[PATCH v3] ARM: multi_v7_defconfig: Enable shmobile platforms

[Arnd Bergmann]

On Wednesday 04 February 2015 11:12:36 Geert Uytterhoeven wrote:
> Hi Arnd,
> 
> On Wed, Jan 21, 2015 at 2:15 PM, Arnd Bergmann <arnd at arndb.de> wrote:
> > On Wednesday 21 January 2015 14:09:56 Geert Uytterhoeven wrote:
> >> On Wed, Jan 21, 2015 at 2:02 PM, Laurent Pinchart
> >> <laurent.pinchart at ideasonboard.com> wrote:
> >> > To avoid bloating the kernel needlessly, couldn't at least some of the drivers
> >> > needed for the shmobile boards be compiled as modules ? Audio, USB, SATA,
> >> > touchscreen are possible candidates, depending on what we need to boot the
> >> > boards to userspace.
> >>
> >> That's a possibility.
> >>
> >> Note that currently there are almost no modules in multi_v7_defconfig:
> >> CONFIG_INET6_AH=m
> >> CONFIG_INET6_ESP=m
> >> CONFIG_INET6_IPCOMP=m
> >> CONFIG_IPV6_MIP6=m
> >> CONFIG_IPV6_TUNNEL=m
> >> CONFIG_CFG80211=m
> >> CONFIG_MAC80211=m
> >> CONFIG_BRCMFMAC=m
> >> CONFIG_RT2X00=m
> >> CONFIG_RT2800USB=m
> >>
> >> Arnd, Olof: What do you prefer?
> >
> > I think it would be nice to move to having more loadable modules. We
> > haven't really asked people to do that in the past, but the size is
> > starting to become a problem and we may end up having to change a lot
> > of the '=y' to '=m' in the future.
> 
> Do you want me to respin this patch with some options as modules?

I think it would be a good start, and I'll ask others to do the same
in the future/

> One disadvantage of that is that it lets us discover less runtime-conflicts
> between platforms using multi_v7_defconfig

Good point. My feeling is that the bloat we get in the kernel image
when everything is built-in weighs more though. 

> (does anyone boot arm allyesconfig on his farm?).

It's currently impossible to run allmodconfig on real hardware, there
are a couple of options that prevent you from doing this.

allyesconfig cannot currently be built even, mostly due to bad interaction
of the way we link the kernel with how the linker is implemented: we
get lots of failed relocations for cross-section calls larger than
32MB away.

	Arnd