All of lore.kernel.org
 help / color / mirror / Atom feed
From: Christoffer Dall <cdall@linaro.org>
To: Andrew Jones <drjones@redhat.com>
Cc: marc.zyngier@arm.com, kvmarm@lists.cs.columbia.edu
Subject: Re: [PATCH 4/5] KVM: arm/arm64: improve kvm_arch_vcpu_runnable
Date: Sat, 14 Oct 2017 21:13:25 +0200	[thread overview]
Message-ID: <20171014191325.GC1929@lvm> (raw)
In-Reply-To: <20170929113041.24371-5-drjones@redhat.com>

On Fri, Sep 29, 2017 at 01:30:40PM +0200, Andrew Jones wrote:
> Conceptually, kvm_arch_vcpu_runnable() should be "not waiting,
> or waiting for interrupts and interrupts are pending",
> 
>   !waiting-uninterruptable &&
>   (!waiting-for-interrupts || interrupts-pending)
> 
> but the implementation was only
> 
>   !waiting-uninterruptable && interrupts-pending
> 
> Thanks to the context of the two callers there wasn't an issue,
> however, to future-proof the function, this patch implements the
> conceptual condition by applying mp_state to track waiting-
> uninterruptable vs. waiting-for-interrupts.
> 
> Signed-off-by: Andrew Jones <drjones@redhat.com>
> ---
>  Documentation/virtual/kvm/api.txt | 10 ++++++----
>  virt/kvm/arm/arm.c                | 13 +++++++++----
>  2 files changed, 15 insertions(+), 8 deletions(-)
> 
> diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
> index e63a35fafef0..7c0bb1ae10a2 100644
> --- a/Documentation/virtual/kvm/api.txt
> +++ b/Documentation/virtual/kvm/api.txt
> @@ -1071,7 +1071,7 @@ Possible values are:
>   - KVM_MP_STATE_INIT_RECEIVED:   the vcpu has received an INIT signal, and is
>                                   now ready for a SIPI [x86]
>   - KVM_MP_STATE_HALTED:          the vcpu has executed a HLT instruction and
> -                                 is waiting for an interrupt [x86]
> +                                 is waiting for an interrupt [x86,arm/arm64]
>   - KVM_MP_STATE_SIPI_RECEIVED:   the vcpu has just received a SIPI (vector
>                                   accessible via KVM_GET_VCPU_EVENTS) [x86]
>   - KVM_MP_STATE_STOPPED:         the vcpu is stopped [s390,arm/arm64]
> @@ -1087,8 +1087,9 @@ these architectures.
>  
>  For arm/arm64:
>  
> -The only states that are valid are KVM_MP_STATE_STOPPED and
> -KVM_MP_STATE_RUNNABLE which reflect if the vcpu is paused or not.
> +The only states that are valid are KVM_MP_STATE_STOPPED, KVM_MP_STATE_HALTED,
> +and KVM_MP_STATE_RUNNABLE which reflect if the vcpu is powered-off, waiting
> +for interrupts, or powered-on and not waiting for interrupts.
>  

Is it valid to introduce another value after we've publicly declared
that it's not valid?  What if userspace has an assert(state ==
KVM_MP_STATE_STOPPED || state == KVM_MP_STATE_RUNNABLE) ?

At least we should make that is not the case.

>  4.39 KVM_SET_MP_STATE
>  
> @@ -1108,7 +1109,8 @@ these architectures.
>  For arm/arm64:
>  
>  The only states that are valid are KVM_MP_STATE_STOPPED and
> -KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be paused or not.
> +KVM_MP_STATE_RUNNABLE which reflect if the vcpu should be
> +powered-off or not.
>  

It looks pretty dodgy to me that we can return an MP state which we
cannot set again.  Feels like this could break migration.

