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;
Sender: Tejun Heo <htejun@gmail.com>
Date:   Fri, 27 May 2022 07:23:47 -1000
From:   Tejun Heo <tj@kernel.org>
To:     Michal =?iso-8859-1?Q?Koutn=FD?= <mkoutny@suse.com>
Cc:     Tadeusz Struk <tadeusz.struk@linaro.org>, cgroups@vger.kernel.org,
        linux-kernel@vger.kernel.org, Zefan Li <lizefan.x@bytedance.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Bui Quang Minh <minhquangbui99@gmail.com>
Subject: Re: [PATCH 2/2] cgroup: Use separate work structs on css release path
Message-ID: <YpEJI7OZDVnxHRA8@slm.duckdns.org>
References: <20220525151517.8430-1-mkoutny@suse.com>
 <20220525151517.8430-3-mkoutny@suse.com>
 <20220525161455.GA16134@blackbody.suse.cz>
 <Yo7KfEOz92kS2z5Y@blackbook>
 <Yo/DtjEU/kYr190u@slm.duckdns.org>
 <904ef8af-13a5-e566-b760-74519f70fa62@linaro.org>
 <20220527165429.GJ11007@blackbody.suse.cz>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20220527165429.GJ11007@blackbody.suse.cz>
Precedence: bulk

On Fri, May 27, 2022 at 06:54:29PM +0200, Michal Koutn? wrote:
> Hello Tadeusz.
> 
> On Fri, May 27, 2022 at 09:39:20AM -0700, Tadeusz Struk <tadeusz.struk@linaro.org> wrote:
> > As far as I can see we are trying to test the same thing suggested by Tejun.
> > I just sent a test request to try this:
> > https://github.com/tstruk/linux/commit/master
> 
> Yup, I've added few more prints to get more fine-grained resolution.
> Also, I decided to use ftrace printk not to interfere with timing too
> much (due to the original race hypothesis).
> 
> > Let me know if you have any more tests to run and I will hold off until
> > you are done.
> 
> My latest attempt is [1] (tip 5500e05d82fd5b5db2203eedb3f786857d3ccbea).
> 
> So far, I'm not convinced, I extract the complete ftrace buffer from the
> syzbot runs, so I'm not drawing any conclusions from the traces I've
> got. I'm not going to continue today. You may have more luck with your
> plain printk (if it's just imbalance and it avoids printk locking
> sensitive paths).

At least for triaging, it may be easier to use bpftrace to attach kprobes to
the get/put functions with the right filter and record the kstack counts.

Thanks.

-- 
tejun