Re: [PATCH v3 1/4] x86/kvm: Add "nopvspin" parameter to disable PV spinlocks

[Sean Christopherson]

On Mon, Sep 30, 2019 at 08:44:36PM +0800, Zhenzhong Duan wrote:
> There are cases where a guest tries to switch spinlocks to bare metal
> behavior (e.g. by setting "xen_nopvspin" on XEN platform and
> "hv_nopvspin" on HYPER_V).
> 
> That feature is missed on KVM, add a new parameter "nopvspin" to disable
> PV spinlocks for KVM guest.
> 
> This new parameter is also used to replace "xen_nopvspin" and
> "hv_nopvspin".

This is confusing as there are no Xen or Hyper-V changes in this patch.
Please make it clear that you're talking about future patches, e.g.:

  The new 'nopvspin' parameter will also replace Xen and Hyper-V specific
  parameters in future patches.

> 
> The global variable pvspin isn't defined as __initdata as it's used at
> runtime by XEN guest.

Same comment as above regarding what this patch is doing versus what will
be done in the future.  Arguably you should even mark it __initdata in
this patch and deal with conflict in the Xen patch, e.g. use it only to
set the existing xen_pvspin variable.

> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@xxxxxxxxxx>
> Cc: Jonathan Corbet <corbet@xxxxxxx>
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: Borislav Petkov <bp@xxxxxxxxx>
> Cc: "H. Peter Anvin" <hpa@xxxxxxxxx>
> Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx>
> Cc: Radim Krcmar <rkrcmar@xxxxxxxxxx>
> Cc: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> Cc: Vitaly Kuznetsov <vkuznets@xxxxxxxxxx>
> Cc: Wanpeng Li <wanpengli@xxxxxxxxxxx>
> Cc: Jim Mattson <jmattson@xxxxxxxxxx>
> Cc: Joerg Roedel <joro@xxxxxxxxxx>
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Cc: Will Deacon <will@xxxxxxxxxx>
> ---
>  Documentation/admin-guide/kernel-parameters.txt | 4 ++++
>  arch/x86/include/asm/qspinlock.h                | 1 +
>  arch/x86/kernel/kvm.c                           | 7 +++++++
>  kernel/locking/qspinlock.c                      | 7 +++++++
>  4 files changed, 19 insertions(+)
> 
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index c7ac2f3..4b956d8 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -5330,6 +5330,10 @@
>  			as generic guest with no PV drivers. Currently support
>  			XEN HVM, KVM, HYPER_V and VMWARE guest.
>  
> +	nopvspin	[X86,KVM] Disables the qspinlock slow path
> +			using PV optimizations which allow the hypervisor to
> +			'idle' the guest on lock contention.
> +
>  	xirc2ps_cs=	[NET,PCMCIA]
>  			Format:
>  			<irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
> diff --git a/arch/x86/include/asm/qspinlock.h b/arch/x86/include/asm/qspinlock.h
> index 444d6fd..34a4484 100644
> --- a/arch/x86/include/asm/qspinlock.h
> +++ b/arch/x86/include/asm/qspinlock.h
> @@ -32,6 +32,7 @@ static __always_inline u32 queued_fetch_set_pending_acquire(struct qspinlock *lo
>  extern void __pv_init_lock_hash(void);
>  extern void __pv_queued_spin_lock_slowpath(struct qspinlock *lock, u32 val);
>  extern void __raw_callee_save___pv_queued_spin_unlock(struct qspinlock *lock);
> +extern bool pvspin;
>  
>  #define	queued_spin_unlock queued_spin_unlock
>  /**
> diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
> index e820568..a4f108d 100644
> --- a/arch/x86/kernel/kvm.c
> +++ b/arch/x86/kernel/kvm.c
> @@ -842,6 +842,13 @@ void __init kvm_spinlock_init(void)
>  	if (num_possible_cpus() == 1)
>  		return;
>  
> +	if (!pvspin) {
> +		pr_info("PV spinlocks disabled\n");
> +		static_branch_disable(&virt_spin_lock_key);
> +		return;
> +	}
> +	pr_info("PV spinlocks enabled\n");

These prints could be confusing as KVM also disables PV spinlocks when it
sees KVM_HINTS_REALTIME.

> +
>  	__pv_init_lock_hash();
>  	pv_ops.lock.queued_spin_lock_slowpath = __pv_queued_spin_lock_slowpath;
>  	pv_ops.lock.queued_spin_unlock =
> diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c
> index 2473f10..945b510 100644
> --- a/kernel/locking/qspinlock.c
> +++ b/kernel/locking/qspinlock.c
> @@ -580,4 +580,11 @@ void queued_spin_lock_slowpath(struct qspinlock *lock, u32 val)
>  #include "qspinlock_paravirt.h"
>  #include "qspinlock.c"
>  
> +bool pvspin = true;

This can be __ro_after_init, or probably better __initdata and have Xen
snapshot the value for its use case.

Personal preference: I'd invert the bool and name it nopvspin to make it
easier to connect the variable to the kernel param.

> +static __init int parse_nopvspin(char *arg)
> +{
> +	pvspin = false;
> +	return 0;
> +}
> +early_param("nopvspin", parse_nopvspin);
>  #endif
> -- 
> 1.8.3.1
>