[U-Boot] [PATCH] arm: Tegra: fix undefined instruction hang immediately after reset

[Aneesh V]

Tom,

On Thursday 08 December 2011 09:07 PM, Tom Warren wrote:
> Aneesh&  Mike,
>
>> -----Original Message-----
>> From: Aneesh V [mailto:aneesh at ti.com]
>> Sent: Wednesday, December 07, 2011 10:51 PM
>> To: Mike Frysinger
>> Cc: u-boot at lists.denx.de; trini at ti.com; Tom Warren; Tom Warren
>> Subject: Re: [U-Boot] [PATCH] arm: Tegra: fix undefined instruction hang
>> immediately after reset
>>
>> Mike,
>>
>> On Thursday 08 December 2011 05:44 AM, Mike Frysinger wrote:
>>> On Tuesday 06 December 2011 18:00:19 Tom Warren wrote:
>>>> commit 0d479b53 (Aneesh V) added code for OMAP4 that doesn't execute
>>>> on Tegra, due to the AVP (ARM7TDI) not having a CP15.
>>>> Result was an undefined instruction hang just after reset.
>>>>
>>>> --- a/arch/arm/cpu/armv7/start.S
>>>> +++ b/arch/arm/cpu/armv7/start.S
>>>>
>>>> +#if !defined(CONFIG_TEGRA2)
>>>>    /*
>>>>     * Setup vector:
>>>>     * (OMAP4 spl TEXT_BASE is not 32 byte aligned.
>>>> @@ -159,6 +160,7 @@ reset:
>>>>    	ldr	r0, =_start
>>>>    	mcr	p15, 0, r0, c12, c0, 0	@Set VBAR
>>>>    #endif
>>>> +#endif	/* !Tegra2 */
>>>
>>> forgive my ignorance, but would it be better to invert the logic ?
>>> have ARM cores that do have a CP15 define CONFIG_ARM_CP15 (or
>>> whatever) and then put all this logic behind that rather than grow a list
>> of SoC's that lack it ?
>>> -mike
>>
>> As far as I understand CP15 is typically available(if not mandatory) on all
>> armv7 processors. Here, IIUC, NVidia has a peculiar architecture that
>> necessitates an armv4 processor supported by armv7 code. IMHO, this is the
>> exceptional case.
>
> Tegra has (at least) 2 ARM processors - the AVP, which is ARM7TDMI-based and hence has no CP15, and the CPU complex, which as 2 or more A9 CPUs.  Since the AVP is the boot processor, it runs the initial U-Boot code (start.S) and will eventually setup and kick off the CPU(s) to run the bulk of U-Boot.
>
> Perhaps we should be in our own arch/arm/cpu/arm7tdmi subdir, but that's a major change from the current structure. Is it possible for this code ("Setup vector") to only be defined for OMAP processor builds? The code it replaces was ifdef'd for OMAP3, and had no affect on Tegra builds.

I think setting up vector is useful for all platforms and I think this
is the right place to do it. Just that we didn't have it till now ecept
for OMAP3 and OMAP3 was doing it in a round-about way.

>
> Unless some other SoC decides to use an ARM7TDMI-S boot proc, though, the CONFIG_TEGRA2 fencing should be the only one necessary.

I agree. I don't think there will be many more exceptions like this. We
can live with that #ifdef. Not sure what Albert thinks though.

>
> Tom
>>
>> br,
>> Aneesh
> -----------------------------------------------------------------------------------
> This email message is for the sole use of the intended recipient(s) and may contain
> confidential information.  Any unauthorized review, use, disclosure or distribution
> is prohibited.  If you are not the intended recipient, please contact the sender by
> reply email and destroy all copies of the original message.
> -----------------------------------------------------------------------------------