X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Fobdclass%2Fllog.c;h=0fdd8feb346cda7632d0f0c48fd7b817c7e08a64;hb=f623f68ae3cbfc6e2d4951aefb3047f72e669564;hp=a69023305e054b5c8ef792dd1cfa9d63efe934d7;hpb=c3d20d0a57e1eebd021ec34323b1223a20e7df16;p=fs%2Flustre-release.git

diff --git a/lustre/obdclass/llog.c b/lustre/obdclass/llog.c
index a690233..0fdd8fe 100644
--- a/lustre/obdclass/llog.c
+++ b/lustre/obdclass/llog.c
@@ -27,7 +27,7 @@
  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
  *
- * Copyright (c) 2012, 2013, Intel Corporation.
+ * Copyright (c) 2012, 2015, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
@@ -47,19 +47,16 @@
 
 #define DEBUG_SUBSYSTEM S_LOG
 
-#ifndef __KERNEL__
-#include <liblustre.h>
-#endif
-
-#include <obd_class.h>
+#include <linux/kthread.h>
+#include <llog_swab.h>
 #include <lustre_log.h>
+#include <obd_class.h>
 #include "llog_internal.h"
-
 /*
  * Allocate a new log or catalog handle
  * Used inside llog_open().
  */
-struct llog_handle *llog_alloc_handle(void)
+static struct llog_handle *llog_alloc_handle(void)
 {
 	struct llog_handle *loghandle;
 
@@ -68,9 +65,9 @@ struct llog_handle *llog_alloc_handle(void)
 		return NULL;
 
 	init_rwsem(&loghandle->lgh_lock);
-	spin_lock_init(&loghandle->lgh_hdr_lock);
-	CFS_INIT_LIST_HEAD(&loghandle->u.phd.phd_entry);
-	cfs_atomic_set(&loghandle->lgh_refcount, 1);
+	mutex_init(&loghandle->lgh_hdr_mutex);
+	INIT_LIST_HEAD(&loghandle->u.phd.phd_entry);
+	atomic_set(&loghandle->lgh_refcount, 1);
 
 	return loghandle;
 }
@@ -78,103 +75,231 @@ struct llog_handle *llog_alloc_handle(void)
 /*
  * Free llog handle and header data if exists. Used in llog_close() only
  */
-void llog_free_handle(struct llog_handle *loghandle)
+static void llog_free_handle(struct llog_handle *loghandle)
 {
 	LASSERT(loghandle != NULL);
 
 	/* failed llog_init_handle */
-	if (!loghandle->lgh_hdr)
+	if (loghandle->lgh_hdr == NULL)
 		goto out;
 
 	if (loghandle->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN)
-		LASSERT(cfs_list_empty(&loghandle->u.phd.phd_entry));
+		LASSERT(list_empty(&loghandle->u.phd.phd_entry));
 	else if (loghandle->lgh_hdr->llh_flags & LLOG_F_IS_CAT)
-		LASSERT(cfs_list_empty(&loghandle->u.chd.chd_head));
-	LASSERT(sizeof(*(loghandle->lgh_hdr)) == LLOG_CHUNK_SIZE);
-	OBD_FREE(loghandle->lgh_hdr, LLOG_CHUNK_SIZE);
+		LASSERT(list_empty(&loghandle->u.chd.chd_head));
+	OBD_FREE_LARGE(loghandle->lgh_hdr, loghandle->lgh_hdr_size);
 out:
 	OBD_FREE_PTR(loghandle);
 }
 
 void llog_handle_get(struct llog_handle *loghandle)
 {
-	cfs_atomic_inc(&loghandle->lgh_refcount);
+	atomic_inc(&loghandle->lgh_refcount);
 }
 
 void llog_handle_put(struct llog_handle *loghandle)
 {
-	LASSERT(cfs_atomic_read(&loghandle->lgh_refcount) > 0);
-	if (cfs_atomic_dec_and_test(&loghandle->lgh_refcount))
+	LASSERT(atomic_read(&loghandle->lgh_refcount) > 0);
+	if (atomic_dec_and_test(&loghandle->lgh_refcount))
 		llog_free_handle(loghandle);
 }
 
