[ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[Geert Uytterhoeven]

	Hi all,

This patchset enables the PM runtime code in drivers/sh/pm_runtime.c when
running a multi-platform ARM kernel including support for shmobile.
Before this code was only enabled for legacy shmobile kernels, leading to
disabled clocks in multiplatform kernels, depending on implicit reset state
or on the bootloader.
It also contains some related cleanups, and removals of workarounds.

  - [01/17] ARM: shmobile: r8a7740: Remove unused
    r8a7740_add_early_devices_dt()
  - [02/17] ARM: shmobile: r8a7740: Make r8a7740_meram_workaround() __init
  - [03/17] ARM: shmobile: sh7372: Call sh7372_add_early_devices() instead of
    open coding
  - [04/17] drivers: sh: pm_runtime does not need idle callback
  - [05/17] drivers: sh: pm_runtime implementation needs to suspend and resume
    devices
  - [06/17] sh: intc: Restrict INTC_USERIMASK to SH4A
  - [07/17] drivers: sh: compile drivers/sh/pm_runtime.c if ARCH_SHMOBILE_MULTI
  - [08/17] ARM: shmobile: Call sh_pm_runtime_init() from platform code on all
    SoCs
  - [09/17] drivers: sh: Remove automatic sh_pm_runtime_init() if
    ARCH_SHMOBILE_LEGACY
  - [10/17] ARM: shmobile: koelsch-reference: Remove workarounds for core clock
    issues
  - [11/17] ARM: shmobile: lager-reference: Remove workarounds for core clock
    issues
  - [12/17] clk: shmobile: mstp: Fix the is_enabled() operation
  - [13/17] ARM: shmobile: r7s72100: Populate .init_machine for generic
    r7s72100
  - [14/17] ARM: shmobile: r8a73a4: Populate .init_machine for generic r8a73a4
  - [15/17] ARM: shmobile: r8a7778: Populate .init_machine for generic r8a7778
  - [16/17] ARM: shmobile: r8a7790: Populate .init_machine for generic r8a7790
  - [17/17] ARM: shmobile: r8a7791: Populate .init_machine for generic r8a7791

Patches 4, 5, and 7 were authored by Ben Dooks, with some changes made by me
(for patch 7).
Patch 8 was based on a patch by Ben for r8a7790.
Patch 12 was authored by Laurent Pinchart.

Besides rebasing, only one patch was changed, compared to previous submission:
  - 8: The call to sh_pm_runtime_init() was moved to the start of
    .init_machine(), to enable the GPIO clocks on R-Car Gen2.

Several patches are new:
  - 1-3 contain some related cleanups to board support code,
  - 5 was added as it touches related code,
  - 10-11 remove workarounds that are no longer needed,
  - 12 was added as it depends on the fixes from this series,
  - 13-17 are more or less RFCs: As we always have to call
    sh_pm_runtime_init(), I added .init_machine hooks in the generic support
    code for all SoCs that didn't have it yet. I also included code to set up
    SoC platform support while at it, as this will be useful later. Booting
    using a generic "renesas,r*" DT was untested.

This was tested on Koelsch, both legacy and multi-platform.

Many thanks to Ben Dooks, Laurent Pinchart, and Magnus Damm for the
valuable input, investigations, and discussions.

Thanks!

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: [ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[Laurent Pinchart]

Hi Geert,

On Tuesday 15 April 2014 14:33:56 Geert Uytterhoeven wrote:
> 	Hi all,
> 
> This patchset enables the PM runtime code in drivers/sh/pm_runtime.c when
> running a multi-platform ARM kernel including support for shmobile.
> Before this code was only enabled for legacy shmobile kernels, leading to
> disabled clocks in multiplatform kernels, depending on implicit reset state
> or on the bootloader.
> It also contains some related cleanups, and removals of workarounds.
>
>   - [01/17] ARM: shmobile: r8a7740: Remove unused
>     r8a7740_add_early_devices_dt()
>   - [02/17] ARM: shmobile: r8a7740: Make r8a7740_meram_workaround() __init
>   - [03/17] ARM: shmobile: sh7372: Call sh7372_add_early_devices() instead
> of open coding
>   - [04/17] drivers: sh: pm_runtime does not need idle callback
>   - [05/17] drivers: sh: pm_runtime implementation needs to suspend and
> resume devices
>   - [06/17] sh: intc: Restrict INTC_USERIMASK to SH4A
>   - [07/17] drivers: sh: compile drivers/sh/pm_runtime.c if
> ARCH_SHMOBILE_MULTI - [08/17] ARM: shmobile: Call sh_pm_runtime_init() from
> platform code on all SoCs
>   - [09/17] drivers: sh: Remove automatic sh_pm_runtime_init() if
>     ARCH_SHMOBILE_LEGACY
>   - [10/17] ARM: shmobile: koelsch-reference: Remove workarounds for core
> clock issues
>   - [11/17] ARM: shmobile: lager-reference: Remove workarounds for core
> clock issues
>   - [12/17] clk: shmobile: mstp: Fix the is_enabled() operation
>   - [13/17] ARM: shmobile: r7s72100: Populate .init_machine for generic
>     r7s72100
>   - [14/17] ARM: shmobile: r8a73a4: Populate .init_machine for generic
> r8a73a4 - [15/17] ARM: shmobile: r8a7778: Populate .init_machine for
> generic r8a7778 - [16/17] ARM: shmobile: r8a7790: Populate .init_machine
> for generic r8a7790 - [17/17] ARM: shmobile: r8a7791: Populate
> .init_machine for generic r8a7791
> 
> Patches 4, 5, and 7 were authored by Ben Dooks, with some changes made by me
> (for patch 7).
> Patch 8 was based on a patch by Ben for r8a7790.
> Patch 12 was authored by Laurent Pinchart.
> 
> Besides rebasing, only one patch was changed, compared to previous
> submission: - 8: The call to sh_pm_runtime_init() was moved to the start of
>     .init_machine(), to enable the GPIO clocks on R-Car Gen2.

Thank you for the patches.

> Several patches are new:
>   - 1-3 contain some related cleanups to board support code,
>   - 5 was added as it touches related code,
>   - 10-11 remove workarounds that are no longer needed,
>   - 12 was added as it depends on the fixes from this series,

For 01/17 to 12/17,

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

>   - 13-17 are more or less RFCs: As we always have to call
>     sh_pm_runtime_init(), I added .init_machine hooks in the generic support
> code for all SoCs that didn't have it yet. I also included code to set up
> SoC platform support while at it, as this will be useful later. Booting
> using a generic "renesas,r*" DT was untested.

I wonder whether we shouldn't wait until we get rid of the *_add_dt_devices() 
functions before applying those.

As the goal is to remove the setup-*.c files, have you thought about where the 
sh_pm_runtime_init() calls should move to ?

> This was tested on Koelsch, both legacy and multi-platform.
> 
> Many thanks to Ben Dooks, Laurent Pinchart, and Magnus Damm for the
> valuable input, investigations, and discussions.

-- 
Regards,

Laurent Pinchart

Re: [ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[Geert Uytterhoeven]

Hi Laurent,

On Tue, Apr 15, 2014 at 4:49 PM, Laurent Pinchart
<laurent.pinchart@ideasonboard.com> wrote:
>>   - 13-17 are more or less RFCs: As we always have to call
>>     sh_pm_runtime_init(), I added .init_machine hooks in the generic support
>> code for all SoCs that didn't have it yet. I also included code to set up
>> SoC platform support while at it, as this will be useful later. Booting
>> using a generic "renesas,r*" DT was untested.
>
> I wonder whether we shouldn't wait until we get rid of the *_add_dt_devices()
> functions before applying those.

Sure, that's why they're marked RFC...

> As the goal is to remove the setup-*.c files, have you thought about where the
> sh_pm_runtime_init() calls should move to ?

cpg_mstp_clocks_init() would be nice ;-) Unfortunately it's way too early.

I'm looking for a way to tie this into the connection between devices and
MSTP gate clocks.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[Simon Horman]

On Tue, Apr 15, 2014 at 04:49:35PM +0200, Laurent Pinchart wrote:
> Hi Geert,
> 
> On Tuesday 15 April 2014 14:33:56 Geert Uytterhoeven wrote:
> > 	Hi all,
> > 
> > This patchset enables the PM runtime code in drivers/sh/pm_runtime.c when
> > running a multi-platform ARM kernel including support for shmobile.
> > Before this code was only enabled for legacy shmobile kernels, leading to
> > disabled clocks in multiplatform kernels, depending on implicit reset state
> > or on the bootloader.
> > It also contains some related cleanups, and removals of workarounds.

I am wondering what kind of overhead this introduces for non shmobile
platforms that use a kernel compiled with ARCH_SHMOBILE_MULTI enabled.

> >
> >   - [01/17] ARM: shmobile: r8a7740: Remove unused
> >     r8a7740_add_early_devices_dt()
> >   - [02/17] ARM: shmobile: r8a7740: Make r8a7740_meram_workaround() __init
> >   - [03/17] ARM: shmobile: sh7372: Call sh7372_add_early_devices() instead
> > of open coding
> >   - [04/17] drivers: sh: pm_runtime does not need idle callback
> >   - [05/17] drivers: sh: pm_runtime implementation needs to suspend and
> > resume devices
> >   - [06/17] sh: intc: Restrict INTC_USERIMASK to SH4A
> >   - [07/17] drivers: sh: compile drivers/sh/pm_runtime.c if
> > ARCH_SHMOBILE_MULTI - [08/17] ARM: shmobile: Call sh_pm_runtime_init() from
> > platform code on all SoCs
> >   - [09/17] drivers: sh: Remove automatic sh_pm_runtime_init() if
> >     ARCH_SHMOBILE_LEGACY
> >   - [10/17] ARM: shmobile: koelsch-reference: Remove workarounds for core
> > clock issues
> >   - [11/17] ARM: shmobile: lager-reference: Remove workarounds for core
> > clock issues
> >   - [12/17] clk: shmobile: mstp: Fix the is_enabled() operation
> >   - [13/17] ARM: shmobile: r7s72100: Populate .init_machine for generic
> >     r7s72100
> >   - [14/17] ARM: shmobile: r8a73a4: Populate .init_machine for generic
> > r8a73a4 - [15/17] ARM: shmobile: r8a7778: Populate .init_machine for
> > generic r8a7778 - [16/17] ARM: shmobile: r8a7790: Populate .init_machine
> > for generic r8a7790 - [17/17] ARM: shmobile: r8a7791: Populate
> > .init_machine for generic r8a7791
> > Patches 4, 5, and 7 were authored by Ben Dooks, with some changes made by me
> > (for patch 7).
> > Patch 8 was based on a patch by Ben for r8a7790.
> > Patch 12 was authored by Laurent Pinchart.
> > 
> > Besides rebasing, only one patch was changed, compared to previous
> > submission: - 8: The call to sh_pm_runtime_init() was moved to the start of
> >     .init_machine(), to enable the GPIO clocks on R-Car Gen2.
> 
> Thank you for the patches.
> 
> > Several patches are new:
> >   - 1-3 contain some related cleanups to board support code,

Thanks, I'll look at queuing them up.

> >   - 5 was added as it touches related code,
> >   - 10-11 remove workarounds that are no longer needed,
> >   - 12 was added as it depends on the fixes from this series,

Magnus, I'd appreciate a review from you of patches 4 - 12.

