Re: [PATCH 3/4] writeback, memcg: Implement cgroup_writeback_by_id()

From: Tejun Heo <tj@kernel.org>
To: Jan Kara <jack@suse.cz>
Cc: axboe@kernel.dk, hannes@cmpxchg.org, mhocko@kernel.org,
	vdavydov.dev@gmail.com, cgroups@vger.kernel.org,
	linux-mm@kvack.org, linux-block@vger.kernel.org,
	linux-kernel@vger.kernel.org, kernel-team@fb.com, guro@fb.com,
	akpm@linux-foundation.org
Subject: Re: [PATCH 3/4] writeback, memcg: Implement cgroup_writeback_by_id()
Date: Thu, 15 Aug 2019 09:12:11 -0700	[thread overview]
Message-ID: <20190815161211.GC588936@devbig004.ftw2.facebook.com> (raw)
In-Reply-To: <20190815145421.GN14313@quack2.suse.cz>

On Thu, Aug 15, 2019 at 04:54:21PM +0200, Jan Kara wrote:
> > +	/* and find the associated wb */
> > +	wb = wb_get_create(bdi, memcg_css, GFP_NOWAIT | __GFP_NOWARN);
> > +	if (!wb) {
> > +		ret = -ENOMEM;
> > +		goto out_css_put;
> > +	}
> 
> One more thought: You don't want the "_create" part here, do you? If
> there's any point in writing back using this wb, it must be attached to
> some inode and thus it must exist. In the normal case wb_get_create() will
> just fetch the reference and be done with it but when you feed garbage into
> this function due to id going stale or frn structures getting corrupted due
> to concurrent access, you can be creating bogus wb structures in bdi...

Yeah, it can create wbs unnecessarily which isn't critical but also is
easy to fix.  Will update.

Thanks.

-- 
tejun