Re: [f2fs-dev] [PATCH 08/10] f2fs: relax async discard commands more

[Chao Yu]

On 2017/1/6 10:42, Jaegeuk Kim wrote:
> Hi Chao,
> 
> On 01/06, Chao Yu wrote:
>> On 2017/1/6 3:46, Jaegeuk Kim wrote:
>>> On 01/05, Chao Yu wrote:
>>>> On 2017/1/4 17:29, Chao Yu wrote:
>>>>> On 2016/12/31 2:51, Jaegeuk Kim wrote:
>>>>>> This patch relaxes async discard commands to avoid waiting its end_io during
>>>>>> checkpoint.
>>>>>> Instead of waiting them during checkpoint, it will be done when actually reusing
>>>>>> them.
>>>>>>
>>>>>> Test on initial partition of nvme drive.
>>>>>>
>>>>>>  # time fstrim /mnt/test
>>>>>>
>>>>>> Before : 6.158s
>>>>>> After : 4.822s
>>>>>>
>>>>>> Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
>>>>>
>>>>> Reviewed-by: Chao Yu <yuchao0@xxxxxxxxxx>
>>>>>
>>>>> One comment below,
>>>>
>>>> I still have a comment on this patch.
>>>>
>>>>>> -void f2fs_wait_all_discard_bio(struct f2fs_sb_info *sbi)
>>>>>> +/* This should be covered by global mutex, &sit_i->sentry_lock */
>>>>>> +void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, unsigned int segno)
>>>>>>  {
>>>>>>  	struct list_head *wait_list = &(SM_I(sbi)->wait_list);
>>>>>>  	struct bio_entry *be, *tmp;
>>>>>> @@ -646,7 +650,15 @@ void f2fs_wait_all_discard_bio(struct f2fs_sb_info *sbi)
>>>>>>  		struct bio *bio = be->bio;
>>>>>>  		int err;
>>>>>>  
>>>>>> -		wait_for_completion_io(&be->event);
>>>>>> +		if (!completion_done(&be->event)) {
>>>>>> +			if ((be->start_segno >= segno &&
>>>>>> +					be->end_segno <= segno) ||
>>>>>
>>>>> segno >= be->start_segno && segno < be->end_segno ?
>>
>> Still can not understand this judgment condition, we should wait completion of
>> discard command only when segno is locate in range of [start_segno, end_segno]?
>>
>> But now, this condition can be true only when segno, start_segno, end_segno have
>> equal value.
>>
>> Please correct me if I'm wrong.
> 
> Urg. I rewrote it to use block addresses.
> 
> How about this?
> 
>>From fe24461eedd62815e0c56317f010a3a6e3004434 Mon Sep 17 00:00:00 2001
> From: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
> Date: Thu, 29 Dec 2016 14:07:53 -0800
> Subject: [PATCH] f2fs: relax async discard commands more
> 
> This patch relaxes async discard commands to avoid waiting its end_io during
> checkpoint.
> Instead of waiting them during checkpoint, it will be done when actually reusing
> them.
> 
> Test on initial partition of nvme drive.
> 
>  # time fstrim /mnt/test
> 
> Before : 6.158s
> After : 4.822s

We should wait for completion of all discard IOs in fstrim, since we can't
guarantee discard command can be handled successfully by device after we return
success to caller. Right?

Thanks,

> 
> Reviewed-by: Chao Yu <yuchao0@xxxxxxxxxx>
> Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
> ---
>  fs/f2fs/checkpoint.c |  7 ++-----
>  fs/f2fs/f2fs.h       |  4 +++-
>  fs/f2fs/segment.c    | 23 +++++++++++++++++++----
>  fs/f2fs/super.c      |  3 +++
>  4 files changed, 27 insertions(+), 10 deletions(-)
> 
> diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
> index f73ee9534d83..1a9ba69a22ba 100644
> --- a/fs/f2fs/checkpoint.c
> +++ b/fs/f2fs/checkpoint.c
> @@ -1254,7 +1254,6 @@ int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>  		f2fs_bug_on(sbi, prefree_segments(sbi));
>  		flush_sit_entries(sbi, cpc);
>  		clear_prefree_segments(sbi, cpc);
> -		f2fs_wait_all_discard_bio(sbi);
>  		unblock_operations(sbi);
>  		goto out;
>  	}
> @@ -1273,12 +1272,10 @@ int write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>  
>  	/* unlock all the fs_lock[] in do_checkpoint() */
>  	err = do_checkpoint(sbi, cpc);
> -	if (err) {
> +	if (err)
>  		release_discard_addrs(sbi);
> -	} else {
> +	else
>  		clear_prefree_segments(sbi, cpc);
> -		f2fs_wait_all_discard_bio(sbi);
> -	}
>  
>  	unblock_operations(sbi);
>  	stat_inc_cp_count(sbi->stat_info);
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index bdcfe2a9b532..e0db895fd84c 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -183,6 +183,8 @@ struct discard_entry {
>  
>  struct bio_entry {
>  	struct list_head list;
> +	block_t lstart;
> +	block_t len;
>  	struct bio *bio;
>  	struct completion event;
>  	int error;
> @@ -2111,7 +2113,7 @@ void destroy_flush_cmd_control(struct f2fs_sb_info *, bool);
>  void invalidate_blocks(struct f2fs_sb_info *, block_t);
>  bool is_checkpointed_data(struct f2fs_sb_info *, block_t);
>  void refresh_sit_entry(struct f2fs_sb_info *, block_t, block_t);
> -void f2fs_wait_all_discard_bio(struct f2fs_sb_info *);
> +void f2fs_wait_discard_bio(struct f2fs_sb_info *, block_t);
>  void clear_prefree_segments(struct f2fs_sb_info *, struct cp_control *);
>  void release_discard_addrs(struct f2fs_sb_info *);
>  int npages_for_summary_flush(struct f2fs_sb_info *, bool);
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index 485b9118b418..672bcdefe836 100644
> --- a/fs/f2fs/segment.c
> +++ b/fs/f2fs/segment.c
> @@ -625,20 +625,23 @@ static void locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno)
>  }
>  
>  static struct bio_entry *__add_bio_entry(struct f2fs_sb_info *sbi,
> -							struct bio *bio)
> +			struct bio *bio, block_t lstart, block_t len)
>  {
>  	struct list_head *wait_list = &(SM_I(sbi)->wait_list);
>  	struct bio_entry *be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS);
>  
>  	INIT_LIST_HEAD(&be->list);
>  	be->bio = bio;
> +	be->lstart = lstart;
> +	be->len = len;
>  	init_completion(&be->event);
>  	list_add_tail(&be->list, wait_list);
>  
>  	return be;
>  }
>  
> -void f2fs_wait_all_discard_bio(struct f2fs_sb_info *sbi)
> +/* This should be covered by global mutex, &sit_i->sentry_lock */
> +void f2fs_wait_discard_bio(struct f2fs_sb_info *sbi, block_t blkaddr)
>  {
>  	struct list_head *wait_list = &(SM_I(sbi)->wait_list);
>  	struct bio_entry *be, *tmp;
> @@ -647,7 +650,15 @@ void f2fs_wait_all_discard_bio(struct f2fs_sb_info *sbi)
>  		struct bio *bio = be->bio;
>  		int err;
>  
> -		wait_for_completion_io(&be->event);
> +		if (!completion_done(&be->event)) {
> +			if ((be->lstart <= blkaddr &&
> +					blkaddr < be->lstart + be->len) ||
> +					blkaddr == NULL_ADDR)
> +				wait_for_completion_io(&be->event);
> +			else
> +				continue;
> +		}
> +
>  		err = be->error;
>  		if (err == -EOPNOTSUPP)
>  			err = 0;
> @@ -675,6 +686,7 @@ static int __f2fs_issue_discard_async(struct f2fs_sb_info *sbi,
>  		struct block_device *bdev, block_t blkstart, block_t blklen)
>  {
>  	struct bio *bio = NULL;
> +	block_t lblkstart = blkstart;
>  	int err;
>  
>  	trace_f2fs_issue_discard(sbi->sb, blkstart, blklen);
> @@ -689,7 +701,8 @@ static int __f2fs_issue_discard_async(struct f2fs_sb_info *sbi,
>  				SECTOR_FROM_BLOCK(blklen),
>  				GFP_NOFS, 0, &bio);
>  	if (!err && bio) {
> -		struct bio_entry *be = __add_bio_entry(sbi, bio);
> +		struct bio_entry *be = __add_bio_entry(sbi, bio,
> +						lblkstart, blklen);
>  
>  		bio->bi_private = be;
>  		bio->bi_end_io = f2fs_submit_bio_wait_endio;
> @@ -1575,6 +1588,8 @@ void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
>  
>  	*new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
>  
> +	f2fs_wait_discard_bio(sbi, *new_blkaddr);
> +
>  	/*
>  	 * __add_sum_entry should be resided under the curseg_mutex
>  	 * because, this function updates a summary entry in the
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index f3697f97e527..461c29043aec 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -770,6 +770,9 @@ static void f2fs_put_super(struct super_block *sb)
>  		write_checkpoint(sbi, &cpc);
>  	}
>  
> +	/* be sure to wait for any on-going discard commands */
> +	f2fs_wait_discard_bio(sbi, NULL_ADDR);
> +
>  	/* write_checkpoint can update stat informaion */
>  	f2fs_destroy_stats(sbi);
>  
>