On Wed, Feb 8, 2017 at 7:40 PM, Jason Gunthorpe <jgunthorpe@xxxxxxxxxxxxxxxxxxxx> wrote: > On Wed, Feb 08, 2017 at 06:02:44PM +0200, Matan Barak wrote: > >> Actually, I think one of the problems here is that we insert GIDs >> which aren't supported by the hardware. > > That certainly sounds like a problem, yes. > >> So, assuming we only insert a combination of netdevices which could >> be actually described by the hardware, we could try to match these >> attributes in the gid cache table. Once we have such a match, we >> could verify that with the ingress route (actually, we could use >> ingress route starting from get_netdev() as well). > > Yes, that is the proper thing to do. > > Presumably the hardware does this, it should not be receiving packets > that do not match the GID table. That would be a huge problem when we > talk about namespaces/etc. > >> Since RoCE is symmetrical (you don't know if an AH is for rx/tx in >> UD and you need to select a single sgid_index for RC connections), >> we need to verify that it matches an egress route bounded to this >> netdevice. > > ?? how is an AH involved in RX? > Well, some applications (maybe a lot of them), creates an AH from the first received packet (of each node) and then just assumes the rest of the packets comes from the same logical netdev. Since there is no binding to a netdev, other packets could differ vastly from this one. That's what I meant by "symmetrical". >> As a quick fix, setting ifindex to get_netdev's ifindex in link local >> addresses and filter out unsupported >> netdevices in the GID cache, seems like a reasonable solution to me. > > That still breaks link local addresses on vlan devices, so it is an > ugly hack, not a solution. > I agree, but currently all link local addresses are broken :( > Jason Matan -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
