Kexec on arm64

[Arun Chandran]

Hi Geoff,

I think we should better discuss this in the arm mailing list as it is getting
more arm64 specific.

I will post my reply there.

--Arun

On Tue, Jul 15, 2014 at 3:35 AM, Geoff Levand <geoff at infradead.org> wrote:
> Hi Arun,
>
> On Fri, 2014-07-11 at 21:13 +0530, Arun Chandran wrote:
>> I tried kexec reboot with my modified kexec-tools and got a nice kernel panic.
>> Please find the error log attached.
>> There is only dtb file related change in the kernel sources, as given below.
>>
>> ###########
>>                         cpu-release-addr = <0x1 0x0000fff8>;
>> -                       cpu-return-addr = <0 0>
>> +                       cpu-return-addr = <0x0 0x0>;
>> ###########
>>
>> I just assigned a random value to "cpu-return-addr" in my dtb file;
>
> You need the proper cpu-return-addr value.  Because cpu-return-addr
> was zero, kexec sent all your secondary processors back to
> secondary_startup, and is why you see that in your error log.
>
> I put in a hack for zero cpu-return-addr in my latest version.  Please
> try it.  If you use the same kernel for both the 1st and 2nd stage it
> should work.
>
>> Do you have any idea how to find out that value for my hardware?
>
> It is a property of your bootloader, not hardware.  Ask your
> bootloader maintainer if you can, or check the bootloader sources.
>
> For the rtsm_ve_aemv8a model bootwrapper I set cpu-return-addr to
> <0x0 0x80000000> because the bootwrapper linker script puts _start
> at PHYS_OFFSET, which is 0x80000000 for that platform.
>
> https://git.kernel.org/cgit/linux/kernel/git/mark/boot-wrapper-aarch64.git/tree/model.lds.S#n30
>
>> My hardware uses spin-table method for SMP. As you said there
>> is a bug in kexec for spin-table method, what about trying it without
>> CONFIG_SMP? Does your code support that?
>
> CONFIG_SMP=n should work, but I haven't tried it in a long time.
>
> -Geoff
>
>