Re: [patch 2/3] KVM: x86: KVM_HC_RT_PRIO hypercall (host-side)

[Konrad Rzeszutek Wilk]

On Thu, Sep 21, 2017 at 08:38:37AM -0300, Marcelo Tosatti wrote:
> When executing guest vcpu-0 with FIFO:1 priority, which is necessary to 
> deal with the following situation:
> 
> VCPU-0 (housekeeping VCPU)		VCPU-1 (realtime VCPU)
> 
> raw_spin_lock(A)
> interrupted, schedule task T-1		raw_spin_lock(A) (spin)
> 
> raw_spin_unlock(A)
> 
> Certain operations must interrupt guest vcpu-0 (see trace below).
> 
> To fix this issue, only change guest vcpu-0 to FIFO priority
> on spinlock critical sections (see patch).
> 
> Hang trace
> ==========
> 
> Without FIFO priority:
> 
> qemu-kvm-6705  [002] ....1.. 767785.648964: kvm_exit: reason IO_INSTRUCTION rip 0xe8fe info 1f00039 0
> qemu-kvm-6705  [002] ....1.. 767785.648965: kvm_exit: reason IO_INSTRUCTION rip 0xe911 info 3f60008 0
> qemu-kvm-6705  [002] ....1.. 767785.648968: kvm_exit: reason IO_INSTRUCTION rip 0x8984 info 608000b 0
> qemu-kvm-6705  [002] ....1.. 767785.648971: kvm_exit: reason IO_INSTRUCTION rip 0xb313 info 1f70008 0
> qemu-kvm-6705  [002] ....1.. 767785.648974: kvm_exit: reason IO_INSTRUCTION rip 0xb514 info 3f60000 0
> qemu-kvm-6705  [002] ....1.. 767785.648977: kvm_exit: reason PENDING_INTERRUPT rip 0x8052 info 0 0
> qemu-kvm-6705  [002] ....1.. 767785.648980: kvm_exit: reason IO_INSTRUCTION rip 0xeee6 info 200040 0
> qemu-kvm-6705  [002] ....1.. 767785.648999: kvm_exit: reason EPT_MISCONFIG rip 0x2120 info 0 0
> 
> With FIFO priority:
> 
> qemu-kvm-7636  [002] ....1.. 768218.205065: kvm_exit: reason IO_INSTRUCTION rip 0xb313 info 1f70008 0
> qemu-kvm-7636  [002] ....1.. 768218.205068: kvm_exit: reason IO_INSTRUCTION rip 0x8984 info 608000b 0
> qemu-kvm-7636  [002] ....1.. 768218.205071: kvm_exit: reason IO_INSTRUCTION rip 0xb313 info 1f70008 0
> qemu-kvm-7636  [002] ....1.. 768218.205074: kvm_exit: reason IO_INSTRUCTION rip 0x8984 info 608000b 0
> qemu-kvm-7636  [002] ....1.. 768218.205077: kvm_exit: reason IO_INSTRUCTION rip 0xb313 info 1f70008 0
> ..
> 
> Performance numbers (kernel compilation with make -j2)
> ======================================================
> 
> With hypercall: 4:40.  (make -j2)
> Without hypercall: 3:38.  (make -j2)
> 
> Note for NFV workloads spinlock performance is not relevant
> since DPDK should not enter the kernel (and housekeeping vcpu
> performance is far from a key factor).
> 
> Signed-off-by: Marcelo Tosatti <mtosatti@xxxxxxxxxx>
> 
> ---
>  Documentation/virtual/kvm/hypercalls.txt |   22 +++++++++++++++
>  arch/x86/kvm/x86.c                       |   43 +++++++++++++++++++++++++++++++
>  include/uapi/linux/kvm_para.h            |    2 +
>  3 files changed, 67 insertions(+)
> 
> Index: kvm.fifopriohc-submit/Documentation/virtual/kvm/hypercalls.txt
> ===================================================================
> --- kvm.fifopriohc-submit.orig/Documentation/virtual/kvm/hypercalls.txt
> +++ kvm.fifopriohc-submit/Documentation/virtual/kvm/hypercalls.txt
> @@ -121,3 +121,25 @@ compute the CLOCK_REALTIME for its clock
>  
>  Returns KVM_EOPNOTSUPP if the host does not use TSC clocksource,
>  or if clock type is different than KVM_CLOCK_PAIRING_WALLCLOCK.
> +
> +6. KVM_HC_RT_PRIO
> +------------------------
> +Architecture: x86
> +Status: active
> +Purpose: Hypercall used to change qemu vcpu process -RT priority.

