Prev: cachefiles: use path_get instead of lone dget
Next: [PATCH] OCFS2: Allow huge (> 16 TiB) volumes to mount
From: Marcelo Tosatti on 6 Jul 2010 16:10
On Tue, Jul 06, 2010 at 06:51:06PM +0800, Xiao Guangrong wrote:
> Combine guest pte read between guest pte check in the fetch path and pte prefetch
>
> Signed-off-by: Xiao Guangrong <xiaoguangrong(a)cn.fujitsu.com>
> ---
> arch/x86/kvm/paging_tmpl.h | 69 ++++++++++++++++++++++++++-----------------
> 1 files changed, 42 insertions(+), 27 deletions(-)
>
> diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
> index e04c1a4..a1e6d91 100644
> --- a/arch/x86/kvm/paging_tmpl.h
> +++ b/arch/x86/kvm/paging_tmpl.h
> @@ -67,6 +67,7 @@ struct guest_walker {
> int level;
> gfn_t table_gfn[PT_MAX_FULL_LEVELS];
> pt_element_t ptes[PT_MAX_FULL_LEVELS];
> + pt_element_t prefetch_ptes[PTE_PREFETCH_NUM];
> gpa_t pte_gpa[PT_MAX_FULL_LEVELS];
> unsigned pt_access;
> unsigned pte_access;
> @@ -291,12 +292,12 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
> gpte_to_gfn(gpte), pfn, true, true);
> }
>
> -static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, u64 *sptep)
> +static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, struct guest_walker *gw,
> + u64 *sptep)
> {
> struct kvm_mmu_page *sp;
> - pt_element_t gptep[PTE_PREFETCH_NUM];
> - gpa_t first_pte_gpa;
> - int offset = 0, index, i, j, max;
> + pt_element_t *gptep;
> + int index, i, j, max;
>
> sp = page_header(__pa(sptep));
> index = sptep - sp->spt;
> @@ -311,15 +312,7 @@ static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, u64 *sptep)
> i = index & ~(PTE_PREFETCH_NUM - 1);
> max = index | (PTE_PREFETCH_NUM - 1);
>
> - if (PTTYPE == 32)
> - offset = sp->role.quadrant << PT64_LEVEL_BITS;
> -
> - first_pte_gpa = gfn_to_gpa(sp->gfn) +
> - (offset + i) * sizeof(pt_element_t);
> -
> - if (kvm_read_guest_atomic(vcpu->kvm, first_pte_gpa, gptep,
> - sizeof(gptep)) < 0)
> - return;
> + gptep = gw->prefetch_ptes;

Where do you reread the gpte in the prefetch path?

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Xiao Guangrong on 6 Jul 2010 21:30


Marcelo Tosatti wrote:

>> -
>> - first_pte_gpa = gfn_to_gpa(sp->gfn) +
>> - (offset + i) * sizeof(pt_element_t);
>> -
>> - if (kvm_read_guest_atomic(vcpu->kvm, first_pte_gpa, gptep,
>> - sizeof(gptep)) < 0)
>> - return;
>> + gptep = gw->prefetch_ptes;
>
> Where do you reread the gpte in the prefetch path?
>

Marcelo,

Thanks for your review.

Below codes read gptes in the prefetch path:

index = sptep - sp->spt;
i = index & ~(PTE_PREFETCH_NUM - 1);
max = index | (PTE_PREFETCH_NUM - 1);

if (PTTYPE == 32)
offset = sp->role.quadrant << PT64_LEVEL_BITS;

first_pte_gpa = gfn_to_gpa(sp->gfn) +
(offset + i) * sizeof(pt_element_t);

if (kvm_read_guest_atomic(vcpu->kvm, first_pte_gpa, gptep,
sizeof(gptep)) < 0)
return;

