[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20241009135252.53276de999d3006a20cad21c@linux-foundation.org>
Date: Wed, 9 Oct 2024 13:52:52 -0700
From: Andrew Morton <akpm@...ux-foundation.org>
To: "zhaoyang.huang" <zhaoyang.huang@...soc.com>
Cc: Yu Zhao <yuzhao@...gle.com>, <linux-mm@...ck.org>,
<linux-kernel@...r.kernel.org>, Zhaoyang Huang <huangzhaoyang@...il.com>,
<steve.kang@...soc.com>
Subject: Re: [PATCH] mm: throttle and inc min_seq when both page types reach
MIN_NR_GENS
On Wed, 9 Oct 2024 15:49:53 +0800 "zhaoyang.huang" <zhaoyang.huang@...soc.com> wrote:
> From: Zhaoyang Huang <zhaoyang.huang@...soc.com>
>
> The test case of [1] leads to system hang which caused by a local
> watchdog thread starved over 20s on a 5.5GB RAM ANDROID15(v6.6)
> system. This commit solve the issue by have the reclaimer be throttled
> and increase min_seq if both page types reach MIN_NR_GENS, which may
> introduce a livelock of switching type with holding lruvec->lru_lock.
>
> [1]
> launch below script 8 times simutanously which allocates 1GB virtual
> memory and access it from user space by each thread.
> $ costmem -c1024000 -b12800 -o0 &
>
That looks like a pretty simple testcase. Do people know where to get
`costmem' from?
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
This is a somewhat serious issue, so I'll add the patch for some
testing, but I'll await feedback from MGLRU developers before
proceeeding further, thanks.
Powered by blists - more mailing lists