X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fmds%2Fmds_log.c;h=f25605371cf61b335e29b0956479c6a18f600e23;hp=549c7604683a57dc98c454a17a5d76dca3883f45;hb=01138321c7ce393c189a7ed11559c0938ce9f17e;hpb=342797139fe89f9efd184227cfbd7d654533685c diff --git a/lustre/mds/mds_log.c b/lustre/mds/mds_log.c index 549c760..f256053 100644 --- a/lustre/mds/mds_log.c +++ b/lustre/mds/mds_log.c @@ -1,163 +1,355 @@ /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- * vim:expandtab:shiftwidth=8:tabstop=8: * - * lustre/mds/mds_log.c + * GPL HEADER START * - * Copyright (c) 2001-2003 Cluster File Systems, Inc. - * Author: Peter Braam - * Author: Andreas Dilger - * Author: Phil Schwan + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * This file is part of Lustre, http://www.lustre.org. + * 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. * - * Lustre is free 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. + * 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). * - * 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 - * GNU General Public License for more details. + * 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 * - * You should have received a copy of the GNU General Public License - * along with Lustre; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * 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 (c) 2003, 2010, Oracle and/or its affiliates. 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/mds/mds_log.c + * + * Author: Peter Braam + * Author: Andreas Dilger + * Author: Phil Schwan */ #define DEBUG_SUBSYSTEM S_MDS +#ifndef AUTOCONF_INCLUDED #include +#endif #include #include -#include -#include -#include -#include - +#include +#include +#include +#include +#include #include "mds_internal.h" -static int mds_llog_origin_add(struct llog_ctxt *ctxt, - struct llog_rec_hdr *rec, struct lov_stripe_md *lsm, - struct llog_cookie *logcookies, int numcookies) +static int mds_llog_origin_add(struct llog_ctxt *ctxt, struct llog_rec_hdr *rec, + struct lov_stripe_md *lsm, + struct llog_cookie *logcookies, int numcookies) { struct obd_device *obd = ctxt->loc_obd; - struct obd_device *lov_obd = obd->u.mds.mds_osc_obd; + struct obd_device *lov_obd = obd->u.mds.mds_lov_obd; struct llog_ctxt *lctxt; int rc; ENTRY; lctxt = llog_get_context(lov_obd, ctxt->loc_idx); rc = llog_add(lctxt, rec, lsm, logcookies, numcookies); + llog_ctxt_put(lctxt); + RETURN(rc); } -static int mds_llog_origin_connect(struct llog_ctxt *ctxt, int count, +static int mds_llog_origin_connect(struct llog_ctxt *ctxt, struct llog_logid *logid, - struct llog_gen *gen) + struct llog_gen *gen, + struct obd_uuid *uuid) { struct obd_device *obd = ctxt->loc_obd; - struct obd_device *lov_obd = obd->u.mds.mds_osc_obd; + struct obd_device *lov_obd = obd->u.mds.mds_lov_obd; struct llog_ctxt *lctxt; int rc; ENTRY; lctxt = llog_get_context(lov_obd, ctxt->loc_idx); - rc = llog_connect(lctxt, count, logid, gen); + rc = llog_connect(lctxt, logid, gen, uuid); + llog_ctxt_put(lctxt); RETURN(rc); } +static struct llog_operations mds_ost_orig_logops = { + lop_add: mds_llog_origin_add, + lop_connect: mds_llog_origin_connect, +}; + static int mds_llog_repl_cancel(struct llog_ctxt *ctxt, struct lov_stripe_md *lsm, int count, struct llog_cookie *cookies, int flags) { struct obd_device *obd = ctxt->loc_obd; - struct obd_device *lov_obd = obd->u.mds.mds_osc_obd; + struct obd_device *lov_obd = obd->u.mds.mds_lov_obd; struct llog_ctxt *lctxt; int rc; ENTRY; lctxt = llog_get_context(lov_obd, ctxt->loc_idx); - rc = llog_cancel(lctxt, lsm, count, cookies,flags); + rc = llog_cancel(lctxt, lsm, count, cookies, flags); + llog_ctxt_put(lctxt); RETURN(rc); } -int mds_log_op_unlink(struct obd_device *obd, struct inode *inode, - struct lustre_msg *repmsg, int offset) +static struct llog_operations mds_size_repl_logops = { + lop_cancel: mds_llog_repl_cancel, +}; + +static