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         {NULL, 0, 0, 0}
  67 };
  68
  69 #ifdef _KERNEL
  70 int
  71 zfs_sa_readlink(znode_t *zp, uio_t *uio)
  72 {
  73         dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
  74         size_t bufsz;
  75         int error;
  76
  77         bufsz = zp->z_size;
  78         if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
  79                 error = uiomove((caddr_t)db->db_data +
  80                     ZFS_OLD_ZNODE_PHYS_SIZE,
  81                     MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
  82         } else {
  83                 dmu_buf_t *dbp;
  84                 if ((error = dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id,
  85                     0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
  86                         error = uiomove(dbp->db_data,
  87                             MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
  88                         dmu_buf_rele(dbp, FTAG);
  89                 }
  90         }
  91         return (error);
  92 }
  93
  94 void
  95 zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
  96 {
  97         dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
  98
  99         if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
 100                 VERIFY(dmu_set_bonus(db,
 101                     len + ZFS_OLD_ZNODE_PHYS_SIZE, tx) == 0);
 102                 if (len) {
 103                         bcopy(link, (caddr_t)db->db_data +
 104                             ZFS_OLD_ZNODE_PHYS_SIZE, len);
 105                 }
 106         } else {
 107                 dmu_buf_t *dbp;
 108
 109                 zfs_grow_blocksize(zp, len, tx);
 110                 VERIFY(0 == dmu_buf_hold(ZTOZSB(zp)->z_os,
 111                     zp->z_id, 0, FTAG, &dbp, DMU_READ_NO_PREFETCH));
 112
 113                 dmu_buf_will_dirty(dbp, tx);
 114
 115                 ASSERT3U(len, <=, dbp->db_size);
 116                 bcopy(link, dbp->db_data, len);
 117                 dmu_buf_rele(dbp, FTAG);
 118         }
 119 }
 120
 121 #ifdef HAVE_SCANSTAMP
 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 #endif /* HAVE_SCANSTAMP */
 187
 188 /*
 189  * I'm not convinced we should do any of this upgrade.
 190  * since the SA code can read both old/new znode formats
 191  * with probably little to know performance difference.
 192  *
 193  * All new files will be created with the new format.
 194  */
 195
 196 void
 197 zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
 198 {
 199         dmu_buf_t *db = sa_get_db(hdl);
 200         znode_t *zp = sa_get_userdata(hdl);
 201         zfs_sb_t *zsb = ZTOZSB(zp);
 202         int count = 0;
 203         sa_bulk_attr_t *bulk, *sa_attrs;
 204         zfs_acl_locator_cb_t locate = { 0 };
 205         uint64_t uid, gid, mode, rdev, xattr, parent;
 206         uint64_t crtime[2], mtime[2], ctime[2];
 207         zfs_acl_phys_t znode_acl;
 208 #ifdef HAVE_SCANSTAMP
 209         char scanstamp[AV_SCANSTAMP_SZ];
 210 #endif /* HAVE_SCANSTAMP */
 211         boolean_t drop_lock = B_FALSE;
 212
 213         /*
 214          * No upgrade if ACL isn't cached
 215          * since we won't know which locks are held
 216          * and ready the ACL would require special "locked"
 217          * interfaces that would be messy
 218          */
 219         if (zp->z_acl_cached == NULL || S_ISLNK(ZTOI(zp)->i_mode))
 220                 return;
 221
 222         /*
 223          * If the z_lock is held and we aren't the owner
 224          * the just return since we don't want to deadlock
 225          * trying to update the status of z_is_sa.  This
 226          * file can then be upgraded at a later time.
 227          *
 228          * Otherwise, we know we are doing the
 229          * sa_update() that caused us to enter this function.
 230          */
 231         if (mutex_owner(&zp->z_lock) != curthread) {
 232                 if (mutex_tryenter(&zp->z_lock) == 0)
 233                         return;
 234                 else
 235                         drop_lock = B_TRUE;
 236         }
 237
 238         /* First do a bulk query of the attributes that aren't cached */
 239         bulk = kmem_alloc(sizeof(sa_bulk_attr_t) * 20, KM_SLEEP);
 240         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
 241         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
 242         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zsb), NULL, &crtime, 16);
 243         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
 244         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, &parent, 8);
 245         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zsb), NULL, &xattr, 8);
 246         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zsb), NULL, &rdev, 8);
 247         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
 248         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
 249         SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
 250             &znode_acl, 88);
 251
 252         if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) {
 253                 kmem_free(bulk, sizeof(sa_bulk_attr_t) * 20);
 254                 goto done;
 255         }
 256
 257         /*
 258          * While the order here doesn't matter its best to try and organize
 259          * it is such a way to pick up an already existing layout number
 260          */
 261         count = 0;
 262         sa_attrs = kmem_zalloc(sizeof(sa_bulk_attr_t) * 20, KM_SLEEP);
 263         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
 264         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zsb), NULL,
 265             &zp->z_size, 8);
 266         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zsb),
 267             NULL, &zp->z_gen, 8);
 268         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
 269         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
 270         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zsb),
 271             NULL, &parent, 8);
 272         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zsb), NULL,
 273             &zp->z_pflags, 8);
 274         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zsb), NULL,
 275             zp->z_atime, 16);
 276         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zsb), NULL,
 277             &mtime, 16);
 278         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zsb), NULL,
 279             &ctime, 16);
 280         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zsb), NULL,
 281             &crtime, 16);
 282         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zsb), NULL,
 283             &zp->z_links, 8);
 284         if (S_ISBLK(ZTOI(zp)->i_mode) || S_ISCHR(ZTOI(zp)->i_mode))
 285                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zsb), NULL,
 286                     &rdev, 8);
 287         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zsb), NULL,
 288             &zp->z_acl_cached->z_acl_count, 8);
 289
 290         if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
 291                 zfs_acl_xform(zp, zp->z_acl_cached, CRED());
 292
 293         locate.cb_aclp = zp->z_acl_cached;
 294         SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zsb),
 295             zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
 296
 297         if (xattr)
 298                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zsb),
 299                     NULL, &xattr, 8);
 300
 301 #ifdef HAVE_SCANSTAMP
 302         /* if scanstamp then add scanstamp */
 303
 304         if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
 305                 bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
 306                     scanstamp, AV_SCANSTAMP_SZ);
 307                 SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zsb),
 308                     NULL, scanstamp, AV_SCANSTAMP_SZ);
 309                 zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
 310         }
 311 #endif /* HAVE_SCANSTAMP */
 312
 313         VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
 314         VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
 315             count, tx) == 0);
 316         if (znode_acl.z_acl_extern_obj)
 317                 VERIFY(0 == dmu_object_free(zsb->z_os,
 318                     znode_acl.z_acl_extern_obj, tx));
 319
 320         zp->z_is_sa = B_TRUE;
 321         kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * 20);
 322         kmem_free(bulk, sizeof(sa_bulk_attr_t) * 20);
 323 done:
 324         if (drop_lock)
 325                 mutex_exit(&zp->z_lock);
 326 }
 327
 328 void
 329 zfs_sa_upgrade_txholds(dmu_tx_t *tx, znode_t *zp)
 330 {
 331         if (!ZTOZSB(zp)->z_use_sa || zp->z_is_sa)
 332                 return;
 333
 334
 335         dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
 336
 337         if (zfs_external_acl(zp)) {
 338                 dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
 339                     DMU_OBJECT_END);
 340         }
 341 }
 342
 343 #endif