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/zfs_context.h>
  26 #include <sys/vnode.h>
  27 #include <sys/sa.h>
  28 #include <sys/zfs_acl.h>
  29 #include <sys/zfs_sa.h>
  30 #include <sys/dmu_objset.h>
  31 #include <sys/sa_impl.h>
  32
  33 /*
  34  * ZPL attribute registration table.
  35  * Order of attributes doesn't matter
  36  * a unique value will be assigned for each
  37  * attribute that is file system specific
  38  *
  39  * This is just the set of ZPL attributes that this
  40  * version of ZFS deals with natively.  The file system
  41  * could have other attributes stored in files, but they will be
  42  * ignored.  The SA framework will preserve them, just that
  43  * this version of ZFS won't change or delete them.
  44  */
  45
  46 sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
  47         {"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
  48         {"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
  49         {"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
  50         {"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
  51         {"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
  52         {"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
  53         {"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
  54         {"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
  55         {"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
  56         {"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
  57         {"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
  58         {"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
  59         {"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
  60         {"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
  61         {"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
  62         {"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
  63         {"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
  64         {"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0},
  65         {"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0},
  66         {"ZPL_DACL_ACES", 0, SA_ACL, 0},
  67         {"ZPL_DXATTR", 0, SA_UINT8_ARRAY, 0},
  68         {"ZPL_PROJID", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
  69         {NULL, 0, 0, 0}
  70 };
  71
  72 #ifdef _KERNEL
  73 int
  74 zfs_sa_readlink(znode_t *zp, zfs_uio_t *uio)
  75 {
  76         dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
  77         size_t bufsz;
  78         int error;
  79
  80         bufsz = zp->z_size;
  81         if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
  82                 error = zfs_uiomove((caddr_t)db->db_data +
  83                     ZFS_OLD_ZNODE_PHYS_SIZE,
  84                     MIN((size_t)bufsz, zfs_uio_resid(uio)), UIO_READ, uio);
  85         } else {
  86                 dmu_buf_t *dbp;
  87                 if ((error = dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id,
  88                     0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
  89                         error = zfs_uiomove(dbp->db_data,
  90                             MIN((size_t)bufsz, zfs_uio_resid(uio)), UIO_READ,
  91                             uio);
  92                         dmu_buf_rele(dbp, FTAG);
  93                 }
  94         }
  95         return (error);
  96 }
  97
  98 void
  99 zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
 100 {
 101         dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
 102
 103         if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
 104                 VERIFY0(dmu_set_bonus(db, len + ZFS_OLD_ZNODE_PHYS_SIZE, tx));
 105                 if (len) {
 106                         bcopy(link, (caddr_t)db->db_data +
 107                             ZFS_OLD_ZNODE_PHYS_SIZE, len);
 108                 }
 109         } else {
 110                 dmu_buf_t *dbp;
 111
 112                 zfs_grow_blocksize(zp, len, tx);
 113                 VERIFY0(dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id, 0, FTAG, &dbp,
 114                     DMU_READ_NO_PREFETCH));
 115
 116                 dmu_buf_will_dirty(dbp, tx);
 117
 118                 ASSERT3U(len, <=, dbp->db_size);
 119                 bcopy(link, dbp->db_data, len);
 120                 dmu_buf_rele(dbp, FTAG);
 121         }
 122 }
 123
 124 void
 125 zfs_sa_get_scanstamp(znode_t *zp, xvattr_t *xvap)
 126 {
 127         zfsvfs_t *zfsvfs = ZTOZSB(zp);
 128         xoptattr_t *xoap;
 129
 130         ASSERT(MUTEX_HELD(&zp->z_lock));
 131         VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
 132         if (zp->z_is_sa) {
 133                 if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
 134                     &xoap->xoa_av_scanstamp,
 135                     sizeof (xoap->xoa_av_scanstamp)) != 0)
 136                         return;
 137         } else {
 138                 dmu_object_info_t doi;
 139                 dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
 140                 int len;
 141
 142                 if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP))
 143                         return;
 144
 145                 sa_object_info(zp->z_sa_hdl, &doi);
 146                 len = sizeof (xoap->xoa_av_scanstamp) +
 147                     ZFS_OLD_ZNODE_PHYS_SIZE;
 148
 149                 if (len <= doi.doi_bonus_size) {
 150                         (void) memcpy(xoap->xoa_av_scanstamp,
 151                             (caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
 152                             sizeof (xoap->xoa_av_scanstamp));
 153                 }
 154         }
 155         XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
 156 }
 157
 158 void
 159 zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
 160 {
 161         zfsvfs_t *zfsvfs = ZTOZSB(zp);
 162         xoptattr_t *xoap;
 163
 164         ASSERT(MUTEX_HELD(&zp->z_lock));
 165         VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
 166         if (zp->z_is_sa)
 167                 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
 168                     &xoap->xoa_av_scanstamp,
 169                     sizeof (xoap->xoa_av_scanstamp), tx));
 170         else {
 171                 dmu_object_info_t doi;
 172                 dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
 173                 int len;
 174
 175                 sa_object_info(zp->z_sa_hdl, &doi);
 176                 len = sizeof (xoap->xoa_av_scanstamp) +
 177                     ZFS_OLD_ZNODE_PHYS_SIZE;
 178                 if (len > doi.doi_bonus_size)
 179                         VERIFY(dmu_set_bonus(db, len, tx) == 0);
 180                 (void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
 181                     xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp));
 182
 183                 zp->z_pflags |= ZFS_BONUS_SCANSTAMP;
 184                 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
 185                     &zp->z_pflags, sizeof (uint64_t), tx));
 186         }
 187 }
 188
 189 int
 190 zfs_sa_get_xattr(znode_t *zp)
 191 {
 192         zfsvfs_t *zfsvfs = ZTOZSB(zp);
 193         char *obj;
 194         int size;
 195         int error;
 196
 197         ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
 198         ASSERT(!zp->z_xattr_cached);
 199         ASSERT(zp->z_is_sa);
 200
 201         error = sa_size(zp->z_sa_hdl, SA_ZPL_DXATTR(zfsvfs), &size);
 202         if (error) {
 203                 if (error == ENOENT)
 204                         return nvlist_alloc(&zp->z_xattr_cached,
 205                             NV_UNIQUE_NAME, KM_SLEEP);
 206                 else
 207                         return (error);
 208         }
 209
 210         obj = vmem_alloc(size, KM_SLEEP);
 211
 212         error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DXATTR(zfsvfs), obj, size);
 213         if (error == 0)
 214                 error = nvlist_unpack(obj, size, &zp->z_xattr_cached, KM_SLEEP);
 215
 216         vmem_free(obj, size);
 217
 218         return (error);
 219 }
 220
 221 int
 222 zfs_sa_set_xattr(znode_t *zp)
 223 {
 224         zfsvfs_t *zfsvfs = ZTOZSB(zp);
 225         dmu_tx_t *tx;
 226         char *obj;
 227         size_t size;
 228         int error;
 229
 230         ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
 231         ASSERT(zp->z_xattr_cached);
 232         ASSERT(zp->z_is_sa);
 233
 234         error = nvlist_size(zp->z_xattr_cached, &size, NV_ENCODE_XDR);
 235         if ((error == 0) && (size > SA_ATTR_MAX_LEN))
 236                 error = SET_ERROR(EFBIG);
 237         if (error)
 238                 goto out;
 239
 240         obj = vmem_alloc(size, KM_SLEEP);
 241
 242         error = nvlist_pack(zp->z_xattr_cached, &obj, &size,
 243             NV_ENCODE_XDR, KM_SLEEP);
 244         if (error)
 245                 goto out_free;
 246
 247         tx = dmu_tx_create(zfsvfs->z_os);
 248         dmu_tx_hold_sa_create(tx, size);
 249         dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
 250
 251         error = dmu_tx_assign(tx, TXG_WAIT);
 252         if (error) {
 253                 dmu_tx_abort(tx);
 254         } else {
 255                 int count = 0;
 256                 sa_bulk_attr_t bulk[2];
 257                 uint64_t ctime[2];
 258
 259                 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
 260                 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DXATTR(zfsvfs),
 261                     NULL, obj, size);
 262                 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
 263                     NULL, &ctime, 16);
 264                 VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
 265
 266                 dmu_tx_commit(tx);
 267         }
 268 out_free:
 269         vmem_free(obj, size);
 270 out:
 271         return (error);
 272 }
 273
 274 /*
 275  * I'm not convinced we should do any of this upgrade.
 276  * since the SA code can read both old/new znode formats
 277  * with probably little to no performance difference.
 278  *
 279  * All new files will be created with the new format.
 280  */
 281
 282 void
 283 zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
 284 {
 285         dmu_buf_t *db = sa_get_db(hdl);
 286         znode_t *zp = sa_get_userdata(hdl);
 287         zfsvfs_t *zfsvfs = ZTOZSB(zp);
 288         int count = 0;
 289         sa_bulk_attr_t *bulk, *sa_attrs;
 290         zfs_acl_locator_cb_t locate = { 0 };
 291         uint64_t uid, gid, mode, rdev, xattr, parent, tmp_gen;
 292         uint64_t crtime[2], mtime[2], ctime[2], atime[2];
 293         uint64_t links;
 294         zfs_acl_phys_t znode_acl;
 295         char scanstamp[AV_SCANSTAMP_SZ];
 296         boolean_t drop_lock = B_FALSE;
 297
 298         /*
 299          * No upgrade if ACL isn't cached
 300          * since we won't know which locks are held
 301          * and ready the ACL would require special "locked"
 302          * interfaces that would be messy
 303          */
 304         if (zp->z_acl_cached == NULL || Z_ISLNK(ZTOTYPE(zp)))
 305                 return;
 306
 307         /*
 308          * If the z_lock is held and we aren't the owner
 309          * the just return since we don't want to deadlock
 310          * trying to update the status of z_is_sa.  This
 311          * file can then be upgraded at a later time.
