Re: [PATCH v4 22/46] KVM: PPC: Book3S HV P9: Stop handling hcalls in real-mode in the P9 path

[Fabiano Rosas <farosas@linux.ibm.com>]

Nicholas Piggin <npiggin@gmail.com> writes:

> In the interest of minimising the amount of code that is run in
> "real-mode", don't handle hcalls in real mode in the P9 path.
>
> POWER8 and earlier are much more expensive to exit from HV real mode
> and switch to host mode, because on those processors HV interrupts get
> to the hypervisor with the MMU off, and the other threads in the core
> need to be pulled out of the guest, and SLBs all need to be saved,
> ERATs invalidated, and host SLB reloaded before the MMU is re-enabled
> in host mode. Hash guests also require a lot of hcalls to run. The
> XICS interrupt controller requires hcalls to run.
>
> By contrast, POWER9 has independent thread switching, and in radix mode
> the hypervisor is already in a host virtual memory mode when the HV
> interrupt is taken. Radix + xive guests don't need hcalls to handle
> interrupts or manage translations.
>
> So it's much less important to handle hcalls in real mode in P9.
>
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---

<snip>

> diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c
> index fa7614c37e08..17739aaee3d8 100644
> --- a/arch/powerpc/kvm/book3s_hv.c
> +++ b/arch/powerpc/kvm/book3s_hv.c
> @@ -1142,12 +1142,13 @@ int kvmppc_pseries_do_hcall(struct kvm_vcpu *vcpu)
>  }
>
>  /*
> - * Handle H_CEDE in the nested virtualization case where we haven't
> - * called the real-mode hcall handlers in book3s_hv_rmhandlers.S.
> + * Handle H_CEDE in the P9 path where we don't call the real-mode hcall
> + * handlers in book3s_hv_rmhandlers.S.
> + *
>   * This has to be done early, not in kvmppc_pseries_do_hcall(), so
>   * that the cede logic in kvmppc_run_single_vcpu() works properly.
>   */
> -static void kvmppc_nested_cede(struct kvm_vcpu *vcpu)
> +static void kvmppc_cede(struct kvm_vcpu *vcpu)
>  {
>  	vcpu->arch.shregs.msr |= MSR_EE;
>  	vcpu->arch.ceded = 1;
> @@ -1403,9 +1404,15 @@ static int kvmppc_handle_exit_hv(struct kvm_vcpu *vcpu,
>  		/* hcall - punt to userspace */
>  		int i;
>
> -		/* hypercall with MSR_PR has already been handled in rmode,
> -		 * and never reaches here.
> -		 */
> +		if (unlikely(vcpu->arch.shregs.msr & MSR_PR)) {
> +			/*
> +			 * Guest userspace executed sc 1, reflect it back as a
> +			 * privileged program check interrupt.
> +			 */
> +			kvmppc_core_queue_program(vcpu, SRR1_PROGPRIV);
> +			r = RESUME_GUEST;
> +			break;
> +		}

This patch bypasses sc_1_fast_return so it breaks KVM-PR. L1 loops with
the following output:

[    9.503929][ T3443] Couldn't emulate instruction 0x4e800020 (op 19 xop 16)
[    9.503990][ T3443] kvmppc_exit_pr_progint: emulation at 48f4 failed (4e800020)
[    9.504080][ T3443] Couldn't emulate instruction 0x4e800020 (op 19 xop 16)
[    9.504170][ T3443] kvmppc_exit_pr_progint: emulation at 48f4 failed (4e800020)

0x4e800020 is a blr after a sc 1 in SLOF.

For KVM-PR we need to inject a 0xc00 at some point, either here or
before branching to no_try_real in book3s_hv_rmhandlers.S.

