From andrew.rybchenko at oktetlabs.ru Mon Sep 15 08:17:13 2025 From: andrew.rybchenko at oktetlabs.ru (Andrew Rybchenko) Date: Mon, 15 Sep 2025 09:17:13 +0300 Subject: Interesting unused file in repo with bridging trace Message-ID: Hi, diving into Grout I found very interesting file modules/infra/control/xconnect.c which contains trace of bridging support (bridge_domain, GR_L2_MODE_SET, gr_l2_iface_mode_req). I'm really interested in this topic and would like to know if you have plans on it. Do you have any WIP which you could share? This bits already give some hints on how it could be added etc, but I'd like to know more if it is possible. Yes, the file is not actually used.? It is not built. Andrew. From cfontain at redhat.com Mon Sep 15 09:40:24 2025 From: cfontain at redhat.com (Christophe Fontaine) Date: Mon, 15 Sep 2025 09:40:24 +0200 Subject: Interesting unused file in repo with bridging trace In-Reply-To: References: Message-ID: <39F7BB09-2114-48F6-8A1A-D958DEC37CF7@redhat.com> > On 15 Sep 2025, at 08:17, Andrew Rybchenko wrote: > > Hi, > > diving into Grout I found very interesting file modules/infra/control/xconnect.c > which contains trace of bridging support (bridge_domain, GR_L2_MODE_SET, > gr_l2_iface_mode_req). I'm really interested in this topic and would like to know > if you have plans on it. Do you have any WIP which you could share? > This bits already give some hints on how it could be added etc, but I'd like to > know more if it is possible. Hi Andew, I started to work on the boiler plate code [1], in a very rough (aka non working) state, just to understand how much work is needed to get a complete feature. ?> I may rewrite that completely when I?ll actually test it. Using a hash table may not be the most efficient for a mac learning switch, and using the lib rte_member may be better for that purpose, emulating a TCAM. I?ll try both implementations, the deciding factor being how many cycles per packet this will consume ! Here are the pending design decisions: - how to handle packets to a routing interface in a bridge - how does this bridge mode integrate in a potential VXLAN implementation, or in EVPN scenarios HTH, Christophe [1] https://github.com/DPDK/grout/commit/fe8704bd4afed254a20f06538a5098df1d524f3d > > Yes, the file is not actually used. It is not built. Indeed, that file isn?t included, my bad: I removed it as part of the commit. > > Andrew. > From andrew.rybchenko at oktetlabs.ru Mon Sep 15 11:32:49 2025 From: andrew.rybchenko at oktetlabs.ru (Andrew Rybchenko) Date: Mon, 15 Sep 2025 12:32:49 +0300 Subject: Interesting unused file in repo with bridging trace In-Reply-To: <39F7BB09-2114-48F6-8A1A-D958DEC37CF7@redhat.com> References: <39F7BB09-2114-48F6-8A1A-D958DEC37CF7@redhat.com> Message-ID: <7f6f868f-de1a-4d3c-9bb1-4569fce7f8ce@oktetlabs.ru> On 9/15/25 10:40 AM, Christophe Fontaine wrote: > > >> On 15 Sep 2025, at 08:17, Andrew Rybchenko wrote: >> >> Hi, >> >> diving into Grout I found very interesting file modules/infra/control/xconnect.c >> which contains trace of bridging support (bridge_domain, GR_L2_MODE_SET, >> gr_l2_iface_mode_req). I'm really interested in this topic and would like to know >> if you have plans on it. Do you have any WIP which you could share? >> This bits already give some hints on how it could be added etc, but I'd like to >> know more if it is possible. > > Hi Andew, > > I started to work on the boiler plate code [1], in a very rough (aka non working) state, just to understand how much work is needed to get a complete feature. > > ?> I may rewrite that completely when I?ll actually test it. > Using a hash table may not be the most efficient for a mac learning switch, and using the lib rte_member may be better for that purpose, emulating a TCAM. > I?ll try both implementations, the deciding factor being how many cycles per packet this will consume ! > > Here are the pending design decisions: > - how to handle packets to a routing interface in a bridge > - how does this bridge mode integrate in a potential VXLAN implementation, or in EVPN scenarios Many thanks for the prompt reply. It is very helpful. Andrew. > > HTH, > Christophe > > [1] https://github.com/DPDK/grout/commit/fe8704bd4afed254a20f06538a5098df1d524f3d > > > >> >> Yes, the file is not actually used. It is not built. > > Indeed, that file isn?t included, my bad: I removed it as part of the commit. > >> >> Andrew. >> >