struct llog_operations changelog_orig_logops; + +static int llog_changelog_cancel_cb(struct llog_handle *llh, + struct llog_rec_hdr *hdr, void *data) { - struct mds_obd *mds = &obd->u.mds; - struct lov_stripe_md *lsm = NULL; - struct llog_ctxt *ctxt; - int rc; + struct llog_changelog_rec *rec = (struct llog_changelog_rec *)hdr; + struct llog_cookie cookie; + long long endrec = *(long long *)data; + int rc, err; + struct obd_device *obd; + void *trans_h; + struct inode *inode; ENTRY; - if (IS_ERR(mds->mds_osc_obd)) - RETURN(PTR_ERR(mds->mds_osc_obd)); + /* This is always a (sub)log, not the catalog */ + LASSERT(llh->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN); - rc = obd_unpackmd(mds->mds_osc_exp, &lsm, - lustre_msg_buf(repmsg, offset, 0), - repmsg->buflens[offset]); - if (rc < 0) - RETURN(rc); + if (rec->cr.cr_index > endrec) + /* records are in order, so we're done */ + RETURN(LLOG_PROC_BREAK); - ctxt = llog_get_context(obd, LLOG_UNLINK_ORIG_CTXT); - rc = llog_add(ctxt, NULL, lsm, lustre_msg_buf(repmsg, offset + 1, 0), - repmsg->buflens[offset + 1] / sizeof(struct llog_cookie)); + cookie.lgc_lgl = llh->lgh_id; + cookie.lgc_index = hdr->lrh_index; + obd = llh->lgh_ctxt->loc_exp->exp_obd; + inode = llh->lgh_file->f_dentry->d_inode; - obd_free_memmd(mds->mds_osc_exp, &lsm); + /* XXX This is a workaround for the deadlock of changelog adding vs. + * changelog cancelling. Changelog adding always start transaction + * before acquiring the catlog lock (lgh_lock), whereas, changelog + * cancelling do start transaction after holding catlog lock. + * + * We start the transaction earlier here to keep the locking ordering: + * 'start transaction -> catlog lock'. LU-81. */ + trans_h = fsfilt_start_log(obd, inode, FSFILT_OP_CANCEL_UNLINK, + NULL, 1); + if (IS_ERR(trans_h)) { + CERROR("fsfilt_start_log failed: %ld\n", PTR_ERR(trans_h)); + RETURN(PTR_ERR(trans_h)); + } + + /* cancel them one at a time. I suppose we could store up the cookies + and cancel them all at once; probably more efficient, but this is + done as a user call, so who cares... */ + rc = llog_cat_cancel_records(llh->u.phd.phd_cat_handle, 1, &cookie); + + err = fsfilt_commit(obd, inode, trans_h, 0); + if (err) { + CERROR("fsfilt_commit failed: %d\n", err); + rc = (rc >= 0) ? err : rc; + } + + RETURN(rc < 0 ? rc : 0); +} + +static int llog_changelog_cancel(struct llog_ctxt *ctxt, + struct lov_stripe_md *lsm, int count, + struct llog_cookie *cookies, int flags) +{ + struct llog_handle *cathandle = ctxt->loc_handle; + int rc; + ENTRY; + + /* This should only be called with the catalog handle */ + LASSERT(cathandle->lgh_hdr->llh_flags & LLOG_F_IS_CAT); + + rc = llog_cat_process(cathandle, llog_changelog_cancel_cb, + (void *)cookies, 0, 0); + if (rc >= 0) + /* 0 or 1 means we're done */ + rc = 0; + else + CERROR("cancel idx %u of catalog "LPX64" rc=%d\n", + cathandle->lgh_last_idx, cathandle->lgh_id.lgl_oid, rc); RETURN(rc); } -static struct llog_operations mds_unlink_orig_logops = { - lop_add: mds_llog_origin_add, - lop_connect: mds_llog_origin_connect, -}; +int mds_changelog_llog_init(struct obd_device *obd, struct obd_device *tgt) +{ + int rc; -static struct llog_operations mds_size_repl_logops = { - lop_cancel: mds_llog_repl_cancel -}; + /* see osc_llog_init */ + changelog_orig_logops = llog_lvfs_ops; + changelog_orig_logops.lop_setup = llog_obd_origin_setup; + changelog_orig_logops.lop_cleanup = llog_obd_origin_cleanup; + changelog_orig_logops.lop_add = llog_obd_origin_add; + changelog_orig_logops.lop_cancel = llog_changelog_cancel; + + rc = llog_setup_named(obd, &obd->obd_olg, LLOG_CHANGELOG_ORIG_CTXT, + tgt, 1, NULL, CHANGELOG_CATALOG, + &changelog_orig_logops); + if (rc) { + CERROR("changelog llog setup failed %d\n", rc); + RETURN(rc); + } + + rc = llog_setup_named(obd, &obd->obd_olg, LLOG_CHANGELOG_USER_ORIG_CTXT, + tgt, 1, NULL, CHANGELOG_USERS, + &changelog_orig_logops); + if (rc) { + CERROR("changelog users llog setup failed %d\n", rc); + RETURN(rc); + } -int mds_llog_init(struct obd_device *obd, struct obd_device *tgt, - int count, struct llog_logid *logid) + RETURN(rc); +} +EXPORT_SYMBOL(mds_changelog_llog_init); + +int mds_llog_init(struct obd_device *obd, struct obd_llog_group *olg, + struct obd_device *disk_obd, int *index) { - struct obd_device *lov_obd = obd->u.mds.mds_osc_obd; + struct obd_device *lov_obd = obd->u.mds.mds_lov_obd; + struct llog_ctxt *ctxt; int rc; ENTRY; - - rc = llog_setup(obd, LLOG_UNLINK_ORIG_CTXT, tgt, 0, NULL, - &mds_unlink_orig_logops); + + LASSERT(olg == &obd->obd_olg); + rc = llog_setup(obd, &obd->obd_olg, LLOG_MDS_OST_ORIG_CTXT, disk_obd, + 0, NULL, &mds_ost_orig_logops); if (rc) RETURN(rc); - rc = llog_setup(obd, LLOG_SIZE_REPL_CTXT, tgt, 0, NULL, - &mds_size_repl_logops); + rc = llog_setup(obd, &obd->obd_olg, LLOG_SIZE_REPL_CTXT, disk_obd, + 0, NULL, &mds_size_repl_logops); if (rc) - RETURN(rc); + GOTO(err_llog, rc); - rc = obd_llog_init(lov_obd, tgt, count, logid); - if (rc) - CERROR("error lov_llog_init\n"); + rc = obd_llog_init(lov_obd, &lov_obd->obd_olg, disk_obd, index); + if (rc) { + CERROR("lov_llog_init err %d\n", rc); + GOTO(err_cleanup, rc); + } RETURN(rc); +err_cleanup: + ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT); + if (ctxt) + llog_cleanup(ctxt); +err_llog: + ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT); + if (ctxt) + llog_cleanup(ctxt); + return rc; } int mds_llog_finish(struct obd_device *obd, int count) { - struct obd_device *lov_obd = obd->u.mds.mds_osc_obd; + struct llog_ctxt *ctxt; + int rc = 0, rc2 = 0; + ENTRY; + + ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT); + if (ctxt) + rc = llog_cleanup(ctxt); + + ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT); + if (ctxt) + rc2 = llog_cleanup(ctxt); + if (!rc) + rc = rc2; + + ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT); + if (ctxt) + rc2 = llog_cleanup(ctxt); + if (!rc) + rc = rc2; + + ctxt = llog_get_context(obd, LLOG_CHANGELOG_USER_ORIG_CTXT); + if (ctxt) + rc2 = llog_cleanup(ctxt); + if (!rc) + rc = rc2; + + RETURN(rc); +} + +static int mds_llog_add_unlink(struct obd_device *obd, + struct lov_stripe_md *lsm, obd_count count, + struct llog_cookie *logcookie, int cookies) +{ + struct llog_unlink_rec *lur; + struct llog_ctxt *ctxt; + int rc; + + if (cookies < lsm->lsm_stripe_count) + RETURN(rc = -EFBIG); + + /* first prepare unlink log record */ + OBD_ALLOC_PTR(lur); + if (!lur) + RETURN(rc = -ENOMEM); + lur->lur_hdr.lrh_len = lur->lur_tail.lrt_len = sizeof(*lur); + lur->lur_hdr.lrh_type = MDS_UNLINK_REC; + lur->lur_count = count; + + ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT); + rc = llog_add(ctxt, &lur->lur_hdr, lsm, logcookie, cookies); + llog_ctxt_put(ctxt); + + OBD_FREE_PTR(lur); + RETURN(rc); +} + +int mds_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; int rc; ENTRY; - - rc = llog_cleanup(llog_get_context(obd, LLOG_UNLINK_ORIG_CTXT)); - if (rc) - RETURN(rc); - rc = llog_cleanup(llog_get_context(obd, LLOG_SIZE_REPL_CTXT)); - if (rc) + if (IS_ERR(mds->mds_lov_obd)) + RETURN(PTR_ERR(mds->mds_lov_obd)); + + rc = obd_unpackmd(mds->mds_lov_exp, &lsm, lmm, lmm_size); + if (rc < 0) RETURN(rc); + rc = mds_llog_add_unlink(obd, lsm, 0, logcookies, + cookies_size / sizeof(struct llog_cookie)); + obd_free_memmd(mds->mds_lov_exp, &lsm); + RETURN(rc); +} +EXPORT_SYMBOL(mds_log_op_unlink); + +int mds_log_op_orphan(struct obd_device *obd, struct lov_stripe_md *lsm, + obd_count count) +{ + struct mds_obd *mds = &obd->u.mds; + struct llog_cookie logcookie; + int rc; + ENTRY; - rc = obd_llog_finish(lov_obd, count); - if (rc) - CERROR("error lov_llog_finish\n"); + if (IS_ERR(mds->mds_lov_obd)) + RETURN(PTR_ERR(mds->mds_lov_obd)); + rc = mds_llog_add_unlink(obd, lsm, count - 1, &logcookie, 1); RETURN(rc); } +