From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: <dev-bounces@dpdk.org> Received: from dpdk.org (dpdk.org [92.243.14.124]) by inbox.dpdk.org (Postfix) with ESMTP id 70CFDA04DF; Wed, 21 Oct 2020 19:24:32 +0200 (CEST) Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id 1F51972E2; Wed, 21 Oct 2020 19:24:31 +0200 (CEST) Received: from mga03.intel.com (mga03.intel.com [134.134.136.65]) by dpdk.org (Postfix) with ESMTP id 318E872E0 for <dev@dpdk.org>; Wed, 21 Oct 2020 19:24:29 +0200 (CEST) IronPort-SDR: iyJxRdI8JmCM/ZHGOys9a6MfyTjm2p623HDfkIY8OVFDG+ZXO04lOh81DGC9Z0ed9bGlJDHw3h fvs3DEA/MA9A== X-IronPort-AV: E=McAfee;i="6000,8403,9780"; a="167497478" X-IronPort-AV: E=Sophos;i="5.77,401,1596524400"; d="scan'208";a="167497478" X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga005.jf.intel.com ([10.7.209.41]) by orsmga103.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 21 Oct 2020 10:24:27 -0700 IronPort-SDR: cMGuOIT8a5U0gSDwTFCbUGmp4To5d5WCBHhaq9PINDCGZCrGWxUdo9sn2nN9rNKW1U3eqJQ2zv iOfd77RHiLAQ== X-IronPort-AV: E=Sophos;i="5.77,401,1596524400"; d="scan'208";a="533615542" Received: from fyigit-mobl1.ger.corp.intel.com (HELO [10.251.86.101]) ([10.251.86.101]) by orsmga005-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 21 Oct 2020 10:24:25 -0700 To: =?UTF-8?B?6LCi5Y2O5LyfKOatpOaXtuatpOWIu++8iQ==?= <huawei.xhw@alibaba-inc.com>, Maxime Coquelin <maxime.coquelin@redhat.com> Cc: dev@dpdk.org, anatoly.burakov@intel.com, david.marchand@redhat.com, grive@u256.net, zhihong.wang@intel.com, chenbo.xia@intel.com References: <68ecd941-9c56-4de7-fae2-2ad15bdfd81a@alibaba-inc.com> <1602578499-13975-1-git-send-email-huawei.xhw@alibaba-inc.com> <1602578499-13975-2-git-send-email-huawei.xhw@alibaba-inc.com> <2298e6a6-2ca8-0f68-6742-1b7d5dde97a7@intel.com> <5d26db1a-6970-6e3a-c2e5-8dc540922028@alibaba-inc.com> From: Ferruh Yigit <ferruh.yigit@intel.com> Message-ID: <e23712f1-7dbb-ce3c-2694-5648eb8369ae@intel.com> Date: Wed, 21 Oct 2020 18:24:21 +0100 MIME-Version: 1.0 In-Reply-To: <5d26db1a-6970-6e3a-c2e5-8dc540922028@alibaba-inc.com> Content-Type: text/plain; charset=utf-8; format=flowed Content-Language: en-US Content-Transfer-Encoding: 8bit Subject: Re: [dpdk-dev] [PATCH v4] pci: support both PIO and MMIO BAR for legacy virtio on x86 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions <dev.dpdk.org> List-Unsubscribe: <https://mails.dpdk.org/options/dev>, <mailto:dev-request@dpdk.org?subject=unsubscribe> List-Archive: <http://mails.dpdk.org/archives/dev/> List-Post: <mailto:dev@dpdk.org> List-Help: <mailto:dev-request@dpdk.org?subject=help> List-Subscribe: <https://mails.dpdk.org/listinfo/dev>, <mailto:dev-request@dpdk.org?subject=subscribe> Errors-To: dev-bounces@dpdk.org Sender: "dev" <dev-bounces@dpdk.org> On 10/21/2020 1:32 PM, 谢华伟(此时此刻） wrote: > > On 2020/10/21 19:49, Ferruh Yigit wrote: >> On 10/13/2020 9:41 AM, 谢华伟(此时此刻） wrote: >>> From: "huawei.xhw" <huawei.xhw@alibaba-inc.com> >>> >>> Legacy virtio-pci only supports PIO BAR resource. As we need to create lots of >>> virtio devices and PIO resource on x86 is very limited, we expose MMIO BAR. >>> >>> Kernel supports both PIO and MMIO BAR for legacy virtio-pci device. We handles >>> different type of BAR in the similar way. >>> >>> In previous implementation, with igb_uio we get PIO address from igb_uio >>> sysfs entry; with uio_pci_generic, we