LAW Config: PCIe0 --- 0xf80000000
Core hang during read address 0xf80000000. It looks like an unrecoverable error while accessing the address.
Root Cause :
OB_CK bit of PCI Express Configuration Register (PEX_CONFIG offset is 0x14) was not set.
It is normal that Machine Check is generated when a PCIe read transaction addressed memory region assigned to a target device which is not present.
In this case no valid data can be returned, so asynchronous Machine Check is generated.
> What is the correct CPU response when accessing 0xf80000000?
What is mapped to this address?
Please provide U-Boot booting log.
ufedor, Thanks for your replay!
>What is mapped to this address?
I just specify the address 0xf80000000(it could be any other address) as PCIe space through LAW and no real PCIe device assigned to this address.
> What is the correct CPU response when accessing 0xf80000000?
Not clear, I'm asking for this because I caught the machine check exception on T2080RDB CPU version 1.1 but CPU hanged on T2080RDB CPU version 1.0, T2080QDS and T4240QDS.
>In this case no valid data can be returned, so asynchronous Machine Check is generated.
I think this is the real reason for this problem, accessing the PCIe address without a hardware populated may cause an asynchronous MCK.
And does the asynchronous MCK will cause CPU to hang?
> And does the asynchronous MCK will cause CPU to hang?
Please use a debugger to determine what the processor is doing during the hang state.
Please consider that it could be a read instruction which can't be completed because a valid data is not provided.
I have done that before and that is why i have posted this question page. When a load assembly instruction "lwz r6,0(r5)" (reading address 0xf80000000)was executed, the CodeWarrior TAP was disconnected and the value of registers MSR/MCSR and other status registers changed to meaningless. And the core has no response anymore.
However thanks for your replay.