vhiz wrote: > I am trying to figure out the distribution of the number of packets that the > receive interrupt processes. With each interrupt, the NIC would transfer a > number of packets together, and that number would depend on the coalescing > factor. In the case of e1000g, rx interrupt coalescing is affected by the > following three parameters: > > - intr_throttling_rate, or > - the combination of rx_intr_delay and rx_intr_abs_delay > > So I thought that before I delve into the code to figure out the details, I > should first develop an understanding of how the received packets are mapped > to the mblks. Also, if someone could tell me whether I could use dtrace to > access the mblks, and figure out the number of received packets. Can this be > done for each interrupt?
I see. I doubt that how packets are mapped to mblks is really the important thing to be looking at here. I think what you want to look at is how many interrupts are taken, what work is done on each one, and what "unusual" events occur to the mblks that are allocated by this driver -- such as pullups and data copies. It sounds to me like to start this you really need an expert in that particular driver -- someone who knows how to tweak those parameters and why to do so. I'd suggest [email protected]. -- James Carlson 42.703N 71.076W <[email protected]> _______________________________________________ networking-discuss mailing list [email protected]
