From 9eedc884416bd55dbffeac4d0d061baf3f288945 Mon Sep 17 00:00:00 2001
From: Leo Yuriev <leo@yuriev.ru>
Date: Mon, 29 Dec 2014 23:57:26 +0300
Subject: [PATCH] lmdb: ITS#7974 oom-handler feature.

Change-Id: I2c56e003fa1e4abe934288581e4c52c80db27c08
---
 lmdb.h |  33 +++++++++++++++++++
 mdb.c  | 100 ++++++++++++++++++++++++++++++++++++++++++++++++++++++---
 2 files changed, 129 insertions(+), 4 deletions(-)

diff --git a/lmdb.h b/lmdb.h
index 9db5d6fe..7d461a68 100644
--- a/lmdb.h
+++ b/lmdb.h
@@ -1585,6 +1585,39 @@ int	mdb_reader_check(MDB_env *env, int *dead);
 	 */
 int  mdb_txn_straggler(MDB_txn *txnm, int *percent);
 
+	/** @brief A callback function for killing a laggard readers,
+	 * called in case of MDB_MAP_FULL error.
+	 *
+	 * @param[in] env An environment handle returned by #mdb_env_create().
+	 * @param[in] pid pid of the reader process.
+	 * @param[in] thread_id thread_id of the reader thread.
+	 * @param[in] txn Transaction number on which stalled.
+	 * @return -1 on failure (reader is not killed),
+	 * 	0 on a race condition (no such reader),
+	 * 	1 on success (reader was killed),
+	 * 	>1 on success (reader was SURE killed).
+	 */
+typedef int (MDB_oom_func)(MDB_env *env, int pid, void* thread_id, size_t txn, unsigned gap, int retry);
+
+	/** @brief Set the OOM callback.
+	 *
+	 * Callback will be called only on out-of-pages case for killing
+	 * a laggard readers to allowing reclaiming of freeDB.
+	 *
+	 * @param[in] env An environment handle returned by #mdb_env_create().
+	 * @param[in] oomfunc A #MDB_oom_func function or NULL to disable.
+	 */
+void mdb_env_set_oomfunc(MDB_env *env, MDB_oom_func *oom_func);
+
+	/** @brief Get the current oom_func callback.
+	 *
+	 * Callback will be called only on out-of-pages case for killing
+	 * a laggard readers to allowing reclaiming of freeDB.
+	 *
+	 * @param[in] env An environment handle returned by #mdb_env_create().
+	 * @return A #MDB_oom_func function or NULL if disabled.
+	 */
+MDB_oom_func* mdb_env_get_oomfunc(MDB_env *env);
 /**	@} */
 
 #ifdef __cplusplus
diff --git a/mdb.c b/mdb.c
index 3fcea42c..9a47ca48 100644
--- a/mdb.c
+++ b/mdb.c
@@ -1191,6 +1191,7 @@ struct MDB_env {
 	uint64_t	me_sync_pending;	/**< Total dirty/commited bytes since the last mdb_env_sync() */
 	uint64_t	me_sync_threshold;	/**< Treshold of above to force synchronous flush */
 	size_t	me_sync_size;	/**< Tracking me_size for FGREW/fsync() */
+	MDB_oom_func	*me_oom_func; /**< Callback for kicking laggard readers */
 };
 
 	/** Nested transaction */
@@ -1947,6 +1948,80 @@ mdb_find_oldest(MDB_txn *txn)
 	return oldest;
 }
 