So the guest can change the scheduling decisions at the host level?
And the host HAS to follow it? There is no policy override for the
host to say - nah, not going to do it?

Also wouldn't the guest want to always be at SCHED_FIFO? [I am thinking
of a guest admin who wants all the CPU resources he can get]


> +
> +Usage: Having a pCPU share a FIFO:1 vcpu and a QEMU emulator thread
> +can be problematic: especially if the vcpu busy-spins on memory waiting
> +for the QEMU emulator thread to write to, which leads to a hang

.. Is the QEMU emulator writing to the spinlock memory?

> +(because the FIFO:1 vcpu is never scheduled in favor of QEMU emulator
> +thread).
> +So this hypercall is supposed to be called by the guest when
> +the OS knows its not going to busy spin on memory thats
> +written by the emulator thread as above.
> +
> +a0: bit 0 contains enable bit, if 0 indicates that SCHED_OTHER
> +priority should be set for vcpu, if 1 indicates SCHED_FIFO
> +priority (the actual value for FIFO priority is decided
> +by the host).
> +
> +
> Index: kvm.fifopriohc-submit/include/uapi/linux/kvm_para.h
> ===================================================================
> --- kvm.fifopriohc-submit.orig/include/uapi/linux/kvm_para.h
> +++ kvm.fifopriohc-submit/include/uapi/linux/kvm_para.h
> @@ -15,6 +15,7 @@
>  #define KVM_E2BIG		E2BIG
>  #define KVM_EPERM		EPERM
>  #define KVM_EOPNOTSUPP		95
> +#define KVM_EINVAL		EINVAL
>  
>  #define KVM_HC_VAPIC_POLL_IRQ		1
>  #define KVM_HC_MMU_OP			2
> @@ -25,6 +26,7 @@
>  #define KVM_HC_MIPS_EXIT_VM		7
>  #define KVM_HC_MIPS_CONSOLE_OUTPUT	8
>  #define KVM_HC_CLOCK_PAIRING		9
> +#define KVM_HC_RT_PRIO			10
>  
>  /*
>   * hypercalls use architecture specific
> Index: kvm.fifopriohc-submit/arch/x86/kvm/x86.c
> ===================================================================
> --- kvm.fifopriohc-submit.orig/arch/x86/kvm/x86.c
> +++ kvm.fifopriohc-submit/arch/x86/kvm/x86.c
> @@ -66,6 +66,8 @@
>  #include <asm/pvclock.h>
>  #include <asm/div64.h>
>  #include <asm/irq_remapping.h>
> +#include <uapi/linux/sched/types.h>
> +#include <uapi/linux/sched.h>
>  
>  #define CREATE_TRACE_POINTS
>  #include "trace.h"
> @@ -6261,6 +6263,44 @@ void kvm_vcpu_deactivate_apicv(struct kv
>  	kvm_x86_ops->refresh_apicv_exec_ctrl(vcpu);
>  }
>  
> +static int convert_to_kvm_errcode(int error)
> +{
> +	switch (error) {
> +	case -EPERM:
> +		return -KVM_EPERM;
> +	case -EINVAL:
> +	default:
> +		return -KVM_EINVAL;
> +	}
> +}
> +
> +int kvm_pv_rt_prio(struct kvm_vcpu *vcpu, unsigned long a0)
> +{
> +	int ret;
> +	bool enable;
> +	struct sched_param param;
> +
> +	memset(&param, 0, sizeof(struct sched_param));
> +	param.sched_priority = vcpu->arch.rt_sched_priority;
> +
> +	enable = a0 & 0x1;
> +
> +	if (vcpu->arch.enable_rt_prio_hc == false)
> +		return -KVM_EPERM;
> +
> +	if (enable) {
> +		ret = sched_setscheduler(current, SCHED_FIFO, &param);
> +	} else {
> +		param.sched_priority = 0;
> +		ret = sched_setscheduler(current, SCHED_NORMAL, &param);
> +	}
> +
> +	if (ret)
> +		ret = convert_to_kvm_errcode(ret);
> +
> +	return ret;
> +}
> +
>  int kvm_emulate_hypercall(struct kvm_vcpu *vcpu)
>  {
>  	unsigned long nr, a0, a1, a2, a3, ret;
> @@ -6306,6 +6346,9 @@ int kvm_emulate_hypercall(struct kvm_vcp
>  		ret = kvm_pv_clock_pairing(vcpu, a0, a1);
>  		break;
>  #endif
> +	case KVM_HC_RT_PRIO:
> +		ret = kvm_pv_rt_prio(vcpu, a0);
> +		break;
>  	default:
>  		ret = -KVM_ENOSYS;
>  		break;
> 
>