> 
> For 01/17 to 12/17,
> 
> Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> 
> >   - 13-17 are more or less RFCs: As we always have to call
> >     sh_pm_runtime_init(), I added .init_machine hooks in the generic support
> > code for all SoCs that didn't have it yet. I also included code to set up
> > SoC platform support while at it, as this will be useful later. Booting
> > using a generic "renesas,r*" DT was untested.
> 
> I wonder whether we shouldn't wait until we get rid of the *_add_dt_devices() 
> functions before applying those.
> 
> As the goal is to remove the setup-*.c files, have you thought about where the 
> sh_pm_runtime_init() calls should move to ?
> 
> > This was tested on Koelsch, both legacy and multi-platform.
> > 
> > Many thanks to Ben Dooks, Laurent Pinchart, and Magnus Damm for the
> > valuable input, investigations, and discussions.
> 
> -- 
> Regards,
> 
> Laurent Pinchart
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-sh" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[Simon Horman]

On Wed, Apr 16, 2014 at 09:18:14AM +0900, Simon Horman wrote:
> On Tue, Apr 15, 2014 at 04:49:35PM +0200, Laurent Pinchart wrote:
> > Hi Geert,
> > 
> > On Tuesday 15 April 2014 14:33:56 Geert Uytterhoeven wrote:
> > > 	Hi all,
> > > 
> > > This patchset enables the PM runtime code in drivers/sh/pm_runtime.c when
> > > running a multi-platform ARM kernel including support for shmobile.
> > > Before this code was only enabled for legacy shmobile kernels, leading to
> > > disabled clocks in multiplatform kernels, depending on implicit reset state
> > > or on the bootloader.
> > > It also contains some related cleanups, and removals of workarounds.
> 
> I am wondering what kind of overhead this introduces for non shmobile
> platforms that use a kernel compiled with ARCH_SHMOBILE_MULTI enabled.
> 
> > >
> > >   - [01/17] ARM: shmobile: r8a7740: Remove unused
> > >     r8a7740_add_early_devices_dt()
> > >   - [02/17] ARM: shmobile: r8a7740: Make r8a7740_meram_workaround() __init
> > >   - [03/17] ARM: shmobile: sh7372: Call sh7372_add_early_devices() instead
> > > of open coding
> > >   - [04/17] drivers: sh: pm_runtime does not need idle callback
> > >   - [05/17] drivers: sh: pm_runtime implementation needs to suspend and
> > > resume devices
> > >   - [06/17] sh: intc: Restrict INTC_USERIMASK to SH4A
> > >   - [07/17] drivers: sh: compile drivers/sh/pm_runtime.c if
> > > ARCH_SHMOBILE_MULTI - [08/17] ARM: shmobile: Call sh_pm_runtime_init() from
> > > platform code on all SoCs
> > >   - [09/17] drivers: sh: Remove automatic sh_pm_runtime_init() if
> > >     ARCH_SHMOBILE_LEGACY
> > >   - [10/17] ARM: shmobile: koelsch-reference: Remove workarounds for core
> > > clock issues
> > >   - [11/17] ARM: shmobile: lager-reference: Remove workarounds for core
> > > clock issues
> > >   - [12/17] clk: shmobile: mstp: Fix the is_enabled() operation
> > >   - [13/17] ARM: shmobile: r7s72100: Populate .init_machine for generic
> > >     r7s72100
> > >   - [14/17] ARM: shmobile: r8a73a4: Populate .init_machine for generic
> > > r8a73a4 - [15/17] ARM: shmobile: r8a7778: Populate .init_machine for
> > > generic r8a7778 - [16/17] ARM: shmobile: r8a7790: Populate .init_machine
> > > for generic r8a7790 - [17/17] ARM: shmobile: r8a7791: Populate
> > > .init_machine for generic r8a7791
> > > Patches 4, 5, and 7 were authored by Ben Dooks, with some changes made by me
> > > (for patch 7).
> > > Patch 8 was based on a patch by Ben for r8a7790.
> > > Patch 12 was authored by Laurent Pinchart.
> > > 
> > > Besides rebasing, only one patch was changed, compared to previous
> > > submission: - 8: The call to sh_pm_runtime_init() was moved to the start of
> > >     .init_machine(), to enable the GPIO clocks on R-Car Gen2.
> > 
> > Thank you for the patches.
> > 
> > > Several patches are new:
> > >   - 1-3 contain some related cleanups to board support code,
> 
> Thanks, I'll look at queuing them up.

