Whamcloud - gitweb
LU-7340 mdd: changelogs garbage collection
[fs/lustre-release.git] / lustre / obdclass / llog_cat.c
index dafcb23..1628cc8 100644 (file)
@@ -1,6 +1,4 @@
-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
- *
+/*
  * GPL HEADER START
  *
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * 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
- *
- * 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.
+ * http://www.gnu.org/licenses/gpl-2.0.html
  *
  * GPL HEADER END
  */
 /*
- * Copyright  2008 Sun Microsystems, Inc. All rights reserved
+ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
+ *
+ * Copyright (c) 2012, 2016, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
  *   if an OST or MDS fails it need only look at log(s) relevant to itself
  *
  * Author: Andreas Dilger <adilger@clusterfs.com>
+ * Author: Alexey Zhuravlev <alexey.zhuravlev@intel.com>
+ * Author: Mikhail Pershin <mike.pershin@intel.com>
  */
 
 #define DEBUG_SUBSYSTEM S_LOG
 
-#ifndef EXPORT_SYMTAB
-#define EXPORT_SYMTAB
-#endif
-
-#ifndef __KERNEL__
-#include <liblustre.h>
-#endif
 
 #include <obd_class.h>
-#include <lustre_log.h>
-#include <libcfs/list.h>
+
+#include "llog_internal.h"
 
 /* Create a new log handle and add it to the open list.
  * This log handle will be closed when all of the records in it are removed.
  *
  * Assumes caller has already pushed us into the kernel context and is locking.
  */
-static struct llog_handle *llog_cat_new_log(struct llog_handle *cathandle)
+static int llog_cat_new_log(const struct lu_env *env,
+                           struct llog_handle *cathandle,
+                           struct llog_handle *loghandle,
+                           struct thandle *th)
 {
-        struct llog_handle *loghandle;
-        struct llog_log_hdr *llh;
-        struct llog_logid_rec rec = { { 0 }, };
-        int rc, index, bitmap_size;
-        ENTRY;
-
-        llh = cathandle->lgh_hdr;
-        bitmap_size = LLOG_BITMAP_SIZE(llh);
-
-        index = (cathandle->lgh_last_idx + 1) % bitmap_size;
-
-        /* maximum number of available slots in catlog is bitmap_size - 2 */
-        if (llh->llh_cat_idx == index) {
-                CERROR("no free catalog slots for log...\n");
-                RETURN(ERR_PTR(-ENOSPC));
-        }
-
-        if (OBD_FAIL_CHECK(OBD_FAIL_MDS_LLOG_CREATE_FAILED))
-                RETURN(ERR_PTR(-ENOSPC));
-
-        rc = llog_create(cathandle->lgh_ctxt, &loghandle, NULL, NULL);
-        if (rc)
-                RETURN(ERR_PTR(rc));
-
-        rc = llog_init_handle(loghandle,
-                              LLOG_F_IS_PLAIN | LLOG_F_ZAP_WHEN_EMPTY,
-                              &cathandle->lgh_hdr->llh_tgtuuid);
-        if (rc)
-                GOTO(out_destroy, rc);
-
-        if (index == 0)
-                index = 1;
-        if (ext2_set_bit(index, llh->llh_bitmap)) {
-                CERROR("argh, index %u already set in log bitmap?\n",
-                       index);
-                LBUG(); /* should never happen */
-        }
-        cathandle->lgh_last_idx = index;
-        llh->llh_count++;
-        llh->llh_tail.lrt_index = index;
-
-        CDEBUG(D_RPCTRACE,"new recovery log "LPX64":%x for index %u of catalog "
-               LPX64"\n", loghandle->lgh_id.lgl_oid, loghandle->lgh_id.lgl_ogen,
-               index, cathandle->lgh_id.lgl_oid);
-        /* build the record for this log in the catalog */
-        rec.lid_hdr.lrh_len = sizeof(rec);
-        rec.lid_hdr.lrh_index = index;
-        rec.lid_hdr.lrh_type = LLOG_LOGID_MAGIC;
-        rec.lid_id = loghandle->lgh_id;
-        rec.lid_tail.lrt_len = sizeof(rec);
-        rec.lid_tail.lrt_index = index;
-
-        /* update the catalog: header and record */
-        rc = llog_write_rec(cathandle, &rec.lid_hdr,
-                            &loghandle->u.phd.phd_cookie, 1, NULL, index);
-        if (rc < 0) {
-                GOTO(out_destroy, rc);
-        }
+       struct llog_thread_info *lgi = llog_info(env);
+       struct llog_logid_rec   *rec = &lgi->lgi_logid;
+       struct thandle *handle = NULL;
+       struct dt_device *dt = NULL;
+       struct llog_log_hdr     *llh = cathandle->lgh_hdr;
+       int                      rc, index;
+
+       ENTRY;
+
+       index = (cathandle->lgh_last_idx + 1) %
+               (OBD_FAIL_PRECHECK(OBD_FAIL_CAT_RECORDS) ? (cfs_fail_val + 1) :
+                                               LLOG_HDR_BITMAP_SIZE(llh));
+
+       /* check that new llog index will not overlap with the first one.
+        * - llh_cat_idx is the index just before the first/oldest still in-use
+        *      index in catalog
+        * - lgh_last_idx is the last/newest used index in catalog
+        *
+        * When catalog is not wrapped yet then lgh_last_idx is always larger
+        * than llh_cat_idx. After the wrap around lgh_last_idx re-starts
+        * from 0 and llh_cat_idx becomes the upper limit for it
+        *
+        * Check if catalog has already wrapped around or not by comparing
+        * last_idx and cat_idx */
+       if ((index == llh->llh_cat_idx + 1 && llh->llh_count > 1) ||
+           (index == 0 && llh->llh_cat_idx == 0)) {
+               if (cathandle->lgh_name == NULL) {
+                       CWARN("%s: there are no more free slots in catalog "
+                             DFID":%x\n",
+                             loghandle->lgh_ctxt->loc_obd->obd_name,
+                             PFID(&cathandle->lgh_id.lgl_oi.oi_fid),
+                             cathandle->lgh_id.lgl_ogen);
+               } else {
+                       CWARN("%s: there are no more free slots in "
+                             "catalog %s\n",
+                             loghandle->lgh_ctxt->loc_obd->obd_name,
+                             cathandle->lgh_name);
+               }
+               RETURN(-ENOSPC);
+       }
+
+       if (OBD_FAIL_CHECK(OBD_FAIL_MDS_LLOG_CREATE_FAILED))
+               RETURN(-ENOSPC);
+
+       if (loghandle->lgh_hdr != NULL) {
+               /* If llog object is remote and creation is failed, lgh_hdr
+                * might be left over here, free it first */
+               LASSERT(!llog_exist(loghandle));
+               OBD_FREE_LARGE(loghandle->lgh_hdr, loghandle->lgh_hdr_size);
+               loghandle->lgh_hdr = NULL;
+       }
+
+       if (th == NULL) {
+               dt = lu2dt_dev(cathandle->lgh_obj->do_lu.lo_dev);
+
+               handle = dt_trans_create(env, dt);
+               if (IS_ERR(handle))
+                       RETURN(PTR_ERR(handle));
+
+               /* Create update llog object synchronously, which
+                * happens during inialization process see
+                * lod_sub_prep_llog(), to make sure the update
+                * llog object is created before corss-MDT writing
+                * updates into the llog object */
+               if (cathandle->lgh_ctxt->loc_flags & LLOG_CTXT_FLAG_NORMAL_FID)
+                       handle->th_sync = 1;
+
+               handle->th_wait_submit = 1;
+
+               rc = llog_declare_create(env, loghandle, handle);
+               if (rc != 0)
+                       GOTO(out, rc);
+
+               rec->lid_hdr.lrh_len = sizeof(*rec);
+               rec->lid_hdr.lrh_type = LLOG_LOGID_MAGIC;
+               rec->lid_id = loghandle->lgh_id;
+               rc = llog_declare_write_rec(env, cathandle, &rec->lid_hdr, -1,
+                                           handle);
+               if (rc != 0)
+                       GOTO(out, rc);
+
+               rc = dt_trans_start_local(env, dt, handle);
+               if (rc != 0)
+                       GOTO(out, rc);
+
+               th = handle;
+       }
+
+       rc = llog_create(env, loghandle, th);
+       /* if llog is already created, no need to initialize it */
+       if (rc == -EEXIST) {
+               GOTO(out, rc = 0);
+       } else if (rc != 0) {
+               CERROR("%s: can't create new plain llog in catalog: rc = %d\n",
+                      loghandle->lgh_ctxt->loc_obd->obd_name, rc);
+               GOTO(out, rc);
+       }
+
+       rc = llog_init_handle(env, loghandle,
+                             LLOG_F_IS_PLAIN | LLOG_F_ZAP_WHEN_EMPTY,
+                             &cathandle->lgh_hdr->llh_tgtuuid);
+       if (rc < 0)
+               GOTO(out, rc);
+
+       /* build the record for this log in the catalog */
+       rec->lid_hdr.lrh_len = sizeof(*rec);
+       rec->lid_hdr.lrh_type = LLOG_LOGID_MAGIC;
+       rec->lid_id = loghandle->lgh_id;
+
+       /* append the new record into catalog. The new index will be
+        * assigned to the record and updated in rec header */
+       rc = llog_write_rec(env, cathandle, &rec->lid_hdr,
+                           &loghandle->u.phd.phd_cookie, LLOG_NEXT_IDX, th);
+       if (rc < 0)
+               GOTO(out_destroy, rc);
+
+       CDEBUG(D_OTHER, "new plain log "DFID".%u of catalog "DFID"\n",
+              PFID(&loghandle->lgh_id.lgl_oi.oi_fid), rec->lid_hdr.lrh_index,
+              PFID(&cathandle->lgh_id.lgl_oi.oi_fid));
+
+       loghandle->lgh_hdr->llh_cat_idx = rec->lid_hdr.lrh_index;
+
+       /* limit max size of plain llog so that space can be
+        * released sooner, especially on small filesystems */
+       /* 2MB for the cases when free space hasn't been learned yet */
+       loghandle->lgh_max_size = 2 << 20;
+       dt = lu2dt_dev(cathandle->lgh_obj->do_lu.lo_dev);
+       rc = dt_statfs(env, dt, &lgi->lgi_statfs);
+       if (rc == 0 && lgi->lgi_statfs.os_bfree > 0) {
+               __u64 freespace = (lgi->lgi_statfs.os_bfree *
+                                 lgi->lgi_statfs.os_bsize) >> 6;
+               if (freespace < loghandle->lgh_max_size)
+                       loghandle->lgh_max_size = freespace;
+               /* shouldn't be > 128MB in any case?
+                * it's 256K records of 512 bytes each */
+               if (freespace > (128 << 20))
+                       loghandle->lgh_max_size = 128 << 20;
+       }
+       rc = 0;
 
