/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * lustre/mdd/mdd_lov.c
- * Lustre Metadata Server (mds) handling of striped file data
+ * GPL HEADER START
*
- * Copyright (C) 2001-2006 Cluster File Systems, Inc.
- * Author: Peter Braam <braam@clusterfs.com>
- * wangdi <wangdi@clusterfs.com>
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * This file is part of the Lustre file system, http://www.lustre.org
- * Lustre is a trademark of Cluster File Systems, Inc.
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
*
- * You may have signed or agreed to another license before downloading
- * this software. If so, you are bound by the terms and conditions
- * of that agreement, and the following does not apply to you. See the
- * LICENSE file included with this distribution for more information.
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
*
- * If you did not agree to a different license, then this copy of Lustre
- * is open source software; you can redistribute it and/or modify it
- * under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
+ * You should have received a copy of the GNU General Public License
+ * version 2 along with this program; If not, see
+ * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
*
- * In either case, Lustre is distributed in the hope that it will be
- * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
- * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * license text for more details.
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ * GPL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved
+ * Use is subject to license terms.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
+ *
+ * lustre/mdd/mdd_lov.c
+ *
+ * Lustre Metadata Server (mds) handling of striped file data
+ *
+ * Author: Peter Braam <braam@clusterfs.com>
+ * Author: wangdi <wangdi@clusterfs.com>
*/
+
#ifndef EXPORT_SYMTAB
# define EXPORT_SYMTAB
#endif
struct obd_device *obd;
ENTRY;
- mds_id = mdd2lu_dev(mdd)->ld_site->ls_node_id;
+ mds_id = lu_site2md(mdd2lu_dev(mdd)->ld_site)->ms_node_id;
name_size = strlen(MDD_OBD_NAME) + 35;
uuid_size = strlen(MDD_OBD_UUID) + 35;
/*
* Add here for obd notify mechanism, when adding a new ost, the mds
- * will notify this mdd.
+ * will notify this mdd. The mds will be used for quota also.
*/
obd->obd_upcall.onu_upcall = mdd_notify;
obd->obd_upcall.onu_owner = mdd;
mdd->mdd_obd_dev = obd;
-
EXIT;
class_detach:
if (rc)
if (rc)
GOTO(lcfg_cleanup, rc);
mdd->mdd_obd_dev = NULL;
-
+
EXIT;
lcfg_cleanup:
return rc;
*md_size = 0;
rc = 0;
} else if (rc < 0) {
- CERROR("Error %d reading eadata \n", rc);
+ CERROR("Error %d reading eadata - %d\n", rc, *md_size);
} else {
/* XXX: Convert lov EA but fixed after verification test. */
*md_size = rc;
void *md, int *md_size, const char *name)
{
int rc = 0;
- mdd_read_lock(env, obj);
+ mdd_read_lock(env, obj, MOR_TGT_CHILD);
rc = mdd_get_md(env, obj, md, md_size, name);
mdd_read_unlock(env, obj);
return rc;
RETURN(rc);
obd_free_memmd(lov_exp, &lsm);