>
>  		run->papr_hcall.nr = kvmppc_get_gpr(vcpu, 3);
>  		for (i = 0; i < 9; ++i)
> @@ -3663,6 +3670,12 @@ static int kvmhv_load_hv_regs_and_go(struct kvm_vcpu *vcpu, u64 time_limit,
>  	return trap;
>  }
>
> +static inline bool hcall_is_xics(unsigned long req)
> +{
> +	return (req == H_EOI || req == H_CPPR || req == H_IPI ||
> +		req == H_IPOLL || req == H_XIRR || req == H_XIRR_X);
> +}
> +
>  /*
>   * Virtual-mode guest entry for POWER9 and later when the host and
>   * guest are both using the radix MMU.  The LPIDR has already been set.
> @@ -3774,15 +3787,36 @@ static int kvmhv_p9_guest_entry(struct kvm_vcpu *vcpu, u64 time_limit,
>  		/* H_CEDE has to be handled now, not later */
>  		if (trap == BOOK3S_INTERRUPT_SYSCALL && !vcpu->arch.nested &&
>  		    kvmppc_get_gpr(vcpu, 3) == H_CEDE) {
> -			kvmppc_nested_cede(vcpu);
> +			kvmppc_cede(vcpu);
>  			kvmppc_set_gpr(vcpu, 3, 0);
>  			trap = 0;
>  		}
>  	} else {
>  		kvmppc_xive_push_vcpu(vcpu);
>  		trap = kvmhv_load_hv_regs_and_go(vcpu, time_limit, lpcr);
> +		if (trap == BOOK3S_INTERRUPT_SYSCALL && !vcpu->arch.nested &&
> +		    !(vcpu->arch.shregs.msr & MSR_PR)) {
> +			unsigned long req = kvmppc_get_gpr(vcpu, 3);
> +
> +			/* H_CEDE has to be handled now, not later */
> +			if (req == H_CEDE) {
> +				kvmppc_cede(vcpu);
> +				kvmppc_xive_cede_vcpu(vcpu); /* may un-cede */
> +				kvmppc_set_gpr(vcpu, 3, 0);
> +				trap = 0;
> +
> +			/* XICS hcalls must be handled before xive is pulled */
> +			} else if (hcall_is_xics(req)) {
> +				int ret;
> +
> +				ret = kvmppc_xive_xics_hcall(vcpu, req);
> +				if (ret != H_TOO_HARD) {
> +					kvmppc_set_gpr(vcpu, 3, ret);
> +					trap = 0;
> +				}
> +			}
> +		}
>  		kvmppc_xive_pull_vcpu(vcpu);
> -
>  	}
>
>  	vcpu->arch.slb_max = 0;
> @@ -4442,8 +4476,11 @@ static int kvmppc_vcpu_run_hv(struct kvm_vcpu *vcpu)
>  		else
>  			r = kvmppc_run_vcpu(vcpu);
>
> -		if (run->exit_reason == KVM_EXIT_PAPR_HCALL &&
> -		    !(vcpu->arch.shregs.msr & MSR_PR)) {
> +		if (run->exit_reason == KVM_EXIT_PAPR_HCALL) {
> +			if (WARN_ON_ONCE(vcpu->arch.shregs.msr & MSR_PR)) {
> +				r = RESUME_GUEST;
> +				continue;
> +			}

Note that this hunk might need to be dropped.

>  			trace_kvm_hcall_enter(vcpu);
>  			r = kvmppc_pseries_do_hcall(vcpu);
>  			trace_kvm_hcall_exit(vcpu, r);
> diff --git a/arch/powerpc/kvm/book3s_hv_rmhandlers.S b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> index c11597f815e4..2d0d14ed1d92 100644
> --- a/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> +++ b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> @@ -1397,9 +1397,14 @@ END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
>  	mr	r4,r9
>  	bge	fast_guest_return
>  2:
> +	/* If we came in through the P9 short path, no real mode hcalls */
> +	lwz	r0, STACK_SLOT_SHORT_PATH(r1)
> +	cmpwi	r0, 0
> +	bne	no_try_real
>  	/* See if this is an hcall we can handle in real mode */
>  	cmpwi	r12,BOOK3S_INTERRUPT_SYSCALL
>  	beq	hcall_try_real_mode
> +no_try_real:
>
>  	/* Hypervisor doorbell - exit only if host IPI flag set */
>  	cmpwi	r12, BOOK3S_INTERRUPT_H_DOORBELL