Re: add SPP_PLPMTUD_ENABLE/DISABLE flag for spp_flags

[Marcelo Ricardo Leitner <mleitner@redhat.com>]

On Thu, May 20, 2021 at 01:16:38AM +0200, Michael Tuexen wrote:
> > On 20. May 2021, at 00:44, mleitner@redhat.com wrote:
> >
> > On Wed, May 19, 2021 at 02:44:20PM -0400, Xin Long wrote:
> >> On Wed, May 19, 2021 at 2:15 PM Michael Tuexen <tuexen@freebsd.org> wrote:
> >>>
> >>>> On 19. May 2021, at 18:18, Xin Long <lucien.xin@gmail.com> wrote:
> >>>>
> >>>> On Tue, May 18, 2021 at 2:33 PM Xin Long <lucien.xin@gmail.com> wrote:
> >>>>>
> >>>>> On Tue, May 18, 2021 at 1:38 PM Michael Tuexen <tuexen@freebsd.org> wrote:
> >>>>>>
> >>>>>>> On 18. May 2021, at 18:43, Xin Long <lucien.xin@gmail.com> wrote:
> >>>>>>>
> >>>>>>> Hi, Michael,
> >>>>>>>
> >>>>>>> We're implementing RFC8899 (PLPMTUD) on Linux SCTP recently,
> >>>>>>> and to make this be controlled by setsockopt with
> >>>>>>> SCTP_PEER_ADDR_PARAMS, as in
> >>>>>>>
> >>>>>>> https://datatracker.ietf.org/doc/html/rfc6458#section-8.1.12:
> >>>>>>>
> >>>>>>> we need another two flags to add for spp_flags:
> >>>>>>>
> >>>>>>> SPP_PLPMTUD_ENABLE
> >>>>>>> SPP_PLPMTUD_DISABLE
> >>>>>>>
> >>>>>>> Do you think it makes sense? if yes, does the RFC6458 need to update?
> >>>>>>> if not, do you have a better suggestion for it?
> >>>>>> It is great new that you want to implement RFC 8899. I plan to do the
> >>>>>> same for the FreeBSD stack.
> >>>>>>
> >>>>>> In my view, RFC 8899 is the right way to implement PMTU discovery.
> >>>>>> So I will just use the SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE. I don't
> >>>>>> think that the user needs to control which method is used.
> >>>>>> I you want to support multiple versions, I would make that
> >>>>>> controllable via a sysctl variable. But I think for FreeBSD, support
> >>>>>> for RFC 8899 will be the only way of doing PMTU discovery. There
> >>>>>> might be multiple choices for details like how to do the searching,
> >>>>>> how long to wait for some events. These will be controllable via
> >>>>>> sysctl.
> >>>>>>
> >>>>>> So in my view, there is no need to extend the socket API. What do you think?
> >>>> I just noticed that with multiple versions supported, and without extending
> >>>> this API, all applications will have to use the same version as it's
> >>>> controlled by
> >>>> sysctl. And when switching to another version by sysctl, all
> >>>> applications will be
> >>>> affected and have to do the switch. that seems not nice.
> >>> That is true, but an application can not expect any specific behaviour
> >>> right now when they are not disabling PMTUD.
> >>>
> >>> What about adding a sysctl variable, which defines the default
> >>> algorithm and a socket option, which allows to get and set
> >>> the algorithm being used.
> >> yes, that's also what I'm thinking.
> >
> > +1
> >
> >> sysctl is always used for the default value for future sockets.
> >> and the socket option should be added for a socket/asoc's setting.
> >
> > Speaking of inheritance, it should also use the SCTP_FUTURE_ASSOC /
> > SCTP_CURRENT_ASSOC / SCTP_ALL_ASSOC mechanism. Like
> > SCTP_PEER_ADDR_PARAMS, for example.
> Yepp.
> >
> > The system can provide defaults but if the application requires
> > something, it should have a good way of requesting it.
> >
> > Speaking of SCTP_PEER_ADDR_PARAMS, maybe reuse spp_pathmtu field?
> > As in, if SPP_PMTUD_ENABLE is enabled, spp_pathmtu of "1" or "2" bytes
> > doesn't make sense, and it could mean the algorithm used. Thing is,
> > the field is currently ignored, and it could lead to some unexpected
> > behavior change. It's probably safer to just add another sockopt, but
> > wanted to share the idea anyway.
> I leave it completely up to you what you implement in Linux. But I
> would prefer to use a separate socket option instead of overloading
> an existing one.

Wait. Somehow I thought we were talking about extending the RFC with
these new definitions here, no? Or at least agreeing on a common
interface. It would be beneficial for the application to be able to
use the same API on FreeBSD or Linux.

Thanks,
Marcelo

>
> Best regards
> Michael
> >
> >>
> >> SCTP_PTMUD_METHOD?
> >
> > s/PTMUD/PMTUD/ :-)
> >
> >> 0: PTB one
> >> 1. PLPMTUD
> >>
> >>>
> >>> Best regards
> >>> Michael
> >>>>
> >>>>> OK, that makes sense to me.
> >>>>>
> >>>>> Another thing I want to know your opinion on is:  do you think the HB
> >>>>> should be created
> >>>>> separately for PLPMTUD probe, instead of reusing the old HB that
> >>>>> checks the link connectivity?
> >>>>> As the HB for PLPMTUD probe might get lost, which we don't want to
> >>>>> affect the link's
> >>>>> connectivity.
> >>>>>
> >>>>>>
> >>>>>> Best regards
> >>>>>> Michael
> >>>>>>>
> >>>>>>> Thanks.
> >>>>>>
> >>>
> >>
> >
>