+static int llog_declare_destroy(const struct lu_env *env,
+				struct llog_handle *handle,
+				struct thandle *th)
+{
+	struct llog_operations *lop;
+	int rc;
+
+	ENTRY;
+
+	rc = llog_handle2ops(handle, &lop);
+	if (rc)
+		RETURN(rc);
+	if (lop->lop_declare_destroy == NULL)
+		RETURN(-EOPNOTSUPP);
+
+	rc = lop->lop_declare_destroy(env, handle, th);
+
+	RETURN(rc);
+}
+
+int llog_trans_destroy(const struct lu_env *env, struct llog_handle *handle,
+		       struct thandle *th)
+{
+	struct llog_operations	*lop;
+	int rc;
+	ENTRY;
+
+	rc = llog_handle2ops(handle, &lop);
+	if (rc < 0)
+		RETURN(rc);
+	if (lop->lop_destroy == NULL)
+		RETURN(-EOPNOTSUPP);
+
+	LASSERT(handle->lgh_obj != NULL);
+	if (!dt_object_exists(handle->lgh_obj))
+		RETURN(0);
+
+	rc = lop->lop_destroy(env, handle, th);
+
+	RETURN(rc);
+}
+
+int llog_destroy(const struct lu_env *env, struct llog_handle *handle)
+{
+	struct llog_operations	*lop;
+	struct dt_device	*dt;
+	struct thandle		*th;
+	int rc;
+
+	ENTRY;
+
+	rc = llog_handle2ops(handle, &lop);
+	if (rc < 0)
+		RETURN(rc);
+	if (lop->lop_destroy == NULL)
+		RETURN(-EOPNOTSUPP);
+
+	if (handle->lgh_obj == NULL) {
+		/* if lgh_obj == NULL, then it is from client side destroy */
+		rc = lop->lop_destroy(env, handle, NULL);
+		RETURN(rc);
+	}
+
+	if (!dt_object_exists(handle->lgh_obj))
+		RETURN(0);
+
+	dt = lu2dt_dev(handle->lgh_obj->do_lu.lo_dev);
+
+	th = dt_trans_create(env, dt);
+	if (IS_ERR(th))
+		RETURN(PTR_ERR(th));
+
+	rc = llog_declare_destroy(env, handle, th);
+	if (rc != 0)
+		GOTO(out_trans, rc);
+
+	rc = dt_trans_start_local(env, dt, th);
+	if (rc < 0)
+		GOTO(out_trans, rc);
+
+	rc = lop->lop_destroy(env, handle, th);
+
+out_trans:
+	dt_trans_stop(env, dt, th);
+
+	RETURN(rc);
+}
+EXPORT_SYMBOL(llog_destroy);
+
 /* returns negative on error; 0 if success; 1 if success & log destroyed */
 int llog_cancel_rec(const struct lu_env *env, struct llog_handle *loghandle,
 		    int index)
 {
-        struct llog_log_hdr *llh = loghandle->lgh_hdr;
-        int rc = 0;
-        ENTRY;
+	struct llog_thread_info *lgi = llog_info(env);
+	struct dt_device	*dt;
+	struct llog_log_hdr	*llh = loghandle->lgh_hdr;
+	struct thandle		*th;
+	int			 rc;
+	int rc1;
+	bool subtract_count = false;
 
-        CDEBUG(D_RPCTRACE, "Canceling %d in log "DOSTID"\n",
-               index, POSTID(&loghandle->lgh_id.lgl_oi));
+	ENTRY;
 
-        if (index == 0) {
-                CERROR("Can't cancel index 0 which is header\n");
-                RETURN(-EINVAL);
-        }
+	CDEBUG(D_RPCTRACE, "Canceling %d in log "DOSTID"\n", index,
+	       POSTID(&loghandle->lgh_id.lgl_oi));
+
+	if (index == 0) {
+		CERROR("Can't cancel index 0 which is header\n");
+		RETURN(-EINVAL);
+	}
+
+	LASSERT(loghandle != NULL);
+	LASSERT(loghandle->lgh_ctxt != NULL);
+	LASSERT(loghandle->lgh_obj != NULL);
+
+	dt = lu2dt_dev(loghandle->lgh_obj->do_lu.lo_dev);
+
+	th = dt_trans_create(env, dt);
+	if (IS_ERR(th))
+		RETURN(PTR_ERR(th));
 
-	spin_lock(&loghandle->lgh_hdr_lock);
-	if (!ext2_clear_bit(index, llh->llh_bitmap)) {
-		spin_unlock(&loghandle->lgh_hdr_lock);
+	rc = llog_declare_write_rec(env, loghandle, &llh->llh_hdr, index, th);
+	if (rc < 0)
+		GOTO(out_trans, rc);
+
+	if ((llh->llh_flags & LLOG_F_ZAP_WHEN_EMPTY))
+		rc = llog_declare_destroy(env, loghandle, th);
+
+	th->th_wait_submit = 1;
+	rc = dt_trans_start_local(env, dt, th);
+	if (rc < 0)
+		GOTO(out_trans, rc);
+
+	down_write(&loghandle->lgh_lock);
+	/* clear bitmap */
+	mutex_lock(&loghandle->lgh_hdr_mutex);
+	if (!ext2_clear_bit(index, LLOG_HDR_BITMAP(llh))) {
 		CDEBUG(D_RPCTRACE, "Catalog index %u already clear?\n", index);
-		RETURN(-ENOENT);
+		GOTO(out_unlock, rc);
 	}
 
-	llh->llh_count--;
+	loghandle->lgh_hdr->llh_count--;
+	subtract_count = true;
+	/* Pass this index to llog_osd_write_rec(), which will use the index
+	 * to only update the necesary bitmap. */
+	lgi->lgi_cookie.lgc_index = index;
+	/* update header */
+	rc = llog_write_rec(env, loghandle, &llh->llh_hdr, &lgi->lgi_cookie,
+			    LLOG_HEADER_IDX, th);
+	if (rc != 0)
+		GOTO(out_unlock, rc);
 
