Linux KVM/arm64 development list
 help / color / mirror / Atom feed
From: Oliver Upton <oliver.upton@linux.dev>
To: kvmarm@lists.linux.dev
Cc: Marc Zyngier <maz@kernel.org>, Joey Gouly <joey.gouly@arm.com>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	Zenghui Yu <yuzenghui@huawei.com>,
	Mingwei Zhang <mizhang@google.com>,
	Colton Lewis <coltonlewis@google.com>,
	Raghavendra Rao Ananta <rananta@google.com>,
	James Clark <james.clark@linaro.org>,
	Oliver Upton <oliver.upton@linux.dev>
Subject: [PATCH v3 05/16] KVM: arm64: Evaluate debug owner at vcpu_load()
Date: Mon,  9 Dec 2024 10:09:15 -0800	[thread overview]
Message-ID: <20241209180926.2161373-6-oliver.upton@linux.dev> (raw)
In-Reply-To: <20241209180926.2161373-1-oliver.upton@linux.dev>

In preparation for tossing the debug_ptr mess, introduce an enumeration
to track the ownership of the debug registers while in the guest. Update
the owner at vcpu_load() based on whether the host needs to steal the
guest's debug context or if breakpoints/watchpoints are actively in use.

Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
---
 arch/arm64/include/asm/kvm_host.h | 11 ++++++++
 arch/arm64/kvm/arm.c              |  1 +
 arch/arm64/kvm/debug.c            | 47 +++++++++++++++++++++++++++++++
 arch/arm64/kvm/sys_regs.c         |  2 ++
 4 files changed, 61 insertions(+)

diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index d196bf0fce52..f65b30bab0ab 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -756,6 +756,12 @@ struct kvm_vcpu_arch {
 	struct kvm_guest_debug_arch vcpu_debug_state;
 	struct kvm_guest_debug_arch external_debug_state;
 
+	enum {
+		VCPU_DEBUG_FREE,
+		VCPU_DEBUG_HOST_OWNED,
+		VCPU_DEBUG_GUEST_OWNED,
+	} debug_owner;
+
 	/* VGIC state */
 	struct vgic_cpu vgic_cpu;
 	struct arch_timer_cpu timer_cpu;
@@ -1345,10 +1351,15 @@ 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);
+void kvm_vcpu_load_debug(struct kvm_vcpu *vcpu);
+void kvm_debug_set_guest_ownership(struct kvm_vcpu *vcpu);
 
 #define kvm_vcpu_os_lock_enabled(vcpu)		\
 	(!!(__vcpu_sys_reg(vcpu, OSLSR_EL1) & OSLSR_EL1_OSLK))
 
+#define kvm_host_owns_debug_regs(vcpu)		\
+	((vcpu)->arch.debug_owner == VCPU_DEBUG_HOST_OWNED)
+
 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/arm.c b/arch/arm64/kvm/arm.c
index 3822774840e1..a068337da52a 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -598,6 +598,7 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 
 	kvm_vgic_load(vcpu);
 	kvm_timer_vcpu_load(vcpu);
+	kvm_vcpu_load_debug(vcpu);
 	if (has_vhe())
 		kvm_vcpu_load_vhe(vcpu);
 	kvm_arch_vcpu_load_fp(vcpu);
diff --git a/arch/arm64/kvm/debug.c b/arch/arm64/kvm/debug.c
index 1ee2fd765b62..da297cc92ed5 100644
--- a/arch/arm64/kvm/debug.c
+++ b/arch/arm64/kvm/debug.c
@@ -313,3 +313,50 @@ void kvm_init_host_debug_data(void)
 	    !(read_sysreg_s(SYS_TRBIDR_EL1) & TRBIDR_EL1_P))
 		host_data_set_flag(HAS_TRBE);
 }
+
+void kvm_vcpu_load_debug(struct kvm_vcpu *vcpu)
+{
+	u64 mdscr;
+
+	/* Must be called before kvm_vcpu_load_vhe() */
+	KVM_BUG_ON(vcpu_get_flag(vcpu, SYSREGS_ON_CPU), vcpu->kvm);
+
+	/*
+	 * Determine which of the possible debug states we're in:
+	 *
+	 *  - VCPU_DEBUG_HOST_OWNED: KVM has taken ownership of the guest's
+	 *    breakpoint/watchpoint registers, or needs to use MDSCR_EL1 to do
+	 *    software step or emulate the effects of the OS Lock being enabled.
+	 *
+	 *  - VCPU_DEBUG_GUEST_OWNED: The guest has debug exceptions enabled, and
+	 *    the breakpoint/watchpoint registers need to be loaded eagerly.
+	 *
+	 *  - VCPU_DEBUG_FREE: Neither of the above apply, no breakpoint/watchpoint
+	 *    context needs to be loaded on the CPU.
+	 */
+	if (vcpu->guest_debug || kvm_vcpu_os_lock_enabled(vcpu)) {
+		vcpu->arch.debug_owner = VCPU_DEBUG_HOST_OWNED;
+	} else {
+		mdscr = vcpu_read_sys_reg(vcpu, MDSCR_EL1);
+
+		if (mdscr & (MDSCR_EL1_KDE | MDSCR_EL1_MDE))
+			vcpu->arch.debug_owner = VCPU_DEBUG_GUEST_OWNED;
+		else
+			vcpu->arch.debug_owner = VCPU_DEBUG_FREE;
+	}
+}
+
+/*
+ * Updates ownership of the debug registers after a trapped guest access to a
+ * breakpoint/watchpoint register. Host ownership of the debug registers is of
+ * strictly higher priority, and it is the responsibility of the VMM to emulate
+ * guest debug exceptions in this configuration.
+ */
+void kvm_debug_set_guest_ownership(struct kvm_vcpu *vcpu)
+{
+	if (kvm_host_owns_debug_regs(vcpu))
+		return;
+
+	WARN_ON_ONCE(vcpu->arch.debug_owner == VCPU_DEBUG_GUEST_OWNED);
+	vcpu->arch.debug_owner = VCPU_DEBUG_GUEST_OWNED;
+}
diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
index 83c6b4a07ef5..b17cd93c1c8d 100644
--- a/arch/arm64/kvm/sys_regs.c
+++ b/arch/arm64/kvm/sys_regs.c
@@ -656,6 +656,7 @@ static bool trap_debug_regs(struct kvm_vcpu *vcpu,
 	if (p->is_write)
 		vcpu_set_flag(vcpu, DEBUG_DIRTY);
 