- rc = mdd_xattr_set_txn(env, obj, buf, MDS_LOV_MD_NAME, 0, handle);
+ rc = mdd_xattr_set_txn(env, obj, buf, XATTR_NAME_LOV, 0, handle);
CDEBUG(D_INFO, "set lov ea of "DFID" rc %d \n", PFID(mdo2fid(obj)), rc);
RETURN(rc);
LASSERT(S_ISDIR(mdd_object_type(obj)));
lum = (struct lov_user_md*)buf->lb_buf;
- /* if { size, offset, count } = { 0, -1, 0 } (i.e. all default
+ /* if { size, offset, count } = { 0, -1, 0 } and no pool (i.e. all default
* values specified) then delete default striping from dir. */
- if ((lum->lmm_stripe_size == 0 && lum->lmm_stripe_count == 0 &&
- lum->lmm_stripe_offset == (typeof(lum->lmm_stripe_offset))(-1)) ||
- /* lmm_stripe_size == -1 is deprecated in 1.4.6 */
- lum->lmm_stripe_size == (typeof(lum->lmm_stripe_size))(-1)){
+ if (lum->lmm_stripe_size == 0 && lum->lmm_stripe_count == 0 &&
+ lum->lmm_stripe_offset == (typeof(lum->lmm_stripe_offset))(-1) &&
+ lum->lmm_magic != LOV_USER_MAGIC_V3) {
rc = mdd_xattr_set_txn(env, obj, &LU_BUF_NULL,
- MDS_LOV_MD_NAME, 0, handle);
+ XATTR_NAME_LOV, 0, handle);
if (rc == -ENODATA)
rc = 0;
CDEBUG(D_INFO, "delete lov ea of "DFID" rc %d \n",
if (rc)
RETURN(rc);
- if ((uc->mu_fsuid != tmp_la->la_uid) && !mdd_capable(uc, CAP_FOWNER))
+ if ((uc->mu_fsuid != tmp_la->la_uid) &&
+ !mdd_capable(uc, CFS_CAP_FOWNER))
rc = mdd_permission_internal_locked(env, obj, tmp_la,
- MAY_WRITE);
+ MAY_WRITE, MOR_TGT_CHILD);
RETURN(rc);
}
rc = mdd_lov_set_stripe_md(env, child, buf, handle);
} else {
rc = mdd_xattr_set_txn(env, child, buf,
- MDS_LOV_MD_NAME, 0, handle);
+ XATTR_NAME_LOV, 0, handle);
}
} else if (S_ISDIR(mode)) {
if (lmmp == NULL && lmm_size == 0) {
struct mdd_device *mdd = mdd_obj2mdd_dev(child);
struct lov_mds_md *lmm = mdd_max_lmm_get(env, mdd);
- int size = sizeof(*lmm);
+ int size = sizeof(struct lov_mds_md_v3);
/* Get parent dir stripe and set */
if (pobj != NULL)
rc = mdd_get_md_locked(env, pobj, lmm, &size,
- MDS_LOV_MD_NAME);
+ XATTR_NAME_LOV);
if (rc > 0) {
buf = mdd_buf_get(env, lmm, size);
rc = mdd_xattr_set_txn(env, child, buf,
- MDS_LOV_MD_NAME, 0, handle);
+ XATTR_NAME_LOV, 0, handle);
if (rc)
CERROR("error on copy stripe info: rc "
"= %d\n", rc);
return fid_flatten(fid);
}
+int mdd_lov_objid_prepare(struct mdd_device *mdd, struct lov_mds_md *lmm)
+{
+ /* copy mds_lov code is using wrong layer */
+ return mds_lov_prepare_objids(mdd->mdd_obd_dev, lmm);
+}
+
void mdd_lov_objid_update(struct mdd_device *mdd, struct lov_mds_md *lmm)
{
+ /* copy mds_lov code is using wrong layer */
mds_lov_update_objids(mdd->mdd_obd_dev, lmm);
}
struct lov_mds_md *lmm, int lmm_size,
const struct md_op_spec *spec)
{
- if (lmm && !spec->u.sp_ea.no_lov_create)
+ if (lmm && !spec->no_create)
OBD_FREE(lmm, lmm_size);
}
{
struct obd_device *obd = mdd2obd_dev(mdd);
struct obd_export *lov_exp = obd->u.mds.mds_osc_exp;
+ struct lu_site *site = mdd2lu_dev(mdd)->ld_site;
struct obdo *oa;
struct lov_stripe_md *lsm = NULL;
const void *eadata = spec->u.sp_ea.eadata;
oti_init(oti, NULL);
/* replay case, has objects already, only get lov from eadata */
- if (spec->u.sp_ea.no_lov_create != 0) {
+ if (spec->no_create != 0) {
*lmm = (struct lov_mds_md *)spec->u.sp_ea.eadata;
*lmm_size = spec->u.sp_ea.eadatalen;
RETURN(0);
oa->o_uid = 0; /* must have 0 uid / gid on OST */
oa->o_gid = 0;
- oa->o_gr = FILTER_GROUP_MDS0 + mdd2lu_dev(mdd)->ld_site->ls_node_id;
+ oa->o_gr = mdt_to_obd_objgrp(lu_site2md(site)->ms_node_id);
oa->o_mode = S_IFREG | 0600;
oa->o_id = mdd_lov_create_id(mdd_object_fid(child));
oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE | OBD_MD_FLFLAGS |
rc = mdd_get_md_locked(env, parent, _lmm,
&_lmm_size,
- MDS_LOV_MD_NAME);
+ XATTR_NAME_LOV);
if (rc > 0)
rc = obd_iocontrol(OBD_IOC_LOV_SETSTRIPE,
lov_exp, 0, &lsm, _lmm);
}
GOTO(out_oti, rc);
}
- LASSERT(lsm->lsm_object_gr >= FILTER_GROUP_MDS0);
+ LASSERT_MDS_GROUP(lsm->lsm_object_gr);
} else {
LASSERT(eadata != NULL);
rc = obd_iocontrol(OBD_IOC_LOV_SETEA, lov_exp, 0, &lsm,
oa->o_valid |= OBD_MD_FLFID | OBD_MD_FLGENER;
oinfo->oi_oa = oa;
oinfo->oi_md = lsm;
- oinfo->oi_capa = mdo_capa_get(env, child, NULL,
- CAPA_OPC_MDS_DEFAULT);
+ oinfo->oi_capa = NULL;
oinfo->oi_policy.l_extent.start = la->la_size;
oinfo->oi_policy.l_extent.end = OBD_OBJECT_EOF;
- if (IS_ERR(oinfo->oi_capa))
- oinfo->oi_capa = NULL;
-
rc = obd_punch_rqset(lov_exp, oinfo, oti);
- capa_put(oinfo->oi_capa);
if (rc) {
CERROR("Error setting attrs for "DFID": rc %d\n",
PFID(mdo2fid(child)), rc);
CERROR("Cannot pack lsm, err = %d\n", rc);
GOTO(out_oti, rc);
}
+ if (mdd_lov_objid_prepare(mdd, *lmm) != 0) {
+ CERROR("Not have memory for update objid\n");
+ OBD_FREE(*lmm, rc);
+ *lmm = NULL;
+ GOTO(out_oti, rc = -ENOMEM);
+ }
*lmm_size = rc;
rc = 0;
EXIT;
struct lov_stripe_md *lsm = NULL;
struct obd_trans_info *oti = &mdd_env_info(env)->mti_oti;
struct obdo *oa = &mdd_env_info(env)->mti_oa;
+ struct lu_site *site = mdd2lu_dev(mdd)->ld_site;
int rc;
ENTRY;
}
oa->o_id = lsm->lsm_object_id;
- oa->o_gr = FILTER_GROUP_MDS0 + mdd2lu_dev(mdd)->ld_site->ls_node_id;
+ oa->o_gr = mdt_to_obd_objgrp(lu_site2md(site)->ms_node_id);
oa->o_mode = la->la_mode & S_IFMT;
oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE | OBD_MD_FLGROUP;
CDEBUG(D_INFO, "destroying OSS object %d/%d\n",
(int)oa->o_id, (int)oa->o_gr);
- rc = obd_destroy(lov_exp, oa, lsm, oti, NULL);
+ rc = obd_destroy(lov_exp, oa, lsm, oti, NULL, NULL);
obd_free_memmd(lov_exp, &lsm);
RETURN(rc);
}
-
/*
* called with obj not locked.
*/
+
int mdd_lov_destroy(const struct lu_env *env, struct mdd_device *mdd,
struct mdd_object *obj, struct lu_attr *la)
{
RETURN(rc = -ENOMEM);
/* get lov ea */
+
rc = mdd_get_md_locked(env, obj, ma->ma_lmm, &ma->ma_lmm_size,
- MDS_LOV_MD_NAME);
- if (rc) {
- CWARN("Get lov ea failed for "DFID"\n", PFID(mdo2fid(obj)));
+ XATTR_NAME_LOV);
+
+ if (rc <= 0) {
+ CWARN("Get lov ea failed for "DFID" rc = %d\n",
+ PFID(mdo2fid(obj)), rc);
+ if (rc == 0)
+ rc = -ENOENT;
RETURN(rc);
}
+
ma->ma_valid = MA_LOV;
rc = mdd_unlink_log(env, mdd, obj, ma);
PFID(mdo2fid(obj)), rc);
RETURN(rc);
}
- if (ma->ma_valid | MA_COOKIE)
