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;
Message-ID: <4a62a93c-683a-0515-2033-26c59b3b6f20@bytedance.com>
Date:   Fri, 2 Sep 2022 11:30:53 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:102.0)
 Gecko/20100101 Thunderbird/102.2.0
Subject: Re: [PATCH v2] iommu/iova: Optimize alloc_iova with rbtree_augmented
To:     John Garry <john.garry@huawei.com>, joro@8bytes.org,
        will@kernel.org
Cc:     iommu@lists.linux.dev, linux-kernel@vger.kernel.org,
        robin.murphy@arm.com, xieyongji@bytedance.com,
        Ethan Zhao <haifeng.zhao@linux.intel.com>
References: <20220824095139.66477-1-zhangpeng.00@bytedance.com>
 <06c215c6-cbae-d6b9-312c-6535e51a3128@linux.intel.com>
 <486fcde8-e9f5-e40c-db5d-fd3ade59b267@bytedance.com>
 <a2479500-479a-22ef-3bd2-90606a26a35e@linux.intel.com>
 <4527c925-076b-a477-4ff8-697d75f16761@bytedance.com>
 <1351998d-4fd7-dffb-c1fa-a0b1ca759123@huawei.com>
From:   Peng Zhang <zhangpeng.00@bytedance.com>
In-Reply-To: <1351998d-4fd7-dffb-c1fa-a0b1ca759123@huawei.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk


> If you really have a performance issue with alloc_iova_fast() -> 
> alloc_iova() then I suggest that you consider trying to use 
> dma_opt_mapping_size() to teach the DMA engine driver to not create 
> requests whose overall size exceeds to the rcache limit.

Yes. But I don't think it essentially solves the problem.
A library for users should run stably ant it shouldn't hold
the spinlock for a long time in some cases. It can even be
said to be a bug.

Like this:
[Wed May 25 05:27:59 2022] watchdog: BUG: soft lockup - CPU#58
stuck for 23s!
[Wed May 25 05:27:59 2022] Call Trace:
[Wed May 25 05:27:59 2022]  alloc_iova+0xf2/0x140
[Wed May 25 05:27:59 2022]  alloc_iova_fast+0x56/0x251

Now we avoid this problem with iommu=pt, but it didn't solve the problem.

Thanks,
Peng