Inspired by what had been done by X86 kvm and kvm common codes, it would
be better to make arm64 kvm consistent with them.
The memory allocations of VM scope should be charged into VM process.
Hence change the allocation flag from GFP_KERNEL to GFP_KERNEL_ACCOUNT.
There remain some GFP_KERNEL unchanged since they are global allocations,
instead of VM scope.
Changes:
v2: address a few comments from Marc Zyngier, including redundant
GFP_KERNEL, incorrect memcg accounting and variable type
v1: https://lkml.org/lkml/2021/9/2/649
Jia He (2):
KVM: arm64: vgic: Add memcg accounting to vgic allocations
KVM: arm64: Add memcg accounting to KVM allocations
arch/arm64/kvm/arm.c | 6 ++++--
arch/arm64/kvm/mmu.c | 2 +-
arch/arm64/kvm/pmu-emul.c | 2 +-
arch/arm64/kvm/reset.c | 2 +-
arch/arm64/kvm/vgic/vgic-init.c | 2 +-
arch/arm64/kvm/vgic/vgic-irqfd.c | 2 +-
arch/arm64/kvm/vgic/vgic-its.c | 14 +++++++-------
arch/arm64/kvm/vgic/vgic-mmio-v3.c | 2 +-
arch/arm64/kvm/vgic/vgic-v4.c | 2 +-
9 files changed, 18 insertions(+), 16 deletions(-)
--
2.17.1
Inspired by commit 254272ce6505 ("kvm: x86: Add memcg accounting to KVM
allocations"), it would be better to make arm64 KVM consistent with
common kvm codes.
The memory allocations of VM scope should be charged into VM process
cgroup, hence change GFP_KERNEL to GFP_KERNEL_ACCOUNT.
There remain a few cases since these allocations are global, not in VM
scope.
Signed-off-by: Jia He <[email protected]>
---
arch/arm64/kvm/arm.c | 6 ++++--
arch/arm64/kvm/mmu.c | 2 +-
arch/arm64/kvm/pmu-emul.c | 2 +-
arch/arm64/kvm/reset.c | 2 +-
4 files changed, 7 insertions(+), 5 deletions(-)
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 0ca72f5cda41..7a7c478b8e6b 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -293,10 +293,12 @@ long kvm_arch_dev_ioctl(struct file *filp,
struct kvm *kvm_arch_alloc_vm(void)
{
+ size_t sz = sizeof(struct kvm);
+
if (!has_vhe())
- return kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ return kzalloc(sz, GFP_KERNEL_ACCOUNT);
- return vzalloc(sizeof(struct kvm));
+ return __vmalloc(sz, GFP_KERNEL_ACCOUNT | __GFP_HIGHMEM | __GFP_ZERO);
}
void kvm_arch_free_vm(struct kvm *kvm)
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 0625bf2353c2..801845ebd9d5 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -465,7 +465,7 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu)
return -EINVAL;
}
- pgt = kzalloc(sizeof(*pgt), GFP_KERNEL);
+ pgt = kzalloc(sizeof(*pgt), GFP_KERNEL_ACCOUNT);
if (!pgt)
return -ENOMEM;
diff --git a/arch/arm64/kvm/pmu-emul.c b/arch/arm64/kvm/pmu-emul.c
index f33825c995cb..05d42f6b89e4 100644
--- a/arch/arm64/kvm/pmu-emul.c
+++ b/arch/arm64/kvm/pmu-emul.c
@@ -971,7 +971,7 @@ int kvm_arm_pmu_v3_set_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)
mutex_lock(&vcpu->kvm->lock);
if (!vcpu->kvm->arch.pmu_filter) {
- vcpu->kvm->arch.pmu_filter = bitmap_alloc(nr_events, GFP_KERNEL);
+ vcpu->kvm->arch.pmu_filter = bitmap_alloc(nr_events, GFP_KERNEL_ACCOUNT);
if (!vcpu->kvm->arch.pmu_filter) {
mutex_unlock(&vcpu->kvm->lock);
return -ENOMEM;
diff --git a/arch/arm64/kvm/reset.c b/arch/arm64/kvm/reset.c
index cba7872d69a8..608c1baaaa63 100644
--- a/arch/arm64/kvm/reset.c
+++ b/arch/arm64/kvm/reset.c
@@ -106,7 +106,7 @@ static int kvm_vcpu_finalize_sve(struct kvm_vcpu *vcpu)
vl > SVE_VL_ARCH_MAX))
return -EIO;
- buf = kzalloc(SVE_SIG_REGS_SIZE(sve_vq_from_vl(vl)), GFP_KERNEL);
+ buf = kzalloc(SVE_SIG_REGS_SIZE(sve_vq_from_vl(vl)), GFP_KERNEL_ACCOUNT);
if (!buf)
return -ENOMEM;
--
2.17.1
Hi Jia,
On Tue, Sep 7, 2021 at 7:33 AM Jia He <[email protected]> wrote:
>
> Inspired by commit 254272ce6505 ("kvm: x86: Add memcg accounting to KVM
> allocations"), it would be better to make arm64 KVM consistent with
> common kvm codes.
