UDP Tx and order of delivery considerations

["Mohan B.G." <mohan9271@xxxxxxxxx>]

Hi,

I am developing a UDP based networking app for an embedded system on a
multicore CPU. I would like to understand the ordering of delivery in
the Tx path as we have unique requirements that packets be on the wire
in the chronological order of send()/write() calls.

1) If UDP corking is not enabled and ARP is already resolved, does
send()/write() return only after the packet is queued in the ethernet
driver?
 * I have analyzed the code paths and this seems to be the case but I
would like to double check to make sure if I haven't missed some
paths.

2) Process A on core0 does a send()/write() at t0 and Process B on the
same core does another send()/write() at later point in time, is it
guaranteed that the packets are delivered on the wire in the order of
send() calls?

3) Process A on core0 does a send()/write() at t0 and Process B on a
different core does another send()/write() at later point in time, are
there any simple tricks to guarantee order of delivery?
  * I am currently planning to implement a FIFO queue for
serialization where sends() across the cores are buffered and drained
from a single process.

The packets are delivered to a custom chip on the same board with a
direct SGMII connection from the CPU and packet loss is a non-issue
for us by design. Also as this is a point-to-point connection out of
order delivery through routers does not arise for us.

Thanks,
Mohan
--
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html