X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fobdclass%2Fllog_cat.c;h=c8c2da3eca18727e4250d2599974a7f108d9da15;hp=00151a69a14187ea7ae4ed7992d06ab35de07251;hb=485f3ba8743365c7fb5823893a7eb45030cdffb5;hpb=fbf5870b9848929d352460f1f005b79c0b5ccc5a

diff --git a/lustre/obdclass/llog_cat.c b/lustre/obdclass/llog_cat.c
index 00151a6..c8c2da3 100644
--- a/lustre/obdclass/llog_cat.c
+++ b/lustre/obdclass/llog_cat.c
@@ -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.
@@ -26,8 +24,10 @@
  * 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, 2014, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
@@ -42,98 +42,74 @@
  *   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);
-        }
-
-        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);
-
+	struct llog_thread_info	*lgi = llog_info(env);
+	struct llog_logid_rec	*rec = &lgi->lgi_logid;
+	int			 rc;
+
+	ENTRY;
+
+	if (OBD_FAIL_CHECK(OBD_FAIL_MDS_LLOG_CREATE_FAILED))
+		RETURN(-ENOSPC);
+
+	rc = llog_create(env, loghandle, th);
+	/* if llog is already created, no need to initialize it */
+	if (rc == -EEXIST) {
+		RETURN(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);
+		RETURN(rc);
+	}
+
+	rc = llog_init_handle(env, loghandle,
+			      LLOG_F_IS_PLAIN | LLOG_F_ZAP_WHEN_EMPTY,
+			      &cathandle->lgh_hdr->llh_tgtuuid);
+	if (rc)
+		GOTO(out_destroy, 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 recovery log "DOSTID":%x for index %u of catalog"
+	       DOSTID"\n", POSTID(&loghandle->lgh_id.lgl_oi),
+	       loghandle->lgh_id.lgl_ogen, rec->lid_hdr.lrh_index,
+	       POSTID(&cathandle->lgh_id.lgl_oi));
+
+	loghandle->lgh_hdr->llh_cat_idx = rec->lid_hdr.lrh_index;
+	RETURN(0);
 out_destroy:
-        if (rc < 0)
-                llog_destroy(loghandle);
-
-        RETURN(loghandle);
+	llog_destroy(env, loghandle);
+	RETURN(rc);
 }
 
 /* Open an existent log handle and add it to the open list.
@@ -141,71 +117,112 @@ 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) {
+				CERROR("%s: log "DOSTID" generation %x != %x\n",
+				       loghandle->lgh_ctxt->loc_obd->obd_name,
+				       POSTID(&logid->lgl_oi), 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 "DOSTID":%x: rc = %d\n",
+		       cathandle->lgh_ctxt->loc_obd->obd_name,
+		       POSTID(&logid->lgl_oi), 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 +243,56 @@ 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);
+	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);
+		struct llog_log_hdr *llh;
+
+		down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
+		llh = loghandle->lgh_hdr;
+		if (llh == NULL ||
+		    loghandle->lgh_last_idx < LLOG_BITMAP_SIZE(llh) - 1) {
+			up_read(&cathandle->lgh_lock);
                         RETURN(loghandle);
                 } else {
-                        up_write(&loghandle->lgh_lock);
+			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);
+	up_read(&cathandle->lgh_lock);
 
-        /* time to create new log */
+	/* 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 = loghandle->lgh_hdr;
-                down_write_nested(&loghandle->lgh_lock, LLOGH_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 (loghandle->lgh_last_idx < LLOG_BITMAP_SIZE(llh) - 1) {
-                        up_write(&cathandle->lgh_lock);
+			up_write(&cathandle->lgh_lock);
                         RETURN(loghandle);
                 } else {
-                        up_write(&loghandle->lgh_lock);
+			up_write(&loghandle->lgh_lock);
                 }
         }
 
-        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);
+	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);
+	up_write(&cathandle->lgh_lock);
+	LASSERT(loghandle);
+	RETURN(loghandle);
 }
 
 /* Add a single record to the recovery log(s) using a catalog
@@ -280,33 +300,163 @@ 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;
 
         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);
+	loghandle = llog_cat_current_log(cathandle, th);
+	LASSERT(!IS_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);
+			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);
+	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);
-        }
-
-        RETURN(rc);
+		/* try to use next log */
+		loghandle = llog_cat_current_log(cathandle, th);
+		LASSERT(!IS_ERR(loghandle));
+		/* new llog can be created concurrently */
+		if (!llog_exist(loghandle)) {
+			rc = llog_cat_new_log(env, cathandle, loghandle, th);
+			if (rc < 0) {
+				up_write(&loghandle->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)
+			CERROR("llog_write_rec %d: lh=%p\n", rc, loghandle);
+		up_write(&loghandle->lgh_lock);
+	}
+
+	RETURN(rc);
 }
 EXPORT_SYMBOL(llog_cat_add_rec);
 
+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) {
+		/* declare next plain llog */
+		down_write(&cathandle->lgh_lock);
+		if (cathandle->u.chd.chd_next_log == NULL) {
+			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)) {
+		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);
+	}
+	/* declare records in the llogs */
+	rc = llog_declare_write_rec(env, cathandle->u.chd.chd_current_log,
+				    rec, -1, th);
+	if (rc)
+		GOTO(out, rc);
+
+	next = cathandle->u.chd.chd_next_log;
+	if (next) {
+		if (!llog_exist(next)) {
+			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);
+
+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);
+
 /* 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
  * - the log is not empty, just write out the log header
@@ -316,54 +466,59 @@ 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) {
+			CERROR("%s: cannot find handle for llog "DOSTID": %d\n",
+			       cathandle->lgh_ctxt->loc_obd->obd_name,
+			       POSTID(&lgl->lgl_oi), rc);
+			failed++;
+			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;
+			rc = llog_cat_cleanup(env, cathandle, loghandle,
+					      index);
+		} else if (lrc == -ENOENT) {
+			if (rc == 0) /* ENOENT shouldn't rewrite any error */
+				rc = lrc;
+		} else if (lrc < 0) {
+			failed++;
+			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_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_logid_rec *lir = (struct llog_logid_rec *)rec;
         struct llog_handle *llh;
+	struct llog_log_hdr *hdr;
         int rc;
 
         ENTRY;
@@ -371,25 +526,75 @@ int llog_cat_process_cb(struct llog_handle *cat_llh, struct llog_rec_hdr *rec,
                 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);
-        }
+	CDEBUG(D_HA, "processing log "DOSTID":%x at index %u of catalog "
+	       DOSTID"\n", POSTID(&lir->lid_id.lgl_oi), lir->lid_id.lgl_ogen,
+	       rec->lrh_index, POSTID(&cat_llh->lgh_id.lgl_oi));
+
+	rc = llog_cat_id2handle(env, cat_llh, &llh, &lir->lid_id);
+	if (rc) {
+		CERROR("%s: cannot find handle for llog "DOSTID": %d\n",
+		       cat_llh->lgh_ctxt->loc_obd->obd_name,
+		       POSTID(&lir->lid_id.lgl_oi), rc);
+		if (rc == -ENOENT || rc == -ESTALE) {
+			/* 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. */
+			rc = llog_cat_cleanup(env, cat_llh, NULL,
+					      rec->lrh_index);
+		}
+
+		RETURN(rc);
+	}
+
+	/* clean old empty llogs, do not consider current llog in use */
+	/* ignore remote (lgh_obj=NULL) llogs */
+	hdr = llh->lgh_hdr;
+	if ((hdr->llh_flags & LLOG_F_ZAP_WHEN_EMPTY) &&
+	    hdr->llh_count == 1 && cat_llh->lgh_obj != NULL &&
+	    llh != cat_llh->u.chd.chd_current_log) {
+		rc = llog_destroy(env, llh);
+		if (rc)
+			CERROR("%s: fail to destroy empty log: rc = %d\n",
+			       llh->lgh_ctxt->loc_obd->obd_name, rc);
+		GOTO(out, rc = LLOG_DEL_PLAIN);
+	}
+
+	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);
+	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 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 +602,106 @@ 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;
+        d.lpd_startcat = startcat;
+        d.lpd_startidx = startidx;
 
         if (llh->llh_cat_idx > cat_llh->lgh_last_idx) {
-                CWARN("catlog "LPX64" crosses index zero\n",
-                      cat_llh->lgh_id.lgl_oid);
+                struct llog_process_cat_data cd;
+
+                CWARN("catlog "DOSTID" crosses index zero\n",
+                      POSTID(&cat_llh->lgh_id.lgl_oi));
 
