Re: [PATCH v4] OMAP3: introduce OMAP3630

[Tony Lindgren <tony@atomide.com>]

* Nishanth Menon <nm@ti.com> [091009 11:47]:
> Tony Lindgren had written, on 10/09/2009 01:03 PM, the following:
>> * Aguirre Rodriguez, Sergio Alberto <saaguirre@ti.com> [091009 08:59]:
>>>  
>>>
>>>> -----Original Message-----
>>>> From: Menon, Nishanth Sent: Friday, October 09, 2009 10:56 AM
>>>> To: linux-omap
>>>> Cc: Menon, Nishanth; Chikkature Rajashekar, Madhusudhan; Pandita, 
>>>> Vikram; Pais, Allen; Gadiyar, Anand; Cousson, Benoit; Felipe Balbi; 
>>>> Kevin Hilman; Premi, Sanjeev; Shilimkar, Santosh; Aguirre 
>>>> Rodriguez, Sergio Alberto; Tony Lindgren
>>>> Subject: [PATCH v4] OMAP3: introduce OMAP3630
>>>>
>>>> **--- SNIP ME --
>>>> Hi,
>>>> A bit of history for this patchset
>>>> V4 - Sergio's improvement to the handling of rev decision as we just
>>>>      change the define to 3630 and remove crapy override logic in
>>>>      omap3_check_revision from v3 patch
>>> Now looks much better ;)
>>>
>>>> V3 - Fixes from Sergio's comments + boot tested on SDP3430+3630.
>>>> V2 - fixes of generic comments from Felipe Balbi+minor cleanups
>>>> V1 - inital implementation of (a) approach
>>>> V0 - original approach introducing a new silicon family
>>>> **--- END OF SNIP ME --
>>>> Device intro:
>>>> OMAP3630 is the latest in the family of OMAP3 devices
>>>> and among the changes it introduces are:
>>>>
>>>> New OPP levels for new voltage and frequency levels. a bunch of
>>>> Bug fixes to various modules feature additions, notably with ISP,
>>>> sDMA etc.
>>>>
>>>> Details about the chip is available here:
>>>> http://focus.ti.com/general/docs/wtbu/wtbuproductcontent.tsp?t
>>>> emplateId=6123&navigationId=12836&contentId=52606
>>>>
>>>> Strategy used:
>>>> Strategy to introduce this device into Linux was discussed here:
>>>> Ref: http://marc.info/?t=125343303400003&r=1&w=2
>>>>
>>>> Two approaches were available:
>>>> a) Consider 3630 generation of devices as a new family of silicon
>>>> b) Consider 3630 as an offshoot of 3430 family of devices
>>>>
>>>> As a common consensus, (b) seems to be more valid for 3630 as:
>>>> * There are changes which are easily handled by using "FEATURES"
>>>>   infrastructure.
>>>>   For details how to do this, see thread:
>>>>   http://marc.info/?t=125050998500001&r=1&w=2
>>>> * Most of existing 34xx infrastructure can be reused(almost 90%+)
>>>> 	- so no ugly if (cpu_is_omap34xx() || cpu_is_omap36xx())
>>>> 	  all over the place
>>>> 	- lesser chance of bugs due to reuse of proven code flow
>>>> 	- 36xx specific handling can still be done where required
>>>> 	  within the existing infrastructure
>>>>
>>>> NOTE:
>>>> * If additional 34xx series are added, OMAP3430_REV_ESXXXX can be
>>>>   added on top of the existing 3630 ones are renumbered
>>>>
>>>> This patch was tested on SDP3430, boot tested on 3630 platform using
>>>> 3430sdp defconfig
>>>>
>>>> Signed-off-by: Madhusudhan Chikkature Rajashekar <madhu.cr@ti.com>
>>>> Signed-off-by: Nishanth Menon <nm@ti.com>
>>>> Signed-off-by: Vikram Pandita <vikram.pandita@ti.com>
>>>> Cc: Allen Pais <allen.pais@ti.com>
>>>> Cc: Anand Gadiyar <gadiyar@ti.com>
>>>> Cc: Benoit Cousson <b-cousson@ti.com>
>>>> Cc: Felipe Balbi <felipe.balbi@nokia.com>
>>>> Cc: Kevin Hilman <khilman@deeprootsystems.com>
>>>> Cc: Sanjeev Premi <premi@ti.com>
>>>> Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
>>>> Cc: Sergio Alberto Aguirre Rodriguez <saaguirre@ti.com>
>>>> Cc: Tony Lindgren <tony@atomide.com>
>>> Acked-by: Sergio Aguirre <saaguirre@ti.com>
>>>
>>>> ---
>>>>  arch/arm/mach-omap2/id.c              |   24 ++++++++++++++++++++++--
>>>>  arch/arm/plat-omap/include/mach/cpu.h |    6 ++++++
>>>>  2 files changed, 28 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/arch/arm/mach-omap2/id.c b/arch/arm/mach-omap2/id.c
>>>> index 03b80f2..ee3bb69 100644
>>>> --- a/arch/arm/mach-omap2/id.c
>>>> +++ b/arch/arm/mach-omap2/id.c
>>>> @@ -210,7 +210,9 @@ void __init omap3_check_revision(void)
>>>>  	hawkeye = (idcode >> 12) & 0xffff;
>>>>  	rev = (idcode >> 28) & 0xff;
>>>>  -	if (hawkeye == 0xb7ae) {
>>>> +	switch (hawkeye) {
>>>> +	case 0xb7ae:
>>>> +		/* Handle 34xx devices */
>>>>  		switch (rev) {
>>>>  		case 0:
>>>>  			omap_revision = OMAP3430_REV_ES2_0;
>>>> @@ -231,8 +233,26 @@ void __init omap3_check_revision(void)
>>>>  		default:
>>>>  			/* Use the latest known revision as default */
>>>>  			omap_revision = OMAP3430_REV_ES3_1;
>>>> -			rev_name = "Unknown revision\n";
>>>> +			rev_name = "Unknown 34xx revision\n";
>>>>  		}
>>>> +		break;
>>>> +	case 0xb891:
>>>> +		/* Handle 36xx devices */
>>>> +		switch (rev) {
>>>> +		case 0:
>>>> +			omap_revision = OMAP3630_REV_ES1_0;
>>>> +			rev_name = "ES1.0";
>>>> +			break;
>>>> +		default:
>>>> +			/* Use the latest known revision as default */
>>>> +			omap_revision = OMAP3630_REV_ES1_0;
>>>> +			rev_name = "Unknown 36xx revision\n";
>>>> +		}
>>>> +		break;
>>>> +	default:
>>>> +		/* Unknown default to latest rev as default*/
>>>> +		omap_revision = OMAP3630_REV_ES1_0;
>>>> +		rev_name = "Unknown revision\n";
>>>>  	}
>>>>   out:
>>>> diff --git a/arch/arm/plat-omap/include/mach/cpu.h  
>>>> b/arch/arm/plat-omap/include/mach/cpu.h
>>>> index 431fec4..65f1882 100644
>>>> --- a/arch/arm/plat-omap/include/mach/cpu.h
>>>> +++ b/arch/arm/plat-omap/include/mach/cpu.h
>>>> @@ -383,6 +383,12 @@ IS_OMAP_TYPE(3430, 0x3430)
>>>>  #define OMAP3430_REV_ES2_1	0x34302034
>>>>  #define OMAP3430_REV_ES3_0	0x34303034
>>>>  #define OMAP3430_REV_ES3_1	0x34304034
>>>> +/* NOTE: Add 36xx series below
>>>> + * If additional 34xx series are added, OMAP3430_REV_ESXXXX can be
>>>> + * added above the 3630 defines and series renumbered to ensure
>>>> + * rev() > checks to work
>>>> + */
>>>> +#define OMAP3630_REV_ES1_0	0x36301034
>>
>> Hmm, to me it looks that this breaks cpu_is_omap34xx() for 36xx, right?
>>
>> How about build a kernel that boots both on 3430sdp and on some
>> 36xx board and make sure things get detected properly by adding
>> some debug printk statements?
> Tested with the following patch:
> diff --git a/arch/arm/mach-omap2/board-3430sdp.c  
> b/arch/arm/mach-omap2/board-3430sdp.c
> index 81aabac..2aac26d 100644
> --- a/arch/arm/mach-omap2/board-3430sdp.c
> +++ b/arch/arm/mach-omap2/board-3430sdp.c
> @@ -498,6 +498,8 @@ static struct ehci_hcd_omap_platform_data ehci_pdata  
> __initconst = {
>
>  static void __init omap_3430sdp_init(void)
>  {
> +       if (cpu_is_omap34xx())
> +               printk(KERN_ERR "_______________------ 34xx detected\n");
>         omap3430_i2c_init();
>         platform_add_devices(sdp3430_devices, ARRAY_SIZE(sdp3430_devices));
>         if (omap_rev() > OMAP3430_REV_ES1_0)
>
>
> and yes, it prints on both 3430 and 3630 platforms (i am hacking at the  
> moment by using 3430sdpdefconfig for both platforms).
>
> Here is why it works:
> arch/arm/plat-omap/include/mach/cpu.h
> cpu_is_omap343x() depends on is_omap34xx() depends on  GET_OMAP_CLASS  
> depends on lower 8 bits of value returned by omap_rev()
> omap_rev()(arch/arm/mach-omap2/id.c) returns omap_revision. this means  
> omap_rev() will return 0x36301034 -> so class is 0x34.. causing  
> cpu_is_omap343x to be true on 3630 also.

Ah, that's right, sorry for being confused. I totally forgot we're using
the lower bits for the class.. I guess I need to start reading the code
more :)

> we might want to consider a follow on patch replacing 0x34 with 0x30 and  
> changing cpu_is_omap343x with cpu_is_omap3xxx -> at a later point 
> ofcourse.

Yeah if necessary. Let's not do it right now though, we already have
quite a pile of clean-up patches coming for the next merge window.

Regards,

Tony