I have now queued them up

> 
> > >   - 5 was added as it touches related code,
> > >   - 10-11 remove workarounds that are no longer needed,
> > >   - 12 was added as it depends on the fixes from this series,
> 
> Magnus, I'd appreciate a review from you of patches 4 - 12.
> 
> > 
> > For 01/17 to 12/17,
> > 
> > Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> > 
> > >   - 13-17 are more or less RFCs: As we always have to call
> > >     sh_pm_runtime_init(), I added .init_machine hooks in the generic support
> > > code for all SoCs that didn't have it yet. I also included code to set up
> > > SoC platform support while at it, as this will be useful later. Booting
> > > using a generic "renesas,r*" DT was untested.
> > 
> > I wonder whether we shouldn't wait until we get rid of the *_add_dt_devices() 
> > functions before applying those.
> > 
> > As the goal is to remove the setup-*.c files, have you thought about where the 
> > sh_pm_runtime_init() calls should move to ?
> > 
> > > This was tested on Koelsch, both legacy and multi-platform.
> > > 
> > > Many thanks to Ben Dooks, Laurent Pinchart, and Magnus Damm for the
> > > valuable input, investigations, and discussions.
> > 
> > -- 
> > Regards,
> > 
> > Laurent Pinchart
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-sh" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-sh" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[Geert Uytterhoeven]

Hi Simon,

On Wed, Apr 16, 2014 at 2:18 AM, Simon Horman <horms@verge.net.au> wrote:
>> > This patchset enables the PM runtime code in drivers/sh/pm_runtime.c when
>> > running a multi-platform ARM kernel including support for shmobile.
>> > Before this code was only enabled for legacy shmobile kernels, leading to
>> > disabled clocks in multiplatform kernels, depending on implicit reset state
>> > or on the bootloader.
>> > It also contains some related cleanups, and removals of workarounds.
>
> I am wondering what kind of overhead this introduces for non shmobile
> platforms that use a kernel compiled with ARCH_SHMOBILE_MULTI enabled.

The only overhead for non-shmobile platforms is the compilation of the code,
and the late_initcall(sh_pm_runtime_late_init), which will return immediately.
It's not enabled unless called from a shmobile .init_machine.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[Simon Horman]

On Wed, Apr 16, 2014 at 09:15:35AM +0200, Geert Uytterhoeven wrote:
> Hi Simon,
> 
> On Wed, Apr 16, 2014 at 2:18 AM, Simon Horman <horms@verge.net.au> wrote:
> >> > This patchset enables the PM runtime code in drivers/sh/pm_runtime.c when
> >> > running a multi-platform ARM kernel including support for shmobile.
> >> > Before this code was only enabled for legacy shmobile kernels, leading to
> >> > disabled clocks in multiplatform kernels, depending on implicit reset state
> >> > or on the bootloader.
> >> > It also contains some related cleanups, and removals of workarounds.
> >
> > I am wondering what kind of overhead this introduces for non shmobile
> > platforms that use a kernel compiled with ARCH_SHMOBILE_MULTI enabled.
> 
> The only overhead for non-shmobile platforms is the compilation of the code,
> and the late_initcall(sh_pm_runtime_late_init), which will return immediately.
> It's not enabled unless called from a shmobile .init_machine.

Thanks, that sounds entirely reasonable to me.

Re: [ PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on multi-platform

[Geert Uytterhoeven]

Hi Laurent,

On Tue, Apr 15, 2014 at 5:52 PM, Geert Uytterhoeven
<geert@linux-m68k.org> wrote:
>> As the goal is to remove the setup-*.c files, have you thought about where the
>> sh_pm_runtime_init() calls should move to ?
>
> cpg_mstp_clocks_init() would be nice ;-) Unfortunately it's way too early.

However, I've just got an idea...

> I'm looking for a way to tie this into the connection between devices and
> MSTP gate clocks.

That's "[PATCH/RFC 0/4] of: Register clocks for Runtime PM with PM core"

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds