diff mbox

ARM: at-91: ARCH_AT91SAM9X5 went missing

Message ID 1363171688.3137.196.camel@x61.thuisdomein (mailing list archive)
State New, archived
Headers show

Commit Message

Paul Bolle March 13, 2013, 10:48 a.m. UTC 
0) Commit 1e3ce2b8545390a2aee8dbfcd49ca4161b636000
("AR[M]: at91: introduce SOC_AT91xxx define to allow to compile SoC core
support") dropped the Kconfig symbol ARCH_AT91SAM9X5. It seems it missed
something like this:


1) But before I submit something along those lines: ARCH_AT91* variants
are still used quite a bit around the tree (both as Kconfig symbols and
as Kconfig macros). How are the current SOC_AT91* symbols wired back to
the old ARCH_AT91* symbols? Or are those old symbols and macros _only_
needed for backwards compatibility?


Paul Bolle

Comments

Nicolas Ferre March 13, 2013, 11:12 a.m. UTC | #1 
On 03/13/2013 11:48 AM, Paul Bolle :
> 0) Commit 1e3ce2b8545390a2aee8dbfcd49ca4161b636000
> ("AR[M]: at91: introduce SOC_AT91xxx define to allow to compile SoC core
> support") dropped the Kconfig symbol ARCH_AT91SAM9X5. It seems it missed
> something like this:
> 
> diff --git a/arch/arm/mach-at91/Kconfig b/arch/arm/mach-at91/Kconfig
> index 6071f4c..2db5370 100644
> --- a/arch/arm/mach-at91/Kconfig
> +++ b/arch/arm/mach-at91/Kconfig
> @@ -131,6 +131,10 @@ config ARCH_AT91SAM9G45
>         bool "AT91SAM9G45"
>         select SOC_AT91SAM9G45
>  
> +config ARCH_AT91SAM9X5
> +       bool "AT91SAM9X5"
> +       select SOC_AT91SAM9X5
> +
>  config ARCH_AT91X40
>         bool "AT91x40"
>         depends on !MMU
> 
> 1) But before I submit something along those lines: ARCH_AT91* variants
> are still used quite a bit around the tree (both as Kconfig symbols and
> as Kconfig macros). How are the current SOC_AT91* symbols wired back to
> the old ARCH_AT91* symbols? Or are those old symbols and macros _only_
> needed for backwards compatibility?

These ARCH_AT91xxx variables are not used anymore for pure Device Tree
SoCs (I mean: at91sam9x5, at91sam9n12, sama5d3 families). They are used
for non-DT SoCs and for chips that support both legacy board files and DT.
The old ARCH_AT91xxx symbols are selecting the corresponding SOC_AT91xxx
symbol.

ARCH_AT91xxx in Kconfig that are specific to a particular SoC are going
to disappear as we tend to allow the compilation of a particular driver
for a wider range of configurations.

Hope that it helps. Best regards,
Paul Bolle March 13, 2013, 11:22 a.m. UTC | #2 
On Wed, 2013-03-13 at 12:12 +0100, Nicolas Ferre wrote:
> These ARCH_AT91xxx variables are not used anymore for pure Device Tree
> SoCs (I mean: at91sam9x5, at91sam9n12, sama5d3 families). They are used
> for non-DT SoCs and for chips that support both legacy board files and DT.
> The old ARCH_AT91xxx symbols are selecting the corresponding SOC_AT91xxx
> symbol.
> 
> ARCH_AT91xxx in Kconfig that are specific to a particular SoC are going
> to disappear as we tend to allow the compilation of a particular driver
> for a wider range of configurations.

So setting one of the SOC_* symbols will not generate the corresponding
ARCH_* symbol, will it? To be specific, for the people entering the
configuration from the new SOC_* side none of this is relevant anymore
(this is all the stuff outside of arch/arm):

drivers/misc/Kconfig:87:        prompt "TC Block" if ARCH_AT91RM9200 || ARCH_AT91SAM9260 || CPU_AT32AP700X
drivers/pcmcia/Kconfig:291:     depends on PCMCIA && ARCH_AT91RM9200
drivers/rtc/Kconfig:977:        depends on ARCH_AT91 && !(ARCH_AT91RM9200 || ARCH_AT91X40)
drivers/tty/serial/atmel_serial.c:268:#ifdef CONFIG_ARCH_AT91RM9200
drivers/watchdog/Kconfig:127:   depends on ARCH_AT91 && !ARCH_AT91RM9200
drivers/misc/Kconfig:87:        prompt "TC Block" if ARCH_AT91RM9200 || ARCH_AT91SAM9260 || CPU_AT32AP700X
drivers/video/Kconfig:1026:     depends on FB_ATMEL && ARCH_AT91SAM9261
drivers/net/can/Kconfig:68:     depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5
drivers/rtc/Kconfig:991:        prompt "RTT module Number" if ARCH_AT91SAM9263
drivers/rtc/Kconfig:1000:       range 0 3 if !ARCH_AT91SAM9263
drivers/rtc/Kconfig:1001:       range 0 15 if ARCH_AT91SAM9263
drivers/usb/gadget/Kconfig:157: depends on AVR32 || ARCH_AT91SAM9RL || ARCH_AT91SAM9G45
drivers/usb/gadget/atmel_usba_udc.c:329:#if defined(CONFIG_ARCH_AT91SAM9RL)
drivers/usb/gadget/atmel_usba_udc.c:349:#endif /* CONFIG_ARCH_AT91SAM9RL */
drivers/mmc/host/Kconfig:305:   depends on MMC_ATMELMCI && (AVR32 || ARCH_AT91SAM9G45) && DMA_ENGINE
drivers/usb/gadget/Kconfig:157: depends on AVR32 || ARCH_AT91SAM9RL || ARCH_AT91SAM9G45
drivers/net/can/Kconfig:68:     depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5


Paul Bolle
Jean-Christophe PLAGNIOL-VILLARD March 13, 2013, 12:26 p.m. UTC | #3 
On 12:22 Wed 13 Mar     , Paul Bolle wrote:
> On Wed, 2013-03-13 at 12:12 +0100, Nicolas Ferre wrote:
> > These ARCH_AT91xxx variables are not used anymore for pure Device Tree
> > SoCs (I mean: at91sam9x5, at91sam9n12, sama5d3 families). They are used
> > for non-DT SoCs and for chips that support both legacy board files and DT.
> > The old ARCH_AT91xxx symbols are selecting the corresponding SOC_AT91xxx
> > symbol.
> > 
> > ARCH_AT91xxx in Kconfig that are specific to a particular SoC are going
> > to disappear as we tend to allow the compilation of a particular driver
> > for a wider range of configurations.
> 
> So setting one of the SOC_* symbols will not generate the corresponding
> ARCH_* symbol, will it? To be specific, for the people entering the
> configuration from the new SOC_* side none of this is relevant anymore
> (this is all the stuff outside of arch/arm):

yes it's as when I prepare the at91 for dt support I split the code in two
world. The old style (one soc at a time) and the st multi soc

the ARCH_AT91xxx are still here as part of the code  are still depending on
old style platform data and not converted to DT