It reads the 16 aligned items around sptep's corresponding gpte and this gpte
is also in this area. :-)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Marcelo Tosatti on 7 Jul 2010 09:10
On Wed, Jul 07, 2010 at 09:23:15AM +0800, Xiao Guangrong wrote:
>
>
> Marcelo Tosatti wrote:
>
> >> -
> >> - first_pte_gpa = gfn_to_gpa(sp->gfn) +
> >> - (offset + i) * sizeof(pt_element_t);
> >> -
> >> - if (kvm_read_guest_atomic(vcpu->kvm, first_pte_gpa, gptep,
> >> - sizeof(gptep)) < 0)
> >> - return;
> >> + gptep = gw->prefetch_ptes;
> >
> > Where do you reread the gpte in the prefetch path?
> >
>
> Marcelo,
>
> Thanks for your review.
>
> Below codes read gptes in the prefetch path:
>
> index = sptep - sp->spt;
> i = index & ~(PTE_PREFETCH_NUM - 1);
> max = index | (PTE_PREFETCH_NUM - 1);
>
> if (PTTYPE == 32)
> offset = sp->role.quadrant << PT64_LEVEL_BITS;
>
> first_pte_gpa = gfn_to_gpa(sp->gfn) +
> (offset + i) * sizeof(pt_element_t);
>
> if (kvm_read_guest_atomic(vcpu->kvm, first_pte_gpa, gptep,
> sizeof(gptep)) < 0)
> return;
>
> It reads the 16 aligned items around sptep's corresponding gpte and this gpte
> is also in this area. :-)

But you removed that in patch 8?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Xiao Guangrong on 7 Jul 2010 09:20


Marcelo Tosatti wrote:

>>
>> Below codes read gptes in the prefetch path:
>>
>> index = sptep - sp->spt;
>> i = index & ~(PTE_PREFETCH_NUM - 1);
>> max = index | (PTE_PREFETCH_NUM - 1);
>>
>> if (PTTYPE == 32)
>> offset = sp->role.quadrant << PT64_LEVEL_BITS;
>>
>> first_pte_gpa = gfn_to_gpa(sp->gfn) +
>> (offset + i) * sizeof(pt_element_t);
>>
>> if (kvm_read_guest_atomic(vcpu->kvm, first_pte_gpa, gptep,
>> sizeof(gptep)) < 0)
>> return;
>>
>> It reads the 16 aligned items around sptep's corresponding gpte and this gpte
>> is also in this area. :-)
>
> But you removed that in patch 8?
>

Oh, i just want it good for review, you mean it's better let patch 7 and 8 in one patch?

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Marcelo Tosatti on 7 Jul 2010 10:00
On Wed, Jul 07, 2010 at 09:11:56PM +0800, Xiao Guangrong wrote:
>
>
> Marcelo Tosatti wrote:
>
> >>
> >> Below codes read gptes in the prefetch path:
> >>
> >> index = sptep - sp->spt;
> >> i = index & ~(PTE_PREFETCH_NUM - 1);
> >> max = index | (PTE_PREFETCH_NUM - 1);
> >>
> >> if (PTTYPE == 32)
> >> offset = sp->role.quadrant << PT64_LEVEL_BITS;
> >>
> >> first_pte_gpa = gfn_to_gpa(sp->gfn) +
> >> (offset + i) * sizeof(pt_element_t);
> >>
> >> if (kvm_read_guest_atomic(vcpu->kvm, first_pte_gpa, gptep,
> >> sizeof(gptep)) < 0)
> >> return;
> >>
> >> It reads the 16 aligned items around sptep's corresponding gpte and this gpte
> >> is also in this area. :-)
> >
> > But you removed that in patch 8?
> >
>
> Oh, i just want it good for review, you mean it's better let patch 7 and 8 in one patch?

I mean patch 7 is wrong because it removes rereading of gpte, with
mmu_lock held, in the prefetch patch.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 |  Next  |  Last
Pages: 1 2
Prev: cachefiles: use path_get instead of lone dget
Next: [PATCH] OCFS2: Allow huge (> 16 TiB) volumes to mount