ixgbe & msi/x

[Vijay Singh]

> that was my case too. I have not gone too far into my investigation but
> should
> note that not _all_ interrupts were lost; my symptoms were queue overflows
> under netmap even at a low 2 Mpps, which with 2k entries in the rx ring
> means
> that the interrupt was delayed for more than 1ms, well above the moderation
> delay.

This would be consistent with what I am seeing. I saw that vmstat -i
reported some interrupt rate for the rx rings but even a simple ping
at that point would lead to input errors - queue overflows.


> So just to clarify, which one of these symptoms did you see
> 1) no rx interrupts at all at any rx rate
> 2) occasional missing interrupts/drops as the rx pps increase
> 3) complete loss of rx interrupts above some pps threshold ?

I think it would be closest to 3. The same HW runs fine when I disable
msi/x and use legacy interrupts.

-vijay