Re: [PATCH v7 4/4] configs: sama5d2: enable option CONFIG_ATMEL_TCB_TIMER

[<Eugen.Hristev@microchip.com>]

On 3/31/22 10:40 AM, Clément Léger wrote:
> Le Thu, 31 Mar 2022 05:59:02 +0000,
> <Eugen.Hristev@microchip.com> a écrit :
> 
>> On 3/30/22 6:49 PM, Clément Léger wrote:
>>> Le Tue, 22 Mar 2022 13:35:08 +0100,
>>> Clément Léger <clement.leger@bootlin.com> a écrit :
>>>
>>>> Le Tue, 22 Mar 2022 11:18:39 +0000,
>>>> <Eugen.Hristev@microchip.com> a écrit :
>>>>
>>>
>>> [...]
>>>
>>>>>
>>>>> Could not initialize timer (err -22)
>>>>>
>>>>> ...
>>>>>
>>>>>
>>>>>
>>>>> Could you investigate this please ?
>>>>
>>>> Hi Eugen,
>>>>
>>>> Ok, I'll try to debug that !
>>>
>>> Hi Eugen, while debugging it, I found that there was actually one major
>>> flaw. The clock that I used was the GLCK. If only this clock is
>>> enabled, the TCB can not work, the tcb0_clk must also be enabled for the
>>> TCB to start counting. It worked on my side because I forgot that I had
>>> a quirk in OP-TEE that was enabling the clocks before starting U-Boot.
>>> Without that, it actually could not work.
>>>
>>> Since the tcb0_clk is needed, I switched the driver to use this one
>>> only and it works well on a "bare" setup (ie without OP-TEE). I even
>>> think it could not work with the tcb0_gclk since this one does not
>>> correctly handles the case when the parent clock is the slow_clk (which
>>> might be the case after booting) and in this case, will return 0 as
>>> rate (this is by the way supported in your CCF branch).
>>>
>>> Anyway, I fixed the driver to use the tcb0_clk and thus it should work
>>> much better. I will send a V8 and it should work with U-Boot and the
>>> SPL.
>>>
>>> By the way, do you have any documentation which explains how to use
>>> U-Boot SPL on sama5d2_xplained (or other) board ?
>>
>> Hi Clement,
>>
>> I don't have any documentation at hand, but what I can say in a few
>> words is that once you build the SPL, you should have a file
>> spl/boot.bin and this binary should replace our stage2 at91 bootloader
>> at91bootstrap.
>>
>> Unlike at91bootstrap, the SPL has all the drivers inside and can boot
>> from all supported media, however, SPL will read a dedicated CPU
>> register where the stage1 has written some bits representing the index
>> of the NVM *used to copy the boot.bin from* . The SPL will then attempt
>> to read the u-boot.bin from the same media. (you can alter the file name
>> in defconfig ).
> 
> Ok, I successfully booted a SPL < 64k from SD card by disabling MMC
> support after fixing the TCB driver:
> 
> ;4��K3K{�KOp�C�0�
> U-Boot SPL 2022.04-rc3-00023-gf9da67b41a-dirty (Mar 31 2022 - 09:23:48
> +0200)
> SPL: Unsupported Boot Device!
> SPL: failed to boot from all boot devices
> ### ERROR ### Please RESET the board ###
> 
> However, as soon as the size is greater than 64K, it does not boot
> anymore. Looks like you already had the same problem. Do you have any
> more documentation which would allow me to boot a u-boot.bin file from
> this SPL easily ?

Hi,

If size is above 64k, it won't fit into the SRAM, thus the stage1 won't 
copy it at all, and abort booting it.

For arm926 there is a build time check for the size in the lds linker 
script, but it looks like for the armv7 it's missing... No idea why , if 
it was removed, I haven't checked.

I don't have any more documentation.

> 
> By the way, since u-boot,dm-pre-reloc is needed in the device-tree to
> be used with the SPL, I will probably add them to my series
> (property which by the way is missing for the PIT timer
> device-tree node).

Yes, definitely. You can send a prequel or separate patch to add this to 
the PIT timer if you like.
And have this property for your node yes.

Normally all u-boot properties should be in a separate dtsi file, but 
this is an old DT in uboot so I think it's fine to have it in the base DT.

Eugen

> 
> Thanks,
> 
> Clément
> 
> --
> Clément Léger,
> Embedded Linux and Kernel engineer at Bootlin
> https://bootlin.com
>