-        loghandle->lgh_hdr->llh_cat_idx = index;
-        cathandle->u.chd.chd_current_log = loghandle;
-        LASSERT(list_empty(&loghandle->u.phd.phd_entry));
-        list_add_tail(&loghandle->u.phd.phd_entry, &cathandle->u.chd.chd_head);
+out:
+       if (handle != NULL) {
+               handle->th_result = rc >= 0 ? 0 : rc;
+               dt_trans_stop(env, dt, handle);
+       }
+       RETURN(rc);
 
 out_destroy:
-        if (rc < 0)
-                llog_destroy(loghandle);
-
-        RETURN(loghandle);
+       /* to signal llog_cat_close() it shouldn't try to destroy the llog,
+        * we want to destroy it in this transaction, otherwise the object
+        * becomes an orphan */
+       loghandle->lgh_hdr->llh_flags &= ~LLOG_F_ZAP_WHEN_EMPTY;
+       /* this is to mimic full log, so another llog_cat_current_log()
+        * can skip it and ask for another onet */
+       loghandle->lgh_last_idx = LLOG_HDR_BITMAP_SIZE(llh) + 1;
+       llog_trans_destroy(env, loghandle, th);
+       RETURN(rc);
 }
 
 /* Open an existent log handle and add it to the open list.
@@ -141,71 +223,112 @@ out_destroy:
  *
  * Assumes caller has already pushed us into the kernel context and is locking.
  * We return a lock on the handle to ensure nobody yanks it from us.
+ *
+ * This takes extra reference on llog_handle via llog_handle_get() and require
+ * this reference to be put by caller using llog_handle_put()
  */
-int llog_cat_id2handle(struct llog_handle *cathandle, struct llog_handle **res,
-                       struct llog_logid *logid)
+int llog_cat_id2handle(const struct lu_env *env, struct llog_handle *cathandle,
+                      struct llog_handle **res, struct llog_logid *logid)
 {
-        struct llog_handle *loghandle;
-        int rc = 0;
-        ENTRY;
-
-        if (cathandle == NULL)
-                RETURN(-EBADF);
-
-        list_for_each_entry(loghandle, &cathandle->u.chd.chd_head,
-                            u.phd.phd_entry) {
-                struct llog_logid *cgl = &loghandle->lgh_id;
-                if (cgl->lgl_oid == logid->lgl_oid) {
-                        if (cgl->lgl_ogen != logid->lgl_ogen) {
-                                CERROR("log "LPX64" generation %x != %x\n",
-                                       logid->lgl_oid, cgl->lgl_ogen,
-                                       logid->lgl_ogen);
-                                continue;
-                        }
-                        loghandle->u.phd.phd_cat_handle = cathandle;
-                        GOTO(out, rc = 0);
-                }
-        }
-
-        rc = llog_create(cathandle->lgh_ctxt, &loghandle, logid, NULL);
-        if (rc) {
-                CERROR("error opening log id "LPX64":%x: rc %d\n",
-                       logid->lgl_oid, logid->lgl_ogen, rc);
-        } else {
-                rc = llog_init_handle(loghandle, LLOG_F_IS_PLAIN, NULL);
-                if (!rc) {
-                        list_add(&loghandle->u.phd.phd_entry,
-                                 &cathandle->u.chd.chd_head);
-                }
-        }
-        if (!rc) {
-                loghandle->u.phd.phd_cat_handle = cathandle;
-                loghandle->u.phd.phd_cookie.lgc_lgl = cathandle->lgh_id;
-                loghandle->u.phd.phd_cookie.lgc_index = 
-                        loghandle->lgh_hdr->llh_cat_idx;
-        }
-
+       struct llog_handle      *loghandle;
+       enum llog_flag           fmt;
+       int                      rc = 0;
+
+       ENTRY;
+
+       if (cathandle == NULL)
+               RETURN(-EBADF);
+
+       fmt = cathandle->lgh_hdr->llh_flags & LLOG_F_EXT_MASK;
+       down_write(&cathandle->lgh_lock);
+       list_for_each_entry(loghandle, &cathandle->u.chd.chd_head,
+                           u.phd.phd_entry) {
+               struct llog_logid *cgl = &loghandle->lgh_id;
+
+               if (ostid_id(&cgl->lgl_oi) == ostid_id(&logid->lgl_oi) &&
+                   ostid_seq(&cgl->lgl_oi) == ostid_seq(&logid->lgl_oi)) {
+                       if (cgl->lgl_ogen != logid->lgl_ogen) {
+                               CWARN("%s: log "DFID" generation %x != %x\n",
+                                     loghandle->lgh_ctxt->loc_obd->obd_name,
+                                     PFID(&logid->lgl_oi.oi_fid),
+                                     cgl->lgl_ogen, logid->lgl_ogen);
+                               continue;
+                       }
+                       loghandle->u.phd.phd_cat_handle = cathandle;
+                       up_write(&cathandle->lgh_lock);
+                       GOTO(out, rc = 0);
+               }
+       }
+       up_write(&cathandle->lgh_lock);
+
+       rc = llog_open(env, cathandle->lgh_ctxt, &loghandle, logid, NULL,
+                      LLOG_OPEN_EXISTS);
+       if (rc < 0) {
+               CERROR("%s: error opening log id "DFID":%x: rc = %d\n",
+                      cathandle->lgh_ctxt->loc_obd->obd_name,
+                      PFID(&logid->lgl_oi.oi_fid), logid->lgl_ogen, rc);
+               RETURN(rc);
+       }
+
+       rc = llog_init_handle(env, loghandle, LLOG_F_IS_PLAIN | fmt, NULL);
+       if (rc < 0) {
+               llog_close(env, loghandle);
+               loghandle = NULL;
+               RETURN(rc);
+       }
+
+       down_write(&cathandle->lgh_lock);
+       list_add_tail(&loghandle->u.phd.phd_entry, &cathandle->u.chd.chd_head);
+       up_write(&cathandle->lgh_lock);
+
+       loghandle->u.phd.phd_cat_handle = cathandle;
+       loghandle->u.phd.phd_cookie.lgc_lgl = cathandle->lgh_id;
+       loghandle->u.phd.phd_cookie.lgc_index =
+                               loghandle->lgh_hdr->llh_cat_idx;
+       EXIT;
 out:
-        *res = loghandle;
-        RETURN(rc);
+       llog_handle_get(loghandle);
+       *res = loghandle;
+       return 0;
 }
 
