linux-kernel - Re: linux-next: manual merge of the akpm tree with the ext4 tree

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]
Message-ID: <20130619074402.GC1990@localhost.localdomain>
Date:	Wed, 19 Jun 2013 11:44:04 +0400
From:	Glauber Costa <glommer@...il.com>
To:	Stephen Rothwell <sfr@...b.auug.org.au>
Cc:	Andrew Morton <akpm@...ux-foundation.org>,
	linux-next@...r.kernel.org, linux-kernel@...r.kernel.org,
	Zheng Liu <wenqing.lz@...bao.com>,
	Theodore Ts'o <tytso@....edu>,
	Dave Chinner <dchinner@...hat.com>,
	Glauber Costa <glommer@...nvz.org>
Subject: Re: linux-next: manual merge of the akpm tree with the ext4 tree

On Wed, Jun 19, 2013 at 05:27:21PM +1000, Stephen Rothwell wrote:
> Hi Andrew,
> 
> Today's linux-next merge of the akpm tree got a conflict in
> fs/ext4/extents_status.c between commit 6480bad916be ("ext4: improve
> extent cache shrink mechanism to avoid to burn CPU time") from the ext
> tree and commit 1f42d0934b4e ("fs: convert fs shrinkers to new scan/count
> API") from the akpm tree.
> 
> I fixed it up (I am not sure if the result makes complete sense - see
> below) and can carry the fix as necessary (no action is required).
> 
> -- 
> Cheers,
> Stephen Rothwell                    sfr@...b.auug.org.au
> 
> diff --cc fs/ext4/extents_status.c
> index 80dcc59,4bce4f0..0000000
> --- a/fs/ext4/extents_status.c
> +++ b/fs/ext4/extents_status.c
> @@@ -876,58 -878,32 +876,63 @@@ int ext4_es_zeroout(struct inode *inode
>   				     EXTENT_STATUS_WRITTEN);
>   }
>   
>  +static int ext4_inode_touch_time_cmp(void *priv, struct list_head *a,
>  +				     struct list_head *b)
>  +{
>  +	struct ext4_inode_info *eia, *eib;
>  +	unsigned long diff;
>  +
>  +	eia = list_entry(a, struct ext4_inode_info, i_es_lru);
>  +	eib = list_entry(b, struct ext4_inode_info, i_es_lru);
>  +
>  +	diff = eia->i_touch_when - eib->i_touch_when;
>  +	if (diff < 0)
>  +		return -1;
>  +	if (diff > 0)
>  +		return 1;
>  +	return 0;
>  +}
>   
This comes straight from ext4, so no problem.

Now we just need to make sure that the shrinker is clearly separated
between a count part and a scan part.

> - static int ext4_es_shrink(struct shrinker *shrink, struct shrink_control *sc)
> + static long ext4_es_count(struct shrinker *shrink, struct shrink_control *sc)
> + {
> + 	long nr;
> + 	struct ext4_sb_info *sbi = container_of(shrink,
> + 					struct ext4_sb_info, s_es_shrinker);
> + 
> + 	nr = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
> + 	trace_ext4_es_shrink_enter(sbi->s_sb, sc->nr_to_scan, nr);
> + 	return nr;
> + }
The count seems okay.

> + 
> + static long ext4_es_scan(struct shrinker *shrink, struct shrink_control *sc)
>   {
>   	struct ext4_sb_info *sbi = container_of(shrink,
>   					struct ext4_sb_info, s_es_shrinker);
>   	struct ext4_inode_info *ei;
>  -	struct list_head *cur, *tmp, scanned;
>  +	struct list_head *cur, *tmp;
>  +	LIST_HEAD(skiped);
>   	int nr_to_scan = sc->nr_to_scan;
> - 	int ret, nr_shrunk = 0;
> - 
> - 	ret = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
> - 	trace_ext4_es_shrink_enter(sbi->s_sb, nr_to_scan, ret);
> - 
> - 	if (!nr_to_scan)
> - 		return ret;
> + 	int ret = 0, nr_shrunk = 0;
>   
>  -	INIT_LIST_HEAD(&scanned);
>  -
>   	spin_lock(&sbi->s_es_lru_lock);
>  +
>  +	/*
>  +	 * If the inode that is at the head of LRU list is newer than
>  +	 * last_sorted time, that means that we need to sort this list.
>  +	 */
>  +	ei = list_first_entry(&sbi->s_es_lru, struct ext4_inode_info, i_es_lru);
>  +	if (sbi->s_es_last_sorted < ei->i_touch_when) {
>  +		list_sort(NULL, &sbi->s_es_lru, ext4_inode_touch_time_cmp);
>  +		sbi->s_es_last_sorted = jiffies;
>  +	}
>  +
They are sorting the list, but that doesn't change the ammount of items they report.
So far it is fine. My only concern here would be locking, but it seems to be all
protected by the s_es_lru_lock.

>   	list_for_each_safe(cur, tmp, &sbi->s_es_lru) {
>  -		list_move_tail(cur, &scanned);
>  +		/*
>  +		 * If we have already reclaimed all extents from extent
>  +		 * status tree, just stop the loop immediately.
>  +		 */
>  +		if (percpu_counter_read_positive(&sbi->s_extent_cache_cnt) == 0)
>  +			break;
>   
>   		ei = list_entry(cur, struct ext4_inode_info, i_es_lru);
>   
> @@@ -951,22 -923,22 +956,22 @@@
>   		if (nr_to_scan == 0)
>   			break;
>   	}
>  -	list_splice_tail(&scanned, &sbi->s_es_lru);
>  +
>  +	/* Move the newer inodes into the tail of the LRU list. */
>  +	list_splice_tail(&skiped, &sbi->s_es_lru);
>   	spin_unlock(&sbi->s_es_lru_lock);
>   
> - 	ret = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
>   	trace_ext4_es_shrink_exit(sbi->s_sb, nr_shrunk, ret);
> - 	return ret;
> + 	return nr_shrunk;
>   }
>   
This part also seems fine.

>  -void ext4_es_register_shrinker(struct super_block *sb)
>  +void ext4_es_register_shrinker(struct ext4_sb_info *sbi)
>   {
>  -	struct ext4_sb_info *sbi;
>  -
>  -	sbi = EXT4_SB(sb);
>   	INIT_LIST_HEAD(&sbi->s_es_lru);
>   	spin_lock_init(&sbi->s_es_lru_lock);
>  +	sbi->s_es_last_sorted = 0;
> - 	sbi->s_es_shrinker.shrink = ext4_es_shrink;
> + 	sbi->s_es_shrinker.scan_objects = ext4_es_scan;
> + 	sbi->s_es_shrinker.count_objects = ext4_es_count;
>   	sbi->s_es_shrinker.seeks = DEFAULT_SEEKS;
>   	register_shrinker(&sbi->s_es_shrinker);
>   }

And so does this.

I believe the resolution is okay, at least from our PoV.

--
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/