Whamcloud - gitweb
LU-1302 llog: pass lu_env as parametr in llog functions
[fs/lustre-release.git] / lustre / obdfilter / filter_log.c
index c271382..d997ae8 100644 (file)
@@ -1,45 +1,60 @@
-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
+/*
+ * GPL HEADER START
  *
- *  linux/fs/obdfilter/filter_log.c
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
- *  Copyright (c) 2001-2003 Cluster File Systems, Inc.
- *   Author: Peter Braam <braam@clusterfs.com>
- *   Author: Andreas Dilger <adilger@clusterfs.com>
- *   Author: Phil Schwan <phil@clusterfs.com>
+ * 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.
  *
- *   This file is part of Lustre, http://www.lustre.org.
+ * 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 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.
+ * 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
  *
- *   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.
+ * 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.
  *
- *   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.
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ *
+ * Copyright (c) 2011, Whamcloud, Inc.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
+ *
+ * lustre/obdfilter/filter_log.c
+ *
+ * Author: Peter Braam <braam@clusterfs.com>
+ * Author: Andreas Dilger <adilger@clusterfs.com>
+ * Author: Phil Schwan <phil@clusterfs.com>
  */
 
 #define DEBUG_SUBSYSTEM S_FILTER
 
-#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/version.h>
 
-#include <portals/list.h>
-#include <linux/obd_class.h>
-#include <linux/lustre_fsfilt.h>
-#include <linux/lustre_commit_confd.h>
-
+#include <libcfs/list.h>
+#include <obd_class.h>
+#include <lustre_log.h>
+#include <lustre_fsfilt.h>
 #include "filter_internal.h"
 
-int filter_log_sz_change(struct llog_handle *cathandle, 
-                         struct lustre_id *id, __u32 io_epoch,
-                         struct llog_cookie *logcookie, 
+int filter_log_sz_change(struct llog_handle *cathandle,
+                         struct ll_fid *mds_fid,
+                         __u32 ioepoch,
+                         struct llog_cookie *logcookie,
                          struct inode *inode)
 {
         struct llog_size_change_rec *lsc;
@@ -47,40 +62,39 @@ int filter_log_sz_change(struct llog_handle *cathandle,
         struct ost_filterdata *ofd;
         ENTRY;
 
-        down(&inode->i_sem);
-        ofd = inode->i_filterdata;
-        
-        if (ofd && ofd->ofd_epoch >= io_epoch) {
-                if (ofd->ofd_epoch > io_epoch)
-                        CERROR("client sent old epoch %d for obj ino %ld\n", 
-                               io_epoch, inode->i_ino);
-                up(&inode->i_sem);
-                RETURN(0);
-        }
+       mutex_lock(&inode->i_mutex);
+       ofd = inode->i_private;
 
-        if (ofd && ofd->ofd_epoch < io_epoch) {
-                ofd->ofd_epoch = io_epoch;
-        } else if (!ofd) {
-                OBD_ALLOC(ofd, sizeof(*ofd));
-                if (!ofd)
-                        GOTO(out, rc = -ENOMEM);
-                igrab(inode);
-                inode->i_filterdata = ofd;
-                ofd->ofd_epoch = io_epoch;
-        }
-        /* the decision to write a record is now made, unlock */
-        up(&inode->i_sem);
+       if (ofd && ofd->ofd_epoch >= ioepoch) {
+               if (ofd->ofd_epoch > ioepoch)
+                       CERROR("client sent old epoch %d for obj ino %ld\n",
+                              ioepoch, inode->i_ino);
+               mutex_unlock(&inode->i_mutex);
+               RETURN(0);
+       }
+
+       if (ofd && ofd->ofd_epoch < ioepoch) {
+               ofd->ofd_epoch = ioepoch;
+       } else if (!ofd) {
+               OBD_ALLOC(ofd, sizeof(*ofd));
+               if (!ofd)
+                       GOTO(out, rc = -ENOMEM);
+               igrab(inode);
+               inode->i_private = ofd;
+               ofd->ofd_epoch = ioepoch;
+       }
+       /* the decision to write a record is now made, unlock */
+       mutex_unlock(&inode->i_mutex);
 