- rc = mdd_lovobj_unlink(env, mdd, obj, la,
+
+ if (ma->ma_valid & MA_COOKIE)
+ rc = mdd_lovobj_unlink(env, mdd, obj, la,
ma->ma_lmm, ma->ma_lmm_size,
ma->ma_cookie, 1);
RETURN(rc);
}
+int mdd_log_op_unlink(struct obd_device *obd,
+ struct lov_mds_md *lmm, int lmm_size,
+ struct llog_cookie *logcookies, int cookies_size)
+{
+ struct mds_obd *mds = &obd->u.mds;
+ struct lov_stripe_md *lsm = NULL;
+ struct llog_unlink_rec *lur;
+ struct llog_ctxt *ctxt;
+ int rc;
+ ENTRY;
+
+ if (IS_ERR(mds->mds_osc_obd))
+ RETURN(PTR_ERR(mds->mds_osc_obd));
+
+ rc = obd_unpackmd(mds->mds_osc_exp, &lsm, lmm, lmm_size);
+ if (rc < 0)
+ RETURN(rc);
+ rc = obd_checkmd(mds->mds_osc_exp, obd->obd_self_export, lsm);
+ if (rc)
+ GOTO(out, rc);
+ /* first prepare unlink log record */
+ OBD_ALLOC(lur, sizeof(*lur));
+ if (!lur)
+ GOTO(out, rc = -ENOMEM);
+ lur->lur_hdr.lrh_len = lur->lur_tail.lrt_len = sizeof(*lur);
+ lur->lur_hdr.lrh_type = MDS_UNLINK_REC;
+
+ ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
+ rc = llog_add(ctxt, &lur->lur_hdr, lsm, logcookies,
+ cookies_size / sizeof(struct llog_cookie));
+ llog_ctxt_put(ctxt);
+
+ OBD_FREE(lur, sizeof(*lur));
+ GOTO(out, rc);
+out:
+ obd_free_memmd(mds->mds_osc_exp, &lsm);
+ return rc;
+}
int mdd_unlink_log(const struct lu_env *env, struct mdd_device *mdd,
struct mdd_object *mdd_cobj, struct md_attr *ma)
LASSERT(ma->ma_valid & MA_LOV);
if ((ma->ma_cookie_size > 0) &&
- (mds_log_op_unlink(obd, ma->ma_lmm, ma->ma_lmm_size,
+ (mdd_log_op_unlink(obd, ma->ma_lmm, ma->ma_lmm_size,
ma->ma_cookie, ma->ma_cookie_size) > 0)) {
ma->ma_valid |= MA_COOKIE;
}
return 0;
}
+int mdd_log_op_setattr(struct obd_device *obd, __u32 uid, __u32 gid,
+ struct lov_mds_md *lmm, int lmm_size,
+ struct llog_cookie *logcookies, int cookies_size)
+{
+ struct mds_obd *mds = &obd->u.mds;
+ struct lov_stripe_md *lsm = NULL;
+ struct llog_setattr64_rec *lsr;
+ struct llog_ctxt *ctxt;
+ int rc;
+ ENTRY;
+
+ if (IS_ERR(mds->mds_osc_obd))
+ RETURN(PTR_ERR(mds->mds_osc_obd));
+
+ rc = obd_unpackmd(mds->mds_osc_exp, &lsm, lmm, lmm_size);
+ if (rc < 0)
+ RETURN(rc);
+
+ rc = obd_checkmd(mds->mds_osc_exp, obd->obd_self_export, lsm);
+ if (rc)
+ GOTO(out, rc);
+
+ OBD_ALLOC(lsr, sizeof(*lsr));
+ if (!lsr)
+ GOTO(out, rc = -ENOMEM);
+
+ /* prepare setattr log record */
+ lsr->lsr_hdr.lrh_len = lsr->lsr_tail.lrt_len = sizeof(*lsr);
+ lsr->lsr_hdr.lrh_type = MDS_SETATTR64_REC;
+ lsr->lsr_uid = uid;
+ lsr->lsr_gid = gid;
+
+ /* write setattr log */
+ ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
+ rc = llog_add(ctxt, &lsr->lsr_hdr, lsm, logcookies,
+ cookies_size / sizeof(struct llog_cookie));
+
+ llog_ctxt_put(ctxt);
+
+ OBD_FREE(lsr, sizeof(*lsr));
+ out:
+ obd_free_memmd(mds->mds_osc_exp, &lsm);
+ RETURN(rc);
+}
+
int mdd_setattr_log(const struct lu_env *env, struct mdd_device *mdd,
const struct md_attr *ma,
struct lov_mds_md *lmm, int lmm_size,
/* journal chown/chgrp in llog, just like unlink */
if (lmm_size > 0) {
CDEBUG(D_INFO, "setattr llog for uid/gid=%lu/%lu\n",
