Re: [PATCH 6/6] powerpc/rtas: constrain user region allocation to RMA

[Nathan Lynch <nathanl@linux.ibm.com>]

Michael Ellerman <mpe@ellerman.id.au> writes:
> Nathan Lynch <nathanl@linux.ibm.com> writes:
>> Michael Ellerman <mpe@ellerman.id.au> writes:
>>> Nathan Lynch <nathanl@linux.ibm.com> writes:
>>>> Memory locations passed as arguments from the OS to RTAS usually need
>>>> to be addressable in 32-bit mode and must reside in the Real Mode
>>>> Area. On PAPR guests, the RMA starts at logical address 0 and is the
>>>> first logical memory block reported in the LPAR’s device tree.
>>>>
>>>> On powerpc targets with RTAS, Linux makes available to user space a
>>>> region of memory suitable for arguments to be passed to RTAS via
>>>> sys_rtas(). This region (rtas_rmo_buf) is allocated via the memblock
>>>> API during boot in order to ensure that it satisfies the requirements
>>>> described above.
>>>>
>>>> With radix MMU, the upper limit supplied to the memblock allocation
>>>> can exceed the bounds of the first logical memory block, since
>>>> ppc64_rma_size is ULONG_MAX and RTAS_INSTANTIATE_MAX is 1GB.
>>>
>>> Why does the size of the first memory block matter for radix?
>>
>> Here is my understanding: in the platform architecture, the size of the
>> first memory block equals the RMA, regardless of the MMU mode. It just
>> so happens that when using radix, Linux can pass ibm,configure-connector
>> a work area address outside of the RMA because the allocation
>> constraints for the work area are computed differently. It would be
>> wrong of the OS to pass RTAS arguments outside of this region with hash
>> MMU as well.
>
> If that's the requirement then shouldn't we be adjusting ppc64_rma_size?
> Otherwise aren't other uses of ppc64_rma_size going to run into similar
> problems.

Not all allocations limited by ppc64_rma_size set up memory that is
passed to RTAS though, do they? e.g. emergency_stack_init and
init_fallback_flush? Those shouldn't be confined to the first LMB
unnecessarily.

That's why I'm thinking what I've written here should be generalized a
bit and placed in an early allocator function that can be used to set up
the user region and the per-cpu reentrant RTAS argument buffers
(see allocate_paca_ptrs/new_rtas_args). So far those two sites are the
only ones I'm convinced need attention.