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: <20240816175020.GB6039@frogsfrogsfrogs>
Date: Fri, 16 Aug 2024 10:50:20 -0700
From: "Darrick J. Wong" <djwong@...nel.org>
To: alexjlzheng@...il.com
Cc: yangerkun@...weicloud.com, adilger.kernel@...ger.ca,
	alexjlzheng@...cent.com, linux-ext4@...r.kernel.org,
	linux-kernel@...r.kernel.org, tytso@....edu
Subject: Re: [PATCH] ext4: disambiguate the return value of
 ext4_dio_write_end_io()

On Sat, Aug 17, 2024 at 12:57:27AM +0800, alexjlzheng@...il.com wrote:
> On Fri, 16 Aug 2024 20:21:22 +0800, yangerkun@...weicloud.com wrote:
> > 在 2024/8/15 19:27, alexjlzheng@...il.com 写道:
> > > From: Jinliang Zheng <alexjlzheng@...cent.com>
> > > 
> > > The commit 91562895f803 ("ext4: properly sync file size update after O_SYNC
> > > direct IO") causes confusion about the meaning of the return value of
> > > ext4_dio_write_end_io().
> > > 
> > > Specifically, when the ext4_handle_inode_extension() operation succeeds,
> > > ext4_dio_write_end_io() directly returns count instead of 0.
> > > 
> > > This does not cause a bug in the current kernel, but the semantics of the
> > > return value of the ext4_dio_write_end_io() function are wrong, which is
> > > likely to introduce bugs in the future code evolution.
> > > 
> > > Signed-off-by: Jinliang Zheng <alexjlzheng@...cent.com>
> > > ---
> > >   fs/ext4/file.c | 5 +++--
> > >   1 file changed, 3 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/fs/ext4/file.c b/fs/ext4/file.c
> > > index c89e434db6b7..6df5a92cec2b 100644
> > > --- a/fs/ext4/file.c
> > > +++ b/fs/ext4/file.c
> > > @@ -392,8 +392,9 @@ static int ext4_dio_write_end_io(struct kiocb *iocb, ssize_t size,
> > >   	 */
> > >   	if (pos + size <= READ_ONCE(EXT4_I(inode)->i_disksize) &&
> > >   	    pos + size <= i_size_read(inode))
> > > -		return size;
> > > -	return ext4_handle_inode_extension(inode, pos, size);
> > > +		return 0;
> > > +	error = ext4_handle_inode_extension(inode, pos, size);
> > > +	return error < 0 ? error : 0;
> > 
> > Why?
> 
> Before commit 91562895f803 ("ext4: properly sync file size update after O_SYNC
> direct IO"), all filesystems' iomap_dio_ops.end_io() return 0 on success and
> negative value on failure.
> 
> Moreover, this confusion of return value semantics caused data corruption when
> this above patch was merged to the stable branch. See
> https://lwn.net/Articles/954285/ for details.
> 
> > 
> > iomap_dio_complete can use the return value directly without any bug. 
> > And I think the code now seems more clearly...
> > 
> 
> In my opinion, clean code should be clearly defined code, especially the
> interface functions connecting various modules. So, what is the return value
> definition of iomap_dio_ops.end_io()? What is the return value definition of
> ext4_dio_write_end_io()?

iomap_dio_ops.end_io() implemenations should return one of:

- negative errno
- 0 if @size bytes were written

I'm not sure why ext4_dio_write_end_io returns size explicitly or why
ext4_handle_inode_extension returns count.  Neither function need to do
that.

--D

> Thanks,
> Jinliang Zheng
> 
> > >   }
> > >   
> > >   static const struct iomap_dio_ops ext4_dio_write_ops = {
> 

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