module/zfs/zpl_super.c

   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_vfsops.h>
  27 #include <sys/zfs_vnops.h>
  28 #include <sys/zfs_znode.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  * Prior to 2.6.36 this eviction was accomplished by the vfs calling
  85  * ->delete_inode().  It was ->delete_inode()'s responsibility to
  86  * truncate the inode pages and call clear_inode().  The call to
  87  * clear_inode() synchronously invalidates all the buffers and
  88  * calls ->clear_inode().  It was ->clear_inode()'s responsibility
  89  * to cleanup and filesystem specific data before freeing the inode.
  90  *
  91  * This elaborate mechanism was replaced by ->evict_inode() which
  92  * does the job of both ->delete_inode() and ->clear_inode().  It
  93  * will be called exactly once, and when it returns the inode must
  94  * be in a state where it can simply be freed.i
  95  *
  96  * The ->evict_inode() callback must minimally truncate the inode pages,
  97  * and call clear_inode().  For 2.6.35 and later kernels this will
  98  * simply update the inode state, with the sync occurring before the
  99  * truncate in evict().  For earlier kernels clear_inode() maps to
 100  * end_writeback() which is responsible for completing all outstanding
 101  * write back.  In either case, once this is done it is safe to cleanup
 102  * any remaining inode specific data via zfs_inactive().
 103  * remaining filesystem specific data.
 104  */
 105 #ifdef HAVE_EVICT_INODE
 106 static void
 107 zpl_evict_inode(struct inode *ip)
 108 {
 109         fstrans_cookie_t cookie;
 110
 111         cookie = spl_fstrans_mark();
 112         truncate_setsize(ip, 0);
 113         clear_inode(ip);
 114         zfs_inactive(ip);
 115         spl_fstrans_unmark(cookie);
 116 }
 117
 118 #else
 119
 120 static void
 121 zpl_drop_inode(struct inode *ip)
 122 {
 123         generic_delete_inode(ip);
 124 }
 125
 126 static void
 127 zpl_clear_inode(struct inode *ip)
 128 {
 129         fstrans_cookie_t cookie;
 130
 131         cookie = spl_fstrans_mark();
 132         zfs_inactive(ip);
 133         spl_fstrans_unmark(cookie);
 134 }
 135
 136 static void
 137 zpl_inode_delete(struct inode *ip)
 138 {
 139         truncate_setsize(ip, 0);
 140         clear_inode(ip);
 141 }
 142 #endif /* HAVE_EVICT_INODE */
 143
 144 static void
 145 zpl_put_super(struct super_block *sb)
 146 {
 147         fstrans_cookie_t cookie;
 148         int error;
 149
 150         cookie = spl_fstrans_mark();
 151         error = -zfs_umount(sb);
 152         spl_fstrans_unmark(cookie);
 153         ASSERT3S(error, <=, 0);
 154 }
 155
 156 static int
 157 zpl_sync_fs(struct super_block *sb, int wait)
 158 {
 159         fstrans_cookie_t cookie;
 160         cred_t *cr = CRED();
 161         int error;
 162
 163         crhold(cr);
 164         cookie = spl_fstrans_mark();
 165         error = -zfs_sync(sb, wait, cr);
 166         spl_fstrans_unmark(cookie);
 167         crfree(cr);
 168         ASSERT3S(error, <=, 0);
 169
 170         return (error);
 171 }
 172
 173 static int
 174 zpl_statfs(struct dentry *dentry, struct kstatfs *statp)
 175 {
 176         fstrans_cookie_t cookie;
 177         int error;
 178
 179         cookie = spl_fstrans_mark();
 180         error = -zfs_statvfs(dentry, statp);
 181         spl_fstrans_unmark(cookie);
 182         ASSERT3S(error, <=, 0);
 183
 184         /*
 185          * If required by a 32-bit system call, dynamically scale the
 186          * block size up to 16MiB and decrease the block counts.  This
 187          * allows for a maximum size of 64EiB to be reported.  The file
 188          * counts must be artificially capped at 2^32-1.
