linux-kernel - Re: [f2fs-dev] [PATCH] f2fs: fix to prevent clearing immutable for large folio supported inode

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: <77f0a77f-532c-40f2-97c9-df1a064e4a89@kernel.org>
Date: Fri, 23 Jan 2026 12:04:04 +0800
From: Chao Yu <chao@...nel.org>
To: Nanzhe Zhao <nzzhao@....com>, Barry Song <21cnbao@...il.com>
Cc: chao@...nel.org, jaegeuk@...nel.org, linux-kernel@...r.kernel.org,
 linux-f2fs-devel@...ts.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH] f2fs: fix to prevent clearing immutable for
 large folio supported inode

On 1/22/2026 10:02 AM, Nanzhe Zhao wrote:
> Hi all:
> 
> 在 2026/1/9 16:44, Chao Yu via Linux-f2fs-devel 写道:
> 
>> w/ above locks, it seems there is still a race condition as below:
>>
>> f2fs_fileattr_set                read
>>    - f2fs_setflags_common
>>     - truncate_pagecache
>>                           - f2fs_read_data_large_folio
>>                           : read large folios
>>     - mapping_set_folio_order_range
>>
>> Thanks,
> 
> I noticed that the comment for `mapping_set_folio_order_range()` says:
> 
>> "Context: This should not be called while the inode is active as it is non-atomic."
> 
> I'm not sure whether cases like f2fs_fileattr_set falls into the inode
> active case.
> 
>>>
>>>
>>> If clearing immutable is indeed rare, we may leave this as is, since
>>> writes are not supported until the page cache is fully dropped.
>>> Eventually, we will support large folios on non-immutable files.
>>>
>>> Thanks
>>> Barry
> 
> Also, I couldn’t find any API in `include/linux/pagemap.h` that can
> atomically set the
> folio order range. However, I think dynamically setting large folio
> order while an inode
> is active is a reasonable requirement. In my non immutable file large
> folio write work, I need to enable large folio support for a file when
> it is converted from an inline file to a non-inline file. The lack of an
> atomic API seems to be a blocker for moving large folio write forward.
> Any idea on that?

IIUC, our concern here is to avoid "echo 3 > drop_caches", right?

If so, can we try this in new ioctl?
- kern_path(target_file_path)
- d_invalidate(path->dentry)
- path_put

> 
> Thanks,
> Nanzhe Zhao
> 
>