 	if ((llh->llh_flags & LLOG_F_ZAP_WHEN_EMPTY) &&
 	    (llh->llh_count == 1) &&
-	    (loghandle->lgh_last_idx == (LLOG_BITMAP_BYTES * 8) - 1)) {
-		spin_unlock(&loghandle->lgh_hdr_lock);
-		rc = llog_destroy(env, loghandle);
+	    ((loghandle->lgh_last_idx == LLOG_HDR_BITMAP_SIZE(llh) - 1) ||
+	     (loghandle->u.phd.phd_cat_handle != NULL &&
+	      loghandle->u.phd.phd_cat_handle->u.chd.chd_current_log !=
+		loghandle))) {
+		rc = llog_trans_destroy(env, loghandle, th);
 		if (rc < 0) {
+			/* Sigh, can not destroy the final plain llog, but
+			 * the bitmap has been clearly, so the record can not
+			 * be accessed anymore, let's return 0 for now, and
+			 * the orphan will be handled by LFSCK. */
 			CERROR("%s: can't destroy empty llog #"DOSTID
 			       "#%08x: rc = %d\n",
 			       loghandle->lgh_ctxt->loc_obd->obd_name,
 			       POSTID(&loghandle->lgh_id.lgl_oi),
 			       loghandle->lgh_id.lgl_ogen, rc);
-			GOTO(out_err, rc);
+			GOTO(out_unlock, rc);
 		}
-		RETURN(1);
+		rc = LLOG_DEL_PLAIN;
 	}
-	spin_unlock(&loghandle->lgh_hdr_lock);
 
-	rc = llog_write(env, loghandle, &llh->llh_hdr, NULL, 0, NULL, 0);
-	if (rc < 0) {
-		CERROR("%s: fail to write header for llog #"DOSTID
-		       "#%08x: rc = %d\n",
-		       loghandle->lgh_ctxt->loc_obd->obd_name,
-		       POSTID(&loghandle->lgh_id.lgl_oi),
-		       loghandle->lgh_id.lgl_ogen, rc);
-		GOTO(out_err, rc);
+out_unlock:
+	mutex_unlock(&loghandle->lgh_hdr_mutex);
+	up_write(&loghandle->lgh_lock);
+out_trans:
+	rc1 = dt_trans_stop(env, dt, th);
+	if (rc == 0)
+		rc = rc1;
+	if (rc < 0 && subtract_count) {
+		mutex_lock(&loghandle->lgh_hdr_mutex);
+		loghandle->lgh_hdr->llh_count++;
+		ext2_set_bit(index, LLOG_HDR_BITMAP(llh));
+		mutex_unlock(&loghandle->lgh_hdr_mutex);
 	}
-	RETURN(0);
-out_err:
-	spin_lock(&loghandle->lgh_hdr_lock);
-	ext2_set_bit(index, llh->llh_bitmap);
-	llh->llh_count++;
-	spin_unlock(&loghandle->lgh_hdr_lock);
-	return rc;
+	RETURN(rc);
 }
-EXPORT_SYMBOL(llog_cancel_rec);
 
-static int llog_read_header(const struct lu_env *env,
-			    struct llog_handle *handle,
-			    struct obd_uuid *uuid)
+int llog_read_header(const struct lu_env *env, struct llog_handle *handle,
+		     const struct obd_uuid *uuid)
 {
 	struct llog_operations *lop;
 	int rc;
+	ENTRY;
 
 	rc = llog_handle2ops(handle, &lop);
 	if (rc)
@@ -187,35 +312,52 @@ static int llog_read_header(const struct lu_env *env,
 	if (rc == LLOG_EEMPTY) {
 		struct llog_log_hdr *llh = handle->lgh_hdr;
 
+		/* lrh_len should be initialized in llog_init_handle */
 		handle->lgh_last_idx = 0; /* header is record with index 0 */
+		handle->lgh_write_offset = 0;
 		llh->llh_count = 1;         /* for the header record */
 		llh->llh_hdr.lrh_type = LLOG_HDR_MAGIC;
-		llh->llh_hdr.lrh_len = llh->llh_tail.lrt_len = LLOG_CHUNK_SIZE;
-		llh->llh_hdr.lrh_index = llh->llh_tail.lrt_index = 0;
+		LASSERT(handle->lgh_ctxt->loc_chunk_size >=
+						LLOG_MIN_CHUNK_SIZE);
+		llh->llh_hdr.lrh_len = handle->lgh_ctxt->loc_chunk_size;
+		llh->llh_hdr.lrh_index = 0;
 		llh->llh_timestamp = cfs_time_current_sec();
 		if (uuid)
 			memcpy(&llh->llh_tgtuuid, uuid,
 			       sizeof(llh->llh_tgtuuid));
 		llh->llh_bitmap_offset = offsetof(typeof(*llh), llh_bitmap);
-		ext2_set_bit(0, llh->llh_bitmap);
+		/* Since update llog header might also call this function,
+		 * let's reset the bitmap to 0 here */
+		memset(LLOG_HDR_BITMAP(llh), 0, llh->llh_hdr.lrh_len -
+						llh->llh_bitmap_offset -
+						sizeof(llh->llh_tail));
+		ext2_set_bit(0, LLOG_HDR_BITMAP(llh));
+		LLOG_HDR_TAIL(llh)->lrt_len = llh->llh_hdr.lrh_len;
+		LLOG_HDR_TAIL(llh)->lrt_index = llh->llh_hdr.lrh_index;
 		rc = 0;
 	}
-	return rc;
+	RETURN(rc);
 }
+EXPORT_SYMBOL(llog_read_header);
 
 int llog_init_handle(const struct lu_env *env, struct llog_handle *handle,
 		     int flags, struct obd_uuid *uuid)
 {
 	struct llog_log_hdr	*llh;
+	enum llog_flag		 fmt = flags & LLOG_F_EXT_MASK;
 	int			 rc;
-
+	int			chunk_size = handle->lgh_ctxt->loc_chunk_size;
 	ENTRY;
+
 	LASSERT(handle->lgh_hdr == NULL);
 