-int llog_cat_put(struct llog_handle *cathandle)
+int llog_cat_close(const struct lu_env *env, struct llog_handle *cathandle)
 {
-        struct llog_handle *loghandle, *n;
-        int rc;
-        ENTRY;
-
-        list_for_each_entry_safe(loghandle, n, &cathandle->u.chd.chd_head,
-                                 u.phd.phd_entry) {
-                int err = llog_close(loghandle);
-                if (err)
-                        CERROR("error closing loghandle\n");
-        }
-        rc = llog_close(cathandle);
-        RETURN(rc);
+       struct llog_handle      *loghandle, *n;
+       int                      rc;
+
+       ENTRY;
+
+       list_for_each_entry_safe(loghandle, n, &cathandle->u.chd.chd_head,
+                                u.phd.phd_entry) {
+               struct llog_log_hdr     *llh = loghandle->lgh_hdr;
+               int                      index;
+
+               /* unlink open-not-created llogs */
+               list_del_init(&loghandle->u.phd.phd_entry);
+               llh = loghandle->lgh_hdr;
+               if (loghandle->lgh_obj != NULL && llh != NULL &&
+                   (llh->llh_flags & LLOG_F_ZAP_WHEN_EMPTY) &&
+                   (llh->llh_count == 1)) {
+                       rc = llog_destroy(env, loghandle);
+                       if (rc)
+                               CERROR("%s: failure destroying log during "
+                                      "cleanup: rc = %d\n",
+                                      loghandle->lgh_ctxt->loc_obd->obd_name,
+                                      rc);
+
+                       index = loghandle->u.phd.phd_cookie.lgc_index;
+                       llog_cat_cleanup(env, cathandle, NULL, index);
+               }
+               llog_close(env, loghandle);
+       }
+       /* if handle was stored in ctxt, remove it too */
+       if (cathandle->lgh_ctxt->loc_handle == cathandle)
+               cathandle->lgh_ctxt->loc_handle = NULL;
+       rc = llog_close(env, cathandle);
+       RETURN(rc);
 }
-EXPORT_SYMBOL(llog_cat_put);
+EXPORT_SYMBOL(llog_cat_close);
 
 /**
  * lockdep markers for nested struct llog_handle::lgh_lock locking.
@@ -226,53 +349,115 @@ enum {
  * NOTE: loghandle is write-locked upon successful return
  */
 static struct llog_handle *llog_cat_current_log(struct llog_handle *cathandle,
-                                                int create)
+                                               struct thandle *th)
 {
         struct llog_handle *loghandle = NULL;
         ENTRY;
 
-        down_read_nested(&cathandle->lgh_lock, LLOGH_CAT);
-        loghandle = cathandle->u.chd.chd_current_log;
-        if (loghandle) {
-                struct llog_log_hdr *llh = loghandle->lgh_hdr;
-                down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
-                if (loghandle->lgh_last_idx < LLOG_BITMAP_SIZE(llh) - 1) {
-                        up_read(&cathandle->lgh_lock);
-                        RETURN(loghandle);
-                } else {
-                        up_write(&loghandle->lgh_lock);
-                }
-        }
-        if (!create) {
-                if (loghandle)
-                        down_write(&loghandle->lgh_lock);
-                up_read(&cathandle->lgh_lock);
-                RETURN(loghandle);
-        }
-        up_read(&cathandle->lgh_lock);
 
-        /* time to create new log */
+       if (OBD_FAIL_CHECK(OBD_FAIL_MDS_LLOG_CREATE_FAILED2)) {
+               down_write_nested(&cathandle->lgh_lock, LLOGH_CAT);
+               GOTO(next, loghandle);
+       }
 
-        /* first, we have to make sure the state hasn't changed */
-        down_write_nested(&cathandle->lgh_lock, LLOGH_CAT);
+       down_read_nested(&cathandle->lgh_lock, LLOGH_CAT);
         loghandle = cathandle->u.chd.chd_current_log;
         if (loghandle) {
-                struct llog_log_hdr *llh = loghandle->lgh_hdr;
-                down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
-                if (loghandle->lgh_last_idx < LLOG_BITMAP_SIZE(llh) - 1) {
-                        up_write(&cathandle->lgh_lock);
+               struct llog_log_hdr *llh;
+
+               down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
+               llh = loghandle->lgh_hdr;
+               if (llh == NULL || !llog_is_full(loghandle)) {
+                       up_read(&cathandle->lgh_lock);
                         RETURN(loghandle);
                 } else {
-                        up_write(&loghandle->lgh_lock);
+                       up_write(&loghandle->lgh_lock);
                 }
         }
+       up_read(&cathandle->lgh_lock);
+
+       /* time to use next log */
+
+       /* first, we have to make sure the state hasn't changed */
+       down_write_nested(&cathandle->lgh_lock, LLOGH_CAT);
+       loghandle = cathandle->u.chd.chd_current_log;
+       if (loghandle) {
+               struct llog_log_hdr *llh;
+
+               down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
+               llh = loghandle->lgh_hdr;
+               LASSERT(llh);
+               if (!llog_is_full(loghandle))
+                       GOTO(out_unlock, loghandle);
+               else
+                       up_write(&loghandle->lgh_lock);
+       }
+
+next:
+       /* Sigh, the chd_next_log and chd_current_log is initialized
+        * in declare phase, and we do not serialize the catlog
+        * accessing, so it might be possible the llog creation
+        * thread (see llog_cat_declare_add_rec()) did not create
+        * llog successfully, then the following thread might
+        * meet this situation. */
+       if (IS_ERR_OR_NULL(cathandle->u.chd.chd_next_log)) {
+               CERROR("%s: next log does not exist!\n",
+                      cathandle->lgh_ctxt->loc_obd->obd_name);
+               loghandle = ERR_PTR(-EIO);
+               if (cathandle->u.chd.chd_next_log == NULL) {
+                       /* Store the error in chd_next_log, so
+                        * the following process can get correct
+                        * failure value */
+                       cathandle->u.chd.chd_next_log = loghandle;
+               }
+               GOTO(out_unlock, loghandle);
+       }
+
+       CDEBUG(D_INODE, "use next log\n");
+
+       loghandle = cathandle->u.chd.chd_next_log;
+       cathandle->u.chd.chd_current_log = loghandle;
+       cathandle->u.chd.chd_next_log = NULL;
+       down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
+
+out_unlock:
+       up_write(&cathandle->lgh_lock);
+       LASSERT(loghandle);
+       RETURN(loghandle);
+}
 
