Re: [PATCH] LoongArch: KVM: Delay secondary mmu tlb flush before guest entry

[Sean Christopherson]

On Sat, Jun 15, 2024, Bibo Mao wrote:
> diff --git a/arch/loongarch/include/asm/kvm_host.h b/arch/loongarch/include/asm/kvm_host.h
> index c87b6ea0ec47..98078e01dd55 100644
> --- a/arch/loongarch/include/asm/kvm_host.h
> +++ b/arch/loongarch/include/asm/kvm_host.h
> @@ -30,6 +30,7 @@
>  #define KVM_PRIVATE_MEM_SLOTS		0
>  
>  #define KVM_HALT_POLL_NS_DEFAULT	500000
> +#define KVM_REQ_TLB_FLUSH_GPA		KVM_ARCH_REQ(0)
>  
>  #define KVM_GUESTDBG_SW_BP_MASK		\
>  	(KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)
> @@ -190,6 +191,7 @@ struct kvm_vcpu_arch {
>  
>  	/* vcpu's vpid */
>  	u64 vpid;
> +	unsigned long flush_gpa;

Side topic, GPAs should really use "gpa_t" instead of "unsigned long", otherwise
32-bit kernels running on CPUs with 64-bit physical addresses will fail miserably
(which may or may not be a problem in practice for LoongArch).

> diff --git a/arch/loongarch/kvm/tlb.c b/arch/loongarch/kvm/tlb.c
> index 02535df6b51f..55f7f3621e38 100644
> --- a/arch/loongarch/kvm/tlb.c
> +++ b/arch/loongarch/kvm/tlb.c
> @@ -21,12 +21,9 @@ void kvm_flush_tlb_all(void)
>  	local_irq_restore(flags);
>  }
>  
> +/* Called with irq disabled */

Rather than add a comment, add:

	lockdep_assert_irqs_disabled()

in the function.

>  void kvm_flush_tlb_gpa(struct kvm_vcpu *vcpu, unsigned long gpa)
>  {
> -	unsigned long flags;
> -
> -	local_irq_save(flags);
>  	gpa &= (PAGE_MASK << 1);
>  	invtlb(INVTLB_GID_ADDR, read_csr_gstat() & CSR_GSTAT_GID, gpa);
> -	local_irq_restore(flags);
>  }
> diff --git a/arch/loongarch/kvm/vcpu.c b/arch/loongarch/kvm/vcpu.c
> index 9e8030d45129..ae9ae88c11db 100644
> --- a/arch/loongarch/kvm/vcpu.c
> +++ b/arch/loongarch/kvm/vcpu.c
> @@ -51,6 +51,16 @@ static int kvm_check_requests(struct kvm_vcpu *vcpu)
>  	return RESUME_GUEST;
>  }
>  
> +/* Check pending request with irq disabled */

Same thing here.