16cbd5570SChris Mason /* 26cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 36cbd5570SChris Mason * 46cbd5570SChris Mason * This program is free software; you can redistribute it and/or 56cbd5570SChris Mason * modify it under the terms of the GNU General Public 66cbd5570SChris Mason * License v2 as published by the Free Software Foundation. 76cbd5570SChris Mason * 86cbd5570SChris Mason * This program is distributed in the hope that it will be useful, 96cbd5570SChris Mason * but WITHOUT ANY WARRANTY; without even the implied warranty of 106cbd5570SChris Mason * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 116cbd5570SChris Mason * General Public License for more details. 126cbd5570SChris Mason * 136cbd5570SChris Mason * You should have received a copy of the GNU General Public 146cbd5570SChris Mason * License along with this program; if not, write to the 156cbd5570SChris Mason * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 166cbd5570SChris Mason * Boston, MA 021110-1307, USA. 176cbd5570SChris Mason */ 186cbd5570SChris Mason 194b82d6e4SYan #include <linux/blkdev.h> 202e635a27SChris Mason #include <linux/module.h> 21e20d96d6SChris Mason #include <linux/buffer_head.h> 222e635a27SChris Mason #include <linux/fs.h> 232e635a27SChris Mason #include <linux/pagemap.h> 242e635a27SChris Mason #include <linux/highmem.h> 252e635a27SChris Mason #include <linux/time.h> 262e635a27SChris Mason #include <linux/init.h> 27a9572a15SEric Paris #include <linux/seq_file.h> 282e635a27SChris Mason #include <linux/string.h> 292e635a27SChris Mason #include <linux/backing-dev.h> 304b82d6e4SYan #include <linux/mount.h> 31dee26a9fSChris Mason #include <linux/mpage.h> 3275dfe396SChris Mason #include <linux/swap.h> 3375dfe396SChris Mason #include <linux/writeback.h> 348fd17795SChris Mason #include <linux/statfs.h> 3508607c1bSChris Mason #include <linux/compat.h> 3695e05289SChris Mason #include <linux/parser.h> 37c59f8951SChris Mason #include <linux/ctype.h> 386da6abaeSChris Mason #include <linux/namei.h> 39a9218f6bSChris Mason #include <linux/miscdevice.h> 401bcbf313SQinghuang Feng #include <linux/magic.h> 415a0e3ad6STejun Heo #include <linux/slab.h> 424b4e25f2SChris Mason #include "compat.h" 432e635a27SChris Mason #include "ctree.h" 44e20d96d6SChris Mason #include "disk-io.h" 45d5719762SChris Mason #include "transaction.h" 462c90e5d6SChris Mason #include "btrfs_inode.h" 47c5739bbaSChris Mason #include "ioctl.h" 483a686375SChris Mason #include "print-tree.h" 495103e947SJosef Bacik #include "xattr.h" 508a4b83ccSChris Mason #include "volumes.h" 51b3c3da71SChris Mason #include "version.h" 52be6e8dc0SBalaji Rao #include "export.h" 53c8b97818SChris Mason #include "compression.h" 542e635a27SChris Mason 55b87221deSAlexey Dobriyan static const struct super_operations btrfs_super_ops; 56e20d96d6SChris Mason 57e20d96d6SChris Mason static void btrfs_put_super(struct super_block *sb) 58e20d96d6SChris Mason { 59e20d96d6SChris Mason struct btrfs_root *root = btrfs_sb(sb); 60e20d96d6SChris Mason int ret; 61e20d96d6SChris Mason 62e20d96d6SChris Mason ret = close_ctree(root); 63e20d96d6SChris Mason sb->s_fs_info = NULL; 64559af821SAndi Kleen 65559af821SAndi Kleen (void)ret; /* FIXME: need to fix VFS to return error? */ 66e20d96d6SChris Mason } 672e635a27SChris Mason 6895e05289SChris Mason enum { 6973f73415SJosef Bacik Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum, 70287a0ab9SJosef Bacik Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd, 71287a0ab9SJosef Bacik Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress, 72261507a0SLi Zefan Opt_compress_type, Opt_compress_force, Opt_compress_force_type, 73261507a0SLi Zefan Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard, 74261507a0SLi Zefan Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, Opt_err, 7595e05289SChris Mason }; 7695e05289SChris Mason 7795e05289SChris Mason static match_table_t tokens = { 78dfe25020SChris Mason {Opt_degraded, "degraded"}, 7995e05289SChris Mason {Opt_subvol, "subvol=%s"}, 8073f73415SJosef Bacik {Opt_subvolid, "subvolid=%d"}, 8143e570b0SChristoph Hellwig {Opt_device, "device=%s"}, 82b6cda9bcSChris Mason {Opt_nodatasum, "nodatasum"}, 83be20aa9dSChris Mason {Opt_nodatacow, "nodatacow"}, 8421ad10cfSChris Mason {Opt_nobarrier, "nobarrier"}, 856f568d35SChris Mason {Opt_max_inline, "max_inline=%s"}, 868f662a76SChris Mason {Opt_alloc_start, "alloc_start=%s"}, 874543df7eSChris Mason {Opt_thread_pool, "thread_pool=%d"}, 88c8b97818SChris Mason {Opt_compress, "compress"}, 89261507a0SLi Zefan {Opt_compress_type, "compress=%s"}, 90a555f810SChris Mason {Opt_compress_force, "compress-force"}, 91261507a0SLi Zefan {Opt_compress_force_type, "compress-force=%s"}, 92e18e4809SChris Mason {Opt_ssd, "ssd"}, 93451d7585SChris Mason {Opt_ssd_spread, "ssd_spread"}, 943b30c22fSChris Mason {Opt_nossd, "nossd"}, 9533268eafSJosef Bacik {Opt_noacl, "noacl"}, 963a5e1404SSage Weil {Opt_notreelog, "notreelog"}, 97dccae999SSage Weil {Opt_flushoncommit, "flushoncommit"}, 9897e728d4SJosef Bacik {Opt_ratio, "metadata_ratio=%d"}, 99e244a0aeSChristoph Hellwig {Opt_discard, "discard"}, 1000af3d00bSJosef Bacik {Opt_space_cache, "space_cache"}, 10188c2ba3bSJosef Bacik {Opt_clear_cache, "clear_cache"}, 1024260f7c7SSage Weil {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"}, 10333268eafSJosef Bacik {Opt_err, NULL}, 10495e05289SChris Mason }; 10595e05289SChris Mason 106edf24abeSChristoph Hellwig /* 107edf24abeSChristoph Hellwig * Regular mount options parser. Everything that is needed only when 108edf24abeSChristoph Hellwig * reading in a new superblock is parsed here. 109edf24abeSChristoph Hellwig */ 110edf24abeSChristoph Hellwig int btrfs_parse_options(struct btrfs_root *root, char *options) 11195e05289SChris Mason { 112edf24abeSChristoph Hellwig struct btrfs_fs_info *info = root->fs_info; 11395e05289SChris Mason substring_t args[MAX_OPT_ARGS]; 114da495eccSJosef Bacik char *p, *num, *orig; 1154543df7eSChris Mason int intarg; 116a7a3f7caSSage Weil int ret = 0; 117261507a0SLi Zefan char *compress_type; 118261507a0SLi Zefan bool compress_force = false; 119b6cda9bcSChris Mason 12095e05289SChris Mason if (!options) 121edf24abeSChristoph Hellwig return 0; 12295e05289SChris Mason 123be20aa9dSChris Mason /* 124be20aa9dSChris Mason * strsep changes the string, duplicate it because parse_options 125be20aa9dSChris Mason * gets called twice 126be20aa9dSChris Mason */ 127be20aa9dSChris Mason options = kstrdup(options, GFP_NOFS); 128be20aa9dSChris Mason if (!options) 129be20aa9dSChris Mason return -ENOMEM; 130be20aa9dSChris Mason 131da495eccSJosef Bacik orig = options; 132be20aa9dSChris Mason 13395e05289SChris Mason while ((p = strsep(&options, ",")) != NULL) { 13495e05289SChris Mason int token; 13595e05289SChris Mason if (!*p) 13695e05289SChris Mason continue; 13795e05289SChris Mason 13895e05289SChris Mason token = match_token(p, tokens, args); 13995e05289SChris Mason switch (token) { 140dfe25020SChris Mason case Opt_degraded: 141edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: allowing degraded mounts\n"); 142dfe25020SChris Mason btrfs_set_opt(info->mount_opt, DEGRADED); 143dfe25020SChris Mason break; 14495e05289SChris Mason case Opt_subvol: 14573f73415SJosef Bacik case Opt_subvolid: 14643e570b0SChristoph Hellwig case Opt_device: 147edf24abeSChristoph Hellwig /* 14843e570b0SChristoph Hellwig * These are parsed by btrfs_parse_early_options 149edf24abeSChristoph Hellwig * and can be happily ignored here. 150edf24abeSChristoph Hellwig */ 15195e05289SChris Mason break; 152b6cda9bcSChris Mason case Opt_nodatasum: 153067c28adSChris Mason printk(KERN_INFO "btrfs: setting nodatasum\n"); 154b6cda9bcSChris Mason btrfs_set_opt(info->mount_opt, NODATASUM); 155be20aa9dSChris Mason break; 156be20aa9dSChris Mason case Opt_nodatacow: 157edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: setting nodatacow\n"); 158be20aa9dSChris Mason btrfs_set_opt(info->mount_opt, NODATACOW); 159be20aa9dSChris Mason btrfs_set_opt(info->mount_opt, NODATASUM); 160b6cda9bcSChris Mason break; 161a555f810SChris Mason case Opt_compress_force: 162261507a0SLi Zefan case Opt_compress_force_type: 163261507a0SLi Zefan compress_force = true; 164261507a0SLi Zefan case Opt_compress: 165261507a0SLi Zefan case Opt_compress_type: 166261507a0SLi Zefan if (token == Opt_compress || 167261507a0SLi Zefan token == Opt_compress_force || 168261507a0SLi Zefan strcmp(args[0].from, "zlib") == 0) { 169261507a0SLi Zefan compress_type = "zlib"; 170261507a0SLi Zefan info->compress_type = BTRFS_COMPRESS_ZLIB; 171a6fa6faeSLi Zefan } else if (strcmp(args[0].from, "lzo") == 0) { 172a6fa6faeSLi Zefan compress_type = "lzo"; 173a6fa6faeSLi Zefan info->compress_type = BTRFS_COMPRESS_LZO; 174261507a0SLi Zefan } else { 175261507a0SLi Zefan ret = -EINVAL; 176261507a0SLi Zefan goto out; 177261507a0SLi Zefan } 178261507a0SLi Zefan 179a555f810SChris Mason btrfs_set_opt(info->mount_opt, COMPRESS); 180261507a0SLi Zefan if (compress_force) { 181261507a0SLi Zefan btrfs_set_opt(info->mount_opt, FORCE_COMPRESS); 182261507a0SLi Zefan pr_info("btrfs: force %s compression\n", 183261507a0SLi Zefan compress_type); 184261507a0SLi Zefan } else 185261507a0SLi Zefan pr_info("btrfs: use %s compression\n", 186261507a0SLi Zefan compress_type); 187a555f810SChris Mason break; 188e18e4809SChris Mason case Opt_ssd: 189edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: use ssd allocation scheme\n"); 190e18e4809SChris Mason btrfs_set_opt(info->mount_opt, SSD); 191e18e4809SChris Mason break; 192451d7585SChris Mason case Opt_ssd_spread: 193451d7585SChris Mason printk(KERN_INFO "btrfs: use spread ssd " 194451d7585SChris Mason "allocation scheme\n"); 195451d7585SChris Mason btrfs_set_opt(info->mount_opt, SSD); 196451d7585SChris Mason btrfs_set_opt(info->mount_opt, SSD_SPREAD); 197451d7585SChris Mason break; 1983b30c22fSChris Mason case Opt_nossd: 199451d7585SChris Mason printk(KERN_INFO "btrfs: not using ssd allocation " 200451d7585SChris Mason "scheme\n"); 201c289811cSChris Mason btrfs_set_opt(info->mount_opt, NOSSD); 2023b30c22fSChris Mason btrfs_clear_opt(info->mount_opt, SSD); 203451d7585SChris Mason btrfs_clear_opt(info->mount_opt, SSD_SPREAD); 2043b30c22fSChris Mason break; 20521ad10cfSChris Mason case Opt_nobarrier: 206edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: turning off barriers\n"); 20721ad10cfSChris Mason btrfs_set_opt(info->mount_opt, NOBARRIER); 20821ad10cfSChris Mason break; 2094543df7eSChris Mason case Opt_thread_pool: 2104543df7eSChris Mason intarg = 0; 2114543df7eSChris Mason match_int(&args[0], &intarg); 2124543df7eSChris Mason if (intarg) { 2134543df7eSChris Mason info->thread_pool_size = intarg; 2144543df7eSChris Mason printk(KERN_INFO "btrfs: thread pool %d\n", 2154543df7eSChris Mason info->thread_pool_size); 2164543df7eSChris Mason } 2174543df7eSChris Mason break; 2186f568d35SChris Mason case Opt_max_inline: 219edf24abeSChristoph Hellwig num = match_strdup(&args[0]); 2206f568d35SChris Mason if (num) { 22191748467SAkinobu Mita info->max_inline = memparse(num, NULL); 2226f568d35SChris Mason kfree(num); 2236f568d35SChris Mason 22415ada040SChris Mason if (info->max_inline) { 2256f568d35SChris Mason info->max_inline = max_t(u64, 22615ada040SChris Mason info->max_inline, 22715ada040SChris Mason root->sectorsize); 22815ada040SChris Mason } 229edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: max_inline at %llu\n", 23021380931SJoel Becker (unsigned long long)info->max_inline); 2316f568d35SChris Mason } 2326f568d35SChris Mason break; 2338f662a76SChris Mason case Opt_alloc_start: 234edf24abeSChristoph Hellwig num = match_strdup(&args[0]); 2358f662a76SChris Mason if (num) { 23691748467SAkinobu Mita info->alloc_start = memparse(num, NULL); 2378f662a76SChris Mason kfree(num); 238edf24abeSChristoph Hellwig printk(KERN_INFO 239edf24abeSChristoph Hellwig "btrfs: allocations start at %llu\n", 24021380931SJoel Becker (unsigned long long)info->alloc_start); 2418f662a76SChris Mason } 2428f662a76SChris Mason break; 24333268eafSJosef Bacik case Opt_noacl: 24433268eafSJosef Bacik root->fs_info->sb->s_flags &= ~MS_POSIXACL; 24533268eafSJosef Bacik break; 2463a5e1404SSage Weil case Opt_notreelog: 2473a5e1404SSage Weil printk(KERN_INFO "btrfs: disabling tree log\n"); 2483a5e1404SSage Weil btrfs_set_opt(info->mount_opt, NOTREELOG); 2493a5e1404SSage Weil break; 250dccae999SSage Weil case Opt_flushoncommit: 251dccae999SSage Weil printk(KERN_INFO "btrfs: turning on flush-on-commit\n"); 252dccae999SSage Weil btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT); 253dccae999SSage Weil break; 25497e728d4SJosef Bacik case Opt_ratio: 25597e728d4SJosef Bacik intarg = 0; 25697e728d4SJosef Bacik match_int(&args[0], &intarg); 25797e728d4SJosef Bacik if (intarg) { 25897e728d4SJosef Bacik info->metadata_ratio = intarg; 25997e728d4SJosef Bacik printk(KERN_INFO "btrfs: metadata ratio %d\n", 26097e728d4SJosef Bacik info->metadata_ratio); 26197e728d4SJosef Bacik } 26297e728d4SJosef Bacik break; 263e244a0aeSChristoph Hellwig case Opt_discard: 264e244a0aeSChristoph Hellwig btrfs_set_opt(info->mount_opt, DISCARD); 265e244a0aeSChristoph Hellwig break; 2660af3d00bSJosef Bacik case Opt_space_cache: 2670af3d00bSJosef Bacik printk(KERN_INFO "btrfs: enabling disk space caching\n"); 2680af3d00bSJosef Bacik btrfs_set_opt(info->mount_opt, SPACE_CACHE); 2690de90876SJosef Bacik break; 27088c2ba3bSJosef Bacik case Opt_clear_cache: 27188c2ba3bSJosef Bacik printk(KERN_INFO "btrfs: force clearing of disk cache\n"); 27288c2ba3bSJosef Bacik btrfs_set_opt(info->mount_opt, CLEAR_CACHE); 2730af3d00bSJosef Bacik break; 2744260f7c7SSage Weil case Opt_user_subvol_rm_allowed: 2754260f7c7SSage Weil btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED); 2764260f7c7SSage Weil break; 277a7a3f7caSSage Weil case Opt_err: 278a7a3f7caSSage Weil printk(KERN_INFO "btrfs: unrecognized mount option " 279a7a3f7caSSage Weil "'%s'\n", p); 280a7a3f7caSSage Weil ret = -EINVAL; 281a7a3f7caSSage Weil goto out; 28295e05289SChris Mason default: 283be20aa9dSChris Mason break; 28495e05289SChris Mason } 28595e05289SChris Mason } 286a7a3f7caSSage Weil out: 287da495eccSJosef Bacik kfree(orig); 288a7a3f7caSSage Weil return ret; 289edf24abeSChristoph Hellwig } 290edf24abeSChristoph Hellwig 291edf24abeSChristoph Hellwig /* 292edf24abeSChristoph Hellwig * Parse mount options that are required early in the mount process. 293edf24abeSChristoph Hellwig * 294edf24abeSChristoph Hellwig * All other options will be parsed on much later in the mount process and 295edf24abeSChristoph Hellwig * only when we need to allocate a new super block. 296edf24abeSChristoph Hellwig */ 29797288f2cSChristoph Hellwig static int btrfs_parse_early_options(const char *options, fmode_t flags, 29873f73415SJosef Bacik void *holder, char **subvol_name, u64 *subvol_objectid, 29943e570b0SChristoph Hellwig struct btrfs_fs_devices **fs_devices) 300edf24abeSChristoph Hellwig { 301edf24abeSChristoph Hellwig substring_t args[MAX_OPT_ARGS]; 302edf24abeSChristoph Hellwig char *opts, *p; 303edf24abeSChristoph Hellwig int error = 0; 30473f73415SJosef Bacik int intarg; 305edf24abeSChristoph Hellwig 306edf24abeSChristoph Hellwig if (!options) 307edf24abeSChristoph Hellwig goto out; 308edf24abeSChristoph Hellwig 309edf24abeSChristoph Hellwig /* 310edf24abeSChristoph Hellwig * strsep changes the string, duplicate it because parse_options 311edf24abeSChristoph Hellwig * gets called twice 312edf24abeSChristoph Hellwig */ 313edf24abeSChristoph Hellwig opts = kstrdup(options, GFP_KERNEL); 314edf24abeSChristoph Hellwig if (!opts) 315edf24abeSChristoph Hellwig return -ENOMEM; 316edf24abeSChristoph Hellwig 317edf24abeSChristoph Hellwig while ((p = strsep(&opts, ",")) != NULL) { 318edf24abeSChristoph Hellwig int token; 319edf24abeSChristoph Hellwig if (!*p) 320edf24abeSChristoph Hellwig continue; 321edf24abeSChristoph Hellwig 322edf24abeSChristoph Hellwig token = match_token(p, tokens, args); 323edf24abeSChristoph Hellwig switch (token) { 324edf24abeSChristoph Hellwig case Opt_subvol: 325edf24abeSChristoph Hellwig *subvol_name = match_strdup(&args[0]); 326edf24abeSChristoph Hellwig break; 32773f73415SJosef Bacik case Opt_subvolid: 32873f73415SJosef Bacik intarg = 0; 3294849f01dSJosef Bacik error = match_int(&args[0], &intarg); 3304849f01dSJosef Bacik if (!error) { 3314849f01dSJosef Bacik /* we want the original fs_tree */ 3324849f01dSJosef Bacik if (!intarg) 3334849f01dSJosef Bacik *subvol_objectid = 3344849f01dSJosef Bacik BTRFS_FS_TREE_OBJECTID; 3354849f01dSJosef Bacik else 33673f73415SJosef Bacik *subvol_objectid = intarg; 3374849f01dSJosef Bacik } 33873f73415SJosef Bacik break; 33943e570b0SChristoph Hellwig case Opt_device: 34043e570b0SChristoph Hellwig error = btrfs_scan_one_device(match_strdup(&args[0]), 34143e570b0SChristoph Hellwig flags, holder, fs_devices); 34243e570b0SChristoph Hellwig if (error) 34343e570b0SChristoph Hellwig goto out_free_opts; 34443e570b0SChristoph Hellwig break; 345edf24abeSChristoph Hellwig default: 346edf24abeSChristoph Hellwig break; 347edf24abeSChristoph Hellwig } 348edf24abeSChristoph Hellwig } 349edf24abeSChristoph Hellwig 35043e570b0SChristoph Hellwig out_free_opts: 351edf24abeSChristoph Hellwig kfree(opts); 352edf24abeSChristoph Hellwig out: 353edf24abeSChristoph Hellwig /* 354edf24abeSChristoph Hellwig * If no subvolume name is specified we use the default one. Allocate 3553de4586cSChris Mason * a copy of the string "." here so that code later in the 356edf24abeSChristoph Hellwig * mount path doesn't care if it's the default volume or another one. 357edf24abeSChristoph Hellwig */ 358edf24abeSChristoph Hellwig if (!*subvol_name) { 3593de4586cSChris Mason *subvol_name = kstrdup(".", GFP_KERNEL); 360edf24abeSChristoph Hellwig if (!*subvol_name) 361edf24abeSChristoph Hellwig return -ENOMEM; 362edf24abeSChristoph Hellwig } 363edf24abeSChristoph Hellwig return error; 36495e05289SChris Mason } 36595e05289SChris Mason 36673f73415SJosef Bacik static struct dentry *get_default_root(struct super_block *sb, 36773f73415SJosef Bacik u64 subvol_objectid) 36873f73415SJosef Bacik { 36973f73415SJosef Bacik struct btrfs_root *root = sb->s_fs_info; 37073f73415SJosef Bacik struct btrfs_root *new_root; 37173f73415SJosef Bacik struct btrfs_dir_item *di; 37273f73415SJosef Bacik struct btrfs_path *path; 37373f73415SJosef Bacik struct btrfs_key location; 37473f73415SJosef Bacik struct inode *inode; 37573f73415SJosef Bacik struct dentry *dentry; 37673f73415SJosef Bacik u64 dir_id; 37773f73415SJosef Bacik int new = 0; 37873f73415SJosef Bacik 37973f73415SJosef Bacik /* 38073f73415SJosef Bacik * We have a specific subvol we want to mount, just setup location and 38173f73415SJosef Bacik * go look up the root. 38273f73415SJosef Bacik */ 38373f73415SJosef Bacik if (subvol_objectid) { 38473f73415SJosef Bacik location.objectid = subvol_objectid; 38573f73415SJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 38673f73415SJosef Bacik location.offset = (u64)-1; 38773f73415SJosef Bacik goto find_root; 38873f73415SJosef Bacik } 38973f73415SJosef Bacik 39073f73415SJosef Bacik path = btrfs_alloc_path(); 39173f73415SJosef Bacik if (!path) 39273f73415SJosef Bacik return ERR_PTR(-ENOMEM); 39373f73415SJosef Bacik path->leave_spinning = 1; 39473f73415SJosef Bacik 39573f73415SJosef Bacik /* 39673f73415SJosef Bacik * Find the "default" dir item which points to the root item that we 39773f73415SJosef Bacik * will mount by default if we haven't been given a specific subvolume 39873f73415SJosef Bacik * to mount. 