         OBD_ALLOC(lsc, sizeof(*lsc));
         if (lsc == NULL)
                 RETURN(-ENOMEM);
         lsc->lsc_hdr.lrh_len = lsc->lsc_tail.lrt_len = sizeof(*lsc);
         lsc->lsc_hdr.lrh_type =  OST_SZ_REC;
-        lsc->lsc_id = *id;
-        lsc->lsc_io_epoch = io_epoch;
+        lsc->lsc_fid = *mds_fid;
+        lsc->lsc_ioepoch = ioepoch;
 
-        rc = llog_cat_add_rec(cathandle, &lsc->lsc_hdr, logcookie,
-                              NULL, NULL, NULL);
+       rc = llog_cat_add_rec(NULL, cathandle, &lsc->lsc_hdr, logcookie, NULL);
         OBD_FREE(lsc, sizeof(*lsc));
 
         if (rc > 0) {
@@ -91,97 +105,203 @@ int filter_log_sz_change(struct llog_handle *cathandle,
         out:
         RETURN(rc);
 }
-struct obd_llogs * filter_grab_llog_for_group(struct obd_device *,
-                                              int, struct obd_export *);
 
 /* When this (destroy) operation is committed, return the cancel cookie */
 void filter_cancel_cookies_cb(struct obd_device *obd, __u64 transno,
                               void *cb_data, int error)
 {
         struct llog_cookie *cookie = cb_data;
-        struct obd_llogs *llogs = NULL;
+        struct obd_llog_group *olg;
         struct llog_ctxt *ctxt;
+        int rc;
 
         /* we have to find context for right group */
-        llogs = filter_grab_llog_for_group(obd, cookie->lgc_lgl.lgl_ogr, NULL);
-
-        if (llogs) {
-                ctxt = llog_get_context(llogs, cookie->lgc_subsys + 1);
-                if (ctxt) {
-                        llog_cancel(ctxt, 1, cookie, 0, NULL);
-                } else
-                        CERROR("no valid context for group "LPU64"\n",
-                               cookie->lgc_lgl.lgl_ogr);
-        } else {
-                CDEBUG(D_HA, "unknown group "LPU64"!\n", cookie->lgc_lgl.lgl_ogr);
+        if (error != 0 || obd->obd_stopping) {
+                CDEBUG(D_INODE, "not cancel logcookie err %d stopping %d \n",
+                       error, obd->obd_stopping);
+                GOTO (out, rc = 0);
         }
 
-        OBD_FREE(cb_data, sizeof(struct llog_cookie));
+        olg = filter_find_olg(obd, cookie->lgc_lgl.lgl_oseq);
+        if (!olg) {
+                CDEBUG(D_HA, "unknown group "LPU64"!\n", cookie->lgc_lgl.lgl_oseq);
+                GOTO(out, rc = 0);
+        }
+
+        ctxt = llog_group_get_ctxt(olg, cookie->lgc_subsys + 1);
+        if (!ctxt) {
+                CERROR("no valid context for group "LPU64"\n",
+                        cookie->lgc_lgl.lgl_oseq);
+                GOTO(out, rc = 0);
+        }
+
+        OBD_FAIL_TIMEOUT(OBD_FAIL_OST_CANCEL_COOKIE_TIMEOUT, 30);
+
+       rc = llog_cancel(NULL, ctxt, NULL, 1, cookie, 0);
+        if (rc)
+                CERROR("error cancelling log cookies: rc = %d\n", rc);
+        llog_ctxt_put(ctxt);
+out:
+        OBD_FREE(cookie, sizeof(*cookie));
 }
 
