MDEV-11802 InnoDB purge does not always run when there is work to do

srv_sys_t::n_threads_active[]: Protect writes by both the mutex and
by atomic memory access.

srv_active_wake_master_thread_low(): Reliably wake up the master
thread if there is work to do. The trick is to atomically read
srv_sys->n_threads_active[].

srv_wake_purge_thread_if_not_active(): Atomically read
srv_sys->n_threads_active[] (and trx_sys->rseg_history_len),
so that the purge should always be triggered when there is work to do.

trx_commit_in_memory(): Invoke srv_wake_purge_thread_if_not_active()
whenever a transaction is committed. Purge could have been prevented by
the read view of the currently committing transaction, even if it is
a read-only transaction.

trx_purge_add_update_undo_to_history(): Do not wake up the purge.
This is only called by trx_undo_update_cleanup(), as part of
trx_write_serialisation_history(), which in turn is only called by
trx_commit_low() which will always call trx_commit_in_memory().
Thus, the added call in trx_commit_in_memory() will cover also
this use case where a committing read-write transaction added
some update_undo log to the purge queue.

trx_rseg_mem_restore(): Atomically modify trx_sys->rseg_history_len.

5 files changed, 30 insertions, 61 deletions

diff --git a/storage/innobase/include/srv0srv.h b/storage/innobase/include/srv0srv.h
index 6803047ed91..230d7b7effe 100644
--- a/storage/innobase/include/srv0srv.h
+++ b/storage/innobase/include/srv0srv.h
@@ -749,36 +749,24 @@ srv_set_io_thread_op_info(
 Resets the info describing an i/o thread current state. */
 void
 srv_reset_io_thread_op_info();
-/*=========================*/
-/*******************************************************************//**
-Tells the purge thread that there has been activity in the database
-and wakes up the purge thread if it is suspended (not sleeping).  Note
-that there is a small chance that the purge thread stays suspended
-(we do not protect our operation with the srv_sys_t:mutex, for
-performance reasons). */
+
+/** Wake up the purge threads if there is work to do. */
 void
-srv_wake_purge_thread_if_not_active(void);
-/*=====================================*/
-/*******************************************************************//**
-Tells the Innobase server that there has been activity in the database
-and wakes up the master thread if it is suspended (not sleeping). Used
-in the MySQL interface. Note that there is a small chance that the master
-thread stays suspended (we do not protect our operation with the kernel
-mutex, for performace reasons). */
+srv_wake_purge_thread_if_not_active();
+/** Wake up the InnoDB master thread if it was suspended (not sleeping). */
 void
-srv_active_wake_master_thread_low(void);
-/*===================================*/
+srv_active_wake_master_thread_low();
+
 #define srv_active_wake_master_thread()					\
 	do {								\
 		if (!srv_read_only_mode) {				\
 			srv_active_wake_master_thread_low();		\
 		}							\
 	} while (0)
-/*******************************************************************//**
-Wakes up the master thread if it is suspended or being suspended. */
+/** Wake up the master thread if it is suspended or being suspended. */
 void
-srv_wake_master_thread(void);
-/*========================*/
+srv_wake_master_thread();
+
 /******************************************************************//**
 Outputs to a file the output of the InnoDB Monitor.
 @return FALSE if not all information printed
diff --git a/storage/innobase/srv/srv0srv.cc b/storage/innobase/srv/srv0srv.cc
index aa795e972a2..8c62cc976b1 100644
--- a/storage/innobase/srv/srv0srv.cc
+++ b/storage/innobase/srv/srv0srv.cc
@@ -648,14 +648,16 @@ struct srv_sys_t{
 
 	ulint		n_threads_active[SRV_MASTER + 1];
 						/*!< number of threads active
-						in a thread class */
+						in a thread class; protected
+						by both my_atomic_addlint()
+						and mutex */
 
 	srv_stats_t::ulint_ctr_1_t
 			activity_count;		/*!< For tracking server
 						activity */
 };
 
-static srv_sys_t*	srv_sys	= NULL;
+static srv_sys_t*	srv_sys;
 
 /** Event to signal srv_monitor_thread. Not protected by a mutex.
 Set after setting srv_print_innodb_monitor. */
@@ -853,7 +855,7 @@ srv_reserve_slot(
 
 	ut_ad(srv_slot_get_type(slot) == type);
 
-	++srv_sys->n_threads_active[type];
+	my_atomic_addlint(&srv_sys->n_threads_active[type], 1);
 
 	srv_sys_mutex_exit();
 
@@ -894,16 +896,15 @@ srv_suspend_thread_low(
 
 	case SRV_WORKER:
 		ut_a(srv_n_purge_threads > 1);
-		ut_a(srv_sys->n_threads_active[type] > 0);
 		break;
 	}
 
 	ut_a(!slot->suspended);
 	slot->suspended = TRUE;
 
-	ut_a(srv_sys->n_threads_active[type] > 0);
-
-	srv_sys->n_threads_active[type]--;
+	if (my_atomic_addlint(&srv_sys->n_threads_active[type], -1) < 0) {
+		ut_error;
+	}
 
 	return(os_event_reset(slot->event));
 }
@@ -958,7 +959,7 @@ srv_resume_thread(srv_slot_t* slot, int64_t sig_count = 0, bool wait = true,
 	ut_ad(slot->suspended);
 
 	slot->suspended = FALSE;
-	++srv_sys->n_threads_active[slot->type];
+	my_atomic_addlint(&srv_sys->n_threads_active[slot->type], 1);
 	srv_sys_mutex_exit();
 	return(timeout);
 }
@@ -2066,22 +2067,16 @@ srv_get_active_thread_type(void)
 	return(ret);
 }
 
