Update core ZFS code from build 121 to build 141.
[zfs.git] / module / zfs / zfs_sa.c
diff --git a/module/zfs/zfs_sa.c b/module/zfs/zfs_sa.c
new file mode 100644 (file)
index 0000000..73a40aa
--- /dev/null
@@ -0,0 +1,312 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
+ */
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/vnode.h>
+#include <sys/sa.h>
+#include <sys/zfs_acl.h>
+#include <sys/zfs_sa.h>
+
+/*
+ * ZPL attribute registration table.
+ * Order of attributes doesn't matter
+ * a unique value will be assigned for each
+ * attribute that is file system specific
+ *
+ * This is just the set of ZPL attributes that this
+ * version of ZFS deals with natively.  The file system
+ * could have other attributes stored in files, but they will be
+ * ignored.  The SA framework will preserve them, just that
+ * this version of ZFS won't change or delete them.
+ */
+
+sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
+       {"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
+       {"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
+       {"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
+       {"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
+       {"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
+       {"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
+       {"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
+       {"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
+       {"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
+       {"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
+       {"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
+       {"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
+       {"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
+       {"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
+       {"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
+       {"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
+       {"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
+       {"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0},
+       {"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0},
+       {"ZPL_DACL_ACES", 0, SA_ACL, 0},
+       {NULL, 0, 0, 0}
+};
+
+#ifdef _KERNEL
+
+int
+zfs_sa_readlink(znode_t *zp, uio_t *uio)
+{
+       dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
+       size_t bufsz;
+       int error;
+
+       bufsz = zp->z_size;
+       if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
+               error = uiomove((caddr_t)db->db_data +
+                   ZFS_OLD_ZNODE_PHYS_SIZE,
+                   MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
+       } else {
+               dmu_buf_t *dbp;
+               if ((error = dmu_buf_hold(zp->z_zfsvfs->z_os, zp->z_id,
+                   0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
+                       error = uiomove(dbp->db_data,
+                           MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
+                       dmu_buf_rele(dbp, FTAG);
+               }
+       }
+       return (error);
+}
+
+void
+zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
+{
+       dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
+
+       if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
+               VERIFY(dmu_set_bonus(db,
+                   len + ZFS_OLD_ZNODE_PHYS_SIZE, tx) == 0);
+               if (len) {
+                       bcopy(link, (caddr_t)db->db_data +
+                           ZFS_OLD_ZNODE_PHYS_SIZE, len);
+               }
+       } else {
+               dmu_buf_t *dbp;
+
+               zfs_grow_blocksize(zp, len, tx);
+               VERIFY(0 == dmu_buf_hold(zp->z_zfsvfs->z_os,
+                   zp->z_id, 0, FTAG, &dbp, DMU_READ_NO_PREFETCH));
+
+               dmu_buf_will_dirty(dbp, tx);
+
+               ASSERT3U(len, <=, dbp->db_size);
+               bcopy(link, dbp->db_data, len);
+               dmu_buf_rele(dbp, FTAG);
+       }
+}
+
+void
+zfs_sa_get_scanstamp(znode_t *zp, xvattr_t *xvap)
+{
+       zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+       xoptattr_t *xoap;
+
+       VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
+       if (zp->z_is_sa) {
+               if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
+                   &xoap->xoa_av_scanstamp,
+                   sizeof (xoap->xoa_av_scanstamp)) != 0)
+                       return;
+       } else {
+               dmu_object_info_t doi;
+               dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
+               int len;
+
+               if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP))
+                       return;
+
+               sa_object_info(zp->z_sa_hdl, &doi);
+               len = sizeof (xoap->xoa_av_scanstamp) +
+                   ZFS_OLD_ZNODE_PHYS_SIZE;
+
+               if (len <= doi.doi_bonus_size) {
+                       (void) memcpy(xoap->xoa_av_scanstamp,
+                           (caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
+                           sizeof (xoap->xoa_av_scanstamp));
+               }
+       }
+       XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
+}
+
+void
+zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
+{
+       zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+       xoptattr_t *xoap;
+
+       VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
+       if (zp->z_is_sa)
+               VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
+                   &xoap->xoa_av_scanstamp,
+                   sizeof (xoap->xoa_av_scanstamp), tx));
+       else {
+               dmu_object_info_t doi;
+               dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
+               int len;
+
+               sa_object_info(zp->z_sa_hdl, &doi);
+               len = sizeof (xoap->xoa_av_scanstamp) +
+                   ZFS_OLD_ZNODE_PHYS_SIZE;
+               if (len > doi.doi_bonus_size)
+                       VERIFY(dmu_set_bonus(db, len, tx) == 0);
+               (void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
+                   xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp));
+
+               zp->z_pflags |= ZFS_BONUS_SCANSTAMP;
+               VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
+                   &zp->z_pflags, sizeof (uint64_t), tx));
+       }
+}
+
+/*
+ * I'm not convinced we should do any of this upgrade.