-	OBD_ALLOC_PTR(llh);
+	LASSERT(chunk_size >= LLOG_MIN_CHUNK_SIZE);
+	OBD_ALLOC_LARGE(llh, chunk_size);
 	if (llh == NULL)
 		RETURN(-ENOMEM);
+
 	handle->lgh_hdr = llh;
+	handle->lgh_hdr_size = chunk_size;
 	/* first assign flags to use llog_client_ops */
 	llh->llh_flags = flags;
 	rc = llog_read_header(env, handle, uuid);
@@ -252,18 +394,20 @@ int llog_init_handle(const struct lu_env *env, struct llog_handle *handle,
 		}
 	}
 	if (flags & LLOG_F_IS_CAT) {
-		LASSERT(cfs_list_empty(&handle->u.chd.chd_head));
-		CFS_INIT_LIST_HEAD(&handle->u.chd.chd_head);
+		LASSERT(list_empty(&handle->u.chd.chd_head));
+		INIT_LIST_HEAD(&handle->u.chd.chd_head);
 		llh->llh_size = sizeof(struct llog_logid_rec);
+		llh->llh_flags |= LLOG_F_IS_FIXSIZE;
 	} else if (!(flags & LLOG_F_IS_PLAIN)) {
 		CERROR("%s: unknown flags: %#x (expected %#x or %#x)\n",
 		       handle->lgh_ctxt->loc_obd->obd_name,
 		       flags, LLOG_F_IS_CAT, LLOG_F_IS_PLAIN);
 		rc = -EINVAL;
 	}
+	llh->llh_flags |= fmt;
 out:
 	if (rc) {
-		OBD_FREE_PTR(llh);
+		OBD_FREE_LARGE(llh, chunk_size);
 		handle->lgh_hdr = NULL;
 	}
 	RETURN(rc);
@@ -277,135 +421,197 @@ static int llog_process_thread(void *arg)
 	struct llog_log_hdr		*llh = loghandle->lgh_hdr;
 	struct llog_process_cat_data	*cd  = lpi->lpi_catdata;
 	char				*buf;
-	__u64				 cur_offset = LLOG_CHUNK_SIZE;
-	__u64				 last_offset;
+	size_t				 chunk_size;
+	__u64				 cur_offset;
 	int				 rc = 0, index = 1, last_index;
 	int				 saved_index = 0;
 	int				 last_called_index = 0;
 
 	ENTRY;
 
-        LASSERT(llh);
+	if (llh == NULL)
+		RETURN(-EINVAL);
 
-        OBD_ALLOC(buf, LLOG_CHUNK_SIZE);
-        if (!buf) {
-                lpi->lpi_rc = -ENOMEM;
+	cur_offset = chunk_size = llh->llh_hdr.lrh_len;
+	/* expect chunk_size to be power of two */
+	LASSERT(is_power_of_2(chunk_size));
+
+	OBD_ALLOC_LARGE(buf, chunk_size);
+	if (buf == NULL) {
+		lpi->lpi_rc = -ENOMEM;
 		RETURN(0);
-        }
+	}
 
-        if (cd != NULL) {
-                last_called_index = cd->lpcd_first_idx;
-                index = cd->lpcd_first_idx + 1;
-        }
-        if (cd != NULL && cd->lpcd_last_idx)
-                last_index = cd->lpcd_last_idx;
-        else
-                last_index = LLOG_BITMAP_BYTES * 8 - 1;
-
-        while (rc == 0) {
-                struct llog_rec_hdr *rec;
-
-                /* skip records not set in bitmap */
-                while (index <= last_index &&
-                       !ext2_test_bit(index, llh->llh_bitmap))
-                        ++index;
-
-                LASSERT(index <= last_index + 1);
-                if (index == last_index + 1)
-                        break;
-repeat:
-                CDEBUG(D_OTHER, "index: %d last_index %d\n",
-                       index, last_index);
+	if (cd != NULL) {
+		last_called_index = cd->lpcd_first_idx;
+		index = cd->lpcd_first_idx + 1;
+	}
+	if (cd != NULL && cd->lpcd_last_idx)
+		last_index = cd->lpcd_last_idx;
+	else
+		last_index = LLOG_HDR_BITMAP_SIZE(llh) - 1;
+
+	while (rc == 0) {
+		struct llog_rec_hdr *rec;
+		off_t chunk_offset;
+		unsigned int buf_offset = 0;
+		bool partial_chunk;
+
+		/* skip records not set in bitmap */
+		while (index <= last_index &&
+		       !ext2_test_bit(index, LLOG_HDR_BITMAP(llh)))
+			++index;
+
+		/* There are no indices prior the last_index */
+		if (index > last_index)
+			break;
 
