linux-kernel - Re: [PATCH v10 06/16] mm/damon: Split regions into 4 subregions if possible

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date:   Thu, 7 May 2020 10:28:36 +0200
From:   SeongJae Park <sjpark@...zon.com>
To:     <Jonathan.Cameron@...wei.com>, SeongJae Park <sjpark@...zon.com>
CC:     <akpm@...ux-foundation.org>, SeongJae Park <sjpark@...zon.de>,
        <aarcange@...hat.com>, <acme@...nel.org>,
        <alexander.shishkin@...ux.intel.com>, <amit@...nel.org>,
        <benh@...nel.crashing.org>, <brendan.d.gregg@...il.com>,
        <brendanhiggins@...gle.com>, <cai@....pw>,
        <colin.king@...onical.com>, <corbet@....net>, <dwmw@...zon.com>,
        <irogers@...gle.com>, <jolsa@...hat.com>, <kirill@...temov.name>,
        <mark.rutland@....com>, <mgorman@...e.de>, <minchan@...nel.org>,
        <mingo@...hat.com>, <namhyung@...nel.org>, <peterz@...radead.org>,
        <rdunlap@...radead.org>, <riel@...riel.com>, <rientjes@...gle.com>,
        <rostedt@...dmis.org>, <sblbir@...zon.com>, <shakeelb@...gle.com>,
        <shuah@...nel.org>, <sj38.park@...il.com>, <snu@...zon.de>,
        <vbabka@...e.cz>, <vdavydov.dev@...il.com>,
        <yang.shi@...ux.alibaba.com>, <ying.huang@...el.com>,
        <linux-damon@...zon.com>, <linux-mm@...ck.org>,
        <linux-doc@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v10 06/16] mm/damon: Split regions into 4 subregions if possible

On Tue, 5 May 2020 13:08:05 +0200 SeongJae Park <sjpark@...zon.com> wrote:

> From: SeongJae Park <sjpark@...zon.de>
> 
> Suppose that there are a very wide and cold region and a hot region, and
> both regions are identified by DAMON.  And then, the middle small region
> inside the very wide and cold region becomes hot.  DAMON will not be
> able to identify this new region because the adaptive regions adjustment
> mechanism splits each region to only two subregions.
> 
> This commit modifies the logic to split each region to 4 subregions if
> possible so that such problematic region can eventually identified.

As you can check with the 'Evaluations' section in the cover letter of this
patchset, the extend of the slowdown DAMON's recording feature made to the
target workload has doulbed (0.55% with v9, 1.02% with this patchset).  It's
still just a small number, but the worst case slowdown is 5.29%.  It was only
1.88% before.  I suspect this patch is the reason, as this is the only one
major change we made after v9.

For the reason, I would like to make this patch to split each region into 3 sub
regions instead of 4 subregions.  It will reduce the overhead a little bit
while still allow the corner case regions be eventually detected.  If you have
other concerns or opinions, please let me know.


Thanks,
SeongJae Park

> 
> Suggested-by: James Cameron <Jonathan.Cameron@...wei.com>
> Signed-off-by: SeongJae Park <sjpark@...zon.de>
> ---
>  mm/damon.c | 44 +++++++++++++++++++++++++++-----------------
>  1 file changed, 27 insertions(+), 17 deletions(-)
> 
> diff --git a/mm/damon.c b/mm/damon.c
> index cec946197e13..342f905927a0 100644
> --- a/mm/damon.c
> +++ b/mm/damon.c
> @@ -650,26 +650,32 @@ static void damon_split_region_at(struct damon_ctx *ctx,
>  	damon_insert_region(new, r, damon_next_region(r));
>  }
>  
> -/* Split every region in the given task into two randomly-sized regions */
> -static void damon_split_regions_of(struct damon_ctx *ctx, struct damon_task *t)
> +/* Split every region in the given task into 'nr_subs' regions */
> +static void damon_split_regions_of(struct damon_ctx *ctx,
> +				     struct damon_task *t, int nr_subs)
>  {
>  	struct damon_region *r, *next;
> -	unsigned long sz_orig_region, sz_left_region;
> +	unsigned long sz_region, sz_sub = 0;
> +	int i;
>  
>  	damon_for_each_region_safe(r, next, t) {
> -		sz_orig_region = r->vm_end - r->vm_start;
> -
> -		/*
> -		 * Randomly select size of left sub-region to be at least
> -		 * 10 percent and at most 90% of original region
> -		 */
> -		sz_left_region = ALIGN_DOWN(damon_rand(1, 10) * sz_orig_region
> -					    / 10, MIN_REGION);
> -		/* Do not allow blank region */
> -		if (sz_left_region == 0 || sz_left_region >= sz_orig_region)
> -			continue;
> -
> -		damon_split_region_at(ctx, r, sz_left_region);
> +		sz_region = r->vm_end - r->vm_start;
> +
> +		for (i = 0; i < nr_subs - 1 &&
> +				sz_region > 2 * MIN_REGION; i++) {
> +			/*
> +			 * Randomly select size of left sub-region to be at
> +			 * least 10 percent and at most 90% of original region
> +			 */
> +			sz_sub = ALIGN_DOWN(damon_rand(1, 10) *
> +					sz_region / 10, MIN_REGION);
> +			/* Do not allow blank region */
> +			if (sz_sub == 0 || sz_sub >= sz_region)
> +				continue;
> +
> +			damon_split_region_at(ctx, r, sz_sub);
> +			sz_region = sz_sub;
> +		}
>  	}
>  }
>  
> @@ -687,14 +693,18 @@ static void kdamond_split_regions(struct damon_ctx *ctx)
>  {
>  	struct damon_task *t;
>  	unsigned int nr_regions = 0;
> +	int nr_subregions = 2;
>  
>  	damon_for_each_task(ctx, t)
>  		nr_regions += nr_damon_regions(t);
>  	if (nr_regions > ctx->max_nr_regions / 2)
>  		return;
>  
> +	if (nr_regions < ctx->max_nr_regions / 4)
> +		nr_subregions = 4;
> +
>  	damon_for_each_task(ctx, t)
> -		damon_split_regions_of(ctx, t);
> +		damon_split_regions_of(ctx, t, nr_subregions);
>  }
>  
>  /*
> -- 
> 2.17.1
>