From db0e1a7eba9926630e0499ab9a843c09f649e6c2 Mon Sep 17 00:00:00 2001

From: Mark Reynolds <mreynolds@redhat.com>

Date: Tue, 8 Apr 2014 14:39:47 -0400

Subject: [PATCH 196/225] Ticket 47771 - Performing deletes during tombstone

 purging results in operation errors

Bug Description:  An operations error can occur when deleting entry while

                  tombstone purging is happening.  The error occurs when it

                  tries the lock the parent entry, but the parent entry was

                  replaced in the cache before it could be locked.

Fix Description:  Return a special error code when cache_lock_entry fails because

                  the entry was marked as deleted.  Then try to grab the entry

                  again and lock it.

https://fedorahosted.org/389/ticket/47771

Reviewed by: rmeggins & nhosoi(Thanks!!)

(cherry picked from commit c5f22dd6c278f670fc36af8029d5c28a89051cfa)

---

 ldap/servers/slapd/back-ldbm/back-ldbm.h   |  1 +

 ldap/servers/slapd/back-ldbm/cache.c       |  6 +++--

 ldap/servers/slapd/back-ldbm/ldbm_delete.c | 38 ++++++++++++++++++++++++------

 3 files changed, 36 insertions(+), 9 deletions(-)

diff --git a/ldap/servers/slapd/back-ldbm/back-ldbm.h b/ldap/servers/slapd/back-ldbm/back-ldbm.h

index 7e5a261..8ad3c20 100644

--- a/ldap/servers/slapd/back-ldbm/back-ldbm.h

+++ b/ldap/servers/slapd/back-ldbm/back-ldbm.h

@@ -215,6 +215,7 @@ typedef unsigned short u_int16_t;

 #define DEFAULT_IMPORT_INDEX_BUFFER_SIZE  0

 #define SUBLEN                   3

 #define LDBM_CACHE_RETRY_COUNT 1000 /* Number of times we re-try a cache operation */

+#define RETRY_CACHE_LOCK 2 /* error code to signal a retry of the cache lock */

 #define IDL_FETCH_RETRY_COUNT 5 /* Number of times we re-try idl_fetch if it returns deadlock */

 #define IMPORT_SUBCOUNT_HASHTABLE_SIZE 500 /* Number of buckets in hash used to accumulate subcount for broody parents */

diff --git a/ldap/servers/slapd/back-ldbm/cache.c b/ldap/servers/slapd/back-ldbm/cache.c

index d97644f..865f1ef 100644

--- a/ldap/servers/slapd/back-ldbm/cache.c

+++ b/ldap/servers/slapd/back-ldbm/cache.c

@@ -1471,7 +1471,9 @@ void cache_unlock(struct cache *cache)

 /* locks an entry so that it can be modified (you should have gotten the

  * entry via cache_find_*).

- * returns 0 on success, 1 if the entry is scheduled for deletion.

+ * returns 0 on success,

+ * returns 1 if the entry lock could not be created

+ * returns 2 (RETRY_CACHE_LOCK) if the entry is scheduled for deletion.

  */

 int cache_lock_entry(struct cache *cache, struct backentry *e)

 {

@@ -1503,7 +1505,7 @@ int cache_lock_entry(struct cache *cache, struct backentry *e)

        PR_Unlock(cache->c_mutex);

        PR_ExitMonitor(e->ep_mutexp);

        LOG("<= cache_lock_entry (DELETED)\n", 0, 0, 0);

-       return 1;

+       return RETRY_CACHE_LOCK;

     }

     PR_Unlock(cache->c_mutex);

diff --git a/ldap/servers/slapd/back-ldbm/ldbm_delete.c b/ldap/servers/slapd/back-ldbm/ldbm_delete.c

index a4b8d8e..7c036df 100644

--- a/ldap/servers/slapd/back-ldbm/ldbm_delete.c

+++ b/ldap/servers/slapd/back-ldbm/ldbm_delete.c

@@ -335,14 +335,37 @@ ldbm_back_delete( Slapi_PBlock *pb )

 			 * (find_entry2modify_only_ext), a wrong parent could be found,

 			 * and numsubordinate count could get confused.

 			 */

-			ID pid = (ID)strtol(pid_str, (char **)NULL, 10);

+			ID pid;

+			int cache_retry_count = 0;

+			int cache_retry = 0;

+

+			pid = (ID)strtol(pid_str, (char **)NULL, 10);

 			slapi_ch_free_string(&pid_str);

-			parent = id2entry(be, pid ,NULL, &retval);

-			if (parent && cache_lock_entry(&inst->inst_cache, parent)) {

-				/* Failed to obtain parent entry's entry lock */

-				CACHE_RETURN(&(inst->inst_cache), &parent);

-				retval = -1;

-				goto error_return;

+

+			/*

+			 * Its possible that the parent entry retrieved from the cache in id2entry

+			 * could be removed before we lock it, because tombstone purging updated/replaced

+			 * the parent.  If we fail to lock the entry, just try again.

+			 */

+			while(1){

+				parent = id2entry(be, pid ,NULL, &retval);

+				if (parent && (cache_retry = cache_lock_entry(&inst->inst_cache, parent))) {

+					/* Failed to obtain parent entry's entry lock */

+					if(cache_retry == RETRY_CACHE_LOCK &&

+					   cache_retry_count < LDBM_CACHE_RETRY_COUNT)

+					{

+						/* try again */

+						DS_Sleep(PR_MillisecondsToInterval(100));

+						cache_retry_count++;

+						continue;

+					}

+					retval = -1;

+					CACHE_RETURN(&(inst->inst_cache), &parent);

+					goto error_return;

+				} else {

+					/* entry locked, move on */

+					break;

+				}

 			}

 		}

 		if (NULL == parent) {

@@ -1224,6 +1247,7 @@ diskfull_return:

 	slapi_ch_free((void**)&errbuf);

 	slapi_sdn_done(&nscpEntrySDN);

 	slapi_ch_free_string(&e_uniqueid);

+	slapi_sdn_done(&parentsdn);

 	if (pb->pb_conn)

 	{

 		slapi_log_error (SLAPI_LOG_TRACE, "ldbm_back_delete", "leave conn=%" NSPRIu64 " op=%d\n", pb->pb_conn->c_connid, operation->o_opid);

-- 

1.8.1.4