Re: [PATCH 1/5] mm/vmscan: Throttle reclaim until some writeback completes if congested

[Hillf Danton <hdanton@xxxxxxxx>]

On Mon, 20 Sep 2021 09:54:32 +0100 Mel Gorman wrote:
> +static void
> +reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason,
> +							long timeout)
> +{
> +	wait_queue_head_t *wqh =3D &pgdat->reclaim_wait;
> +	unsigned long start =3D jiffies;
> +	long ret;
> +	DEFINE_WAIT(wait);
> +
> +	atomic_inc(&pgdat->nr_reclaim_throttled);
> +	WRITE_ONCE(pgdat->nr_reclaim_start,
> +		 node_page_state(pgdat, NR_THROTTLED_WRITTEN));

Missing wakeup could happen if the current sleeper overwrites 
pgdat->nr_reclaim_start set by the existing sleeper.

	if (1 == atomic_inc_and_return(&pgdat->nr_reclaim_throttled))
		WRITE_ONCE(pgdat->nr_reclaim_start,
				node_page_state(pgdat, NR_THROTTLED_WRITTEN));

> +
> +	prepare_to_wait(wqh, &wait, TASK_INTERRUPTIBLE);
> +	ret =3D schedule_timeout(timeout);
> +	finish_wait(wqh, &wait);
> +	atomic_dec(&pgdat->nr_reclaim_throttled);
> +
> +	trace_mm_vmscan_throttled(pgdat->node_id, jiffies_to_usecs(timeout),
> +				jiffies_to_usecs(jiffies - start),
> +				reason);
> +}
> +
> +/*
> + * Account for pages written if tasks are throttled waiting on dirty
> + * pages to clean. If enough pages have been cleaned since throttling
> + * started then wakeup the throttled tasks.
> + */
> +void __acct_reclaim_writeback(pg_data_t *pgdat, struct page *page)
> +{
> +	unsigned long nr_written;
> +	int nr_throttled =3D atomic_read(&pgdat->nr_reclaim_throttled);
> +
> +	__inc_node_page_state(page, NR_THROTTLED_WRITTEN);
> +	nr_written =3D node_page_state(pgdat, NR_THROTTLED_WRITTEN) -
> +		READ_ONCE(pgdat->nr_reclaim_start);
> +
> +	if (nr_written > SWAP_CLUSTER_MAX * nr_throttled)
> +		wake_up_interruptible_all(&pgdat->reclaim_wait);
> +}