Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753225Ab0FMI7B (ORCPT ); Sun, 13 Jun 2010 04:59:01 -0400 Received: from mga11.intel.com ([192.55.52.93]:28586 "EHLO mga11.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753084Ab0FMI66 convert rfc822-to-8bit (ORCPT ); Sun, 13 Jun 2010 04:58:58 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.53,410,1272870000"; d="scan'208";a="807455352" From: "Xin, Xiaohui" To: "Xin, Xiaohui" , Herbert Xu CC: Stephen Hemminger , "netdev@vger.kernel.org" , "kvm@vger.kernel.org" , "linux-kernel@vger.kernel.org" , "mst@redhat.com" , "mingo@elte.hu" , "davem@davemloft.net" , "jdike@linux.intel.com" Date: Sun, 13 Jun 2010 16:58:36 +0800 Subject: RE: [RFC PATCH v7 01/19] Add a new structure for skb buffer from external. Thread-Topic: [RFC PATCH v7 01/19] Add a new structure for skb buffer from external. Thread-Index: AcsJJe1B7Nc7C/pfSVyCPcMLZUukkgA4bXJgADLVSAA= Message-ID: References: <1275732899-5423-1-git-send-email-xiaohui.xin@intel.com> <20100606161348.427822fb@nehalam> <20100608052744.GA21547@gondor.apana.org.au> <20100611052112.GA25649@gondor.apana.org.au> In-Reply-To: Accept-Language: en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: acceptlanguage: en-US Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8BIT MIME-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5383 Lines: 121 >-----Original Message----- >From: netdev-owner@vger.kernel.org [mailto:netdev-owner@vger.kernel.org] On Behalf Of >Xin, Xiaohui >Sent: Saturday, June 12, 2010 5:31 PM >To: Herbert Xu >Cc: Stephen Hemminger; netdev@vger.kernel.org; kvm@vger.kernel.org; >linux-kernel@vger.kernel.org; mst@redhat.com; mingo@elte.hu; davem@davemloft.net; >jdike@linux.intel.com >Subject: RE: [RFC PATCH v7 01/19] Add a new structure for skb buffer from external. > >>-----Original Message----- >>From: Herbert Xu [mailto:herbert@gondor.apana.org.au] >>Sent: Friday, June 11, 2010 1:21 PM >>To: Xin, Xiaohui >>Cc: Stephen Hemminger; netdev@vger.kernel.org; kvm@vger.kernel.org; >>linux-kernel@vger.kernel.org; mst@redhat.com; mingo@elte.hu; davem@davemloft.net; >>jdike@linux.intel.com >>Subject: Re: [RFC PATCH v7 01/19] Add a new structure for skb buffer from external. >> >>On Wed, Jun 09, 2010 at 05:54:02PM +0800, Xin, Xiaohui wrote: >>> >>> I'm not sure if I understand your way correctly: >>> 1) Does the way only deal with driver with SG feature? Since packet >>> is non-linear... >> >>No the hardware doesn't have to support SG. You just need to >>place the entire packet contents in a page instead of skb->head. >> >>> 2) Is skb->data still pointing to guest user buffers? >>> If yes, how to avoid the modifications to net core change to skb? >> >>skb->data would not point to guest user buffers. In the common >>case the packet is not modified on its way to the guest so this >>is not an issue. >> >>In the rare case where it is modified, you only have to copy the >>bits which are modified and the cost of that is inconsequential >>since you have to write to that memory anyway. >> >>> 3) In our way only parts of drivers need be modified to support zero-copy. >>> and here, need we modify all the drivers? >> >>If you're asking the portion of each driver supporting zero-copy >>that needs to be modified, then AFAICS this doesn't change that >>very much at all. >> >>> I think to make skb->head empty at first will cause more effort to pass the check with >>> skb header. Have I missed something here? I really make the skb->head NULL >>> just before kfree(skb) in skb_release_data(), it's done by callback we have made for skb. >> >>No I'm not suggesting you set it to NULL. It should have some >>memory allocated, but skb_headlen(skb) should be zero. >> >>Please have a look at how the napi_gro_frags interface works (e.g., >>in drivers/net/cxgb3/sge.c). This is exactly the model that I am >>suggesting. >> >>Cheers, >>-- >>Visit Openswan at http://www.openswan.org/ >>Email: Herbert Xu ~{PmV>HI~} >>Home Page: http://gondor.apana.org.au/~herbert/ >>PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt > >Herbert, >I explained what I think the thought in your mind here, please clarify if >something missed. > >1) Modify driver from netdev_alloc_skb() to alloc user pages if dev is zero-copyed. > If the driver support PS mode, then modify alloc_page() too. >2) Add napi_gro_frags() in driver to receive the user pages instead of driver's receiving >function. >3) napi_gro_frags() will allocate small skb and pull the header data from >the first page to skb->data. > >Is above the way what you have suggested? >I have thought something in detail about the way. > >1) The first page will have an offset after the header is copied into allocated kernel skb. >The offset should be recalculated when the user page data is transferred to guest. This >may modify some of the gro code. > >2) napi_gro_frags() may remove a page when it's data is totally be pulled, but we cannot >put a user page as normally. This may modify the gro code too. > >3) When the user buffer returned to guest, some of them need to be appended a vnet header. >That means for some pages, the vnet header room should be reserved when allocated. >But we cannot know which one will be used as the first page when allocated. If we reserved >vnet header for each page, since the set_skb_frag() in guest driver only use the offset 0 for >second pages, then page data will be wrong. > >4) Since the user buffer pages should be released, so we still need a dtor callback to do that, >and then I still need a place to hold it. How do you think about to put it in skb_shinfo? > >Currently I can only think of this. >How do you think about then? > >Thanks >Xiaohui Herbert, In this way, I think we should create 3 functions at least in drivers to allocate rx buffer, to receive the rx buffers, and to clean the rx buffers. We can also have another way here. We can provide a function to only substitute alloc_page(), and a function to release the pages when cleaning the rx buffers. The skb for the rx buffer can be allocated in original way, and when pushing the data to guest, the header data will be copied to guest buffer. In this way, we should reserve sufficient room for the header in the first guest user buffers. That need modifications to guest virtio-net kernel. And this way only suitable for PS mode supported driver. Considered the advanced driver mostly has PS mode. So it should be not a critical issue. Thanks Xiaohui -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/