>  4.40 KVM_SET_IDENTITY_MAP_ADDR
>  
> diff --git a/virt/kvm/arm/arm.c b/virt/kvm/arm/arm.c
> index 220a3dbda76c..5bc9b0d2fd0f 100644
> --- a/virt/kvm/arm/arm.c
> +++ b/virt/kvm/arm/arm.c
> @@ -414,13 +414,16 @@ static bool vcpu_should_sleep(struct kvm_vcpu *vcpu)
>   * kvm_arch_vcpu_runnable - determine if the vcpu can be scheduled
>   * @vcpu:	The VCPU pointer
>   *
> - * If the guest CPU is not waiting for interrupts or an interrupt line is
> - * asserted, the CPU is by definition runnable.
> + * If the VCPU is not waiting at all (including sleeping, which is waiting
> + * uninterruptably), or it's waiting for interrupts but interrupts are
> + * pending, then it is by definition runnable.
>   */
>  int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
>  {
> -	return (!!vcpu->arch.irq_lines || kvm_vgic_vcpu_pending_irq(vcpu))
> -		&& !vcpu_should_sleep(vcpu);
> +	return !vcpu_should_sleep(vcpu) &&
> +	       (vcpu->arch.mp_state != KVM_MP_STATE_HALTED ||
> +		(!!vcpu->arch.irq_lines ||
> +		 kvm_vgic_vcpu_pending_irq(vcpu)));

This is hard to read.  How about:

	bool irq_pending = !!vcpu->arch.irq_lines ||
			   kvm_vgic_vcpu_pending_irq(vcpu);
	
	if (vcpu_should_sleep(vcpu)
		return false;
	else if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED && !irq_pending)
		return false;
	return true;

>  }
>  
>  bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu)
> @@ -582,7 +585,9 @@ void kvm_arm_emulate_wfe(struct kvm_vcpu *vcpu)
>  void kvm_arm_emulate_wfi(struct kvm_vcpu *vcpu)
>  {
>  	vcpu->stat.wfi_exit_stat++;
> +	vcpu->arch.mp_state = KVM_MP_STATE_HALTED;
>  	kvm_vcpu_block(vcpu);
> +	vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
>  	kvm_clear_request(KVM_REQ_UNHALT, vcpu);
>  }
>  
> -- 
> 2.13.5
> 

Thanks,
-Christoffer

  parent reply	other threads:[~2017-10-14 19:12 UTC|newest]

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-09-29 11:30 [PATCH 0/5] kvm_arch_vcpu_runnable related improvements Andrew Jones
2017-09-29 11:30 ` [PATCH 1/5] KVM: arm/arm64: tidy 'should sleep' conditions Andrew Jones
2017-10-05  8:13   ` Marc Zyngier
2017-09-29 11:30 ` [PATCH 2/5] KVM: arm/arm64: replace power_off with mp_state=STOPPED Andrew Jones
2017-10-05  8:32   ` Marc Zyngier
2017-10-10 13:26     ` Andrew Jones
2017-10-14 19:12   ` Christoffer Dall
2017-10-18 12:04     ` Andrew Jones
2017-09-29 11:30 ` [PATCH 3/5] KVM: arm/arm64: factor out common wfe/wfi emulation code Andrew Jones
2017-10-05  8:36   ` Marc Zyngier
2017-10-14 19:13   ` Christoffer Dall
2017-10-18 12:06     ` Andrew Jones
2017-09-29 11:30 ` [PATCH 4/5] KVM: arm/arm64: improve kvm_arch_vcpu_runnable Andrew Jones
2017-10-05  9:19   ` Marc Zyngier
2017-10-14 19:13   ` Christoffer Dall [this message]
2017-10-18 12:09     ` Andrew Jones
2017-09-29 11:30 ` [PATCH 5/5] KVM: arm/arm64: kvm_arch_vcpu_runnable: don't miss injected irqs Andrew Jones
2017-10-05  9:37   ` Marc Zyngier
2017-10-10 13:28     ` Andrew Jones
2017-10-14 19:13   ` Christoffer Dall
2017-10-18 12:13     ` Andrew Jones
2017-10-18 13:18       ` Christoffer Dall
2017-10-18 13:55         ` Andrew Jones
2017-10-18 14:14           ` Christoffer Dall
2017-10-02  8:31 ` [PATCH 6/5] KVM: arm/arm64: make kvm_vgic_vcpu_pending_irq static Andrew Jones
2017-10-05  9:37   ` Marc Zyngier

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20171014191325.GC1929@lvm \
    --to=cdall@linaro.org \
    --cc=drjones@redhat.com \
    --cc=kvmarm@lists.cs.columbia.edu \
    --cc=marc.zyngier@arm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.