>
> The memory allocations of VM scope should be charged into VM process
> cgroup, hence change GFP_KERNEL to GFP_KERNEL_ACCOUNT.
>
> There remain a few cases since these allocations are global, not in VM
> scope.
I believe there are more memory allocations that could be switched to
GFP_KERNEL_ACCOUNT. For non-pKVM kernels, we probably should charge
all stage-2 paging structure allocations to the VM process. Your patch
appears to only change the allocation of the kvm_pgtable structure,
but not descendent paging structures.
--
Thanks,
Oliver
Hi Oliver, thanks for the review, please my comments below:
> -----Original Message-----
> From: Oliver Upton <[email protected]>
> Sent: Wednesday, September 8, 2021 6:29 AM
> To: Justin He <[email protected]>
> Cc: Marc Zyngier <[email protected]>; James Morse <[email protected]>;
> Alexandru Elisei <[email protected]>; Suzuki Poulose
> <[email protected]>; Xiaoming Ni <[email protected]>; Lorenzo
> Pieralisi <[email protected]>; Kees Cook <[email protected]>;
> Catalin Marinas <[email protected]>; Nick Desaulniers
> <[email protected]>; [email protected]; Liu Shixin
> <[email protected]>; Sami Tolvanen <[email protected]>; Will
> Deacon <[email protected]>; [email protected]; linux-arm-
> [email protected]
> Subject: Re: [PATCH v2 2/2] KVM: arm64: Add memcg accounting to KVM
> allocations
>
> Hi Jia,
>
> On Tue, Sep 7, 2021 at 7:33 AM Jia He <[email protected]> wrote:
> >
> > Inspired by commit 254272ce6505 ("kvm: x86: Add memcg accounting to KVM
> > allocations"), it would be better to make arm64 KVM consistent with
> > common kvm codes.
> >
> > The memory allocations of VM scope should be charged into VM process
> > cgroup, hence change GFP_KERNEL to GFP_KERNEL_ACCOUNT.
> >
> > There remain a few cases since these allocations are global, not in VM
> > scope.
>
> I believe there are more memory allocations that could be switched to
> GFP_KERNEL_ACCOUNT. For non-pKVM kernels, we probably should charge
> all stage-2 paging structure allocations to the VM process. Your patch
> appears to only change the allocation of the kvm_pgtable structure,
> but not descendent paging structures.
>
Do you mean kvm_hyp_zalloc_page() here?
Seems kvm_hyp_zalloc_page() is in both global and VM scopes.
I will replace GFP_KERNEL with XXX_ACCOUNT in next version if no one
objects that.