+static txnid_t
+mdb_laggard_reader(MDB_env *env, int *laggard)
+{
+	txnid_t tail = 0;
+	if (laggard)
+		*laggard = -1;
+	if (env->me_txns->mti_txnid > 1) {
+		int i;
+		MDB_reader *r = env->me_txns->mti_readers;
+
+		tail = env->me_txns->mti_txnid - 1;
+		for (i = env->me_txns->mti_numreaders; --i >= 0; ) {
+			if (r[i].mr_pid) {
+				txnid_t mr = r[i].mr_txnid;
+				if (tail > mr) {
+					tail = mr;
+					if (laggard)
+						*laggard = i;
+				}
+			}
+		}
+	}
+
+	return tail;
+}
+
+static int
+mdb_oomkick_laggard(MDB_env *env)
+{
+	int idx, retry;
+	txnid_t snap, tail = mdb_laggard_reader(env, &idx);
+	if (idx < 0)
+		return 0;
+
+	for(retry = 0; ; ++retry) {
+		MDB_reader *r;
+		MDB_THR_T tid;
+		pid_t pid;
+		int rc;
+
+		if (mdb_reader_check(env, NULL))
+			break;
+
+		snap = mdb_laggard_reader(env, NULL);
+		if (tail < snap)
+			return 1;
+
+		if (!env->me_oom_func)
+			break;
+
+		r = &env->me_txns->mti_readers[ idx ];
+		pid = r->mr_pid;
+		tid = r->mr_tid;
+		if (r->mr_txnid != tail || pid <= 0)
+			continue;
+
+		rc = env->me_oom_func(env, pid, (void*) tid, tail,
+			env->me_metas[ mdb_env_pick_meta(env) ]->mm_txnid - tail, retry);
+		if (rc < 0)
+			break;
+
+		if (rc) {
+			r->mr_txnid = (txnid_t)-1;
+			if (rc > 1) {
+				r->mr_tid = 0;
+				r->mr_pid = 0;
+			}
+		}
+	}
+
+	snap = mdb_laggard_reader(env, NULL);
+	return tail < snap;
+}
+
 /** Add a page to the txn's dirty list */
 static void
 mdb_page_dirty(MDB_txn *txn, MDB_page *mp)
@@ -2025,6 +2100,7 @@ mdb_page_alloc(MDB_cursor *mc, int num, MDB_page **mp)
 		goto fail;
 	}
 
+oomkick_retry:;
 	for (op = MDB_FIRST;; op = MDB_NEXT) {
 		MDB_val key, data;
 		MDB_node *leaf;
@@ -2120,9 +2196,11 @@ mdb_page_alloc(MDB_cursor *mc, int num, MDB_page **mp)
 	i = 0;
 	pgno = txn->mt_next_pgno;
 	if (pgno + num > env->me_maxpg) {
-			DPUTS("DB size maxed out");
-			rc = MDB_MAP_FULL;
-			goto fail;
+		DPUTS("DB size maxed out");
+		if (mdb_oomkick_laggard(env))
+			goto oomkick_retry;
+		rc = MDB_MAP_FULL;
+		goto fail;
 	}
 
 search_done:
@@ -2911,7 +2989,7 @@ mdb_txn_straggler(MDB_txn *txn, int *percent)
 		*percent = (meta->mm_last_pg + cent / 2 + 1) / (cent ? cent : 1);
 	}
 	lag = meta->mm_txnid - txn->mt_u.reader->mr_txnid;
-	return (0 > (int) lag) ? ~0u >> 1: lag;
+	return (0 > (long) lag) ? ~0u >> 1: lag;
 }
 
 /** Common code for #mdb_txn_reset() and #mdb_txn_abort().
@@ -9719,4 +9797,18 @@ static int mdb_mutex_failed(MDB_env *env, mdb_mutex_t *mutex, int rc)
 	return rc;
 }
 #endif	/* MDB_ROBUST_SUPPORTED */
+
+void
+mdb_env_set_oomfunc(MDB_env *env, MDB_oom_func *oomfunc)
+{
+	if (env)
+		env->me_oom_func = oomfunc;
+}
+
+MDB_oom_func*
+mdb_env_get_oomfunc(MDB_env *env)
+{
+	return env ? env->me_oom_func : NULL;
+}
+
 /** @} */