+ * since the SA code can read both old/new znode formats
+ * with probably little to know performance difference.
+ *
+ * All new files will be created with the new format.
+ */
+
+void
+zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
+{
+       dmu_buf_t *db = sa_get_db(hdl);
+       znode_t *zp = sa_get_userdata(hdl);
+       zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+       sa_bulk_attr_t bulk[20];
+       int count = 0;
+       sa_bulk_attr_t sa_attrs[20] = { 0 };
+       zfs_acl_locator_cb_t locate = { 0 };
+       uint64_t uid, gid, mode, rdev, xattr, parent;
+       uint64_t crtime[2], mtime[2], ctime[2];
+       zfs_acl_phys_t znode_acl;
+       char scanstamp[AV_SCANSTAMP_SZ];
+
+       /*
+        * No upgrade if ACL isn't cached
+        * since we won't know which locks are held
+        * and ready the ACL would require special "locked"
+        * interfaces that would be messy
+        */
+       if (zp->z_acl_cached == NULL || ZTOV(zp)->v_type == VLNK)
+               return;
+
+       /* First do a bulk query of the attributes that aren't cached */
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
+       SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
+           &znode_acl, 88);
+
+       if (sa_bulk_lookup_locked(hdl, bulk, count) != 0)
+               return;
+
+
+       /*
+        * While the order here doesn't matter its best to try and organize
+        * it is such a way to pick up an already existing layout number
+        */
+       count = 0;
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
+           &zp->z_size, 8);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs),
+           NULL, &zp->z_gen, 8);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs),
+           NULL, &parent, 8);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
+           &zp->z_pflags, 8);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL,
+           zp->z_atime, 16);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL,
+           &mtime, 16);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL,
+           &ctime, 16);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
+           &crtime, 16);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL,
+           &zp->z_links, 8);
+       if (zp->z_vnode->v_type == VBLK || zp->z_vnode->v_type == VCHR)
+               SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
+                   &rdev, 8);
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
+           &zp->z_acl_cached->z_acl_count, 8);
+
+       if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
+               zfs_acl_xform(zp, zp->z_acl_cached, CRED());
+
+       locate.cb_aclp = zp->z_acl_cached;
+       SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
+           zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
+       if (xattr)
+               SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs),
+                   NULL, &rdev, 8);
+
+       /* if scanstamp then add scanstamp */
+
+       if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
+               bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
+                   scanstamp, AV_SCANSTAMP_SZ);
+               SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zfsvfs),
+                   NULL, scanstamp, AV_SCANSTAMP_SZ);
+               zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
+       }
+
+       VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
+       VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
+           count, tx) == 0);
+       if (znode_acl.z_acl_extern_obj)
+               VERIFY(0 == dmu_object_free(zfsvfs->z_os,
+                   znode_acl.z_acl_extern_obj, tx));
+
+       zp->z_is_sa = B_TRUE;
+}
+
+void
+zfs_sa_upgrade_txholds(dmu_tx_t *tx, znode_t *zp)
+{
+       if (!zp->z_zfsvfs->z_use_sa || zp->z_is_sa)
+               return;
+
+       ASSERT(!zp->z_is_sa);
+
+       dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
+
+       if (ZFS_EXTERNAL_ACL(zp)) {
+               dmu_tx_hold_free(tx, ZFS_EXTERNAL_ACL(zp), 0,
+                   DMU_OBJECT_END);
+       }
+}
+
+#endif