39973f73415SJosef Bacik */ 40073f73415SJosef Bacik dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); 40173f73415SJosef Bacik di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0); 402fb4f6f91SDan Carpenter if (IS_ERR(di)) 403fb4f6f91SDan Carpenter return ERR_CAST(di); 40473f73415SJosef Bacik if (!di) { 40573f73415SJosef Bacik /* 40673f73415SJosef Bacik * Ok the default dir item isn't there. This is weird since 40773f73415SJosef Bacik * it's always been there, but don't freak out, just try and 40873f73415SJosef Bacik * mount to root most subvolume. 40973f73415SJosef Bacik */ 41073f73415SJosef Bacik btrfs_free_path(path); 41173f73415SJosef Bacik dir_id = BTRFS_FIRST_FREE_OBJECTID; 41273f73415SJosef Bacik new_root = root->fs_info->fs_root; 41373f73415SJosef Bacik goto setup_root; 41473f73415SJosef Bacik } 41573f73415SJosef Bacik 41673f73415SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); 41773f73415SJosef Bacik btrfs_free_path(path); 41873f73415SJosef Bacik 41973f73415SJosef Bacik find_root: 42073f73415SJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 42173f73415SJosef Bacik if (IS_ERR(new_root)) 422d0b678cbSJulia Lawall return ERR_CAST(new_root); 42373f73415SJosef Bacik 42473f73415SJosef Bacik if (btrfs_root_refs(&new_root->root_item) == 0) 42573f73415SJosef Bacik return ERR_PTR(-ENOENT); 42673f73415SJosef Bacik 42773f73415SJosef Bacik dir_id = btrfs_root_dirid(&new_root->root_item); 42873f73415SJosef Bacik setup_root: 42973f73415SJosef Bacik location.objectid = dir_id; 43073f73415SJosef Bacik location.type = BTRFS_INODE_ITEM_KEY; 43173f73415SJosef Bacik location.offset = 0; 43273f73415SJosef Bacik 43373f73415SJosef Bacik inode = btrfs_iget(sb, &location, new_root, &new); 4344cbd1149SDan Carpenter if (IS_ERR(inode)) 4354cbd1149SDan Carpenter return ERR_CAST(inode); 43673f73415SJosef Bacik 43773f73415SJosef Bacik /* 43873f73415SJosef Bacik * If we're just mounting the root most subvol put the inode and return 43973f73415SJosef Bacik * a reference to the dentry. We will have already gotten a reference 44073f73415SJosef Bacik * to the inode in btrfs_fill_super so we're good to go. 44173f73415SJosef Bacik */ 44273f73415SJosef Bacik if (!new && sb->s_root->d_inode == inode) { 44373f73415SJosef Bacik iput(inode); 44473f73415SJosef Bacik return dget(sb->s_root); 44573f73415SJosef Bacik } 44673f73415SJosef Bacik 44773f73415SJosef Bacik if (new) { 44873f73415SJosef Bacik const struct qstr name = { .name = "/", .len = 1 }; 44973f73415SJosef Bacik 45073f73415SJosef Bacik /* 45173f73415SJosef Bacik * New inode, we need to make the dentry a sibling of s_root so 45273f73415SJosef Bacik * everything gets cleaned up properly on unmount. 45373f73415SJosef Bacik */ 45473f73415SJosef Bacik dentry = d_alloc(sb->s_root, &name); 45573f73415SJosef Bacik if (!dentry) { 45673f73415SJosef Bacik iput(inode); 45773f73415SJosef Bacik return ERR_PTR(-ENOMEM); 45873f73415SJosef Bacik } 45973f73415SJosef Bacik d_splice_alias(inode, dentry); 46073f73415SJosef Bacik } else { 46173f73415SJosef Bacik /* 46273f73415SJosef Bacik * We found the inode in cache, just find a dentry for it and 46373f73415SJosef Bacik * put the reference to the inode we just got. 46473f73415SJosef Bacik */ 46573f73415SJosef Bacik dentry = d_find_alias(inode); 46673f73415SJosef Bacik iput(inode); 46773f73415SJosef Bacik } 46873f73415SJosef Bacik 46973f73415SJosef Bacik return dentry; 47073f73415SJosef Bacik } 47173f73415SJosef Bacik 4728a4b83ccSChris Mason static int btrfs_fill_super(struct super_block *sb, 4738a4b83ccSChris Mason struct btrfs_fs_devices *fs_devices, 4748a4b83ccSChris Mason void *data, int silent) 4752e635a27SChris Mason { 4762e635a27SChris Mason struct inode *inode; 477e20d96d6SChris Mason struct dentry *root_dentry; 4780f7d52f4SChris Mason struct btrfs_root *tree_root; 4795d4f98a2SYan Zheng struct btrfs_key key; 48039279cc3SChris Mason int err; 4812e635a27SChris Mason 4822e635a27SChris Mason sb->s_maxbytes = MAX_LFS_FILESIZE; 4832e635a27SChris Mason sb->s_magic = BTRFS_SUPER_MAGIC; 484e20d96d6SChris Mason sb->s_op = &btrfs_super_ops; 485be6e8dc0SBalaji Rao sb->s_export_op = &btrfs_export_ops; 4865103e947SJosef Bacik sb->s_xattr = btrfs_xattr_handlers; 4872e635a27SChris Mason sb->s_time_gran = 1; 4880eda294dSChris Mason #ifdef CONFIG_BTRFS_FS_POSIX_ACL 48933268eafSJosef Bacik sb->s_flags |= MS_POSIXACL; 49049cf6f45SChris Ball #endif 491e20d96d6SChris Mason 492dfe25020SChris Mason tree_root = open_ctree(sb, fs_devices, (char *)data); 493d98237b3SChris Mason 494e58ca020SYan if (IS_ERR(tree_root)) { 495e20d96d6SChris Mason printk("btrfs: open_ctree failed\n"); 496e58ca020SYan return PTR_ERR(tree_root); 497e20d96d6SChris Mason } 4980f7d52f4SChris Mason sb->s_fs_info = tree_root; 499b888db2bSChris Mason 5005d4f98a2SYan Zheng key.objectid = BTRFS_FIRST_FREE_OBJECTID; 5015d4f98a2SYan Zheng key.type = BTRFS_INODE_ITEM_KEY; 5025d4f98a2SYan Zheng key.offset = 0; 50373f73415SJosef Bacik inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL); 5045d4f98a2SYan Zheng if (IS_ERR(inode)) { 5055d4f98a2SYan Zheng err = PTR_ERR(inode); 50639279cc3SChris Mason goto fail_close; 50739279cc3SChris Mason } 5082e635a27SChris Mason 509e20d96d6SChris Mason root_dentry = d_alloc_root(inode); 510e20d96d6SChris Mason if (!root_dentry) { 5112e635a27SChris Mason iput(inode); 51239279cc3SChris Mason err = -ENOMEM; 51339279cc3SChris Mason goto fail_close; 5142e635a27SChris Mason } 51558176a96SJosef Bacik 516e20d96d6SChris Mason sb->s_root = root_dentry; 5176885f308SChris Mason 5186885f308SChris Mason save_mount_options(sb, data); 5192e635a27SChris Mason return 0; 5202e635a27SChris Mason 52139279cc3SChris Mason fail_close: 52239279cc3SChris Mason close_ctree(tree_root); 523d5719762SChris Mason return err; 524d5719762SChris Mason } 525d5719762SChris Mason 5266bf13c0cSSage Weil int btrfs_sync_fs(struct super_block *sb, int wait) 527d5719762SChris Mason { 528d5719762SChris Mason struct btrfs_trans_handle *trans; 529dccae999SSage Weil struct btrfs_root *root = btrfs_sb(sb); 530d5719762SChris Mason int ret; 531df2ce34cSChris Mason 532d561c025SChris Mason if (!wait) { 5337cfcc17eSChris Mason filemap_flush(root->fs_info->btree_inode->i_mapping); 534df2ce34cSChris Mason return 0; 535d561c025SChris Mason } 536771ed689SChris Mason 53724bbcf04SYan, Zheng btrfs_start_delalloc_inodes(root, 0); 53824bbcf04SYan, Zheng btrfs_wait_ordered_extents(root, 0, 0); 539771ed689SChris Mason 540a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 541d5719762SChris Mason ret = btrfs_commit_transaction(trans, root); 54254aa1f4dSChris Mason return ret; 543d5719762SChris Mason } 544d5719762SChris Mason 545a9572a15SEric Paris static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs) 546a9572a15SEric Paris { 547a9572a15SEric Paris struct btrfs_root *root = btrfs_sb(vfs->mnt_sb); 548a9572a15SEric Paris