Re: [RFC] remove arch/sh?

[Rich Felker]

On Fri, Jul 05, 2019 at 10:51:55PM +0900, Yoshinori Sato wrote:
> On Thu, 27 Jun 2019 00:38:21 +0900,
> Rich Felker wrote:
> > 
> > On Wed, Jun 26, 2019 at 08:25:20PM +0900, Yoshinori Sato wrote:
> > > On Wed, 26 Jun 2019 00:48:09 +0900,
> > > Arnd Bergmann wrote:
> > > > 
> > > > On Tue, Jun 25, 2019 at 4:28 PM Rich Felker <dalias@xxxxxxxx> wrote:
> > > > > On Tue, Jun 25, 2019 at 02:50:01PM +0200, Arnd Bergmann wrote:
> > > > > > don't build, or are incomplete and not worked on for a long
> > > > > > time, compared to the bits that are known to work and that someone
> > > > > > is still using or at least playing with.
> > > > > > I guess a lot of the SoCs that have no board support other than
> > > > > > the Hitachi/Renesas reference platform can go away too, as any products
> > > > > > based on those boards have long stopped updating their kernels.
> > > > >
> > > > > My intent here was always, after getting device tree theoretically
> > > > > working for some reasonable subset of socs/boards, drop the rest and
> > > > > add them back as dts files (possibly plus some small drivers) only if
> > > > > there's demand/complaint about regression.
> > > > 
> > > > Do you still think that this is a likely scenario for the future though?
> > > > 
> > > > If nobody's actively working on the DT support for the old chips and
> > > > this is unlikely to change soon, removing the known-broken bits earlier
> > > > should at least make it easier to keep maintaining the working bits
> > > > afterwards.
> > > > 
> > > > FWIW, I went through the SH2, SH2A and SH3 based boards that
> > > > are supported in the kernel and found almost all of them to
> > > > be just reference platforms, with no actual product ever merged.
> > > > IIRC the idea back then was that users would supply their
> > > > own board files as an add-on patch, but I would consider all the
> > > > ones that did to be obsolete now.
> > > > 
> > > > HP Jornada 6xx is the main machine that was once supported, but
> > > > given that according to the defconfig file it only comes with 4MB
> > > > of RAM, it is unlikely to still boot any 5.x kernel, let alone user
> > > > space (wikipedia claims there were models with 16MB of RAM,
> > > > but that is still not a lot these days).
> > > > 
> > > > "Magicpanel" was another product that is supported in theory, but
> > > > the google search showed the 2007 patch for the required
> > > > flash storage driver that was never merged.
> > > > 
> > > > Maybe everything but J2 and SH4(a) can just get retired?
> > > > 
> > > >      Arnd
> > > 
> > > I also have some boards, so it's possible to rewrite more.
> > > I can not rewrite the target I do not have, so I think that
> > > there is nothing but to retire.
> > 
> > To clarify, are you agreeing with Arnd's suggestion to retire/remove
> > everything but jcore and sh4[a]?
> > 
> > Rich
> 
> I have SH2/2A/3 target board.
> So can mantain CPU support.
> But with board support it will be difficult.
> I would like to make the transition to a common framework.
> I also have to fix the parts that depend on each board for migration,
> so I would like to limit the target for maintenance to only those
> that can be used now.

Do you still have a working version of your device tree patches that
applies to current kernel? If not, could I post the forward-ported
versions I have right now (they're not up to current kernel but newer)
for you to take a look and see what might be wrong?

Your original version with the kernel version it applied to worked,
but my forward-port one doesn't. PCI is crashing during boot with the
qemu-emulated board, so I can't get disks or network, and I eventually
got frustrated trying to fix it and set it aside.

Rich