Re: [PATCH v3 16/20] kvm: arm64: Switch to per VM IPA limit

From: Suzuki K Poulose
Date: Mon Jul 02 2018 - 09:53:18 EST

Hi Marc,

On 02/07/18 14:32, Marc Zyngier wrote:
On 29/06/18 12:15, Suzuki K Poulose wrote:
Now that we can manage the stage2 page table per VM, switch the
configuration details to per VM instance. We keep track of the
IPA bits, number of page table levels and the VTCR bits (which
depends on the IPA and the number of levels). While at it, remove
unused pgd_lock field from kvm_arch for arm64.

Cc: Marc Zyngier <marc.zyngier@xxxxxxx>
Cc: Christoffer Dall <cdall@xxxxxxxxxx>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@xxxxxxx>

diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index 328f472..9a15860 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -61,13 +61,23 @@ struct kvm_arch {
u64 vmid_gen;
u32 vmid;
- /* 1-level 2nd stage table and lock */
- spinlock_t pgd_lock;
+ /* stage-2 page table */
pgd_t *pgd;
/* VTTBR value associated with above pgd and vmid */
u64 vttbr;
+ /* Private bits of VTCR_EL2 for this VM */
+ u64 vtcr_private;

As I said in another email, this should become a full VTCR_EL2 copy.


+ /* Size of the PA size for this guest */
+ u8 phys_shift;
+ /*
+ * Number of levels in page table. We could always calculate
+ * it from phys_shift above. We cache it for faster switches
+ * in stage2 page table helpers.
+ */
+ u8 s2_levels;

And these two fields feel like they should be derived from the VTCR
itself, instead of being there on their own. Any chance you could look
into this?

Yes, the VTCR is computed from the above two values and we could compute
them back from the VTCR. I will give it a try.

diff --git a/arch/arm64/include/asm/stage2_pgtable.h b/arch/arm64/include/asm/stage2_pgtable.h
index ffc37cc..91d7936 100644
--- a/arch/arm64/include/asm/stage2_pgtable.h
+++ b/arch/arm64/include/asm/stage2_pgtable.h
@@ -65,7 +65,6 @@
#define __s2_pgd_ptrs(pa, lvls) (1 << ((pa) - pt_levels_pgdir_shift((lvls))))
#define __s2_pgd_size(pa, lvls) (__s2_pgd_ptrs((pa), (lvls)) * sizeof(pgd_t))
-#define kvm_stage2_levels(kvm) stage2_pt_levels(kvm_phys_shift(kvm))
#define stage2_pgdir_shift(kvm) \
#define stage2_pgdir_size(kvm) (_AC(1, UL) << stage2_pgdir_shift((kvm)))
diff --git a/virt/kvm/arm/mmu.c b/virt/kvm/arm/mmu.c
index a339e00..d7822e1 100644
--- a/virt/kvm/arm/mmu.c
+++ b/virt/kvm/arm/mmu.c
@@ -867,6 +867,10 @@ int kvm_alloc_stage2_pgd(struct kvm *kvm)
return -EINVAL;
+ /* Make sure we have the stage2 configured for this VM */
+ if (WARN_ON(!kvm_phys_shift(kvm)))

Can this be triggered from userspace?

No. As we initialise the phys shift before we get here. If type is left
blank (i.e, 0), we default to 40bits. So there should be something there.
The check is to make sure we have indeed past the configuration step.

+ return -EINVAL;
/* Allocate the HW PGD, making sure that each page gets its own refcount */
pgd = stage2_alloc_pgd(kvm);
if (!pgd)