Re: [PATCH] KVM: ARM: Remove pgtable page standard functions from stage-2 page tables

From: Suzuki K Poulose <suzuki.poulose@arm.com>
To: anshuman.khandual@arm.com, catalin.marinas@arm.com,
	will.deacon@arm.com, mark.rutland@arm.com, yuzhao@google.com
Cc: linux-mm@kvack.org, linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH] KVM: ARM: Remove pgtable page standard functions from stage-2 page tables
Date: Tue, 12 Mar 2019 11:43:02 +0000	[thread overview]
Message-ID: <3be0b7e0-2ef8-babb-88c9-d229e0fdd220@arm.com> (raw)
In-Reply-To: <2d00c35d-ae10-ba4a-9b34-939fcf2b2f49@arm.com>

On 12/03/2019 11:31, Anshuman Khandual wrote:
> 
> 
> On 03/12/2019 04:07 PM, Suzuki K Poulose wrote:
>> Hi Anshuman,
>>
>> On 12/03/2019 02:19, Anshuman Khandual wrote:
>>> ARM64 standard pgtable functions are going to use pgtable_page_[ctor|dtor]
>>> or pgtable_pmd_page_[ctor|dtor] constructs. At present KVM guest stage-2
>>> PUD|PMD|PTE level page tabe pages are allocated with __get_free_page()
>>> via mmu_memory_cache_alloc() but released with standard pud|pmd_free() or
>>> pte_free_kernel(). These will fail once they start calling into pgtable_
>>> [pmd]_page_dtor() for pages which never originally went through respective
>>> constructor functions. Hence convert all stage-2 page table page release
>>> functions to call buddy directly while freeing pages.
>>>
>>> Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
>>> ---
>>>    arch/arm/include/asm/stage2_pgtable.h   | 4 ++--
>>>    arch/arm64/include/asm/stage2_pgtable.h | 4 ++--
>>>    virt/kvm/arm/mmu.c                      | 2 +-
>>>    3 files changed, 5 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/arch/arm/include/asm/stage2_pgtable.h b/arch/arm/include/asm/stage2_pgtable.h
>>> index de2089501b8b..417a3be00718 100644
>>> --- a/arch/arm/include/asm/stage2_pgtable.h
>>> +++ b/arch/arm/include/asm/stage2_pgtable.h
>>> @@ -32,14 +32,14 @@
>>>    #define stage2_pgd_present(kvm, pgd)        pgd_present(pgd)
>>>    #define stage2_pgd_populate(kvm, pgd, pud)    pgd_populate(NULL, pgd, pud)
>>>    #define stage2_pud_offset(kvm, pgd, address)    pud_offset(pgd, address)
>>> -#define stage2_pud_free(kvm, pud)        pud_free(NULL, pud)
>>> +#define stage2_pud_free(kvm, pud)        free_page((unsigned long)pud)
>>
>> That must be a NOP, as we don't have pud on arm32 (we have 3 level table).
>> The pud_* helpers here all fallback to the generic no-pud helpers.
> Which is the following here for pud_free()
> 
> #define pud_free(mm, x)                         do { } while (0)
> 
> On arm64 its protected by kvm_stage2_has_pud() helper before calling into pud_free().
> In this case even though applicable pud_free() is NOP, it is still misleading. If we
> are sure about page table level will always remain three it can directly have a NOP
> (do/while) in there.
> 

Yes, it is fixed for arm32 and you could have it as do {} while (0), which is
what I meant by NOP. On arm64, we had varied number of levels depending on the
PAGE_SIZE and now due to the dynamic IPA, hence the check.

Cheers
Suzuki