writeback: optimize periodic bdi thread wakeups
Whe the first inode for a bdi is marked dirty, we wake up the bdi thread which
should take care of the periodic background write-out. However, the write-out
will actually start only 'dirty_writeback_interval' centisecs later, so we can
delay the wake-up.
This change was requested by Nick Piggin who pointed out that if we delay the
wake-up, we weed out 2 unnecessary contex switches, which matters because
'__mark_inode_dirty()' is a hot-path function.
This patch introduces a new function - 'bdi_wakeup_thread_delayed()', which
sets up a timer to wake-up the bdi thread and returns. So the wake-up is
delayed.
We also delete the timer in bdi threads just before writing-back. And
synchronously delete it when unregistering bdi. At the unregister point the bdi
does not have any users, so no one can arm it again.
Since now we take 'bdi->wb_lock' in the timer, which can execute in softirq
context, we have to use 'spin_lock_bh()' for 'bdi->wb_lock'. This patch makes
this change as well.
This patch also moves the 'bdi_wb_init()' function down in the file to avoid
forward-declaration of 'bdi_wakeup_thread_delayed()'.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index a9a08d8..cfff722 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -248,17 +248,6 @@
}
subsys_initcall(default_bdi_init);
-static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
-{
- memset(wb, 0, sizeof(*wb));
-
- wb->bdi = bdi;
- wb->last_old_flush = jiffies;
- INIT_LIST_HEAD(&wb->b_dirty);
- INIT_LIST_HEAD(&wb->b_io);
- INIT_LIST_HEAD(&wb->b_more_io);
-}
-
int bdi_has_dirty_io(struct backing_dev_info *bdi)
{
return wb_has_dirty_io(&bdi->wb);
@@ -316,6 +305,43 @@
bdi_arm_supers_timer();
}
+static void wakeup_timer_fn(unsigned long data)
+{
+ struct backing_dev_info *bdi = (struct backing_dev_info *)data;
+
+ spin_lock_bh(&bdi->wb_lock);
+ if (bdi->wb.task) {
+ wake_up_process(bdi->wb.task);
+ } else {
+ /*
+ * When bdi tasks are inactive for long time, they are killed.
+ * In this case we have to wake-up the forker thread which
+ * should create and run the bdi thread.
+ */
+ wake_up_process(default_backing_dev_info.wb.task);
+ }
+ spin_unlock_bh(&bdi->wb_lock);
+}
+
+/*
+ * This function is used when the first inode for this bdi is marked dirty. It
+ * wakes-up the corresponding bdi thread which should then take care of the
+ * periodic background write-out of dirty inodes. Since the write-out would
+ * starts only 'dirty_writeback_interval' centisecs from now anyway, we just
+ * set up a timer which wakes the bdi thread up later.
+ *
+ * Note, we wouldn't bother setting up the timer, but this function is on the
+ * fast-path (used by '__mark_inode_dirty()'), so we save few context switches
+ * by delaying the wake-up.
+ */
+void bdi_wakeup_thread_delayed(struct backing_dev_info *bdi)
+{
+ unsigned long timeout;
+
+ timeout = msecs_to_jiffies(dirty_writeback_interval * 10);
+ mod_timer(&bdi->wb.wakeup_timer, jiffies + timeout);
+}
+
/*
* Calculate the longest interval (jiffies) bdi threads are allowed to be
* inactive.
@@ -353,8 +379,10 @@
* Temporary measure, we want to make sure we don't see
* dirty data on the default backing_dev_info
*/
- if (wb_has_dirty_io(me) || !list_empty(&me->bdi->work_list))
+ if (wb_has_dirty_io(me) || !list_empty(&me->bdi->work_list)) {
+ del_timer(&me->wakeup_timer);
wb_do_writeback(me, 0);
+ }
spin_lock_bh(&bdi_lock);
set_current_state(TASK_INTERRUPTIBLE);
@@ -386,7 +414,7 @@
break;
}
- spin_lock(&bdi->wb_lock);
+ spin_lock_bh(&bdi->wb_lock);
/*
* If there is no work to do and the bdi thread was
* inactive long enough - kill it. The wb_lock is taken
@@ -403,7 +431,7 @@
action = KILL_THREAD;
break;
}
- spin_unlock(&bdi->wb_lock);
+ spin_unlock_bh(&bdi->wb_lock);
}
spin_unlock_bh(&bdi_lock);
@@ -427,9 +455,9 @@
* The spinlock makes sure we do not lose
* wake-ups when racing with 'bdi_queue_work()'.
*/
- spin_lock(&bdi->wb_lock);
+ spin_lock_bh(&bdi->wb_lock);
bdi->wb.task = task;
- spin_unlock(&bdi->wb_lock);
+ spin_unlock_bh(&bdi->wb_lock);
}
break;
@@ -586,6 +614,7 @@
if (bdi->dev) {
trace_writeback_bdi_unregister(bdi);
bdi_prune_sb(bdi);
+ del_timer_sync(&bdi->wb.wakeup_timer);
if (!bdi_cap_flush_forker(bdi))
bdi_wb_shutdown(bdi);
@@ -596,6 +625,18 @@
}
EXPORT_SYMBOL(bdi_unregister);
+static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
+{
+ memset(wb, 0, sizeof(*wb));
+
+ wb->bdi = bdi;
+ wb->last_old_flush = jiffies;
+ INIT_LIST_HEAD(&wb->b_dirty);
+ INIT_LIST_HEAD(&wb->b_io);
+ INIT_LIST_HEAD(&wb->b_more_io);
+ setup_timer(&wb->wakeup_timer, wakeup_timer_fn, (unsigned long)bdi);
+}
+
int bdi_init(struct backing_dev_info *bdi)
{
int i, err;