linux-kernel - Re: [PATCH v2] mm, compaction: properly signal and act upon lock and need

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] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 21 May 2014 13:11:57 -0700
From:	Andrew Morton <akpm@...ux-foundation.org>
To:	Vlastimil Babka <vbabka@...e.cz>
Cc:	Joonsoo Kim <iamjoonsoo.kim@....com>,
	David Rientjes <rientjes@...gle.com>,
	Hugh Dickins <hughd@...gle.com>,
	Greg Thelen <gthelen@...gle.com>, linux-kernel@...r.kernel.org,
	linux-mm@...ck.org, Minchan Kim <minchan@...nel.org>,
	Mel Gorman <mgorman@...e.de>,
	Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>,
	Michal Nazarewicz <mina86@...a86.com>,
	Christoph Lameter <cl@...ux.com>,
	Rik van Riel <riel@...hat.com>
Subject: Re: [PATCH v2] mm, compaction: properly signal and act upon lock
 and need_sched() contention

On Wed, 21 May 2014 16:13:39 +0200 Vlastimil Babka <vbabka@...e.cz> wrote:

> >>
> >> ...
> >>
> >> @@ -718,9 +739,11 @@ static void isolate_freepages(struct zone *zone,
> >>   		/*
> >>   		 * This can iterate a massively long zone without finding any
> >>   		 * suitable migration targets, so periodically check if we need
> >> -		 * to schedule.
> >> +		 * to schedule, or even abort async compaction.
> >>   		 */
> >> -		cond_resched();
> >> +		if (!(block_start_pfn % (SWAP_CLUSTER_MAX * pageblock_nr_pages))
> >> +						&& compact_should_abort(cc))
> > 
> > This seems rather gratuitously inefficient and isn't terribly clear.
> > What's wrong with
> > 
> > 	if ((++foo % SWAP_CLUSTER_MAX) == 0 && compact_should_abort(cc))
> 
> It's a new variable and it differs from how isolate_migratepages_range() does this.
> But yeah, I might change it later there as well. There it makes even more sense.
> E.g. when skipping whole pageblock there, pfn % SWAP_CLUSTER_MAX will be always zero
> so the periodicity varies.
>  
> > ?
> > 
> > (Assumes that SWAP_CLUSTER_MAX is power-of-2 and that the compiler will
> > use &)
>  
> I hoped that compiler would be smart enough about SWAP_CLUSTER_MAX * pageblock_nr_pages
> as well, as those are constants and also power-of-2. But I didn't check the assembly.

Always check the assembly!  Just a quick `size mm/compaction.o' is
enough tell if you're on the right track.

> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -779,6 +779,7 @@ static void isolate_freepages(struct zone *zone,
>  	unsigned long block_start_pfn;	/* start of current pageblock */
>  	unsigned long block_end_pfn;	/* end of current pageblock */
>  	unsigned long low_pfn;	     /* lowest pfn scanner is able to scan */
> +	unsigned long nr_blocks_scanned = 0; /* for periodical abort checks */
>  	int nr_freepages = cc->nr_freepages;
>  	struct list_head *freelist = &cc->freepages;
>  
> @@ -813,7 +814,7 @@ static void isolate_freepages(struct zone *zone,
>  		 * suitable migration targets, so periodically check if we need
>  		 * to schedule, or even abort async compaction.
>  		 */
> -		if (!(block_start_pfn % (SWAP_CLUSTER_MAX * pageblock_nr_pages))
> +		if ((++nr_blocks_scanned % SWAP_CLUSTER_MAX) == 0
>  						&& compact_should_abort(cc))
>  			break;

This change actually makes the code worse, and the .o file gets larger.

For some stupid reason we went and make pageblock_nr_pages all lower
case but surprise surprise, it's actually a literal constant.  So the
compiler does the multiplication at compile time and converts the
modulus operation into a bitwise AND.  Duh.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/