On Tue, Sep 20, 2011 at 02:59:18PM -0400, Andy Gospodarek wrote:

On Tue, Sep 20, 2011 at 01:10:01AM +0100, Ben Hutchings wrote:

quoted

On Mon, 2011-09-19 at 16:42 -0700, Stephen Hemminger wrote:

quoted

I would like to see a general solution to allow configuring
napi weight. The Rx weight isn't easily configurable either.

Indeed.

quoted

Probably needs to be through ethtool callback since actual value range
and dev -> napi relationship is device specific.

The maximum meaningful value is device specific but I'm not sure that
really matters.

And as David said it's really a many-to-one mapping of queue -> NAPI.
At netconf we talked about having 'irq' as an attribute of each queue
but maybe we should expose NAPI contexts through sysfs and make queues
refer to them instead.  NAPI contexts would be named (in the same way as
the corresponding IRQ handlers) and have irq, weight, etc.

(Still short of time to work on this myself, alas.)

I've been having a similar discussion with Neil Horman about how we can
better control all sorts of device values (interrupts, number of queues,
queue to node mapping, etc.).  Some of this is based on what I would
like to see and some is from Stephen's talk at LPC two weeks ago.  I
think the sysfs work Neil has done and posted to lkml can easily be
expanded to allow enhanced configuration of each device.  Having napi
weight in there too seems like a reasonable addition to this.

This is the work Andy is referring to for those interested:
http://marc.info/?l=linux-kernel&m=131644727521409&w=2

This version has Gregs Ack, and is waiting for an Ack from Jesse Barnes at the
moment.  I think Andy's probably right, theres room here for expansion to create
a relationship between a given interrupt and a napi wieght.  I expect what would
be most direct would be adding a napi_weight attribute that was conditional on
the class of the pci device allocating the irqs (make it visible for class 0x200
devs, invisible for others).

Thoughts?
Neil