *buf_size, 1955ba346b35SGerhard Heift char __user *ubuf, 1956ac8e9819SChris Mason unsigned long *sk_offset, 1957ac8e9819SChris Mason int *num_found) 1958ac8e9819SChris Mason { 1959ac8e9819SChris Mason u64 found_transid; 1960ac8e9819SChris Mason struct extent_buffer *leaf; 1961ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1962dd81d459SNaohiro Aota struct btrfs_key test; 1963ac8e9819SChris Mason unsigned long item_off; 1964ac8e9819SChris Mason unsigned long item_len; 1965ac8e9819SChris Mason int nritems; 1966ac8e9819SChris Mason int i; 1967ac8e9819SChris Mason int slot; 1968ac8e9819SChris Mason int ret = 0; 1969ac8e9819SChris Mason 1970ac8e9819SChris Mason leaf = path->nodes[0]; 1971ac8e9819SChris Mason slot = path->slots[0]; 1972ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1973ac8e9819SChris Mason 1974ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1975ac8e9819SChris Mason i = nritems; 1976ac8e9819SChris Mason goto advance_key; 1977ac8e9819SChris Mason } 1978ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1979ac8e9819SChris Mason 1980ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1981ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1982ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1983ac8e9819SChris Mason 198403b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 198503b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 198603b71c6cSGabriel de Perthuis continue; 198703b71c6cSGabriel de Perthuis 19889b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 19898f5f6178SGerhard Heift if (*num_found) { 1990ac8e9819SChris Mason ret = 1; 19918f5f6178SGerhard Heift goto out; 19928f5f6178SGerhard Heift } 19938f5f6178SGerhard Heift 19948f5f6178SGerhard Heift /* 19958f5f6178SGerhard Heift * return one empty item back for v1, which does not 19968f5f6178SGerhard Heift * handle -EOVERFLOW 19978f5f6178SGerhard Heift */ 19988f5f6178SGerhard Heift 19999b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2000ac8e9819SChris Mason item_len = 0; 20018f5f6178SGerhard Heift ret = -EOVERFLOW; 20028f5f6178SGerhard Heift } 2003ac8e9819SChris Mason 20049b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2005ac8e9819SChris Mason ret = 1; 200625c9bc2eSGerhard Heift goto out; 2007ac8e9819SChris Mason } 2008ac8e9819SChris Mason 2009ac8e9819SChris Mason sh.objectid = key->objectid; 2010ac8e9819SChris Mason sh.offset = key->offset; 2011ac8e9819SChris Mason sh.type = key->type; 2012ac8e9819SChris Mason sh.len = item_len; 2013ac8e9819SChris Mason sh.transid = found_transid; 2014ac8e9819SChris Mason 2015ac8e9819SChris Mason /* copy search result header */ 2016ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2017ba346b35SGerhard Heift ret = -EFAULT; 2018ba346b35SGerhard Heift goto out; 2019ba346b35SGerhard Heift } 2020ba346b35SGerhard Heift 2021ac8e9819SChris Mason *sk_offset += sizeof(sh); 2022ac8e9819SChris Mason 2023ac8e9819SChris Mason if (item_len) { 2024ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2025ac8e9819SChris Mason /* copy the item */ 2026ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2027ba346b35SGerhard Heift item_off, item_len)) { 2028ba346b35SGerhard Heift ret = -EFAULT; 2029ba346b35SGerhard Heift goto out; 2030ba346b35SGerhard Heift } 2031ba346b35SGerhard Heift 2032ac8e9819SChris Mason *sk_offset += item_len; 2033ac8e9819SChris Mason } 2034e2156867SHugo Mills (*num_found)++; 2035ac8e9819SChris Mason 20368f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20378f5f6178SGerhard Heift goto out; 20388f5f6178SGerhard Heift 203925c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 204025c9bc2eSGerhard Heift ret = 1; 204125c9bc2eSGerhard Heift goto out; 204225c9bc2eSGerhard Heift } 2043ac8e9819SChris Mason } 2044ac8e9819SChris Mason advance_key: 2045ac8e9819SChris Mason ret = 0; 2046dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2047dd81d459SNaohiro Aota test.type = sk->max_type; 2048dd81d459SNaohiro Aota test.offset = sk->max_offset; 2049dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2050dd81d459SNaohiro Aota ret = 1; 2051dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2052abc6e134SChris Mason key->offset++; 2053dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2054abc6e134SChris Mason key->offset = 0; 2055abc6e134SChris Mason key->type++; 2056dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2057abc6e134SChris Mason key->offset = 0; 2058abc6e134SChris Mason key->type = 0; 2059abc6e134SChris Mason key->objectid++; 2060abc6e134SChris Mason } else 2061abc6e134SChris Mason ret = 1; 206225c9bc2eSGerhard Heift out: 2063ba346b35SGerhard Heift /* 2064ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2065ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2066ba346b35SGerhard Heift * * all items were found 2067ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2068ba346b35SGerhard Heift * leaf 2069ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2070ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2071ba346b35SGerhard Heift */ 2072ac8e9819SChris Mason return ret; 2073ac8e9819SChris Mason } 2074ac8e9819SChris Mason 2075ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 207612544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 20779b6e817dSGerhard Heift size_t *buf_size, 2078ba346b35SGerhard Heift char __user *ubuf) 2079ac8e9819SChris Mason { 2080ac8e9819SChris Mason struct btrfs_root *root; 2081ac8e9819SChris Mason struct btrfs_key key; 2082ac8e9819SChris Mason struct btrfs_path *path; 2083ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 2084ac8e9819SChris Mason int ret; 2085ac8e9819SChris Mason int num_found = 0; 2086ac8e9819SChris Mason unsigned long sk_offset = 0; 2087ac8e9819SChris Mason 20889b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 20899b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 209012544442SGerhard Heift return -EOVERFLOW; 20919b6e817dSGerhard Heift } 209212544442SGerhard Heift 2093ac8e9819SChris Mason path = btrfs_alloc_path(); 2094ac8e9819SChris Mason if (!path) 2095ac8e9819SChris Mason return -ENOMEM; 2096ac8e9819SChris Mason 2097ac8e9819SChris Mason if (sk->tree_id == 0) { 2098ac8e9819SChris Mason /* search the root of the inode that was passed */ 2099ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2100ac8e9819SChris Mason } else { 2101ac8e9819SChris Mason key.objectid = sk->tree_id; 2102ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2103ac8e9819SChris Mason key.offset = (u64)-1; 2104ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2105ac8e9819SChris Mason if (IS_ERR(root)) { 2106ac8e9819SChris Mason btrfs_free_path(path); 2107ac8e9819SChris Mason return -ENOENT; 2108ac8e9819SChris Mason } 2109ac8e9819SChris Mason } 2110ac8e9819SChris Mason 2111ac8e9819SChris Mason key.objectid = sk->min_objectid; 2112ac8e9819SChris Mason key.type = sk->min_type; 2113ac8e9819SChris Mason key.offset = sk->min_offset; 2114ac8e9819SChris Mason 2115ac8e9819SChris Mason while (1) { 21166174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2117ac8e9819SChris Mason if (ret != 0) { 2118ac8e9819SChris Mason if (ret > 0) 2119ac8e9819SChris Mason ret = 0; 2120ac8e9819SChris Mason goto err; 2121ac8e9819SChris Mason } 2122df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2123ac8e9819SChris Mason &sk_offset, &num_found); 2124b3b4aa74SDavid Sterba btrfs_release_path(path); 212525c9bc2eSGerhard Heift if (ret) 2126ac8e9819SChris Mason break; 2127ac8e9819SChris Mason 2128ac8e9819SChris Mason } 21298f5f6178SGerhard Heift if (ret > 0) 2130ac8e9819SChris Mason ret = 0; 2131ac8e9819SChris Mason err: 2132ac8e9819SChris Mason sk->nr_items = num_found; 2133ac8e9819SChris Mason btrfs_free_path(path); 2134ac8e9819SChris Mason return ret; 2135ac8e9819SChris Mason } 2136ac8e9819SChris Mason 2137ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2138ac8e9819SChris Mason void __user *argp) 2139ac8e9819SChris Mason { 2140ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2141ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2142ac8e9819SChris Mason struct inode *inode; 2143ac8e9819SChris Mason int ret; 21449b6e817dSGerhard Heift size_t buf_size; 2145ac8e9819SChris Mason 2146ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2147ac8e9819SChris Mason return -EPERM; 2148ac8e9819SChris Mason 2149ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2150ac8e9819SChris Mason 2151ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2152ba346b35SGerhard Heift return -EFAULT; 2153ba346b35SGerhard Heift 2154ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2155ac8e9819SChris Mason 2156496ad9aaSAl Viro inode = file_inode(file); 2157ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21588f5f6178SGerhard Heift 21598f5f6178SGerhard Heift /* 21608f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21618f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21628f5f6178SGerhard Heift */ 21638f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21648f5f6178SGerhard Heift ret = 0; 21658f5f6178SGerhard Heift 2166ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2167ac8e9819SChris Mason ret = -EFAULT; 2168ac8e9819SChris Mason return ret; 2169ac8e9819SChris Mason } 2170ac8e9819SChris Mason 2171cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2172cc68a8a5SGerhard Heift void __user *argp) 2173cc68a8a5SGerhard Heift { 2174cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2175cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2176cc68a8a5SGerhard Heift struct inode *inode; 2177cc68a8a5SGerhard Heift int ret; 2178cc68a8a5SGerhard Heift size_t buf_size; 2179ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2180cc68a8a5SGerhard Heift 2181cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2182cc68a8a5SGerhard Heift return -EPERM; 2183cc68a8a5SGerhard Heift 2184cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2185cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2186cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2187cc68a8a5SGerhard Heift return -EFAULT; 2188cc68a8a5SGerhard Heift 2189cc68a8a5SGerhard Heift buf_size = args.buf_size; 2190cc68a8a5SGerhard Heift 2191cc68a8a5SGerhard Heift if (buf_size < sizeof(struct btrfs_ioctl_search_header)) 2192cc68a8a5SGerhard Heift return -EOVERFLOW; 2193cc68a8a5SGerhard Heift 2194cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2195cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2196cc68a8a5SGerhard Heift buf_size = buf_limit; 2197cc68a8a5SGerhard Heift 2198cc68a8a5SGerhard Heift inode = file_inode(file); 2199cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2200cc68a8a5SGerhard Heift (char *)(&uarg->buf[0])); 2201cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2202cc68a8a5SGerhard Heift ret = -EFAULT; 2203cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2204cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2205cc68a8a5SGerhard Heift ret = -EFAULT; 2206cc68a8a5SGerhard Heift 220776dda93cSYan, Zheng return ret; 220876dda93cSYan, Zheng } 220976dda93cSYan, Zheng 221098d377a0STARUISI Hiroaki /* 2211ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2212ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 221398d377a0STARUISI Hiroaki */ 221498d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 221598d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 221698d377a0STARUISI Hiroaki { 221798d377a0STARUISI Hiroaki struct btrfs_root *root; 221898d377a0STARUISI Hiroaki struct btrfs_key key; 2219ac8e9819SChris Mason char *ptr; 222098d377a0STARUISI Hiroaki int ret = -1; 222198d377a0STARUISI Hiroaki int slot; 222298d377a0STARUISI Hiroaki int len; 222398d377a0STARUISI Hiroaki int total_len = 0; 222498d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 222598d377a0STARUISI Hiroaki struct extent_buffer *l; 222698d377a0STARUISI Hiroaki struct btrfs_path *path; 222798d377a0STARUISI Hiroaki 222898d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 222998d377a0STARUISI Hiroaki name[0]='\0'; 223098d377a0STARUISI Hiroaki return 0; 223198d377a0STARUISI Hiroaki } 223298d377a0STARUISI Hiroaki 223398d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 223498d377a0STARUISI Hiroaki if (!path) 223598d377a0STARUISI Hiroaki return -ENOMEM; 223698d377a0STARUISI Hiroaki 2237ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 223898d377a0STARUISI Hiroaki 223998d377a0STARUISI Hiroaki key.objectid = tree_id; 224098d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 224198d377a0STARUISI Hiroaki key.offset = (u64)-1; 224298d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 224398d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2244f14d104dSDavid Sterba btrfs_err(info, "could not find root %llu", tree_id); 22458ad6fcabSChris Mason ret = -ENOENT; 22468ad6fcabSChris Mason goto out; 224798d377a0STARUISI Hiroaki } 224898d377a0STARUISI Hiroaki 224998d377a0STARUISI Hiroaki key.objectid = dirid; 225098d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22518ad6fcabSChris Mason key.offset = (u64)-1; 225298d377a0STARUISI Hiroaki 225398d377a0STARUISI Hiroaki while (1) { 225498d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 225598d377a0STARUISI Hiroaki if (ret < 0) 225698d377a0STARUISI Hiroaki goto out; 225718674c6cSFilipe David Borba Manana else if (ret > 0) { 225818674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 225918674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 226018674c6cSFilipe David Borba Manana if (ret < 0) 226118674c6cSFilipe David Borba Manana goto out; 226218674c6cSFilipe David Borba Manana else if (ret > 0) { 2263ac8e9819SChris Mason ret = -ENOENT; 226498d377a0STARUISI Hiroaki goto out; 2265ac8e9819SChris Mason } 226618674c6cSFilipe David Borba Manana } 226718674c6cSFilipe David Borba Manana 226818674c6cSFilipe David Borba Manana l = path->nodes[0]; 226918674c6cSFilipe David Borba Manana slot = path->slots[0]; 227018674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 227198d377a0STARUISI Hiroaki 227298d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 227398d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 227498d377a0STARUISI Hiroaki ptr -= len + 1; 227598d377a0STARUISI Hiroaki total_len += len + 1; 2276a696cf35SFilipe David Borba Manana if (ptr < name) { 2277a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 227898d377a0STARUISI Hiroaki goto out; 2279a696cf35SFilipe David Borba Manana } 228098d377a0STARUISI Hiroaki 228198d377a0STARUISI Hiroaki *(ptr + len) = '/'; 228298d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 228398d377a0STARUISI Hiroaki 228498d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 228598d377a0STARUISI Hiroaki break; 228698d377a0STARUISI Hiroaki 2287b3b4aa74SDavid Sterba btrfs_release_path(path); 228898d377a0STARUISI Hiroaki key.objectid = key.offset; 22898ad6fcabSChris Mason key.offset = (u64)-1; 229098d377a0STARUISI Hiroaki dirid = key.objectid; 229198d377a0STARUISI Hiroaki } 229277906a50SLi Zefan memmove(name, ptr, total_len); 229398d377a0STARUISI Hiroaki name[total_len] = '\0'; 229498d377a0STARUISI Hiroaki ret = 0; 229598d377a0STARUISI Hiroaki out: 229698d377a0STARUISI Hiroaki btrfs_free_path(path); 2297ac8e9819SChris Mason return ret; 2298ac8e9819SChris Mason } 2299ac8e9819SChris Mason 2300ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2301ac8e9819SChris Mason void __user *argp) 2302ac8e9819SChris Mason { 2303ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2304ac8e9819SChris Mason struct inode *inode; 230501b810b8SDavid Sterba int ret = 0; 2306ac8e9819SChris Mason 23072354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 23082354d08fSJulia Lawall if (IS_ERR(args)) 23092354d08fSJulia Lawall return PTR_ERR(args); 2310c2b96929SDan Carpenter 2311496ad9aaSAl Viro inode = file_inode(file); 2312ac8e9819SChris Mason 231301b810b8SDavid Sterba /* 231401b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 231501b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 231601b810b8SDavid Sterba */ 23171b53ac4dSChris Mason if (args->treeid == 0) 23181b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 23191b53ac4dSChris Mason 232001b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 232101b810b8SDavid Sterba args->name[0] = 0; 232201b810b8SDavid Sterba goto out; 232301b810b8SDavid Sterba } 232401b810b8SDavid Sterba 232501b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 232601b810b8SDavid Sterba ret = -EPERM; 232701b810b8SDavid Sterba goto out; 232801b810b8SDavid Sterba } 232901b810b8SDavid Sterba 2330ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2331ac8e9819SChris Mason args->treeid, args->objectid, 2332ac8e9819SChris Mason args->name); 2333ac8e9819SChris Mason 233401b810b8SDavid Sterba out: 2335ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2336ac8e9819SChris Mason ret = -EFAULT; 2337ac8e9819SChris Mason 2338ac8e9819SChris Mason kfree(args); 233998d377a0STARUISI Hiroaki return ret; 234098d377a0STARUISI Hiroaki } 234198d377a0STARUISI Hiroaki 234276dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 234376dda93cSYan, Zheng void __user *arg) 234476dda93cSYan, Zheng { 234554563d41SAl Viro struct dentry *parent = file->f_path.dentry; 234676dda93cSYan, Zheng struct dentry *dentry; 23472b0143b5SDavid Howells struct inode *dir = d_inode(parent); 234876dda93cSYan, Zheng struct inode *inode; 234976dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 235076dda93cSYan, Zheng struct btrfs_root *dest = NULL; 235176dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 235276dda93cSYan, Zheng struct btrfs_trans_handle *trans; 2353c58aaad2SMiao Xie struct btrfs_block_rsv block_rsv; 2354521e0546SDavid Sterba u64 root_flags; 2355c58aaad2SMiao Xie u64 qgroup_reserved; 235676dda93cSYan, Zheng int namelen; 235776dda93cSYan, Zheng int ret; 235876dda93cSYan, Zheng int err = 0; 235976dda93cSYan, Zheng 236076dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 236176dda93cSYan, Zheng if (IS_ERR(vol_args)) 236276dda93cSYan, Zheng return PTR_ERR(vol_args); 236376dda93cSYan, Zheng 236476dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 236576dda93cSYan, Zheng namelen = strlen(vol_args->name); 236676dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 236776dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 236876dda93cSYan, Zheng err = -EINVAL; 236976dda93cSYan, Zheng goto out; 237076dda93cSYan, Zheng } 237176dda93cSYan, Zheng 2372a561be71SAl Viro err = mnt_want_write_file(file); 237376dda93cSYan, Zheng if (err) 237476dda93cSYan, Zheng goto out; 237576dda93cSYan, Zheng 2376521e0546SDavid Sterba 237700235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 237800235411SAl Viro if (err == -EINTR) 237900235411SAl Viro goto out_drop_write; 238076dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 238176dda93cSYan, Zheng if (IS_ERR(dentry)) { 238276dda93cSYan, Zheng err = PTR_ERR(dentry); 238376dda93cSYan, Zheng goto out_unlock_dir; 238476dda93cSYan, Zheng } 238576dda93cSYan, Zheng 23862b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 238776dda93cSYan, Zheng err = -ENOENT; 238876dda93cSYan, Zheng goto out_dput; 238976dda93cSYan, Zheng } 239076dda93cSYan, Zheng 23912b0143b5SDavid Howells inode = d_inode(dentry); 23924260f7c7SSage Weil dest = BTRFS_I(inode)->root; 23934260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 23944260f7c7SSage Weil /* 23954260f7c7SSage Weil * Regular user. Only allow this with a special mount 23964260f7c7SSage Weil * option, when the user has write+exec access to the 23974260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 23984260f7c7SSage Weil * allowed. 23994260f7c7SSage Weil * 24004260f7c7SSage Weil * Note that this is _not_ check that the subvol is 24014260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 24024260f7c7SSage Weil * otherwise be able to delete. 24034260f7c7SSage Weil * 24044260f7c7SSage Weil * Users who want to delete empty subvols should try 24054260f7c7SSage Weil * rmdir(2). 24064260f7c7SSage Weil */ 24074260f7c7SSage Weil err = -EPERM; 24083cdde224SJeff Mahoney if (!btrfs_test_opt(root->fs_info, USER_SUBVOL_RM_ALLOWED)) 24094260f7c7SSage Weil goto out_dput; 24104260f7c7SSage Weil 24114260f7c7SSage Weil /* 24124260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 24134260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 24144260f7c7SSage Weil * must be called on the dentry referencing the root 24154260f7c7SSage Weil * of the subvol, not a random directory contained 24164260f7c7SSage Weil * within it. 24174260f7c7SSage Weil */ 24184260f7c7SSage Weil err = -EINVAL; 24194260f7c7SSage Weil if (root == dest) 24204260f7c7SSage Weil goto out_dput; 24214260f7c7SSage Weil 24224260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 24234260f7c7SSage Weil if (err) 24244260f7c7SSage Weil goto out_dput; 24255c39da5bSMiao Xie } 24264260f7c7SSage Weil 24275c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24284260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24294260f7c7SSage Weil if (err) 24304260f7c7SSage Weil goto out_dput; 24314260f7c7SSage Weil 243233345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 243376dda93cSYan, Zheng err = -EINVAL; 243476dda93cSYan, Zheng goto out_dput; 243576dda93cSYan, Zheng } 243676dda93cSYan, Zheng 24375955102cSAl Viro inode_lock(inode); 2438521e0546SDavid Sterba 2439521e0546SDavid Sterba /* 2440521e0546SDavid Sterba * Don't allow to delete a subvolume with send in progress. This is 2441521e0546SDavid Sterba * inside the i_mutex so the error handling that has to drop the bit 2442521e0546SDavid Sterba * again is not run concurrently. 