Re: [PATCH v2 08/28] ARM: KVM: Add TLB invalidation code

[Marc Zyngier <marc.zyngier@arm.com>]

On 09/02/16 18:42, Christoffer Dall wrote:
> On Thu, Feb 04, 2016 at 11:00:25AM +0000, Marc Zyngier wrote:
>> Convert the TLB invalidation code to C, hooking it into the
>> build system whilst we're at it.
>>
>> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
>> ---
>>  arch/arm/kvm/Makefile     |  1 +
>>  arch/arm/kvm/hyp/Makefile |  5 ++++
>>  arch/arm/kvm/hyp/hyp.h    |  5 ++++
>>  arch/arm/kvm/hyp/tlb.c    | 71 +++++++++++++++++++++++++++++++++++++++++++++++
>>  4 files changed, 82 insertions(+)
>>  create mode 100644 arch/arm/kvm/hyp/Makefile
>>  create mode 100644 arch/arm/kvm/hyp/tlb.c
>>
>> diff --git a/arch/arm/kvm/Makefile b/arch/arm/kvm/Makefile
>> index c5eef02c..eb1bf43 100644
>> --- a/arch/arm/kvm/Makefile
>> +++ b/arch/arm/kvm/Makefile
>> @@ -17,6 +17,7 @@ AFLAGS_interrupts.o := -Wa,-march=armv7-a$(plus_virt)
>>  KVM := ../../../virt/kvm
>>  kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o $(KVM)/vfio.o
>>  
>> +obj-$(CONFIG_KVM_ARM_HOST) += hyp/
>>  obj-y += kvm-arm.o init.o interrupts.o
>>  obj-y += arm.o handle_exit.o guest.o mmu.o emulate.o reset.o
>>  obj-y += coproc.o coproc_a15.o coproc_a7.o mmio.o psci.o perf.o
>> diff --git a/arch/arm/kvm/hyp/Makefile b/arch/arm/kvm/hyp/Makefile
>> new file mode 100644
>> index 0000000..36c760d
>> --- /dev/null
>> +++ b/arch/arm/kvm/hyp/Makefile
>> @@ -0,0 +1,5 @@
>> +#
>> +# Makefile for Kernel-based Virtual Machine module, HYP part
>> +#
>> +
>> +obj-$(CONFIG_KVM_ARM_HOST) += tlb.o
>> diff --git a/arch/arm/kvm/hyp/hyp.h b/arch/arm/kvm/hyp/hyp.h
>> index 727089f..5808bbd 100644
>> --- a/arch/arm/kvm/hyp/hyp.h
>> +++ b/arch/arm/kvm/hyp/hyp.h
>> @@ -42,4 +42,9 @@
>>  })
>>  #define read_sysreg(...)		__read_sysreg(__VA_ARGS__)
>>  
>> +#define VTTBR		__ACCESS_CP15_64(6, c2)
>> +#define ICIALLUIS	__ACCESS_CP15(c7, 0, c1, 0)
>> +#define TLBIALLIS	__ACCESS_CP15(c8, 0, c3, 0)
>> +#define TLBIALLNSNHIS	__ACCESS_CP15(c8, 4, c3, 4)
>> +
>>  #endif /* __ARM_KVM_HYP_H__ */
>> diff --git a/arch/arm/kvm/hyp/tlb.c b/arch/arm/kvm/hyp/tlb.c
>> new file mode 100644
>> index 0000000..993fe89
>> --- /dev/null
>> +++ b/arch/arm/kvm/hyp/tlb.c
>> @@ -0,0 +1,71 @@
>> +/*
>> + * Original code:
>> + * Copyright (C) 2012 - Virtual Open Systems and Columbia University
>> + * Author: Christoffer Dall <c.dall@virtualopensystems.com>
>> + *
>> + * Mostly rewritten in C by Marc Zyngier <marc.zyngier@arm.com>
>> + *
>> + * This program is free software; you can redistribute it and/or modify
>> + * it under the terms of the GNU General Public License version 2 as
>> + * published by the Free Software Foundation.
>> + *
>> + * This program is distributed in the hope that it will be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> + * GNU General Public License for more details.
>> + *
>> + * You should have received a copy of the GNU General Public License
>> + * along with this program.  If not, see <http://www.gnu.org/licenses/>.
>> + */
>> +
>> +#include "hyp.h"
>> +
>> +/**
>> + * Flush per-VMID TLBs
>> + *
>> + * __kvm_tlb_flush_vmid(struct kvm *kvm);
>> + *
>> + * We rely on the hardware to broadcast the TLB invalidation to all CPUs
>> + * inside the inner-shareable domain (which is the case for all v7
>> + * implementations).  If we come across a non-IS SMP implementation, we'll
>> + * have to use an IPI based mechanism. Until then, we stick to the simple
>> + * hardware assisted version.
>> + *
>> + * As v7 does not support flushing per IPA, just nuke the whole TLB
>> + * instead, ignoring the ipa value.
>> + */
>> +static void __hyp_text __tlb_flush_vmid(struct kvm *kvm)
>> +{
>> +	dsb(ishst);
>> +
>> +	/* Switch to requested VMID */
>> +	kvm = kern_hyp_va(kvm);
>> +	write_sysreg(kvm->arch.vttbr, VTTBR);
>> +	isb();
>> +
>> +	write_sysreg(0, TLBIALLIS);
>> +	dsb(ish);
>> +	isb();
>> +
>> +	write_sysreg(0, VTTBR);
>> +}
>> +
>> +__alias(__tlb_flush_vmid) void __weak __kvm_tlb_flush_vmid(struct kvm *kvm);
>> +
>> +static void __hyp_text __tlb_flush_vmid_ipa(struct kvm *kvm, phys_addr_t ipa)
>> +{
>> +	__tlb_flush_vmid(kvm);
>> +}
>> +
>> +__alias(__tlb_flush_vmid_ipa) void __weak __kvm_tlb_flush_vmid_ipa(struct kvm *kvm,
>> +							    phys_addr_t ipa);
>> +
>> +static void __hyp_text __tlb_flush_vm_context(void)
>> +{
>> +	dsb(ishst);
> 
> do we need this initial dsb?

I'm a copy-paste muppet. I'll drop that.

>> +	write_sysreg(0, TLBIALLNSNHIS);
>> +	write_sysreg(0, ICIALLUIS);
>> +	dsb(ish);
> 
> we used to have an isb here, but we got rid of this because it's always
> followed by eret?

Indeed. We were super extra cautious in the old code, and eret does the
right thing.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...