 189          */
 190         if (unlikely(zpl_is_32bit_api())) {
 191                 while (statp->f_blocks > UINT32_MAX &&
 192                     statp->f_bsize < SPA_MAXBLOCKSIZE) {
 193                         statp->f_frsize <<= 1;
 194                         statp->f_bsize <<= 1;
 195
 196                         statp->f_blocks >>= 1;
 197                         statp->f_bfree >>= 1;
 198                         statp->f_bavail >>= 1;
 199                 }
 200
 201                 uint64_t usedobjs = statp->f_files - statp->f_ffree;
 202                 statp->f_ffree = MIN(statp->f_ffree, UINT32_MAX - usedobjs);
 203                 statp->f_files = statp->f_ffree + usedobjs;
 204         }
 205
 206         return (error);
 207 }
 208
 209 static int
 210 zpl_remount_fs(struct super_block *sb, int *flags, char *data)
 211 {
 212         zfs_mnt_t zm = { .mnt_osname = NULL, .mnt_data = data };
 213         fstrans_cookie_t cookie;
 214         int error;
 215
 216         cookie = spl_fstrans_mark();
 217         error = -zfs_remount(sb, flags, &zm);
 218         spl_fstrans_unmark(cookie);
 219         ASSERT3S(error, <=, 0);
 220
 221         return (error);
 222 }
 223
 224 static int
 225 __zpl_show_options(struct seq_file *seq, zfsvfs_t *zfsvfs)
 226 {
 227         seq_printf(seq, ",%s",
 228             zfsvfs->z_flags & ZSB_XATTR ? "xattr" : "noxattr");
 229
 230 #ifdef CONFIG_FS_POSIX_ACL
 231         switch (zfsvfs->z_acl_type) {
 232         case ZFS_ACLTYPE_POSIXACL:
 233                 seq_puts(seq, ",posixacl");
 234                 break;
 235         default:
 236                 seq_puts(seq, ",noacl");
 237                 break;
 238         }
 239 #endif /* CONFIG_FS_POSIX_ACL */
 240
 241         return (0);
 242 }
 243
 244 #ifdef HAVE_SHOW_OPTIONS_WITH_DENTRY
 245 static int
 246 zpl_show_options(struct seq_file *seq, struct dentry *root)
 247 {
 248         return (__zpl_show_options(seq, root->d_sb->s_fs_info));
 249 }
 250 #else
 251 static int
 252 zpl_show_options(struct seq_file *seq, struct vfsmount *vfsp)
 253 {
 254         return (__zpl_show_options(seq, vfsp->mnt_sb->s_fs_info));
 255 }
 256 #endif /* HAVE_SHOW_OPTIONS_WITH_DENTRY */
 257
 258 static int
 259 zpl_fill_super(struct super_block *sb, void *data, int silent)
 260 {
 261         zfs_mnt_t *zm = (zfs_mnt_t *)data;
 262         fstrans_cookie_t cookie;
 263         int error;
 264
 265         cookie = spl_fstrans_mark();
 266         error = -zfs_domount(sb, zm, silent);
 267         spl_fstrans_unmark(cookie);
 268         ASSERT3S(error, <=, 0);
 269
 270         return (error);
 271 }
 272
 273 static int
 274 zpl_test_super(struct super_block *s, void *data)
 275 {
 276         zfsvfs_t *zfsvfs = s->s_fs_info;
 277         objset_t *os = data;
 278
 279         if (zfsvfs == NULL)
 280                 return (0);
 281
 282         return (os == zfsvfs->z_os);
 283 }
 284
 285 static struct super_block *
 286 zpl_mount_impl(struct file_system_type *fs_type, int flags, zfs_mnt_t *zm)
 287 {
 288         struct super_block *s;
 289         objset_t *os;
 290         int err;
 291
 292         err = dmu_objset_hold(zm->mnt_osname, FTAG, &os);
 293         if (err)
 294                 return (ERR_PTR(-err));
 295
 296         /*
 297          * The dsl pool lock must be released prior to calling sget().
 298          * It is possible sget() may block on the lock in grab_super()
 299          * while deactivate_super() holds that same lock and waits for
 300          * a txg sync.  If the dsl_pool lock is held over over sget()
 301          * this can prevent the pool sync and cause a deadlock.
 302          */
 303         dsl_pool_rele(dmu_objset_pool(os), FTAG);
 304         s = zpl_sget(fs_type, zpl_test_super, set_anon_super, flags, os);
 305         dsl_dataset_rele(dmu_objset_ds(os), FTAG);
 306
 307         if (IS_ERR(s))
 308                 return (ERR_CAST(s));
 309
 310         if (s->s_root == NULL) {
 311                 err = zpl_fill_super(s, zm, flags & SB_SILENT ? 1 : 0);
 312                 if (err) {
 313                         deactivate_locked_super(s);
 314                         return (ERR_PTR(err));
 315                 }
 316                 s->s_flags |= SB_ACTIVE;
 317         } else if ((flags ^ s->s_flags) & SB_RDONLY) {
 318                 deactivate_locked_super(s);
 319                 return (ERR_PTR(-EBUSY));
 320         }
 321
 322         return (s);
 323 }
 324
 325 #ifdef HAVE_FST_MOUNT
 326 static struct dentry *
 327 zpl_mount(struct file_system_type *fs_type, int flags,
 328     const char *osname, void *data)
 329 {
 330         zfs_mnt_t zm = { .mnt_osname = osname, .mnt_data = data };
 331
 332         struct super_block *sb = zpl_mount_impl(fs_type, flags, &zm);
 333         if (IS_ERR(sb))
 334                 return (ERR_CAST(sb));
 335
 336         return (dget(sb->s_root));
 337 }
 338 #else
 339 static int
 340 zpl_get_sb(struct file_system_type *fs_type, int flags,
 341     const char *osname, void *data, struct vfsmount *mnt)
 342 {
 343         zfs_mnt_t zm = { .mnt_osname = osname, .mnt_data = data };
 344
 345         struct super_block *sb = zpl_mount_impl(fs_type, flags, &zm);
 346         if (IS_ERR(sb))
 347                 return (PTR_ERR(sb));
 348
 349         (void) simple_set_mnt(mnt, sb);
 350
 351         return (0);
 352 }
 353 #endif /* HAVE_FST_MOUNT */
 354
 355 static void
 356 zpl_kill_sb(struct super_block *sb)
 357 {
 358         zfs_preumount(sb);
 359         kill_anon_super(sb);
 360
 361 #ifdef HAVE_S_INSTANCES_LIST_HEAD
 362         sb->s_instances.next = &(zpl_fs_type.fs_supers);
 363 #endif /* HAVE_S_INSTANCES_LIST_HEAD */
 364 }
 365
 366 void
 367 zpl_prune_sb(int64_t nr_to_scan, void *arg)
 368 {
 369         struct super_block *sb = (struct super_block *)arg;
 370         int objects = 0;
 371
 372         (void) -zfs_prune(sb, nr_to_scan, &objects);
 373 }
 374
 375 #ifdef HAVE_NR_CACHED_OBJECTS
 376 static int
 377 zpl_nr_cached_objects(struct super_block *sb)
 378 {
 379         return (0);
 380 }
 381 #endif /* HAVE_NR_CACHED_OBJECTS */
 382
 383 #ifdef HAVE_FREE_CACHED_OBJECTS
 384 static void
 385 zpl_free_cached_objects(struct super_block *sb, int nr_to_scan)
 386 {
 387         /* noop */
 388 }
 389 #endif /* HAVE_FREE_CACHED_OBJECTS */
 390
 391 const struct super_operations zpl_super_operations = {
 392         .alloc_inode            = zpl_inode_alloc,
 393         .destroy_inode          = zpl_inode_destroy,
 394         .dirty_inode            = zpl_dirty_inode,
 395         .write_inode            = NULL,
 396 #ifdef HAVE_EVICT_INODE
 397         .evict_inode            = zpl_evict_inode,
 398 #else
 399         .drop_inode             = zpl_drop_inode,
 400         .clear_inode            = zpl_clear_inode,
 401         .delete_inode           = zpl_inode_delete,
 402 #endif /* HAVE_EVICT_INODE */
 403         .put_super              = zpl_put_super,
 404         .sync_fs                = zpl_sync_fs,
 405         .statfs                 = zpl_statfs,
 406         .remount_fs             = zpl_remount_fs,
 407         .show_options           = zpl_show_options,
 408         .show_stats             = NULL,
 409 #ifdef HAVE_NR_CACHED_OBJECTS
 410         .nr_cached_objects      = zpl_nr_cached_objects,
 411 #endif /* HAVE_NR_CACHED_OBJECTS */
 412 #ifdef HAVE_FREE_CACHED_OBJECTS
 413         .free_cached_objects    = zpl_free_cached_objects,
 414 #endif /* HAVE_FREE_CACHED_OBJECTS */
 415 };
 416
 417 struct file_system_type zpl_fs_type = {
 418         .owner                  = THIS_MODULE,
 419         .name                   = ZFS_DRIVER,
 420 #ifdef HAVE_FST_MOUNT
 421         .mount                  = zpl_mount,
 422 #else
 423         .get_sb                 = zpl_get_sb,
 424 #endif /* HAVE_FST_MOUNT */
 425         .kill_sb                = zpl_kill_sb,
 426 };