module/zfs/zfs_sa.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) 2010, Oracle and/or its affiliates. All rights reserved.
  23  */
  24
  25 #include <sys/types.h>
  26 #include <sys/param.h>
  27 #include <sys/vnode.h>
  28 #include <sys/sa.h>
  29 #include <sys/zfs_acl.h>
  30 #include <sys/zfs_sa.h>
  31
  32 /*
  33  * ZPL attribute registration table.
  34  * Order of attributes doesn't matter
  35  * a unique value will be assigned for each
  36  * attribute that is file system specific
  37  *
  38  * This is just the set of ZPL attributes that this
  39  * version of ZFS deals with natively.  The file system
  40  * could have other attributes stored in files, but they will be
  41  * ignored.  The SA framework will preserve them, just that
  42  * this version of ZFS won't change or delete them.
  43  */
  44
  45 sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
  46         {"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
  47         {"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
  48         {"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
  49         {"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
  50         {"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
  51         {"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
  52         {"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
  53         {"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
  54         {"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
  55         {"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
  56         {"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
  57         {"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
  58         {"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
  59         {"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
  60         {"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
  61         {"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
  62         {"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
  63         {"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0},
  64         {"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0},
  65         {"ZPL_DACL_ACES", 0, SA_ACL, 0},
  66         {"ZPL_DXATTR", 0, SA_UINT8_ARRAY, 0},
  67         {NULL, 0, 0, 0}
  68 };
  69
  70 #ifdef _KERNEL
  71 int
  72 zfs_sa_readlink(znode_t *zp, uio_t *uio)
  73 {
  74         dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
  75         size_t bufsz;
  76         int error;
  77
  78         bufsz = zp->z_size;
  79         if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
  80                 error = uiomove((caddr_t)db->db_data +
  81                     ZFS_OLD_ZNODE_PHYS_SIZE,
  82                     MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
  83         } else {
  84                 dmu_buf_t *dbp;
  85                 if ((error = dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id,
  86                     0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
  87                         error = uiomove(dbp->db_data,
  88                             MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
  89                         dmu_buf_rele(dbp, FTAG);
  90                 }
  91         }
  92         return (error);
  93 }
  94
  95 void
  96 zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
  97 {
  98         dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
  99
 100         if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
 101                 VERIFY(dmu_set_bonus(db,
 102                     len + ZFS_OLD_ZNODE_PHYS_SIZE, tx) == 0);
 103                 if (len) {
 104                         bcopy(link, (caddr_t)db->db_data +
 105                             ZFS_OLD_ZNODE_PHYS_SIZE, len);
 106                 }
 107         } else {
 108                 dmu_buf_t *dbp;
 109
 110                 zfs_grow_blocksize(zp, len, tx);
 111                 VERIFY(0 == dmu_buf_hold(ZTOZSB(zp)->z_os,
 112                     zp->z_id, 0, FTAG, &dbp, DMU_READ_NO_PREFETCH));
 113
 114                 dmu_buf_will_dirty(dbp, tx);
 115
 116                 ASSERT3U(len, <=, dbp->db_size);
 117                 bcopy(link, dbp->db_data, len);
 118                 dmu_buf_rele(dbp, FTAG);
 119         }
 120 }
 121
 122 void
 123 zfs_sa_get_scanstamp(znode_t *zp, xvattr_t *xvap)
 124 {
 125         zfs_sb_t *zsb = ZTOZSB(zp);
 126         xoptattr_t *xoap;
 127
 128         ASSERT(MUTEX_HELD(&zp->z_lock));
 129         VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
 130         if (zp->z_is_sa) {
 131                 if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
 132                     &xoap->xoa_av_scanstamp,
 133                     sizeof (xoap->xoa_av_scanstamp)) != 0)
 134                         return;
 135         } else {
 136                 dmu_object_info_t doi;
 137                 dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
 138                 int len;
 139
 140                 if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP))
 141                         return;
 142
 143                 sa_object_info(zp->z_sa_hdl, &doi);
 144                 len = sizeof (xoap->xoa_av_scanstamp) +
 145                     ZFS_OLD_ZNODE_PHYS_SIZE;
 146
 147                 if (len <= doi.doi_bonus_size) {
 148                         (void) memcpy(xoap->xoa_av_scanstamp,
 149                             (caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
 150                             sizeof (xoap->xoa_av_scanstamp));
 151                 }
 152         }
 153         XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
 154 }
 155
 156 void
 157 zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
 158 {
 159         zfs_sb_t *zsb = ZTOZSB(zp);
 160         xoptattr_t *xoap;
 161
 162         ASSERT(MUTEX_HELD(&zp->z_lock));
 163         VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
 164         if (zp->z_is_sa)
 165                 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
 166                     &xoap->xoa_av_scanstamp,
 167                     sizeof (xoap->xoa_av_scanstamp), tx));
 168         else {
 169                 dmu_object_info_t doi;
 170                 dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
 171                 int len;
 172
 173                 sa_object_info(zp->z_sa_hdl, &doi);
 174                 len = sizeof (xoap->xoa_av_scanstamp) +
 175                     ZFS_OLD_ZNODE_PHYS_SIZE;
 176                 if (len > doi.doi_bonus_size)
 177                         VERIFY(dmu_set_bonus(db, len, tx) == 0);
 178                 (void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
 179                     xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp));
 180
 181                 zp->z_pflags |= ZFS_BONUS_SCANSTAMP;
 182                 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zsb),
 183                     &zp->z_pflags, sizeof (uint64_t), tx));
 184         }
 185 }
 186
 187 int
 188 zfs_sa_get_xattr(znode_t *zp)
 189 {
 190         zfs_sb_t *zsb = ZTOZSB(zp);
 191         char *obj;
 192         int size;
 193         int error;
 194
 195         ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
 196         ASSERT(!zp->z_xattr_cached);
 197         ASSERT(zp->z_is_sa);
 198
 199         error = sa_size(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb), &size);
 200         if (error) {
 201                 if (error == ENOENT)
 202                         return nvlist_alloc(&zp->z_xattr_cached,
 203                             NV_UNIQUE_NAME, KM_SLEEP);
 204                 else
 205                         return (error);
 206         }
 207
 208         obj = zio_buf_alloc(size);
 209
 210         error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb), obj, size);
 211         if (error == 0)
 212                 error = nvlist_unpack(obj, size, &zp->z_xattr_cached, KM_SLEEP);
 213
 214         zio_buf_free(obj, size);
 215
 216         return (error);
 217 }
 218
 219 int
 220 zfs_sa_set_xattr(znode_t *zp)
 221 {
 222         zfs_sb_t *zsb = ZTOZSB(zp);
 223         dmu_tx_t *tx;
 224         char *obj;
 225         size_t size;
 226         int error;
 227
 228         ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
 229         ASSERT(zp->z_xattr_cached);
 230         ASSERT(zp->z_is_sa);
 231
 232         error = nvlist_size(zp->z_xattr_cached, &size, NV_ENCODE_XDR);
 233         if (error)
 234                 goto out;
 235
 236         obj = zio_buf_alloc(size);
 237
 238         error = nvlist_pack(zp->z_xattr_cached, &obj, &size,
 239             NV_ENCODE_XDR, KM_SLEEP);
 240         if (error)
 241                 goto out_free;
 242
 243         tx = dmu_tx_create(zsb->z_os);
 244         dmu_tx_hold_sa_create(tx, size);
 245         dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
 246
 247         error = dmu_tx_assign(tx, TXG_WAIT);
 248         if (error) {
 249                 dmu_tx_abort(tx);
 250         } else {
 251                 error = sa_update(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb),
 252                     obj, size, tx);
 253                 if (error)
 254                         dmu_tx_abort(tx);
 255                 else
 256                         dmu_tx_commit(tx);
 257         }
 258 out_free:
 259         zio_buf_free(obj, size);
 260 out:
 261         return (error);
 262 }
 263
 264 /*
 265  * I'm not convinced we should do any of this upgrade.
 266  * since the SA code can read both old/new znode formats
 267  * with probably little to no performance difference.
 268  *
 269  * All new files will be created with the new format.
 270  */
 271
 272 void
 273 zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
 274 {
 275         dmu_buf_t *db = sa_get_db(hdl);
 276         znode_t *zp = sa_get_userdata(hdl);
 277         zfs_sb_t *zsb = ZTOZSB(zp);
 278         int count = 0;
 279         sa_bulk_attr_t *bulk, *sa_attrs;
 280         zfs_acl_locator_cb_t locate = { 0 };
 281         uint64_t uid, gid, mode, rdev, xattr, parent;
 282         uint64_t crtime[2], mtime[2], ctime[2];
 283         zfs_acl_phys_t znode_acl;
 284         char scanstamp[AV_SCANSTAMP_SZ];
 285         boolean_t drop_lock = B_FALSE;
 286
 287         /*
 288          * No upgrade if ACL isn't cached
 289          * since we won't know which locks are held
 290          * and ready the ACL would require special "locked"
 291          * interfaces that would be messy
 292          */
 293         if (zp->z_acl_cached == NULL || S_ISLNK(ZTOI(zp)->i_mode))
 294                 return;
 295
 296         /*
 297          * If the z_lock is held and we aren't the owner
 298          * the just return since we don't want to deadlock
 299          * trying to update the status of z_is_sa.  This
 300          * file can then be upgraded at a later time.
