Skip to contentSkip to conversation

gpio: Add GPIO driver framework for Marvell SoCs

PATCH Superseded Ajay Bhargav 2 participants 2 messages Aug 6, 2011 – Aug 7, 2011 1 message not in the archive

v2 of 7 Superseded Latest is v7 →

lore
Share & tools
lore
#1 lore raw

PATCH v2 1/2 gpio: Add GPIO driver framework for Marvell SoCs

In reply to a message not in the archive
Quoted 130 lines · ----- "Prafulla Wadaskar
Quoted 2 lines

-----Original Message-----
From: u-boot-bounces at lists.denx.de

[mailto:u-boot-bounces at lists.denx.de]

Quoted 5 lines

On Behalf Of Ajay Bhargav
Sent: Thursday, August 04, 2011 4:21 PM
To: Lei Wen
Cc: u-boot at lists.denx.de
Subject: Re: [U-Boot] [PATCH v2 1/2] gpio: Add GPIO driver framework

for

Quoted 15 lines

Marvell SoCs

Quoted 11 lines · ----- "Lei Wen
Quoted 9 lines · Albert ARIBAUD

Hi Simon,

Quoted 6 lines · Simon Guinot

Hi Ajay,

Quoted 3 lines · Ajay Bhargav
Quoted 1 line · ----- "Simon Guinot"<[email protected]> ?wrote:

AFAIK, Orion and Kirkwood SoCs don't provide bitwise

set/clear

Quoted 7 lines
Quoted 7 lines

for

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

GPIO output/direction registers. Instead, a register must be

read

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

first to leave other bits unchanged (see __set_direction in kw_gpio.c).

Is it possible to handle Armada SoCs GPIOs in a same way ?

maybe

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

using
the pin registers (gpxx in the Armada struct gpio_reg array)

?

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

If not, this code is not Marvell generic but rather specific

for

Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 6 lines

Armada
SoCs and then maybe armada_gpio is a better name...

Regards,

Simon

Hi Simon,

Yes its possible to implement code that way, Armada SoC does

have

Quoted 2 lines
Quoted 2 lines

GPIO

Quoted 1 line
Quoted 1 line
Quoted 1 line

registers for set/clear. what about register naming?? I think

they

Quoted 4 lines
Quoted 4 lines

are

Quoted 3 lines
Quoted 3 lines
Quoted 1 line

different for Kirkwood and Orion.

I think that the register names could be OK. But here is a

most

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

important problem: On Orion/Kirkwood SoCs, a single GPIO

output

Quoted 2 lines
Quoted 2 lines

register

Quoted 1 line
Quoted 1 line

is available (no set/clear variants as for Armada). I missed

that

Quoted 4 lines
Quoted 4 lines

point

Quoted 1 line
Quoted 1 line

at my first look. It is quite problematic because only two

registers are

Quoted 1 line
Quoted 1 line

shared between the different Marvell SoCs: level and direction.

In

Quoted 2 lines
Quoted 2 lines

fact,

Quoted 1 line
Quoted 1 line

this registers are probably relevant on every machines

providing

Quoted 6 lines
Quoted 6 lines

GPIOs...

Quoted 5 lines
Quoted 3 lines

Maybe that having two common registers is not enough to add Orion/Kirkwood support to the mvgpio driver ?

?>

Quoted 1 line
Quoted 1 line

One more thing which can be done to make this code generic is

to

Quoted 2 lines
Quoted 2 lines

have

Quoted 1 line
Quoted 1 line
Quoted 1 line

some macros which can be defined by individual arch for

specific

Quoted 9 lines
Quoted 9 lines

registers

Quoted 8 lines
Quoted 8 lines
Quoted 6 lines

which are going to be in use e.g.

#define GPIO_PIN_LEVEL_REG
#define GPIO_DIR_REG
#define GPIO_PIN_SET_REG
#define GPIO_PIN_CLR_REG

Yes, but how to handle both a single GPI0 output register and

some

Quoted 6 lines
Quoted 6 lines

GPIO

Quoted 3 lines
Quoted 1 line

{set,clear} output registers (in a nice way) ?