-                /* get the buf with our target record; avoid old garbage */
-                memset(buf, 0, LLOG_CHUNK_SIZE);
-                last_offset = cur_offset;
+		CDEBUG(D_OTHER, "index: %d last_index %d\n", index,
+		       last_index);
+
+repeat:
+		/* get the buf with our target record; avoid old garbage */
+		memset(buf, 0, chunk_size);
 		rc = llog_next_block(lpi->lpi_env, loghandle, &saved_index,
-				     index, &cur_offset, buf, LLOG_CHUNK_SIZE);
-                if (rc)
-                        GOTO(out, rc);
+				     index, &cur_offset, buf, chunk_size);
+		if (rc != 0)
+			GOTO(out, rc);
+
+		/* NB: after llog_next_block() call the cur_offset is the
+		 * offset of the next block after read one.
+		 * The absolute offset of the current chunk is calculated
+		 * from cur_offset value and stored in chunk_offset variable.
+		 */
+		if (cur_offset % chunk_size != 0) {
+			partial_chunk = true;
+			chunk_offset = cur_offset & ~(chunk_size - 1);
+		} else {
+			partial_chunk = false;
+			chunk_offset = cur_offset - chunk_size;
+		}
 
-                /* NB: when rec->lrh_len is accessed it is already swabbed
-                 * since it is used at the "end" of the loop and the rec
-                 * swabbing is done at the beginning of the loop. */
-                for (rec = (struct llog_rec_hdr *)buf;
-                     (char *)rec < buf + LLOG_CHUNK_SIZE;
-                     rec = (struct llog_rec_hdr *)((char *)rec + rec->lrh_len)){
+		/* NB: when rec->lrh_len is accessed it is already swabbed
+		 * since it is used at the "end" of the loop and the rec
+		 * swabbing is done at the beginning of the loop. */
+		for (rec = (struct llog_rec_hdr *)(buf + buf_offset);
+		     (char *)rec < buf + chunk_size;
+		     rec = llog_rec_hdr_next(rec)) {
 
-                        CDEBUG(D_OTHER, "processing rec 0x%p type %#x\n",
-                               rec, rec->lrh_type);
+			CDEBUG(D_OTHER, "processing rec 0x%p type %#x\n",
+			       rec, rec->lrh_type);
 
-                        if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
+			if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
 				lustre_swab_llog_rec(rec);
 
-                        CDEBUG(D_OTHER, "after swabbing, type=%#x idx=%d\n",
-                               rec->lrh_type, rec->lrh_index);
+			CDEBUG(D_OTHER, "after swabbing, type=%#x idx=%d\n",
+			       rec->lrh_type, rec->lrh_index);
+
+			/* for partial chunk the end of it is zeroed, check
+			 * for index 0 to distinguish it. */
+			if (partial_chunk && rec->lrh_index == 0) {
+				/* concurrent llog_add() might add new records
+				 * while llog_processing, check this is not
+				 * the case and re-read the current chunk
+				 * otherwise. */
+				if (index > loghandle->lgh_last_idx)
+					GOTO(out, rc = 0);
+				CDEBUG(D_OTHER, "Re-read last llog buffer for "
+				       "new records, index %u, last %u\n",
+				       index, loghandle->lgh_last_idx);
+				/* save offset inside buffer for the re-read */
+				buf_offset = (char *)rec - (char *)buf;
+				cur_offset = chunk_offset;
+				goto repeat;
+			}
 
-			if (rec->lrh_index == 0) {
-				/* probably another rec just got added? */
-				if (index <= loghandle->lgh_last_idx)
-					GOTO(repeat, rc = 0);
-				GOTO(out, rc = 0); /* no more records */
+			if (rec->lrh_len == 0 || rec->lrh_len > chunk_size) {
+				CWARN("invalid length %d in llog record for "
+				      "index %d/%d\n", rec->lrh_len,
+				      rec->lrh_index, index);
+				GOTO(out, rc = -EINVAL);
 			}
-			if (rec->lrh_len == 0 ||
-			    rec->lrh_len > LLOG_CHUNK_SIZE) {
-                                CWARN("invalid length %d in llog record for "
-                                      "index %d/%d\n", rec->lrh_len,
-                                      rec->lrh_index, index);
-                                GOTO(out, rc = -EINVAL);
-                        }
 
-                        if (rec->lrh_index < index) {
-                                CDEBUG(D_OTHER, "skipping lrh_index %d\n",
-                                       rec->lrh_index);
-                                continue;
-                        }
+			if (rec->lrh_index < index) {
+				CDEBUG(D_OTHER, "skipping lrh_index %d\n",
+				       rec->lrh_index);
+				continue;
+			}
 
-                        CDEBUG(D_OTHER,
-                               "lrh_index: %d lrh_len: %d (%d remains)\n",
-                               rec->lrh_index, rec->lrh_len,
-                               (int)(buf + LLOG_CHUNK_SIZE - (char *)rec));
+			if (rec->lrh_index != index) {
+				CERROR("%s: Invalid record: index %u but "
+				       "expected %u\n",
+				       loghandle->lgh_ctxt->loc_obd->obd_name,
+				       rec->lrh_index, index);
+				GOTO(out, rc = -ERANGE);
+			}
 
-                        loghandle->lgh_cur_idx = rec->lrh_index;
-                        loghandle->lgh_cur_offset = (char *)rec - (char *)buf +
-                                                    last_offset;
+			CDEBUG(D_OTHER,
+			       "lrh_index: %d lrh_len: %d (%d remains)\n",
+			       rec->lrh_index, rec->lrh_len,
+			       (int)(buf + chunk_size - (char *)rec));
 