 301          *
 302          * Otherwise, we know we are doing the
 303          * sa_update() that caused us to enter this function.
 304          */
 305         if (mutex_owner(&zp->z_lock) != curthread) {
 306                 if (mutex_tryenter(&zp->z_lock) == 0)
 307                         return;
 308                 else
 309                         drop_lock = B_TRUE;
 310         }
 311
 312         /* First do a bulk query of the attributes that aren't cached */
 313         bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
 314         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
 315         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
 316         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zsb), NULL, &crtime, 16);
 317         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
 318         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, &parent, 8);
 319         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zsb), NULL, &xattr, 8);
 320         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zsb), NULL, &rdev, 8);
 321         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
 322         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
 323         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
 324             &znode_acl, 88);
 325
 326         if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) {
 327                 kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20);
 328                 goto done;
 329         }
 330
 331         /*
 332          * While the order here doesn't matter its best to try and organize
 333          * it is such a way to pick up an already existing layout number
 334          */
 335         count = 0;
 336         sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
 337         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
 338         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zsb), NULL,
 339             &zp->z_size, 8);
 340         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zsb),
 341             NULL, &zp->z_gen, 8);
 342         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
 343         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
 344         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zsb),
 345             NULL, &parent, 8);
 346         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zsb), NULL,
 347             &zp->z_pflags, 8);
 348         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zsb), NULL,
 349             zp->z_atime, 16);
 350         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zsb), NULL,
 351             &mtime, 16);
 352         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zsb), NULL,
 353             &ctime, 16);
 354         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zsb), NULL,
 355             &crtime, 16);
 356         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zsb), NULL,
 357             &zp->z_links, 8);
 358         if (S_ISBLK(ZTOI(zp)->i_mode) || S_ISCHR(ZTOI(zp)->i_mode))
 359                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zsb), NULL,
 360                     &rdev, 8);
 361         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zsb), NULL,
 362             &zp->z_acl_cached->z_acl_count, 8);
 363
 364         if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
 365                 zfs_acl_xform(zp, zp->z_acl_cached, CRED());
 366
 367         locate.cb_aclp = zp->z_acl_cached;
 368         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zsb),
 369             zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
 370
 371         if (xattr)
 372                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zsb),
 373                     NULL, &xattr, 8);
 374
 375         /* if scanstamp then add scanstamp */
 376
 377         if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
 378                 bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
 379                     scanstamp, AV_SCANSTAMP_SZ);
 380                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zsb),
 381                     NULL, scanstamp, AV_SCANSTAMP_SZ);
 382                 zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
 383         }
 384
 385         VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
 386         VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
 387             count, tx) == 0);
 388         if (znode_acl.z_acl_extern_obj)
 389                 VERIFY(0 == dmu_object_free(zsb->z_os,
 390                     znode_acl.z_acl_extern_obj, tx));
 391
 392         zp->z_is_sa = B_TRUE;
 393         kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * 20);
 394         kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20);
 395 done:
 396         if (drop_lock)
 397                 mutex_exit(&zp->z_lock);
 398 }
 399
 400 void
 401 zfs_sa_upgrade_txholds(dmu_tx_t *tx, znode_t *zp)
 402 {
 403         if (!ZTOZSB(zp)->z_use_sa || zp->z_is_sa)
 404                 return;
 405
 406
 407         dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
 408
 409         if (zfs_external_acl(zp)) {
 410                 dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
 411                     DMU_OBJECT_END);
 412         }
 413 }
 414
 415 EXPORT_SYMBOL(zfs_attr_table);
 416 EXPORT_SYMBOL(zfs_sa_readlink);
 417 EXPORT_SYMBOL(zfs_sa_symlink);
 418 EXPORT_SYMBOL(zfs_sa_get_scanstamp);
 419 EXPORT_SYMBOL(zfs_sa_set_scanstamp);
 420 EXPORT_SYMBOL(zfs_sa_get_xattr);
 421 EXPORT_SYMBOL(zfs_sa_set_xattr);
 422 EXPORT_SYMBOL(zfs_sa_upgrade);
 423 EXPORT_SYMBOL(zfs_sa_upgrade_txholds);
 424
 425 #endif