Two distinct gipo drivers for the two marvell variants?

If let I choose, I'd prefer two, since the register set is

different.

Quoted 8 lines
Quoted 4 lines

Best regards,
Lei

Hi Simon,

For Armada minimum 3 registers are required and available for

armada

Quoted 6 lines

1. Direction (read/write)
2. Pin level set (write only)
3. Pin level clear (write only)

@lei
How bout if we check for architecture and use specific code or

defines?

Quoted 8 lines

i.e.
#ifdef CONFIG_KIRKWOOD
//KW code
#elif CONFIG_ARMADA100
//Armada code
#else
//orion or other?
#endif

Let's avoid this, because there will be several SoC architectures that uses similar GPIO register definitions, like kirkwood/orion have similar definition and armada/mmp/pantheon/etc.. have different one.

So we will end up having several #ifdefs. Ideally #ifdefs are discouraged for better coding practices.

Quoted 8 lines

Instead,
I would suggest to use macros for this code segments or alternatively inlined functions and those should be defined in mvgpio.h, #ifdefed with CPU core subversion (i.e. CONFIG_FEROCEION, CONFIG_SHEEVA_88SV331xV5)

Regards..
Prafulla . .

Hi Prafulla,

I think it will be better to keep two driver files. Let this patch be for Armada/mmp/pantheon and other compatible SoCs. Should the common GPIO struct for armada/mmp etc.. be moved out of GPIO.h to mvgpio.h?

Regards,
Ajay Bhargav

#2 lore raw

PATCH v2 1/2 gpio: Add GPIO driver framework for Marvell SoCs

Quoted 10 lines

-----Original Message-----
From: Ajay Bhargav [mailto:ajay.bhargav at einfochips.com]
Sent: Saturday, August 06, 2011 10:40 AM
To: Prafulla Wadaskar
Cc: u-boot at lists.denx.de; Lei Wen; Simon Guinot
Subject: Re: [U-Boot] [PATCH v2 1/2] gpio: Add GPIO driver framework for Marvell SoCs

...snip...

Quoted 29 lines
Quoted 20 lines
Quoted 3 lines

@lei
How bout if we check for architecture and use specific code or

defines?

Quoted 8 lines

i.e.
#ifdef CONFIG_KIRKWOOD
//KW code
#elif CONFIG_ARMADA100
//Armada code
#else
//orion or other?
#endif

Let's avoid this, because there will be several SoC architectures that uses similar GPIO register definitions, like kirkwood/orion have similar definition and armada/mmp/pantheon/etc.. have different one.

So we will end up having several #ifdefs. Ideally #ifdefs are discouraged for better coding practices.

Quoted 8 lines

Instead,
I would suggest to use macros for this code segments or alternatively inlined functions and those should be defined in mvgpio.h, #ifdefed with CPU core subversion (i.e. CONFIG_FEROCEION, CONFIG_SHEEVA_88SV331xV5)

Regards..
Prafulla . .

Quoted 6 lines

Hi Prafulla,

I think it will be better to keep two driver files. Let this patch be for
Armada/mmp/pantheon and other compatible SoCs. Should the common GPIO struct for armada/mmp etc.. be moved out of GPIO.h to mvgpio.h?

Hi Ajay
This is straight forward solution, but the idea here to share the code, better design. To simplify further, you can go ahead with below arch for this design.

1. Generic driver skeleton in mvgpio.c
2. Generic driver header mvgpio.h (optional you can avoid this)
3. Arch specific header asm/arch/gpio.h, to hold MAX GPIOs, register definition and gpio set/clr,config inlined functions or macros only.
4. Base address in asm/arch/<soc_name>.h and this file to be included in asm/arch/gpio.h

In future if we need to add support for other SoC we will add gpio.h.

Regards..
Prafulla . .