-                        /* if set, process the callback on this record */
-                        if (ext2_test_bit(index, llh->llh_bitmap)) {
+			loghandle->lgh_cur_idx = rec->lrh_index;
+			loghandle->lgh_cur_offset = (char *)rec - (char *)buf +
+						    chunk_offset;
+
+			/* if set, process the callback on this record */
+			if (ext2_test_bit(index, LLOG_HDR_BITMAP(llh))) {
 				rc = lpi->lpi_cb(lpi->lpi_env, loghandle, rec,
 						 lpi->lpi_cbdata);
 				last_called_index = index;
 				if (rc == LLOG_PROC_BREAK) {
 					GOTO(out, rc);
 				} else if (rc == LLOG_DEL_RECORD) {
-					llog_cancel_rec(lpi->lpi_env,
-							loghandle,
-							rec->lrh_index);
-                                        rc = 0;
-                                }
-                                if (rc)
-                                        GOTO(out, rc);
-                        } else {
-                                CDEBUG(D_OTHER, "Skipped index %d\n", index);
-                        }
-
-                        /* next record, still in buffer? */
-                        ++index;
-                        if (index > last_index)
-                                GOTO(out, rc = 0);
-                }
-        }
+					rc = llog_cancel_rec(lpi->lpi_env,
+							     loghandle,
+							     rec->lrh_index);
+				}
+				if (rc)
+					GOTO(out, rc);
+			}
+			/* exit if the last index is reached */
+			if (index >= last_index)
+				GOTO(out, rc = 0);
+			++index;
+		}
+	}
 
 out:
-        if (cd != NULL)
-                cd->lpcd_last_idx = last_called_index;
+	if (cd != NULL)
+		cd->lpcd_last_idx = last_called_index;
+
+	if (unlikely(rc == -EIO && loghandle->lgh_obj != NULL)) {
+		if (dt_object_remote(loghandle->lgh_obj)) {
+			/* If it is remote object, then -EIO might means
+			 * disconnection or eviction, let's return -EAGAIN,
+			 * so for update recovery log processing, it will
+			 * retry until the umount or abort recovery, see
+			 * lod_sub_recovery_thread() */
+			CERROR("%s retry remote llog process\n",
+			       loghandle->lgh_ctxt->loc_obd->obd_name);
+			rc = -EAGAIN;
+		} else {
+			/* something bad happened to the processing of a local
+			 * llog file, probably I/O error or the log got
+			 * corrupted to be able to finally release the log we
+			 * discard any remaining bits in the header */
+			CERROR("Local llog found corrupted\n");
+			while (index <= last_index) {
+				if (ext2_test_bit(index,
+						  LLOG_HDR_BITMAP(llh)) != 0)
+					llog_cancel_rec(lpi->lpi_env, loghandle,
+							index);
+				index++;
+			}
+			rc = 0;
+		}
+	}
 
-	OBD_FREE(buf, LLOG_CHUNK_SIZE);
+	OBD_FREE_LARGE(buf, chunk_size);
         lpi->lpi_rc = rc;
         return 0;
 }
 
-#ifdef __KERNEL__
 static int llog_process_thread_daemonize(void *arg)
 {
 	struct llog_process_info	*lpi = arg;
@@ -427,7 +633,6 @@ out:
 	complete(&lpi->lpi_completion);
 	return rc;
 }
-#endif
 
 int llog_process_or_fork(const struct lu_env *env,
 			 struct llog_handle *loghandle,
@@ -448,39 +653,40 @@ int llog_process_or_fork(const struct lu_env *env,
         lpi->lpi_cbdata    = data;
         lpi->lpi_catdata   = catdata;
 
-#ifdef __KERNEL__
 	if (fork) {
+		struct task_struct *task;
+
 		/* The new thread can't use parent env,
 		 * init the new one in llog_process_thread_daemonize. */
 		lpi->lpi_env = NULL;
 		init_completion(&lpi->lpi_completion);
-		rc = PTR_ERR(kthread_run(llog_process_thread_daemonize, lpi,
-					     "llog_process_thread"));
-		if (IS_ERR_VALUE(rc)) {
+		task = kthread_run(llog_process_thread_daemonize, lpi,
+				   "llog_process_thread");
+		if (IS_ERR(task)) {
+			rc = PTR_ERR(task);
 			CERROR("%s: cannot start thread: rc = %d\n",
 			       loghandle->lgh_ctxt->loc_obd->obd_name, rc);
-			OBD_FREE_PTR(lpi);
-			RETURN(rc);
+			GOTO(out_lpi, rc);
 		}
 		wait_for_completion(&lpi->lpi_completion);
 	} else {
 		lpi->lpi_env = env;
 		llog_process_thread(lpi);
 	}
-#else
-	lpi->lpi_env = env;
-	llog_process_thread(lpi);
-#endif
-        rc = lpi->lpi_rc;
-        OBD_FREE_PTR(lpi);
-        RETURN(rc);
+	rc = lpi->lpi_rc;
+
+out_lpi:
+	OBD_FREE_PTR(lpi);
+	RETURN(rc);
 }
 EXPORT_SYMBOL(llog_process_or_fork);
 
 int llog_process(const struct lu_env *env, struct llog_handle *loghandle,
 		 llog_cb_t cb, void *data, void *catdata)
 {
-	return llog_process_or_fork(env, loghandle, cb, data, catdata, true);
+	int rc;
+	rc = llog_process_or_fork(env, loghandle, cb, data, catdata, true);
+	return rc == LLOG_DEL_PLAIN ? 0 : rc;
 }
 EXPORT_SYMBOL(llog_process);
 
