Ecovec in current sh-latest

Ecovec in current sh-latest

[Guennadi Liakhovetski]

Hi

To get ecovec to run with today's sh-latest (pretty much the same with 
Linux' tree) the following patches have to be reverted:

commit 794d78fea51504bad3880d14f354a9847f318f25
Author: Magnus Damm <damm@opensource.se>
Date:   Tue Jun 21 07:55:12 2011 +0000

    drivers: sh: late disabling of clocks V2

commit 3f9b8520b06013939ad247ba08b69529b5f14be1
Author: Paul Mundt <lethal@linux-sh.org>
Date:   Tue May 31 14:38:29 2011 +0900

    sh64: Move from P1SEG to CAC_ADDR for consistent sync.

commit cb3da5b837a410a1fb90019549b18178f3c87258
Author: Magnus Damm <damm@opensource.se>
Date:   Wed Aug 3 12:47:36 2011 +0900

    serial: sh-sci: console Runtime PM support

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/

Re: Ecovec in current sh-latest

[Paul Mundt]

On Fri, Sep 02, 2011 at 10:17:54AM +0200, Guennadi Liakhovetski wrote:
> To get ecovec to run with today's sh-latest (pretty much the same with 
> Linux' tree) the following patches have to be reverted:
> 
> commit 794d78fea51504bad3880d14f354a9847f318f25
> Author: Magnus Damm <damm@opensource.se>
> Date:   Tue Jun 21 07:55:12 2011 +0000
> 
>     drivers: sh: late disabling of clocks V2
> 
Do we know which driver that's the problem for the ecovec case?

We can of course pile on another CLK_ENABLE_ON_INIT workaround for 3.1,
so long as there's a plan to get the driver properly fixed for 3.2.

> commit 3f9b8520b06013939ad247ba08b69529b5f14be1
> Author: Paul Mundt <lethal@linux-sh.org>
> Date:   Tue May 31 14:38:29 2011 +0900
> 
>     sh64: Move from P1SEG to CAC_ADDR for consistent sync.
> 
I'll queue a revert for this for 3.1. There's quite a bit of bogosity
going on in some of the drivers that this exposed, but we're not going to
get it all fixed in one shot, either.

> commit cb3da5b837a410a1fb90019549b18178f3c87258
> Author: Magnus Damm <damm@opensource.se>
> Date:   Wed Aug 3 12:47:36 2011 +0900
> 
>     serial: sh-sci: console Runtime PM support
> 
What's the issue with this one? I'm assuming this and the clock disabling
revert are related.

Re: Ecovec in current sh-latest

[Guennadi Liakhovetski]

On Mon, 5 Sep 2011, Paul Mundt wrote:

> On Fri, Sep 02, 2011 at 10:17:54AM +0200, Guennadi Liakhovetski wrote:
> > To get ecovec to run with today's sh-latest (pretty much the same with 
> > Linux' tree) the following patches have to be reverted:
> > 
> > commit 794d78fea51504bad3880d14f354a9847f318f25
> > Author: Magnus Damm <damm@opensource.se>
> > Date:   Tue Jun 21 07:55:12 2011 +0000
> > 
> >     drivers: sh: late disabling of clocks V2
> > 
> Do we know which driver that's the problem for the ecovec case?

I see at least two drivers misbehaving: ethernet is failing to perform 
dhcp for nfs root-mount, and LCD gets an image at boot and then slowly 
fade out.

> We can of course pile on another CLK_ENABLE_ON_INIT workaround for 3.1,
> so long as there's a plan to get the driver properly fixed for 3.2.
> 
> > commit 3f9b8520b06013939ad247ba08b69529b5f14be1
> > Author: Paul Mundt <lethal@linux-sh.org>
> > Date:   Tue May 31 14:38:29 2011 +0900
> > 
> >     sh64: Move from P1SEG to CAC_ADDR for consistent sync.
> > 
> I'll queue a revert for this for 3.1. There's quite a bit of bogosity
> going on in some of the drivers that this exposed, but we're not going to
> get it all fixed in one shot, either.
> 
> > commit cb3da5b837a410a1fb90019549b18178f3c87258
> > Author: Magnus Damm <damm@opensource.se>
> > Date:   Wed Aug 3 12:47:36 2011 +0900
> > 
> >     serial: sh-sci: console Runtime PM support
> > 
> What's the issue with this one? I'm assuming this and the clock disabling
> revert are related.

IIRC, the boot process stops after

console [ttySC0] enabled

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/

Re: Ecovec in current sh-latest

[Kuninori Morimoto]

Hi Magnus, Paul

> > > To get ecovec to run with today's sh-latest (pretty much the same with 
> > > Linux' tree) the following patches have to be reverted:
> > > 
> > > commit 794d78fea51504bad3880d14f354a9847f318f25
> > > Author: Magnus Damm <damm@opensource.se>
> > > Date:   Tue Jun 21 07:55:12 2011 +0000
> > > 
> > >     drivers: sh: late disabling of clocks V2
> > > 
> > Do we know which driver that's the problem for the ecovec case?
> 
> I see at least two drivers misbehaving: ethernet is failing to perform 
> dhcp for nfs root-mount, and LCD gets an image at boot and then slowly 
> fade out.

My renesas_usbhs driver breaks too in SH7724 ecovec board.
So I debuged it.

when my driver call pm_runtime_get_sync(),
then, arch/sh/kernel/cpu/hwblk.c :: hwblk_enable() are called,
and it enabled my driver's MSTP bit.

But clk_late_init() disabled it.

Does hwblk/pm_runtime (?) didn't care clk->usecount ?

Best regards
---
Kuninori Morimoto

Re: Ecovec in current sh-latest

[Kuninori Morimoto]

Hi Magnus, Paul

> > > To get ecovec to run with today's sh-latest (pretty much the same with 
> > > Linux' tree) the following patches have to be reverted:
> > > 
> > > commit 794d78fea51504bad3880d14f354a9847f318f25
> > > Author: Magnus Damm <damm@opensource.se>
> > > Date:   Tue Jun 21 07:55:12 2011 +0000
> > > 
> > >     drivers: sh: late disabling of clocks V2
> > > 
> > Do we know which driver that's the problem for the ecovec case?
> 
> I see at least two drivers misbehaving: ethernet is failing to perform 
> dhcp for nfs root-mount, and LCD gets an image at boot and then slowly 
> fade out.

My renesas_usbhs driver breaks too in SH7724 ecovec board.
So I debuged it.

when my driver call pm_runtime_get_sync(),
then, arch/sh/kernel/cpu/hwblk.c :: hwblk_enable() are called,
and it enabled my driver's MSTP bit.

But clk_late_init() disabled it.

Does hwblk/pm_runtime (?) didn't care clk->usecount ?

Best regards
---
Kuninori Morimoto