From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-5.5 required=3.0 tests=BAYES_00,DKIMWL_WL_HIGH, DKIM_SIGNED,DKIM_VALID,HEADER_FROM_DIFFERENT_DOMAINS,MAILING_LIST_MULTI, NICE_REPLY_A,SPF_HELO_NONE,SPF_PASS,URIBL_BLOCKED,USER_AGENT_SANE_1 autolearn=no autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 495CEC433E4 for ; Thu, 16 Jul 2020 16:57:26 +0000 (UTC) Received: from merlin.infradead.org (merlin.infradead.org [205.233.59.134]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 1A00E207CB for ; Thu, 16 Jul 2020 16:57:26 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=lists.infradead.org header.i=@lists.infradead.org header.b="rQBJja2v" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 1A00E207CB Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=huawei.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-arm-kernel-bounces+linux-arm-kernel=archiver.kernel.org@lists.infradead.org DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=merlin.20170209; h=Sender:Content-Type: Content-Transfer-Encoding:Cc:List-Subscribe:List-Help:List-Post:List-Archive: List-Unsubscribe:List-Id:In-Reply-To:MIME-Version:Date:Message-ID:From: References:To:Subject:Reply-To:Content-ID:Content-Description:Resent-Date: Resent-From:Resent-Sender:Resent-To:Resent-Cc:Resent-Message-ID:List-Owner; bh=42p7uRRJJFeqsjV9RW9YwK+SAdqWjlt5lEGE0OLWOAI=; b=rQBJja2vgmTFVb/MGbqtXxM+v zFhbQ4ChYZXjSRaHn0GhgH8Z4lH2LRmpSV3a7dF0giW6ZLybUOjitbduScqrIowaKXOilIM9CODgA oXc8S3KEf3Ygpg28FwPQpQYn0qMLpzMUpRt56LiHbvh7JOgG5qwYjcyjN0ZrifdC6j1ImZibFt2Ze N+lh4l0AjjocJKJAYkulOrZ75y/vxBDX6pkl1+EyV1w+7Uo17rHxWLnpsO5c0Env1jeM0eH8sHLm6 6qlrPuR04uIB8OVEXsrlWZcNFfzqLfHX9DQMYbwUiq0TicuXy7WxtphvNNEMm608JCbdC3GqE4dex MgkBIZHSw==; Received: from localhost ([::1] helo=merlin.infradead.org) by merlin.infradead.org with esmtp (Exim 4.92.3 #3 (Red Hat Linux)) id 1jw77e-0003bx-7F; Thu, 16 Jul 2020 16:52:38 +0000 Received: from lhrrgout.huawei.com ([185.176.76.210] helo=huawei.com) by merlin.infradead.org with esmtps (Exim 4.92.3 #3 (Red Hat Linux)) id 1jw77a-0003au-Hn for linux-arm-kernel@lists.infradead.org; Thu, 16 Jul 2020 16:52:35 +0000 Received: from lhreml724-chm.china.huawei.com (unknown [172.18.7.107]) by Forcepoint Email with ESMTP id 845DA1299124E69D3125; Thu, 16 Jul 2020 17:52:30 +0100 (IST) Received: from [127.0.0.1] (10.210.168.254) by lhreml724-chm.china.huawei.com (10.201.108.75) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.1913.5; Thu, 16 Jul 2020 17:52:29 +0100 Subject: Re: [PATCH 0/4] iommu/arm-smmu-v3: Improve cmdq lock efficiency To: Will Deacon , Robin Murphy References: <1592846920-45338-1-git-send-email-john.garry@huawei.com> <20200716101940.GA7036@willie-the-truck> <20200716102233.GC7036@willie-the-truck> <20200716102814.GD7036@willie-the-truck> <20200716113234.GA7290@willie-the-truck> From: John Garry Message-ID: Date: Thu, 16 Jul 2020 17:50:41 +0100 User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:68.0) Gecko/20100101 Thunderbird/68.1.2 MIME-Version: 1.0 In-Reply-To: <20200716113234.GA7290@willie-the-truck> Content-Language: en-US X-Originating-IP: [10.210.168.254] X-ClientProxiedBy: lhreml714-chm.china.huawei.com (10.201.108.65) To lhreml724-chm.china.huawei.com (10.201.108.75) X-CFilter-Loop: Reflected X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20200716_125234_740338_BF43A4E1 X-CRM114-Status: GOOD ( 16.91 ) X-BeenThere: linux-arm-kernel@lists.infradead.