Re: [Qemu-arm] [PATCH v3] hw/arm: Add arm SBSA reference machine

[Hongbo Zhang]

On 27 September 2018 at 18:30, Peter Maydell <address@hidden> wrote:
> On 27 September 2018 at 11:23, Hongbo Zhang <address@hidden> wrote:
>> I would mention that there are some issues in the base qemu tree to
>> boot this (and also 'virt') machine.
>>
>> The MPIDR isn't emulated well, currently only 4 cores on one cluster
>> can be booted, if we have more cores / more clusters / different
>> number of cores on one clusters, SMP cannot boot correctly because
>> PSCI checks MPIDR for booting secondary cores.
>
> 8-core SMP boot works fine for me on 'virt'...
>
Also depends on GIC version, and firmware usage, e.g. QEMU built-in
PSCI or external ARMTF PSCI, let me check and summarize.

>> And I have to revert "device_tree: Increase FDT_MAX_SIZE to 1 MiB" and
>> revert "target/arm: Implement new do_transaction_failed hook" to boot
>> the system, these commits not only block this new sbsa machine , but
>> also the 'virt' machine, maybe the patches need to be checked or maybe
>> firmware needed some adapt to these commits, currently I just revert
>> them to boot.
>
> Problems with do_transaction_failed mean the guest tried
> to access some physical address where there is no device. This
> is always a guest bug or misconfiguration. The commit you cite
> there is the one that means we stop ignoring this and instead
> generate guest bus errors, as real hardware usually does.
>
> Not sure about the device_tree one, but again, 'virt' machine
> works OK for me.
>
You didn't boot with ARMTF and UEFI I think, when both firmwares used,
you likely get blocked at UEFI with message:
(so I guess it is due to UEFI to do some adaption)

ASSERT_EFI_ERROR (Status = Not Found)
ASSERT [ResetSystemRuntimeDxe]
/home/hongbo/work/edk2/Build/ArmVirtQemuKernel-AARCH64/DEBUG_GCC5/AARCH64/MdeModulePkg/Universal/ResetSystemRuntimeDxe/ResetSystemRuntimeDxe/DEBUG/AutoGen.c(362):
!EFI_ERROR (Status)

>> I will add a cover letter for the next iteration included explanation
>> of these issues.
>
> Thanks. Where they affect the 'virt' board, instructions to reproduce
> there would be useful.
>
Yes, sure, will explain issues in detail. (when back from my National
Day vocation)

> -- PMM