2443521e0546SDavid Sterba */ 2444521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2445c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2446c55bfa67SFilipe Manana if (dest->send_in_progress == 0) { 2447c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2448521e0546SDavid Sterba root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2449521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2450521e0546SDavid Sterba } else { 2451521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2452521e0546SDavid Sterba btrfs_warn(root->fs_info, 2453521e0546SDavid Sterba "Attempt to delete subvolume %llu during send", 2454c55bfa67SFilipe Manana dest->root_key.objectid); 2455521e0546SDavid Sterba err = -EPERM; 2456909e26dcSOmar Sandoval goto out_unlock_inode; 2457521e0546SDavid Sterba } 2458521e0546SDavid Sterba 245976dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 246076dda93cSYan, Zheng 246176dda93cSYan, Zheng err = may_destroy_subvol(dest); 246276dda93cSYan, Zheng if (err) 246376dda93cSYan, Zheng goto out_up_write; 246476dda93cSYan, Zheng 2465c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2466c58aaad2SMiao Xie /* 2467c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2468c58aaad2SMiao Xie * ref/backref. 2469c58aaad2SMiao Xie */ 2470c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2471ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2472c58aaad2SMiao Xie if (err) 2473c58aaad2SMiao Xie goto out_up_write; 2474c58aaad2SMiao Xie 2475a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2476a22285a6SYan, Zheng if (IS_ERR(trans)) { 2477a22285a6SYan, Zheng err = PTR_ERR(trans); 2478c58aaad2SMiao Xie goto out_release; 2479a22285a6SYan, Zheng } 2480c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2481c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2482a22285a6SYan, Zheng 24832be63d5cSFilipe Manana btrfs_record_snapshot_destroy(trans, dir); 24842be63d5cSFilipe Manana 248576dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 248676dda93cSYan, Zheng dest->root_key.objectid, 248776dda93cSYan, Zheng dentry->d_name.name, 248876dda93cSYan, Zheng dentry->d_name.len); 248979787eaaSJeff Mahoney if (ret) { 249079787eaaSJeff Mahoney err = ret; 249179787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 249279787eaaSJeff Mahoney goto out_end_trans; 249379787eaaSJeff Mahoney } 249476dda93cSYan, Zheng 249576dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 249676dda93cSYan, Zheng 249776dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 249876dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 249976dda93cSYan, Zheng dest->root_item.drop_level = 0; 250076dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 250176dda93cSYan, Zheng 250227cdeb70SMiao Xie if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 250376dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 250476dda93cSYan, Zheng root->fs_info->tree_root, 250576dda93cSYan, Zheng dest->root_key.objectid); 250679787eaaSJeff Mahoney if (ret) { 250779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 250879787eaaSJeff Mahoney err = ret; 250979787eaaSJeff Mahoney goto out_end_trans; 2510d68fc57bSYan, Zheng } 251179787eaaSJeff Mahoney } 2512dd5f9615SStefan Behrens 2513dd5f9615SStefan Behrens ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2514dd5f9615SStefan Behrens dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL, 2515dd5f9615SStefan Behrens dest->root_key.objectid); 2516dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 2517dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 2518dd5f9615SStefan Behrens err = ret; 2519dd5f9615SStefan Behrens goto out_end_trans; 2520dd5f9615SStefan Behrens } 2521dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 2522dd5f9615SStefan Behrens ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2523dd5f9615SStefan Behrens dest->root_item.received_uuid, 2524dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2525dd5f9615SStefan Behrens dest->root_key.objectid); 2526dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 2527dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 2528dd5f9615SStefan Behrens err = ret; 2529dd5f9615SStefan Behrens goto out_end_trans; 2530dd5f9615SStefan Behrens } 2531dd5f9615SStefan Behrens } 2532dd5f9615SStefan Behrens 253379787eaaSJeff Mahoney out_end_trans: 2534c58aaad2SMiao Xie trans->block_rsv = NULL; 2535c58aaad2SMiao Xie trans->bytes_reserved = 0; 2536531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 253779787eaaSJeff Mahoney if (ret && !err) 253879787eaaSJeff Mahoney err = ret; 253976dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2540c58aaad2SMiao Xie out_release: 2541c58aaad2SMiao Xie btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 254276dda93cSYan, Zheng out_up_write: 254376dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 2544521e0546SDavid Sterba if (err) { 2545521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2546c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2547c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2548521e0546SDavid Sterba root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2549521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2550521e0546SDavid Sterba } 2551909e26dcSOmar Sandoval out_unlock_inode: 25525955102cSAl Viro inode_unlock(inode); 255376dda93cSYan, Zheng if (!err) { 255464ad6c48SOmar Sandoval d_invalidate(dentry); 255576dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 255676dda93cSYan, Zheng d_delete(dentry); 255761155aa0SDavid Sterba ASSERT(dest->send_in_progress == 0); 2558fa6ac876SLiu Bo 2559fa6ac876SLiu Bo /* the last ref */ 256057cdc8dbSDavid Sterba if (dest->ino_cache_inode) { 256157cdc8dbSDavid Sterba iput(dest->ino_cache_inode); 256257cdc8dbSDavid Sterba dest->ino_cache_inode = NULL; 2563fa6ac876SLiu Bo } 256476dda93cSYan, Zheng } 256576dda93cSYan, Zheng out_dput: 256676dda93cSYan, Zheng dput(dentry); 256776dda93cSYan, Zheng out_unlock_dir: 25685955102cSAl Viro inode_unlock(dir); 256900235411SAl Viro out_drop_write: 25702a79f17eSAl Viro mnt_drop_write_file(file); 257176dda93cSYan, Zheng out: 257276dda93cSYan, Zheng kfree(vol_args); 257376dda93cSYan, Zheng return err; 257476dda93cSYan, Zheng } 257576dda93cSYan, Zheng 25761e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2577f46b5a66SChristoph Hellwig { 2578496ad9aaSAl Viro struct inode *inode = file_inode(file); 2579f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25801e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2581c146afadSYan Zheng int ret; 2582c146afadSYan Zheng 258325122d15SIlya Dryomov ret = mnt_want_write_file(file); 258425122d15SIlya Dryomov if (ret) 258525122d15SIlya Dryomov return ret; 2586b83cc969SLi Zefan 258725122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 258825122d15SIlya Dryomov ret = -EROFS; 258925122d15SIlya Dryomov goto out; 25905ac00addSStefan Behrens } 2591f46b5a66SChristoph Hellwig 2592f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2593f46b5a66SChristoph Hellwig case S_IFDIR: 2594e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2595e441d54dSChris Mason ret = -EPERM; 2596e441d54dSChris Mason goto out; 2597e441d54dSChris Mason } 2598de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 25998929ecfaSYan, Zheng if (ret) 26008929ecfaSYan, Zheng goto out; 2601de78b51aSEric Sandeen ret = btrfs_defrag_root(root->fs_info->extent_root); 2602f46b5a66SChristoph Hellwig break; 2603f46b5a66SChristoph Hellwig case S_IFREG: 2604e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2605e441d54dSChris Mason ret = -EINVAL; 2606e441d54dSChris Mason goto out; 2607e441d54dSChris Mason } 26081e701a32SChris Mason 26091e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 26101e701a32SChris Mason if (!range) { 26111e701a32SChris Mason ret = -ENOMEM; 26121e701a32SChris Mason goto out; 26131e701a32SChris Mason } 26141e701a32SChris Mason 26151e701a32SChris Mason if (argp) { 26161e701a32SChris Mason if (copy_from_user(range, argp, 26171e701a32SChris Mason sizeof(*range))) { 26181e701a32SChris Mason ret = -EFAULT; 26191e701a32SChris Mason kfree(range); 2620683be16eSDan Carpenter goto out; 26211e701a32SChris Mason } 26221e701a32SChris Mason /* compression requires us to start the IO */ 26231e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 26241e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 26251e701a32SChris Mason range->extent_thresh = (u32)-1; 26261e701a32SChris Mason } 26271e701a32SChris Mason } else { 26281e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 26291e701a32SChris Mason range->len = (u64)-1; 26301e701a32SChris Mason } 2631496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 26324cb5300bSChris Mason range, 0, 0); 26334cb5300bSChris Mason if (ret > 0) 26344cb5300bSChris Mason ret = 0; 26351e701a32SChris Mason kfree(range); 2636f46b5a66SChristoph Hellwig break; 26378929ecfaSYan, Zheng default: 26388929ecfaSYan, Zheng ret = -EINVAL; 2639f46b5a66SChristoph Hellwig } 2640e441d54dSChris Mason out: 264125122d15SIlya Dryomov mnt_drop_write_file(file); 2642e441d54dSChris Mason return ret; 2643f46b5a66SChristoph Hellwig } 2644f46b5a66SChristoph Hellwig 2645b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2646f46b5a66SChristoph Hellwig { 2647f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2648f46b5a66SChristoph Hellwig int ret; 2649f46b5a66SChristoph Hellwig 2650e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2651e441d54dSChris Mason return -EPERM; 2652e441d54dSChris Mason 26535ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 26545ac00addSStefan Behrens 1)) { 2655e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2656c9e9f97bSIlya Dryomov } 2657c9e9f97bSIlya Dryomov 26585ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 2659dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2660c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2661c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2662c9e9f97bSIlya Dryomov goto out; 2663c9e9f97bSIlya Dryomov } 2664f46b5a66SChristoph Hellwig 26655516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2666f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2667f46b5a66SChristoph Hellwig 266843d20761SAnand Jain if (!ret) 266943d20761SAnand Jain btrfs_info(root->fs_info, "disk added %s",vol_args->name); 267043d20761SAnand Jain 2671f46b5a66SChristoph Hellwig kfree(vol_args); 2672c9e9f97bSIlya Dryomov out: 2673c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 26745ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2675f46b5a66SChristoph Hellwig return ret; 2676f46b5a66SChristoph Hellwig } 2677f46b5a66SChristoph Hellwig 26786b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2679f46b5a66SChristoph Hellwig { 2680496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 26816b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 2682f46b5a66SChristoph Hellwig int ret; 2683f46b5a66SChristoph Hellwig 2684e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2685e441d54dSChris Mason return -EPERM; 2686e441d54dSChris Mason 2687da24927bSMiao Xie ret = mnt_want_write_file(file); 2688da24927bSMiao Xie if (ret) 2689da24927bSMiao Xie return ret; 2690c146afadSYan Zheng 2691dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2692c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2693c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2694c47ca32dSDan Carpenter goto err_drop; 2695c9e9f97bSIlya Dryomov } 2696f46b5a66SChristoph Hellwig 26976b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 2698735654eaSDavid Sterba if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) 2699735654eaSDavid Sterba return -EOPNOTSUPP; 2700f46b5a66SChristoph Hellwig 2701183860f6SAnand Jain if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2702183860f6SAnand Jain 1)) { 2703183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2704183860f6SAnand Jain goto out; 2705183860f6SAnand Jain } 2706183860f6SAnand Jain 2707183860f6SAnand Jain mutex_lock(&root->fs_info->volume_mutex); 2708735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 27096b526ed7SAnand Jain ret = btrfs_rm_device(root, NULL, vol_args->devid); 27106b526ed7SAnand Jain } else { 27116b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 27126b526ed7SAnand Jain ret = btrfs_rm_device(root, vol_args->name, 0); 27136b526ed7SAnand Jain } 2714c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 27155ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2716183860f6SAnand Jain 27176b526ed7SAnand Jain if (!ret) { 2718735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 27196b526ed7SAnand Jain btrfs_info(root->fs_info, "device deleted: id %llu", 27206b526ed7SAnand Jain vol_args->devid); 27216b526ed7SAnand Jain else 27226b526ed7SAnand Jain btrfs_info(root->fs_info, "device deleted: %s", 27236b526ed7SAnand Jain vol_args->name); 27246b526ed7SAnand Jain } 2725183860f6SAnand Jain out: 2726183860f6SAnand Jain kfree(vol_args); 2727c47ca32dSDan Carpenter err_drop: 27284ac20c70SIlya Dryomov mnt_drop_write_file(file); 2729f46b5a66SChristoph Hellwig return ret; 2730f46b5a66SChristoph Hellwig } 2731f46b5a66SChristoph Hellwig 2732f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2733f46b5a66SChristoph Hellwig { 2734f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 2735f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2736f46b5a66SChristoph Hellwig int ret; 2737f46b5a66SChristoph Hellwig 2738f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2739f46b5a66SChristoph Hellwig return -EPERM; 2740f46b5a66SChristoph Hellwig 2741f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 2742f46b5a66SChristoph Hellwig if (ret) 2743f46b5a66SChristoph Hellwig return ret; 2744f46b5a66SChristoph Hellwig 2745f46b5a66SChristoph Hellwig if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2746f46b5a66SChristoph Hellwig 1)) { 2747f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 274858d7bbf8SDavid Sterba goto out_drop_write; 274958d7bbf8SDavid Sterba } 275058d7bbf8SDavid Sterba 275158d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 275258d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 275358d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 2754f46b5a66SChristoph Hellwig goto out; 2755f46b5a66SChristoph Hellwig } 2756f46b5a66SChristoph Hellwig 275758d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2758f46b5a66SChristoph Hellwig mutex_lock(&root->fs_info->volume_mutex); 27596b526ed7SAnand Jain ret = btrfs_rm_device(root, vol_args->name, 0); 2760f46b5a66SChristoph Hellwig mutex_unlock(&root->fs_info->volume_mutex); 2761f46b5a66SChristoph Hellwig 2762f46b5a66SChristoph Hellwig if (!ret) 2763f46b5a66SChristoph Hellwig btrfs_info(root->fs_info, "disk deleted %s",vol_args->name); 2764f46b5a66SChristoph Hellwig kfree(vol_args); 276558d7bbf8SDavid Sterba out: 276658d7bbf8SDavid Sterba atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 276758d7bbf8SDavid Sterba out_drop_write: 2768f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 276958d7bbf8SDavid Sterba 2770f46b5a66SChristoph Hellwig return ret; 2771f46b5a66SChristoph Hellwig } 2772f46b5a66SChristoph Hellwig 2773475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2774475f6387SJan Schmidt { 2775027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2776475f6387SJan Schmidt struct btrfs_device *device; 2777475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2778027ed2f0SLi Zefan int ret = 0; 2779475f6387SJan Schmidt 2780027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2781027ed2f0SLi Zefan if (!fi_args) 2782027ed2f0SLi Zefan return -ENOMEM; 2783027ed2f0SLi Zefan 2784f7171750SFilipe David Borba Manana mutex_lock(&fs_devices->device_list_mutex); 2785027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2786027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2787475f6387SJan Schmidt 2788d7641a49SByongho Lee list_for_each_entry(device, &fs_devices->devices, dev_list) { 2789027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2790027ed2f0SLi Zefan fi_args->max_id = device->devid; 2791475f6387SJan Schmidt } 2792475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2793475f6387SJan Schmidt 279480a773fbSDavid Sterba fi_args->nodesize = root->fs_info->super_copy->nodesize; 279580a773fbSDavid Sterba fi_args->sectorsize = root->fs_info->super_copy->sectorsize; 279680a773fbSDavid Sterba fi_args->clone_alignment = root->fs_info->super_copy->sectorsize; 279780a773fbSDavid Sterba 2798027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2799027ed2f0SLi Zefan ret = -EFAULT; 2800475f6387SJan Schmidt 2801027ed2f0SLi Zefan kfree(fi_args); 2802027ed2f0SLi Zefan return ret; 2803475f6387SJan Schmidt } 2804475f6387SJan Schmidt 2805475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2806475f6387SJan Schmidt { 2807475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2808475f6387SJan Schmidt struct btrfs_device *dev; 2809475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2810475f6387SJan Schmidt int ret = 0; 2811475f6387SJan Schmidt char *s_uuid = NULL; 2812475f6387SJan Schmidt 2813475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2814475f6387SJan Schmidt if (IS_ERR(di_args)) 2815475f6387SJan Schmidt return PTR_ERR(di_args); 2816475f6387SJan Schmidt 2817dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2818475f6387SJan Schmidt s_uuid = di_args->uuid; 2819475f6387SJan Schmidt 2820475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2821aa1b8cd4SStefan Behrens dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); 2822475f6387SJan Schmidt 2823475f6387SJan Schmidt if (!dev) { 2824475f6387SJan Schmidt ret = -ENODEV; 2825475f6387SJan Schmidt goto out; 2826475f6387SJan Schmidt } 2827475f6387SJan Schmidt 2828475f6387SJan Schmidt di_args->devid = dev->devid; 28297cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 28307cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2831475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2832a27202fbSJim Meyering if (dev->name) { 2833606686eeSJosef Bacik struct rcu_string *name; 2834606686eeSJosef Bacik 2835606686eeSJosef Bacik rcu_read_lock(); 2836606686eeSJosef Bacik name = rcu_dereference(dev->name); 2837606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2838606686eeSJosef Bacik rcu_read_unlock(); 2839a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2840a27202fbSJim Meyering } else { 284199ba55adSStefan Behrens di_args->path[0] = '\0'; 2842a27202fbSJim Meyering } 2843475f6387SJan Schmidt 2844475f6387SJan Schmidt out: 284555793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2846475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2847475f6387SJan Schmidt ret = -EFAULT; 2848475f6387SJan Schmidt 2849475f6387SJan Schmidt kfree(di_args); 2850475f6387SJan Schmidt return ret; 2851475f6387SJan Schmidt } 2852475f6387SJan Schmidt 2853f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2854416161dbSMark Fasheh { 2855416161dbSMark Fasheh struct page *page; 2856416161dbSMark Fasheh 2857416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2858416161dbSMark Fasheh if (!page) 285931314002SFilipe Manana return ERR_PTR(-ENOMEM); 2860416161dbSMark Fasheh 2861416161dbSMark Fasheh if (!PageUptodate(page)) { 286231314002SFilipe Manana int ret; 286331314002SFilipe Manana 286431314002SFilipe Manana ret = btrfs_readpage(NULL, page); 286531314002SFilipe Manana if (ret) 286631314002SFilipe Manana return ERR_PTR(ret); 2867416161dbSMark Fasheh lock_page(page); 2868416161dbSMark Fasheh if (!PageUptodate(page)) { 2869416161dbSMark Fasheh unlock_page(page); 287009cbfeafSKirill A. Shutemov put_page(page); 287131314002SFilipe Manana return ERR_PTR(-EIO); 2872416161dbSMark Fasheh } 287331314002SFilipe Manana if (page->mapping != inode->i_mapping) { 2874416161dbSMark Fasheh unlock_page(page); 287509cbfeafSKirill A. Shutemov put_page(page); 287631314002SFilipe Manana return ERR_PTR(-EAGAIN); 2877416161dbSMark Fasheh } 2878416161dbSMark Fasheh } 2879416161dbSMark Fasheh 2880416161dbSMark Fasheh return page; 2881416161dbSMark Fasheh } 2882416161dbSMark Fasheh 2883f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2884f4414602SMark Fasheh int num_pages, u64 off) 2885f4414602SMark Fasheh { 2886f4414602SMark Fasheh int i; 288709cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 2888f4414602SMark Fasheh 2889f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 289031314002SFilipe Manana again: 2891f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 289231314002SFilipe Manana if (IS_ERR(pages[i])) { 289331314002SFilipe Manana int err = PTR_ERR(pages[i]); 289431314002SFilipe Manana 289531314002SFilipe Manana if (err == -EAGAIN) 289631314002SFilipe Manana goto again; 289731314002SFilipe Manana pages[i] = NULL; 289831314002SFilipe Manana return err; 289931314002SFilipe Manana } 2900f4414602SMark Fasheh } 2901f4414602SMark Fasheh return 0; 2902f4414602SMark Fasheh } 2903f4414602SMark Fasheh 2904e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 2905e0bd70c6SFilipe Manana bool retry_range_locking) 290677fe20dcSMark Fasheh { 2907e0bd70c6SFilipe Manana /* 2908e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 2909e0bd70c6SFilipe Manana * another, and lock file content. 