but the ARCH_AT91SAM9x5 never exist at the mainline

and all the old style as now mark as deprecated and for removal

Best Regards,
J.
> 
> drivers/misc/Kconfig:87:        prompt "TC Block" if ARCH_AT91RM9200 || ARCH_AT91SAM9260 || CPU_AT32AP700X
> drivers/pcmcia/Kconfig:291:     depends on PCMCIA && ARCH_AT91RM9200
> drivers/rtc/Kconfig:977:        depends on ARCH_AT91 && !(ARCH_AT91RM9200 || ARCH_AT91X40)
> drivers/tty/serial/atmel_serial.c:268:#ifdef CONFIG_ARCH_AT91RM9200
> drivers/watchdog/Kconfig:127:   depends on ARCH_AT91 && !ARCH_AT91RM9200
> drivers/misc/Kconfig:87:        prompt "TC Block" if ARCH_AT91RM9200 || ARCH_AT91SAM9260 || CPU_AT32AP700X
> drivers/video/Kconfig:1026:     depends on FB_ATMEL && ARCH_AT91SAM9261
> drivers/net/can/Kconfig:68:     depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5
> drivers/rtc/Kconfig:991:        prompt "RTT module Number" if ARCH_AT91SAM9263
> drivers/rtc/Kconfig:1000:       range 0 3 if !ARCH_AT91SAM9263
> drivers/rtc/Kconfig:1001:       range 0 15 if ARCH_AT91SAM9263
> drivers/usb/gadget/Kconfig:157: depends on AVR32 || ARCH_AT91SAM9RL || ARCH_AT91SAM9G45
> drivers/usb/gadget/atmel_usba_udc.c:329:#if defined(CONFIG_ARCH_AT91SAM9RL)
> drivers/usb/gadget/atmel_usba_udc.c:349:#endif /* CONFIG_ARCH_AT91SAM9RL */
> drivers/mmc/host/Kconfig:305:   depends on MMC_ATMELMCI && (AVR32 || ARCH_AT91SAM9G45) && DMA_ENGINE
> drivers/usb/gadget/Kconfig:157: depends on AVR32 || ARCH_AT91SAM9RL || ARCH_AT91SAM9G45
> drivers/net/can/Kconfig:68:     depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5
> 
> 
> Paul Bolle
>
Nicolas Ferre March 13, 2013, 12:30 p.m. UTC | #4 
On 03/13/2013 12:22 PM, Paul Bolle :
> On Wed, 2013-03-13 at 12:12 +0100, Nicolas Ferre wrote:
>> These ARCH_AT91xxx variables are not used anymore for pure Device Tree
>> SoCs (I mean: at91sam9x5, at91sam9n12, sama5d3 families). They are used
>> for non-DT SoCs and for chips that support both legacy board files and DT.
>> The old ARCH_AT91xxx symbols are selecting the corresponding SOC_AT91xxx
>> symbol.
>>
>> ARCH_AT91xxx in Kconfig that are specific to a particular SoC are going
>> to disappear as we tend to allow the compilation of a particular driver
>> for a wider range of configurations.
> 
> So setting one of the SOC_* symbols will not generate the corresponding
> ARCH_* symbol, will it? To be specific, for the people entering the
> configuration from the new SOC_* side none of this is relevant anymore
> (this is all the stuff outside of arch/arm):
> 
> drivers/misc/Kconfig:87:        prompt "TC Block" if ARCH_AT91RM9200 || ARCH_AT91SAM9260 || CPU_AT32AP700X
> drivers/pcmcia/Kconfig:291:     depends on PCMCIA && ARCH_AT91RM9200
> drivers/rtc/Kconfig:977:        depends on ARCH_AT91 && !(ARCH_AT91RM9200 || ARCH_AT91X40)
> drivers/tty/serial/atmel_serial.c:268:#ifdef CONFIG_ARCH_AT91RM9200
> drivers/watchdog/Kconfig:127:   depends on ARCH_AT91 && !ARCH_AT91RM9200
> drivers/misc/Kconfig:87:        prompt "TC Block" if ARCH_AT91RM9200 || ARCH_AT91SAM9260 || CPU_AT32AP700X
> drivers/video/Kconfig:1026:     depends on FB_ATMEL && ARCH_AT91SAM9261
> drivers/net/can/Kconfig:68:     depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5
> drivers/rtc/Kconfig:991:        prompt "RTT module Number" if ARCH_AT91SAM9263
> drivers/rtc/Kconfig:1000:       range 0 3 if !ARCH_AT91SAM9263
> drivers/rtc/Kconfig:1001:       range 0 15 if ARCH_AT91SAM9263
> drivers/usb/gadget/Kconfig:157: depends on AVR32 || ARCH_AT91SAM9RL || ARCH_AT91SAM9G45
> drivers/usb/gadget/atmel_usba_udc.c:329:#if defined(CONFIG_ARCH_AT91SAM9RL)
> drivers/usb/gadget/atmel_usba_udc.c:349:#endif /* CONFIG_ARCH_AT91SAM9RL */
> drivers/mmc/host/Kconfig:305:   depends on MMC_ATMELMCI && (AVR32 || ARCH_AT91SAM9G45) && DMA_ENGINE
> drivers/usb/gadget/Kconfig:157: depends on AVR32 || ARCH_AT91SAM9RL || ARCH_AT91SAM9G45
> drivers/net/can/Kconfig:68:     depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5

They are being removed one by one and replaced by ARCH_AT91, ARM or even
simply HAVE_CLK...

Have a look at recent patch "[PATCH v2 2/2] can: Kconfig: CAN_AT91
depends on ARM".

But it is true that if you find good patches to remove them, I am in
favor for this move.

Best regards,
Paul Bolle March 14, 2013, 8:17 a.m. UTC | #5 
On Wed, 2013-03-13 at 13:26 +0100, Jean-Christophe PLAGNIOL-VILLARD
wrote:
> but the ARCH_AT91SAM9x5 never exist at the mainline

I'm not sure I understand this.