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: trivial@kernel.org, maz@kernel.org, joro@8bytes.org, linuxarm@huawei.com, linux-kernel@vger.kernel.org, iommu@lists.linux-foundation.org, linux-arm-kernel@lists.infradead.org Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="us-ascii"; Format="flowed" Sender: "linux-arm-kernel" Errors-To: linux-arm-kernel-bounces+linux-arm-kernel=archiver.kernel.org@lists.infradead.org >> >> Perhaps a silly question (I'm too engrossed in PMU world ATM to get properly >> back up to speed on this), but couldn't this be done without cmpxchg anyway? >> Instinctively it feels like instead of maintaining a literal software copy >> of the prod value, we could resolve the "claim my slot in the queue" part >> with atomic_fetch_add on a free-running 32-bit "pseudo-prod" index, then >> whoever updates the hardware deals with the truncation and wrap bit to >> convert it to an actual register value. > > Maybe, but it's easier said than done. The hard part is figuring how that > you have space and there's no way I'm touching that logic without a way to > test this. > > I'm also not keen to complicate the code because of systems that don't scale > well with contended CAS [1]. If you put down loads of cores, you need an > interconnect/coherence protocol that can handle it. JFYI, I added some debug to the driver to get the cmpxchg() attempt rate while running a testharness (below): cpus rate 2 1.1 4 1.1 8 1.3 16 3.6 32 8.1 64 12.6 96 14.7 Ideal rate is 1. So it's not crazy high for many CPUs, but still drifting away from 1. John > > Will > > [1] https://lore.kernel.org/lkml/20190607072652.GA5522@hc/T/#m0d00f107c29223045933292a8b5b90d2ca9b7e5c > . > //copied from Barry, thanks :) static int ways = 64; module_param(ways, int, S_IRUGO); static int seconds = 20; module_param(seconds, int, S_IRUGO); int mappings[NR_CPUS]; struct semaphore sem[NR_CPUS]; #define COMPLETIONS_SIZE 50 static noinline dma_addr_t test_mapsingle(struct device *dev, void *buf, int size) { dma_addr_t dma_addr = dma_map_single(dev, buf, size, DMA_TO_DEVICE); return dma_addr; } static noinline void test_unmapsingle(struct device *dev, void *buf, int size, dma_addr_t dma_addr) { dma_unmap_single(dev, dma_addr, size, DMA_TO_DEVICE); } static noinline void test_memcpy(void *out, void *in, int size) { memcpy(out, in, size); } //just a hack to get a dev h behind a SMMU extern struct device *hisi_dev; static int testthread(void *data) { unsigned long stop = jiffies +seconds*HZ; struct device *dev = hisi_dev; char *inputs[COMPLETIONS_SIZE]; char *outputs[COMPLETIONS_SIZE]; dma_addr_t dma_addr[COMPLETIONS_SIZE]; int i, cpu = smp_processor_id(); struct semaphore *sem = data; for (i = 0; i < COMPLETIONS_SIZE; i++) { inputs[i] = kzalloc(4096, GFP_KERNEL); if (!inputs[i]) return -ENOMEM; } for (i = 0; i < COMPLETIONS_SIZE; i++) { outputs[i] = kzalloc(4096, GFP_KERNEL); if (!outputs[i]) return -ENOMEM; } while (time_before(jiffies, stop)) { for (i = 0; i < COMPLETIONS_SIZE; i++) { dma_addr[i] = test_mapsingle(dev, inputs[i], 4096); test_memcpy(outputs[i], inputs[i], 4096); } for (i = 0; i < COMPLETIONS_SIZE; i++) { test_unmapsingle(dev, inputs[i], 4096, dma_addr[i]); } mappings[cpu] += COMPLETIONS_SIZE; } for (i = 0; i < COMPLETIONS_SIZE; i++) { kfree(outputs[i]); kfree(inputs[i]); } up(sem); return 0; } void smmu_test_core(int cpus) { struct task_struct *tsk; int i; int total_mappings = 0; ways = cpus; for(i=0;i