@@ -492,36 +698,36 @@ int llog_reverse_process(const struct lu_env *env,
         struct llog_process_cat_data *cd = catdata;
         void *buf;
         int rc = 0, first_index = 1, index, idx;
+	__u32	chunk_size = llh->llh_hdr.lrh_len;
         ENTRY;
 
-        OBD_ALLOC(buf, LLOG_CHUNK_SIZE);
-        if (!buf)
-                RETURN(-ENOMEM);
+	OBD_ALLOC_LARGE(buf, chunk_size);
+	if (buf == NULL)
+		RETURN(-ENOMEM);
 
-        if (cd != NULL)
-                first_index = cd->lpcd_first_idx + 1;
-        if (cd != NULL && cd->lpcd_last_idx)
-                index = cd->lpcd_last_idx;
-        else
-                index = LLOG_BITMAP_BYTES * 8 - 1;
+	if (cd != NULL)
+		first_index = cd->lpcd_first_idx + 1;
+	if (cd != NULL && cd->lpcd_last_idx)
+		index = cd->lpcd_last_idx;
+	else
+		index = LLOG_HDR_BITMAP_SIZE(llh) - 1;
 
-        while (rc == 0) {
-                struct llog_rec_hdr *rec;
-                struct llog_rec_tail *tail;
+	while (rc == 0) {
+		struct llog_rec_hdr *rec;
+		struct llog_rec_tail *tail;
 
-                /* skip records not set in bitmap */
-                while (index >= first_index &&
-                       !ext2_test_bit(index, llh->llh_bitmap))
-                        --index;
+		/* skip records not set in bitmap */
+		while (index >= first_index &&
+		       !ext2_test_bit(index, LLOG_HDR_BITMAP(llh)))
+			--index;
 
-                LASSERT(index >= first_index - 1);
-                if (index == first_index - 1)
-                        break;
+		LASSERT(index >= first_index - 1);
+		if (index == first_index - 1)
+			break;
 
-                /* get the buf with our target record; avoid old garbage */
-                memset(buf, 0, LLOG_CHUNK_SIZE);
-		rc = llog_prev_block(env, loghandle, index, buf,
-				     LLOG_CHUNK_SIZE);
+		/* get the buf with our target record; avoid old garbage */
+		memset(buf, 0, chunk_size);
+		rc = llog_prev_block(env, loghandle, index, buf, chunk_size);
 		if (rc)
 			GOTO(out, rc);
 
@@ -537,13 +743,13 @@ int llog_reverse_process(const struct lu_env *env,
 		LASSERT(idx == index);
 		tail = (void *)rec + rec->lrh_len - sizeof(*tail);
 
-                /* process records in buffer, starting where we found one */
-                while ((void *)tail > buf) {
+		/* process records in buffer, starting where we found one */
+		while ((void *)tail > buf) {
 			if (tail->lrt_index == 0)
 				GOTO(out, rc = 0); /* no more records */
 
-                        /* if set, process the callback on this record */
-                        if (ext2_test_bit(index, llh->llh_bitmap)) {
+			/* if set, process the callback on this record */
+			if (ext2_test_bit(index, LLOG_HDR_BITMAP(llh))) {
 				rec = (void *)tail - tail->lrt_len +
 				      sizeof(*tail);
 
@@ -551,9 +757,8 @@ int llog_reverse_process(const struct lu_env *env,
 				if (rc == LLOG_PROC_BREAK) {
 					GOTO(out, rc);
 				} else if (rc == LLOG_DEL_RECORD) {
-					llog_cancel_rec(env, loghandle,
-							tail->lrt_index);
-					rc = 0;
+					rc = llog_cancel_rec(env, loghandle,
+							     tail->lrt_index);
 				}
                                 if (rc)
                                         GOTO(out, rc);
@@ -568,8 +773,8 @@ int llog_reverse_process(const struct lu_env *env,
         }
 
 out:
-        if (buf)
-                OBD_FREE(buf, LLOG_CHUNK_SIZE);
+	if (buf != NULL)
+		OBD_FREE_LARGE(buf, chunk_size);
         RETURN(rc);
 }
 EXPORT_SYMBOL(llog_reverse_process);
@@ -628,7 +833,6 @@ int llog_declare_create(const struct lu_env *env,
 		cfs_cap_lower(CFS_CAP_SYS_RESOURCE);
 	RETURN(rc);
 }
-EXPORT_SYMBOL(llog_declare_create);
 
 int llog_create(const struct lu_env *env, struct llog_handle *handle,
 		struct thandle *th)
@@ -652,7 +856,6 @@ int llog_create(const struct lu_env *env, struct llog_handle *handle,
 		cfs_cap_lower(CFS_CAP_SYS_RESOURCE);
 	RETURN(rc);
 }
-EXPORT_SYMBOL(llog_create);
 
 int llog_declare_write_rec(const struct lu_env *env,
 			   struct llog_handle *handle,
@@ -679,11 +882,10 @@ int llog_declare_write_rec(const struct lu_env *env,
 		cfs_cap_lower(CFS_CAP_SYS_RESOURCE);
 	RETURN(rc);
 }
-EXPORT_SYMBOL(llog_declare_write_rec);
 
 int llog_write_rec(const struct lu_env *env, struct llog_handle *handle,
 		   struct llog_rec_hdr *rec, struct llog_cookie *logcookies,
-		   int numcookies, void *buf, int idx, struct thandle *th)
+		   int idx, struct thandle *th)
 {
 	struct llog_operations	*lop;
 	int			 raised, rc, buflen;
@@ -698,27 +900,21 @@ int llog_write_rec(const struct lu_env *env, struct llog_handle *handle,
 	if (lop->lop_write_rec == NULL)
 		RETURN(-EOPNOTSUPP);
 