We're discussing the effects of commit
1e3ce2b8545390a2aee8dbfcd49ca4161b636000 ("AR[M]: at91: introduce
SOC_AT91xxx define to allow to compile SoC core support") here. It did
remove ARCH_AT91SAM9X5:
    $ git show 1e3ce2b8545390a2aee8dbfcd49ca4161b636000 | grep ARCH_AT91SAM9X5
    -config ARCH_AT91SAM9X5
    @@ -93,8 +91,47 @@ config ARCH_AT91SAM9X5
    -obj-$(CONFIG_ARCH_AT91SAM9X5)	+= at91sam9x5.o
    -#ifdef CONFIG_ARCH_AT91SAM9X5
    -#if !defined(CONFIG_ARCH_AT91SAM9X5)

It also added SOC_AT91SAM9X5, and that symbol even triggers the
compilation of a file named at91sam9x5.c:
    $ git show 1e3ce2b8545390a2aee8dbfcd49ca4161b636000 | grep SOC_AT91SAM9X5
    +config SOC_AT91SAM9X5
    +obj-$(CONFIG_SOC_AT91SAM9X5)	+= at91sam9x5.o
    +#ifdef CONFIG_SOC_AT91SAM9X5
    +#if !defined(CONFIG_SOC_AT91SAM9X5)

And also note that ARCH_AT91SAM9X5 still is used as an (optional) dependency:
    $ git grep -n ARCH_AT91SAM9X5 v3.9-rc2
    v3.9-rc2:drivers/net/can/Kconfig:68:    depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5


Paul Bolle
Nicolas Ferre March 14, 2013, 8:30 a.m. UTC | #6 
On 03/14/2013 09:17 AM, Paul Bolle :
> On Wed, 2013-03-13 at 13:26 +0100, Jean-Christophe PLAGNIOL-VILLARD
> wrote:
>> but the ARCH_AT91SAM9x5 never exist at the mainline
> 
> I'm not sure I understand this.
> 
> We're discussing the effects of commit
> 1e3ce2b8545390a2aee8dbfcd49ca4161b636000 ("AR[M]: at91: introduce
> SOC_AT91xxx define to allow to compile SoC core support") here. It did
> remove ARCH_AT91SAM9X5:
>     $ git show 1e3ce2b8545390a2aee8dbfcd49ca4161b636000 | grep ARCH_AT91SAM9X5
>     -config ARCH_AT91SAM9X5
>     @@ -93,8 +91,47 @@ config ARCH_AT91SAM9X5
>     -obj-$(CONFIG_ARCH_AT91SAM9X5)	+= at91sam9x5.o
>     -#ifdef CONFIG_ARCH_AT91SAM9X5
>     -#if !defined(CONFIG_ARCH_AT91SAM9X5)
> 
> It also added SOC_AT91SAM9X5, and that symbol even triggers the
> compilation of a file named at91sam9x5.c:
>     $ git show 1e3ce2b8545390a2aee8dbfcd49ca4161b636000 | grep SOC_AT91SAM9X5
>     +config SOC_AT91SAM9X5
>     +obj-$(CONFIG_SOC_AT91SAM9X5)	+= at91sam9x5.o
>     +#ifdef CONFIG_SOC_AT91SAM9X5
>     +#if !defined(CONFIG_SOC_AT91SAM9X5)
> 
> And also note that ARCH_AT91SAM9X5 still is used as an (optional) dependency:
>     $ git grep -n ARCH_AT91SAM9X5 v3.9-rc2
>     v3.9-rc2:drivers/net/can/Kconfig:68:    depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5

Which is removed by patch:
"[PATCH v2 2/2] can: Kconfig: CAN_AT91 depends on ARM"

It should appear in CAN git tree soon.

Best regards,
Paul Bolle March 14, 2013, 8:52 a.m. UTC | #7 
On Thu, 2013-03-14 at 09:30 +0100, Nicolas Ferre wrote:
> On 03/14/2013 09:17 AM, Paul Bolle :
> > And also note that ARCH_AT91SAM9X5 still is used as an (optional) dependency:
> >     $ git grep -n ARCH_AT91SAM9X5 v3.9-rc2
> >     v3.9-rc2:drivers/net/can/Kconfig:68:    depends on ARCH_AT91SAM9263 || ARCH_AT91SAM9X5
> 
> Which is removed by patch:
> "[PATCH v2 2/2] can: Kconfig: CAN_AT91 depends on ARM"
> 
> It should appear in CAN git tree soon.

Nice. Thanks for the clearing things up.


Paul Bolle
diff mbox

Patch

diff --git a/arch/arm/mach-at91/Kconfig b/arch/arm/mach-at91/Kconfig
index 6071f4c..2db5370 100644
--- a/arch/arm/mach-at91/Kconfig
+++ b/arch/arm/mach-at91/Kconfig
@@ -131,6 +131,10 @@  config ARCH_AT91SAM9G45
        bool "AT91SAM9G45"
        select SOC_AT91SAM9G45
 
+config ARCH_AT91SAM9X5
+       bool "AT91SAM9X5"
+       select SOC_AT91SAM9X5
+
 config ARCH_AT91X40
        bool "AT91x40"
        depends on !MMU