-/* Callback for processing the unlink log record received from MDS by 
- * llog_client_api.
- */
-int filter_recov_log_unlink_cb(struct llog_handle *llh, 
-                               struct llog_rec_hdr *rec, void *data)
+/* Callback for processing the unlink log record received from MDS by
+ * llog_client_api. */
+static int filter_recov_log_unlink_cb(struct llog_ctxt *ctxt,
+                                      struct llog_rec_hdr *rec,
+                                      struct llog_cookie *cookie)
 {
-        struct llog_ctxt *ctxt = llh->lgh_ctxt;
-        struct obd_device *obd = ctxt->loc_obd;
-        struct obd_export *exp = obd->obd_self_export;
-        struct llog_cookie cookie;
-        struct llog_gen_rec *lgr;
+        struct obd_export *exp = ctxt->loc_obd->obd_self_export;
         struct llog_unlink_rec *lur;
         struct obdo *oa;
         obd_id oid;
+        obd_count count;
         int rc = 0;
         ENTRY;
-    
-        if (!(le32_to_cpu(llh->lgh_hdr->llh_flags) & LLOG_F_IS_PLAIN)) {
-                CERROR("log is not plain\n");
-                RETURN(-EINVAL);
-        }
-        if (rec->lrh_type != MDS_UNLINK_REC &&
-            rec->lrh_type != LLOG_GEN_REC) {
-                CERROR("log record type error\n");
-                RETURN(-EINVAL);
-        }
-        cookie.lgc_lgl = llh->lgh_id;
-        cookie.lgc_subsys = LLOG_UNLINK_ORIG_CTXT;
-        cookie.lgc_index = le32_to_cpu(rec->lrh_index);
-
-        if (rec->lrh_type == LLOG_GEN_REC) {
-                lgr = (struct llog_gen_rec *)rec;
-                if (llog_gen_lt(lgr->lgr_gen, ctxt->loc_gen))
-                        rc = 0;
-                else
-                        rc = LLOG_PROC_BREAK;
-                CWARN("fetch generation log, send cookie\n");
-                llog_cancel(ctxt, 1, &cookie, 0, NULL);
-                RETURN(rc);
-        }
 
         lur = (struct llog_unlink_rec *)rec;
-        oa = obdo_alloc();
-        if (oa == NULL) 
+        OBDO_ALLOC(oa);
+        if (oa == NULL)
                 RETURN(-ENOMEM);
         oa->o_valid |= OBD_MD_FLCOOKIE;
         oa->o_id = lur->lur_oid;
-        oa->o_gr = lur->lur_ogen;
+        oa->o_seq = lur->lur_oseq;
         oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP;
-        memcpy(obdo_logcookie(oa), &cookie, sizeof(cookie));
+        oa->o_lcookie = *cookie;
         oid = oa->o_id;
+        /* objid gap may require to destroy several objects in row */
+        count = lur->lur_count + 1;
+
+        /* This check is only valid before FID-on-OST and it should
+         * be removed after FID-on-OST is implemented */
+        if (oa->o_seq > FID_SEQ_OST_MAX) {
+                CERROR("%s: invalid group number "LPU64" > MAX_CMD_GROUP %u\n",
+                        exp->exp_obd->obd_name, oa->o_seq, FID_SEQ_OST_MAX);
+                RETURN(-EINVAL);
+        }
+
+        while (count > 0) {
+                rc = filter_destroy(NULL, exp, oa, NULL, NULL, NULL, NULL);
+                if (rc == 0)
+                        CDEBUG(D_RPCTRACE, "object "LPU64" is destroyed\n",
+                               oid);
+                else if (rc != -ENOENT)
+                        CEMERG("error destroying object "LPU64": %d\n",
+                               oid, rc);
+                else
+                        rc = 0;
+                count--;
+                oid++;
+        }
+        OBDO_FREE(oa);
+
+        RETURN(rc);
+}
+
+/* Callback for processing the setattr log record received from MDS by
+ * llog_client_api. */
+static int filter_recov_log_setattr_cb(struct llog_ctxt *ctxt,
+                                       struct llog_rec_hdr *rec,
+                                       struct llog_cookie *cookie)
+{
+        struct obd_device *obd = ctxt->loc_obd;
+        struct obd_export *exp = obd->obd_self_export;
+        struct obd_info oinfo = { { { 0 } } };
+        obd_id oid;
+        int rc = 0;
+        ENTRY;
+
+        OBDO_ALLOC(oinfo.oi_oa);
+        if (oinfo.oi_oa == NULL)
+                RETURN(-ENOMEM);
+
+       if (rec->lrh_type == MDS_SETATTR64_REC) {
+                struct llog_setattr64_rec *lsr = (struct llog_setattr64_rec *)rec;
+
+                oinfo.oi_oa->o_id = lsr->lsr_oid;
+                oinfo.oi_oa->o_seq = lsr->lsr_oseq;
+                oinfo.oi_oa->o_uid = lsr->lsr_uid;
+                oinfo.oi_oa->o_gid = lsr->lsr_gid;
+       } else {
+               CERROR("%s: wrong llog type %#x\n", obd->obd_name,
+                      rec->lrh_type);
+               RETURN(-EINVAL);
+       }
+
+        oinfo.oi_oa->o_valid |= (OBD_MD_FLID | OBD_MD_FLUID | OBD_MD_FLGID |
+                                 OBD_MD_FLCOOKIE);
+        oinfo.oi_oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP;
+        oinfo.oi_oa->o_lcookie = *cookie;
+        oid = oinfo.oi_oa->o_id;
+
+        rc = filter_setattr(NULL, exp, &oinfo, NULL);
+        OBDO_FREE(oinfo.oi_oa);
 