-        CDEBUG(D_INODE, "creating new log\n");
-        loghandle = llog_cat_new_log(cathandle);
-        if (!IS_ERR(loghandle))
-                down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
-        up_write(&cathandle->lgh_lock);
-        RETURN(loghandle);
+static int llog_cat_update_header(const struct lu_env *env,
+                          struct llog_handle *cathandle)
+{
+       struct llog_handle *loghandle;
+       int rc;
+       ENTRY;
+
+       /* refresh llog */
+       down_write(&cathandle->lgh_lock);
+       if (!cathandle->lgh_stale) {
+               up_write(&cathandle->lgh_lock);
+               RETURN(0);
+       }
+       list_for_each_entry(loghandle, &cathandle->u.chd.chd_head,
+                           u.phd.phd_entry) {
+               if (!llog_exist(loghandle))
+                       continue;
+
+               rc = llog_read_header(env, loghandle, NULL);
+               if (rc != 0) {
+                       up_write(&cathandle->lgh_lock);
+                       GOTO(out, rc);
+               }
+       }
+       rc = llog_read_header(env, cathandle, NULL);
+       if (rc == 0)
+               cathandle->lgh_stale = 0;
+       up_write(&cathandle->lgh_lock);
+       if (rc != 0)
+               GOTO(out, rc);
+out:
+       RETURN(rc);
 }
 
 /* Add a single record to the recovery log(s) using a catalog
@@ -280,32 +465,262 @@ static struct llog_handle *llog_cat_current_log(struct llog_handle *cathandle,
  *
  * Assumes caller has already pushed us into the kernel context.
  */
-int llog_cat_add_rec(struct llog_handle *cathandle, struct llog_rec_hdr *rec,
-                     struct llog_cookie *reccookie, void *buf)
+int llog_cat_add_rec(const struct lu_env *env, struct llog_handle *cathandle,
+                    struct llog_rec_hdr *rec, struct llog_cookie *reccookie,
+                    struct thandle *th)
 {
         struct llog_handle *loghandle;
-        int rc;
-        ENTRY;
+       int rc, retried = 0;
+       ENTRY;
+
+       LASSERT(rec->lrh_len <= cathandle->lgh_ctxt->loc_chunk_size);
+
+retry:
+       loghandle = llog_cat_current_log(cathandle, th);
+       if (IS_ERR(loghandle))
+               RETURN(PTR_ERR(loghandle));
+
+       /* loghandle is already locked by llog_cat_current_log() for us */
+       if (!llog_exist(loghandle)) {
+               rc = llog_cat_new_log(env, cathandle, loghandle, th);
+               if (rc < 0) {
+                       up_write(&loghandle->lgh_lock);
+                       /* nobody should be trying to use this llog */
+                       down_write(&cathandle->lgh_lock);
+                       if (cathandle->u.chd.chd_current_log == loghandle)
+                               cathandle->u.chd.chd_current_log = NULL;
+                       up_write(&cathandle->lgh_lock);
+                       RETURN(rc);
+               }
+       }
+       /* now let's try to add the record */
+       rc = llog_write_rec(env, loghandle, rec, reccookie, LLOG_NEXT_IDX, th);
+       if (rc < 0) {
+               CDEBUG_LIMIT(rc == -ENOSPC ? D_HA : D_ERROR,
+                            "llog_write_rec %d: lh=%p\n", rc, loghandle);
+               /* -ENOSPC is returned if no empty records left
+                * and when it's lack of space on the stogage.
+                * there is no point to try again if it's the second
+                * case. many callers (like llog test) expect ENOSPC,
+                * so we preserve this error code, but look for the
+                * actual cause here */
+               if (rc == -ENOSPC && llog_is_full(loghandle))
+                       rc = -ENOBUFS;
+       }
+       up_write(&loghandle->lgh_lock);
+
+       if (rc == -ENOBUFS) {
+               if (retried++ == 0)
+                       GOTO(retry, rc);
+               CERROR("%s: error on 2nd llog: rc = %d\n",
+                      cathandle->lgh_ctxt->loc_obd->obd_name, rc);
+       }
+
+       RETURN(rc);
+}
+EXPORT_SYMBOL(llog_cat_add_rec);
 
-        LASSERT(rec->lrh_len <= LLOG_CHUNK_SIZE);
-        loghandle = llog_cat_current_log(cathandle, 1);
-        if (IS_ERR(loghandle))
-                RETURN(PTR_ERR(loghandle));
-        /* loghandle is already locked by llog_cat_current_log() for us */
-        rc = llog_write_rec(loghandle, rec, reccookie, 1, buf, -1);
-        up_write(&loghandle->lgh_lock);
-        if (rc == -ENOSPC) {
-                /* to create a new plain log */
-                loghandle = llog_cat_current_log(cathandle, 1);
-                if (IS_ERR(loghandle))
-                        RETURN(PTR_ERR(loghandle));
-                rc = llog_write_rec(loghandle, rec, reccookie, 1, buf, -1);
-                up_write(&loghandle->lgh_lock);
-        }
+int llog_cat_declare_add_rec(const struct lu_env *env,
+                            struct llog_handle *cathandle,
+                            struct llog_rec_hdr *rec, struct thandle *th)
+{
+       struct llog_thread_info *lgi = llog_info(env);
+       struct llog_logid_rec   *lirec = &lgi->lgi_logid;
+       struct llog_handle      *loghandle, *next;
+       int                      rc = 0;
+
+       ENTRY;
+
+       if (cathandle->u.chd.chd_current_log == NULL) {
+               /* declare new plain llog */
+               down_write(&cathandle->lgh_lock);
+               if (cathandle->u.chd.chd_current_log == NULL) {
+                       rc = llog_open(env, cathandle->lgh_ctxt, &loghandle,
+                                      NULL, NULL, LLOG_OPEN_NEW);
+                       if (rc == 0) {
+                               cathandle->u.chd.chd_current_log = loghandle;
+                               list_add_tail(&loghandle->u.phd.phd_entry,
+                                             &cathandle->u.chd.chd_head);
+                       }
+               }
+               up_write(&cathandle->lgh_lock);
+       } else if (cathandle->u.chd.chd_next_log == NULL ||
+                  IS_ERR(cathandle->u.chd.chd_next_log)) {
+               /* declare next plain llog */
+               down_write(&cathandle->lgh_lock);
+               if (cathandle->u.chd.chd_next_log == NULL ||
+                   IS_ERR(cathandle->u.chd.chd_next_log)) {
+                       rc = llog_open(env, cathandle->lgh_ctxt, &loghandle,
+                                      NULL, NULL, LLOG_OPEN_NEW);
+                       if (rc == 0) {
+                               cathandle->u.chd.chd_next_log = loghandle;
+                               list_add_tail(&loghandle->u.phd.phd_entry,
+                                             &cathandle->u.chd.chd_head);
+                       }
+               }
+               up_write(&cathandle->lgh_lock);
+       }
+       if (rc)
+               GOTO(out, rc);
+
+       lirec->lid_hdr.lrh_len = sizeof(*lirec);
+
+       if (!llog_exist(cathandle->u.chd.chd_current_log)) {
+               if (dt_object_remote(cathandle->lgh_obj)) {
+                       /* For remote operation, if we put the llog object
+                        * creation in the current transaction, then the
+                        * llog object will not be created on the remote
+                        * target until the transaction stop, if other
+                        * operations start before the transaction stop,
+                        * and use the same llog object, will be dependent
+                        * on the success of this transaction. So let's
+                        * create the llog object synchronously here to
+                        * remove the dependency. */
+create_again:
+                       down_read_nested(&cathandle->lgh_lock, LLOGH_CAT);
+                       loghandle = cathandle->u.chd.chd_current_log;
+                       down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
+                       if (cathandle->lgh_stale) {
+                               up_write(&loghandle->lgh_lock);
+                               up_read(&cathandle->lgh_lock);
+                               GOTO(out, rc = -EIO);
+                       }
+                       if (!llog_exist(loghandle)) {
+                               rc = llog_cat_new_log(env, cathandle, loghandle,
+                                                     NULL);
+                               if (rc == -ESTALE)
+                                       cathandle->lgh_stale = 1;
+                       }
+                       up_write(&loghandle->lgh_lock);
+                       up_read(&cathandle->lgh_lock);
+                       if (rc == -ESTALE) {
+                               rc = llog_cat_update_header(env, cathandle);
+                               if (rc != 0)
+                                       GOTO(out, rc);
+                               goto create_again;
+                       } else if (rc < 0) {
+                               GOTO(out, rc);
+                       }
+               } else {
+                       rc = llog_declare_create(env,
+                                       cathandle->u.chd.chd_current_log, th);
+                       if (rc)
+                               GOTO(out, rc);
+                       llog_declare_write_rec(env, cathandle,
+                                              &lirec->lid_hdr, -1, th);
+               }
+       }
+
+write_again:
+       /* declare records in the llogs */
+       rc = llog_declare_write_rec(env, cathandle->u.chd.chd_current_log,
+                                   rec, -1, th);
+       if (rc == -ESTALE) {
+               down_write(&cathandle->lgh_lock);
+               if (cathandle->lgh_stale) {
+                       up_write(&cathandle->lgh_lock);
+                       GOTO(out, rc = -EIO);
+               }
+
+               cathandle->lgh_stale = 1;
+               up_write(&cathandle->lgh_lock);
+               rc = llog_cat_update_header(env, cathandle);
+               if (rc != 0)
+                       GOTO(out, rc);
+               goto write_again;
+       } else if (rc < 0) {
+               GOTO(out, rc);
+       }
+
+       next = cathandle->u.chd.chd_next_log;
+       if (!IS_ERR_OR_NULL(next)) {
+               if (!llog_exist(next)) {
+                       if (dt_object_remote(cathandle->lgh_obj)) {
+                               /* For remote operation, if we put the llog
+                                * object creation in the current transaction,
+                                * then the llog object will not be created on
+                                * the remote target until the transaction stop,
+                                * if other operations start before the
+                                * transaction stop, and use the same llog
+                                * object, will be dependent on the success of
+                                * this transaction. So let's create the llog
+                                * object synchronously here to remove the
+                                * dependency. */
+                               down_write_nested(&cathandle->lgh_lock,
+                                                LLOGH_CAT);
+                               next = cathandle->u.chd.chd_next_log;
+                               if (IS_ERR_OR_NULL(next)) {
+                                       /* Sigh, another thread just tried,
+                                        * let's fail as well */
+                                       up_write(&cathandle->lgh_lock);
+                                       if (next == NULL)
+                                               rc = -EIO;
+                                       else
+                                               rc = PTR_ERR(next);
+                                       GOTO(out, rc);
+                               }
+
+                               down_write_nested(&next->lgh_lock, LLOGH_LOG);
+                               if (!llog_exist(next)) {
+                                       rc = llog_cat_new_log(env, cathandle,
+                                                             next, NULL);
+                                       if (rc < 0)
+                                               cathandle->u.chd.chd_next_log =
+                                                               ERR_PTR(rc);
+                               }
+                               up_write(&next->lgh_lock);
+                               up_write(&cathandle->lgh_lock);
+                               if (rc < 0)
+                                       GOTO(out, rc);
+                       } else {
+                               rc = llog_declare_create(env, next, th);
+                               llog_declare_write_rec(env, cathandle,
+                                               &lirec->lid_hdr, -1, th);
+                       }
+               }
+               /* XXX: we hope for declarations made for existing llog
+                *      this might be not correct with some backends
+                *      where declarations are expected against specific
+                *      object like ZFS with full debugging enabled */
+               /*llog_declare_write_rec(env, next, rec, -1, th);*/
+       }
+out:
+       RETURN(rc);
+}
+EXPORT_SYMBOL(llog_cat_declare_add_rec);
 