2910e0bd70c6SFilipe Manana * The locking order is: 2911e0bd70c6SFilipe Manana * 2912e0bd70c6SFilipe Manana * 1) pages 2913e0bd70c6SFilipe Manana * 2) range in the inode's io tree 2914e0bd70c6SFilipe Manana */ 291577fe20dcSMark Fasheh while (1) { 291677fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 291777fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 291877fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 291977fe20dcSMark Fasheh off + len - 1); 2920ff5df9b8SFilipe Manana if ((!ordered || 2921ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2922ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 292377fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2924ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2925ff5df9b8SFilipe Manana if (ordered) 2926ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 292777fe20dcSMark Fasheh break; 2928ff5df9b8SFilipe Manana } 292977fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 293077fe20dcSMark Fasheh if (ordered) 293177fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 2932e0bd70c6SFilipe Manana if (!retry_range_locking) 2933e0bd70c6SFilipe Manana return -EAGAIN; 293477fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 293577fe20dcSMark Fasheh } 2936e0bd70c6SFilipe Manana return 0; 293777fe20dcSMark Fasheh } 293877fe20dcSMark Fasheh 2939f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2940416161dbSMark Fasheh { 29415955102cSAl Viro inode_unlock(inode1); 29425955102cSAl Viro inode_unlock(inode2); 2943416161dbSMark Fasheh } 2944416161dbSMark Fasheh 2945f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2946f4414602SMark Fasheh { 2947f4414602SMark Fasheh if (inode1 < inode2) 2948f4414602SMark Fasheh swap(inode1, inode2); 2949f4414602SMark Fasheh 29505955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 29515955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 2952f4414602SMark Fasheh } 2953f4414602SMark Fasheh 2954f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2955f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2956f4414602SMark Fasheh { 2957f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2958f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2959f4414602SMark Fasheh } 2960f4414602SMark Fasheh 2961e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2962e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 2963e0bd70c6SFilipe Manana bool retry_range_locking) 2964416161dbSMark Fasheh { 2965e0bd70c6SFilipe Manana int ret; 2966e0bd70c6SFilipe Manana 2967416161dbSMark Fasheh if (inode1 < inode2) { 2968416161dbSMark Fasheh swap(inode1, inode2); 2969416161dbSMark Fasheh swap(loff1, loff2); 2970416161dbSMark Fasheh } 2971e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 2972e0bd70c6SFilipe Manana if (ret) 2973e0bd70c6SFilipe Manana return ret; 2974e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 2975e0bd70c6SFilipe Manana if (ret) 2976e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 2977e0bd70c6SFilipe Manana loff1 + len - 1); 2978e0bd70c6SFilipe Manana return ret; 2979416161dbSMark Fasheh } 2980f4414602SMark Fasheh 2981f4414602SMark Fasheh struct cmp_pages { 2982f4414602SMark Fasheh int num_pages; 2983f4414602SMark Fasheh struct page **src_pages; 2984f4414602SMark Fasheh struct page **dst_pages; 2985f4414602SMark Fasheh }; 2986f4414602SMark Fasheh 2987f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 2988f4414602SMark Fasheh { 2989f4414602SMark Fasheh int i; 2990f4414602SMark Fasheh struct page *pg; 2991f4414602SMark Fasheh 2992f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 2993f4414602SMark Fasheh pg = cmp->src_pages[i]; 2994e0bd70c6SFilipe Manana if (pg) { 2995e0bd70c6SFilipe Manana unlock_page(pg); 299609cbfeafSKirill A. Shutemov put_page(pg); 2997e0bd70c6SFilipe Manana } 2998f4414602SMark Fasheh pg = cmp->dst_pages[i]; 2999e0bd70c6SFilipe Manana if (pg) { 3000e0bd70c6SFilipe Manana unlock_page(pg); 300109cbfeafSKirill A. Shutemov put_page(pg); 3002f4414602SMark Fasheh } 3003e0bd70c6SFilipe Manana } 3004f4414602SMark Fasheh kfree(cmp->src_pages); 3005f4414602SMark Fasheh kfree(cmp->dst_pages); 3006f4414602SMark Fasheh } 3007f4414602SMark Fasheh 3008f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 3009f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 3010f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 3011f4414602SMark Fasheh { 3012f4414602SMark Fasheh int ret; 301309cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 3014f4414602SMark Fasheh struct page **src_pgarr, **dst_pgarr; 3015f4414602SMark Fasheh 3016f4414602SMark Fasheh /* 3017f4414602SMark Fasheh * We must gather up all the pages before we initiate our 3018f4414602SMark Fasheh * extent locking. We use an array for the page pointers. Size 3019f4414602SMark Fasheh * of the array is bounded by len, which is in turn bounded by 3020f4414602SMark Fasheh * BTRFS_MAX_DEDUPE_LEN. 3021f4414602SMark Fasheh */ 302266722f7cSDavid Sterba src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 302366722f7cSDavid Sterba dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 3024f4414602SMark Fasheh if (!src_pgarr || !dst_pgarr) { 3025f4414602SMark Fasheh kfree(src_pgarr); 3026f4414602SMark Fasheh kfree(dst_pgarr); 3027f4414602SMark Fasheh return -ENOMEM; 3028f4414602SMark Fasheh } 3029f4414602SMark Fasheh cmp->num_pages = num_pages; 3030f4414602SMark Fasheh cmp->src_pages = src_pgarr; 3031f4414602SMark Fasheh cmp->dst_pages = dst_pgarr; 3032f4414602SMark Fasheh 3033f4414602SMark Fasheh ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff); 3034f4414602SMark Fasheh if (ret) 3035f4414602SMark Fasheh goto out; 3036f4414602SMark Fasheh 3037f4414602SMark Fasheh ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff); 3038f4414602SMark Fasheh 3039f4414602SMark Fasheh out: 3040f4414602SMark Fasheh if (ret) 3041f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 3042f4414602SMark Fasheh return 0; 3043416161dbSMark Fasheh } 3044416161dbSMark Fasheh 3045416161dbSMark Fasheh static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, 3046f4414602SMark Fasheh u64 dst_loff, u64 len, struct cmp_pages *cmp) 3047416161dbSMark Fasheh { 3048416161dbSMark Fasheh int ret = 0; 3049f4414602SMark Fasheh int i; 3050416161dbSMark Fasheh struct page *src_page, *dst_page; 305109cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 3052416161dbSMark Fasheh void *addr, *dst_addr; 3053416161dbSMark Fasheh 3054f4414602SMark Fasheh i = 0; 3055416161dbSMark Fasheh while (len) { 305609cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 3057416161dbSMark Fasheh cmp_len = len; 3058416161dbSMark Fasheh 3059f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 3060f4414602SMark Fasheh 3061f4414602SMark Fasheh src_page = cmp->src_pages[i]; 3062f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 3063e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 3064e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 3065f4414602SMark Fasheh 3066416161dbSMark Fasheh addr = kmap_atomic(src_page); 3067416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 3068416161dbSMark Fasheh 3069416161dbSMark Fasheh flush_dcache_page(src_page); 3070416161dbSMark Fasheh flush_dcache_page(dst_page); 3071416161dbSMark Fasheh 3072416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 30732b3909f8SDarrick J. Wong ret = -EBADE; 3074416161dbSMark Fasheh 3075416161dbSMark Fasheh kunmap_atomic(addr); 3076416161dbSMark Fasheh kunmap_atomic(dst_addr); 3077416161dbSMark Fasheh 3078416161dbSMark Fasheh if (ret) 3079416161dbSMark Fasheh break; 3080416161dbSMark Fasheh 3081416161dbSMark Fasheh len -= cmp_len; 3082f4414602SMark Fasheh i++; 3083416161dbSMark Fasheh } 3084416161dbSMark Fasheh 3085416161dbSMark Fasheh return ret; 3086416161dbSMark Fasheh } 3087416161dbSMark Fasheh 3088e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3089e1d227a4SMark Fasheh u64 olen) 3090416161dbSMark Fasheh { 3091e1d227a4SMark Fasheh u64 len = *plen; 3092416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3093416161dbSMark Fasheh 3094e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 3095416161dbSMark Fasheh return -EINVAL; 3096e1d227a4SMark Fasheh 3097e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 3098e1d227a4SMark Fasheh if (off + len == inode->i_size) 3099e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 3100e1d227a4SMark Fasheh 3101416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 3102416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3103416161dbSMark Fasheh return -EINVAL; 3104416161dbSMark Fasheh 3105416161dbSMark Fasheh return 0; 3106416161dbSMark Fasheh } 3107416161dbSMark Fasheh 3108e1d227a4SMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 3109416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 3110416161dbSMark Fasheh { 3111416161dbSMark Fasheh int ret; 3112e1d227a4SMark Fasheh u64 len = olen; 3113f4414602SMark Fasheh struct cmp_pages cmp; 31140efa9f48SMark Fasheh int same_inode = 0; 31150efa9f48SMark Fasheh u64 same_lock_start = 0; 31160efa9f48SMark Fasheh u64 same_lock_len = 0; 3117416161dbSMark Fasheh 3118416161dbSMark Fasheh if (src == dst) 31190efa9f48SMark Fasheh same_inode = 1; 3120416161dbSMark Fasheh 3121113e8283SFilipe Manana if (len == 0) 3122113e8283SFilipe Manana return 0; 3123113e8283SFilipe Manana 31240efa9f48SMark Fasheh if (same_inode) { 31255955102cSAl Viro inode_lock(src); 31260efa9f48SMark Fasheh 31270efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 31280efa9f48SMark Fasheh if (ret) 31290efa9f48SMark Fasheh goto out_unlock; 3130f4dfe687SFilipe Manana ret = extent_same_check_offsets(src, dst_loff, &len, olen); 3131f4dfe687SFilipe Manana if (ret) 3132f4dfe687SFilipe Manana goto out_unlock; 31330efa9f48SMark Fasheh 31340efa9f48SMark Fasheh /* 31350efa9f48SMark Fasheh * Single inode case wants the same checks, except we 31360efa9f48SMark Fasheh * don't want our length pushed out past i_size as 31370efa9f48SMark Fasheh * comparing that data range makes no sense. 31380efa9f48SMark Fasheh * 31390efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 31400efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 31410efa9f48SMark Fasheh * reject the request. 31420efa9f48SMark Fasheh * 31430efa9f48SMark Fasheh * This effectively means we require aligned extents 31440efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 31450efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 31460efa9f48SMark Fasheh * i_size. 31470efa9f48SMark Fasheh */ 31480efa9f48SMark Fasheh if (len != olen) { 31490efa9f48SMark Fasheh ret = -EINVAL; 31500efa9f48SMark Fasheh goto out_unlock; 31510efa9f48SMark Fasheh } 31520efa9f48SMark Fasheh 31530efa9f48SMark Fasheh /* Check for overlapping ranges */ 31540efa9f48SMark Fasheh if (dst_loff + len > loff && dst_loff < loff + len) { 31550efa9f48SMark Fasheh ret = -EINVAL; 31560efa9f48SMark Fasheh goto out_unlock; 31570efa9f48SMark Fasheh } 31580efa9f48SMark Fasheh 31590efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 31600efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 31610efa9f48SMark Fasheh } else { 3162f4414602SMark Fasheh btrfs_double_inode_lock(src, dst); 3163416161dbSMark Fasheh 3164e1d227a4SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 3165416161dbSMark Fasheh if (ret) 3166416161dbSMark Fasheh goto out_unlock; 3167416161dbSMark Fasheh 3168e1d227a4SMark Fasheh ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3169416161dbSMark Fasheh if (ret) 3170416161dbSMark Fasheh goto out_unlock; 31710efa9f48SMark Fasheh } 3172416161dbSMark Fasheh 3173416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 3174416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3175416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 3176416161dbSMark Fasheh ret = -EINVAL; 3177416161dbSMark Fasheh goto out_unlock; 3178416161dbSMark Fasheh } 3179416161dbSMark Fasheh 3180e0bd70c6SFilipe Manana again: 3181f4414602SMark Fasheh ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 3182f4414602SMark Fasheh if (ret) 3183f4414602SMark Fasheh goto out_unlock; 3184f4414602SMark Fasheh 31850efa9f48SMark Fasheh if (same_inode) 3186e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3187e0bd70c6SFilipe Manana false); 31880efa9f48SMark Fasheh else 3189e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3190e0bd70c6SFilipe Manana false); 3191e0bd70c6SFilipe Manana /* 3192e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3193e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3194e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3195e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3196e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3197e0bd70c6SFilipe Manana * range locking). 3198e0bd70c6SFilipe Manana */ 3199e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3200e0bd70c6SFilipe Manana /* 3201e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3202e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3203e0bd70c6SFilipe Manana */ 3204e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3205e0bd70c6SFilipe Manana if (same_inode) { 3206e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3207e0bd70c6SFilipe Manana same_lock_len); 3208e0bd70c6SFilipe Manana } else { 3209e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3210e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3211e0bd70c6SFilipe Manana } 3212e0bd70c6SFilipe Manana goto again; 3213e0bd70c6SFilipe Manana } 3214e0bd70c6SFilipe Manana ASSERT(ret == 0); 3215e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3216e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3217e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3218e0bd70c6SFilipe Manana return ret; 3219e0bd70c6SFilipe Manana } 3220f4414602SMark Fasheh 3221207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 3222f4414602SMark Fasheh ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); 3223416161dbSMark Fasheh if (ret == 0) 32241c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3225416161dbSMark Fasheh 32260efa9f48SMark Fasheh if (same_inode) 32270efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 32280efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 32290efa9f48SMark Fasheh else 3230f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3231f4414602SMark Fasheh 3232f4414602SMark Fasheh btrfs_cmp_data_free(&cmp); 3233416161dbSMark Fasheh out_unlock: 32340efa9f48SMark Fasheh if (same_inode) 32355955102cSAl Viro inode_unlock(src); 32360efa9f48SMark Fasheh else 3237f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3238416161dbSMark Fasheh 3239416161dbSMark Fasheh return ret; 3240416161dbSMark Fasheh } 3241416161dbSMark Fasheh 3242ee22184bSByongho Lee #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3243416161dbSMark Fasheh 32442b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 32452b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3246416161dbSMark Fasheh { 32472b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 32482b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3249416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 32502b3909f8SDarrick J. Wong ssize_t res; 3251416161dbSMark Fasheh 32522b3909f8SDarrick J. Wong if (olen > BTRFS_MAX_DEDUPE_LEN) 32532b3909f8SDarrick J. Wong olen = BTRFS_MAX_DEDUPE_LEN; 3254416161dbSMark Fasheh 325509cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3256416161dbSMark Fasheh /* 3257416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3258416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3259416161dbSMark Fasheh * this situation without an update. 3260416161dbSMark Fasheh */ 32612b3909f8SDarrick J. Wong return -EINVAL; 3262416161dbSMark Fasheh } 3263416161dbSMark Fasheh 32642b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 32652b3909f8SDarrick J. Wong if (res) 32662b3909f8SDarrick J. Wong return res; 32672b3909f8SDarrick J. Wong return olen; 3268416161dbSMark Fasheh } 3269416161dbSMark Fasheh 3270f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3271f82a9901SFilipe Manana struct inode *inode, 3272f82a9901SFilipe Manana u64 endoff, 3273f82a9901SFilipe Manana const u64 destoff, 32741c919a5eSMark Fasheh const u64 olen, 32751c919a5eSMark Fasheh int no_time_update) 3276f82a9901SFilipe Manana { 3277f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3278f82a9901SFilipe Manana int ret; 3279f82a9901SFilipe Manana 3280f82a9901SFilipe Manana inode_inc_iversion(inode); 32811c919a5eSMark Fasheh if (!no_time_update) 328204b285f3SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_fs_time(inode->i_sb); 3283f82a9901SFilipe Manana /* 3284f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3285f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3286f82a9901SFilipe Manana */ 3287f82a9901SFilipe Manana if (endoff > destoff + olen) 3288f82a9901SFilipe Manana endoff = destoff + olen; 3289f82a9901SFilipe Manana if (endoff > inode->i_size) 3290f82a9901SFilipe Manana btrfs_i_size_write(inode, endoff); 3291f82a9901SFilipe Manana 3292f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3293f82a9901SFilipe Manana if (ret) { 3294f82a9901SFilipe Manana btrfs_abort_transaction(trans, root, ret); 3295f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3296f82a9901SFilipe Manana goto out; 3297f82a9901SFilipe Manana } 3298f82a9901SFilipe Manana ret = btrfs_end_transaction(trans, root); 3299f82a9901SFilipe Manana out: 3300f82a9901SFilipe Manana return ret; 3301f82a9901SFilipe Manana } 3302f82a9901SFilipe Manana 33037ffbb598SFilipe Manana static void clone_update_extent_map(struct inode *inode, 33047ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 33057ffbb598SFilipe Manana const struct btrfs_path *path, 33067ffbb598SFilipe Manana const u64 hole_offset, 33077ffbb598SFilipe Manana const u64 hole_len) 33087ffbb598SFilipe Manana { 33097ffbb598SFilipe Manana struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 33107ffbb598SFilipe Manana struct extent_map *em; 33117ffbb598SFilipe Manana int ret; 33127ffbb598SFilipe Manana 33137ffbb598SFilipe Manana em = alloc_extent_map(); 33147ffbb598SFilipe Manana if (!em) { 33157ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 33167ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 33177ffbb598SFilipe Manana return; 33187ffbb598SFilipe Manana } 33197ffbb598SFilipe Manana 332014f59796SFilipe Manana if (path) { 332114f59796SFilipe Manana struct btrfs_file_extent_item *fi; 332214f59796SFilipe Manana 332314f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 332414f59796SFilipe Manana struct btrfs_file_extent_item); 33257ffbb598SFilipe Manana btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 33267ffbb598SFilipe Manana em->generation = -1; 33277ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 33287ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33297ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 33307ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 33317ffbb598SFilipe Manana } else { 33327ffbb598SFilipe Manana em->start = hole_offset; 33337ffbb598SFilipe Manana em->len = hole_len; 33347ffbb598SFilipe Manana em->ram_bytes = em->len; 33357ffbb598SFilipe Manana em->orig_start = hole_offset; 33367ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33377ffbb598SFilipe Manana em->block_len = 0; 33387ffbb598SFilipe Manana em->orig_block_len = 0; 33397ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33407ffbb598SFilipe Manana em->generation = trans->transid; 33417ffbb598SFilipe Manana } 33427ffbb598SFilipe Manana 33437ffbb598SFilipe Manana while (1) { 33447ffbb598SFilipe Manana write_lock(&em_tree->lock); 33457ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33467ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33477ffbb598SFilipe Manana if (ret != -EEXIST) { 33487ffbb598SFilipe Manana free_extent_map(em); 33497ffbb598SFilipe Manana break; 33507ffbb598SFilipe Manana } 33517ffbb598SFilipe Manana btrfs_drop_extent_cache(inode, em->start, 33527ffbb598SFilipe Manana em->start + em->len - 1, 0); 33537ffbb598SFilipe Manana } 33547ffbb598SFilipe Manana 3355ee39b432SDavid Sterba if (ret) 33567ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 33577ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 33587ffbb598SFilipe Manana } 33597ffbb598SFilipe Manana 33608039d87dSFilipe Manana /* 33618039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33628039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33638039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33648039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33658039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33668039d87dSFilipe Manana * 33678039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33688039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33698039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33708039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33718039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33728039d87dSFilipe Manana * i_mutex, which we are holding here). 