-/*******************************************************************//**
-Tells the InnoDB server that there has been activity in the database
-and wakes up the master thread if it is suspended (not sleeping). Used
-in the MySQL interface. Note that there is a small chance that the master
-thread stays suspended (we do not protect our operation with the
-srv_sys_t->mutex, for performance reasons). */
+/** Wake up the InnoDB master thread if it was suspended (not sleeping). */
 void
 srv_active_wake_master_thread_low()
-/*===============================*/
 {
 	ut_ad(!srv_read_only_mode);
 	ut_ad(!srv_sys_mutex_own());
 
 	srv_inc_activity_count();
 
-	if (srv_sys->n_threads_active[SRV_MASTER] == 0) {
+	if (my_atomic_loadlint(&srv_sys->n_threads_active[SRV_MASTER]) == 0) {
 		srv_slot_t*	slot;
 
 		srv_sys_mutex_enter();
@@ -2099,35 +2094,25 @@ srv_active_wake_master_thread_low()
 	}
 }
 
-/*******************************************************************//**
-Tells the purge thread that there has been activity in the database
-and wakes up the purge thread if it is suspended (not sleeping).  Note
-that there is a small chance that the purge thread stays suspended
-(we do not protect our check with the srv_sys_t:mutex and the
-purge_sys->latch, for performance reasons). */
+/** Wake up the purge threads if there is work to do. */
 void
-srv_wake_purge_thread_if_not_active(void)
-/*=====================================*/
+srv_wake_purge_thread_if_not_active()
 {
 	ut_ad(!srv_sys_mutex_own());
 
 	if (purge_sys->state == PURGE_STATE_RUN
-	    && srv_sys->n_threads_active[SRV_PURGE] == 0) {
+	    && !my_atomic_loadlint(&srv_sys->n_threads_active[SRV_PURGE])
+	    && my_atomic_loadlint(&trx_sys->rseg_history_len)) {
 
 		srv_release_threads(SRV_PURGE, 1);
 	}
 }
 
-/*******************************************************************//**
-Wakes up the master thread if it is suspended or being suspended. */
+/** Wake up the master thread if it is suspended or being suspended. */
 void
-srv_wake_master_thread(void)
-/*========================*/
+srv_wake_master_thread()
 {
-	ut_ad(!srv_sys_mutex_own());
-
 	srv_inc_activity_count();
-
 	srv_release_threads(SRV_MASTER, 1);
 }
 
@@ -2713,12 +2698,8 @@ DECLARE_THREAD(srv_worker_thread)(
 	slot = srv_reserve_slot(SRV_WORKER);
 
 	ut_a(srv_n_purge_threads > 1);
-
-	srv_sys_mutex_enter();
-
-	ut_a(srv_sys->n_threads_active[SRV_WORKER] < srv_n_purge_threads);
-
-	srv_sys_mutex_exit();
+	ut_a(my_atomic_loadlint(&srv_sys->n_threads_active[SRV_WORKER])
+	     < static_cast<lint>(srv_n_purge_threads));
 
 	/* We need to ensure that the worker threads exit after the
 	purge coordinator thread. Otherwise the purge coordinator can
diff --git a/storage/innobase/trx/trx0purge.cc b/storage/innobase/trx/trx0purge.cc
index 42ae4890cab..b21ec75c3a6 100644
--- a/storage/innobase/trx/trx0purge.cc
+++ b/storage/innobase/trx/trx0purge.cc
@@ -291,7 +291,6 @@ trx_purge_add_update_undo_to_history(
 		       undo_header + TRX_UNDO_HISTORY_NODE, mtr);
 
 	my_atomic_addlint(&trx_sys->rseg_history_len, 1);
-	srv_wake_purge_thread_if_not_active();
 
 	/* Write the trx number to the undo log header */
 	mlog_write_ull(undo_header + TRX_UNDO_TRX_NO, trx->no, mtr);
diff --git a/storage/innobase/trx/trx0rseg.cc b/storage/innobase/trx/trx0rseg.cc
index 3393a0464a8..663566cf26f 100644
--- a/storage/innobase/trx/trx0rseg.cc
+++ b/storage/innobase/trx/trx0rseg.cc
@@ -204,7 +204,7 @@ trx_rseg_mem_restore(trx_rseg_t* rseg, mtr_t* mtr)
 	len = flst_get_len(rseg_header + TRX_RSEG_HISTORY);
 
 	if (len > 0) {
-		trx_sys->rseg_history_len += len;
+		my_atomic_addlint(&trx_sys->rseg_history_len, len);
 
 		node_addr = trx_purge_get_log_from_hist(
 			flst_get_last(rseg_header + TRX_RSEG_HISTORY, mtr));
diff --git a/storage/innobase/trx/trx0trx.cc b/storage/innobase/trx/trx0trx.cc
index 36324c43970..d6755fca7c0 100644
--- a/storage/innobase/trx/trx0trx.cc
+++ b/storage/innobase/trx/trx0trx.cc
@@ -1908,6 +1908,7 @@ trx_commit_in_memory(
 	trx_mutex_exit(trx);
 
 	ut_a(trx->error_state == DB_SUCCESS);
+	srv_wake_purge_thread_if_not_active();
 }
 
 /****************************************************************//**