Re: [PATCH 1/1] mm: vmscan: Reduce throttling due to a failure to make progress

From: Mike Galbraith <efault@gmx.de>
To: Alexey Avramov <hakavlad@inbox.lv>,
	Mel Gorman <mgorman@techsingularity.net>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Michal Hocko <mhocko@suse.com>,  Vlastimil Babka <vbabka@suse.cz>,
	Rik van Riel <riel@surriel.com>, Darrick Wong <djwong@kernel.org>,
	regressions@lists.linux.dev,
	Linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	Linux-MM <linux-mm@kvack.org>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 1/1] mm: vmscan: Reduce throttling due to a failure to make progress
Date: Sun, 28 Nov 2021 19:38:37 +0100	[thread overview]
Message-ID: <1c1835271e0ea093dd169d19038b477cf8563c32.camel@gmx.de> (raw)
In-Reply-To: <dd39029d93bb4de5ed485b5d4181fc19d4c0c4f0.camel@gmx.de>

On Sun, 2021-11-28 at 13:35 +0100, Mike Galbraith wrote:
>
> Bandaid likely just break the writeback bits all to pieces, but that's
> ok, if breakage makes things work better, they need more breaking :)

@@ -1048,6 +1050,10 @@ void reclaim_throttle(pg_data_t *pgdat,
 	 */
 	switch(reason) {
 	case VMSCAN_THROTTLE_WRITEBACK:
+		if (!async_bdi_congested()) {
+			cond_resched();
+			return;
+		}

And indeed, that's the only THROTTLE_WRITEBACK path I've seen taken.
Not pulling the plug made no difference to the tail /dev/zero test,
leaving only the bits I swiped from defunct wait_iff_congested() as
behavior delta cause.  Actually sleeping there apparently stings.

	-Mike