struct btrfs_fs_info *info = root->fs_info; 549a9572a15SEric Paris 550a9572a15SEric Paris if (btrfs_test_opt(root, DEGRADED)) 551a9572a15SEric Paris seq_puts(seq, ",degraded"); 552a9572a15SEric Paris if (btrfs_test_opt(root, NODATASUM)) 553a9572a15SEric Paris seq_puts(seq, ",nodatasum"); 554a9572a15SEric Paris if (btrfs_test_opt(root, NODATACOW)) 555a9572a15SEric Paris seq_puts(seq, ",nodatacow"); 556a9572a15SEric Paris if (btrfs_test_opt(root, NOBARRIER)) 557a9572a15SEric Paris seq_puts(seq, ",nobarrier"); 558a9572a15SEric Paris if (info->max_inline != 8192 * 1024) 55921380931SJoel Becker seq_printf(seq, ",max_inline=%llu", 56021380931SJoel Becker (unsigned long long)info->max_inline); 561a9572a15SEric Paris if (info->alloc_start != 0) 56221380931SJoel Becker seq_printf(seq, ",alloc_start=%llu", 56321380931SJoel Becker (unsigned long long)info->alloc_start); 564a9572a15SEric Paris if (info->thread_pool_size != min_t(unsigned long, 565a9572a15SEric Paris num_online_cpus() + 2, 8)) 566a9572a15SEric Paris seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); 567a9572a15SEric Paris if (btrfs_test_opt(root, COMPRESS)) 568a9572a15SEric Paris seq_puts(seq, ",compress"); 569c289811cSChris Mason if (btrfs_test_opt(root, NOSSD)) 570c289811cSChris Mason seq_puts(seq, ",nossd"); 571451d7585SChris Mason if (btrfs_test_opt(root, SSD_SPREAD)) 572451d7585SChris Mason seq_puts(seq, ",ssd_spread"); 573451d7585SChris Mason else if (btrfs_test_opt(root, SSD)) 574a9572a15SEric Paris seq_puts(seq, ",ssd"); 5753a5e1404SSage Weil if (btrfs_test_opt(root, NOTREELOG)) 5766b65c5c6SSage Weil seq_puts(seq, ",notreelog"); 577dccae999SSage Weil if (btrfs_test_opt(root, FLUSHONCOMMIT)) 5786b65c5c6SSage Weil seq_puts(seq, ",flushoncommit"); 57920a5239aSMatthew Wilcox if (btrfs_test_opt(root, DISCARD)) 58020a5239aSMatthew Wilcox seq_puts(seq, ",discard"); 581a9572a15SEric Paris if (!(root->fs_info->sb->s_flags & MS_POSIXACL)) 582a9572a15SEric Paris seq_puts(seq, ",noacl"); 583a9572a15SEric Paris return 0; 584a9572a15SEric Paris } 585a9572a15SEric Paris 586a061fc8dSChris Mason static int btrfs_test_super(struct super_block *s, void *data) 5872e635a27SChris Mason { 588450ba0eaSJosef Bacik struct btrfs_root *test_root = data; 589a061fc8dSChris Mason struct btrfs_root *root = btrfs_sb(s); 5904b82d6e4SYan 591619c8c76SIan Kent /* 592619c8c76SIan Kent * If this super block is going away, return false as it 593619c8c76SIan Kent * can't match as an existing super block. 594619c8c76SIan Kent */ 595619c8c76SIan Kent if (!atomic_read(&s->s_active)) 596619c8c76SIan Kent return 0; 597450ba0eaSJosef Bacik return root->fs_info->fs_devices == test_root->fs_info->fs_devices; 5984b82d6e4SYan } 5994b82d6e4SYan 600450ba0eaSJosef Bacik static int btrfs_set_super(struct super_block *s, void *data) 601450ba0eaSJosef Bacik { 602450ba0eaSJosef Bacik s->s_fs_info = data; 603450ba0eaSJosef Bacik 604450ba0eaSJosef Bacik return set_anon_super(s, data); 605450ba0eaSJosef Bacik } 606450ba0eaSJosef Bacik 607450ba0eaSJosef Bacik 608edf24abeSChristoph Hellwig /* 609edf24abeSChristoph Hellwig * Find a superblock for the given device / mount point. 610edf24abeSChristoph Hellwig * 611edf24abeSChristoph Hellwig * Note: This is based on get_sb_bdev from fs/super.c with a few additions 612edf24abeSChristoph Hellwig * for multiple device setup. Make sure to keep it in sync. 613edf24abeSChristoph Hellwig */ 614edf24abeSChristoph Hellwig static int btrfs_get_sb(struct file_system_type *fs_type, int flags, 615edf24abeSChristoph Hellwig const char *dev_name, void *data, struct vfsmount *mnt) 6164b82d6e4SYan { 6174b82d6e4SYan struct block_device *bdev = NULL; 6184b82d6e4SYan struct super_block *s; 6194b82d6e4SYan struct dentry *root; 6208a4b83ccSChris Mason struct btrfs_fs_devices *fs_devices = NULL; 621450ba0eaSJosef Bacik struct btrfs_root *tree_root = NULL; 622450ba0eaSJosef Bacik struct btrfs_fs_info *fs_info = NULL; 62397288f2cSChristoph Hellwig fmode_t mode = FMODE_READ; 62473f73415SJosef Bacik char *subvol_name = NULL; 62573f73415SJosef Bacik u64 subvol_objectid = 0; 6264b82d6e4SYan int error = 0; 6274b82d6e4SYan 62897288f2cSChristoph Hellwig if (!(flags & MS_RDONLY)) 62997288f2cSChristoph Hellwig mode |= FMODE_WRITE; 63097288f2cSChristoph Hellwig 63197288f2cSChristoph Hellwig error = btrfs_parse_early_options(data, mode, fs_type, 63273f73415SJosef Bacik &subvol_name, &subvol_objectid, 63373f73415SJosef Bacik &fs_devices); 634edf24abeSChristoph Hellwig if (error) 6351f483660SShen Feng return error; 636edf24abeSChristoph Hellwig 63797288f2cSChristoph Hellwig error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices); 6388a4b83ccSChris Mason if (error) 639edf24abeSChristoph Hellwig goto error_free_subvol_name; 6404b82d6e4SYan 64197288f2cSChristoph Hellwig error = btrfs_open_devices(fs_devices, mode, fs_type); 6428a4b83ccSChris Mason if (error) 643edf24abeSChristoph Hellwig goto error_free_subvol_name; 6448a4b83ccSChris Mason 6452b82032cSYan Zheng if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) { 6462b82032cSYan Zheng error = -EACCES; 6472b82032cSYan Zheng goto error_close_devices; 6482b82032cSYan Zheng } 6492b82032cSYan Zheng 650450ba0eaSJosef Bacik /* 651450ba0eaSJosef Bacik * Setup a dummy root and fs_info for test/set super. This is because 652450ba0eaSJosef Bacik * we don't actually fill this stuff out until open_ctree, but we need 653450ba0eaSJosef Bacik * it for searching for existing supers, so this lets us do that and 654450ba0eaSJosef Bacik * then open_ctree will properly initialize everything later. 655450ba0eaSJosef Bacik */ 656450ba0eaSJosef Bacik fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS); 657450ba0eaSJosef Bacik tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); 658450ba0eaSJosef Bacik if (!fs_info || !tree_root) { 659450ba0eaSJosef Bacik error = -ENOMEM; 660450ba0eaSJosef Bacik goto error_close_devices; 661450ba0eaSJosef Bacik } 662450ba0eaSJosef Bacik fs_info->tree_root = tree_root; 663450ba0eaSJosef Bacik fs_info->fs_devices = fs_devices; 664450ba0eaSJosef Bacik tree_root->fs_info = fs_info; 665450ba0eaSJosef Bacik 666dfe25020SChris Mason bdev = fs_devices->latest_bdev; 667450ba0eaSJosef Bacik s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root); 6684b82d6e4SYan if (IS_ERR(s)) 6694b82d6e4SYan goto error_s; 6704b82d6e4SYan 6714b82d6e4SYan if (s->s_root) { 6724b82d6e4SYan if ((flags ^ s->s_flags) & MS_RDONLY) { 6736f5bbff9SAl Viro deactivate_locked_super(s); 6744b82d6e4SYan error = -EBUSY; 675c146afadSYan Zheng goto error_close_devices; 6764b82d6e4SYan } 6774b82d6e4SYan 6782b82032cSYan Zheng btrfs_close_devices(fs_devices); 6794b82d6e4SYan } else { 6804b82d6e4SYan char b[BDEVNAME_SIZE]; 6814b82d6e4SYan 6824b82d6e4SYan s->s_flags = flags; 6834b82d6e4SYan strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); 6848a4b83ccSChris Mason error = btrfs_fill_super(s, fs_devices, data, 6858a4b83ccSChris Mason flags & MS_SILENT ? 