Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 13 Dec 2022 10:21:00 +0800
From:   zhoudan <zhuqiandann@gmail.com>
To:     Jaegeuk Kim <jaegeuk@kernel.org>
Cc:     chao@kernel.org, linux-f2fs-devel@lists.sourceforge.net,
        linux-kernel@vger.kernel.org, zhoudan8@xiaomi.com
Subject: Re: [PATCH] f2fs: don't set FI_COMPRESS_RELEASED if file is not
 compressed
Message-ID: <20221213022100.GA841758@mi-ThinkStation-K>
References: <20221208050808.2448146-1-zhoudan8@xiaomi.com>
 <Y5OYYJYx9G2LbRmc@google.com>
 <20221212122109.GA714122@mi-ThinkStation-K>
 <Y5ezpOwd8sOTTaW5@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Y5ezpOwd8sOTTaW5@google.com>
Precedence: bulk

However, 'f2fs_compressed_file()' only determines whether the file can 
be compressed, not whether the file has been compressed. As far as I 
know, when compress_mode is user, files marked FI_COMPRESSED_FILE 
will be compressed only after 'f2fs_ioc_compress_file()' is called.
On Mon, Dec 12, 2022 at 03:05:08PM -0800, Jaegeuk Kim wrote:
> On 12/12, zhoudan wrote:
> > Maybe I'm not describing it clearly enough, but I think there is 
> > something wrong with the logic here.The 'f2fs_release_compress_blocks'
> > method does not determine if the file is compressed, but simply adds 
> > the FI_COMPRESS_RELEASED flag. 
> 
> I firstly lost your point since f2fs_release_compress_blocks() checked
> f2fs_compressed_file().
> 
> > In particular, in the current Android system, when the application is 
> > installed, the release interface is called by default to release the 
> > storage marked as compressed,  without checking whether the file is 
> > actually compressed. In this case, when compress_mode is set to user, 
> > calling the compress interface returns ENVAL and the file cannot be 
> > compressed.
> > So I think the implementation of release needs to be modified, and 
> > only set FI_COMPRESS_RELEASED when it's really compressed and the 
> > storage is released.
> > 
> > On Fri, Dec 09, 2022 at 12:19:44PM -0800, Jaegeuk Kim wrote:
> > > On 12/08, zhoudan8 wrote:
> > > > In compress_mode=user, f2fs_release_compress_blocks()
> > > >  does not verify whether it has been compressed and
> > > >  sets FI_COMPRESS_RELEASED directly. which will lead to
> > > > return -EINVAL after calling compress.
> > > > To fix it,let's do not set FI_COMPRESS_RELEASED if file
> > > > is not compressed.
> > > 
> > > Do you mean you want to avoid EINVAL on a file having FI_COMPRESS_RELEASED
> > > with zero i_compr_blokcs?
> > > 
> > > I think the current logic is giving the error on a released file already.
> > > 
> > > > 
> > > > Signed-off-by: zhoudan8 <zhoudan8@xiaomi.com>
> > > > ---
> > > >  fs/f2fs/file.c | 3 +--
> > > >  1 file changed, 1 insertion(+), 2 deletions(-)
> > > > 
> > > > diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> > > > index 82cda1258227..f32910077df6 100644
> > > > --- a/fs/f2fs/file.c
> > > > +++ b/fs/f2fs/file.c
> > > > @@ -3451,14 +3451,13 @@ static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
> > > >  	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
> > > >  	if (ret)
> > > >  		goto out;
> > > > -
> > > > -	set_inode_flag(inode, FI_COMPRESS_RELEASED);
> > > >  	inode->i_ctime = current_time(inode);
> > > >  	f2fs_mark_inode_dirty_sync(inode, true);
> > > >  
> > > >  	if (!atomic_read(&F2FS_I(inode)->i_compr_blocks))
> > > >  		goto out;
> > > >  
> > > > +	set_inode_flag(inode, FI_COMPRESS_RELEASED);
> > > >  	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> > > >  	filemap_invalidate_lock(inode->i_mapping);
> > > >  
> > > > -- 
> > > > 2.38.1