Re: [PATCH v2 2/2] xen/ppc: Implement a basic exception handler

[Jan Beulich <jbeulich@xxxxxxxx>]

On 19.10.2023 22:02, Shawn Anastasio wrote:
> On 10/18/23 10:43 AM, Jan Beulich wrote:
>> On 13.10.2023 20:13, Shawn Anastasio wrote:
>>> --- a/xen/arch/ppc/setup.c
>>> +++ b/xen/arch/ppc/setup.c
>>> @@ -11,6 +11,15 @@
>>>  /* Xen stack for bringing up the first CPU. */
>>>  unsigned char __initdata cpu0_boot_stack[STACK_SIZE] __aligned(STACK_SIZE);
>>>
>>> +void setup_exceptions(void)
>>> +{
>>> +    unsigned long lpcr;
>>> +
>>> +    /* Set appropriate interrupt location in LPCR */
>>> +    lpcr = mfspr(SPRN_LPCR);
>>> +    mtspr(SPRN_LPCR, lpcr | LPCR_AIL_3);
>>> +}
>>
>> Please forgive my lack of PPC knowledge: There's no use of any address
>> here afaict. Where's the link to (presumably) AIL_VECTOR_BASE? If that
>> address is kind of magic, I'm then lacking a connection between
>> XEN_VIRT_START and that constant. (If Xen needed moving around in
>> address space, it would be nice if changing a single constant would
>> suffice.)
>>
> 
> AIL_VECTOR_BASE is indeed a magic address defined by the ISA for AIL=3.
> As for the second part of your question, I'm a bit confused as to what
> you're asking. The ISRs are placed at a position relative to
> the start of the .text.exceptions section (EXCEPTION_VECTORS_START), so
> Xen can be arbitrarily shuffled around in address space as long as
> EXCEPTION_VECTORS_START lies at or before AIL_VECTOR_BASE.

Well, AIL_VECTOR_BASE is #define-d to a plain constant, not derived
from EXCEPTION_VECTORS_START. In turn EXCEPTION_VECTORS_START is
#define-d to a plain constant in patch 1, not derived from
XEN_VIRT_START. Therefore moving Xen around would require to change
(at least) 3 #define-s afaict.

Jan