Re: [PATCH v5 4/7] can: Kconfig: add CONFIG_CAN_RX_OFFLOAD

[Vincent MAILHOL]

On Wed. 8 juin 2022 at 09:14, Jakub Kicinski <kuba@xxxxxxxxxx> wrote:
> On Wed, 8 Jun 2022 01:43:54 +0200 Max Staudt wrote:
> > It seems strange to me to magically build some extra features into
> > can_dev.ko, depending on whether some other .ko files are built in that
> > very same moment, or not. By "magically", I mean an invisible Kconfig
> > option. This is why I think Vincent's approach is best here, by making
> > the drivers a clearly visible subset of the RX_OFFLOAD option in
> > Kconfig, and RX_OFFLOAD user-selectable.
>
> Sorry for a chunked response, vger becoming unresponsive the week after
> the merge window seems to become a tradition :/
>
> We have a ton of "magical" / hidden Kconfigs in networking, take a look
> at net/Kconfig. Quick grep, likely not very accurate but FWIW:
>
> # not-hidden
> $ git grep -c -E '(bool|tristate)..' net/Kconfig
> net/Kconfig:23
>
> # hidden
> $ git grep -c -E '(bool|tristate)$' net/Kconfig
> net/Kconfig:20

OK. So we have a proposal to make CAN_RX_OFFLOAD an hidden
configuration. I did not consider this approach before because the CAN
subsystem *never* relies on this and I did not really explore other
Kbuild files.
| $ git grep -c -E '(bool|tristate)$' net/can/Kconfig
| <no output>

Before pushing my driver upstream, it was also an out of tree module
for about one year and I relate a lot to what Max said. But Jakub
explanations are consistent and reflect the best practices of the
kernel development.
Also, mainstream distribution would do an allyesconfig and ship the
can-dev.ko with everything built in. So the lambda user would still
have everything built-in.

I will let people continue to comment for a couple of days before
making the final choice and sending the next version. But so far, I am
leading toward Jakub’s idea to make it a hidden feature.

> > How about making RX_OFFLOAD a separate .ko file, so we don't have
> > various possible versions of can_dev.ko?
> >
> > @Vincent, I think you suggested that some time ago, IIRC?
> >
> > (I know, I was against a ton of little modules, but I'm changing my
> > ways here now since it seems to help...)
>
> A separate module wouldn't help with my objections, I don't think.