[openib-general] IPOIB NAPI

Eli Cohen eli at dev.mellanox.co.il
Mon Oct 16 09:59:23 PDT 2006


On Mon, 2006-10-16 at 09:48 -0700, Roland Dreier wrote:
>     Eli> Please diff to see my comments. Generaly it looks like the
>     Eli> condition on netif_rx_reschedule() should be inverted.
> 
> Why?  A return value of 0 means that the reschedule failed (probably
> because the poll routine is already running somewhere else) and the
> poll routine should just return.  I think the code is correct as it stands.
> 
>     Eli> Also ou need to set max to some large value since you don't
>     Eli> know if how many completions you missed and you want to make
>     Eli> sure you get all the ones the sneaked from the last poll to
>     Eli> the request notify.
> 
> Why?  max is there to limit us from doing more work than the quota
> passed in from the networking stack.  If we fail to drain the CQ
> because we exhaust max, then the poll routine will return 1 and will
> remain scheduled, so the networking stack will call the poll routine
> again to continue grabbing completions.
> 
>  - R.

OK I see what you mean. So I guess it's OK then.





More information about the general mailing list