Re: [RFC 12/13] mm, compaction: more reliably increase direct compaction priority

From	Vlastimil Babka <vbabka@suse.cz>
Newsgroups	linux.kernel
Subject	Re: [RFC 12/13] mm, compaction: more reliably increase direct compaction priority
Date	2016-05-31 14:30 +0200
Message-ID	<rEQY3-2DM-57@gated-at.bofh.it> (permalink)
References	<rxaqR-6fU-5@gated-at.bofh.it> <rxaqS-6fU-31@gated-at.bofh.it> <rELvk-7hB-57@gated-at.bofh.it> <rEQEH-2x9-47@gated-at.bofh.it>
Organization	linux.* mail to news gateway

Show all headers | View raw

On 05/31/2016 02:07 PM, Vlastimil Babka wrote:
> On 05/31/2016 08:37 AM, Joonsoo Kim wrote:
>>> @@ -3695,22 +3695,22 @@ __alloc_pages_slowpath(gfp_t gfp_mask, unsigned int order,
>>>  	else
>>>  		no_progress_loops++;
>>>
>>> -	if (should_reclaim_retry(gfp_mask, order, ac, alloc_flags,
>>> -				 did_some_progress > 0, no_progress_loops))
>>> -		goto retry;
>>> -
>>> +	should_retry = should_reclaim_retry(gfp_mask, order, ac, alloc_flags,
>>> +				 did_some_progress > 0, no_progress_loops);
>>>  	/*
>>>  	 * It doesn't make any sense to retry for the compaction if the order-0
>>>  	 * reclaim is not able to make any progress because the current
>>>  	 * implementation of the compaction depends on the sufficient amount
>>>  	 * of free memory (see __compaction_suitable)
>>>  	 */
>>> -	if (did_some_progress > 0 &&
>>> -			should_compact_retry(ac, order, alloc_flags,
>>> +	if (did_some_progress > 0)
>>> +		should_retry |= should_compact_retry(ac, order, alloc_flags,
>>>  				compact_result, &compact_priority,
>>> -				compaction_retries))
>>> +				compaction_retries);
>>> +	if (should_retry)
>>>  		goto retry;
>>
>> Hmm... it looks odd that we check should_compact_retry() when
>> did_some_progress > 0. If system is full of anonymous memory and we
>> don't have swap, we can't reclaim anything but we can compact.
>
> Right, thanks.

Hmm on the other hand, should_compact_retry will assume (in 
compaction_zonelist_suitable()) that reclaimable memory is actually 
reclaimable. If there's nothing to tell us that it actually isn't, if we 
drop the reclaim progress requirement. That's risking an infinite loop?

Thread

Re: [RFC 12/13] mm, compaction: more reliably increase direct  compaction priority Joonsoo Kim <iamjoonsoo.kim@lge.com> - 2016-05-31 08:40 +0200
  Re: [RFC 12/13] mm, compaction: more reliably increase direct  compaction priority Vlastimil Babka <vbabka@suse.cz> - 2016-05-31 14:10 +0200
    Re: [RFC 12/13] mm, compaction: more reliably increase direct  compaction priority Vlastimil Babka <vbabka@suse.cz> - 2016-05-31 14:30 +0200
      Re: [RFC 12/13] mm, compaction: more reliably increase direct  compaction priority Joonsoo Kim <iamjoonsoo.kim@lge.com> - 2016-06-02 04:50 +0200

csiph-web