On 15/03/18 19:15, James Morse wrote:
> On 14/03/18 16:50, Marc Zyngier wrote:
>> Although the idmap section of KVM can only be at most 4kB and
>> must be aligned on a 4kB boundary, the rest of the code expects
>> it to be page aligned. Things get messy when tearing down the
>> HYP page tables when PAGE_SIZE is 64K, and the idmap section isn't
>> 64K aligned.
>>
>> Let's fix this by computing aligned boundaries that the HYP code
>> will use.
>
>> diff --git a/virt/kvm/arm/mmu.c b/virt/kvm/arm/mmu.c
>> index 0e5cfffb4c21..a9e0513868e9 100644
>> --- a/virt/kvm/arm/mmu.c
>> +++ b/virt/kvm/arm/mmu.c
>> @@ -1815,7 +1815,9 @@ int kvm_mmu_init(void)
>> int err;
>>
>> hyp_idmap_start = kvm_virt_to_phys(__hyp_idmap_text_start);
>> + hyp_idmap_start = ALIGN_DOWN(hyp_idmap_start, PAGE_SIZE);
>> hyp_idmap_end = kvm_virt_to_phys(__hyp_idmap_text_end);
>> + hyp_idmap_end = ALIGN(hyp_idmap_end, PAGE_SIZE);
>> hyp_idmap_vector = kvm_virt_to_phys(__kvm_hyp_init);
>>
>> /*
>
>
> This makes the:
> | hyp_idmap_start != (unsigned long)__hyp_idmap_text_start
>
> check below look funny, but that must be for 32bit which only has 4K pages (I
> think), so its behaviour hasn't changed.
Yes, that's to handle the 2/2 split on 32bit where some funnies may
happen, see d2896d4b55b2 ("arm: KVM: Fix idmap overlap detection when
the kernel is idmap'ed").
> Reviewed-by: James Morse <james.morse@xxxxxxx>
Thanks,
M.
--
Jazz is not dead. It just smells funny...