1 : 0); 6864b82d6e4SYan if (error) { 6876f5bbff9SAl Viro deactivate_locked_super(s); 6881f483660SShen Feng goto error_free_subvol_name; 6894b82d6e4SYan } 6904b82d6e4SYan 691788f20ebSChris Mason btrfs_sb(s)->fs_info->bdev_holder = fs_type; 6924b82d6e4SYan s->s_flags |= MS_ACTIVE; 6934b82d6e4SYan } 6944b82d6e4SYan 69573f73415SJosef Bacik root = get_default_root(s, subvol_objectid); 6964b82d6e4SYan if (IS_ERR(root)) { 6974b82d6e4SYan error = PTR_ERR(root); 69873f73415SJosef Bacik deactivate_locked_super(s); 6990e78340fSJosef Bacik goto error_free_subvol_name; 7004b82d6e4SYan } 70173f73415SJosef Bacik /* if they gave us a subvolume name bind mount into that */ 70273f73415SJosef Bacik if (strcmp(subvol_name, ".")) { 70373f73415SJosef Bacik struct dentry *new_root; 70473f73415SJosef Bacik mutex_lock(&root->d_inode->i_mutex); 70573f73415SJosef Bacik new_root = lookup_one_len(subvol_name, root, 70673f73415SJosef Bacik strlen(subvol_name)); 70773f73415SJosef Bacik mutex_unlock(&root->d_inode->i_mutex); 70873f73415SJosef Bacik 70973f73415SJosef Bacik if (IS_ERR(new_root)) { 710f106e82cSLi Zefan dput(root); 71173f73415SJosef Bacik deactivate_locked_super(s); 71273f73415SJosef Bacik error = PTR_ERR(new_root); 7130e78340fSJosef Bacik goto error_free_subvol_name; 71473f73415SJosef Bacik } 71573f73415SJosef Bacik if (!new_root->d_inode) { 71673f73415SJosef Bacik dput(root); 71773f73415SJosef Bacik dput(new_root); 7186f5bbff9SAl Viro deactivate_locked_super(s); 7194b82d6e4SYan error = -ENXIO; 7200e78340fSJosef Bacik goto error_free_subvol_name; 7214b82d6e4SYan } 72273f73415SJosef Bacik dput(root); 72373f73415SJosef Bacik root = new_root; 72476fcef19SDavid Woodhouse } 7254b82d6e4SYan 7264b82d6e4SYan mnt->mnt_sb = s; 7274b82d6e4SYan mnt->mnt_root = root; 728edf24abeSChristoph Hellwig 729edf24abeSChristoph Hellwig kfree(subvol_name); 7304b82d6e4SYan return 0; 7314b82d6e4SYan 7324b82d6e4SYan error_s: 7334b82d6e4SYan error = PTR_ERR(s); 734c146afadSYan Zheng error_close_devices: 7358a4b83ccSChris Mason btrfs_close_devices(fs_devices); 736450ba0eaSJosef Bacik kfree(fs_info); 737450ba0eaSJosef Bacik kfree(tree_root); 738edf24abeSChristoph Hellwig error_free_subvol_name: 739edf24abeSChristoph Hellwig kfree(subvol_name); 7404b82d6e4SYan return error; 7414b82d6e4SYan } 7422e635a27SChris Mason 743c146afadSYan Zheng static int btrfs_remount(struct super_block *sb, int *flags, char *data) 744c146afadSYan Zheng { 745c146afadSYan Zheng struct btrfs_root *root = btrfs_sb(sb); 746c146afadSYan Zheng int ret; 747c146afadSYan Zheng 748b288052eSChris Mason ret = btrfs_parse_options(root, data); 749b288052eSChris Mason if (ret) 750b288052eSChris Mason return -EINVAL; 751b288052eSChris Mason 752c146afadSYan Zheng if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) 753c146afadSYan Zheng return 0; 754c146afadSYan Zheng 755c146afadSYan Zheng if (*flags & MS_RDONLY) { 756c146afadSYan Zheng sb->s_flags |= MS_RDONLY; 757c146afadSYan Zheng 758c146afadSYan Zheng ret = btrfs_commit_super(root); 759c146afadSYan Zheng WARN_ON(ret); 760c146afadSYan Zheng } else { 7612b82032cSYan Zheng if (root->fs_info->fs_devices->rw_devices == 0) 7622b82032cSYan Zheng return -EACCES; 7632b82032cSYan Zheng 764c146afadSYan Zheng if (btrfs_super_log_root(&root->fs_info->super_copy) != 0) 765c146afadSYan Zheng return -EINVAL; 766c146afadSYan Zheng 767d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(root->fs_info); 768c146afadSYan Zheng WARN_ON(ret); 769c146afadSYan Zheng 770d68fc57bSYan, Zheng /* recover relocation */ 771d68fc57bSYan, Zheng ret = btrfs_recover_relocation(root); 772c146afadSYan Zheng WARN_ON(ret); 773c146afadSYan Zheng 774c146afadSYan Zheng sb->s_flags &= ~MS_RDONLY; 775c146afadSYan Zheng } 776c146afadSYan Zheng 777c146afadSYan Zheng return 0; 778c146afadSYan Zheng } 779c146afadSYan Zheng 780*6d07bcecSMiao Xie /* 781*6d07bcecSMiao Xie * The helper to calc the free space on the devices that can be used to store 782*6d07bcecSMiao Xie * file data. 783*6d07bcecSMiao Xie */ 784*6d07bcecSMiao Xie static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes) 785*6d07bcecSMiao Xie { 786*6d07bcecSMiao Xie struct btrfs_fs_info *fs_info = root->fs_info; 787*6d07bcecSMiao Xie struct btrfs_device_info *devices_info; 788*6d07bcecSMiao Xie struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 789*6d07bcecSMiao Xie struct btrfs_device *device; 790*6d07bcecSMiao Xie u64 skip_space; 791*6d07bcecSMiao Xie u64 type; 792*6d07bcecSMiao Xie u64 avail_space; 793*6d07bcecSMiao Xie u64 used_space; 794*6d07bcecSMiao Xie u64 min_stripe_size; 795*6d07bcecSMiao Xie int min_stripes = 1; 796*6d07bcecSMiao Xie int i = 0, nr_devices; 797*6d07bcecSMiao Xie int ret; 798*6d07bcecSMiao Xie 799*6d07bcecSMiao Xie nr_devices = fs_info->fs_devices->rw_devices; 800*6d07bcecSMiao Xie BUG_ON(!nr_devices); 801*6d07bcecSMiao Xie 802*6d07bcecSMiao Xie devices_info = kmalloc(sizeof(*devices_info) * nr_devices, 803*6d07bcecSMiao Xie GFP_NOFS); 804*6d07bcecSMiao Xie if (!devices_info) 805*6d07bcecSMiao Xie return -ENOMEM; 806*6d07bcecSMiao Xie 807*6d07bcecSMiao Xie /* calc min stripe number for data space alloction */ 808*6d07bcecSMiao Xie type = btrfs_get_alloc_profile(root, 1); 809*6d07bcecSMiao Xie if (type & BTRFS_BLOCK_GROUP_RAID0) 810*6d07bcecSMiao Xie min_stripes = 2; 811*6d07bcecSMiao Xie else if (type & BTRFS_BLOCK_GROUP_RAID1) 812*6d07bcecSMiao Xie min_stripes = 2; 813*6d07bcecSMiao Xie else if (type & BTRFS_BLOCK_GROUP_RAID10) 814*6d07bcecSMiao Xie min_stripes = 4; 815*6d07bcecSMiao Xie 816*6d07bcecSMiao Xie if (type & BTRFS_BLOCK_GROUP_DUP) 817*6d07bcecSMiao Xie min_stripe_size = 2 * BTRFS_STRIPE_LEN; 818*6d07bcecSMiao Xie else 819*6d07bcecSMiao Xie min_stripe_size = BTRFS_STRIPE_LEN; 820*6d07bcecSMiao Xie 821*6d07bcecSMiao Xie list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) { 822*6d07bcecSMiao Xie if (!device->in_fs_metadata) 823*6d07bcecSMiao Xie continue; 824*6d07bcecSMiao Xie 825*6d07bcecSMiao Xie avail_space = device->total_bytes - device->bytes_used; 826*6d07bcecSMiao Xie 827*6d07bcecSMiao Xie /* align with stripe_len */ 828*6d07bcecSMiao Xie do_div(avail_space, BTRFS_STRIPE_LEN); 829*6d07bcecSMiao Xie avail_space *= BTRFS_STRIPE_LEN; 830*6d07bcecSMiao Xie 831*6d07bcecSMiao Xie /* 832*6d07bcecSMiao Xie * In order to avoid overwritting the superblock on the drive, 833*6d07bcecSMiao Xie * btrfs starts at an offset of at least 1MB when doing chunk 834*6d07bcecSMiao Xie * allocation. 