33738039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33748039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33758039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33768039d87dSFilipe Manana * 33778039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33788039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33798039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33808039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33818039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33828039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33838039d87dSFilipe Manana * the former. 33848039d87dSFilipe Manana */ 33858039d87dSFilipe Manana static int clone_copy_inline_extent(struct inode *src, 33868039d87dSFilipe Manana struct inode *dst, 33878039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33888039d87dSFilipe Manana struct btrfs_path *path, 33898039d87dSFilipe Manana struct btrfs_key *new_key, 33908039d87dSFilipe Manana const u64 drop_start, 33918039d87dSFilipe Manana const u64 datal, 33928039d87dSFilipe Manana const u64 skip, 33938039d87dSFilipe Manana const u64 size, 33948039d87dSFilipe Manana char *inline_data) 33958039d87dSFilipe Manana { 33968039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 33978039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 33988039d87dSFilipe Manana root->sectorsize); 33998039d87dSFilipe Manana int ret; 34008039d87dSFilipe Manana struct btrfs_key key; 34018039d87dSFilipe Manana 34028039d87dSFilipe Manana if (new_key->offset > 0) 34038039d87dSFilipe Manana return -EOPNOTSUPP; 34048039d87dSFilipe Manana 34058039d87dSFilipe Manana key.objectid = btrfs_ino(dst); 34068039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 34078039d87dSFilipe Manana key.offset = 0; 34088039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 34098039d87dSFilipe Manana if (ret < 0) { 34108039d87dSFilipe Manana return ret; 34118039d87dSFilipe Manana } else if (ret > 0) { 34128039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 34138039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 34148039d87dSFilipe Manana if (ret < 0) 34158039d87dSFilipe Manana return ret; 34168039d87dSFilipe Manana else if (ret > 0) 34178039d87dSFilipe Manana goto copy_inline_extent; 34188039d87dSFilipe Manana } 34198039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 34208039d87dSFilipe Manana if (key.objectid == btrfs_ino(dst) && 34218039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 34228039d87dSFilipe Manana ASSERT(key.offset > 0); 34238039d87dSFilipe Manana return -EOPNOTSUPP; 34248039d87dSFilipe Manana } 34258039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34268039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34278039d87dSFilipe Manana u64 ext_len; 34288039d87dSFilipe Manana 34298039d87dSFilipe Manana /* 34308039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34318039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34328039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34338039d87dSFilipe Manana */ 34348039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34358039d87dSFilipe Manana struct btrfs_file_extent_item); 34368039d87dSFilipe Manana /* 34378039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34388039d87dSFilipe Manana * following it. 34398039d87dSFilipe Manana */ 34408039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34418039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34428039d87dSFilipe Manana goto copy_inline_extent; 34438039d87dSFilipe Manana 34448039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34458039d87dSFilipe Manana if (ext_len > aligned_end) 34468039d87dSFilipe Manana return -EOPNOTSUPP; 34478039d87dSFilipe Manana 34488039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34498039d87dSFilipe Manana if (ret < 0) { 34508039d87dSFilipe Manana return ret; 34518039d87dSFilipe Manana } else if (ret == 0) { 34528039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34538039d87dSFilipe Manana path->slots[0]); 34548039d87dSFilipe Manana if (key.objectid == btrfs_ino(dst) && 34558039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34568039d87dSFilipe Manana return -EOPNOTSUPP; 34578039d87dSFilipe Manana } 34588039d87dSFilipe Manana } 34598039d87dSFilipe Manana 34608039d87dSFilipe Manana copy_inline_extent: 34618039d87dSFilipe Manana /* 34628039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34638039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34648039d87dSFilipe Manana */ 34658039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34668039d87dSFilipe Manana /* 34678039d87dSFilipe Manana * If the destination inode has an inline extent... 34688039d87dSFilipe Manana * This would require copying the data from the source inline 34698039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34708039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34718039d87dSFilipe Manana * or just one of them, which would require decompressing and 34728039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34738039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34748039d87dSFilipe Manana * inline extent). 34758039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34768039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34778039d87dSFilipe Manana */ 34788039d87dSFilipe Manana return -EOPNOTSUPP; 34798039d87dSFilipe Manana } 34808039d87dSFilipe Manana 34818039d87dSFilipe Manana btrfs_release_path(path); 34828039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34838039d87dSFilipe Manana if (ret) 34848039d87dSFilipe Manana return ret; 34858039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34868039d87dSFilipe Manana if (ret) 34878039d87dSFilipe Manana return ret; 34888039d87dSFilipe Manana 34898039d87dSFilipe Manana if (skip) { 34908039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 34918039d87dSFilipe Manana 34928039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 34938039d87dSFilipe Manana } 34948039d87dSFilipe Manana 34958039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 34968039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 34978039d87dSFilipe Manana path->slots[0]), 34988039d87dSFilipe Manana size); 34998039d87dSFilipe Manana inode_add_bytes(dst, datal); 35008039d87dSFilipe Manana 35018039d87dSFilipe Manana return 0; 35028039d87dSFilipe Manana } 35038039d87dSFilipe Manana 350432b7c687SMark Fasheh /** 350532b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 350632b7c687SMark Fasheh * 350732b7c687SMark Fasheh * @src: Inode to clone from 350832b7c687SMark Fasheh * @inode: Inode to clone to 350932b7c687SMark Fasheh * @off: Offset within source to start clone from 351032b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 35111c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 351232b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 35131c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 351432b7c687SMark Fasheh */ 351532b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3516f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 35171c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3518f46b5a66SChristoph Hellwig { 3519f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 352032b7c687SMark Fasheh struct btrfs_path *path = NULL; 3521f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 352232b7c687SMark Fasheh struct btrfs_trans_handle *trans; 352332b7c687SMark Fasheh char *buf = NULL; 3524ae01a0abSYan Zheng struct btrfs_key key; 3525f46b5a66SChristoph Hellwig u32 nritems; 3526f46b5a66SChristoph Hellwig int slot; 3527ae01a0abSYan Zheng int ret; 3528f82a9901SFilipe Manana const u64 len = olen_aligned; 3529f82a9901SFilipe Manana u64 last_dest_end = destoff; 3530ae01a0abSYan Zheng 3531ae01a0abSYan Zheng ret = -ENOMEM; 353215351955SDavid Sterba buf = kmalloc(root->nodesize, GFP_KERNEL | __GFP_NOWARN); 353315351955SDavid Sterba if (!buf) { 3534707e8a07SDavid Sterba buf = vmalloc(root->nodesize); 3535ae01a0abSYan Zheng if (!buf) 353632b7c687SMark Fasheh return ret; 353715351955SDavid Sterba } 3538ae01a0abSYan Zheng 3539ae01a0abSYan Zheng path = btrfs_alloc_path(); 3540ae01a0abSYan Zheng if (!path) { 354115351955SDavid Sterba kvfree(buf); 354232b7c687SMark Fasheh return ret; 3543ae01a0abSYan Zheng } 354432b7c687SMark Fasheh 3545e4058b54SDavid Sterba path->reada = READA_FORWARD; 3546c5c9cd4dSSage Weil /* clone data */ 354733345d01SLi Zefan key.objectid = btrfs_ino(src); 3548ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35492c463823SFilipe Manana key.offset = off; 3550f46b5a66SChristoph Hellwig 3551f46b5a66SChristoph Hellwig while (1) { 3552de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3553df858e76SFilipe Manana 3554f46b5a66SChristoph Hellwig /* 3555f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3556f46b5a66SChristoph Hellwig * tree. 3557f46b5a66SChristoph Hellwig */ 3558e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3559362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3560362a20c5SDavid Sterba 0, 0); 3561f46b5a66SChristoph Hellwig if (ret < 0) 3562f46b5a66SChristoph Hellwig goto out; 35632c463823SFilipe Manana /* 35642c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35652c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35662c463823SFilipe Manana * it might overlap our target range, therefore process it. 35672c463823SFilipe Manana */ 35682c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35692c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35702c463823SFilipe Manana path->slots[0] - 1); 35712c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35722c463823SFilipe Manana path->slots[0]--; 35732c463823SFilipe Manana } 3574f46b5a66SChristoph Hellwig 3575ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3576e4355f34SFilipe David Borba Manana process_slot: 3577ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3578362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3579f46b5a66SChristoph Hellwig if (ret < 0) 3580f46b5a66SChristoph Hellwig goto out; 3581f46b5a66SChristoph Hellwig if (ret > 0) 3582f46b5a66SChristoph Hellwig break; 3583ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3584f46b5a66SChristoph Hellwig } 3585f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3586f46b5a66SChristoph Hellwig slot = path->slots[0]; 3587f46b5a66SChristoph Hellwig 3588ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3589962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 359033345d01SLi Zefan key.objectid != btrfs_ino(src)) 3591f46b5a66SChristoph Hellwig break; 3592f46b5a66SChristoph Hellwig 3593962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3594c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3595c5c9cd4dSSage Weil int type; 359631840ae1SZheng Yan u32 size; 359731840ae1SZheng Yan struct btrfs_key new_key; 3598c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3599c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3600c5c9cd4dSSage Weil u8 comp; 3601f82a9901SFilipe Manana u64 drop_start; 360231840ae1SZheng Yan 3603c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3604c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3605c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3606c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3607c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3608c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3609d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3610d397712bSChris Mason extent); 3611d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3612d397712bSChris Mason extent); 3613c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3614d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3615d397712bSChris Mason extent); 3616c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3617c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3618c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3619c5c9cd4dSSage Weil extent); 3620c5c9cd4dSSage Weil } 362131840ae1SZheng Yan 36222c463823SFilipe Manana /* 36232c463823SFilipe Manana * The first search might have left us at an extent 36242c463823SFilipe Manana * item that ends before our target range's start, can 36252c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 36262c463823SFilipe Manana */ 36272c463823SFilipe Manana if (key.offset + datal <= off) { 3628e4355f34SFilipe David Borba Manana path->slots[0]++; 3629e4355f34SFilipe David Borba Manana goto process_slot; 36302c463823SFilipe Manana } else if (key.offset >= off + len) { 36312c463823SFilipe Manana break; 3632e4355f34SFilipe David Borba Manana } 3633df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3634e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3635e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3636e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3637e4355f34SFilipe David Borba Manana size); 3638e4355f34SFilipe David Borba Manana 3639e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3640e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3641c5c9cd4dSSage Weil 364231840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 364333345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 36444d728ec7SLi Zefan if (off <= key.offset) 3645c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36464d728ec7SLi Zefan else 36474d728ec7SLi Zefan new_key.offset = destoff; 3648c5c9cd4dSSage Weil 3649b6f3409bSSage Weil /* 3650f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3651f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3652f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3653f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3654f82a9901SFilipe Manana * partially overlaps it). 3655f82a9901SFilipe Manana */ 3656f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3657f82a9901SFilipe Manana drop_start = last_dest_end; 3658f82a9901SFilipe Manana else 3659f82a9901SFilipe Manana drop_start = new_key.offset; 3660f82a9901SFilipe Manana 3661f82a9901SFilipe Manana /* 3662b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3663b6f3409bSSage Weil * 1 - add new extent 3664b6f3409bSSage Weil * 1 - inode update 3665b6f3409bSSage Weil */ 3666b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3667a22285a6SYan, Zheng if (IS_ERR(trans)) { 3668a22285a6SYan, Zheng ret = PTR_ERR(trans); 3669a22285a6SYan, Zheng goto out; 3670a22285a6SYan, Zheng } 3671a22285a6SYan, Zheng 3672c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3673c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3674d72c0842SLi Zefan /* 3675d72c0842SLi Zefan * a | --- range to clone ---| b 3676d72c0842SLi Zefan * | ------------- extent ------------- | 3677d72c0842SLi Zefan */ 3678d72c0842SLi Zefan 367993915584SAntonio Ospite /* subtract range b */ 3680d72c0842SLi Zefan if (key.offset + datal > off + len) 3681d72c0842SLi Zefan datal = off + len - key.offset; 3682d72c0842SLi Zefan 368393915584SAntonio Ospite /* subtract range a */ 3684a22285a6SYan, Zheng if (off > key.offset) { 3685a22285a6SYan, Zheng datao += off - key.offset; 3686a22285a6SYan, Zheng datal -= off - key.offset; 3687a22285a6SYan, Zheng } 3688a22285a6SYan, Zheng 36895dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3690f82a9901SFilipe Manana drop_start, 3691a22285a6SYan, Zheng new_key.offset + datal, 36922671485dSJosef Bacik 1); 369379787eaaSJeff Mahoney if (ret) { 36943f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 369500fdf13aSLiu Bo btrfs_abort_transaction(trans, 369600fdf13aSLiu Bo root, ret); 369779787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 369879787eaaSJeff Mahoney goto out; 369979787eaaSJeff Mahoney } 3700a22285a6SYan, Zheng 370131840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 370231840ae1SZheng Yan &new_key, size); 370379787eaaSJeff Mahoney if (ret) { 370479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 370579787eaaSJeff Mahoney ret); 370679787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 370779787eaaSJeff Mahoney goto out; 370879787eaaSJeff Mahoney } 370931840ae1SZheng Yan 371031840ae1SZheng Yan leaf = path->nodes[0]; 371131840ae1SZheng Yan slot = path->slots[0]; 371231840ae1SZheng Yan write_extent_buffer(leaf, buf, 371331840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 371431840ae1SZheng Yan size); 3715ae01a0abSYan Zheng 3716f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3717f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3718c5c9cd4dSSage Weil 3719c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3720c5c9cd4dSSage Weil if (!disko) 3721c5c9cd4dSSage Weil datao = 0; 3722c5c9cd4dSSage Weil 3723c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3724c5c9cd4dSSage Weil datao); 3725c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3726c5c9cd4dSSage Weil datal); 3727fcebe456SJosef Bacik 3728c5c9cd4dSSage Weil if (disko) { 3729c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 3730ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 37315d4f98a2SYan Zheng disko, diskl, 0, 3732f46b5a66SChristoph Hellwig root->root_key.objectid, 373333345d01SLi Zefan btrfs_ino(inode), 3734b06c4bf5SFilipe Manana new_key.offset - datao); 373579787eaaSJeff Mahoney if (ret) { 373679787eaaSJeff Mahoney btrfs_abort_transaction(trans, 373779787eaaSJeff Mahoney root, 373879787eaaSJeff Mahoney ret); 373979787eaaSJeff Mahoney btrfs_end_transaction(trans, 374079787eaaSJeff Mahoney root); 374179787eaaSJeff Mahoney goto out; 374279787eaaSJeff Mahoney 374379787eaaSJeff Mahoney } 3744f46b5a66SChristoph Hellwig } 3745c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3746c5c9cd4dSSage Weil u64 skip = 0; 3747c5c9cd4dSSage Weil u64 trim = 0; 3748ed958762SFilipe Manana 3749c5c9cd4dSSage Weil if (off > key.offset) { 3750c5c9cd4dSSage Weil skip = off - key.offset; 3751c5c9cd4dSSage Weil new_key.offset += skip; 375231840ae1SZheng Yan } 3753d397712bSChris Mason 3754c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3755c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3756d397712bSChris Mason 3757c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3758c5c9cd4dSSage Weil ret = -EINVAL; 3759a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 3760c5c9cd4dSSage Weil goto out; 376131840ae1SZheng Yan } 3762c5c9cd4dSSage Weil size -= skip + trim; 3763c5c9cd4dSSage Weil datal -= skip + trim; 3764a22285a6SYan, Zheng 37658039d87dSFilipe Manana ret = clone_copy_inline_extent(src, inode, 37668039d87dSFilipe Manana trans, path, 37678039d87dSFilipe Manana &new_key, 3768f82a9901SFilipe Manana drop_start, 37698039d87dSFilipe Manana datal, 37708039d87dSFilipe Manana skip, size, buf); 377179787eaaSJeff Mahoney if (ret) { 37723f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37733a29bc09SChris Mason btrfs_abort_transaction(trans, 37748039d87dSFilipe Manana root, 377579787eaaSJeff Mahoney ret); 377679787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 377779787eaaSJeff Mahoney goto out; 377879787eaaSJeff Mahoney } 3779c5c9cd4dSSage Weil leaf = path->nodes[0]; 3780c5c9cd4dSSage Weil slot = path->slots[0]; 3781c5c9cd4dSSage Weil } 3782c5c9cd4dSSage Weil 37837ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 37847ffbb598SFilipe Manana if (drop_start < new_key.offset) 37857ffbb598SFilipe Manana clone_update_extent_map(inode, trans, 378614f59796SFilipe Manana NULL, drop_start, 37877ffbb598SFilipe Manana new_key.offset - drop_start); 37887ffbb598SFilipe Manana 378914f59796SFilipe Manana clone_update_extent_map(inode, trans, path, 0, 0); 37907ffbb598SFilipe Manana 3791c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3792b3b4aa74SDavid Sterba btrfs_release_path(path); 3793c5c9cd4dSSage Weil 379462e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 379562e2390eSFilipe Manana root->sectorsize); 3796f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3797f82a9901SFilipe Manana last_dest_end, 37981c919a5eSMark Fasheh destoff, olen, 37991c919a5eSMark Fasheh no_time_update); 3800f82a9901SFilipe Manana if (ret) 380179787eaaSJeff Mahoney goto out; 38022c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 38032c463823SFilipe Manana break; 3804a22285a6SYan, Zheng } 3805b3b4aa74SDavid Sterba btrfs_release_path(path); 3806df858e76SFilipe Manana key.offset = next_key_min_offset; 3807ae01a0abSYan Zheng } 3808f46b5a66SChristoph Hellwig ret = 0; 380932b7c687SMark Fasheh 3810f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3811f82a9901SFilipe Manana /* 3812f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3813f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3814f82a9901SFilipe Manana */ 3815b3b4aa74SDavid Sterba btrfs_release_path(path); 3816f82a9901SFilipe Manana 3817f82a9901SFilipe Manana /* 3818f82a9901SFilipe Manana * 1 - remove extent(s) 3819f82a9901SFilipe Manana * 1 - inode update 3820f82a9901SFilipe Manana */ 3821f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3822f82a9901SFilipe Manana if (IS_ERR(trans)) { 3823f82a9901SFilipe Manana ret = PTR_ERR(trans); 3824f82a9901SFilipe Manana goto out; 3825f82a9901SFilipe Manana } 3826f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3827f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3828f82a9901SFilipe Manana if (ret) { 3829f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 3830f82a9901SFilipe Manana btrfs_abort_transaction(trans, root, ret); 3831f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3832f82a9901SFilipe Manana goto out; 3833f82a9901SFilipe Manana } 383414f59796SFilipe Manana clone_update_extent_map(inode, trans, NULL, last_dest_end, 383514f59796SFilipe Manana destoff + len - last_dest_end); 3836f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38371c919a5eSMark Fasheh destoff, olen, no_time_update); 3838f82a9901SFilipe Manana } 3839f82a9901SFilipe Manana 3840f46b5a66SChristoph Hellwig out: 384132b7c687SMark Fasheh btrfs_free_path(path); 384215351955SDavid Sterba kvfree(buf); 384332b7c687SMark Fasheh return ret; 384432b7c687SMark Fasheh } 384532b7c687SMark Fasheh 38463db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 384732b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 384832b7c687SMark Fasheh { 384954563d41SAl Viro struct inode *inode = file_inode(file); 38503db11b2eSZach Brown struct inode *src = file_inode(file_src); 385132b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 385232b7c687SMark Fasheh int ret; 385332b7c687SMark Fasheh u64 len = olen; 385432b7c687SMark Fasheh u64 bs = root->fs_info->sb->s_blocksize; 38553db11b2eSZach Brown int same_inode = src == inode; 385632b7c687SMark Fasheh 385732b7c687SMark Fasheh /* 385832b7c687SMark Fasheh * TODO: 385932b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 386032b7c687SMark Fasheh * decompress into destination's address_space (the file offset 386132b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 386232b7c687SMark Fasheh * otherwise reinsert) a subrange. 386300fdf13aSLiu Bo * 386400fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 386500fdf13aSLiu Bo * be either compressed or non-compressed. 386632b7c687SMark Fasheh */ 386732b7c687SMark Fasheh 386832b7c687SMark Fasheh if (btrfs_root_readonly(root)) 386932b7c687SMark Fasheh return -EROFS; 387032b7c687SMark Fasheh 38713db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 38723db11b2eSZach Brown src->i_sb != inode->i_sb) 38733db11b2eSZach Brown return -EXDEV; 387432b7c687SMark Fasheh 387532b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 387632b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 387732b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 38783db11b2eSZach Brown return -EINVAL; 387932b7c687SMark Fasheh 388032b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 38813db11b2eSZach Brown return -EISDIR; 388232b7c687SMark Fasheh 388332b7c687SMark Fasheh if (!same_inode) { 3884293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 388532b7c687SMark Fasheh } else { 38865955102cSAl Viro inode_lock(src); 388732b7c687SMark Fasheh } 388832b7c687SMark Fasheh 388932b7c687SMark Fasheh /* determine range to clone */ 389032b7c687SMark Fasheh ret = -EINVAL; 389132b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 389232b7c687SMark Fasheh goto out_unlock; 389332b7c687SMark Fasheh if (len == 0) 389432b7c687SMark Fasheh olen = len = src->i_size - off; 389532b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 389632b7c687SMark Fasheh if (off + len == src->i_size) 389732b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 389832b7c687SMark Fasheh 3899ccccf3d6SFilipe Manana if (len == 0) { 3900ccccf3d6SFilipe Manana ret = 0; 3901ccccf3d6SFilipe Manana goto out_unlock; 3902ccccf3d6SFilipe Manana } 3903ccccf3d6SFilipe Manana 390432b7c687SMark Fasheh /* verify the end result is block aligned */ 390532b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 390632b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 390732b7c687SMark Fasheh goto out_unlock; 390832b7c687SMark Fasheh 390932b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 391032b7c687SMark Fasheh if (same_inode) { 391132b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 391232b7c687SMark Fasheh goto out_unlock; 391332b7c687SMark Fasheh } 391432b7c687SMark Fasheh 391532b7c687SMark Fasheh if (destoff > inode->i_size) { 391632b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 391732b7c687SMark Fasheh if (ret) 391832b7c687SMark Fasheh goto out_unlock; 391932b7c687SMark Fasheh } 392032b7c687SMark Fasheh 3921c125b8bfSFilipe Manana /* 3922c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3923c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3924c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3925c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3926c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3927c125b8bfSFilipe Manana */ 3928c125b8bfSFilipe Manana if (same_inode) { 3929c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3930c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 393132b7c687SMark Fasheh 3932e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 3933c125b8bfSFilipe Manana } else { 3934e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 3935e0bd70c6SFilipe Manana true); 3936e0bd70c6SFilipe Manana } 3937e0bd70c6SFilipe Manana ASSERT(ret == 0); 3938e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3939e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3940e0bd70c6SFilipe Manana goto out_unlock; 3941c125b8bfSFilipe Manana } 394232b7c687SMark Fasheh 39431c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 394432b7c687SMark Fasheh 3945c125b8bfSFilipe Manana if (same_inode) { 3946c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3947c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3948c125b8bfSFilipe Manana 3949c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3950c125b8bfSFilipe Manana } else { 3951293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 3952c125b8bfSFilipe Manana } 3953c125b8bfSFilipe Manana /* 3954c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3955c125b8bfSFilipe Manana * data immediately and not the previous data. 3956c125b8bfSFilipe Manana */ 395765bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 395809cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 395909cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 3960f46b5a66SChristoph Hellwig out_unlock: 3961293a8489SMark Fasheh if (!same_inode) 3962293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 3963293a8489SMark Fasheh else 39645955102cSAl Viro inode_unlock(src); 39653db11b2eSZach Brown return ret; 39663db11b2eSZach Brown } 39673db11b2eSZach Brown 39683db11b2eSZach Brown ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in, 39693db11b2eSZach Brown struct file *file_out, loff_t pos_out, 39703db11b2eSZach Brown size_t len, unsigned int flags) 39713db11b2eSZach Brown { 39723db11b2eSZach Brown ssize_t ret; 39733db11b2eSZach Brown 39743db11b2eSZach Brown ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out); 39753db11b2eSZach Brown if (ret == 0) 39763db11b2eSZach Brown ret = len; 39773db11b2eSZach Brown return ret; 39783db11b2eSZach Brown } 39793db11b2eSZach Brown 398004b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 398104b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 39823db11b2eSZach Brown { 398304b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 3984c5c9cd4dSSage Weil } 3985c5c9cd4dSSage Weil 3986f46b5a66SChristoph Hellwig /* 3987f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3988f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3989f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3990f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3991f46b5a66SChristoph Hellwig */ 3992b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 3993f46b5a66SChristoph Hellwig { 3994496ad9aaSAl Viro struct inode *inode = file_inode(file); 3995f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3996f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 39971ab86aedSSage Weil int ret; 3998f46b5a66SChristoph Hellwig 39991ab86aedSSage Weil ret = -EPERM; 4000df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 4001f46b5a66SChristoph Hellwig goto out; 40021ab86aedSSage Weil 40031ab86aedSSage Weil ret = -EINPROGRESS; 40041ab86aedSSage Weil if (file->private_data) 40051ab86aedSSage Weil goto out; 40069ca9ee09SSage Weil 4007b83cc969SLi Zefan ret = -EROFS; 4008b83cc969SLi Zefan if (btrfs_root_readonly(root)) 4009b83cc969SLi Zefan goto out; 4010b83cc969SLi Zefan 4011a561be71SAl Viro ret = mnt_want_write_file(file); 4012c146afadSYan Zheng if (ret) 4013c146afadSYan Zheng goto out; 4014c146afadSYan Zheng 4015a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 40169ca9ee09SSage Weil 4017f46b5a66SChristoph Hellwig ret = -ENOMEM; 40187a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 4019abd30bb0STsutomu Itoh if (IS_ERR(trans)) 40201ab86aedSSage Weil goto out_drop; 40211ab86aedSSage Weil 40221ab86aedSSage Weil file->private_data = trans; 40231ab86aedSSage Weil return 0; 40241ab86aedSSage Weil 40251ab86aedSSage Weil out_drop: 4026a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 40272a79f17eSAl Viro mnt_drop_write_file(file); 4028f46b5a66SChristoph Hellwig out: 4029f46b5a66SChristoph Hellwig return ret; 4030f46b5a66SChristoph Hellwig } 4031f46b5a66SChristoph Hellwig 40326ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40336ef5ed0dSJosef Bacik { 4034496ad9aaSAl Viro struct inode *inode = file_inode(file); 40356ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40366ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40376ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40386ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40396ef5ed0dSJosef Bacik struct btrfs_path *path; 40406ef5ed0dSJosef Bacik struct btrfs_key location; 40416ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40426ef5ed0dSJosef Bacik u64 objectid = 0; 40436ef5ed0dSJosef Bacik u64 dir_id; 40443c04ce01SMiao Xie int ret; 40456ef5ed0dSJosef Bacik 40466ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40476ef5ed0dSJosef Bacik return -EPERM; 40486ef5ed0dSJosef Bacik 40493c04ce01SMiao Xie ret = mnt_want_write_file(file); 40503c04ce01SMiao Xie if (ret) 40513c04ce01SMiao Xie return ret; 40523c04ce01SMiao Xie 40533c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40543c04ce01SMiao Xie ret = -EFAULT; 40553c04ce01SMiao Xie goto out; 40563c04ce01SMiao Xie } 40576ef5ed0dSJosef Bacik 40586ef5ed0dSJosef Bacik if (!objectid) 40591cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40606ef5ed0dSJosef Bacik 40616ef5ed0dSJosef Bacik location.objectid = objectid; 40626ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40636ef5ed0dSJosef Bacik location.offset = (u64)-1; 40646ef5ed0dSJosef Bacik 40656ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 40663c04ce01SMiao Xie if (IS_ERR(new_root)) { 40673c04ce01SMiao Xie ret = PTR_ERR(new_root); 40683c04ce01SMiao Xie goto out; 40693c04ce01SMiao Xie } 40706ef5ed0dSJosef Bacik 40716ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40723c04ce01SMiao Xie if (!path) { 40733c04ce01SMiao Xie ret = -ENOMEM; 40743c04ce01SMiao Xie goto out; 40753c04ce01SMiao Xie } 40766ef5ed0dSJosef Bacik path->leave_spinning = 1; 40776ef5ed0dSJosef Bacik 40786ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 407998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40806ef5ed0dSJosef Bacik btrfs_free_path(path); 40813c04ce01SMiao Xie ret = PTR_ERR(trans); 40823c04ce01SMiao Xie goto out; 40836ef5ed0dSJosef Bacik } 40846ef5ed0dSJosef Bacik 40856c41761fSDavid Sterba dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 40866ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 40876ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4088cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40896ef5ed0dSJosef Bacik btrfs_free_path(path); 40906ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 4091efe120a0SFrank Holton btrfs_err(new_root->fs_info, "Umm, you don't have the default dir" 4092efe120a0SFrank Holton "item, this isn't going to work"); 40933c04ce01SMiao Xie ret = -ENOENT; 40943c04ce01SMiao Xie goto out; 40956ef5ed0dSJosef Bacik } 40966ef5ed0dSJosef Bacik 40976ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 40986ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 40996ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 41006ef5ed0dSJosef Bacik btrfs_free_path(path); 41016ef5ed0dSJosef Bacik 41022b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); 41036ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 41043c04ce01SMiao Xie out: 41053c04ce01SMiao Xie mnt_drop_write_file(file); 41063c04ce01SMiao Xie return ret; 41076ef5ed0dSJosef Bacik } 41086ef5ed0dSJosef Bacik 41095af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 4110bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4111bf5fc093SJosef Bacik { 4112bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4113bf5fc093SJosef Bacik 4114bf5fc093SJosef Bacik space->total_bytes = 0; 4115bf5fc093SJosef Bacik space->used_bytes = 0; 4116bf5fc093SJosef Bacik space->flags = 0; 4117bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4118bf5fc093SJosef Bacik space->flags = block_group->flags; 4119bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4120bf5fc093SJosef Bacik space->used_bytes += 4121bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4122bf5fc093SJosef Bacik } 4123bf5fc093SJosef Bacik } 4124bf5fc093SJosef Bacik 412548a3b636SEric Sandeen static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 41261406e432SJosef Bacik { 41271406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41281406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41291406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41307fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 413113f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41321406e432SJosef Bacik struct btrfs_space_info *info; 4133bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 4134bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4135bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4136bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 4137bf5fc093SJosef Bacik int num_types = 4; 41387fde62bfSChris Mason int alloc_size; 41391406e432SJosef Bacik int ret = 0; 414051788b1bSDan Rosenberg u64 slot_count = 0; 4141bf5fc093SJosef Bacik int i, c; 41421406e432SJosef Bacik 41431406e432SJosef Bacik if (copy_from_user(&space_args, 41441406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41451406e432SJosef Bacik sizeof(space_args))) 41461406e432SJosef Bacik return -EFAULT; 41471406e432SJosef Bacik 4148bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4149bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4150bf5fc093SJosef Bacik 4151bf5fc093SJosef Bacik info = NULL; 41527fde62bfSChris Mason rcu_read_lock(); 4153bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 4154bf5fc093SJosef Bacik list) { 4155bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4156bf5fc093SJosef Bacik info = tmp; 4157bf5fc093SJosef Bacik break; 4158bf5fc093SJosef Bacik } 4159bf5fc093SJosef Bacik } 41607fde62bfSChris Mason rcu_read_unlock(); 41611406e432SJosef Bacik 4162bf5fc093SJosef Bacik if (!info) 4163bf5fc093SJosef Bacik continue; 4164bf5fc093SJosef Bacik 4165bf5fc093SJosef Bacik down_read(&info->groups_sem); 4166bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4167bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4168bf5fc093SJosef Bacik slot_count++; 4169bf5fc093SJosef Bacik } 4170bf5fc093SJosef Bacik up_read(&info->groups_sem); 4171bf5fc093SJosef Bacik } 4172bf5fc093SJosef Bacik 417336523e95SDavid Sterba /* 417436523e95SDavid Sterba * Global block reserve, exported as a space_info 417536523e95SDavid Sterba */ 417636523e95SDavid Sterba slot_count++; 417736523e95SDavid Sterba 41787fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 41797fde62bfSChris Mason if (space_args.space_slots == 0) { 41807fde62bfSChris Mason space_args.total_spaces = slot_count; 41817fde62bfSChris Mason goto out; 41827fde62bfSChris Mason } 4183bf5fc093SJosef Bacik 418451788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4185bf5fc093SJosef Bacik 41867fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4187bf5fc093SJosef Bacik 41887fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41897fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41907fde62bfSChris Mason */ 419109cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 41927fde62bfSChris Mason return -ENOMEM; 41937fde62bfSChris Mason 41947fde62bfSChris Mason space_args.total_spaces = 0; 41958d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 41967fde62bfSChris Mason if (!dest) 41977fde62bfSChris Mason return -ENOMEM; 41987fde62bfSChris Mason dest_orig = dest; 41997fde62bfSChris Mason 42007fde62bfSChris Mason /* now we have a buffer to copy into */ 4201bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4202bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4203bf5fc093SJosef Bacik 420451788b1bSDan Rosenberg if (!slot_count) 420551788b1bSDan Rosenberg break; 420651788b1bSDan Rosenberg 4207bf5fc093SJosef Bacik info = NULL; 42081406e432SJosef Bacik rcu_read_lock(); 4209bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 4210bf5fc093SJosef Bacik list) { 4211bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4212bf5fc093SJosef Bacik info = tmp; 42137fde62bfSChris Mason break; 4214bf5fc093SJosef Bacik } 4215bf5fc093SJosef Bacik } 4216bf5fc093SJosef Bacik rcu_read_unlock(); 42177fde62bfSChris Mason 4218bf5fc093SJosef Bacik if (!info) 4219bf5fc093SJosef Bacik continue; 4220bf5fc093SJosef Bacik down_read(&info->groups_sem); 4221bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4222bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 42235af3e8ccSStefan Behrens btrfs_get_block_group_info( 42245af3e8ccSStefan Behrens &info->block_groups[c], &space); 42257fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42261406e432SJosef Bacik dest++; 42271406e432SJosef Bacik space_args.total_spaces++; 422851788b1bSDan Rosenberg slot_count--; 42291406e432SJosef Bacik } 423051788b1bSDan Rosenberg if (!slot_count) 423151788b1bSDan Rosenberg break; 4232bf5fc093SJosef Bacik } 4233bf5fc093SJosef Bacik up_read(&info->groups_sem); 4234bf5fc093SJosef Bacik } 42351406e432SJosef Bacik 423636523e95SDavid Sterba /* 423736523e95SDavid Sterba * Add global block reserve 423836523e95SDavid Sterba */ 423936523e95SDavid Sterba if (slot_count) { 424036523e95SDavid Sterba struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv; 424136523e95SDavid Sterba 424236523e95SDavid Sterba spin_lock(&block_rsv->lock); 424336523e95SDavid Sterba space.total_bytes = block_rsv->size; 424436523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 424536523e95SDavid Sterba spin_unlock(&block_rsv->lock); 424636523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 424736523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 424836523e95SDavid Sterba space_args.total_spaces++; 424936523e95SDavid Sterba } 425036523e95SDavid Sterba 42512eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42527fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42537fde62bfSChris Mason 42547fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42557fde62bfSChris Mason ret = -EFAULT; 42567fde62bfSChris Mason 42577fde62bfSChris Mason kfree(dest_orig); 42587fde62bfSChris Mason out: 42597fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42601406e432SJosef Bacik ret = -EFAULT; 42611406e432SJosef Bacik 42621406e432SJosef Bacik return ret; 42631406e432SJosef Bacik } 42641406e432SJosef Bacik 4265f46b5a66SChristoph Hellwig /* 4266f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4267f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4268f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4269f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4270f46b5a66SChristoph Hellwig */ 