 312          *
 313          * Otherwise, we know we are doing the
 314          * sa_update() that caused us to enter this function.
 315          */
 316         if (MUTEX_NOT_HELD(&zp->z_lock)) {
 317                 if (mutex_tryenter(&zp->z_lock) == 0)
 318                         return;
 319                 else
 320                         drop_lock = B_TRUE;
 321         }
 322
 323         /* First do a bulk query of the attributes that aren't cached */
 324         bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
 325         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16);
 326         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
 327         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
 328         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
 329         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
 330         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
 331         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8);
 332         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8);
 333         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
 334         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
 335         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8);
 336         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
 337             &znode_acl, 88);
 338
 339         if (sa_bulk_lookup_locked(hdl, bulk, count) != 0)
 340                 goto done;
 341
 342         if (dmu_objset_projectquota_enabled(hdl->sa_os) &&
 343             !(zp->z_pflags & ZFS_PROJID)) {
 344                 zp->z_pflags |= ZFS_PROJID;
 345                 zp->z_projid = ZFS_DEFAULT_PROJID;
 346         }
 347
 348         /*
 349          * While the order here doesn't matter its best to try and organize
 350          * it is such a way to pick up an already existing layout number
 351          */
 352         count = 0;
 353         sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
 354         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
 355         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
 356             &zp->z_size, 8);
 357         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs),
 358             NULL, &tmp_gen, 8);
 359         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
 360         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
 361         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs),
 362             NULL, &parent, 8);
 363         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
 364             &zp->z_pflags, 8);
 365         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL,
 366             &atime, 16);
 367         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL,
 368             &mtime, 16);
 369         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL,
 370             &ctime, 16);
 371         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
 372             &crtime, 16);
 373         links = ZTONLNK(zp);
 374         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL,
 375             &links, 8);
 376         if (dmu_objset_projectquota_enabled(hdl->sa_os))
 377                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PROJID(zfsvfs), NULL,
 378                     &zp->z_projid, 8);
 379         if (Z_ISBLK(ZTOTYPE(zp)) || Z_ISCHR(ZTOTYPE(zp)))
 380                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
 381                     &rdev, 8);
 382         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
 383             &zp->z_acl_cached->z_acl_count, 8);
 384
 385         if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
 386                 zfs_acl_xform(zp, zp->z_acl_cached, CRED());
 387
 388         locate.cb_aclp = zp->z_acl_cached;
 389         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
 390             zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
 391
 392         if (xattr)
 393                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zfsvfs),
 394                     NULL, &xattr, 8);
 395
 396         /* if scanstamp then add scanstamp */
 397
 398         if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
 399                 bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
 400                     scanstamp, AV_SCANSTAMP_SZ);
 401                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zfsvfs),
 402                     NULL, scanstamp, AV_SCANSTAMP_SZ);
 403                 zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
 404         }
 405
 406         VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
 407         VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
 408             count, tx) == 0);
 409         if (znode_acl.z_acl_extern_obj)
 410                 VERIFY(0 == dmu_object_free(zfsvfs->z_os,
 411                     znode_acl.z_acl_extern_obj, tx));
 412
 413         zp->z_is_sa = B_TRUE;
 414         kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
 415 done:
 416         kmem_free(bulk, sizeof (sa_bulk_attr_t) * ZPL_END);
 417         if (drop_lock)
 418                 mutex_exit(&zp->z_lock);
 419 }
 420
 421 void
 422 zfs_sa_upgrade_txholds(dmu_tx_t *tx, znode_t *zp)
 423 {
 424         if (!ZTOZSB(zp)->z_use_sa || zp->z_is_sa)
 425                 return;
 426
 427
 428         dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
 429
 430         if (zfs_external_acl(zp)) {
 431                 dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
 432                     DMU_OBJECT_END);
 433         }
 434 }
 435
 436 EXPORT_SYMBOL(zfs_attr_table);
 437 EXPORT_SYMBOL(zfs_sa_readlink);
 438 EXPORT_SYMBOL(zfs_sa_symlink);
 439 EXPORT_SYMBOL(zfs_sa_get_scanstamp);
 440 EXPORT_SYMBOL(zfs_sa_set_scanstamp);
 441 EXPORT_SYMBOL(zfs_sa_get_xattr);
 442 EXPORT_SYMBOL(zfs_sa_set_xattr);
 443 EXPORT_SYMBOL(zfs_sa_upgrade);
 444 EXPORT_SYMBOL(zfs_sa_upgrade_txholds);
 445
 446 #endif