linux-kernel - Re: [PATCH 2/2] f2fs: Enhance the subsequent logic of valid_thresh_ratio to prevent unnecessary background GC

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: <b7b20efb-8cdc-4cf0-a057-d4d41ae66aba@kernel.org>
Date: Thu, 18 Sep 2025 10:16:52 +0800
From: Chao Yu <chao@...nel.org>
To: Liao Yuanhong <liaoyuanhong@...o.com>, Jaegeuk Kim <jaegeuk@...nel.org>,
 "open list:F2FS FILE SYSTEM" <linux-f2fs-devel@...ts.sourceforge.net>,
 open list <linux-kernel@...r.kernel.org>
Cc: chao@...nel.org
Subject: Re: [PATCH 2/2] f2fs: Enhance the subsequent logic of
 valid_thresh_ratio to prevent unnecessary background GC

On 9/17/25 16:13, Liao Yuanhong wrote:
> 
> On 9/17/2025 3:57 PM, Chao Yu wrote:
>> On 9/17/25 15:08, Liao Yuanhong wrote:
>>> On 9/15/2025 4:36 PM, Chao Yu wrote:
>>>> On 9/9/25 21:44, Liao Yuanhong wrote:
>>>>> When the proportion of dirty segments within a section exceeds the
>>>>> valid_thresh_ratio, the gc_cost of that section is set to UINT_MAX,
>>>>> indicating that these sections should not be released. However, if all
>>>>> section costs within the scanning range of get_victim() are UINT_MAX,
>>>>> background GC will still occur. Add a condition to prevent this situation.
>>>> For this case, f2fs_get_victim() will return 0, and f2fs_gc() will use unchanged
>>>> segno for GC?
>>>>
>>>> Thanks,
>>> You're right, segno won't update in this scenario, and this patch feature is redundant.
>> Oh, I meant, if f2fs_get_victim() fails to select a valid victim due to the reason you
>> described, f2fs_get_victim() will return 0, and f2fs_gc() will migrate segment #NULL_SEGNO?
>> Or am I missing something?
>>
>> Thanks,
> 
> Yes. In this scenario, since it won't enter the|p.min_cost > cost|condition,|p.min_segno|will retain its initial value|||NULL_SEGNO|. This is consistent with what you described.

Do you have a script to reproduce this bug?

Thanks,

> 
> 
> Thanks,
> 
> Liao
> 
>>>
>>> Thanks,
>>>
>>> Liao
>>>
>>>>> Signed-off-by: Liao Yuanhong <liaoyuanhong@...o.com>
>>>>> ---
>>>>>    fs/f2fs/gc.c | 5 +++++
>>>>>    1 file changed, 5 insertions(+)
>>>>>
>>>>> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
>>>>> index 4a8c08f970e3..ffc3188416f4 100644
>>>>> --- a/fs/f2fs/gc.c
>>>>> +++ b/fs/f2fs/gc.c
>>>>> @@ -936,6 +936,11 @@ int f2fs_get_victim(struct f2fs_sb_info *sbi, unsigned int *result,
>>>>>            }
>>>>>        }
>>>>>    +    if (f2fs_sb_has_blkzoned(sbi) && p.min_cost == UINT_MAX) {
>>>>> +        ret = -ENODATA;
>>>>> +        goto out;
>>>>> +    }
>>>>> +
>>>>>        /* get victim for GC_AT/AT_SSR */
>>>>>        if (is_atgc) {
>>>>>            lookup_victim_by_age(sbi, &p);