On Mon, Nov 10, 2014 at 12:05 PM, Ard Biesheuvel
<ard.biesheuvel@xxxxxxxxxx> wrote:
> On 10 November 2014 11:53, Christoffer Dall <christoffer.dall@xxxxxxxxxx> wrote:
>> Hi Ard,
>>
>> On Mon, Nov 10, 2014 at 09:33:56AM +0100, Ard Biesheuvel wrote:
>>> This reverts commit 85c8555ff0 ("KVM: check for !is_zero_pfn() in
>>> kvm_is_mmio_pfn()") and renames the function to kvm_is_reserved_pfn.
>>>
>>> The problem being addressed by the patch above was that some ARM code
>>> based the memory mapping attributes of a pfn on the return value of
>>> kvm_is_mmio_pfn(), whose name indeed suggests that such pfns should
>>> be mapped as device memory.
>>>
>>> However, kvm_is_mmio_pfn() doesn't do quite what it says on the tin,
>>> and the existing non-ARM users were already using it in a way which
>>> suggests that its name should probably have been 'kvm_is_reserved_pfn'
>>> from the beginning, e.g., whether or not to call get_page/put_page on
>>> it etc. This means that returning false for the zero page is a mistake
>>> and the patch above should be reverted.
>>>
>>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>
>>> ---
>>> arch/ia64/kvm/kvm-ia64.c | 2 +-
>>> arch/x86/kvm/mmu.c | 6 +++---
>>> include/linux/kvm_host.h | 2 +-
>>> virt/kvm/kvm_main.c | 16 ++++++++--------
>>> 4 files changed, 13 insertions(+), 13 deletions(-)
>>>
>>> diff --git a/arch/ia64/kvm/kvm-ia64.c b/arch/ia64/kvm/kvm-ia64.c
>>> index ec6b9acb6bea..dbe46f43884d 100644
>>> --- a/arch/ia64/kvm/kvm-ia64.c
>>> +++ b/arch/ia64/kvm/kvm-ia64.c
>>> @@ -1563,7 +1563,7 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
>>>
>>> for (i = 0; i < npages; i++) {
>>> pfn = gfn_to_pfn(kvm, base_gfn + i);
>>> - if (!kvm_is_mmio_pfn(pfn)) {
>>> + if (!kvm_is_reserved_pfn(pfn)) {
>>> kvm_set_pmt_entry(kvm, base_gfn + i,
>>> pfn << PAGE_SHIFT,
>>> _PAGE_AR_RWX | _PAGE_MA_WB);
>>> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
>>> index ac1c4de3a484..978f402006ee 100644
>>> --- a/arch/x86/kvm/mmu.c
>>> +++ b/arch/x86/kvm/mmu.c
>>> @@ -630,7 +630,7 @@ static int mmu_spte_clear_track_bits(u64 *sptep)
>>> * kvm mmu, before reclaiming the page, we should
>>> * unmap it from mmu first.
>>> */
>>> - WARN_ON(!kvm_is_mmio_pfn(pfn) && !page_count(pfn_to_page(pfn)));
>>> + WARN_ON(!kvm_is_reserved_pfn(pfn) && !page_count(pfn_to_page(pfn)));
>>>
>>> if (!shadow_accessed_mask || old_spte & shadow_accessed_mask)
>>> kvm_set_pfn_accessed(pfn);
>>> @@ -2461,7 +2461,7 @@ static int set_spte(struct kvm_vcpu *vcpu, u64 *sptep,
>>> spte |= PT_PAGE_SIZE_MASK;
>>> if (tdp_enabled)
>>> spte |= kvm_x86_ops->get_mt_mask(vcpu, gfn,
>>> - kvm_is_mmio_pfn(pfn));
>>> + kvm_is_reserved_pfn(pfn));
>>>
>>> if (host_writable)
>>> spte |= SPTE_HOST_WRITEABLE;
>>> @@ -2737,7 +2737,7 @@ static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
>>> * PT_PAGE_TABLE_LEVEL and there would be no adjustment done
>>> * here.
>>> */
>>> - if (!is_error_noslot_pfn(pfn) && !kvm_is_mmio_pfn(pfn) &&
>>> + if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
>>> level == PT_PAGE_TABLE_LEVEL &&
>>> PageTransCompound(pfn_to_page(pfn)) &&
>>> !has_wrprotected_page(vcpu->kvm, gfn, PT_DIRECTORY_LEVEL)) {
>>> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
>>> index ea53b04993f2..a6059bdf7b03 100644
>>> --- a/include/linux/kvm_host.h
>>> +++ b/include/linux/kvm_host.h
>>> @@ -703,7 +703,7 @@ void kvm_arch_sync_events(struct kvm *kvm);
>>> int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
>>> void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
>>>
>>> -bool kvm_is_mmio_pfn(pfn_t pfn);
>>> +bool kvm_is_reserved_pfn(pfn_t pfn);
>>>
>>> struct kvm_irq_ack_notifier {
>>> struct hlist_node link;
>>> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
>>> index 25ffac9e947d..3cee7b167052 100644
>>> --- a/virt/kvm/kvm_main.c
>>> +++ b/virt/kvm/kvm_main.c
>>> @@ -107,10 +107,10 @@ EXPORT_SYMBOL_GPL(kvm_rebooting);
>>>
>>> static bool largepages_enabled = true;
>>>
>>> -bool kvm_is_mmio_pfn(pfn_t pfn)
>>> +bool kvm_is_reserved_pfn(pfn_t pfn)
>>> {
>>> if (pfn_valid(pfn))
>>> - return !is_zero_pfn(pfn) && PageReserved(pfn_to_page(pfn));
>>> + return PageReserved(pfn_to_page(pfn));
>>
>> so we return true for !pfn_valid(pfn), is this still semantically
>> correct with the rename?
>>
>
> I guess it is still debatable, but is arguably more correct than
> 'kvm_is_mmio_pfn'
>
agreed
> I was reluctant to choose something like 'kvm_is_special_pfn' because
> 'special' is not very discriminating here, and at least 'reserved' has
> a very clear meaning wrt pages, and treating non-struct page backed
> pfn's as reserved implicitly is not counter-intuitive imo.
>
just wanted to make sure we thought everything through, and it sounds
like we did.
>>>
>>> return true;
>>> }
>>> @@ -1321,7 +1321,7 @@ static pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async,
>>> else if ((vma->vm_flags & VM_PFNMAP)) {
>>> pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) +
>>> vma->vm_pgoff;
>>> - BUG_ON(!kvm_is_mmio_pfn(pfn));
>>> + BUG_ON(!kvm_is_reserved_pfn(pfn));
>>> } else {
>>> if (async && vma_is_valid(vma, write_fault))
>>> *async = true;
>>> @@ -1427,7 +1427,7 @@ static struct page *kvm_pfn_to_page(pfn_t pfn)
>>> if (is_error_noslot_pfn(pfn))
>>> return KVM_ERR_PTR_BAD_PAGE;
>>>
>>> - if (kvm_is_mmio_pfn(pfn)) {
>>> + if (kvm_is_reserved_pfn(pfn)) {
>>> WARN_ON(1);
>>> return KVM_ERR_PTR_BAD_PAGE;
>>> }
>>> @@ -1456,7 +1456,7 @@ EXPORT_SYMBOL_GPL(kvm_release_page_clean);
>>>
>>> void kvm_release_pfn_clean(pfn_t pfn)
>>> {
>>> - if (!is_error_noslot_pfn(pfn) && !kvm_is_mmio_pfn(pfn))
>>> + if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn))
>>> put_page(pfn_to_page(pfn));
>>> }
>>> EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
>>> @@ -1477,7 +1477,7 @@ static void kvm_release_pfn_dirty(pfn_t pfn)
>>>
>>> void kvm_set_pfn_dirty(pfn_t pfn)
>>> {
>>> - if (!kvm_is_mmio_pfn(pfn)) {
>>> + if (!kvm_is_reserved_pfn(pfn)) {
>>> struct page *page = pfn_to_page(pfn);
>>> if (!PageReserved(page))
>>> SetPageDirty(page);
>>
>> this looks rather redundant now then? Or is it catering specifically to
>> the situation where !pfn_valid(pfn) ?
>>
>
> I hadn't spotted this myself, to be honest, but that second test was
> redundant to begin with, i.e., we never enter the block for a reserved
> page.
> I can remove it here, or in a followup patch, as you [plural] prefer.
>
I have no preference either way, just happened to spot it purely from
the diff context.
Thanks,
-Christoffer
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
