On Thu, Jan 19, 2017 at 1:23 PM, Ryota Ozaki <[email protected]> wrote: > On Wed, Jan 18, 2017 at 6:25 PM, Martin Husemann <[email protected]> wrote: >> On Wed, Jan 18, 2017 at 04:48:32PM +0900, Ryota Ozaki wrote: >>> Note that it of course consumes more CPU time for softint >>> and more memory for m_dup. That said, I think that this is >>> a suitable workaround for now until someone makes them >>> softint-based. >> >> Could you provide some TODO like document (e.g. on the wiki) where >> the poor soul wading into ISDN drivers (i.e.) could find out what >> exactly needs to be done to the drivers? (Even if this is just two >> or three sentences, it should be spelled out explicitly.) > > I added doc/TODO.smpnet for it. It will include TODOs of MP-safe > network stack. > > Updated the patch: http://www.netbsd.org/~ozaki-r/bpf_mtap_softint.diff > >> >> Did I understand correctly that you will add the defered mtap to all >> drivers for now, and after the driver has been fixed it will return to >> old bpf_mtap? > > Yes. > >> So if we ask nxr for users of the defered tap and find none, >> we know that the task is done? > > Yes, we can know it by searching bpf_mtap_softint.
I added KASSERT(cpu_intr_p()) to bpf_mtap_softint, so it triggers once we change a remaining component to use softint on Rx. ozaki-r