-        RETURN(rc);
+int llog_cat_add(const struct lu_env *env, struct llog_handle *cathandle,
+                struct llog_rec_hdr *rec, struct llog_cookie *reccookie)
+{
+       struct llog_ctxt        *ctxt;
+       struct dt_device        *dt;
+       struct thandle          *th = NULL;
+       int                      rc;
+
+       ctxt = cathandle->lgh_ctxt;
+       LASSERT(ctxt);
+       LASSERT(ctxt->loc_exp);
+
+       LASSERT(cathandle->lgh_obj != NULL);
+       dt = lu2dt_dev(cathandle->lgh_obj->do_lu.lo_dev);
+
+       th = dt_trans_create(env, dt);
+       if (IS_ERR(th))
+               RETURN(PTR_ERR(th));
+
+       rc = llog_cat_declare_add_rec(env, cathandle, rec, th);
+       if (rc)
+               GOTO(out_trans, rc);
+
+       rc = dt_trans_start_local(env, dt, th);
+       if (rc)
+               GOTO(out_trans, rc);
+       rc = llog_cat_add_rec(env, cathandle, rec, reccookie, th);
+out_trans:
+       dt_trans_stop(env, dt, th);
+       RETURN(rc);
 }
-EXPORT_SYMBOL(llog_cat_add_rec);
+EXPORT_SYMBOL(llog_cat_add);
 
 /* For each cookie in the cookie array, we clear the log in-use bit and either:
  * - the log is empty, so mark it free in the catalog header and delete it
@@ -316,80 +731,182 @@ EXPORT_SYMBOL(llog_cat_add_rec);
  *
  * Assumes caller has already pushed us into the kernel context.
  */
-int llog_cat_cancel_records(struct llog_handle *cathandle, int count,
-                        struct llog_cookie *cookies)
+int llog_cat_cancel_records(const struct lu_env *env,
+                           struct llog_handle *cathandle, int count,
+                           struct llog_cookie *cookies)
 {
-        int i, index, rc = 0;
-        ENTRY;
-
-        down_write_nested(&cathandle->lgh_lock, LLOGH_CAT);
-        for (i = 0; i < count; i++, cookies++) {
-                struct llog_handle *loghandle;
-                struct llog_logid *lgl = &cookies->lgc_lgl;
-
-                rc = llog_cat_id2handle(cathandle, &loghandle, lgl);
-                if (rc) {
-                        CERROR("Cannot find log "LPX64"\n", lgl->lgl_oid);
-                        break;
-                }
-
-                down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
-                rc = llog_cancel_rec(loghandle, cookies->lgc_index);
-                up_write(&loghandle->lgh_lock);
-
-                if (rc == 1) {          /* log has been destroyed */
-                        index = loghandle->u.phd.phd_cookie.lgc_index;
-                        if (cathandle->u.chd.chd_current_log == loghandle)
-                                cathandle->u.chd.chd_current_log = NULL;
-                        llog_free_handle(loghandle);
-
-                        LASSERT(index);
-                        llog_cat_set_first_idx(cathandle, index);
-                        rc = llog_cancel_rec(cathandle, index);
-                        if (rc == 0)
-                                CDEBUG(D_RPCTRACE,"cancel plain log at index %u"
-                                       " of catalog "LPX64"\n",
-                                       index, cathandle->lgh_id.lgl_oid);
-                }
-        }
-        up_write(&cathandle->lgh_lock);
-
-        RETURN(rc);
+       int i, index, rc = 0, failed = 0;
+
+       ENTRY;
+
+       for (i = 0; i < count; i++, cookies++) {
+               struct llog_handle *loghandle;
+               struct llog_logid *lgl = &cookies->lgc_lgl;
+               int  lrc;
+
+               rc = llog_cat_id2handle(env, cathandle, &loghandle, lgl);
+               if (rc) {
+                       CDEBUG(D_HA, "%s: cannot find llog for handle "DFID":%x"
+                              ": rc = %d\n",
+                              cathandle->lgh_ctxt->loc_obd->obd_name,
+                              PFID(&lgl->lgl_oi.oi_fid), lgl->lgl_ogen, rc);
+                       failed++;
+                       continue;
+               }
+
+               if ((cathandle->lgh_ctxt->loc_flags &
+                    LLOG_CTXT_FLAG_NORMAL_FID) && !llog_exist(loghandle)) {
+                       /* For update log, some of loghandles of cathandle
+                        * might not exist because remote llog creation might
+                        * be failed, so let's skip the record cancellation
+                        * for these non-exist llogs.
+                        */
+                       lrc = -ENOENT;
+                       CDEBUG(D_HA, "%s: llog "DFID":%x does not exist"
+                              ": rc = %d\n",
+                              cathandle->lgh_ctxt->loc_obd->obd_name,
+                              PFID(&lgl->lgl_oi.oi_fid), lgl->lgl_ogen, lrc);
+                       failed++;
+                       if (rc == 0)
+                               rc = lrc;
+                       continue;
+               }
+
+               lrc = llog_cancel_rec(env, loghandle, cookies->lgc_index);
+               if (lrc == LLOG_DEL_PLAIN) { /* log has been destroyed */
+                       index = loghandle->u.phd.phd_cookie.lgc_index;
+                       lrc = llog_cat_cleanup(env, cathandle, loghandle,
+                                              index);
+                       if (rc == 0)
+                               rc = lrc;
+               } else if (lrc == -ENOENT) {
+                       if (rc == 0) /* ENOENT shouldn't rewrite any error */
+                               rc = lrc;
+               } else if (lrc < 0) {
+                       failed++;
+                       if (rc == 0)
+                               rc = lrc;
+               }
+               llog_handle_put(loghandle);
+       }
+       if (rc)
+               CERROR("%s: fail to cancel %d of %d llog-records: rc = %d\n",
+                      cathandle->lgh_ctxt->loc_obd->obd_name, failed, count,
+                      rc);
+
+       RETURN(rc);
 }
 EXPORT_SYMBOL(llog_cat_cancel_records);
 
-int llog_cat_process_cb(struct llog_handle *cat_llh, struct llog_rec_hdr *rec,
-                        void *data)
+static int llog_cat_process_common(const struct lu_env *env,
+                                  struct llog_handle *cat_llh,
+                                  struct llog_rec_hdr *rec,
+                                  struct llog_handle **llhp)
 {
-        struct llog_process_data *d = data;
-        struct llog_logid_rec *lir = (struct llog_logid_rec *)rec;
-        struct llog_handle *llh;
-        int rc;
+       struct llog_logid_rec *lir = container_of(rec, typeof(*lir), lid_hdr);
+       struct llog_log_hdr *hdr;
+       int rc;
+
+       ENTRY;
+       if (rec->lrh_type != le32_to_cpu(LLOG_LOGID_MAGIC)) {
+               rc = -EINVAL;
+               CWARN("%s: invalid record in catalog "DFID":%x: rc = %d\n",
+                     cat_llh->lgh_ctxt->loc_obd->obd_name,
+                     PFID(&cat_llh->lgh_id.lgl_oi.oi_fid),
+                     cat_llh->lgh_id.lgl_ogen, rc);
+               RETURN(rc);
+       }
+       CDEBUG(D_HA, "processing log "DFID":%x at index %u of catalog "DFID"\n",
+              PFID(&lir->lid_id.lgl_oi.oi_fid), lir->lid_id.lgl_ogen,
+              le32_to_cpu(rec->lrh_index),
+              PFID(&cat_llh->lgh_id.lgl_oi.oi_fid));
+
+       rc = llog_cat_id2handle(env, cat_llh, llhp, &lir->lid_id);
+       if (rc) {
+               /* After a server crash, a stub of index record in catlog could
+                * be kept, because plain log destroy + catlog index record
+                * deletion are not atomic. So we end up with an index but no
+                * actual record. Destroy the index and move on. */
+               if (rc == -ENOENT || rc == -ESTALE)
+                       rc = LLOG_DEL_RECORD;
+               else if (rc)
+                       CWARN("%s: can't find llog handle "DFID":%x: rc = %d\n",
+                             cat_llh->lgh_ctxt->loc_obd->obd_name,
+                             PFID(&lir->lid_id.lgl_oi.oi_fid),
+                             lir->lid_id.lgl_ogen, rc);
+
+               RETURN(rc);
+       }
+
+       /* clean old empty llogs, do not consider current llog in use */
+       /* ignore remote (lgh_obj == NULL) llogs */
+       hdr = (*llhp)->lgh_hdr;
+       if ((hdr->llh_flags & LLOG_F_ZAP_WHEN_EMPTY) &&
+           hdr->llh_count == 1 && cat_llh->lgh_obj != NULL &&
+           *llhp != cat_llh->u.chd.chd_current_log) {
+               rc = llog_destroy(env, *llhp);
+               if (rc)
+                       CWARN("%s: can't destroy empty log "DFID": rc = %d\n",
+                             (*llhp)->lgh_ctxt->loc_obd->obd_name,
+                             PFID(&lir->lid_id.lgl_oi.oi_fid), rc);
+               rc = LLOG_DEL_PLAIN;
+       }
+
+       RETURN(rc);
+}
 
-        ENTRY;
-        if (rec->lrh_type != LLOG_LOGID_MAGIC) {
-                CERROR("invalid record in catalog\n");
-                RETURN(-EINVAL);
-        }
-        CDEBUG(D_HA, "processing log "LPX64":%x at index %u of catalog "
-               LPX64"\n", lir->lid_id.lgl_oid, lir->lid_id.lgl_ogen,
-               rec->lrh_index, cat_llh->lgh_id.lgl_oid);
-
-        rc = llog_cat_id2handle(cat_llh, &llh, &lir->lid_id);
-        if (rc) {
-                CERROR("Cannot find handle for log "LPX64"\n",
-                       lir->lid_id.lgl_oid);
-                RETURN(rc);
-        }
+static int llog_cat_process_cb(const struct lu_env *env,
+                              struct llog_handle *cat_llh,
+                              struct llog_rec_hdr *rec, void *data)
+{
+       struct llog_process_data *d = data;
+       struct llog_handle *llh = NULL;
+       int rc;
+
+       ENTRY;
+       rc = llog_cat_process_common(env, cat_llh, rec, &llh);
+       if (rc)
+               GOTO(out, rc);
+
+       if (rec->lrh_index < d->lpd_startcat) {
+               /* Skip processing of the logs until startcat */
+               rc = 0;
+       } else if (d->lpd_startidx > 0) {
+                struct llog_process_cat_data cd;
+
+                cd.lpcd_first_idx = d->lpd_startidx;
+                cd.lpcd_last_idx = 0;
+               rc = llog_process_or_fork(env, llh, d->lpd_cb, d->lpd_data,
+                                         &cd, false);
+               /* Continue processing the next log from idx 0 */
+               d->lpd_startidx = 0;
+       } else {
+               rc = llog_process_or_fork(env, llh, d->lpd_cb, d->lpd_data,
+                                         NULL, false);
+       }
 
-        rc = llog_process(llh, d->lpd_cb, d->lpd_data, NULL);
-        RETURN(rc);
+out:
+       /* The empty plain log was destroyed while processing */
+       if (rc == LLOG_DEL_PLAIN) {
+               rc = llog_cat_cleanup(env, cat_llh, llh,
+                                     llh->u.phd.phd_cookie.lgc_index);
+       } else if (rc == LLOG_DEL_RECORD) {
+               /* clear wrong catalog entry */
+               rc = llog_cat_cleanup(env, cat_llh, NULL, rec->lrh_index);
+       }
+
+       if (llh)
+               llog_handle_put(llh);
+
+       RETURN(rc);
 }
 
