Message-ID: <65da1862-364b-9500-4be7-a463a12e6a7f@bytedance.com>
Date:   Thu, 16 Feb 2023 19:02:59 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:102.0)
 Gecko/20100101 Thunderbird/102.7.2
Subject: Re: [PATCH] iommu: Avoid softlockup and rcu stall in
 fq_flush_timeout().
To:     Hillf Danton <hdanton@sina.com>
Cc:     robin.murphy@arm.com, joro@8bytes.org, will@kernel.org,
        iommu@lists.linux.dev, linux-kernel@vger.kernel.org,
        Li Bin <huawei.libin@huawei.com>,
        Xie XiuQi <xiexiuqi@huawei.com>,
        Yang Yingliang <yangyingliang@huawei.com>
References: <20230216084902.1486-1-hdanton@sina.com>
From:   Peng Zhang <zhangpeng.00@bytedance.com>
In-Reply-To: <20230216084902.1486-1-hdanton@sina.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


在 2023/2/16 16:49, Hillf Danton 写道:
> On Thu, 16 Feb 2023 15:11:48 +0800 Peng Zhang <zhangpeng.00@bytedance.com>
>> There is softlockup under fio pressure test with smmu enabled:
>> watchdog: BUG: soft lockup - CPU#81 stuck for 22s!  [swapper/81:0]
> 
> What is your kernel version?
Rcu stall occurs in kernel version 5.4.
The test where the softlockup happened was not done by me, so I don't 
know the kernel version.
However, it is the same as the code logic of fq_flush_timeout in the 
mainline kernel.

>> This is because the timer callback fq_flush_timeout may run more than
>> 10ms, and timer may be processed continuously in the softirq so trigger
>> softlockup and rcu stall. We can use work to deal with fq_ring_free for
>> each cpu which may take long time, that to avoid triggering softlockup
>> and rcu stall.
>>
>> This patch is modified from the patch[1] of openEuler.
> 
> Because of a timer hog observed on your system with 128 CPUs for instance
> does it make any sense to ask Peter to apply the patch for his 2-CPU box?
What is 2-CPU box?