Message-ID: <20250430143714.GA2020@cmpxchg.org>
Date: Wed, 30 Apr 2025 10:37:14 -0400
From: Johannes Weiner <hannes@...xchg.org>
To: Muchun Song <songmuchun@...edance.com>
Cc: mhocko@...nel.org, roman.gushchin@...ux.dev, shakeel.butt@...ux.dev,
	muchun.song@...ux.dev, akpm@...ux-foundation.org,
	david@...morbit.com, zhengqi.arch@...edance.com,
	yosry.ahmed@...ux.dev, nphamcs@...il.com, chengming.zhou@...ux.dev,
	linux-kernel@...r.kernel.org, cgroups@...r.kernel.org,
	linux-mm@...ck.org, hamzamahfooz@...ux.microsoft.com,
	apais@...ux.microsoft.com, Hugh Dickins <hughd@...gle.com>
Subject: Re: [PATCH RFC 07/28] mm: thp: use folio_batch to handle THP
 splitting in deferred_split_scan()

On Tue, Apr 15, 2025 at 10:45:11AM +0800, Muchun Song wrote:
> The maintenance of the folio->_deferred_list is intricate because it's
> reused in a local list.
> 
> Here are some peculiarities:
> 
>    1) When a folio is removed from its split queue and added to a local
>       on-stack list in deferred_split_scan(), the ->split_queue_len isn't
>       updated, leading to an inconsistency between it and the actual
>       number of folios in the split queue.
> 
>    2) When the folio is split via split_folio() later, it's removed from
>       the local list while holding the split queue lock. At this time,
>       this lock protects the local list, not the split queue.
> 
>    3) To handle the race condition with a third-party freeing or migrating
>       the preceding folio, we must ensure there's always one safe (with
>       raised refcount) folio before by delaying its folio_put(). More
>       details can be found in commit e66f3185fa04. It's rather tricky.
> 
> We can use the folio_batch infrastructure to handle this clearly. In this
> case, ->split_queue_len will be consistent with the real number of folios
> in the split queue. If list_empty(&folio->_deferred_list) returns false,
> it's clear the folio must be in its split queue (not in a local list
> anymore).
> 
> In the future, we aim to reparent LRU folios during memcg offline to
> eliminate dying memory cgroups. This patch prepares for using
> folio_split_queue_lock_irqsave() as folio memcg may change then.
> 
> Signed-off-by: Muchun Song <songmuchun@...edance.com>

This is a very nice simplification. And getting rid of the stack list
and its subtle implication on all the various current and future
list_empty(&folio->_deferred_list) checks should be much more robust.

However, I think there is one snag related to this:

> ---
>  mm/huge_memory.c | 69 +++++++++++++++++++++---------------------------
>  1 file changed, 30 insertions(+), 39 deletions(-)
> 
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 70820fa75c1f..d2bc943a40e8 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -4220,40 +4220,47 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
>  	struct pglist_data *pgdata = NODE_DATA(sc->nid);
>  	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>  	unsigned long flags;
> -	LIST_HEAD(list);
> -	struct folio *folio, *next, *prev = NULL;
> -	int split = 0, removed = 0;
> +	struct folio *folio, *next;
> +	int split = 0, i;
> +	struct folio_batch fbatch;
> +	bool done;
>  
>  #ifdef CONFIG_MEMCG
>  	if (sc->memcg)
>  		ds_queue = &sc->memcg->deferred_split_queue;
>  #endif
> -
> +	folio_batch_init(&fbatch);
> +retry:
> +	done = true;
>  	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>  	/* Take pin on all head pages to avoid freeing them under us */
>  	list_for_each_entry_safe(folio, next, &ds_queue->split_queue,
>  							_deferred_list) {
>  		if (folio_try_get(folio)) {
> -			list_move(&folio->_deferred_list, &list);
> -		} else {
> +			folio_batch_add(&fbatch, folio);
> +		} else if (folio_test_partially_mapped(folio)) {
>  			/* We lost race with folio_put() */
> -			if (folio_test_partially_mapped(folio)) {
> -				folio_clear_partially_mapped(folio);
> -				mod_mthp_stat(folio_order(folio),
> -					      MTHP_STAT_NR_ANON_PARTIALLY_MAPPED, -1);
> -			}
> -			list_del_init(&folio->_deferred_list);
> -			ds_queue->split_queue_len--;
> +			folio_clear_partially_mapped(folio);
> +			mod_mthp_stat(folio_order(folio),
> +				      MTHP_STAT_NR_ANON_PARTIALLY_MAPPED, -1);
>  		}
> +		list_del_init(&folio->_deferred_list);
> +		ds_queue->split_queue_len--;
>  		if (!--sc->nr_to_scan)
>  			break;
> +		if (folio_batch_space(&fbatch) == 0) {
> +			done = false;
> +			break;
> +		}
>  	}
>  	split_queue_unlock_irqrestore(ds_queue, flags);
>  
> -	list_for_each_entry_safe(folio, next, &list, _deferred_list) {
> +	for (i = 0; i < folio_batch_count(&fbatch); i++) {
>  		bool did_split = false;
>  		bool underused = false;
> +		struct deferred_split *fqueue;
>  
> +		folio = fbatch.folios[i];
>  		if (!folio_test_partially_mapped(folio)) {
>  			underused = thp_underused(folio);
>  			if (!underused)
> @@ -4269,39 +4276,23 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
>  		}
>  		folio_unlock(folio);
>  next:
> +		if (did_split || !folio_test_partially_mapped(folio))
> +			continue;

There IS a list_empty() check in the splitting code that we actually
relied on, for cleaning up the partially_mapped state and counter:

		    !list_empty(&folio->_deferred_list)) {
			ds_queue->split_queue_len--;
			if (folio_test_partially_mapped(folio)) {
				folio_clear_partially_mapped(folio);
				mod_mthp_stat(folio_order(folio),
					      MTHP_STAT_NR_ANON_PARTIALLY_MAPPED, -1);
			}
			/*
			 * Reinitialize page_deferred_list after removing the
			 * page from the split_queue, otherwise a subsequent
			 * split will see list corruption when checking the
			 * page_deferred_list.
			 */
			list_del_init(&folio->_deferred_list);

With the folios isolated up front, it looks like you need to handle
this from the shrinker.

Otherwise this looks correct to me. But this code is subtle, I would
feel much better if Hugh (CC-ed) could take a look as well.

Thanks!

>  		/*
> -		 * split_folio() removes folio from list on success.
>  		 * Only add back to the queue if folio is partially mapped.
>  		 * If thp_underused returns false, or if split_folio fails
>  		 * in the case it was underused, then consider it used and
>  		 * don't add it back to split_queue.
>  		 */
> -		if (did_split) {
> -			; /* folio already removed from list */
> -		} else if (!folio_test_partially_mapped(folio)) {
> -			list_del_init(&folio->_deferred_list);
> -			removed++;
> -		} else {
> -			/*
> -			 * That unlocked list_del_init() above would be unsafe,
> -			 * unless its folio is separated from any earlier folios
> -			 * left on the list (which may be concurrently unqueued)
> -			 * by one safe folio with refcount still raised.
> -			 */
> -			swap(folio, prev);
> -		}
> -		if (folio)
> -			folio_put(folio);
> +		fqueue = folio_split_queue_lock_irqsave(folio, &flags);
> +		list_add_tail(&folio->_deferred_list, &fqueue->split_queue);
> +		fqueue->split_queue_len++;
> +		split_queue_unlock_irqrestore(fqueue, flags);
>  	}
> +	folios_put(&fbatch);
>  
> -	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
> -	list_splice_tail(&list, &ds_queue->split_queue);
> -	ds_queue->split_queue_len -= removed;
> -	split_queue_unlock_irqrestore(ds_queue, flags);
> -
> -	if (prev)
> -		folio_put(prev);
> -
> +	if (!done)
> +		goto retry;
>  	/*
>  	 * Stop shrinker if we didn't split any page, but the queue is empty.
>  	 * This can happen if pages were freed under us.
> -- 
> 2.20.1

Powered by blists - more mailing lists