X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftarget%2Fupdate_recovery.c;h=85764cfa073cf82f795137350e5571492cc5090b;hp=8d7f5cc857907438f8a60308833b50ce4b5e2b27;hb=a41c6fad4672a60166088b9ad8aeb4f1b51c38e7;hpb=9543df37cdfd35980c888440265d161e350d166d

diff --git a/lustre/target/update_recovery.c b/lustre/target/update_recovery.c
index 8d7f5cc..85764cf 100644
--- a/lustre/target/update_recovery.c
+++ b/lustre/target/update_recovery.c
@@ -20,7 +20,7 @@
  * GPL HEADER END
  */
 /*
- * Copyright (c) 2014, Intel Corporation.
+ * Copyright (c) 2014, 2015, Intel Corporation.
  */
 
 /*
@@ -44,10 +44,13 @@
 #define DEBUG_SUBSYSTEM S_CLASS
 
 #include <lu_target.h>
-#include <md_object.h>
+#include <lustre_obdo.h>
 #include <lustre_update.h>
+#include <lustre_swab.h>
+#include <md_object.h>
 #include <obd.h>
 #include <obd_class.h>
+
 #include "tgt_internal.h"
 
 /**
@@ -442,7 +445,7 @@ again:
 	 * we do not need lock here */
 	if (replace_record) {
 		/* Replace the update record and master transno */
-		OBD_FREE(dtrq->dtrq_lur, dtrq->dtrq_lur_size);
+		OBD_FREE_LARGE(dtrq->dtrq_lur, dtrq->dtrq_lur_size);
 		dtrq->dtrq_lur = NULL;
 		dtrq->dtrq_lur_size = llog_update_record_size(lur);
 		OBD_ALLOC_LARGE(dtrq->dtrq_lur, dtrq->dtrq_lur_size);
@@ -539,6 +542,11 @@ void dtrq_list_destroy(struct target_distribute_txn_data *tdtd)
 		list_del_init(&dtrq->dtrq_list);
 		dtrq_destroy(dtrq);
 	}
+	list_for_each_entry_safe(dtrq, tmp, &tdtd->tdtd_replay_finish_list,
+				 dtrq_list) {
+		list_del_init(&dtrq->dtrq_list);
+		dtrq_destroy(dtrq);
+	}
 	spin_unlock(&tdtd->tdtd_replay_list_lock);
 }
 EXPORT_SYMBOL(dtrq_list_destroy);
@@ -597,6 +605,40 @@ __u64 distribute_txn_get_next_transno(struct target_distribute_txn_data *tdtd)
 }
 EXPORT_SYMBOL(distribute_txn_get_next_transno);
 
+struct distribute_txn_replay_req *
+distribute_txn_lookup_finish_list(struct target_distribute_txn_data *tdtd,
+				  __u64 xid)
+{
+	struct distribute_txn_replay_req *dtrq = NULL;
+	struct distribute_txn_replay_req *iter;
+
+	spin_lock(&tdtd->tdtd_replay_list_lock);
+	list_for_each_entry(iter, &tdtd->tdtd_replay_finish_list, dtrq_list) {
+		if (iter->dtrq_xid == xid) {
+			dtrq = iter;
+			break;
+		}
+	}
+	spin_unlock(&tdtd->tdtd_replay_list_lock);
+	return dtrq;
+}
+
+bool is_req_replayed_by_update(struct ptlrpc_request *req)
+{
+	struct lu_target *tgt = class_exp2tgt(req->rq_export);
+	struct distribute_txn_replay_req *dtrq;
+
+	if (tgt->lut_tdtd == NULL)
+		return false;
+
+	dtrq = distribute_txn_lookup_finish_list(tgt->lut_tdtd, req->rq_xid);
+	if (dtrq == NULL)
+		return false;
+
+	return true;
+}
+EXPORT_SYMBOL(is_req_replayed_by_update);
+
 /**
  * Check if the update of one object is committed
  *
@@ -1025,6 +1067,7 @@ static void update_recovery_update_ses(struct lu_env *env,
 				      struct target_distribute_txn_data *tdtd,
 				      struct thandle *th,
 				      struct thandle *master_th,
+				      struct distribute_txn_replay_req *dtrq,
 				      struct tx_arg *ta_arg)
 {
 	struct tgt_session_info	*tsi;
@@ -1068,6 +1111,7 @@ static void update_recovery_update_ses(struct lu_env *env,
 	tsi->tsi_opdata = lrd->lrd_data;
 	tsi->tsi_result = lrd->lrd_result;
 	tsi->tsi_client_gen = lrd->lrd_client_gen;
+	dtrq->dtrq_xid = lrd->lrd_xid;
 	top_th = container_of(th, struct top_thandle, tt_super);
 	top_th->tt_master_sub_thandle = master_th;
 	cfs_hash_putref(hash);
@@ -1123,6 +1167,17 @@ static int update_recovery_exec(const struct lu_env *env,
 		dt_obj = dt_locate(env, tdtd->tdtd_dt, fid);
 		if (IS_ERR(dt_obj)) {
 			rc = PTR_ERR(dt_obj);
+			if (rc == -EREMCHG)
+				LCONSOLE_WARN("%.16s: hit invalid OI mapping "
+					      "for "DFID" during recovering, "
+					      "that may because auto scrub is "
+					      "disabled on related MDT, and "
+					      "will cause recovery failure. "
+					      "Please enable auto scrub and "
+					      "retry the recovery.\n",
+					      tdtd->tdtd_lut->lut_obd->obd_name,
+					      PFID(fid));
+
 			break;
 		}
 		sub_dt_obj = dt_object_child(dt_obj);
@@ -1335,7 +1390,7 @@ int distribute_txn_replay_handle(struct lu_env *env,
 		 * tgt_last_rcvd_update() can be called correctly */
 		if (rc == 0 && dt_obj == tdtd->tdtd_lut->lut_reply_data)
 			update_recovery_update_ses(env, tdtd, th,
-						   st->st_sub_th, ta_arg);
+						   st->st_sub_th, dtrq, ta_arg);
 
 		if (unlikely(rc < 0)) {
 			CDEBUG(D_HA, "error during execution of #%u from"