4271f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 4272f46b5a66SChristoph Hellwig { 4273496ad9aaSAl Viro struct inode *inode = file_inode(file); 4274f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4275f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4276f46b5a66SChristoph Hellwig 4277f46b5a66SChristoph Hellwig trans = file->private_data; 42781ab86aedSSage Weil if (!trans) 42791ab86aedSSage Weil return -EINVAL; 4280b214107eSChristoph Hellwig file->private_data = NULL; 42819ca9ee09SSage Weil 42821ab86aedSSage Weil btrfs_end_transaction(trans, root); 42831ab86aedSSage Weil 4284a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 42859ca9ee09SSage Weil 42862a79f17eSAl Viro mnt_drop_write_file(file); 42871ab86aedSSage Weil return 0; 4288f46b5a66SChristoph Hellwig } 4289f46b5a66SChristoph Hellwig 42909a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42919a8c28beSMiao Xie void __user *argp) 429246204592SSage Weil { 429346204592SSage Weil struct btrfs_trans_handle *trans; 429446204592SSage Weil u64 transid; 4295db5b493aSTsutomu Itoh int ret; 429646204592SSage Weil 4297d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4298ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4299ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 430098d5dc13STsutomu Itoh return PTR_ERR(trans); 4301ff7c1d33SMiao Xie 4302ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4303ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4304ff7c1d33SMiao Xie goto out; 4305ff7c1d33SMiao Xie } 430646204592SSage Weil transid = trans->transid; 4307db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 43088b2b2d3cSTsutomu Itoh if (ret) { 43098b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 4310db5b493aSTsutomu Itoh return ret; 43118b2b2d3cSTsutomu Itoh } 4312ff7c1d33SMiao Xie out: 431346204592SSage Weil if (argp) 431446204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 431546204592SSage Weil return -EFAULT; 431646204592SSage Weil return 0; 431746204592SSage Weil } 431846204592SSage Weil 43199a8c28beSMiao Xie static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, 43209a8c28beSMiao Xie void __user *argp) 432146204592SSage Weil { 432246204592SSage Weil u64 transid; 432346204592SSage Weil 432446204592SSage Weil if (argp) { 432546204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 432646204592SSage Weil return -EFAULT; 432746204592SSage Weil } else { 432846204592SSage Weil transid = 0; /* current trans */ 432946204592SSage Weil } 433046204592SSage Weil return btrfs_wait_for_commit(root, transid); 433146204592SSage Weil } 433246204592SSage Weil 4333b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4334475f6387SJan Schmidt { 4335496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4336475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4337b8e95489SMiao Xie int ret; 4338475f6387SJan Schmidt 4339475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4340475f6387SJan Schmidt return -EPERM; 4341475f6387SJan Schmidt 4342475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4343475f6387SJan Schmidt if (IS_ERR(sa)) 4344475f6387SJan Schmidt return PTR_ERR(sa); 4345475f6387SJan Schmidt 4346b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4347b8e95489SMiao Xie ret = mnt_want_write_file(file); 4348b8e95489SMiao Xie if (ret) 4349b8e95489SMiao Xie goto out; 4350b8e95489SMiao Xie } 4351b8e95489SMiao Xie 4352aa1b8cd4SStefan Behrens ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, 435363a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 435463a212abSStefan Behrens 0); 4355475f6387SJan Schmidt 4356475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4357475f6387SJan Schmidt ret = -EFAULT; 4358475f6387SJan Schmidt 4359b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4360b8e95489SMiao Xie mnt_drop_write_file(file); 4361b8e95489SMiao Xie out: 4362475f6387SJan Schmidt kfree(sa); 4363475f6387SJan Schmidt return ret; 4364475f6387SJan Schmidt } 4365475f6387SJan Schmidt 4366475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 4367475f6387SJan Schmidt { 4368475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4369475f6387SJan Schmidt return -EPERM; 4370475f6387SJan Schmidt 4371aa1b8cd4SStefan Behrens return btrfs_scrub_cancel(root->fs_info); 4372475f6387SJan Schmidt } 4373475f6387SJan Schmidt 4374475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 4375475f6387SJan Schmidt void __user *arg) 4376475f6387SJan Schmidt { 4377475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4378475f6387SJan Schmidt int ret; 4379475f6387SJan Schmidt 4380475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4381475f6387SJan Schmidt return -EPERM; 4382475f6387SJan Schmidt 4383475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4384475f6387SJan Schmidt if (IS_ERR(sa)) 4385475f6387SJan Schmidt return PTR_ERR(sa); 4386475f6387SJan Schmidt 4387475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 4388475f6387SJan Schmidt 4389475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4390475f6387SJan Schmidt ret = -EFAULT; 4391475f6387SJan Schmidt 4392475f6387SJan Schmidt kfree(sa); 4393475f6387SJan Schmidt return ret; 4394475f6387SJan Schmidt } 4395475f6387SJan Schmidt 4396c11d2c23SStefan Behrens static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, 4397b27f7c0cSDavid Sterba void __user *arg) 4398c11d2c23SStefan Behrens { 4399c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4400c11d2c23SStefan Behrens int ret; 4401c11d2c23SStefan Behrens 4402c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4403c11d2c23SStefan Behrens if (IS_ERR(sa)) 4404c11d2c23SStefan Behrens return PTR_ERR(sa); 4405c11d2c23SStefan Behrens 4406b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4407b27f7c0cSDavid Sterba kfree(sa); 4408b27f7c0cSDavid Sterba return -EPERM; 4409b27f7c0cSDavid Sterba } 4410b27f7c0cSDavid Sterba 4411b27f7c0cSDavid Sterba ret = btrfs_get_dev_stats(root, sa); 4412c11d2c23SStefan Behrens 4413c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4414c11d2c23SStefan Behrens ret = -EFAULT; 4415c11d2c23SStefan Behrens 4416c11d2c23SStefan Behrens kfree(sa); 4417c11d2c23SStefan Behrens return ret; 4418c11d2c23SStefan Behrens } 4419c11d2c23SStefan Behrens 44203f6bcfbdSStefan Behrens static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) 44213f6bcfbdSStefan Behrens { 44223f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 44233f6bcfbdSStefan Behrens int ret; 44243f6bcfbdSStefan Behrens 44253f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44263f6bcfbdSStefan Behrens return -EPERM; 44273f6bcfbdSStefan Behrens 44283f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44293f6bcfbdSStefan Behrens if (IS_ERR(p)) 44303f6bcfbdSStefan Behrens return PTR_ERR(p); 44313f6bcfbdSStefan Behrens 44323f6bcfbdSStefan Behrens switch (p->cmd) { 44333f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4434adfa97cbSIlya Dryomov if (root->fs_info->sb->s_flags & MS_RDONLY) { 4435adfa97cbSIlya Dryomov ret = -EROFS; 4436adfa97cbSIlya Dryomov goto out; 4437adfa97cbSIlya Dryomov } 44383f6bcfbdSStefan Behrens if (atomic_xchg( 44393f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 44403f6bcfbdSStefan Behrens 1)) { 4441e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44423f6bcfbdSStefan Behrens } else { 4443b5255456SAnand Jain ret = btrfs_dev_replace_by_ioctl(root, p); 44443f6bcfbdSStefan Behrens atomic_set( 44453f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 44463f6bcfbdSStefan Behrens 0); 44473f6bcfbdSStefan Behrens } 44483f6bcfbdSStefan Behrens break; 44493f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44503f6bcfbdSStefan Behrens btrfs_dev_replace_status(root->fs_info, p); 44513f6bcfbdSStefan Behrens ret = 0; 44523f6bcfbdSStefan Behrens break; 44533f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 44543f6bcfbdSStefan Behrens ret = btrfs_dev_replace_cancel(root->fs_info, p); 44553f6bcfbdSStefan Behrens break; 44563f6bcfbdSStefan Behrens default: 44573f6bcfbdSStefan Behrens ret = -EINVAL; 44583f6bcfbdSStefan Behrens break; 44593f6bcfbdSStefan Behrens } 44603f6bcfbdSStefan Behrens 44613f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 44623f6bcfbdSStefan Behrens ret = -EFAULT; 4463adfa97cbSIlya Dryomov out: 44643f6bcfbdSStefan Behrens kfree(p); 44653f6bcfbdSStefan Behrens return ret; 44663f6bcfbdSStefan Behrens } 44673f6bcfbdSStefan Behrens 4468d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4469d7728c96SJan Schmidt { 4470d7728c96SJan Schmidt int ret = 0; 4471d7728c96SJan Schmidt int i; 4472740c3d22SChris Mason u64 rel_ptr; 4473d7728c96SJan Schmidt int size; 4474806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4475d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4476d7728c96SJan Schmidt struct btrfs_path *path; 4477d7728c96SJan Schmidt 447882b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4479d7728c96SJan Schmidt return -EPERM; 4480d7728c96SJan Schmidt 4481d7728c96SJan Schmidt path = btrfs_alloc_path(); 4482d7728c96SJan Schmidt if (!path) { 4483d7728c96SJan Schmidt ret = -ENOMEM; 4484d7728c96SJan Schmidt goto out; 4485d7728c96SJan Schmidt } 4486d7728c96SJan Schmidt 4487d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4488d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4489d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4490d7728c96SJan Schmidt ipa = NULL; 4491d7728c96SJan Schmidt goto out; 4492d7728c96SJan Schmidt } 4493d7728c96SJan Schmidt 4494d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4495d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4496d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4497d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4498d7728c96SJan Schmidt ipath = NULL; 4499d7728c96SJan Schmidt goto out; 4500d7728c96SJan Schmidt } 4501d7728c96SJan Schmidt 4502d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4503d7728c96SJan Schmidt if (ret < 0) 4504d7728c96SJan Schmidt goto out; 4505d7728c96SJan Schmidt 4506d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4507745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4508745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4509740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4510d7728c96SJan Schmidt } 4511d7728c96SJan Schmidt 4512745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 4513745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 4514d7728c96SJan Schmidt if (ret) { 4515d7728c96SJan Schmidt ret = -EFAULT; 4516d7728c96SJan Schmidt goto out; 4517d7728c96SJan Schmidt } 4518d7728c96SJan Schmidt 4519d7728c96SJan Schmidt out: 4520d7728c96SJan Schmidt btrfs_free_path(path); 4521d7728c96SJan Schmidt free_ipath(ipath); 4522d7728c96SJan Schmidt kfree(ipa); 4523d7728c96SJan Schmidt 4524d7728c96SJan Schmidt return ret; 4525d7728c96SJan Schmidt } 4526d7728c96SJan Schmidt 4527d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4528d7728c96SJan Schmidt { 4529d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4530d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4531d7728c96SJan Schmidt 4532d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4533d7728c96SJan Schmidt inodes->bytes_left -= c; 4534d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4535d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4536d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4537d7728c96SJan Schmidt inodes->elem_cnt += 3; 4538d7728c96SJan Schmidt } else { 4539d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4540d7728c96SJan Schmidt inodes->bytes_left = 0; 4541d7728c96SJan Schmidt inodes->elem_missed += 3; 4542d7728c96SJan Schmidt } 4543d7728c96SJan Schmidt 4544d7728c96SJan Schmidt return 0; 4545d7728c96SJan Schmidt } 4546d7728c96SJan Schmidt 4547d7728c96SJan Schmidt static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 4548d7728c96SJan Schmidt void __user *arg) 4549d7728c96SJan Schmidt { 4550d7728c96SJan Schmidt int ret = 0; 4551d7728c96SJan Schmidt int size; 4552d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4553d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4554d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4555d7728c96SJan Schmidt 4556d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4557d7728c96SJan Schmidt return -EPERM; 4558d7728c96SJan Schmidt 4559d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 4560d7728c96SJan Schmidt if (IS_ERR(loi)) { 4561d7728c96SJan Schmidt ret = PTR_ERR(loi); 4562d7728c96SJan Schmidt loi = NULL; 4563d7728c96SJan Schmidt goto out; 4564d7728c96SJan Schmidt } 4565d7728c96SJan Schmidt 4566d7728c96SJan Schmidt path = btrfs_alloc_path(); 4567d7728c96SJan Schmidt if (!path) { 4568d7728c96SJan Schmidt ret = -ENOMEM; 4569d7728c96SJan Schmidt goto out; 4570d7728c96SJan Schmidt } 4571d7728c96SJan Schmidt 4572ee22184bSByongho Lee size = min_t(u32, loi->size, SZ_64K); 4573d7728c96SJan Schmidt inodes = init_data_container(size); 4574d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4575d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4576d7728c96SJan Schmidt inodes = NULL; 4577d7728c96SJan Schmidt goto out; 4578d7728c96SJan Schmidt } 4579d7728c96SJan Schmidt 4580df031f07SLiu Bo ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, 4581df031f07SLiu Bo build_ino_list, inodes); 4582df031f07SLiu Bo if (ret == -EINVAL) 4583d7728c96SJan Schmidt ret = -ENOENT; 4584d7728c96SJan Schmidt if (ret < 0) 4585d7728c96SJan Schmidt goto out; 4586d7728c96SJan Schmidt 4587745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 4588745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 4589d7728c96SJan Schmidt if (ret) 4590d7728c96SJan Schmidt ret = -EFAULT; 4591d7728c96SJan Schmidt 4592d7728c96SJan Schmidt out: 4593d7728c96SJan Schmidt btrfs_free_path(path); 4594425d17a2SLiu Bo vfree(inodes); 4595d7728c96SJan Schmidt kfree(loi); 4596d7728c96SJan Schmidt 4597d7728c96SJan Schmidt return ret; 4598d7728c96SJan Schmidt } 4599d7728c96SJan Schmidt 460019a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4601c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4602c9e9f97bSIlya Dryomov { 4603c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4604c9e9f97bSIlya Dryomov 4605c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4606c9e9f97bSIlya Dryomov 4607837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 4608837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4609837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4610837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4611a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4612a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4613837d5b6eSIlya Dryomov 4614c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4615c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4616c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 461719a39dceSIlya Dryomov 461819a39dceSIlya Dryomov if (lock) { 461919a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 462019a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 462119a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 462219a39dceSIlya Dryomov } else { 462319a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 462419a39dceSIlya Dryomov } 4625c9e9f97bSIlya Dryomov } 4626c9e9f97bSIlya Dryomov 46279ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4628c9e9f97bSIlya Dryomov { 4629496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4630c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4631c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4632c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4633ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4634c9e9f97bSIlya Dryomov int ret; 4635c9e9f97bSIlya Dryomov 4636c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4637c9e9f97bSIlya Dryomov return -EPERM; 4638c9e9f97bSIlya Dryomov 4639e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46409ba1f6e4SLiu Bo if (ret) 46419ba1f6e4SLiu Bo return ret; 46429ba1f6e4SLiu Bo 4643ed0fb78fSIlya Dryomov again: 4644ed0fb78fSIlya Dryomov if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 4645c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 4646c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4647ed0fb78fSIlya Dryomov need_unlock = true; 4648ed0fb78fSIlya Dryomov goto locked; 4649ed0fb78fSIlya Dryomov } 4650ed0fb78fSIlya Dryomov 4651ed0fb78fSIlya Dryomov /* 465201327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4653ed0fb78fSIlya Dryomov * (1) some other op is running 4654ed0fb78fSIlya Dryomov * (2) balance is running 4655ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4656ed0fb78fSIlya Dryomov */ 4657ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4658ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4659ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 4660ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 4661ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4662ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 4663ed0fb78fSIlya Dryomov goto again; 4664ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4665ed0fb78fSIlya Dryomov 4666ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 4667ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 4668ed0fb78fSIlya Dryomov /* this is (3) */ 4669ed0fb78fSIlya Dryomov need_unlock = false; 4670ed0fb78fSIlya Dryomov goto locked; 4671ed0fb78fSIlya Dryomov } 4672ed0fb78fSIlya Dryomov 4673ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4674ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4675ed0fb78fSIlya Dryomov goto again; 4676ed0fb78fSIlya Dryomov } else { 4677ed0fb78fSIlya Dryomov /* this is (2) */ 4678ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4679ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4680ed0fb78fSIlya Dryomov goto out; 4681ed0fb78fSIlya Dryomov } 4682ed0fb78fSIlya Dryomov } else { 4683ed0fb78fSIlya Dryomov /* this is (1) */ 4684ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4685e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4686ed0fb78fSIlya Dryomov goto out; 4687ed0fb78fSIlya Dryomov } 4688ed0fb78fSIlya Dryomov 4689ed0fb78fSIlya Dryomov locked: 4690ed0fb78fSIlya Dryomov BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); 4691c9e9f97bSIlya Dryomov 4692c9e9f97bSIlya Dryomov if (arg) { 4693c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4694c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4695c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4696ed0fb78fSIlya Dryomov goto out_unlock; 4697c9e9f97bSIlya Dryomov } 4698de322263SIlya Dryomov 4699de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4700de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4701de322263SIlya Dryomov ret = -ENOTCONN; 4702de322263SIlya Dryomov goto out_bargs; 4703de322263SIlya Dryomov } 4704de322263SIlya Dryomov 4705de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4706de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4707de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4708de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4709de322263SIlya Dryomov 4710de322263SIlya Dryomov goto do_balance; 4711de322263SIlya Dryomov } 4712c9e9f97bSIlya Dryomov } else { 4713c9e9f97bSIlya Dryomov bargs = NULL; 4714c9e9f97bSIlya Dryomov } 4715c9e9f97bSIlya Dryomov 4716ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4717837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4718837d5b6eSIlya Dryomov goto out_bargs; 4719837d5b6eSIlya Dryomov } 4720837d5b6eSIlya Dryomov 47218d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4722c9e9f97bSIlya Dryomov if (!bctl) { 4723c9e9f97bSIlya Dryomov ret = -ENOMEM; 4724c9e9f97bSIlya Dryomov goto out_bargs; 4725c9e9f97bSIlya Dryomov } 4726c9e9f97bSIlya Dryomov 4727c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 4728c9e9f97bSIlya Dryomov if (arg) { 4729c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4730c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4731c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4732c9e9f97bSIlya Dryomov 4733c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4734f43ffb60SIlya Dryomov } else { 4735f43ffb60SIlya Dryomov /* balance everything - no filters */ 4736f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4737c9e9f97bSIlya Dryomov } 4738c9e9f97bSIlya Dryomov 47398eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47408eb93459SDavid Sterba ret = -EINVAL; 47410f89abf5SChristian Engelmayer goto out_bctl; 47428eb93459SDavid Sterba } 47438eb93459SDavid Sterba 4744de322263SIlya Dryomov do_balance: 4745c9e9f97bSIlya Dryomov /* 4746ed0fb78fSIlya Dryomov * Ownership of bctl and mutually_exclusive_operation_running 4747ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4748ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 4749ed0fb78fSIlya Dryomov * free_fs_info. mutually_exclusive_operation_running is 4750ed0fb78fSIlya Dryomov * cleared in __cancel_balance. 