From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:51940) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1d6lOu-0005yb-VB for qemu-devel@nongnu.org; Fri, 05 May 2017 18:08:38 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1d6lOq-0008EU-Gk for qemu-devel@nongnu.org; Fri, 05 May 2017 18:08:36 -0400 Received: from mx1.redhat.com ([209.132.183.28]:60622) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1d6lOq-0008AI-82 for qemu-devel@nongnu.org; Fri, 05 May 2017 18:08:32 -0400 Date: Sat, 6 May 2017 01:08:30 +0300 From: "Michael S. Tsirkin" Message-ID: <20170506010644-mutt-send-email-mst@kernel.org> References: <286AC319A985734F985F78AFA26841F7391FDD30@shsmsx102.ccr.corp.intel.com> <056500d7-6a91-12e5-be1d-2b2beebd0430@redhat.com> <590C1353.7070501@intel.com> <6b96612b-2fd9-cf65-023e-f72561ec936a@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <6b96612b-2fd9-cf65-023e-f72561ec936a@redhat.com> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [virtio-dev] Re: virtio-net: configurable TX queue size List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Jason Wang Cc: Wei Wang , Stefan Hajnoczi , =?iso-8859-1?Q?Marc-Andr=E9?= Lureau , "pbonzini@redhat.com" , "virtio-dev@lists.oasis-open.org" , "qemu-devel@nongnu.org" , Jan Scheurich On Fri, May 05, 2017 at 05:20:07PM +0800, Jason Wang wrote: >=20 >=20 > On 2017=E5=B9=B405=E6=9C=8805=E6=97=A5 13:53, Wei Wang wrote: > > On 05/05/2017 10:27 AM, Jason Wang wrote: > > >=20 > > >=20 > > > On 2017=E5=B9=B405=E6=9C=8804=E6=97=A5 18:58, Wang, Wei W wrote: > > > > Hi, > > > >=20 > > > > I want to re-open the discussion left long time ago: > > > > https://lists.gnu.org/archive/html/qemu-devel/2015-11/msg06194.ht= ml > > > > , and discuss the possibility of changing the hardcoded (256) TX = queue > > > > size to be configurable between 256 and 1024. > > >=20 > > > Yes, I think we probably need this. > >=20 > > That's great, thanks. > >=20 > > >=20 > > > >=20 > > > > The reason to propose this request is that a severe issue of > > > > packet drops in > > > > TX direction was observed with the existing hardcoded 256 queue s= ize, > > > > which causes performance issues for packet drop sensitive guest > > > > applications that cannot use indirect descriptor tables. The > > > > issue goes away > > > > with 1K queue size. > > >=20 > > > Do we need even more, what if we find 1K is even not sufficient in > > > the future? Modern nics has size up to ~8192. > >=20 > > Yes. Probably, we can also set the RX queue size to 8192 (currently i= t's > > 1K) as well. > >=20 > > >=20 > > > >=20 > > > > The concern mentioned in the previous discussion (please check th= e link > > > > above) is that the number of chained descriptors would exceed > > > > UIO_MAXIOV (1024) supported by the Linux. > > >=20 > > > We could try to address this limitation but probably need a new > > > feature bit to allow more than UIO_MAXIOV sgs. > >=20 > > I think we should first discuss whether it would be an issue below. > >=20 > > >=20 > > > >=20 > > > > From the code, I think the number of the chained descriptors > > > > is limited to > > > > MAX_SKB_FRAGS + 2 (~18), which is much less than UIO_MAXIOV. > > >=20 > > > This is the limitation of #page frags for skb, not the iov limitati= on. > >=20 > > I think the number of page frags are filled into the same number of > > descriptors > > in the virtio-net driver (e.g. use 10 descriptors for 10 page frags).= On > > the other > > side, the virtio-net backend uses the same number of iov for the > > descriptors. > >=20 > > Since the number of page frags is limited to 18, I think there wouldn= 't > > be more > > than 18 iovs to be passed to writev, right? >=20 > Looks not, see skb_copy_datagram_from_iter(). >=20 > Thanks Besides, guests don't all use linux drivers. Some use e.g. dpdk pmd ones. It's best not to make assumptions outside the spec - if you need to make an assumption, it's best to add a flag to specify it. > >=20 > > Best, > > Wei > >=20 > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: virtio-dev-unsubscribe@lists.oasis-open.org > > For additional commands, e-mail: virtio-dev-help@lists.oasis-open.org > >=20