- (unsigned long)ma->ma_attr.la_uid,
+ (unsigned long)ma->ma_attr.la_uid,
(unsigned long)ma->ma_attr.la_gid);
- return mds_log_op_setattr(obd, ma->ma_attr.la_uid,
- ma->ma_attr.la_gid, lmm,
+ return mdd_log_op_setattr(obd, ma->ma_attr.la_uid,
+ ma->ma_attr.la_gid, lmm,
lmm_size, logcookies,
cookies_size);
} else
return 0;
}
+static int mdd_osc_setattr_async(struct obd_device *obd, __u32 uid, __u32 gid,
+ struct lov_mds_md *lmm, int lmm_size,
+ struct llog_cookie *logcookies, __u64 id, __u32 gen,
+ struct obd_capa *oc)
+{
+ struct mds_obd *mds = &obd->u.mds;
+ struct obd_trans_info oti = { 0 };
+ struct obd_info oinfo = { { { 0 } } };
+ int rc;
+ ENTRY;
+
+ if (OBD_FAIL_CHECK(OBD_FAIL_MDS_OST_SETATTR))
+ RETURN(0);
+
+ /* first get memory EA */
+ OBDO_ALLOC(oinfo.oi_oa);
+ if (!oinfo.oi_oa)
+ RETURN(-ENOMEM);
+
+ LASSERT(lmm);
+
+ rc = obd_unpackmd(mds->mds_osc_exp, &oinfo.oi_md, lmm, lmm_size);
+ if (rc < 0) {
+ CERROR("Error unpack md %p for inode "LPU64"\n", lmm, id);
+ GOTO(out, rc);
+ }
+
+ rc = obd_checkmd(mds->mds_osc_exp, obd->obd_self_export, oinfo.oi_md);
+ if (rc) {
+ CERROR("Error revalidate lsm %p \n", oinfo.oi_md);
+ GOTO(out, rc);
+ }
+
+ /* then fill oa */
+ oinfo.oi_oa->o_uid = uid;
+ oinfo.oi_oa->o_gid = gid;
+ oinfo.oi_oa->o_id = oinfo.oi_md->lsm_object_id;
+ oinfo.oi_oa->o_gr = oinfo.oi_md->lsm_object_gr;
+ oinfo.oi_oa->o_valid |= OBD_MD_FLID | OBD_MD_FLGROUP |
+ OBD_MD_FLUID | OBD_MD_FLGID;
+ if (logcookies) {
+ oinfo.oi_oa->o_valid |= OBD_MD_FLCOOKIE;
+ oti.oti_logcookies = logcookies;
+ }
+
+ oinfo.oi_oa->o_fid = id;
+ oinfo.oi_oa->o_generation = gen;
+ oinfo.oi_oa->o_valid |= OBD_MD_FLFID | OBD_MD_FLGENER;
+ oinfo.oi_capa = oc;
+
+ /* do async setattr from mds to ost not waiting for responses. */
+ rc = obd_setattr_async(mds->mds_osc_exp, &oinfo, &oti, NULL);
+ if (rc)
+ CDEBUG(D_INODE, "mds to ost setattr objid 0x"LPX64
+ " on ost error %d\n", oinfo.oi_md->lsm_object_id, rc);
+out:
+ if (oinfo.oi_md)
+ obd_free_memmd(mds->mds_osc_exp, &oinfo.oi_md);
+ OBDO_FREE(oinfo.oi_oa);
+ RETURN(rc);
+}
+
int mdd_lov_setattr_async(const struct lu_env *env, struct mdd_object *obj,
- struct lov_mds_md *lmm, int lmm_size,
+ struct lov_mds_md *lmm, int lmm_size,
struct llog_cookie *logcookies)
{
struct mdd_device *mdd = mdo2mdd(&obj->mod_obj);
struct obd_device *obd = mdd2obd_dev(mdd);
struct lu_attr *tmp_la = &mdd_env_info(env)->mti_la;
const struct lu_fid *fid = mdd_object_fid(obj);
- struct obd_capa *oc;
int rc = 0;
ENTRY;
- mdd_read_lock(env, obj);
+ mdd_read_lock(env, obj, MOR_TGT_CHILD);
rc = mdo_attr_get(env, obj, tmp_la, mdd_object_capa(env, obj));
mdd_read_unlock(env, obj);
if (rc)
RETURN(rc);
- oc = mdo_capa_get(env, obj, NULL, CAPA_OPC_MDS_DEFAULT);
- if (IS_ERR(oc))
- oc = NULL;
-
- rc = mds_osc_setattr_async(obd, tmp_la->la_uid, tmp_la->la_gid, lmm,
+ rc = mdd_osc_setattr_async(obd, tmp_la->la_uid, tmp_la->la_gid, lmm,
lmm_size, logcookies, fid_seq(fid),
- fid_oid(fid), oc);
-
- capa_put(oc);
-
+ fid_oid(fid), NULL);
RETURN(rc);
}
-