On 28 August 2012 16:00, Christoffer Dall <c.dall@xxxxxxxxxxxxxxxxxxxxxx> wrote:
> On Thu, Aug 23, 2012 at 7:20 AM, Christoffer Dall
> <c.dall@xxxxxxxxxxxxxxxxxxxxxx> wrote:
>> On Thu, Aug 23, 2012 at 3:19 AM, Rusty Russell <rusty.russell@xxxxxxxxxx> wrote:
>>> + if (par & (1U << 11)) {
>>> + /* LPAE PAR format */
>>> + /*
>>> + * TODO: Check if this ever happens
>>> + * - called from Hyp mode
>>> + */
>>
>> Getting a chance to revisit this, surely it has to happen if the guest
>> uses LPAE - otherwise you wouldn't be able to get the top bits from
>> the VA to PA operation executed from Hyp mode.
>>
>> Peter/Marc, can you verify?
__kvm_va_to_pa always switches to hyp mode to do the translation,
right? ATS1Cxx operations from Hyp mode always return the answer
using the 64-bit PAR format (see ARM ARM B3.14.3). So bit 11 is
always set. (If you do the operation from PL1 then the format of
the PAR depends on whether the TTBR.EAE bit is set, but this isn't
true for operations done from Hyp mode, to avoid the hypervisor
having to care whether the guest is using long format descriptor
or not.)
-- PMM
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
