Re: IMX8M and Optee support

[Ahmad Fatoum <a.fatoum@pengutronix.de>]

Hello Hans,

On 18.10.23 16:06, Hans Christian Lønstad wrote:
> The ATF for imx8 is quite convoluted requiring fixed memory locations for images and device trees. The ATF itself supports transferring parameters
> Using 4 unsigned long arguments. These are unfortunately zeroed out by a misconfiguration for IMX8:
> 
> RESET_TO_BL31 := 1
> and not at the same time setting
> RESET_TO_BL31_WITH_PARAMS := 1
> 
> As demoed below this can be made to work.
> ———><——
> 
> Uart initialized
> Run level 3
> Init power
> Init DDR
> Handover to ATF
> imx8mp_load_and_start_image_via_tfa: Expect OPTEE at 0xbe000000
> CH e2a7175151fe3e842116990e62c864334e1bb030ca146b749d8e6b0c02481357
> IH e2a7175151fe3e842116990e62c864334e1bb030ca146b749d8e6b0c02481357
>                                                                    CAAM early init successful
> imx8m_atf_start_bl31: Callout BL31 with 1, 2, 3, 4
> NOTICE:  Do not release JR0 to NS as it can be used by HAB
> INFO:    bl31_early_platform_setup2: args are 0x00000001, 0x00000002, 0x00000003, 0x00000004
> NOTICE:  BL31: v2.8(debug):lf-6.1.36-2.1.0-0-g1a3beeab6-dirty
> NOTICE:  BL31: Built : 15:49:38, Oct 18 2023
> 
> ——><——
> 
> The "best" approach is likely to patch ATFs imx8mX_bl31_setup.c replacing preprocessor values with values
> handed over from Barebox. Adding these parameters will be compatible with unpatched ATFs anyway.

There's void bl31_entry(uintptr_t bl31_entry, uintptr_t bl32_entry,
		        uintptr_t bl33_entry, uintptr_t fdt_addr)

which is used for RK3568/RK3588 platforms. AFAIU, future SoCs will likely
use a linked list of hand over blocks to transfer this same information.

Any of these two approaches would be better than what we have now on i.MX.

I am still not sure how this relates to OP-TEE hanging. Can you configure
your OP-TEE to not expect a FDT?

Cheers,
Ahmad

> 
> Hans
> 
> 
>> 18. okt. 2023 kl. 15:04 skrev Ahmad Fatoum <a.fatoum@pengutronix.de>:
>>
>> Hello,
>>
>> On 18.10.23 14:31, Hans Christian Lønstad wrote:
>>> I may have gotten this wrong, but should not the device tree be passed over to ATF which again pass it over to Optee?
>>>
>>> The bl31() wrapper in imx8m_atf_start_bl31 should do this using register call parameters?
>>
>> I am using a quite antiquated imx-atf v2.4 and optee-os-3.2.0.imx in my setup and
>> haven't come around to update them. I know that barebox works fine with newer TF-A
>> versions (especially after https://review.trustedfirmware.org/c/TF-A/trusted-firmware-a/+/16370),
>> but I haven't tested a newer OP-TEE, so if your OP-TEE configuration/version expects
>> a device tree, this may indeed be the culprit.
>>
>>>> 18. okt. 2023 kl. 13:29 skrev Hans Christian Lønstad <hcl@datarespons.no>:
>>>>
>>>> We have a 2GB IMX8MP system using 32MB OPTEEE_SIZE, so expect to find Optee blob at 0xBE000000.
>>>>
>>>> ATF compiled with:
>>>> make PLAT=imx8mp BL32_BASE=0xBE000000 IMX_BOOT_UART_BASE=0x30890000 SPD=opteed DEBUG=1 -j
>>>>
>>>> Optee compiled with:
>>>>
>>>> CFG_DDR_SIZE ?= UL(0x80000000)
>>>> CFG_UART_BASE ?= UART2_BASE
>>>> CFG_TZDRAM_START ?= 0xBE000000
>>>> undefine CFG_NS_ENTRY_ADDR
>>>>
>>>> Barebox compiled with:
>>>> CONFIG_HAVE_OPTEE=y CONFIG_OPTEE_SIZE=0x02000000 # CONFIG_BOOTM_OPTEE is not set CONFIG_PBL_OPTEE=y CONFIG_FIRMWARE_IMX8MP_OPTEE=y
>>
>> barebox configuration looks ok assuming board code calls imx8mp_load_and_start_image_via_tfa().
>>
>>>>
>>>> This produces the following on boot:
>>>> —————————————————><--------------
>>>>
>>>> Uart initialized
>>>> Run level 3
>>>> Init power
>>>> Init DDR
>>>> Handover to ATF
>>>> imx8mp_load_and_start_image_via_tfa: Expect OPTEE at 0xbe000000
>>>> CH e2a7175151fe3e842116990e62c864334e1bb030ca146b749d8e6b0c02481357
>>>> IH e2a7175151fe3e842116990e62c864334e1bb030ca146b749d8e6b0c02481357
>>>>                                                                  NOTICE:  Do not release JR0 to NS as it can be used by HAB
>>>> NOTICE:  BL31: v2.8(debug):lf-6.1.36-2.1.0-0-g1a3beeab6-dirty
>>>> NOTICE:  BL31: Built : 13:14:09, Oct 18 2023
>>>> INFO:    GICv3 with legacy support detected.
>>>> INFO:    ARM GICv3 driver initialized in EL3
>>>> INFO:    Maximum SPI INTID supported: 191
>>>> INFO:    BL31: Initializing runtime services
>>>> INFO:    bl31_plat_get_next_image_ep_info: want image 0
>>>> INFO:    bl31_plat_get_next_image_ep_info: bl32 PC is 0xbe000000
>>>> INFO:    BL31: cortex_a53: CPU workaround for 855873 was applied
>>>> WARNING: BL31: cortex_a53: CPU workaround for 1530924 was missing!
>>>> INFO:    BL31: Initializing BL32
>>>> INFO:    bl31_plat_get_next_image_ep_info: want image 0
>>>> INFO:    bl31_plat_get_next_image_ep_info: bl32 PC is 0xbe000000
>>>> INFO:    opteed_init: 176 - calling <opteed_synchronous_sp_entry>
>>>> INFO:    opteed_synchronous_sp_entry: 79 - calling <opteed_enter_sp>
>>
>> Try compiling op-tee with CFG_TEE_CORE_LOG_LEVEL=4 and see how far it gets.
>>
>>>> I´m not sure why it asks for the same image twice and if this implies anything …
>>
>> No idea..
>>
>> Cheers,
>> Ahmad
>>
>> -- 
>> Pengutronix e.K.                           |                             |
>> Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
>> 31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
>> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
>>
> 

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |