Re: [PATCH 1/2] wan/hdlc_x25: make lapb params configurable

[Martin Schiller]

On 2020-01-14 13:51, Jakub Kicinski wrote:
> On Tue, 14 Jan 2020 06:37:03 +0100, Martin Schiller wrote:
> > >> diff --git a/include/uapi/linux/hdlc/ioctl.h
> > >> b/include/uapi/linux/hdlc/ioctl.h
> > >> index 0fe4238e8246..3656ce8b8af0 100644
> > >> --- a/include/uapi/linux/hdlc/ioctl.h
> > >> +++ b/include/uapi/linux/hdlc/ioctl.h
> > >> @@ -3,7 +3,7 @@
> > >>  #define __HDLC_IOCTL_H__
> > >>
> > >>
> > >> -#define GENERIC_HDLC_VERSION 4	/* For synchronization with sethdlc
> > >> utility */
> > >> +#define GENERIC_HDLC_VERSION 5	/* For synchronization with sethdlc
> > >> utility */
> > >
> > > What's the backward compatibility story in this code?
> >
> > Well, I thought I have to increment the version to keep the kernel 
> > code
> > and the sethdlc utility in sync (like the comment says).
>
> Perhaps I chose the wrong place for asking this question, IOCTL code
> was my real worry. I don't think this version number is validated so
> I think bumping it shouldn't break anything?

sethdlc validates the GENERIC_HDLC_VERSION at compile time.

https://mirrors.edge.kernel.org/pub/linux/utils/net/hdlc/

Another question:
Where do I have to send my patch for sethdlc to?

> > > The IOCTL handling at least looks like it may start returning errors
> > > to existing user space which could have expected the parameters to
> > > IF_PROTO_X25 (other than just ifr_settings.type) to be ignored.
> >
> > I could also try to implement it without incrementing the version by
> > looking at ifr_settings.size and using the former defaults if the size
> > doesn't match.
>
> Sounds good, thank you!

OK, I will send a v2 of the patch.