-int llog_cat_process(struct llog_handle *cat_llh, llog_cb_t cb, void *data)
+int llog_cat_process_or_fork(const struct lu_env *env,
+                            struct llog_handle *cat_llh, llog_cb_t cat_cb,
+                            llog_cb_t cb, void *data, int startcat,
+                            int startidx, bool fork)
 {
         struct llog_process_data d;
-        struct llog_process_cat_data cd;
         struct llog_log_hdr *llh = cat_llh->lgh_hdr;
         int rc;
         ENTRY;
@@ -397,107 +914,147 @@ int llog_cat_process(struct llog_handle *cat_llh, llog_cb_t cb, void *data)
         LASSERT(llh->llh_flags & LLOG_F_IS_CAT);
         d.lpd_data = data;
         d.lpd_cb = cb;
-
-        if (llh->llh_cat_idx > cat_llh->lgh_last_idx) {
-                CWARN("catlog "LPX64" crosses index zero\n",
-                      cat_llh->lgh_id.lgl_oid);
-
-                cd.lpcd_first_idx = llh->llh_cat_idx;
-                cd.lpcd_last_idx = 0;
-                rc = llog_process(cat_llh, llog_cat_process_cb, &d, &cd);
-                if (rc != 0)
-                        RETURN(rc);
-
-                cd.lpcd_first_idx = 0;
-                cd.lpcd_last_idx = cat_llh->lgh_last_idx;
-                rc = llog_process(cat_llh, llog_cat_process_cb, &d, &cd);
+        d.lpd_startcat = startcat;
+        d.lpd_startidx = startidx;
+
+       if (llh->llh_cat_idx >= cat_llh->lgh_last_idx &&
+           llh->llh_count > 1) {
+               struct llog_process_cat_data cd;
+
+               CWARN("%s: catlog "DFID" crosses index zero\n",
+                     cat_llh->lgh_ctxt->loc_obd->obd_name,
+                     PFID(&cat_llh->lgh_id.lgl_oi.oi_fid));
+
+               cd.lpcd_first_idx = llh->llh_cat_idx;
+               cd.lpcd_last_idx = 0;
+               rc = llog_process_or_fork(env, cat_llh, cat_cb,
+                                         &d, &cd, fork);
+               if (rc != 0)
+                       RETURN(rc);
+
+               cd.lpcd_first_idx = 0;
+               cd.lpcd_last_idx = cat_llh->lgh_last_idx;
+               rc = llog_process_or_fork(env, cat_llh, cat_cb,
+                                         &d, &cd, fork);
         } else {
-                rc = llog_process(cat_llh, llog_cat_process_cb, &d, NULL);
+               rc = llog_process_or_fork(env, cat_llh, cat_cb,
+                                         &d, NULL, fork);
         }
 
         RETURN(rc);
 }
-EXPORT_SYMBOL(llog_cat_process);
 
-#ifdef __KERNEL__
-int llog_cat_process_thread(void *data)
+int llog_cat_process(const struct lu_env *env, struct llog_handle *cat_llh,
+                    llog_cb_t cb, void *data, int startcat, int startidx)
 {
-        struct llog_process_cat_args *args = data;
-        struct llog_ctxt *ctxt = args->lpca_ctxt;
-        struct llog_handle *llh = NULL;
-        void  *cb = args->lpca_cb;
-        struct llog_logid logid;
-        int rc;
-        ENTRY;
+       return llog_cat_process_or_fork(env, cat_llh, llog_cat_process_cb,
+                                       cb, data, startcat, startidx, false);
+}
+EXPORT_SYMBOL(llog_cat_process);
 
-        cfs_daemonize_ctxt("ll_log_process");
+static int llog_cat_size_cb(const struct lu_env *env,
+                            struct llog_handle *cat_llh,
+                            struct llog_rec_hdr *rec, void *data)
+{
+       struct llog_process_data *d = data;
+       struct llog_handle *llh = NULL;
+       __u64 *cum_size = d->lpd_data;
+       __u64 size;
+       int rc;
+
+       ENTRY;
+       rc = llog_cat_process_common(env, cat_llh, rec, &llh);
+
+       if (rc == LLOG_DEL_PLAIN) {
+               /* empty log was deleted, don't count it */
+               rc = llog_cat_cleanup(env, cat_llh, llh,
+                                     llh->u.phd.phd_cookie.lgc_index);
+       } else if (rc == LLOG_DEL_RECORD) {
+               /* clear wrong catalog entry */
+               rc = llog_cat_cleanup(env, cat_llh, NULL, rec->lrh_index);
+       } else {
+               size = llog_size(env, llh);
+               *cum_size += size;
+
+               CDEBUG(D_INFO, "Add llog entry "DFID" size=%llu, tot=%llu\n",
+                      PFID(&llh->lgh_id.lgl_oi.oi_fid), size, *cum_size);
+       }
+
+       if (llh != NULL)
+               llog_handle_put(llh);
+
+       RETURN(0);
+}
 
-        logid = *(struct llog_logid *)(args->lpca_arg);
-        rc = llog_create(ctxt, &llh, &logid, NULL);
-        if (rc) {
-                CERROR("llog_create() failed %d\n", rc);
-                GOTO(out, rc);
-        }
-        rc = llog_init_handle(llh, LLOG_F_IS_CAT, NULL);
-        if (rc) {
-                CERROR("llog_init_handle failed %d\n", rc);
-                GOTO(release_llh, rc);
-        }
+__u64 llog_cat_size(const struct lu_env *env, struct llog_handle *cat_llh)
+{
+       __u64 size = llog_size(env, cat_llh);
 
-        if (cb) {
-                rc = llog_cat_process(llh, (llog_cb_t)cb, NULL);
-                if (rc != LLOG_PROC_BREAK)
-                        CERROR("llog_cat_process() failed %d\n", rc);
-        } else {
-                CWARN("No callback function for recovery\n");
-        }
+       llog_cat_process_or_fork(env, cat_llh, llog_cat_size_cb,
+                                NULL, &size, 0, 0, false);
 
-        /* 
-         * Make sure that all cached data is sent. 
-         */
-        llog_sync(ctxt, NULL);
-        GOTO(release_llh, rc);
-release_llh:
-        rc = llog_cat_put(llh);
-        if (rc)
-                CERROR("llog_cat_put() failed %d\n", rc);
-out:
-        llog_ctxt_put(ctxt);
-        OBD_FREE_PTR(args);
-        return rc;
+       return size;
 }
-EXPORT_SYMBOL(llog_cat_process_thread);
-#endif
+EXPORT_SYMBOL(llog_cat_size);
 
-static int llog_cat_reverse_process_cb(struct llog_handle *cat_llh,
-                                       struct llog_rec_hdr *rec, void *data)
+/* currently returns the number of "free" entries in catalog,
+ * ie the available entries for a new plain LLOG file creation,
+ * even if catalog has wrapped
+ */
+__u32 llog_cat_free_space(struct llog_handle *cat_llh)
 {
-        struct llog_process_data *d = data;
-        struct llog_logid_rec *lir = (struct llog_logid_rec *)rec;
-        struct llog_handle *llh;
-        int rc;
+       /* simulate almost full Catalog */
+       if (OBD_FAIL_CHECK(OBD_FAIL_CAT_FREE_RECORDS))
+               return cfs_fail_val;
 
-        if (le32_to_cpu(rec->lrh_type) != LLOG_LOGID_MAGIC) {
-                CERROR("invalid record in catalog\n");
-                RETURN(-EINVAL);
-        }
-        CDEBUG(D_HA, "processing log "LPX64":%x at index %u of catalog "
-               LPX64"\n", lir->lid_id.lgl_oid, lir->lid_id.lgl_ogen,
-               le32_to_cpu(rec->lrh_index), cat_llh->lgh_id.lgl_oid);
-
-        rc = llog_cat_id2handle(cat_llh, &llh, &lir->lid_id);
-        if (rc) {
-                CERROR("Cannot find handle for log "LPX64"\n",
-                       lir->lid_id.lgl_oid);
-                RETURN(rc);
-        }
+       if (cat_llh->lgh_hdr->llh_count == 1)
+               return LLOG_HDR_BITMAP_SIZE(cat_llh->lgh_hdr) - 1;
 
