On 10/17/2024 2:38 AM, Stephen Hemminger wrote: > On Sun, 13 Oct 2024 16:59:47 +0300 > Gur Stavi <gur.st...@huawei.com> wrote: > >> When creating AF_PACKET socket with specified protocol it is >> immediately implicitly bound to any existing interface and >> becomes RUNNING. Calling bind on such socket is affectively unbind >> from "any interface", then bind to the specific interface. >> >> When creating socket with 0 as protocol, it is created in non-RUNNING >> state, then it can be bound to interface and protocol in a single bind >> call and switch to RUNNING state. >> >> Especially with ETH_P_ALL, binding to any interface is not a good idea. >> It is safer and faster to use the 2nd approach. >> >> This patch replaces protocol in socket creation from ETH_P_ALL to 0. >> >> Signed-off-by: Gur Stavi <gur.st...@huawei.com> > > This makes sense see packet(7) man page: > > By default, all packets of the specified protocol type are passed to > a > packet socket. To get packets only from a specific interface > use > bind(2) specifying an address in a struct sockaddr_ll to bind the > packet > socket to an interface. Fields used for binding are sll_family > (should > be AF_PACKET), sll_protocol, and sll_ifindex. > > So there is a small window where the packet socket could pick up junk before > the bind from other interfaces. >
Thanks, this answers some of my questions in this thread, also I did some experiment and verified the same. I agree this change is more close the intention of the driver (driver is not to get packets from all interfaces), hence: Acked-by: Ferruh Yigit <ferruh.yi...@amd.com>
