1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 2011, Lawrence Livermore National Security, LLC. 23 */ 24 25 26 #include <sys/zfs_znode.h> 27 #include <sys/zfs_vfsops.h> 28 #include <sys/zfs_vnops.h> 29 #include <sys/zfs_ctldir.h> 30 #include <sys/zpl.h> 31 32 33 static struct inode * 34 zpl_inode_alloc(struct super_block *sb) 35 { 36 struct inode *ip; 37 38 VERIFY3S(zfs_inode_alloc(sb, &ip), ==, 0); 39 inode_set_iversion(ip, 1); 40 41 return (ip); 42 } 43 44 static void 45 zpl_inode_destroy(struct inode *ip) 46 { 47 ASSERT(atomic_read(&ip->i_count) == 0); 48 zfs_inode_destroy(ip); 49 } 50 51 /* 52 * Called from __mark_inode_dirty() to reflect that something in the 53 * inode has changed. We use it to ensure the znode system attributes 54 * are always strictly update to date with respect to the inode. 55 */ 56 #ifdef HAVE_DIRTY_INODE_WITH_FLAGS 57 static void 58 zpl_dirty_inode(struct inode *ip, int flags) 59 { 60 fstrans_cookie_t cookie; 61 62 cookie = spl_fstrans_mark(); 63 zfs_dirty_inode(ip, flags); 64 spl_fstrans_unmark(cookie); 65 } 66 #else 67 static void 68 zpl_dirty_inode(struct inode *ip) 69 { 70 fstrans_cookie_t cookie; 71 72 cookie = spl_fstrans_mark(); 73 zfs_dirty_inode(ip, 0); 74 spl_fstrans_unmark(cookie); 75 } 76 #endif /* HAVE_DIRTY_INODE_WITH_FLAGS */ 77 78 /* 79 * When ->drop_inode() is called its return value indicates if the 80 * inode should be evicted from the inode cache. If the inode is 81 * unhashed and has no links the default policy is to evict it 82 * immediately. 83 * 84 * The ->evict_inode() callback must minimally truncate the inode pages, 85 * and call clear_inode(). For 2.6.35 and later kernels this will 86 * simply update the inode state, with the sync occurring before the 87 * truncate in evict(). For earlier kernels clear_inode() maps to 88 * end_writeback() which is responsible for completing all outstanding 89 * write back. In either case, once this is done it is safe to cleanup 90 * any remaining inode specific data via zfs_inactive(). 91 * remaining filesystem specific data. 92 */ 93 static void 94 zpl_evict_inode(struct inode *ip) 95 { 96 fstrans_cookie_t cookie; 97 98 cookie = spl_fstrans_mark(); 99 truncate_setsize(ip, 0); 100 clear_inode(ip); 101 zfs_inactive(ip); 102 spl_fstrans_unmark(cookie); 103 } 104 105 static void 106 zpl_put_super(struct super_block *sb) 107 { 108 fstrans_cookie_t cookie; 109 int error; 110 111 cookie = spl_fstrans_mark(); 112 error = -zfs_umount(sb); 113 spl_fstrans_unmark(cookie); 114 ASSERT3S(error, <=, 0); 115 } 116 117 static int 118 zpl_sync_fs(struct super_block *sb, int wait) 119 { 120 fstrans_cookie_t cookie; 121 cred_t *cr = CRED(); 122 int error; 123 124 crhold(cr); 125 cookie = spl_fstrans_mark(); 126 error = -zfs_sync(sb, wait, cr); 127 spl_fstrans_unmark(cookie); 128 crfree(cr); 129 ASSERT3S(error, <=, 0); 130 131 return (error); 132 } 133 134 static int 135 zpl_statfs(struct dentry *dentry, struct kstatfs *statp) 136 { 137 fstrans_cookie_t cookie; 138 int error; 139 140 cookie = spl_fstrans_mark(); 141 error = -zfs_statvfs(dentry->d_inode, statp); 142 spl_fstrans_unmark(cookie); 143 ASSERT3S(error, <=, 0); 144 145 /* 146 * If required by a 32-bit system call, dynamically scale the 147 * block size up to 16MiB and decrease the block counts. This 148 * allows for a maximum size of 64EiB to be reported. The file 149 * counts must be artificially capped at 2^32-1. 150 */ 151 if (unlikely(zpl_is_32bit_api())) { 152 while (statp->f_blocks > UINT32_MAX && 153 statp->f_bsize < SPA_MAXBLOCKSIZE) { 154 statp->f_frsize <<= 1; 155 statp->f_bsize <<= 1; 156 157 statp->f_blocks >>= 1; 158 statp->f_bfree >>= 1; 159 statp->f_bavail >>= 1; 160 } 161 162 uint64_t usedobjs = statp->f_files - statp->f_ffree; 163 statp->f_ffree = MIN(statp->f_ffree, UINT32_MAX - usedobjs); 164 statp->f_files = statp->f_ffree + usedobjs; 165 } 166 167 return (error); 168 } 169 170 static int 171 zpl_remount_fs(struct super_block *sb, int *flags, char *data) 172 { 173 zfs_mnt_t zm = { .mnt_osname = NULL, .mnt_data = data }; 174 fstrans_cookie_t cookie; 175 int error; 176 177 cookie = spl_fstrans_mark(); 178 error = -zfs_remount(sb, flags, &zm); 179 spl_fstrans_unmark(cookie); 180 ASSERT3S(error, <=, 0); 181 182 return (error); 183 } 184 185 static int 186 __zpl_show_devname(struct seq_file *seq, zfsvfs_t *zfsvfs) 187 { 188 char *fsname; 189 190 ZFS_ENTER(zfsvfs); 191 fsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); 192 dmu_objset_name(zfsvfs->z_os, fsname); 193 seq_puts(seq, fsname); 194 kmem_free(fsname, ZFS_MAX_DATASET_NAME_LEN); 195 ZFS_EXIT(zfsvfs); 196 197 return (0); 198 } 199 200 static int 201 zpl_show_devname(struct seq_file *seq, struct dentry *root) 202 { 203 return (__zpl_show_devname(seq, root->d_sb->s_fs_info)); 204 } 205 206 static int 207 __zpl_show_options(struct seq_file *seq, zfsvfs_t *zfsvfs) 208 { 209 seq_printf(seq, ",%s", 210 zfsvfs->z_flags & ZSB_XATTR ? "xattr" : "noxattr"); 211 212 #ifdef CONFIG_FS_POSIX_ACL 213 switch (zfsvfs->z_acl_type) { 214 case ZFS_ACLTYPE_POSIX: 215 seq_puts(seq, ",posixacl"); 216 break; 217 default: 218 seq_puts(seq, ",noacl"); 219 break; 220 } 221 #endif /* CONFIG_FS_POSIX_ACL */ 222 223 return (0); 224 } 225 226 static int 227 zpl_show_options(struct seq_file *seq, struct dentry *root) 228 { 229 return (__zpl_show_options(seq, root->d_sb->s_fs_info)); 230 } 231 232 static int 233 zpl_fill_super(struct super_block *sb, void *data, int silent) 234 { 235 zfs_mnt_t *zm = (zfs_mnt_t *)data; 236 fstrans_cookie_t cookie; 237 int error; 238 239 cookie = spl_fstrans_mark(); 240 error = -zfs_domount(sb, zm, silent); 241 spl_fstrans_unmark(cookie); 242 ASSERT3S(error, <=, 0); 243 244 return (error); 245 } 246 247 static int 248 zpl_test_super(struct super_block *s, void *data) 249 { 250 zfsvfs_t *zfsvfs = s->s_fs_info; 251 objset_t *os = data; 252 253 if (zfsvfs == NULL) 254 return (0); 255 256 return (os == zfsvfs->z_os); 257 } 258 259 static struct super_block * 260 zpl_mount_impl(struct file_system_type *fs_type, int flags, zfs_mnt_t *zm) 261 { 262 struct super_block *s; 263 objset_t *os; 264 int err; 265 266 err = dmu_objset_hold(zm->mnt_osname, FTAG, &os); 267 if (err) 268 return (ERR_PTR(-err)); 269 270 /* 271 * The dsl pool lock must be released prior to calling sget(). 272 * It is possible sget() may block on the lock in grab_super() 273 * while deactivate_super() holds that same lock and waits for 274 * a txg sync. If the dsl_pool lock is held over sget() 275 * this can prevent the pool sync and cause a deadlock. 276 */ 277 dsl_dataset_long_hold(dmu_objset_ds(os), FTAG); 278 dsl_pool_rele(dmu_objset_pool(os), FTAG); 279 280 s = sget(fs_type, zpl_test_super, set_anon_super, flags, os); 281 282 dsl_dataset_long_rele(dmu_objset_ds(os), FTAG); 283 dsl_dataset_rele(dmu_objset_ds(os), FTAG); 284 285 if (IS_ERR(s)) 286 return (ERR_CAST(s)); 287 288 if (s->s_root == NULL) { 289 err = zpl_fill_super(s, zm, flags & SB_SILENT ? 1 : 0); 290 if (err) { 291 deactivate_locked_super(s); 292 return (ERR_PTR(err)); 293 } 294 s->s_flags |= SB_ACTIVE; 295 } else if ((flags ^ s->s_flags) & SB_RDONLY) { 296 deactivate_locked_super(s); 297 return (ERR_PTR(-EBUSY)); 298 } 299 300 return (s); 301 } 302 303 static struct dentry * 304 zpl_mount(struct file_system_type *fs_type, int flags, 305 const char *osname, void *data) 306 { 307 zfs_mnt_t zm = { .mnt_osname = osname, .mnt_data = data }; 308 309 struct super_block *sb = zpl_mount_impl(fs_type, flags, &zm); 310 if (IS_ERR(sb)) 311 return (ERR_CAST(sb)); 312 313 return (dget(sb->s_root)); 314 } 315 316 static void 317 zpl_kill_sb(struct super_block *sb) 318 { 319 zfs_preumount(sb); 320 kill_anon_super(sb); 321 } 322 323 void 324 zpl_prune_sb(int64_t nr_to_scan, void *arg) 325 { 326 struct super_block *sb = (struct super_block *)arg; 327 int objects = 0; 328 329 (void) -zfs_prune(sb, nr_to_scan, &objects); 330 } 331 332 const struct super_operations zpl_super_operations = { 333 .alloc_inode = zpl_inode_alloc, 334 .destroy_inode = zpl_inode_destroy, 335 .dirty_inode = zpl_dirty_inode, 336 .write_inode = NULL, 337 .evict_inode = zpl_evict_inode, 338 .put_super = zpl_put_super, 339 .sync_fs = zpl_sync_fs, 340 .statfs = zpl_statfs, 341 .remount_fs = zpl_remount_fs, 342 .show_devname = zpl_show_devname, 343 .show_options = zpl_show_options, 344 .show_stats = NULL, 345 }; 346 347 struct file_system_type zpl_fs_type = { 348 .owner = THIS_MODULE, 349 .name = ZFS_DRIVER, 350 .mount = zpl_mount, 351 .kill_sb = zpl_kill_sb, 352 }; 353