On Mon, 27 Feb 2017 17:48:54 -0500 (EST) David Miller <da...@davemloft.net> wrote:
> From: Stephen Hemminger <step...@networkplumber.org> > Date: Mon, 27 Feb 2017 14:44:55 -0800 > > > On Mon, 27 Feb 2017 14:35:17 -0800 > > Eric Dumazet <eric.duma...@gmail.com> wrote: > > > >> On Mon, 2017-02-27 at 14:14 -0800, Stephen Hemminger wrote: > >> > >> > The original design (as Davem mentioned) was that IRQ's must be disabled > >> > during device polling. If that was true, then the race above > >> > would be impossible. > >> > >> I would love to see an alternative patch. > > > > Turn off busy poll? > > The poll stuff runs risk of breaking more things. > > Eric is exactly trying to make busy poll even more prominent in > the stack, not less prominent. > > It's an important component of some performance improvements he is > working on. Maybe making IRQ controlled as part of the network device model (instead of a side effect left to device driver to handle) would be less problematic. Really just shooting in the dark because I don't have any of the problematic hardware to play with.