-	if (buf)
-		buflen = rec->lrh_len + sizeof(struct llog_rec_hdr) +
-			 sizeof(struct llog_rec_tail);
-	else
-		buflen = rec->lrh_len;
+	buflen = rec->lrh_len;
 	LASSERT(cfs_size_round(buflen) == buflen);
 
 	raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE);
 	if (!raised)
 		cfs_cap_raise(CFS_CAP_SYS_RESOURCE);
-	rc = lop->lop_write_rec(env, handle, rec, logcookies, numcookies,
-				buf, idx, th);
+	rc = lop->lop_write_rec(env, handle, rec, logcookies, idx, th);
 	if (!raised)
 		cfs_cap_lower(CFS_CAP_SYS_RESOURCE);
 	RETURN(rc);
 }
-EXPORT_SYMBOL(llog_write_rec);
 
 int llog_add(const struct lu_env *env, struct llog_handle *lgh,
 	     struct llog_rec_hdr *rec, struct llog_cookie *logcookies,
-	     void *buf, struct thandle *th)
+	     struct thandle *th)
 {
 	int raised, rc;
 
@@ -730,7 +926,7 @@ int llog_add(const struct lu_env *env, struct llog_handle *lgh,
 	raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE);
 	if (!raised)
 		cfs_cap_raise(CFS_CAP_SYS_RESOURCE);
-	rc = lgh->lgh_logops->lop_add(env, lgh, rec, logcookies, buf, th);
+	rc = lgh->lgh_logops->lop_add(env, lgh, rec, logcookies, th);
 	if (!raised)
 		cfs_cap_lower(CFS_CAP_SYS_RESOURCE);
 	RETURN(rc);
@@ -786,6 +982,15 @@ int llog_open_create(const struct lu_env *env, struct llog_ctxt *ctxt,
 	if (IS_ERR(th))
 		GOTO(out, rc = PTR_ERR(th));
 
+	/* 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 (ctxt->loc_flags & LLOG_CTXT_FLAG_NORMAL_FID)
+		th->th_sync = 1;
+
+	th->th_wait_submit = 1;
 	rc = llog_declare_create(env, *res, th);
 	if (rc == 0) {
 		rc = dt_trans_start_local(env, d, th);
@@ -836,8 +1041,7 @@ EXPORT_SYMBOL(llog_erase);
  * Valid only with local llog.
  */
 int llog_write(const struct lu_env *env, struct llog_handle *loghandle,
-	       struct llog_rec_hdr *rec, struct llog_cookie *reccookie,
-	       int cookiecount, void *buf, int idx)
+	       struct llog_rec_hdr *rec, int idx)
 {
 	struct dt_device	*dt;
 	struct thandle		*th;
@@ -859,13 +1063,13 @@ int llog_write(const struct lu_env *env, struct llog_handle *loghandle,
 	if (rc)
 		GOTO(out_trans, rc);
 
+	th->th_wait_submit = 1;
 	rc = dt_trans_start_local(env, dt, th);
 	if (rc)
 		GOTO(out_trans, rc);
 
 	down_write(&loghandle->lgh_lock);
-	rc = llog_write_rec(env, loghandle, rec, reccookie,
-			    cookiecount, buf, idx, th);
+	rc = llog_write_rec(env, loghandle, rec, NULL, idx, th);
 	up_write(&loghandle->lgh_lock);
 out_trans:
 	dt_trans_stop(env, dt, th);
@@ -929,6 +1133,17 @@ out:
 }
 EXPORT_SYMBOL(llog_close);
 
+/**
+ * Helper function to get the llog size in records. It is used by MGS
+ * mostly to check that config llog exists and contains data.
+ *
+ * \param[in] env	execution environment
+ * \param[in] ctxt	llog context
+ * \param[in] name	llog name
+ *
+ * \retval		true if there are records in llog besides a header
+ * \retval		false on error or llog without records
+ */
 int llog_is_empty(const struct lu_env *env, struct llog_ctxt *ctxt,
 		  char *name)
 {
@@ -950,7 +1165,8 @@ int llog_is_empty(const struct lu_env *env, struct llog_ctxt *ctxt,
 out_close:
 	llog_close(env, llh);
 out:
-	/* header is record 1 */
+	/* The header is record 1, the llog is still considered as empty
+	 * if there is only header */
 	return (rc <= 1);
 }
 EXPORT_SYMBOL(llog_is_empty);
@@ -961,9 +1177,8 @@ int llog_copy_handler(const struct lu_env *env, struct llog_handle *llh,
 	struct llog_handle	*copy_llh = data;
 
 	/* Append all records */
-	return llog_write(env, copy_llh, rec, NULL, 0, NULL, -1);
+	return llog_write(env, copy_llh, rec, LLOG_NEXT_IDX);
 }
-EXPORT_SYMBOL(llog_copy_handler);
 
 /* backup plain llog */
 int llog_backup(const struct lu_env *env, struct obd_device *obd,