From mboxrd@z Thu Jan 1 00:00:00 1970 From: Wisam Monther Subject: Re: [PATCH 1/2] app/testpmd: add custom topology command Date: Wed, 16 May 2018 12:43:11 +0000 Message-ID: References: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Cc: Raslan Darawsheh , "dev@dpdk.org" , Shahaf Shuler To: Kevin Traynor , "jingjing.wu@intel.com" , "wenzhuo.lu@intel.com" , Thomas Monjalon Return-path: Received: from EUR01-HE1-obe.outbound.protection.outlook.com (mail-he1eur01on0047.outbound.protection.outlook.com [104.47.0.47]) by dpdk.org (Postfix) with ESMTP id 05B081B1A5 for ; Wed, 16 May 2018 14:43:15 +0200 (CEST) In-Reply-To: Content-Language: en-US List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" PSB. BRs, Wisam Jaddo -----Original Message----- From: Kevin Traynor [mailto:ktraynor@redhat.com]=20 Sent: Wednesday, May 16, 2018 3:41 PM To: Wisam Monther; jingjing.wu@intel.com; wenzhuo.lu@intel.com; Thomas Monj= alon Cc: Raslan Darawsheh; dev@dpdk.org; Shahaf Shuler Subject: Re: [dpdk-dev] [PATCH 1/2] app/testpmd: add custom topology comman= d On 05/15/2018 03:05 PM, Wisam Monther wrote: > Hey Kevin, >=20 > What if I have 10 ports and want all 9 ports to forward packets to=20 > port 0, this cannot done in paired. > But with this topo the user can overwrite the topo. > E.g: > - Set custom-topo 1 0 > Will make those paired and those two as active only. >=20 > - set custom-topo 2 0 > Will make those two as paired, but still port one will forward packets=20 > to 0. > And the active ports will be 3 ports (0,1,2) And so on... > With this topo you can create your own topo. (Mixed between all topos=20 > or even non-pattern topo for any usage) In my case its needed for=20 > representors and virtio. >=20 > Any traffic coming from representors/vhost (from VM) need to be=20 > forwarded from PF/phy port to the wire. >=20 Thanks for sharing the use case. Will it be an issue that multiple cores may be sending to the same TxQ? I didn't check to see if there is locking provided in the app. [Wisam] I've tested such case recently, and yes it will cause a seg. Fault = in this solution, since multiple cores will access same queues. I'm currently working on a fix for it, I'm planning to map single queues fo= r each core with custom topology to avoid such scenario. > This new topo is useful for such cases. >=20 > BRs, > Wisam jaddo >=20