-        rc = llog_reverse_process(llh, d->lpd_cb, d->lpd_data, NULL);
-        RETURN(rc);
+       if (cat_llh->lgh_last_idx > cat_llh->lgh_hdr->llh_cat_idx)
+               return LLOG_HDR_BITMAP_SIZE(cat_llh->lgh_hdr) - 1 +
+                      cat_llh->lgh_hdr->llh_cat_idx - cat_llh->lgh_last_idx;
+
+       /* catalog is presently wrapped */
+       return cat_llh->lgh_hdr->llh_cat_idx - cat_llh->lgh_last_idx;
+}
+EXPORT_SYMBOL(llog_cat_free_space);
+
+static int llog_cat_reverse_process_cb(const struct lu_env *env,
+                                      struct llog_handle *cat_llh,
+                                      struct llog_rec_hdr *rec, void *data)
+{
+       struct llog_process_data *d = data;
+       struct llog_handle *llh;
+       int rc;
+
+       ENTRY;
+       rc = llog_cat_process_common(env, cat_llh, rec, &llh);
+
+       /* The empty plain log was destroyed while processing */
+       if (rc == LLOG_DEL_PLAIN) {
+               rc = llog_cat_cleanup(env, cat_llh, llh,
+                                     llh->u.phd.phd_cookie.lgc_index);
+       } else if (rc == LLOG_DEL_RECORD) {
+               /* clear wrong catalog entry */
+               rc = llog_cat_cleanup(env, cat_llh, NULL, rec->lrh_index);
+       }
+       if (rc)
+               RETURN(rc);
+
+       rc = llog_reverse_process(env, llh, d->lpd_cb, d->lpd_data, NULL);
+
+       /* The empty plain was destroyed while processing */
+       if (rc == LLOG_DEL_PLAIN)
+               rc = llog_cat_cleanup(env, cat_llh, llh,
+                                     llh->u.phd.phd_cookie.lgc_index);
+
+       llog_handle_put(llh);
+       RETURN(rc);
 }
 
-int llog_cat_reverse_process(struct llog_handle *cat_llh,
-                             llog_cb_t cb, void *data)
+int llog_cat_reverse_process(const struct lu_env *env,
+                            struct llog_handle *cat_llh,
+                            llog_cb_t cb, void *data)
 {
         struct llog_process_data d;
         struct llog_process_cat_data cd;
@@ -509,58 +1066,104 @@ int llog_cat_reverse_process(struct llog_handle *cat_llh,
         d.lpd_data = data;
         d.lpd_cb = cb;
 
-        if (llh->llh_cat_idx > cat_llh->lgh_last_idx) {
-                CWARN("catalog "LPX64" crosses index zero\n",
-                      cat_llh->lgh_id.lgl_oid);
-
-                cd.lpcd_first_idx = 0;
-                cd.lpcd_last_idx = cat_llh->lgh_last_idx;
-                rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
-                                          &d, &cd);
-                if (rc != 0)
-                        RETURN(rc);
-
-                cd.lpcd_first_idx = le32_to_cpu(llh->llh_cat_idx);
-                cd.lpcd_last_idx = 0;
-                rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
-                                          &d, &cd);
+       if (llh->llh_cat_idx >= cat_llh->lgh_last_idx &&
+           llh->llh_count > 1) {
+               CWARN("%s: catalog "DFID" crosses index zero\n",
+                     cat_llh->lgh_ctxt->loc_obd->obd_name,
+                     PFID(&cat_llh->lgh_id.lgl_oi.oi_fid));
+
+               cd.lpcd_first_idx = 0;
+               cd.lpcd_last_idx = cat_llh->lgh_last_idx;
+               rc = llog_reverse_process(env, cat_llh,
+                                         llog_cat_reverse_process_cb,
+                                         &d, &cd);
+               if (rc != 0)
+                       RETURN(rc);
+
+               cd.lpcd_first_idx = le32_to_cpu(llh->llh_cat_idx);
+               cd.lpcd_last_idx = 0;
+               rc = llog_reverse_process(env, cat_llh,
+                                         llog_cat_reverse_process_cb,
+                                         &d, &cd);
         } else {
-                rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
-                                          &d, NULL);
+               rc = llog_reverse_process(env, cat_llh,
+                                         llog_cat_reverse_process_cb,
+                                         &d, NULL);
         }
 
         RETURN(rc);
 }
 EXPORT_SYMBOL(llog_cat_reverse_process);
 
-int llog_cat_set_first_idx(struct llog_handle *cathandle, int index)
+static int llog_cat_set_first_idx(struct llog_handle *cathandle, int idx)
 {
-        struct llog_log_hdr *llh = cathandle->lgh_hdr;
-        int i, bitmap_size, idx;
-        ENTRY;
-
-        bitmap_size = LLOG_BITMAP_SIZE(llh);
-        if (llh->llh_cat_idx == (index - 1)) {
-                idx = llh->llh_cat_idx + 1;
-                llh->llh_cat_idx = idx;
-                if (idx == cathandle->lgh_last_idx)
-                        goto out;
-                for (i = (index + 1) % bitmap_size;
-                     i != cathandle->lgh_last_idx;
-                     i = (i + 1) % bitmap_size) {
-                        if (!ext2_test_bit(i, llh->llh_bitmap)) {
-                                idx = llh->llh_cat_idx + 1;
-                                llh->llh_cat_idx = idx;
-                        } else if (i == 0) {
-                                llh->llh_cat_idx = 0;
-                        } else {
-                                break;
-                        }
-                }
-out:
-                CDEBUG(D_RPCTRACE, "set catlog "LPX64" first idx %u\n",
-                       cathandle->lgh_id.lgl_oid, llh->llh_cat_idx);
-        }
+       struct llog_log_hdr *llh = cathandle->lgh_hdr;
+       int bitmap_size;
+
+       ENTRY;
+
+       bitmap_size = LLOG_HDR_BITMAP_SIZE(llh);
+       /*
+        * The llh_cat_idx equals to the first used index minus 1
+        * so if we canceled the first index then llh_cat_idx
+        * must be renewed.
+        */
+       if (llh->llh_cat_idx == (idx - 1)) {
+               llh->llh_cat_idx = idx;
+
+               while (idx != cathandle->lgh_last_idx) {
+                       idx = (idx + 1) % bitmap_size;
+                       if (!ext2_test_bit(idx, LLOG_HDR_BITMAP(llh))) {
+                               /* update llh_cat_idx for each unset bit,
+                                * expecting the next one is set */
+                               llh->llh_cat_idx = idx;
+                       } else if (idx == 0) {
+                               /* skip header bit */
+                               llh->llh_cat_idx = 0;
+                               continue;
+                       } else {
+                               /* the first index is found */
+                               break;
+                       }
+               }
+
+               CDEBUG(D_RPCTRACE, "catlog "DFID" first idx %u, last_idx %u\n",
+                      PFID(&cathandle->lgh_id.lgl_oi.oi_fid),
+                      llh->llh_cat_idx, cathandle->lgh_last_idx);
+       }
+
+       RETURN(0);
+}
 
-        RETURN(0);
+/* Cleanup deleted plain llog traces from catalog */
+int llog_cat_cleanup(const struct lu_env *env, struct llog_handle *cathandle,
+                    struct llog_handle *loghandle, int index)
+{
+       int rc;
+
+       LASSERT(index);
+       if (loghandle != NULL) {
+               /* remove destroyed llog from catalog list and
+                * chd_current_log variable */
+               down_write(&cathandle->lgh_lock);
+               if (cathandle->u.chd.chd_current_log == loghandle)
+                       cathandle->u.chd.chd_current_log = NULL;
+               list_del_init(&loghandle->u.phd.phd_entry);
+               up_write(&cathandle->lgh_lock);
+               LASSERT(index == loghandle->u.phd.phd_cookie.lgc_index);
+               /* llog was opened and keep in a list, close it now */
+               llog_close(env, loghandle);
+       }
+
+       /* do not attempt to cleanup on-disk llog if on client side */
+       if (cathandle->lgh_obj == NULL)
+               return 0;
+
+       /* remove plain llog entry from catalog by index */
+       llog_cat_set_first_idx(cathandle, index);
+       rc = llog_cancel_rec(env, cathandle, index);
+       if (rc == 0)
+               CDEBUG(D_HA, "cancel plain log at index %u of catalog "DFID"\n",
+                      index, PFID(&cathandle->lgh_id.lgl_oi.oi_fid));
+       return rc;
 }