-        rc = obd_destroy(exp, oa, NULL, NULL);
-        obdo_free(oa);
         if (rc == -ENOENT) {
-                CDEBUG(D_HA, "object already removed, send cookie\n");
-                llog_cancel(ctxt, 1, &cookie, 0, NULL);
+                CDEBUG(D_RPCTRACE, "object already removed, send cookie\n");
+               llog_cancel(NULL, ctxt, NULL, 1, cookie, 0);
                 RETURN(0);
         }
 
         if (rc == 0)
-                CDEBUG(D_HA, "object: "LPU64" in record is destroyed\n", oid);
+                CDEBUG(D_RPCTRACE, "object "LPU64" is chown/chgrp\n", oid);
+
+        RETURN(rc);
+}
+
+int filter_recov_log_mds_ost_cb(const struct lu_env *env,
+                               struct llog_handle *llh,
+                               struct llog_rec_hdr *rec, void *data)
+{
+        struct llog_ctxt *ctxt = llh->lgh_ctxt;
+        struct llog_cookie cookie;
+        int rc = 0;
+        ENTRY;
+
+        if (ctxt->loc_obd->obd_stopping)
+                RETURN(LLOG_PROC_BREAK);
+
+        if (rec == NULL) {
+                cfs_spin_lock(&ctxt->loc_obd->u.filter.fo_flags_lock);
+                ctxt->loc_obd->u.filter.fo_mds_ost_sync = 0;
+                cfs_spin_unlock(&ctxt->loc_obd->u.filter.fo_flags_lock);
+                RETURN(0);
+        }
+
+        if (!(llh->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN)) {
+                CERROR("log is not plain\n");
+                RETURN(-EINVAL);
+        }
+
+        OBD_FAIL_TIMEOUT(OBD_FAIL_OST_LLOG_RECOVERY_TIMEOUT, 30);
+        cookie.lgc_lgl = llh->lgh_id;
+        cookie.lgc_subsys = LLOG_MDS_OST_ORIG_CTXT;
+        cookie.lgc_index = rec->lrh_index;
+
+        switch (rec->lrh_type) {
+        case MDS_UNLINK_REC:
+                rc = filter_recov_log_unlink_cb(ctxt, rec, &cookie);
+                break;
+        case MDS_SETATTR64_REC:
+                rc = filter_recov_log_setattr_cb(ctxt, rec, &cookie);
+                break;
+        case LLOG_GEN_REC: {
+                struct llog_gen_rec *lgr = (struct llog_gen_rec *)rec;
+
+                if (llog_gen_lt(lgr->lgr_gen, ctxt->loc_gen))
+                        rc = 0;
+                else
+                        rc = LLOG_PROC_BREAK;
+                CDEBUG(D_HA, "fetch generation log, send cookie\n");
+               llog_cancel(NULL, ctxt, NULL, 1, &cookie, 0);
+                RETURN(rc);
+                }
+                break;
+        default:
+                CERROR("log record type %08x unknown\n", rec->lrh_type);
+                RETURN(-EINVAL);
+                break;
+        }
 
         RETURN(rc);
 }