[PATCH v5 1/3] riscv: Move kernel mapping outside of linear mapping

[Palmer Dabbelt]

On Fri, 11 Jun 2021 04:00:19 PDT (-0700), linux at roeck-us.net wrote:
> On Thu, Jun 10, 2021 at 07:29:15PM +0200, Andreas Schwab wrote:
>> On Jun 10 2021, Guenter Roeck wrote:
>>
>> > On Thu, Jun 10, 2021 at 07:11:38PM +0200, Andreas Schwab wrote:
>> >> On Jun 10 2021, Guenter Roeck wrote:
>> >>
>> >> > On Thu, Jun 10, 2021 at 06:39:39PM +0200, Andreas Schwab wrote:
>> >> >> On Apr 18 2021, Alex Ghiti wrote:
>> >> >>
>> >> >> > To sum up, there are 3 patches that fix this series:
>> >> >> >
>> >> >> > https://patchwork.kernel.org/project/linux-riscv/patch/20210415110426.2238-1-alex@ghiti.fr/
>> >> >> >
>> >> >> > https://patchwork.kernel.org/project/linux-riscv/patch/20210417172159.32085-1-alex@ghiti.fr/
>> >> >> >
>> >> >> > https://patchwork.kernel.org/project/linux-riscv/patch/20210418112856.15078-1-alex@ghiti.fr/
>> >> >>
>> >> >> Has this been fixed yet?  Booting is still broken here.
>> >> >>
>> >> >
>> >> > In -next ?
>> >>
>> >> No, -rc5.
>> >>
>> > Booting v5.13-rc5 in qemu works for me for riscv32 and riscv64,
>> > but of course that doesn't mean much. Just wondering, not knowing
>> > the context - did you provide details ?
>>
>> Does that work for you:
>>
>> https://github.com/openSUSE/kernel-source/blob/master/config/riscv64/default
>>
>
> That isn't an upstream kernel configuration; it looks like includes suse
> patches. But, yes, it does crash almost immediately if I build an upstream
> kernel based on it and try to run that kernel in qemu. I did not try to
> track it down further; after all, it might just be that the configuration
> is inappropriate for use with qemu. But the configuration isn't really
> what I had asked.

This seems a long way off from defconfig.  It's entirly possible I'm 
missing something, but at least CONFIG_SOC_VIRT is jumping out as 
something that's disabled in the SUSE config but enabled upstream.  That 
alone shouldn't actually do anything, but it does ensure we have all the 
drivers necessary to boot on QEMU.

It's entierly possible there's a real bug here, though, as I don't 
really see what these relocatable patches would have to do with that.