--
Cheers,
Justin (Jia He)
Justin,
On Thu, Sep 9, 2021 at 8:57 AM Justin He <[email protected]> wrote:
>
> Hi Oliver, thanks for the review, please my comments below:
>
> > -----Original Message-----
> > From: Oliver Upton <[email protected]>
> > Sent: Wednesday, September 8, 2021 6:29 AM
> > To: Justin He <[email protected]>
> > Cc: Marc Zyngier <[email protected]>; James Morse <[email protected]>;
> > Alexandru Elisei <[email protected]>; Suzuki Poulose
> > <[email protected]>; Xiaoming Ni <[email protected]>; Lorenzo
> > Pieralisi <[email protected]>; Kees Cook <[email protected]>;
> > Catalin Marinas <[email protected]>; Nick Desaulniers
> > <[email protected]>; [email protected]; Liu Shixin
> > <[email protected]>; Sami Tolvanen <[email protected]>; Will
> > Deacon <[email protected]>; [email protected]; linux-arm-
> > [email protected]
> > Subject: Re: [PATCH v2 2/2] KVM: arm64: Add memcg accounting to KVM
> > allocations
> >
> > Hi Jia,
> >
> > On Tue, Sep 7, 2021 at 7:33 AM Jia He <[email protected]> wrote:
> > >
> > > Inspired by commit 254272ce6505 ("kvm: x86: Add memcg accounting to KVM
> > > allocations"), it would be better to make arm64 KVM consistent with
> > > common kvm codes.
> > >
> > > The memory allocations of VM scope should be charged into VM process
> > > cgroup, hence change GFP_KERNEL to GFP_KERNEL_ACCOUNT.
> > >
> > > There remain a few cases since these allocations are global, not in VM
> > > scope.
> >
> > I believe there are more memory allocations that could be switched to
> > GFP_KERNEL_ACCOUNT. For non-pKVM kernels, we probably should charge
> > all stage-2 paging structure allocations to the VM process. Your patch
> > appears to only change the allocation of the kvm_pgtable structure,
> > but not descendent paging structures.
> >
> Do you mean kvm_hyp_zalloc_page() here?
> Seems kvm_hyp_zalloc_page() is in both global and VM scopes.
Doh! kvm_hyp_zalloc_page() is only for the hyp's page tables, hence I
believe your patch is correct as it stands. As such:
Reviewed-by: Oliver Upton <[email protected]>
Hi Marc
> -----Original Message-----
> From: Oliver Upton <[email protected]>
> Sent: Monday, September 13, 2021 8:21 PM
> To: Justin He <[email protected]>
> Cc: Marc Zyngier <[email protected]>; James Morse <[email protected]>;
> Alexandru Elisei <[email protected]>; Suzuki Poulose
> <[email protected]>; Xiaoming Ni <[email protected]>; Lorenzo
> Pieralisi <[email protected]>; Kees Cook <[email protected]>;
> Catalin Marinas <[email protected]>; Nick Desaulniers
> <[email protected]>; [email protected]; Liu Shixin
> <[email protected]>; Sami Tolvanen <[email protected]>; Will
> Deacon <[email protected]>; [email protected]; linux-arm-
> [email protected]
> Subject: Re: [PATCH v2 2/2] KVM: arm64: Add memcg accounting to KVM
> allocations
>
[...]
> > Do you mean kvm_hyp_zalloc_page() here?
> > Seems kvm_hyp_zalloc_page() is in both global and VM scopes.
>
> Doh! kvm_hyp_zalloc_page() is only for the hyp's page tables, hence I
> believe your patch is correct as it stands. As such:
>
> Reviewed-by: Oliver Upton <[email protected]>
Kindly ping for the whole series. ????
--
Cheers,
Justin (Jia He)
On Tue, 7 Sep 2021 20:31:10 +0800, Jia He wrote:
> Inspired by what had been done by X86 kvm and kvm common codes, it would
> be better to make arm64 kvm consistent with them.
>
> The memory allocations of VM scope should be charged into VM process.
> Hence change the allocation flag from GFP_KERNEL to GFP_KERNEL_ACCOUNT.
>
> There remain some GFP_KERNEL unchanged since they are global allocations,
> instead of VM scope.
>
> [...]
Applied to next, thanks!
[1/2] KVM: arm64: vgic: Add memcg accounting to vgic allocations
commit: 3ef231670b9e9001316a426e794b2c74b8f6b4f6
[2/2] KVM: arm64: Add memcg accounting to KVM allocations
commit: 115bae923ac8bb29ee635e0ed6b4d5a3eec9371e
Cheers,
M.
--
Without deviation from the norm, progress is not possible.