get PIO address from >>> /proc/ioports. >>> For PIO/MMIO RW, there is different path for different drivers and arch. >>> For VFIO, PIO/MMIO RW is through syscall, which has big performance >>> issue. >>> On X86, it assumes only PIO is supported. >>> >>> All of the above is too much twisted. >>> This patch unifies the way to get both PIO and MMIO address for different driver >>> and arch, all from standard resource attr under pci sysfs. >>> >> >> As mentined above this patch does multiple things. >> >> The main target is, as far as I understand, you have a legacy virtio device >> which supports "memory-mapped I/O" and "port-mapped I/O", but virtio logic >> forces legacy devices to use the PIO but you want to be able to use the MMIO >> with this device. > yes. >> >> The solution below is adding MMIO support in the PIO funciton, and distinguish >> MMIO or PIO based on their address check. > Yes, kernel does this in the similar way. >> >> >> Instead of this, can't this be resolved in the virtio side, like if the legacy >> device supports MMIO (detect this somehow) use the MMIO istead of hacking PIO >> mapping to support MMIO? > > Get your concern. > > 1> > > If we move, I think we should move all those PCI codes into virtio side, not > just the mmio part. > > Without my patch, those PCI codes are virtio-pci device specific, not generic. > > With this patch, those pci ioport map/rw code could also be used for other > devices if they support both PIO and MMIO. > I was not suggesting moving any code into virtio, but within 'vtpci_init()' what happens when "hw->modern = 1;" is set? And if this is set for your device, will it work without change? > Every option is ok. Hope i make myself clear. > > 2> I don't think this is hacking. for rte_pci_ioport_map/read/write, if ioport > could be both PIO and MMIO, then everything is reasonable. > > Take how kernel does port map for example: > > vp_dev->ioaddr = pci_iomap(pci_dev, 0, 0); > > Here io doesn't mean PIO only. It could also be MMIO. Kernel then uses > ioread/write to access PIO/MMIO port. > > Actually we are pretty much the same in the interface. > > I think this patch extends rather then hacks the ioport interface to support MMIO. > >> >> I have other concerns, specially mergin VFIO mapping too, but lets clarify >> above first. > > vfio doesn't affect other driver but only virtio. > Why it doesn't affect other drivers, can't there be other driver using PIO? > igb_uio, uio_pci_generic and vfio-pci all uses the same way to map/rw ioport. > For vfio, code changes 'pci_vfio_ioport_read()' to the direct address read, first I don't know if this is always safe, and my question why there is a syscall introduced at first place if you can read from address directly? Is your device works as expected when vfio-pci kernel module used? Since it is not suffering from PIO limitation, right? And I wonder if the patch can be done as three patches to simply it, as: 1) Combine 'RTE_PCI_KDRV_IGB_UIO' & 'RTE_PCI_KDRV_UIO_GENERIC' (remove pci_ioport_map) 2) Update 'pci_uio_ioport_map()' to add memory map support (and update read/write functions according) 3) Combine vfio & uio >> >> Thanks, >> ferruh >> >> >> >>> We distinguish PIO and MMIO by their address like how kernel does. It is ugly >>> but works. >>> >>> Signed-off-by: huawei.xhw <huawei.xhw@alibaba-inc.com> <...>