835*6d07bcecSMiao Xie */ 836*6d07bcecSMiao Xie skip_space = 1024 * 1024; 837*6d07bcecSMiao Xie 838*6d07bcecSMiao Xie /* user can set the offset in fs_info->alloc_start. */ 839*6d07bcecSMiao Xie if (fs_info->alloc_start + BTRFS_STRIPE_LEN <= 840*6d07bcecSMiao Xie device->total_bytes) 841*6d07bcecSMiao Xie skip_space = max(fs_info->alloc_start, skip_space); 842*6d07bcecSMiao Xie 843*6d07bcecSMiao Xie /* 844*6d07bcecSMiao Xie * btrfs can not use the free space in [0, skip_space - 1], 845*6d07bcecSMiao Xie * we must subtract it from the total. In order to implement 846*6d07bcecSMiao Xie * it, we account the used space in this range first. 847*6d07bcecSMiao Xie */ 848*6d07bcecSMiao Xie ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1, 849*6d07bcecSMiao Xie &used_space); 850*6d07bcecSMiao Xie if (ret) { 851*6d07bcecSMiao Xie kfree(devices_info); 852*6d07bcecSMiao Xie return ret; 853*6d07bcecSMiao Xie } 854*6d07bcecSMiao Xie 855*6d07bcecSMiao Xie /* calc the free space in [0, skip_space - 1] */ 856*6d07bcecSMiao Xie skip_space -= used_space; 857*6d07bcecSMiao Xie 858*6d07bcecSMiao Xie /* 859*6d07bcecSMiao Xie * we can use the free space in [0, skip_space - 1], subtract 860*6d07bcecSMiao Xie * it from the total. 861*6d07bcecSMiao Xie */ 862*6d07bcecSMiao Xie if (avail_space && avail_space >= skip_space) 863*6d07bcecSMiao Xie avail_space -= skip_space; 864*6d07bcecSMiao Xie else 865*6d07bcecSMiao Xie avail_space = 0; 866*6d07bcecSMiao Xie 867*6d07bcecSMiao Xie if (avail_space < min_stripe_size) 868*6d07bcecSMiao Xie continue; 869*6d07bcecSMiao Xie 870*6d07bcecSMiao Xie devices_info[i].dev = device; 871*6d07bcecSMiao Xie devices_info[i].max_avail = avail_space; 872*6d07bcecSMiao Xie 873*6d07bcecSMiao Xie i++; 874*6d07bcecSMiao Xie } 875*6d07bcecSMiao Xie 876*6d07bcecSMiao Xie nr_devices = i; 877*6d07bcecSMiao Xie 878*6d07bcecSMiao Xie btrfs_descending_sort_devices(devices_info, nr_devices); 879*6d07bcecSMiao Xie 880*6d07bcecSMiao Xie i = nr_devices - 1; 881*6d07bcecSMiao Xie avail_space = 0; 882*6d07bcecSMiao Xie while (nr_devices >= min_stripes) { 883*6d07bcecSMiao Xie if (devices_info[i].max_avail >= min_stripe_size) { 884*6d07bcecSMiao Xie int j; 885*6d07bcecSMiao Xie u64 alloc_size; 886*6d07bcecSMiao Xie 887*6d07bcecSMiao Xie avail_space += devices_info[i].max_avail * min_stripes; 888*6d07bcecSMiao Xie alloc_size = devices_info[i].max_avail; 889*6d07bcecSMiao Xie for (j = i + 1 - min_stripes; j <= i; j++) 890*6d07bcecSMiao Xie devices_info[j].max_avail -= alloc_size; 891*6d07bcecSMiao Xie } 892*6d07bcecSMiao Xie i--; 893*6d07bcecSMiao Xie nr_devices--; 894*6d07bcecSMiao Xie } 895*6d07bcecSMiao Xie 896*6d07bcecSMiao Xie kfree(devices_info); 897*6d07bcecSMiao Xie *free_bytes = avail_space; 898*6d07bcecSMiao Xie return 0; 899*6d07bcecSMiao Xie } 900*6d07bcecSMiao Xie 9018fd17795SChris Mason static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) 9028fd17795SChris Mason { 9038fd17795SChris Mason struct btrfs_root *root = btrfs_sb(dentry->d_sb); 9044b52dff6SChris Mason struct btrfs_super_block *disk_super = &root->fs_info->super_copy; 905bd4d1088SJosef Bacik struct list_head *head = &root->fs_info->space_info; 906bd4d1088SJosef Bacik struct btrfs_space_info *found; 907bd4d1088SJosef Bacik u64 total_used = 0; 908*6d07bcecSMiao Xie u64 total_free_data = 0; 909db94535dSChris Mason int bits = dentry->d_sb->s_blocksize_bits; 9109d03632eSDavid Woodhouse __be32 *fsid = (__be32 *)root->fs_info->fsid; 911*6d07bcecSMiao Xie int ret; 9128fd17795SChris Mason 913*6d07bcecSMiao Xie /* holding chunk_muext to avoid allocating new chunks */ 914*6d07bcecSMiao Xie mutex_lock(&root->fs_info->chunk_mutex); 915bd4d1088SJosef Bacik rcu_read_lock(); 91689a55897SJosef Bacik list_for_each_entry_rcu(found, head, list) { 917*6d07bcecSMiao Xie if (found->flags & BTRFS_BLOCK_GROUP_DATA) { 918*6d07bcecSMiao Xie total_free_data += found->disk_total - found->disk_used; 919*6d07bcecSMiao Xie total_free_data -= 920*6d07bcecSMiao Xie btrfs_account_ro_block_groups_free_space(found); 921*6d07bcecSMiao Xie } 922*6d07bcecSMiao Xie 923b742bb82SYan, Zheng total_used += found->disk_used; 92489a55897SJosef Bacik } 925bd4d1088SJosef Bacik rcu_read_unlock(); 926bd4d1088SJosef Bacik 9278fd17795SChris Mason buf->f_namelen = BTRFS_NAME_LEN; 928db94535dSChris Mason buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; 929bd4d1088SJosef Bacik buf->f_bfree = buf->f_blocks - (total_used >> bits); 9308fd17795SChris Mason buf->f_bsize = dentry->d_sb->s_blocksize; 9318fd17795SChris Mason buf->f_type = BTRFS_SUPER_MAGIC; 932*6d07bcecSMiao Xie buf->f_bavail = total_free_data; 933*6d07bcecSMiao Xie ret = btrfs_calc_avail_data_space(root, &total_free_data); 934*6d07bcecSMiao Xie if (ret) { 935*6d07bcecSMiao Xie mutex_unlock(&root->fs_info->chunk_mutex); 936*6d07bcecSMiao Xie return ret; 937*6d07bcecSMiao Xie } 938*6d07bcecSMiao Xie buf->f_bavail += total_free_data; 939*6d07bcecSMiao Xie buf->f_bavail = buf->f_bavail >> bits; 940*6d07bcecSMiao Xie mutex_unlock(&root->fs_info->chunk_mutex); 941d397712bSChris Mason 9429d03632eSDavid Woodhouse /* We treat it as constant endianness (it doesn't matter _which_) 9439d03632eSDavid Woodhouse because we want the fsid to come out the same whether mounted 9449d03632eSDavid Woodhouse on a big-endian or little-endian host */ 9459d03632eSDavid Woodhouse buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); 9469d03632eSDavid Woodhouse buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); 94732d48fa1SDavid Woodhouse /* Mask in the root object ID too, to disambiguate subvols */ 94832d48fa1SDavid Woodhouse buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32; 94932d48fa1SDavid Woodhouse buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid; 95032d48fa1SDavid Woodhouse 9518fd17795SChris Mason return 0; 9528fd17795SChris Mason } 953b5133862SChris Mason 9542e635a27SChris Mason static struct file_system_type btrfs_fs_type = { 9552e635a27SChris Mason .owner = THIS_MODULE, 9562e635a27SChris Mason .name = "btrfs", 9572e635a27SChris Mason .get_sb = btrfs_get_sb, 958a061fc8dSChris Mason .kill_sb = kill_anon_super, 9592e635a27SChris Mason .fs_flags = FS_REQUIRES_DEV, 9602e635a27SChris Mason }; 961a9218f6bSChris Mason 962d352ac68SChris Mason /* 963d352ac68SChris Mason * used by btrfsctl to scan devices when no FS is mounted 964d352ac68SChris Mason */ 9658a4b83ccSChris Mason static long btrfs_control_ioctl(struct file *file, unsigned int cmd, 9668a4b83ccSChris Mason unsigned long arg) 9678a4b83ccSChris Mason { 9688a4b83ccSChris Mason struct btrfs_ioctl_vol_args *vol; 9698a4b83ccSChris Mason struct btrfs_fs_devices *fs_devices; 970c071fcfdSChris Mason int ret = -ENOTTY; 9718a4b83ccSChris Mason 972e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 973e441d54dSChris Mason return -EPERM; 974e441d54dSChris Mason 975dae7b665SLi Zefan vol = memdup_user((void __user *)arg, sizeof(*vol)); 976dae7b665SLi Zefan if (IS_ERR(vol)) 977dae7b665SLi Zefan return PTR_ERR(vol); 978c071fcfdSChris Mason 9798a4b83ccSChris Mason switch (cmd) { 9808a4b83ccSChris Mason case BTRFS_IOC_SCAN_DEV: 98197288f2cSChristoph Hellwig ret = btrfs_scan_one_device(vol->name, FMODE_READ, 9828a4b83ccSChris Mason &btrfs_fs_type, &fs_devices); 9838a4b83ccSChris Mason break; 9848a4b83ccSChris Mason } 985dae7b665SLi Zefan 9868a4b83ccSChris Mason kfree(vol); 987f819d837SLinda Knippers return ret; 9888a4b83ccSChris Mason } 9898a4b83ccSChris Mason 9900176260fSLinus Torvalds static int btrfs_freeze(struct super_block *sb) 991ed0dab6bSYan { 992ed0dab6bSYan struct btrfs_root *root = btrfs_sb(sb); 993a74a4b97SChris Mason mutex_lock(&root->fs_info->transaction_kthread_mutex); 994a74a4b97SChris Mason mutex_lock(&root->fs_info->cleaner_mutex); 9950176260fSLinus Torvalds return 0; 996ed0dab6bSYan } 997ed0dab6bSYan 9980176260fSLinus Torvalds static int btrfs_unfreeze(struct super_block *sb) 999ed0dab6bSYan { 1000ed0dab6bSYan struct btrfs_root *root = btrfs_sb(sb); 1001a74a4b97SChris Mason mutex_unlock(&root->fs_info->cleaner_mutex); 1002a74a4b97SChris Mason mutex_unlock(&root->fs_info->transaction_kthread_mutex); 10030176260fSLinus Torvalds return 0; 1004ed0dab6bSYan } 10052e635a27SChris Mason 1006b87221deSAlexey Dobriyan static const struct super_operations btrfs_super_ops = { 100776dda93cSYan, Zheng .drop_inode = btrfs_drop_inode, 1008bd555975SAl Viro .evict_inode = btrfs_evict_inode, 1009e20d96d6SChris Mason .put_super = btrfs_put_super, 1010d5719762SChris Mason .sync_fs = btrfs_sync_fs, 1011a9572a15SEric Paris .show_options = btrfs_show_options, 10124730a4bcSChris Mason .write_inode = btrfs_write_inode, 1013b5133862SChris Mason .dirty_inode = btrfs_dirty_inode, 10142c90e5d6SChris Mason .alloc_inode = btrfs_alloc_inode, 10152c90e5d6SChris Mason .destroy_inode = btrfs_destroy_inode, 10168fd17795SChris Mason .statfs = btrfs_statfs, 1017c146afadSYan Zheng .remount_fs = btrfs_remount, 10180176260fSLinus Torvalds .freeze_fs = btrfs_freeze, 10190176260fSLinus Torvalds .unfreeze_fs = btrfs_unfreeze, 1020e20d96d6SChris Mason }; 1021a9218f6bSChris Mason 1022a9218f6bSChris Mason static const struct file_operations btrfs_ctl_fops = { 1023a9218f6bSChris Mason .unlocked_ioctl = btrfs_control_ioctl, 1024a9218f6bSChris Mason .compat_ioctl = btrfs_control_ioctl, 1025a9218f6bSChris Mason .owner = THIS_MODULE, 1026a9218f6bSChris Mason }; 1027a9218f6bSChris Mason 1028a9218f6bSChris Mason static struct miscdevice btrfs_misc = { 1029578454ffSKay Sievers .minor = BTRFS_MINOR, 1030a9218f6bSChris Mason .name = "btrfs-control", 1031a9218f6bSChris Mason .fops = &btrfs_ctl_fops 1032a9218f6bSChris Mason }; 1033a9218f6bSChris Mason 1034578454ffSKay Sievers MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 1035578454ffSKay Sievers MODULE_ALIAS("devname:btrfs-control"); 1036578454ffSKay Sievers 1037a9218f6bSChris Mason static int btrfs_interface_init(void) 1038a9218f6bSChris Mason { 1039a9218f6bSChris Mason return misc_register(&btrfs_misc); 1040a9218f6bSChris Mason } 1041a9218f6bSChris Mason 1042b2950863SChristoph Hellwig static void btrfs_interface_exit(void) 1043a9218f6bSChris Mason { 1044a9218f6bSChris Mason if (misc_deregister(&btrfs_misc) < 0) 1045d397712bSChris Mason printk(KERN_INFO "misc_deregister failed for control device"); 1046a9218f6bSChris Mason } 1047a9218f6bSChris Mason 10482e635a27SChris Mason static int __init init_btrfs_fs(void) 10492e635a27SChris Mason { 10502c90e5d6SChris Mason int err; 105158176a96SJosef Bacik 105258176a96SJosef Bacik err = btrfs_init_sysfs(); 105358176a96SJosef Bacik if (err) 105458176a96SJosef Bacik return err; 105558176a96SJosef Bacik 1056261507a0SLi Zefan err = btrfs_init_compress(); 10572c90e5d6SChris Mason if (err) 1058a74a4b97SChris Mason goto free_sysfs; 1059d1310b2eSChris Mason 1060261507a0SLi Zefan err = btrfs_init_cachep(); 1061261507a0SLi Zefan if (err) 1062261507a0SLi Zefan goto free_compress; 1063261507a0SLi Zefan 1064d1310b2eSChris Mason err = extent_io_init(); 10652f4cbe64SWyatt Banks if (err) 10662f4cbe64SWyatt Banks goto free_cachep; 10672f4cbe64SWyatt Banks 1068d1310b2eSChris Mason err = extent_map_init(); 1069d1310b2eSChris Mason if (err) 1070d1310b2eSChris Mason goto free_extent_io; 1071d1310b2eSChris Mason 1072a9218f6bSChris Mason err = btrfs_interface_init(); 10732f4cbe64SWyatt Banks if (err) 10742f4cbe64SWyatt Banks goto free_extent_map; 1075c8b97818SChris Mason 1076a9218f6bSChris Mason err = register_filesystem(&btrfs_fs_type); 1077a9218f6bSChris Mason if (err) 1078a9218f6bSChris Mason goto unregister_ioctl; 1079b3c3da71SChris Mason 1080b3c3da71SChris Mason printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION); 10812f4cbe64SWyatt Banks return 0; 10822f4cbe64SWyatt Banks 1083a9218f6bSChris Mason unregister_ioctl: 1084a9218f6bSChris Mason btrfs_interface_exit(); 10852f4cbe64SWyatt Banks free_extent_map: 10862f4cbe64SWyatt Banks extent_map_exit(); 1087d1310b2eSChris Mason free_extent_io: 1088d1310b2eSChris Mason extent_io_exit(); 10892f4cbe64SWyatt Banks free_cachep: 10902f4cbe64SWyatt Banks btrfs_destroy_cachep(); 1091261507a0SLi Zefan free_compress: 1092261507a0SLi Zefan btrfs_exit_compress(); 1093a74a4b97SChris Mason free_sysfs: 10942f4cbe64SWyatt Banks btrfs_exit_sysfs(); 10952c90e5d6SChris Mason return err; 10962e635a27SChris Mason } 10972e635a27SChris Mason 10982e635a27SChris Mason static void __exit exit_btrfs_fs(void) 10992e635a27SChris Mason { 110039279cc3SChris Mason btrfs_destroy_cachep(); 1101a52d9a80SChris Mason extent_map_exit(); 1102d1310b2eSChris Mason extent_io_exit(); 1103a9218f6bSChris Mason btrfs_interface_exit(); 11042e635a27SChris Mason unregister_filesystem(&btrfs_fs_type); 110558176a96SJosef Bacik btrfs_exit_sysfs(); 11068a4b83ccSChris Mason btrfs_cleanup_fs_uuids(); 1107261507a0SLi Zefan btrfs_exit_compress(); 11082e635a27SChris Mason } 11092e635a27SChris Mason 11102e635a27SChris Mason module_init(init_btrfs_fs) 11112e635a27SChris Mason module_exit(exit_btrfs_fs) 11122e635a27SChris Mason 11132e635a27SChris Mason MODULE_LICENSE("GPL"); 1114