+	kvm_debug_set_guest_ownership(vcpu);
 	trace_trap_reg(__func__, r->reg, p->is_write, p->regval);
 
 	return true;
@@ -709,6 +710,7 @@ static bool trap_bvr(struct kvm_vcpu *vcpu,
 	else
 		dbg_to_reg(vcpu, p, rd, dbg_reg);
 
+	kvm_debug_set_guest_ownership(vcpu);
 	trace_trap_reg(__func__, rd->CRm, p->is_write, *dbg_reg);
 
 	return true;
-- 
2.39.5


  parent reply	other threads:[~2024-12-09 18:09 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-12-09 18:09 [PATCH v3 00/16] KVM: arm64: Debug cleanups Oliver Upton
2024-12-09 18:09 ` [PATCH v3 01/16] KVM: arm64: Drop MDSCR_EL1_DEBUG_MASK Oliver Upton
2024-12-09 18:09 ` [PATCH v3 02/16] KVM: arm64: Get rid of __kvm_get_mdcr_el2() and related warts Oliver Upton
2024-12-09 18:09 ` [PATCH v3 03/16] KVM: arm64: Track presence of SPE/TRBE in kvm_host_data instead of vCPU Oliver Upton
2024-12-09 18:09 ` [PATCH v3 04/16] KVM: arm64: Move host SME/SVE tracking flags to host data Oliver Upton
2024-12-09 18:09 ` Oliver Upton [this message]
2024-12-09 18:09 ` [PATCH v3 06/16] KVM: arm64: Clean up KVM_SET_GUEST_DEBUG handler Oliver Upton
2024-12-09 18:09 ` [PATCH v3 07/16] KVM: arm64: Select debug state to save/restore based on debug owner Oliver Upton
2024-12-09 18:09 ` [PATCH v3 08/16] KVM: arm64: Remove debug tracepoints Oliver Upton
2024-12-09 18:09 ` [PATCH v3 09/16] KVM: arm64: Remove vestiges of debug_ptr Oliver Upton
2024-12-09 18:09 ` [PATCH v3 10/16] KVM: arm64: Use debug_owner to track if debug regs need save/restore Oliver Upton
2024-12-18 18:11   ` Marc Zyngier
2024-12-18 20:36     ` Oliver Upton
2024-12-09 18:09 ` [PATCH v3 11/16] KVM: arm64: Reload vCPU for accesses to OSLAR_EL1 Oliver Upton
2024-12-09 18:09 ` [PATCH v3 12/16] KVM: arm64: Compute MDCR_EL2 at vcpu_load() Oliver Upton
2024-12-18 18:40   ` Marc Zyngier
2024-12-19 18:00     ` Oliver Upton
2024-12-19 18:16       ` Marc Zyngier
2024-12-09 18:09 ` [PATCH v3 13/16] KVM: arm64: Don't hijack guest context MDSCR_EL1 Oliver Upton
2024-12-09 18:09 ` [PATCH v3 14/16] KVM: arm64: Manage software step state at load/put Oliver Upton
2024-12-09 18:09 ` [PATCH v3 15/16] KVM: arm64: nv: Honor MDCR_EL2.TDE routing for debug exceptions Oliver Upton
2024-12-09 18:09 ` [PATCH v3 16/16] KVM: arm64: Avoid reading ID_AA64DFR0_EL1 for debug save/restore Oliver Upton
2024-12-10 10:30 ` [PATCH v3 00/16] KVM: arm64: Debug cleanups James Clark

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=20241209180926.2161373-6-oliver.upton@linux.dev \
    --to=oliver.upton@linux.dev \
    --cc=coltonlewis@google.com \
    --cc=james.clark@linaro.org \
    --cc=joey.gouly@arm.com \
    --cc=kvmarm@lists.linux.dev \
    --cc=maz@kernel.org \
    --cc=mizhang@google.com \
    --cc=rananta@google.com \
    --cc=suzuki.poulose@arm.com \
    --cc=yuzenghui@huawei.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox