On Wed, Jun 07, 2006 at 11:05:22AM -0400, Neil Horman wrote: > > > > > Matt, any ideas on this? > > > > Not at the moment. > > how about this for a solution? It doesn't make netpoll any more robust, but I > think in the interests of efficiency it would be fair to require that, when > netpolled, a driver must receive frames on the same net device for which it > was > polled. With this patch we detect that condition and handle it accordingly in > e1000_intr. This eliminates the need for us to call the clean_rx method from > the poll_controller when napi is configured, instead allowing the poll method > to > be called from napi_poll, as the netpoll model currently does. This fixes the > netdump regression, and eliminates the layering violation and the potential > race > that we've been discussing. I've just tested it with netdump here and it > works > quite well. > > Thoughts appreciated.
This looks pretty reasonable, mostly from the perspective that it doesn't put any further ugliness in netpoll. We might want to add a comment somewhere in netpoll of the new rule we're now observing. I'll let the e1000 guys comment on the particulars of the driver change. > Signed-off-by: Neil Horman <[EMAIL PROTECTED]> Signed-off-by: Matt Mackall <[EMAIL PROTECTED]> -- Mathematics is the supreme nostalgia of our time. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
