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;
Subject: Re: [PATCH RFC] f2fs: fix to avoid selecting full segment w/ {AT,}SSR
 allocator
To:     Jaegeuk Kim <jaegeuk@kernel.org>
CC:     <linux-f2fs-devel@lists.sourceforge.net>,
        <linux-kernel@vger.kernel.org>, <chao@kernel.org>
References: <20210220094052.64905-1-yuchao0@huawei.com>
 <YFOLNGo+/8sKQ7si@google.com>
From:   Chao Yu <yuchao0@huawei.com>
Message-ID: <ec5cda53-d3f4-450c-7567-7bfc68e224f9@huawei.com>
Date:   Fri, 19 Mar 2021 10:31:09 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:52.0) Gecko/20100101
 Thunderbird/52.9.1
MIME-Version: 1.0
In-Reply-To: <YFOLNGo+/8sKQ7si@google.com>
Content-Type: text/plain; charset="windows-1252"; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2021/3/19 1:17, Jaegeuk Kim wrote:
> On 02/20, Chao Yu wrote:
>> In cp disabling mode, there could be a condition
>> - target segment has 128 ckpt valid blocks
>> - GC migrates 128 valid blocks to other segment (segment is still in
>> dirty list)
>> - GC migrates 384 blocks to target segment (segment has 128 cp_vblocks
>> and 384 vblocks)
>> - If GC selects target segment via {AT,}SSR allocator, however there is
>> no free space in targe segment.
>>
>> Fixes: 4354994f097d ("f2fs: checkpoint disabling")
>> Fixes: 093749e296e2 ("f2fs: support age threshold based garbage collection")
>> Signed-off-by: Chao Yu <yuchao0@huawei.com>
>> ---
>>   fs/f2fs/f2fs.h    |  1 +
>>   fs/f2fs/gc.c      | 17 +++++++++++++----
>>   fs/f2fs/segment.c | 20 ++++++++++++++++++++
>>   3 files changed, 34 insertions(+), 4 deletions(-)
>>
>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
>> index ed7807103c8e..9c753eff0814 100644
>> --- a/fs/f2fs/f2fs.h
>> +++ b/fs/f2fs/f2fs.h
>> @@ -3376,6 +3376,7 @@ block_t f2fs_get_unusable_blocks(struct f2fs_sb_info *sbi);
>>   int f2fs_disable_cp_again(struct f2fs_sb_info *sbi, block_t unusable);
>>   void f2fs_release_discard_addrs(struct f2fs_sb_info *sbi);
>>   int f2fs_npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra);
>> +bool segment_has_free_slot(struct f2fs_sb_info *sbi, int segno);
>>   void f2fs_init_inmem_curseg(struct f2fs_sb_info *sbi);
>>   void f2fs_save_inmem_curseg(struct f2fs_sb_info *sbi);
>>   void f2fs_restore_inmem_curseg(struct f2fs_sb_info *sbi);
>> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
>> index 86ba8ed0b8a7..a1d8062cdace 100644
>> --- a/fs/f2fs/gc.c
>> +++ b/fs/f2fs/gc.c
>> @@ -392,10 +392,6 @@ static void add_victim_entry(struct f2fs_sb_info *sbi,
>>   		if (p->gc_mode == GC_AT &&
>>   			get_valid_blocks(sbi, segno, true) == 0)
>>   			return;
>> -
>> -		if (p->alloc_mode == AT_SSR &&
>> -			get_seg_entry(sbi, segno)->ckpt_valid_blocks == 0)
>> -			return;
>>   	}
>>   
>>   	for (i = 0; i < sbi->segs_per_sec; i++)
>> @@ -736,6 +732,19 @@ static int get_victim_by_default(struct f2fs_sb_info *sbi,
>>   		if (gc_type == BG_GC && test_bit(secno, dirty_i->victim_secmap))
>>   			goto next;
>>   
>> +		if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
>> +			/*
>> +			 * to avoid selecting candidate which has below valid
>> +			 * block distribution:
>> +			 * partial blocks are valid and all left ones are valid
>> +			 * in previous checkpoint.
>> +			 */
>> +			if (p.alloc_mode == SSR || p.alloc_mode == AT_SSR) {
>> +				if (!segment_has_free_slot(sbi, segno))
>> +					goto next;
> 
> Do we need to change this to check free_slot instead of get_ckpt_valid_blocks()?

Jaegeuk,

LFS was assigned only for GC case, in this case we are trying to select source
section, rather than target segment for SSR/AT_SSR case, so we don't need to
check free_slot.

- f2fs_gc
  - __get_victim
   - get_victim(sbi, victim, gc_type, NO_CHECK_TYPE, LFS, 0);

> 
>   732                 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
>   733                                         get_ckpt_valid_blocks(sbi, segno) &&
>   734                                         p.alloc_mode == LFS))

BTW, in LFS mode, GC wants to find source section rather than segment, so we
should change to check valid ckpt blocks in every segment of targe section here?

Thanks,

> 
> 
>> +			}
>> +		}
>> +
>>   		if (is_atgc) {
>>   			add_victim_entry(sbi, &p, segno);
>>   			goto next;
>> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
>> index 2d5a82c4ca15..deaf57e13125 100644
>> --- a/fs/f2fs/segment.c
>> +++ b/fs/f2fs/segment.c
>> @@ -2650,6 +2650,26 @@ static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
>>   		seg->next_blkoff++;
>>   }
>>   
>> +bool segment_has_free_slot(struct f2fs_sb_info *sbi, int segno)
>> +{
>> +	struct sit_info *sit = SIT_I(sbi);
>> +	struct seg_entry *se = get_seg_entry(sbi, segno);
>> +	int entries = SIT_VBLOCK_MAP_SIZE / sizeof(unsigned long);
>> +	unsigned long *target_map = SIT_I(sbi)->tmp_map;
>> +	unsigned long *ckpt_map = (unsigned long *)se->ckpt_valid_map;
>> +	unsigned long *cur_map = (unsigned long *)se->cur_valid_map;
>> +	int i, pos;
>> +
>> +	down_write(&sit->sentry_lock);
>> +	for (i = 0; i < entries; i++)
>> +		target_map[i] = ckpt_map[i] | cur_map[i];
>> +
>> +	pos = __find_rev_next_zero_bit(target_map, sbi->blocks_per_seg, 0);
>> +	up_write(&sit->sentry_lock);
>> +
>> +	return pos < sbi->blocks_per_seg;
>> +}
>> +
>>   /*
>>    * This function always allocates a used segment(from dirty seglist) by SSR
>>    * manner, so it should recover the existing segment information of valid blocks
>> -- 
>> 2.29.2
> .
>