On Wed, 20 Feb 2013 15:37:26 -0800, Christoffer Dall
<cdall@xxxxxxxxxxxxxxx> wrote:
> On Wed, Feb 13, 2013 at 03:47:00PM +0000, Marc Zyngier wrote:
>> Simplify stage2_set_pte() by removing the IO mapping handling.
>> kvm_phys_addr_ioremap() now directly calls stage2_get_pte() and
>> stage2_set_pte_at().
>>
>> This allows for the removal of the iomap flag in stage2_set_pte()
>> prototype.
>>
>> Signed-off-by: Marc Zyngier <marc.zyngier@xxxxxxx>
>> ---
>> arch/arm/kvm/mmu.c | 22 +++++++++++-----------
>> 1 file changed, 11 insertions(+), 11 deletions(-)
>>
>> diff --git a/arch/arm/kvm/mmu.c b/arch/arm/kvm/mmu.c
>> index ab35906..8fde75b 100644
>> --- a/arch/arm/kvm/mmu.c
>> +++ b/arch/arm/kvm/mmu.c
>> @@ -459,18 +459,13 @@ static void stage2_set_pte_at(struct kvm *kvm,
>> phys_addr_t addr,
>> }
>>
>> static int stage2_set_pte(struct kvm *kvm, struct kvm_mmu_memory_cache
>> *cache,
>> - phys_addr_t addr, const pte_t *new_pte, bool iomap)
>> + phys_addr_t addr, const pte_t *new_pte)
>> {
>> pte_t *pte;
>>
>> pte = stage2_get_pte(kvm, cache, addr);
>> - if (!pte)
>> - return 0;
>> -
>> - if (iomap && pte_present(*pte))
>> - return -EFAULT;
>> -
>> - stage2_set_pte_at(kvm, addr, pte, *new_pte);
>> + if (pte)
>> + stage2_set_pte_at(kvm, addr, pte, *new_pte);
>
> this should never happen now right? I think we can get rid of this
> function (but I think stage2_set_pte_at should be named stage2_set_pte).
It does happen. See the call site in kvm_set_spte_handler, which calls
stage_set_pte with a NULL memory cache.
As for the name, stage2_set_pte is what I'm trying to get rid off, so
calling the new function the same would be counter productive... ;-)
Furthermore, set_pte_at is the kernel function to do a similar operation
on a stage-1 page table. Consistency is important IMHO.
M.
--
Fast, cheap, reliable. Pick two.
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
