From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga14.intel.com (mga14.intel.com [192.55.52.115]) by dpdk.org (Postfix) with ESMTP id C3813201 for ; Tue, 24 Jan 2017 20:45:42 +0100 (CET) Received: from orsmga005.jf.intel.com ([10.7.209.41]) by fmsmga103.fm.intel.com with ESMTP; 24 Jan 2017 11:45:41 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.33,279,1477983600"; d="scan'208";a="56766705" Received: from irsmsx101.ger.corp.intel.com ([163.33.3.153]) by orsmga005.jf.intel.com with ESMTP; 24 Jan 2017 11:45:40 -0800 Received: from irsmsx105.ger.corp.intel.com ([169.254.7.38]) by IRSMSX101.ger.corp.intel.com ([169.254.1.112]) with mapi id 14.03.0248.002; Tue, 24 Jan 2017 19:45:39 +0000 From: "Ananyev, Konstantin" To: "Wiles, Keith" CC: Stephen Hemminger , "Hu, Jiayu" , "dev@dpdk.org" , "Kinsella, Ray" , "Gilmore, Walter E" , "Venkatesan, Venky" , "yuanhan.liu@linux.intel.com" Thread-Topic: [dpdk-dev] [RFC] Add GRO support in DPDK Thread-Index: AQHSdXke8G3CoYE5S0ao0Pji4hqADqFGTb0AgABNfoCAAD8VoIAAP1oAgABT/hCAAElGgIAAPOdA Date: Tue, 24 Jan 2017 19:45:39 +0000 Message-ID: <2601191342CEEE43887BDE71AB9772583F10C6DC@irsmsx105.ger.corp.intel.com> References: <1485176592-111525-1-git-send-email-jiayu.hu@intel.com> <20170123091550.212dca35@xeon-e3> <6B5C6BED-CAD4-4C51-8FB7-8509663B813B@intel.com> <2601191342CEEE43887BDE71AB9772583F10AD94@irsmsx105.ger.corp.intel.com> <1F520FF1-C38B-483B-95E1-FBD4C631E7D2@intel.com> <2601191342CEEE43887BDE71AB9772583F10AEBD@irsmsx105.ger.corp.intel.com> <6D277342-5212-462F-A507-93B63E86DA90@intel.com> In-Reply-To: <6D277342-5212-462F-A507-93B63E86DA90@intel.com> Accept-Language: en-IE, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-originating-ip: [163.33.239.181] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Subject: Re: [dpdk-dev] [RFC] Add GRO support in DPDK X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 24 Jan 2017 19:45:43 -0000 > -----Original Message----- > From: Wiles, Keith > Sent: Tuesday, January 24, 2017 2:49 PM > To: Ananyev, Konstantin > Cc: Stephen Hemminger ; Hu, Jiayu ; dev@dpdk.org; Kinsella, Ray > ; Gilmore, Walter E ;= Venkatesan, Venky ; > yuanhan.liu@linux.intel.com > Subject: Re: [dpdk-dev] [RFC] Add GRO support in DPDK >=20 >=20 > > On Jan 24, 2017, at 3:33 AM, Ananyev, Konstantin wrote: > > > > > > > >> -----Original Message----- > >> From: Wiles, Keith > >> Sent: Tuesday, January 24, 2017 5:26 AM > >> To: Ananyev, Konstantin > >> Cc: Stephen Hemminger ; Hu, Jiayu ; dev@dpdk.org; Kinsella, Ray > >> ; Gilmore, Walter E ; Venkatesan, Venky ; > >> yuanhan.liu@linux.intel.com > >> Subject: Re: [dpdk-dev] [RFC] Add GRO support in DPDK > >> > >> > >>> On Jan 23, 2017, at 6:43 PM, Ananyev, Konstantin wrote: > >>> > >>> > >>> > >>>> -----Original Message----- > >>>> From: Wiles, Keith > >>>> Sent: Monday, January 23, 2017 9:53 PM > >>>> To: Stephen Hemminger > >>>> Cc: Hu, Jiayu ; dev@dpdk.org; Kinsella, Ray ; Ananyev, Konstantin > >>>> ; Gilmore, Walter E ; Venkatesan, Venky > >> ; > >>>> yuanhan.liu@linux.intel.com > >>>> Subject: Re: [dpdk-dev] [RFC] Add GRO support in DPDK > >>>> > >>>> > >>>>> On Jan 23, 2017, at 10:15 AM, Stephen Hemminger wrote: > >>>>> > >>>>> On Mon, 23 Jan 2017 21:03:12 +0800 > >>>>> Jiayu Hu wrote: > >>>>> > >>>>>> With the support of hardware segmentation techniques in DPDK, the > >>>>>> networking stack overheads of send-side of applications, which dir= ectly > >>>>>> leverage DPDK, have been greatly reduced. But for receive-side, nu= mbers of > >>>>>> segmented packets seriously burden the networking stack of applica= tions. > >>>>>> Generic Receive Offload (GRO) is a widely used method to solve the > >>>>>> receive-side issue, which gains performance by reducing the amount= of > >>>>>> packets processed by the networking stack. But currently, DPDK doe= sn't > >>>>>> support GRO. Therefore, we propose to add GRO support in DPDK, and= this > >>>>>> RFC is used to explain the basic DPDK GRO design. > >>>>>> > >>>>>> DPDK GRO is a SW-based packets assembly library, which provides GR= O > >>>>>> abilities for numbers of protocols. In DPDK GRO, packets are merge= d > >>>>>> before returning to applications and after receiving from drivers. > >>>>>> > >>>>>> In DPDK, GRO is a capability of NIC drivers. That support GRO or n= ot and > >>>>>> what GRO types are supported are up to NIC drivers. Different driv= ers may > >>>>>> support different GRO types. By default, drivers enable all suppor= ted GRO > >>>>>> types. For applications, they can inquire the supported GRO types = by > >>>>>> each driver, and can control what GRO types are applied. For examp= le, > >>>>>> ixgbe supports TCP and UDP GRO, but the application just needs TCP= GRO. > >>>>>> The application can disable ixgbe UDP GRO. > >>>>>> > >>>>>> To support GRO, a driver should provide a way to tell applications= what > >>>>>> GRO types are supported, and provides a GRO function, which is in = charge > >>>>>> of assembling packets. Since different drivers may support differe= nt GRO > >>>>>> types, their GRO functions may be different. For applications, the= y don't > >>>>>> need extra operations to enable GRO. But if there are some GRO typ= es that > >>>>>> are not needed, applications can use an API, like > >>>>>> rte_eth_gro_disable_protocols, to disable them. Besides, they can > >>>>>> re-enable the disabled ones. > >>>>>> > >>>>>> The GRO function processes numbers of packets at a time. In each > >>>>>> invocation, what GRO types are applied depends on applications, an= d the > >>>>>> amount of packets to merge depends on the networking status and > >>>>>> applications. Specifically, applications determine the maximum num= ber of > >>>>>> packets to be processed by the GRO function, but how many packets = are > >>>>>> actually processed depends on if there are available packets to re= ceive. > >>>>>> For example, the receive-side application asks the GRO function to > >>>>>> process 64 packets, but the sender only sends 40 packets. At this = time, > >>>>>> the GRO function returns after processing 40 packets. To reassembl= e the > >>>>>> given packets, the GRO function performs an "assembly procedure" o= n each > >>>>>> packet. We use an example to demonstrate this procedure. Supposing= the > >>>>>> GRO function is going to process packetX, it will do the following= two > >>>>>> things: > >>>>>> a. Find a L4 assembly function according to the packet type of > >>>>>> packetX. A L4 assembly function is in charge of merging packets o= f a > >>>>>> specific type. For example, TCPv4 assembly function merges packet= s > >>>>>> whose L3 IPv4 and L4 is TCP. Each L4 assembly function has a pack= et > >>>>>> array, which keeps the packets that are unable to assemble. > >>>>>> Initially, the packet array is empty; > >>>>>> b. The L4 assembly function traverses own packet array to find a > >>>>>> mergeable packet (comparing Ethernet, IP and L4 header fields). I= f > >>>>>> finds, merges it and packetX via chaining them together; if doesn= 't, > >>>>>> allocates a new array element to store packetX and updates elemen= t > >>>>>> number of the array. > >>>>>> After performing the assembly procedure to all packets, the GRO fu= nction > >>>>>> combines the results of all packet arrays, and returns these packe= ts to > >>>>>> applications. > >>>>>> > >>>>>> There are lots of ways to implement the above design in DPDK. One = of the > >>>>>> ways is: > >>>>>> a. Drivers tell applications what GRO types are supported via > >>>>>> dev->dev_ops->dev_infos_get; > >>>>>> b. When initialize, drivers register own GRO function as a RX > >>>>>> callback, which is invoked inside rte_eth_rx_burst. The name of t= he > >>>>>> GRO function should be like xxx_gro_receive (e.g. ixgbe_gro_recei= ve). > >>>>>> Currently, the RX callback can only process the packets returned = by > >>>>>> dev->rx_pkt_burst each time, and the maximum packet number > >>>>>> dev->rx_pkt_burst returns is determined by each driver, which can= 't > >>>>>> be interfered by applications. Therefore, to implement the above = GRO > >>>>>> design, we have to modify current RX implementation to make drive= r > >>>>>> return packets as many as possible until the packet number meets = the > >>>>>> demand of applications or there are not available packets to rece= ive. > >>>>>> This modification is also proposed in patch: > >>>>>> http://dpdk.org/ml/archives/dev/2017-January/055887.html; > >>>>>> c. The GRO types to apply and the maximum number of packets to me= rge > >>>>>> are passed by resetting RX callback parameters. It can be achieve= d by > >>>>>> invoking rte_eth_rx_callback; > >>>>>> d. Simply, we can just store packet addresses into the packet arr= ay. > >>>>>> To check one element, we need to fetch the packet via its address= . > >>>>>> However, this simple design is not efficient enough. Since whenev= er > >>>>>> checking one packet, one pointer dereference is generated. And a > >>>>>> pointer dereference always causes a cache line miss. A better way= is > >>>>>> to store some rules in each array element. The rules must be the > >>>>>> prerequisites of merging two packets, like the sequence number of= TCP > >>>>>> packets. We first compare the rules, then retrieve the packet if = the > >>>>>> rules match. If storing the rules causes the packet array structu= re > >>>>>> is cache-unfriendly, we can store a fixed-length signature of the > >>>>>> rules instead. For example, the signature can be calculated by > >>>>>> performing XOR operation on IP addresses. Both design can avoid > >>>>>> unnecessary pointer dereferences. > >>>>> > >>>>> > >>>>> Since DPDK does burst mode already, GRO is a lot less relevant. > >>>>> GRO in Linux was invented because there is no burst mode in the rec= eive API. > >>>>> > >>>>> If you look at VPP in FD.io you will see they already do aggregrati= on and > >>>>> steering at the higher level in the stack. > >>>>> > >>>>> The point of GRO is that it is generic, no driver changes are neces= sary. > >>>>> Your proposal would add a lot of overhead, and cause drivers to hav= e to > >>>>> be aware of higher level flows. > >>>> > >>>> NACK > >>>> > >>>> The design is not super clear to me here and we need to understand t= he impact to DPDK, performance and the application. I would > like > >> to > >>>> have a clean transparent design to the application and as little imp= act on performance as possible. > >>>> > >>>> Let discuss this as I am not sure my previous concerns were addresse= d in this RFC. > >>>> > >>> > >>> I would agree that design looks overcomplicated and strange: > >>> If GRO can (and supposed to be) done fully in SW, why do we need to m= odify PMDs at all, > >>> why it can't be just a standalone DPDK library that user can use on h= is/her convenience? > >>> I'd suggest to start with some simple and most widespread case (TCP?)= and try to implement > >>> a library for it first: something similar to what we have for ip reas= sembly. > >> > >> The reason this should not be a library the application calls is to al= low for a transparent design for HW and SW support of this feature. > Using > >> the SW version the application should not need to understand (other th= en performance) that GRO is being done for this port. > >> > > > > Why is that? > > Let say we have ip reassembly library that is called explicitly by the = application. > > I think for L4 grouping we can do the same. > > After all it is a pure SW feature, so to me it makes sense to allow app= lication to decide > > when/where to call it. > > Again it would allow people to develop/use it without any modifications= in current PMDs. >=20 > I guess I did not make it clear, we need to support HW and this SW versio= n transparently just as we handle other features in HW/SW under a > generic API for DPDK. Ok, I probably wasn't very clear too. What I meant: Let's try to implement GRO (in SW) as a standalone DPDK library, with clean & simple interface and see how fast and useful it would be. We can refer to it as step 1. When (if) we'll have step 1 in place, then we can start thinking about adding combined HW/SW solution for it (step 2). I think at that stage it would be much clearer: is there any point in it at all, and if yes, how it should be done: -changes at rte_ethedev or on PMD layers or both - would changes at rte_ethdev API be needed and if yes what particular, e= tc. >>From my perspective, without step 1 in place, there is no much point in ap= proaching step 2. BTW, any particular HW you have in mind? Currently, as I can see LRO (HW) is supported only by ixgbe and probably by= viritual PMDs (virtio/vmxent3). Though even for ixgbe there are plenty of limitations: SRIOV mode should be= off, HW CRC stropping should be off, etc. So my guess, right now step 1 is much more useful and feasible. =20 >=20 > > > >> As I was told the Linux kernel hides this features and make it transpa= rent. > > > > Yes, but DPDK does a lot things in a different way. > > So it doesn't look like a compelling reason for me :) >=20 > Just looking at different options here and it is a compelling reason to m= e as it enforces the design can be transparent to the application. > Having the application in a NFV deciding on hw or sw or both is not a goo= d place to put that logic IMO. Actually could you provide an example of linux NIC driver, that uses HW off= loads (and which) to implement GRO? I presume some might use HW generated hashes, but apart from that, when HW = performs actual packet grouping? >>From what I've seen Intel ones rely SW implementation for that. But I am not a linux/GRO expert, so feel free to correct me here. Konstantin=20