4751c9e9f97bSIlya Dryomov */ 4752ed0fb78fSIlya Dryomov need_unlock = false; 4753ed0fb78fSIlya Dryomov 4754ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 47550f89abf5SChristian Engelmayer bctl = NULL; 4756ed0fb78fSIlya Dryomov 4757c9e9f97bSIlya Dryomov if (arg) { 4758c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4759c9e9f97bSIlya Dryomov ret = -EFAULT; 4760c9e9f97bSIlya Dryomov } 4761c9e9f97bSIlya Dryomov 47620f89abf5SChristian Engelmayer out_bctl: 47630f89abf5SChristian Engelmayer kfree(bctl); 4764c9e9f97bSIlya Dryomov out_bargs: 4765c9e9f97bSIlya Dryomov kfree(bargs); 4766ed0fb78fSIlya Dryomov out_unlock: 4767c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4768c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4769ed0fb78fSIlya Dryomov if (need_unlock) 4770ed0fb78fSIlya Dryomov atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 4771ed0fb78fSIlya Dryomov out: 4772e54bfa31SLiu Bo mnt_drop_write_file(file); 4773c9e9f97bSIlya Dryomov return ret; 4774c9e9f97bSIlya Dryomov } 4775c9e9f97bSIlya Dryomov 4776837d5b6eSIlya Dryomov static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 4777837d5b6eSIlya Dryomov { 4778837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4779837d5b6eSIlya Dryomov return -EPERM; 4780837d5b6eSIlya Dryomov 4781837d5b6eSIlya Dryomov switch (cmd) { 4782837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 4783837d5b6eSIlya Dryomov return btrfs_pause_balance(root->fs_info); 4784a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 4785a7e99c69SIlya Dryomov return btrfs_cancel_balance(root->fs_info); 4786837d5b6eSIlya Dryomov } 4787837d5b6eSIlya Dryomov 4788837d5b6eSIlya Dryomov return -EINVAL; 4789837d5b6eSIlya Dryomov } 4790837d5b6eSIlya Dryomov 479119a39dceSIlya Dryomov static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 479219a39dceSIlya Dryomov void __user *arg) 479319a39dceSIlya Dryomov { 479419a39dceSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 479519a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 479619a39dceSIlya Dryomov int ret = 0; 479719a39dceSIlya Dryomov 479819a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 479919a39dceSIlya Dryomov return -EPERM; 480019a39dceSIlya Dryomov 480119a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 480219a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 480319a39dceSIlya Dryomov ret = -ENOTCONN; 480419a39dceSIlya Dryomov goto out; 480519a39dceSIlya Dryomov } 480619a39dceSIlya Dryomov 48078d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 480819a39dceSIlya Dryomov if (!bargs) { 480919a39dceSIlya Dryomov ret = -ENOMEM; 481019a39dceSIlya Dryomov goto out; 481119a39dceSIlya Dryomov } 481219a39dceSIlya Dryomov 481319a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 481419a39dceSIlya Dryomov 481519a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 481619a39dceSIlya Dryomov ret = -EFAULT; 481719a39dceSIlya Dryomov 481819a39dceSIlya Dryomov kfree(bargs); 481919a39dceSIlya Dryomov out: 482019a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 482119a39dceSIlya Dryomov return ret; 482219a39dceSIlya Dryomov } 482319a39dceSIlya Dryomov 4824905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 48255d13a37bSArne Jansen { 4826496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 48275d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 48285d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 48295d13a37bSArne Jansen int ret; 48305d13a37bSArne Jansen int err; 48315d13a37bSArne Jansen 48325d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48335d13a37bSArne Jansen return -EPERM; 48345d13a37bSArne Jansen 4835905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4836905b0ddaSMiao Xie if (ret) 4837905b0ddaSMiao Xie return ret; 48385d13a37bSArne Jansen 48395d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4840905b0ddaSMiao Xie if (IS_ERR(sa)) { 4841905b0ddaSMiao Xie ret = PTR_ERR(sa); 4842905b0ddaSMiao Xie goto drop_write; 4843905b0ddaSMiao Xie } 48445d13a37bSArne Jansen 48457708f029SWang Shilong down_write(&root->fs_info->subvol_sem); 484692f183aaSWang Shilong trans = btrfs_start_transaction(root->fs_info->tree_root, 2); 48475d13a37bSArne Jansen if (IS_ERR(trans)) { 48485d13a37bSArne Jansen ret = PTR_ERR(trans); 48495d13a37bSArne Jansen goto out; 48505d13a37bSArne Jansen } 48515d13a37bSArne Jansen 48525d13a37bSArne Jansen switch (sa->cmd) { 48535d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 48545d13a37bSArne Jansen ret = btrfs_quota_enable(trans, root->fs_info); 48555d13a37bSArne Jansen break; 48565d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 48575d13a37bSArne Jansen ret = btrfs_quota_disable(trans, root->fs_info); 48585d13a37bSArne Jansen break; 48595d13a37bSArne Jansen default: 48605d13a37bSArne Jansen ret = -EINVAL; 48615d13a37bSArne Jansen break; 48625d13a37bSArne Jansen } 48635d13a37bSArne Jansen 486492f183aaSWang Shilong err = btrfs_commit_transaction(trans, root->fs_info->tree_root); 48655d13a37bSArne Jansen if (err && !ret) 48665d13a37bSArne Jansen ret = err; 48675d13a37bSArne Jansen out: 48685d13a37bSArne Jansen kfree(sa); 48697708f029SWang Shilong up_write(&root->fs_info->subvol_sem); 4870905b0ddaSMiao Xie drop_write: 4871905b0ddaSMiao Xie mnt_drop_write_file(file); 48725d13a37bSArne Jansen return ret; 48735d13a37bSArne Jansen } 48745d13a37bSArne Jansen 4875905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48765d13a37bSArne Jansen { 4877496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 48785d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48795d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48805d13a37bSArne Jansen int ret; 48815d13a37bSArne Jansen int err; 48825d13a37bSArne Jansen 48835d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48845d13a37bSArne Jansen return -EPERM; 48855d13a37bSArne Jansen 4886905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4887905b0ddaSMiao Xie if (ret) 4888905b0ddaSMiao Xie return ret; 48895d13a37bSArne Jansen 48905d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4891905b0ddaSMiao Xie if (IS_ERR(sa)) { 4892905b0ddaSMiao Xie ret = PTR_ERR(sa); 4893905b0ddaSMiao Xie goto drop_write; 4894905b0ddaSMiao Xie } 48955d13a37bSArne Jansen 48965d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48975d13a37bSArne Jansen if (IS_ERR(trans)) { 48985d13a37bSArne Jansen ret = PTR_ERR(trans); 48995d13a37bSArne Jansen goto out; 49005d13a37bSArne Jansen } 49015d13a37bSArne Jansen 49025d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49035d13a37bSArne Jansen if (sa->assign) { 49045d13a37bSArne Jansen ret = btrfs_add_qgroup_relation(trans, root->fs_info, 49055d13a37bSArne Jansen sa->src, sa->dst); 49065d13a37bSArne Jansen } else { 49075d13a37bSArne Jansen ret = btrfs_del_qgroup_relation(trans, root->fs_info, 49085d13a37bSArne Jansen sa->src, sa->dst); 49095d13a37bSArne Jansen } 49105d13a37bSArne Jansen 4911e082f563SQu Wenruo /* update qgroup status and info */ 4912e082f563SQu Wenruo err = btrfs_run_qgroups(trans, root->fs_info); 4913e082f563SQu Wenruo if (err < 0) 4914ad8403dfSAnand Jain btrfs_handle_fs_error(root->fs_info, err, 4915ad8403dfSAnand Jain "failed to update qgroup status and info"); 49165d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 49175d13a37bSArne Jansen if (err && !ret) 49185d13a37bSArne Jansen ret = err; 49195d13a37bSArne Jansen 49205d13a37bSArne Jansen out: 49215d13a37bSArne Jansen kfree(sa); 4922905b0ddaSMiao Xie drop_write: 4923905b0ddaSMiao Xie mnt_drop_write_file(file); 49245d13a37bSArne Jansen return ret; 49255d13a37bSArne Jansen } 49265d13a37bSArne Jansen 4927905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 49285d13a37bSArne Jansen { 4929496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 49305d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49315d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49325d13a37bSArne Jansen int ret; 49335d13a37bSArne Jansen int err; 49345d13a37bSArne Jansen 49355d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49365d13a37bSArne Jansen return -EPERM; 49375d13a37bSArne Jansen 4938905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4939905b0ddaSMiao Xie if (ret) 4940905b0ddaSMiao Xie return ret; 49415d13a37bSArne Jansen 49425d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4943905b0ddaSMiao Xie if (IS_ERR(sa)) { 4944905b0ddaSMiao Xie ret = PTR_ERR(sa); 4945905b0ddaSMiao Xie goto drop_write; 4946905b0ddaSMiao Xie } 49475d13a37bSArne Jansen 4948d86e56cfSMiao Xie if (!sa->qgroupid) { 4949d86e56cfSMiao Xie ret = -EINVAL; 4950d86e56cfSMiao Xie goto out; 4951d86e56cfSMiao Xie } 4952d86e56cfSMiao Xie 49535d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49545d13a37bSArne Jansen if (IS_ERR(trans)) { 49555d13a37bSArne Jansen ret = PTR_ERR(trans); 49565d13a37bSArne Jansen goto out; 49575d13a37bSArne Jansen } 49585d13a37bSArne Jansen 49595d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49605d13a37bSArne Jansen if (sa->create) { 49614087cf24SDongsheng Yang ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid); 49625d13a37bSArne Jansen } else { 49635d13a37bSArne Jansen ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); 49645d13a37bSArne Jansen } 49655d13a37bSArne Jansen 49665d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 49675d13a37bSArne Jansen if (err && !ret) 49685d13a37bSArne Jansen ret = err; 49695d13a37bSArne Jansen 49705d13a37bSArne Jansen out: 49715d13a37bSArne Jansen kfree(sa); 4972905b0ddaSMiao Xie drop_write: 4973905b0ddaSMiao Xie mnt_drop_write_file(file); 49745d13a37bSArne Jansen return ret; 49755d13a37bSArne Jansen } 49765d13a37bSArne Jansen 4977905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49785d13a37bSArne Jansen { 4979496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 49805d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49815d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49825d13a37bSArne Jansen int ret; 49835d13a37bSArne Jansen int err; 49845d13a37bSArne Jansen u64 qgroupid; 49855d13a37bSArne Jansen 49865d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49875d13a37bSArne Jansen return -EPERM; 49885d13a37bSArne Jansen 4989905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4990905b0ddaSMiao Xie if (ret) 4991905b0ddaSMiao Xie return ret; 49925d13a37bSArne Jansen 49935d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4994905b0ddaSMiao Xie if (IS_ERR(sa)) { 4995905b0ddaSMiao Xie ret = PTR_ERR(sa); 4996905b0ddaSMiao Xie goto drop_write; 4997905b0ddaSMiao Xie } 49985d13a37bSArne Jansen 49995d13a37bSArne Jansen trans = btrfs_join_transaction(root); 50005d13a37bSArne Jansen if (IS_ERR(trans)) { 50015d13a37bSArne Jansen ret = PTR_ERR(trans); 50025d13a37bSArne Jansen goto out; 50035d13a37bSArne Jansen } 50045d13a37bSArne Jansen 50055d13a37bSArne Jansen qgroupid = sa->qgroupid; 50065d13a37bSArne Jansen if (!qgroupid) { 50075d13a37bSArne Jansen /* take the current subvol as qgroup */ 50085d13a37bSArne Jansen qgroupid = root->root_key.objectid; 50095d13a37bSArne Jansen } 50105d13a37bSArne Jansen 50115d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 50125d13a37bSArne Jansen ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); 50135d13a37bSArne Jansen 50145d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 50155d13a37bSArne Jansen if (err && !ret) 50165d13a37bSArne Jansen ret = err; 50175d13a37bSArne Jansen 50185d13a37bSArne Jansen out: 50195d13a37bSArne Jansen kfree(sa); 5020905b0ddaSMiao Xie drop_write: 5021905b0ddaSMiao Xie mnt_drop_write_file(file); 50225d13a37bSArne Jansen return ret; 50235d13a37bSArne Jansen } 50245d13a37bSArne Jansen 50252f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 50262f232036SJan Schmidt { 50276d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 50282f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50292f232036SJan Schmidt int ret; 50302f232036SJan Schmidt 50312f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50322f232036SJan Schmidt return -EPERM; 50332f232036SJan Schmidt 50342f232036SJan Schmidt ret = mnt_want_write_file(file); 50352f232036SJan Schmidt if (ret) 50362f232036SJan Schmidt return ret; 50372f232036SJan Schmidt 50382f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50392f232036SJan Schmidt if (IS_ERR(qsa)) { 50402f232036SJan Schmidt ret = PTR_ERR(qsa); 50412f232036SJan Schmidt goto drop_write; 50422f232036SJan Schmidt } 50432f232036SJan Schmidt 50442f232036SJan Schmidt if (qsa->flags) { 50452f232036SJan Schmidt ret = -EINVAL; 50462f232036SJan Schmidt goto out; 50472f232036SJan Schmidt } 50482f232036SJan Schmidt 50492f232036SJan Schmidt ret = btrfs_qgroup_rescan(root->fs_info); 50502f232036SJan Schmidt 50512f232036SJan Schmidt out: 50522f232036SJan Schmidt kfree(qsa); 50532f232036SJan Schmidt drop_write: 50542f232036SJan Schmidt mnt_drop_write_file(file); 50552f232036SJan Schmidt return ret; 50562f232036SJan Schmidt } 50572f232036SJan Schmidt 50582f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50592f232036SJan Schmidt { 50606d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 50612f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50622f232036SJan Schmidt int ret = 0; 50632f232036SJan Schmidt 50642f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50652f232036SJan Schmidt return -EPERM; 50662f232036SJan Schmidt 50678d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 50682f232036SJan Schmidt if (!qsa) 50692f232036SJan Schmidt return -ENOMEM; 50702f232036SJan Schmidt 50712f232036SJan Schmidt if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50722f232036SJan Schmidt qsa->flags = 1; 50732f232036SJan Schmidt qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; 50742f232036SJan Schmidt } 50752f232036SJan Schmidt 50762f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50772f232036SJan Schmidt ret = -EFAULT; 50782f232036SJan Schmidt 50792f232036SJan Schmidt kfree(qsa); 50802f232036SJan Schmidt return ret; 50812f232036SJan Schmidt } 50822f232036SJan Schmidt 508357254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 508457254b6eSJan Schmidt { 508554563d41SAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 508657254b6eSJan Schmidt 508757254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 508857254b6eSJan Schmidt return -EPERM; 508957254b6eSJan Schmidt 509057254b6eSJan Schmidt return btrfs_qgroup_wait_for_completion(root->fs_info); 509157254b6eSJan Schmidt } 509257254b6eSJan Schmidt 5093abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5094abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50958ea05e3aSAlexander Block { 5096496ad9aaSAl Viro struct inode *inode = file_inode(file); 50978ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50988ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50998ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 510004b285f3SDeepa Dinamani struct timespec ct = current_fs_time(inode->i_sb); 51018ea05e3aSAlexander Block int ret = 0; 5102dd5f9615SStefan Behrens int received_uuid_changed; 51038ea05e3aSAlexander Block 5104bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5105bd60ea0fSDavid Sterba return -EPERM; 5106bd60ea0fSDavid Sterba 51078ea05e3aSAlexander Block ret = mnt_want_write_file(file); 51088ea05e3aSAlexander Block if (ret < 0) 51098ea05e3aSAlexander Block return ret; 51108ea05e3aSAlexander Block 51118ea05e3aSAlexander Block down_write(&root->fs_info->subvol_sem); 51128ea05e3aSAlexander Block 51138ea05e3aSAlexander Block if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 51148ea05e3aSAlexander Block ret = -EINVAL; 51158ea05e3aSAlexander Block goto out; 51168ea05e3aSAlexander Block } 51178ea05e3aSAlexander Block 51188ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 51198ea05e3aSAlexander Block ret = -EROFS; 51208ea05e3aSAlexander Block goto out; 51218ea05e3aSAlexander Block } 51228ea05e3aSAlexander Block 5123dd5f9615SStefan Behrens /* 5124dd5f9615SStefan Behrens * 1 - root item 5125dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5126dd5f9615SStefan Behrens */ 5127dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51288ea05e3aSAlexander Block if (IS_ERR(trans)) { 51298ea05e3aSAlexander Block ret = PTR_ERR(trans); 51308ea05e3aSAlexander Block trans = NULL; 51318ea05e3aSAlexander Block goto out; 51328ea05e3aSAlexander Block } 51338ea05e3aSAlexander Block 51348ea05e3aSAlexander Block sa->rtransid = trans->transid; 51358ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51368ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51378ea05e3aSAlexander Block 5138dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5139dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5140dd5f9615SStefan Behrens if (received_uuid_changed && 5141dd5f9615SStefan Behrens !btrfs_is_empty_uuid(root_item->received_uuid)) 5142dd5f9615SStefan Behrens btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 5143dd5f9615SStefan Behrens root_item->received_uuid, 5144dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5145dd5f9615SStefan Behrens root->root_key.objectid); 51468ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51478ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51488ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51493cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51503cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51513cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51523cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51538ea05e3aSAlexander Block 51548ea05e3aSAlexander Block ret = btrfs_update_root(trans, root->fs_info->tree_root, 51558ea05e3aSAlexander Block &root->root_key, &root->root_item); 51568ea05e3aSAlexander Block if (ret < 0) { 51578ea05e3aSAlexander Block btrfs_end_transaction(trans, root); 51588ea05e3aSAlexander Block goto out; 5159dd5f9615SStefan Behrens } 5160dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5161dd5f9615SStefan Behrens ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 5162dd5f9615SStefan Behrens sa->uuid, 5163dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5164dd5f9615SStefan Behrens root->root_key.objectid); 5165dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 5166dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 5167dd5f9615SStefan Behrens goto out; 5168dd5f9615SStefan Behrens } 5169dd5f9615SStefan Behrens } 51708ea05e3aSAlexander Block ret = btrfs_commit_transaction(trans, root); 5171dd5f9615SStefan Behrens if (ret < 0) { 5172dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 51738ea05e3aSAlexander Block goto out; 51748ea05e3aSAlexander Block } 51758ea05e3aSAlexander Block 5176abccd00fSHugo Mills out: 5177abccd00fSHugo Mills up_write(&root->fs_info->subvol_sem); 5178abccd00fSHugo Mills mnt_drop_write_file(file); 5179abccd00fSHugo Mills return ret; 5180abccd00fSHugo Mills } 5181abccd00fSHugo Mills 5182abccd00fSHugo Mills #ifdef CONFIG_64BIT 5183abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5184abccd00fSHugo Mills void __user *arg) 5185abccd00fSHugo Mills { 5186abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5187abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5188abccd00fSHugo Mills int ret = 0; 5189abccd00fSHugo Mills 5190abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 5191abccd00fSHugo Mills if (IS_ERR(args32)) { 5192abccd00fSHugo Mills ret = PTR_ERR(args32); 5193abccd00fSHugo Mills args32 = NULL; 5194abccd00fSHugo Mills goto out; 5195abccd00fSHugo Mills } 5196abccd00fSHugo Mills 51978d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 519884dbeb87SDan Carpenter if (!args64) { 519984dbeb87SDan Carpenter ret = -ENOMEM; 5200abccd00fSHugo Mills goto out; 5201abccd00fSHugo Mills } 5202abccd00fSHugo Mills 5203abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5204abccd00fSHugo Mills args64->stransid = args32->stransid; 5205abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5206abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5207abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5208abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5209abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5210abccd00fSHugo Mills args64->flags = args32->flags; 5211abccd00fSHugo Mills 5212abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5213abccd00fSHugo Mills if (ret) 5214abccd00fSHugo Mills goto out; 5215abccd00fSHugo Mills 5216abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5217abccd00fSHugo Mills args32->stransid = args64->stransid; 5218abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5219abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5220abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5221abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5222abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5223abccd00fSHugo Mills args32->flags = args64->flags; 5224abccd00fSHugo Mills 5225abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5226abccd00fSHugo Mills if (ret) 5227abccd00fSHugo Mills ret = -EFAULT; 5228abccd00fSHugo Mills 5229abccd00fSHugo Mills out: 5230abccd00fSHugo Mills kfree(args32); 5231abccd00fSHugo Mills kfree(args64); 5232abccd00fSHugo Mills return ret; 5233abccd00fSHugo Mills } 5234abccd00fSHugo Mills #endif 5235abccd00fSHugo Mills 5236abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5237abccd00fSHugo Mills void __user *arg) 5238abccd00fSHugo Mills { 5239abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5240abccd00fSHugo Mills int ret = 0; 5241abccd00fSHugo Mills 5242abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 5243abccd00fSHugo Mills if (IS_ERR(sa)) { 5244abccd00fSHugo Mills ret = PTR_ERR(sa); 5245abccd00fSHugo Mills sa = NULL; 5246abccd00fSHugo Mills goto out; 5247abccd00fSHugo Mills } 5248abccd00fSHugo Mills 5249abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5250abccd00fSHugo Mills 5251abccd00fSHugo Mills if (ret) 5252abccd00fSHugo Mills goto out; 5253abccd00fSHugo Mills 52548ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52558ea05e3aSAlexander Block if (ret) 52568ea05e3aSAlexander Block ret = -EFAULT; 52578ea05e3aSAlexander Block 52588ea05e3aSAlexander Block out: 52598ea05e3aSAlexander Block kfree(sa); 52608ea05e3aSAlexander Block return ret; 52618ea05e3aSAlexander Block } 52628ea05e3aSAlexander Block 5263867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5264867ab667Sjeff.liu { 52656d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5266a1b83ac5SAnand Jain size_t len; 5267867ab667Sjeff.liu int ret; 5268a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5269a1b83ac5SAnand Jain 5270a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 5271a1b83ac5SAnand Jain memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); 5272a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 5273a1b83ac5SAnand Jain 5274a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5275867ab667Sjeff.liu 5276867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 5277efe120a0SFrank Holton btrfs_warn(root->fs_info, 5278efe120a0SFrank Holton "label is too long, return the first %zu bytes", --len); 5279867ab667Sjeff.liu } 5280867ab667Sjeff.liu 5281867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5282867ab667Sjeff.liu 5283867ab667Sjeff.liu return ret ? -EFAULT : 0; 5284867ab667Sjeff.liu } 5285867ab667Sjeff.liu 5286a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5287a8bfd4abSjeff.liu { 52886d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5289a8bfd4abSjeff.liu struct btrfs_super_block *super_block = root->fs_info->super_copy; 5290a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5291a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5292a8bfd4abSjeff.liu int ret; 5293a8bfd4abSjeff.liu 5294a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5295a8bfd4abSjeff.liu return -EPERM; 5296a8bfd4abSjeff.liu 5297a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5298a8bfd4abSjeff.liu return -EFAULT; 5299a8bfd4abSjeff.liu 5300a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 5301efe120a0SFrank Holton btrfs_err(root->fs_info, "unable to set label with more than %d bytes", 5302a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5303a8bfd4abSjeff.liu return -EINVAL; 5304a8bfd4abSjeff.liu } 5305a8bfd4abSjeff.liu 5306a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5307a8bfd4abSjeff.liu if (ret) 5308a8bfd4abSjeff.liu return ret; 5309a8bfd4abSjeff.liu 5310a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5311a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5312a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5313a8bfd4abSjeff.liu goto out_unlock; 5314a8bfd4abSjeff.liu } 5315a8bfd4abSjeff.liu 5316a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 5317a8bfd4abSjeff.liu strcpy(super_block->label, label); 5318a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 5319d0270acaSJeff Mahoney ret = btrfs_commit_transaction(trans, root); 5320a8bfd4abSjeff.liu 5321a8bfd4abSjeff.liu out_unlock: 5322a8bfd4abSjeff.liu mnt_drop_write_file(file); 5323a8bfd4abSjeff.liu return ret; 5324a8bfd4abSjeff.liu } 5325a8bfd4abSjeff.liu 53262eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 53272eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 53282eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53292eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53302eaa055fSJeff Mahoney 5331d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53322eaa055fSJeff Mahoney { 53334d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53342eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53352eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53362eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53372eaa055fSJeff Mahoney }; 53382eaa055fSJeff Mahoney 53392eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53402eaa055fSJeff Mahoney return -EFAULT; 53412eaa055fSJeff Mahoney 53422eaa055fSJeff Mahoney return 0; 53432eaa055fSJeff Mahoney } 53442eaa055fSJeff Mahoney 53452eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53462eaa055fSJeff Mahoney { 53472eaa055fSJeff Mahoney struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 53482eaa055fSJeff Mahoney struct btrfs_super_block *super_block = root->fs_info->super_copy; 53492eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53502eaa055fSJeff Mahoney 53512eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53522eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53532eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53542eaa055fSJeff Mahoney 53552eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53562eaa055fSJeff Mahoney return -EFAULT; 53572eaa055fSJeff Mahoney 53582eaa055fSJeff Mahoney return 0; 53592eaa055fSJeff Mahoney } 53602eaa055fSJeff Mahoney 53613b02a68aSJeff Mahoney static int check_feature_bits(struct btrfs_root *root, 53623b02a68aSJeff Mahoney enum btrfs_feature_set set, 53632eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53642eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53652eaa055fSJeff Mahoney { 53663b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 53673b02a68aSJeff Mahoney char *names; 53682eaa055fSJeff Mahoney u64 disallowed, unsupported; 53692eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53702eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53712eaa055fSJeff Mahoney 53722eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53732eaa055fSJeff Mahoney if (unsupported) { 53743b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53753b02a68aSJeff Mahoney if (names) { 53763b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 53773b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53783b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53793b02a68aSJeff Mahoney kfree(names); 53803b02a68aSJeff Mahoney } else 53812eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 53822eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53832eaa055fSJeff Mahoney type, unsupported); 53842eaa055fSJeff Mahoney return -EOPNOTSUPP; 53852eaa055fSJeff Mahoney } 53862eaa055fSJeff Mahoney 53872eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53882eaa055fSJeff Mahoney if (disallowed) { 53893b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53903b02a68aSJeff Mahoney if (names) { 53913b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 53923b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53933b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53943b02a68aSJeff Mahoney kfree(names); 53953b02a68aSJeff Mahoney } else 53962eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 53972eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53982eaa055fSJeff Mahoney type, disallowed); 53992eaa055fSJeff Mahoney return -EPERM; 54002eaa055fSJeff Mahoney } 54012eaa055fSJeff Mahoney 54022eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 54032eaa055fSJeff Mahoney if (disallowed) { 54043b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 54053b02a68aSJeff Mahoney if (names) { 54063b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 54073b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 54083b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 54093b02a68aSJeff Mahoney kfree(names); 54103b02a68aSJeff Mahoney } else 54112eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 54122eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 54132eaa055fSJeff Mahoney type, disallowed); 54142eaa055fSJeff Mahoney return -EPERM; 54152eaa055fSJeff Mahoney } 54162eaa055fSJeff Mahoney 54172eaa055fSJeff Mahoney return 0; 54182eaa055fSJeff Mahoney } 54192eaa055fSJeff Mahoney 54202eaa055fSJeff Mahoney #define check_feature(root, change_mask, flags, mask_base) \ 54213b02a68aSJeff Mahoney check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ 54222eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 54232eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 54242eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 54252eaa055fSJeff Mahoney 54262eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 54272eaa055fSJeff Mahoney { 54282eaa055fSJeff Mahoney struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 54292eaa055fSJeff Mahoney struct btrfs_super_block *super_block = root->fs_info->super_copy; 54302eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54312eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54322eaa055fSJeff Mahoney u64 newflags; 54332eaa055fSJeff Mahoney int ret; 54342eaa055fSJeff Mahoney 54352eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54362eaa055fSJeff Mahoney return -EPERM; 54372eaa055fSJeff Mahoney 54382eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54392eaa055fSJeff Mahoney return -EFAULT; 54402eaa055fSJeff Mahoney 54412eaa055fSJeff Mahoney /* Nothing to do */ 54422eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54432eaa055fSJeff Mahoney !flags[0].incompat_flags) 54442eaa055fSJeff Mahoney return 0; 54452eaa055fSJeff Mahoney 54462eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_flags, 54472eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54482eaa055fSJeff Mahoney if (ret) 54492eaa055fSJeff Mahoney return ret; 54502eaa055fSJeff Mahoney 54512eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_ro_flags, 54522eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54532eaa055fSJeff Mahoney if (ret) 54542eaa055fSJeff Mahoney return ret; 54552eaa055fSJeff Mahoney 54562eaa055fSJeff Mahoney ret = check_feature(root, flags[0].incompat_flags, 54572eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54582eaa055fSJeff Mahoney if (ret) 54592eaa055fSJeff Mahoney return ret; 54602eaa055fSJeff Mahoney 54617ab19625SDavid Sterba ret = mnt_want_write_file(file); 54627ab19625SDavid Sterba if (ret) 54637ab19625SDavid Sterba return ret; 54647ab19625SDavid Sterba 54658051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54667ab19625SDavid Sterba if (IS_ERR(trans)) { 54677ab19625SDavid Sterba ret = PTR_ERR(trans); 54687ab19625SDavid Sterba goto out_drop_write; 54697ab19625SDavid Sterba } 54702eaa055fSJeff Mahoney 54712eaa055fSJeff Mahoney spin_lock(&root->fs_info->super_lock); 54722eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54732eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54742eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54752eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54762eaa055fSJeff Mahoney 54772eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54782eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54792eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54802eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54812eaa055fSJeff Mahoney 54822eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54832eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54842eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54852eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54862eaa055fSJeff Mahoney spin_unlock(&root->fs_info->super_lock); 54872eaa055fSJeff Mahoney 54887ab19625SDavid Sterba ret = btrfs_commit_transaction(trans, root); 54897ab19625SDavid Sterba out_drop_write: 54907ab19625SDavid Sterba mnt_drop_write_file(file); 54917ab19625SDavid Sterba 54927ab19625SDavid Sterba return ret; 54932eaa055fSJeff Mahoney } 54942eaa055fSJeff Mahoney 5495f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5496f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5497f46b5a66SChristoph Hellwig { 5498496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 54994bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5500f46b5a66SChristoph Hellwig 5501f46b5a66SChristoph Hellwig switch (cmd) { 55026cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 55036cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 55046cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 55056cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 55066cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 55076cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5508f7039b1dSLi Dongyang case FITRIM: 5509f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5510f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5511fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5512fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5513fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 55143de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5515fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 55166f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 55176f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 551876dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 551976dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 55200caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 55210caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 55220caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 55230caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55246ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55256ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5526f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55271e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55281e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55291e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5530f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5531198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5532f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55334bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 5534f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5535da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55366b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55376b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5538475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 5539475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 5540475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 5541475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 5542f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55439ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5544f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 5545f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 5546f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 5547f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 5548ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5549ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5550cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5551cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5552ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5553ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5554d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5555d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5556d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5557d7728c96SJan Schmidt return btrfs_ioctl_logical_to_ino(root, argp); 55581406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55591406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 55609b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55619b199859SFilipe David Borba Manana int ret; 55629b199859SFilipe David Borba Manana 55636c255e67SMiao Xie ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1); 55649b199859SFilipe David Borba Manana if (ret) 55659b199859SFilipe David Borba Manana return ret; 5566ddb52f4fSAl Viro ret = btrfs_sync_fs(file_inode(file)->i_sb, 1); 55672fad4e83SDavid Sterba /* 55682fad4e83SDavid Sterba * The transaction thread may want to do more work, 556901327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 55702fad4e83SDavid Sterba * processing uncleaned subvols. 55712fad4e83SDavid Sterba */ 55722fad4e83SDavid Sterba wake_up_process(root->fs_info->transaction_kthread); 55739b199859SFilipe David Borba Manana return ret; 55749b199859SFilipe David Borba Manana } 557546204592SSage Weil case BTRFS_IOC_START_SYNC: 55769a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 557746204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55789a8c28beSMiao Xie return btrfs_ioctl_wait_sync(root, argp); 5579475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5580b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5581475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 5582475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 5583475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 5584475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 5585c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55869ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5587837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 5588837d5b6eSIlya Dryomov return btrfs_ioctl_balance_ctl(root, arg); 558919a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 559019a39dceSIlya Dryomov return btrfs_ioctl_balance_progress(root, argp); 55918ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55928ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5593abccd00fSHugo Mills #ifdef CONFIG_64BIT 5594abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5595abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5596abccd00fSHugo Mills #endif 559731db9f7cSAlexander Block case BTRFS_IOC_SEND: 559831db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 5599c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 5600b27f7c0cSDavid Sterba return btrfs_ioctl_get_dev_stats(root, argp); 56015d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5602905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 56035d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5604905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 56055d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5606905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 56075d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5608905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 56092f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 56102f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 56112f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 56122f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 561357254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 561457254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 56153f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 56163f6bcfbdSStefan Behrens return btrfs_ioctl_dev_replace(root, argp); 5617867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5618867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5619a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5620a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 56212eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5622d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 56232eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 56242eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 56252eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56262eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5627f46b5a66SChristoph Hellwig } 5628f46b5a66SChristoph Hellwig 5629f46b5a66SChristoph Hellwig return -ENOTTY; 5630f46b5a66SChristoph Hellwig } 56314c63c245SLuke Dashjr 56324c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56334c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56344c63c245SLuke Dashjr { 56354c63c245SLuke Dashjr switch (cmd) { 56364c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56374c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56384c63c245SLuke Dashjr break; 56394c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56404c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56414c63c245SLuke Dashjr break; 56424c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56434c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56444c63c245SLuke Dashjr break; 56454c63c245SLuke Dashjr default: 56464c63c245SLuke Dashjr return -ENOIOCTLCMD; 56474c63c245SLuke Dashjr } 56484c63c245SLuke Dashjr 56494c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56504c63c245SLuke Dashjr } 56514c63c245SLuke Dashjr #endif 5652