Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 27 Jun 2022 13:38:22 +0200
From:   Jan Kara <jack@suse.cz>
To:     Zhang Yi <yi.zhang@huawei.com>
Cc:     Jan Kara <jack@suse.cz>, linux-ext4@vger.kernel.org, tytso@mit.edu,
        adilger.kernel@dilger.ca, yukuai3@huawei.com
Subject: Re: [PATCH] ext4: silence the warning when evicting inode with
 dioread_nolock
Message-ID: <20220627113822.g55mttsy2nx6yuyr@quack3.lan>
References: <20220624070404.763603-1-yi.zhang@huawei.com>
 <20220624125117.bi5o4ovuhhtgs44x@quack3.lan>
 <22096be1-d77a-a7e4-cb72-6378e76ae6cd@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <22096be1-d77a-a7e4-cb72-6378e76ae6cd@huawei.com>
Precedence: bulk

On Sat 25-06-22 17:33:50, Zhang Yi wrote:
> > probably we should add:
> > 
> > 	WARN_ON_ONCE(!list_empty(&inode->i_io_list));
> > 
> > to the no_delete: case of ext4_evict_inode()? Race like you mention above
> > does not seem possible for that case but seeing the complicated
> > interactions I'd rather have the assertion in place.
> > 
> 
> For the no_delete case, I did some tests and IIUC, it's true that the race could
> not happen, because inode_lru_isolate() make sure inode->i_data.nrpages is zero
> before adding inode into the freeable list, so the evict() procedure could not be
> invoked before the page cache have been dropped (it could only happened after
> ext4_end_io_end() has been finished).
> 
> We don't have a !list_empty(&inode->i_io_list) check for the no_delete case now.
> But I am not quite get the purpose of adding it, do you want to detect inode
> use-after-free issue in advance?

Yes, I wanted to check that we didn't accidentally dirty the inode
somewhere on the eviction path which would cause use-after-free issues
which are always hard to debug...

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR