On Mon, Aug 31, 2015 at 10:05 AM, David Ahern <d...@cumulusnetworks.com> wrote: > On 8/31/15 11:02 AM, Tom Herbert wrote: >> >> On Mon, Aug 31, 2015 at 9:29 AM, David Ahern <d...@cumulusnetworks.com> >> wrote: >>> >>> Remove the VRF change in udp_sendmsg to set the source address. The VRF >>> driver already has access to the packet on the TX path via the dst. It >>> can be used to update the source address in the header. >>> >> >> I don't understand this. The previous code was about selecting a >> source address for packets being sourced ed on a socket, but this new >> patch seems to essentially be doing SNAT in the VRF transmit path >> which seems like a fundamentally different behavior. Is this really >> your intention? >> > > The original code and this new code are only controlling FIB lookups which > in turn set the source address. Functionally both versions do the same > thing. > It's a major departure from current convention. The source address of the packet should be set before doing ip_send_skb. In UDP unconnected case ip_route_output_flow calls inet_select_addr. AKAIK there is no provision for not setting the source address and relying on the output device to do this in its transmit routine. I still think it would be better to call into the VRF device from inet_select_addr using an ndo function.
Tom > David > -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
