(arm-)global platform device registration functions

(arm-)global platform device registration functions

[Uwe Kleine-König]

Hello,

some time ago I started to dynamically allocate platform_devices for
arm/plat-mxc.  For that purpose I added many functions, see
arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
want to do the same, so I wonder if it would be sensible to do something
like this more globally (either for arm or better kernel global).

Thoughts?

Best regards
Uwe

[1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=arch/arm/plat-mxc/include/mach/devices-common.h;hb=HEAD

-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

(arm-)global platform device registration functions

[Uwe Kleine-König]

Hello,

some time ago I started to dynamically allocate platform_devices for
arm/plat-mxc.  For that purpose I added many functions, see
arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
want to do the same, so I wonder if it would be sensible to do something
like this more globally (either for arm or better kernel global).

Thoughts?

Best regards
Uwe

[1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=arch/arm/plat-mxc/include/mach/devices-common.h;hb=HEAD

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

(arm-)global platform device registration functions

[Andrew Morton]

On Wed, 29 Dec 2010 17:25:12 +0100
Uwe Kleine-K__nig <u.kleine-koenig@pengutronix.de> wrote:

> Hello,
> 
> some time ago I started to dynamically allocate platform_devices for
> arm/plat-mxc.  For that purpose I added many functions, see
> arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
> want to do the same, so I wonder if it would be sensible to do something
> like this more globally (either for arm or better kernel global).
> 

Sounds sensible to me, although I don't really understand what you've
done - didn't platform_device_alloc() already "dynamically allocate
platform_devices"?

If you were to implement this common infrastructure, there may be
already-existing code which could use it.  How would we locate that?

Re: (arm-)global platform device registration functions

[Andrew Morton]

On Wed, 29 Dec 2010 17:25:12 +0100
Uwe Kleine-K__nig <u.kleine-koenig@pengutronix.de> wrote:

> Hello,
> 
> some time ago I started to dynamically allocate platform_devices for
> arm/plat-mxc.  For that purpose I added many functions, see
> arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
> want to do the same, so I wonder if it would be sensible to do something
> like this more globally (either for arm or better kernel global).
> 

Sounds sensible to me, although I don't really understand what you've
done - didn't platform_device_alloc() already "dynamically allocate
platform_devices"?

If you were to implement this common infrastructure, there may be
already-existing code which could use it.  How would we locate that?

(arm-)global platform device registration functions

[Uwe Kleine-König]

On Wed, Dec 29, 2010 at 12:37:40PM -0800, Andrew Morton wrote:
> On Wed, 29 Dec 2010 17:25:12 +0100
> Uwe Kleine-K__nig <u.kleine-koenig@pengutronix.de> wrote:
> 
> > Hello,
> > 
> > some time ago I started to dynamically allocate platform_devices for
> > arm/plat-mxc.  For that purpose I added many functions, see
> > arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
> > want to do the same, so I wonder if it would be sensible to do something
> > like this more globally (either for arm or better kernel global).
> > 
> 
> Sounds sensible to me, although I don't really understand what you've
> done - didn't platform_device_alloc() already "dynamically allocate
> platform_devices"?
my code is a bit more specialised than platform_device_alloc.

Look for example at arch/arm/plat-mxc/devices/platform-fec.c.  Calling
imx51_add_fec() creates a fec device with the resources needed for
i.MX51 and adds platform data.  (imx_add_platform_device does the same
as platform_device_register_resndata[2])

> If you were to implement this common infrastructure, there may be
> already-existing code which could use it.  How would we locate that?
coccinelle probably could.  For imx51_add_fec it could match for a
struct platform_device with name = "fec".

Best regards
Uwe

[2] actually platform_device_register_resndata was created by me to
substitute imx_add_platform_device, but I need to "fix" the former to
handle dma stuff (not merged yet) before imx_add_platform_device can go
away.
-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

Re: (arm-)global platform device registration functions

[Uwe Kleine-König]

On Wed, Dec 29, 2010 at 12:37:40PM -0800, Andrew Morton wrote:
> On Wed, 29 Dec 2010 17:25:12 +0100
> Uwe Kleine-K__nig <u.kleine-koenig@pengutronix.de> wrote:
> 
> > Hello,
> > 
> > some time ago I started to dynamically allocate platform_devices for
> > arm/plat-mxc.  For that purpose I added many functions, see
> > arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
> > want to do the same, so I wonder if it would be sensible to do something
> > like this more globally (either for arm or better kernel global).
> > 
> 
> Sounds sensible to me, although I don't really understand what you've
> done - didn't platform_device_alloc() already "dynamically allocate
> platform_devices"?
my code is a bit more specialised than platform_device_alloc.

Look for example at arch/arm/plat-mxc/devices/platform-fec.c.  Calling
imx51_add_fec() creates a fec device with the resources needed for
i.MX51 and adds platform data.  (imx_add_platform_device does the same
as platform_device_register_resndata[2])

> If you were to implement this common infrastructure, there may be
> already-existing code which could use it.  How would we locate that?
coccinelle probably could.  For imx51_add_fec it could match for a
struct platform_device with name = "fec".

Best regards
Uwe

[2] actually platform_device_register_resndata was created by me to
substitute imx_add_platform_device, but I need to "fix" the former to
handle dma stuff (not merged yet) before imx_add_platform_device can go
away.
-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

(arm-)global platform device registration functions

[Arnd Bergmann]

On Wednesday 29 December 2010 17:25:12 Uwe Kleine-K?nig wrote:
> some time ago I started to dynamically allocate platform_devices for
> arm/plat-mxc.  For that purpose I added many functions, see
> arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
> want to do the same, so I wonder if it would be sensible to do something
> like this more globally (either for arm or better kernel global).

Adding platform devices dynamically is definitely the right way forward
for new platforms, as Greg explained recently. The imx_add_platform_device()
function you have defined looks useful for this, and I guess it could
be added globally, next to the existing platform_device_register_resndata,
platform_device_register_data and platform_device_register_simple functions
that we already have doing similar things.

I'm not convinced that the various wrappers you have on imx are the
best solution, though I have not come up with anything better either ;-)

This will also get a lot easier once we have the ability to define
devices in using dtc, because that takes care of the resources.

	Arnd

Re: (arm-)global platform device registration functions

[Arnd Bergmann]

On Wednesday 29 December 2010 17:25:12 Uwe Kleine-König wrote:
> some time ago I started to dynamically allocate platform_devices for
> arm/plat-mxc.  For that purpose I added many functions, see
> arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
> want to do the same, so I wonder if it would be sensible to do something
> like this more globally (either for arm or better kernel global).

Adding platform devices dynamically is definitely the right way forward
for new platforms, as Greg explained recently. The imx_add_platform_device()
function you have defined looks useful for this, and I guess it could
be added globally, next to the existing platform_device_register_resndata,
platform_device_register_data and platform_device_register_simple functions
that we already have doing similar things.

I'm not convinced that the various wrappers you have on imx are the
best solution, though I have not come up with anything better either ;-)

This will also get a lot easier once we have the ability to define
devices in using dtc, because that takes care of the resources.

	Arnd

(arm-)global platform device registration functions

[Uwe Kleine-König]

Hello Arnd,

On Thu, Dec 30, 2010 at 11:03:31AM +0100, Arnd Bergmann wrote:
> On Wednesday 29 December 2010 17:25:12 Uwe Kleine-K?nig wrote:
> > some time ago I started to dynamically allocate platform_devices for
> > arm/plat-mxc.  For that purpose I added many functions, see
> > arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
> > want to do the same, so I wonder if it would be sensible to do something
> > like this more globally (either for arm or better kernel global).
> 
> Adding platform devices dynamically is definitely the right way forward
> for new platforms, as Greg explained recently. The imx_add_platform_device()
> function you have defined looks useful for this, and I guess it could
> be added globally, next to the existing platform_device_register_resndata,
imx_add_platform_device could be implemented as

	platform_device_register_resndata(NULL, name, id, res, num_resources, data, size_data)

in next[1] the imx variant got support to handle a dmamask.

> platform_device_register_data and platform_device_register_simple functions
> that we already have doing similar things.
> 
> I'm not convinced that the various wrappers you have on imx are the
> best solution, though I have not come up with anything better either ;-)
I came up with those because many machines (sharing the same SoC) share
the same devices.

> This will also get a lot easier once we have the ability to define
> devices in using dtc, because that takes care of the resources.
Well this is a matter of taste, I'm quite happy with the current
approach.

[1]
http://git.kernel.org/?p=linux/kernel/git/sfr/linux-next.git;a=blob;f=arch/arm/plat-mxc/devices.c;hb=e635401443d9d256a0bfd1ea7864ac56034c6dca
-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

Re: (arm-)global platform device registration functions

[Uwe Kleine-König]

Hello Arnd,

On Thu, Dec 30, 2010 at 11:03:31AM +0100, Arnd Bergmann wrote:
> On Wednesday 29 December 2010 17:25:12 Uwe Kleine-König wrote:
> > some time ago I started to dynamically allocate platform_devices for
> > arm/plat-mxc.  For that purpose I added many functions, see
> > arch/arm/plat-mxc/include/mach/devices-common.h [1].  Now arm/arch-mxs
> > want to do the same, so I wonder if it would be sensible to do something
> > like this more globally (either for arm or better kernel global).
> 
> Adding platform devices dynamically is definitely the right way forward
> for new platforms, as Greg explained recently. The imx_add_platform_device()
> function you have defined looks useful for this, and I guess it could
> be added globally, next to the existing platform_device_register_resndata,
imx_add_platform_device could be implemented as

	platform_device_register_resndata(NULL, name, id, res, num_resources, data, size_data)

in next[1] the imx variant got support to handle a dmamask.

> platform_device_register_data and platform_device_register_simple functions
> that we already have doing similar things.
> 
> I'm not convinced that the various wrappers you have on imx are the
> best solution, though I have not come up with anything better either ;-)
I came up with those because many machines (sharing the same SoC) share
the same devices.

> This will also get a lot easier once we have the ability to define
> devices in using dtc, because that takes care of the resources.
Well this is a matter of taste, I'm quite happy with the current
approach.

[1]
http://git.kernel.org/?p=linux/kernel/git/sfr/linux-next.git;a=blob;f=arch/arm/plat-mxc/devices.c;hb=e635401443d9d256a0bfd1ea7864ac56034c6dca
-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |