Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 15 Jul 2020 10:28:12 +0200
From:   Michal Hocko <mhocko@kernel.org>
To:     js1304@gmail.com
Cc:     Andrew Morton <akpm@linux-foundation.org>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, kernel-team@lge.com,
        Vlastimil Babka <vbabka@suse.cz>,
        Christoph Hellwig <hch@infradead.org>,
        Roman Gushchin <guro@fb.com>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>,
        "Aneesh Kumar K . V" <aneesh.kumar@linux.ibm.com>,
        Joonsoo Kim <iamjoonsoo.kim@lge.com>, stable@vger.kernel.org
Subject: Re: [PATCH 1/4] mm/page_alloc: fix non cma alloc context
Message-ID: <20200715082812.GD5451@dhcp22.suse.cz>
References: <1594789529-6206-1-git-send-email-iamjoonsoo.kim@lge.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1594789529-6206-1-git-send-email-iamjoonsoo.kim@lge.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed 15-07-20 14:05:26, Joonsoo Kim wrote:
> From: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> 
> Currently, preventing cma area in page allocation is implemented by using
> current_gfp_context(). However, there are two problems of this
> implementation.
> 
> First, this doesn't work for allocation fastpath. In the fastpath,
> original gfp_mask is used since current_gfp_context() is introduced in
> order to control reclaim and it is on slowpath.
> Second, clearing __GFP_MOVABLE has a side effect to exclude the memory
> on the ZONE_MOVABLE for allocation target.

This can be especially a problem with movable_node configurations where
a large portion of the memory is in movable zones.

> To fix these problems, this patch changes the implementation to exclude
> cma area in page allocation. Main point of this change is using the
> alloc_flags. alloc_flags is mainly used to control allocation so it fits
> for excluding cma area in allocation.

The approach is sensible and the patch makes sense to me from a quick
glance but I am not really familiar with all subtle details about cma
integration with the allocator so I do not feel confident to provide my
ack.

Thanks!

> Fixes: d7fefcc (mm/cma: add PF flag to force non cma alloc)
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> ---
>  include/linux/sched/mm.h |  4 ----
>  mm/page_alloc.c          | 27 +++++++++++++++------------
>  2 files changed, 15 insertions(+), 16 deletions(-)
> 
> diff --git a/include/linux/sched/mm.h b/include/linux/sched/mm.h
> index 44ad5b7..a73847a 100644
> --- a/include/linux/sched/mm.h
> +++ b/include/linux/sched/mm.h
> @@ -191,10 +191,6 @@ static inline gfp_t current_gfp_context(gfp_t flags)
>  			flags &= ~(__GFP_IO | __GFP_FS);
>  		else if (pflags & PF_MEMALLOC_NOFS)
>  			flags &= ~__GFP_FS;
> -#ifdef CONFIG_CMA
> -		if (pflags & PF_MEMALLOC_NOCMA)
> -			flags &= ~__GFP_MOVABLE;
> -#endif
>  	}
>  	return flags;
>  }
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 6416d08..cd53894 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -2791,7 +2791,7 @@ __rmqueue(struct zone *zone, unsigned int order, int migratetype,
>  	 * allocating from CMA when over half of the zone's free memory
>  	 * is in the CMA area.
>  	 */
> -	if (migratetype == MIGRATE_MOVABLE &&
> +	if (alloc_flags & ALLOC_CMA &&
>  	    zone_page_state(zone, NR_FREE_CMA_PAGES) >
>  	    zone_page_state(zone, NR_FREE_PAGES) / 2) {
>  		page = __rmqueue_cma_fallback(zone, order);
> @@ -2802,7 +2802,7 @@ __rmqueue(struct zone *zone, unsigned int order, int migratetype,
>  retry:
>  	page = __rmqueue_smallest(zone, order, migratetype);
>  	if (unlikely(!page)) {
> -		if (migratetype == MIGRATE_MOVABLE)
> +		if (alloc_flags & ALLOC_CMA)
>  			page = __rmqueue_cma_fallback(zone, order);
>  
>  		if (!page && __rmqueue_fallback(zone, order, migratetype,
> @@ -3502,11 +3502,9 @@ static inline long __zone_watermark_unusable_free(struct zone *z,
>  	if (likely(!alloc_harder))
>  		unusable_free += z->nr_reserved_highatomic;
>  
> -#ifdef CONFIG_CMA
>  	/* If allocation can't use CMA areas don't use free CMA pages */
> -	if (!(alloc_flags & ALLOC_CMA))
> +	if (IS_ENABLED(CONFIG_CMA) && !(alloc_flags & ALLOC_CMA))
>  		unusable_free += zone_page_state(z, NR_FREE_CMA_PAGES);
> -#endif
>  
>  	return unusable_free;
>  }
> @@ -3693,6 +3691,16 @@ alloc_flags_nofragment(struct zone *zone, gfp_t gfp_mask)
>  	return alloc_flags;
>  }
>  
> +static inline void current_alloc_flags(gfp_t gfp_mask,
> +				unsigned int *alloc_flags)
> +{
> +	unsigned int pflags = READ_ONCE(current->flags);
> +
> +	if (!(pflags & PF_MEMALLOC_NOCMA) &&
> +		gfp_migratetype(gfp_mask) == MIGRATE_MOVABLE)
> +		*alloc_flags |= ALLOC_CMA;
> +}
> +
>  /*
>   * get_page_from_freelist goes through the zonelist trying to allocate
>   * a page.
> @@ -3706,6 +3714,8 @@ get_page_from_freelist(gfp_t gfp_mask, unsigned int order, int alloc_flags,
>  	struct pglist_data *last_pgdat_dirty_limit = NULL;
>  	bool no_fallback;
>  
> +	current_alloc_flags(gfp_mask, &alloc_flags);
> +
>  retry:
>  	/*
>  	 * Scan zonelist, looking for a zone with enough free.
> @@ -4339,10 +4349,6 @@ gfp_to_alloc_flags(gfp_t gfp_mask)
>  	} else if (unlikely(rt_task(current)) && !in_interrupt())
>  		alloc_flags |= ALLOC_HARDER;
>  
> -#ifdef CONFIG_CMA
> -	if (gfp_migratetype(gfp_mask) == MIGRATE_MOVABLE)
> -		alloc_flags |= ALLOC_CMA;
> -#endif
>  	return alloc_flags;
>  }
>  
> @@ -4808,9 +4814,6 @@ static inline bool prepare_alloc_pages(gfp_t gfp_mask, unsigned int order,
>  	if (should_fail_alloc_page(gfp_mask, order))
>  		return false;
>  
> -	if (IS_ENABLED(CONFIG_CMA) && ac->migratetype == MIGRATE_MOVABLE)
> -		*alloc_flags |= ALLOC_CMA;
> -
>  	return true;
>  }
>  
> -- 
> 2.7.4

-- 
Michal Hocko
SUSE Labs