Sebastien Roy wrote:
> ip.c:
> 
> 12771: if we hit this condition (IP and UDP headers in separate mblks), 
> it looks like we'll never check for ESP-in-UDP because we'll never call 
> ip_udp_check().  We'll be kicked over to the "slow path" and directly to 
> ip_fanout_udp() without passing go.

Hmm, I think I see where we eventually catch these "slow path" packets. 
Eventually, we end up in ip_fanout_udp_conn(), which then calls 
zero_spi_check() to take care of ESP-in-UDP.  The ip module's 
multiple-personality disorder when it comes to "fast" and "slow" paths is 
just plain confusing, but that's not this project's problem.

-Seb
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to