Re: [PATCH 1/3] kvm-s390: Implement the directed yield (diag 9c) hypervisor call for KVM

[Avi Kivity <avi@redhat.com>]

On 04/24/2012 10:24 AM, Christian Borntraeger wrote:
> From: Konstantin Weitz <WEITZKON@de.ibm.com>
>
> This patch implements the directed yield hypercall found on other
> System z hypervisors. It delegates execution time to the virtual cpu
> specified in the instruction's parameter.
>
> Useful to avoid long spinlock waits in the guest.
>
>
> diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h
> index 7343872..dd17537 100644
> --- a/arch/s390/include/asm/kvm_host.h
> +++ b/arch/s390/include/asm/kvm_host.h
> @@ -148,6 +148,7 @@ struct kvm_vcpu_stat {
>  	u32 instruction_sigp_restart;
>  	u32 diagnose_10;
>  	u32 diagnose_44;
> +	u32 diagnose_9c;
>  };
>  
>  struct kvm_s390_io_info {
> diff --git a/arch/s390/kvm/diag.c b/arch/s390/kvm/diag.c
> index a353f0e..8991009 100644
> --- a/arch/s390/kvm/diag.c
> +++ b/arch/s390/kvm/diag.c
> @@ -53,6 +53,48 @@ static int __diag_time_slice_end(struct kvm_vcpu *vcpu)
>  	return 0;
>  }
>  
> +static int __diag_time_slice_end_directed(struct kvm_vcpu *vcpu)
> +{
> +	struct kvm *kvm = vcpu->kvm;
> +	struct kvm_vcpu *tcpu;
> +	int tid;
> +	int i;
> +
> +	tid = vcpu->run->s.regs.gprs[(vcpu->arch.sie_block->ipa & 0xf0) >> 4];
> +	vcpu->stat.diagnose_9c++;
> +	VCPU_EVENT(vcpu, 5, "diag time slice end directed to %d", tid);
> +
> +	if (tid == vcpu->vcpu_id)
> +		return 0;
> +
> +	kvm_for_each_vcpu(i, tcpu, kvm) {
> +		if (tcpu->vcpu_id == tid) {

Can two vcpus match the same tid?

> +			struct task_struct *task = NULL;
> +			struct pid *pid;
> +			rcu_read_lock();
> +			pid = rcu_dereference(tcpu->pid);
> +			if (pid)
> +				task = get_pid_task(tcpu->pid, PIDTYPE_PID);
> +			rcu_read_unlock();
> +			if (!task)
> +				break;
> +			if (task->flags & PF_VCPU) {
> +				put_task_struct(task);
> +				break;
> +			}
> +			vcpu_put(vcpu);
> +			if (yield_to(task, 1)) {
> +				vcpu_load(vcpu);

vcpu_put()/_load()s are unneeded here, preempt notifiers will
automatically call kvm_arch_vcpu_put()/_load() if we schedule.

> +				put_task_struct(task);
> +				return 0;
> +			}
> +			vcpu_load(vcpu);
> +			put_task_struct(task);
> +		}
> +	}
> +	return 0;
> +}
> +
> diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
> index d30c835..fd98914 100644
> --- a/arch/s390/kvm/kvm-s390.c
> +++ b/arch/s390/kvm/kvm-s390.c
> @@ -74,6 +74,7 @@ struct kvm_stats_debugfs_item debugfs_entries[] = {
>  	{ "instruction_sigp_restart", VCPU_STAT(instruction_sigp_restart) },
>  	{ "diagnose_10", VCPU_STAT(diagnose_10) },
>  	{ "diagnose_44", VCPU_STAT(diagnose_44) },
> +	{ "diagnose_9c", VCPU_STAT(diagnose_9c) },
>  	{ NULL }
>  };
>  

We're switching to tracepoints instead of homebrewed statistics.  It's
in feature-removal-schedule.txt already.  It's okay to add them now, but
please prepare for their removal.

-- 
error compiling committee.c: too many arguments to function