                 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);
+		rc = llog_process_or_fork(env, cat_llh, llog_cat_process_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, llog_cat_process_cb,
+					  &d, &cd, fork);
         } else {
-                rc = llog_process(cat_llh, llog_cat_process_cb, &d, NULL);
+		rc = llog_process_or_fork(env, cat_llh, llog_cat_process_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;
-
-        cfs_daemonize_ctxt("ll_log_process");
-
-        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);
-        }
-
-        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");
-        }
-
-        /*
-         * 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 llog_cat_process_or_fork(env, cat_llh, cb, data, startcat,
+					startidx, false);
 }
-EXPORT_SYMBOL(llog_cat_process_thread);
-#endif
+EXPORT_SYMBOL(llog_cat_process);
 
-static int llog_cat_reverse_process_cb(struct llog_handle *cat_llh,
-                                       struct llog_rec_hdr *rec, void *data)
+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_logid_rec *lir = (struct llog_logid_rec *)rec;
-        struct llog_handle *llh;
-        int rc;
+	struct llog_process_data *d = data;
+	struct llog_logid_rec *lir = (struct llog_logid_rec *)rec;
+	struct llog_handle *llh;
+	struct llog_log_hdr *hdr;
+	int rc;
+
+	if (le32_to_cpu(rec->lrh_type) != LLOG_LOGID_MAGIC) {
+		CERROR("invalid record in catalog\n");
+		RETURN(-EINVAL);
+	}
+	CDEBUG(D_HA, "processing log "DOSTID":%x at index %u of catalog "
+	       DOSTID"\n", POSTID(&lir->lid_id.lgl_oi), lir->lid_id.lgl_ogen,
+	       le32_to_cpu(rec->lrh_index), POSTID(&cat_llh->lgh_id.lgl_oi));
+
+	rc = llog_cat_id2handle(env, cat_llh, &llh, &lir->lid_id);
+	if (rc) {
+		CERROR("%s: cannot find handle for llog "DOSTID": %d\n",
+		       cat_llh->lgh_ctxt->loc_obd->obd_name,
+		       POSTID(&lir->lid_id.lgl_oi), rc);
+		if (rc == -ENOENT || rc == -ESTALE) {
+			/* 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. */
+			rc = llog_cat_cleanup(env, cat_llh, NULL,
+					      rec->lrh_index);
+		}
+
+		RETURN(rc);
+	}
+
+	/* clean old empty llogs, do not consider current llog in use */
+	hdr = llh->lgh_hdr;
+	if ((hdr->llh_flags & LLOG_F_ZAP_WHEN_EMPTY) &&
+	    hdr->llh_count == 1 &&
+	    llh != cat_llh->u.chd.chd_current_log) {
+		rc = llog_destroy(env, llh);
+		if (rc)
+			CERROR("%s: fail to destroy empty log: rc = %d\n",
+			       llh->lgh_ctxt->loc_obd->obd_name, rc);
+		GOTO(out, rc = LLOG_DEL_PLAIN);
+	}
+
+	rc = llog_reverse_process(env, llh, d->lpd_cb, d->lpd_data, NULL);
 
-        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);
-        }
+out:
+	/* 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);
 
-        rc = llog_reverse_process(llh, d->lpd_cb, d->lpd_data, NULL);
-        RETURN(rc);
+	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;
@@ -509,31 +713,34 @@ 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);
+	if (llh->llh_cat_idx > cat_llh->lgh_last_idx) {
+		CWARN("catalog "DOSTID" crosses index zero\n",
+		      POSTID(&cat_llh->lgh_id.lgl_oi));
 
                 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);
+		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 index)
 {
         struct llog_log_hdr *llh = cathandle->lgh_hdr;
         int i, bitmap_size, idx;
@@ -558,9 +765,57 @@ int llog_cat_set_first_idx(struct llog_handle *cathandle, int index)
                         }
                 }
 out:
-                CDEBUG(D_RPCTRACE, "set catlog "LPX64" first idx %u\n",
-                       cathandle->lgh_id.lgl_oid, llh->llh_cat_idx);
-        }
+		CDEBUG(D_RPCTRACE, "set catlog "DOSTID" first idx %u\n",
+		       POSTID(&cathandle->lgh_id.lgl_oi), llh->llh_cat_idx);
+	}
+
+	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 "DOSTID"\n",
+		       index, POSTID(&cathandle->lgh_id.lgl_oi));
+	return rc;
+}
+
+/* helper to initialize catalog llog and process it to cancel */
+int llog_cat_init_and_process(const struct lu_env *env,
+			      struct llog_handle *llh)
+{
+	int rc;
+
+	rc = llog_init_handle(env, llh, LLOG_F_IS_CAT, NULL);
+	if (rc)
+		RETURN(rc);
 
-        RETURN(0);
+	RETURN(0);
 }
+EXPORT_SYMBOL(llog_cat_init_and_process);