[PATCH v3 4/6] KVM: arm64: Emulate the OS Lock

[Marc Zyngier]

On Tue, 23 Nov 2021 21:01:07 +0000,
Oliver Upton <oupton at google.com> wrote:
> 
> The OS lock blocks all debug exceptions at every EL. To date, KVM has
> not implemented the OS lock for its guests, despite the fact that it is
> mandatory per the architecture. Simple context switching between the
> guest and host is not appropriate, as its effects are not constrained to
> the guest context.
> 
> Emulate the OS Lock by clearing MDE and SS in MDSCR_EL1, thereby
> blocking all but software breakpoint instructions. To handle breakpoint
> instructions, trap debug exceptions to EL2 and skip the instruction.

Skipping breakpoint instructions? I don't think you can do that, as
the guest does rely on BRK always being effective. I also don't see
where you do that...

> 
> Signed-off-by: Oliver Upton <oupton at google.com>
> ---
>  arch/arm64/include/asm/kvm_host.h |  4 ++++
>  arch/arm64/kvm/debug.c            | 27 +++++++++++++++++++++++----
>  arch/arm64/kvm/sys_regs.c         |  6 +++---
>  3 files changed, 30 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index 53fc8a6eaf1c..e5a06ff1cba6 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -726,6 +726,10 @@ void kvm_arm_vcpu_init_debug(struct kvm_vcpu *vcpu);
>  void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
>  void kvm_arm_clear_debug(struct kvm_vcpu *vcpu);
>  void kvm_arm_reset_debug_ptr(struct kvm_vcpu *vcpu);
> +
> +#define kvm_vcpu_os_lock_enabled(vcpu)		\
> +	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & SYS_OSLSR_OSLK))
> +
>  int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
>  			       struct kvm_device_attr *attr);
>  int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
> diff --git a/arch/arm64/kvm/debug.c b/arch/arm64/kvm/debug.c
> index db9361338b2a..7835c76347ce 100644
> --- a/arch/arm64/kvm/debug.c
> +++ b/arch/arm64/kvm/debug.c
> @@ -53,6 +53,14 @@ static void restore_guest_debug_regs(struct kvm_vcpu *vcpu)
>  				vcpu_read_sys_reg(vcpu, MDSCR_EL1));
>  }
>  
> +/*
> + * Returns true if the host needs to use the debug registers.
> + */
> +static inline bool host_using_debug_regs(struct kvm_vcpu *vcpu)
> +{
> +	return vcpu->guest_debug || kvm_vcpu_os_lock_enabled(vcpu);

Just the name of the function has sent my head spinning. Even if the
*effects* of the host debug and the OS Lock are vaguely similar from
the guest PoV, they really are different things, and I'd rather not
lob them together.

> +}
> +
>  /**
>   * kvm_arm_init_debug - grab what we need for debug
>   *
> @@ -105,9 +113,11 @@ static void kvm_arm_setup_mdcr_el2(struct kvm_vcpu *vcpu)
>  	 *  - Userspace is using the hardware to debug the guest
>  	 *  (KVM_GUESTDBG_USE_HW is set).
>  	 *  - The guest is not using debug (KVM_ARM64_DEBUG_DIRTY is clear).
> +	 *  - The guest has enabled the OS Lock (debug exceptions are blocked).
>  	 */
>  	if ((vcpu->guest_debug & KVM_GUESTDBG_USE_HW) ||
> -	    !(vcpu->arch.flags & KVM_ARM64_DEBUG_DIRTY))
> +	    !(vcpu->arch.flags & KVM_ARM64_DEBUG_DIRTY) ||
> +	    kvm_vcpu_os_lock_enabled(vcpu))
>  		vcpu->arch.mdcr_el2 |= MDCR_EL2_TDA;
>  
>  	trace_kvm_arm_set_dreg32("MDCR_EL2", vcpu->arch.mdcr_el2);
> @@ -160,8 +170,10 @@ void kvm_arm_setup_debug(struct kvm_vcpu *vcpu)
>  
>  	kvm_arm_setup_mdcr_el2(vcpu);
>  
> -	/* Is Guest debugging in effect? */
> -	if (vcpu->guest_debug) {
> +	/*
> +	 * Check if we need to use the debug registers.
> +	 */
> +	if (host_using_debug_regs(vcpu)) {

I'd rather you expand the helper here and add the comment you have in
the commit message explaining the machine-wide effect of the OS Lock.

>  		/* Save guest debug state */
>  		save_guest_debug_regs(vcpu);
>  
> @@ -223,6 +235,10 @@ void kvm_arm_setup_debug(struct kvm_vcpu *vcpu)
>  			trace_kvm_arm_set_regset("WAPTS", get_num_wrps(),
>  						&vcpu->arch.debug_ptr->dbg_wcr[0],
>  						&vcpu->arch.debug_ptr->dbg_wvr[0]);
> +		} else if (kvm_vcpu_os_lock_enabled(vcpu)) {
> +			mdscr = vcpu_read_sys_reg(vcpu, MDSCR_EL1);
> +			mdscr &= ~DBG_MDSCR_MDE;
> +			vcpu_write_sys_reg(vcpu, mdscr, MDSCR_EL1);
>  		}
>  	}
>  
> @@ -244,7 +260,10 @@ void kvm_arm_clear_debug(struct kvm_vcpu *vcpu)
>  {
>  	trace_kvm_arm_clear_debug(vcpu->guest_debug);
>  
> -	if (vcpu->guest_debug) {
> +	/*
> +	 * Restore the guest's debug registers if we were using them.
> +	 */
> +	if (host_using_debug_regs(vcpu)) {
>  		restore_guest_debug_regs(vcpu);
>  
>  		/*
> diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
> index 5dbdb45d6d44..1346906f5c46 100644
> --- a/arch/arm64/kvm/sys_regs.c
> +++ b/arch/arm64/kvm/sys_regs.c
> @@ -1453,9 +1453,9 @@ static unsigned int mte_visibility(const struct kvm_vcpu *vcpu,
>   * Debug handling: We do trap most, if not all debug related system
>   * registers. The implementation is good enough to ensure that a guest
>   * can use these with minimal performance degradation. The drawback is
> - * that we don't implement any of the external debug, none of the
> - * OSlock protocol. This should be revisited if we ever encounter a
> - * more demanding guest...
> + * that we don't implement any of the external debug architecture.
> + * This should be revisited if we ever encounter a more demanding
> + * guest...
>   */
>  static const struct sys_reg_desc sys_reg_descs[] = {
>  	{ SYS_DESC(SYS_DC_ISW), access_dcsw },

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.