Received-SPF: pass (google.com: domain of linux-kernel+bounces-81680-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
From: Lai Jiangshan <jiangshanlai@gmail.com>
To: linux-kernel@vger.kernel.org
Cc: Lai Jiangshan <jiangshan.ljs@antgroup.com>,
	Hou Wenlong <houwenlong.hwl@antgroup.com>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Sean Christopherson <seanjc@google.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Borislav Petkov <bp@alien8.de>,
	Ingo Molnar <mingo@redhat.com>,
	kvm@vger.kernel.org,
	Paolo Bonzini <pbonzini@redhat.com>,
	x86@kernel.org,
	Kees Cook <keescook@chromium.org>,
	Juergen Gross <jgross@suse.com>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	"H. Peter Anvin" <hpa@zytor.com>
Subject: [RFC PATCH 34/73] KVM: x86/PVM: Handle ERETU/ERETS synthetic instruction
Date: Mon, 26 Feb 2024 22:35:51 +0800
Message-Id: <20240226143630.33643-35-jiangshanlai@gmail.com>
In-Reply-To: <20240226143630.33643-1-jiangshanlai@gmail.com>
References: <20240226143630.33643-1-jiangshanlai@gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

PVM uses the ERETU synthetic instruction to return to user mode and the
ERETS instruction to return to supervisor mode. Similar to event
injection, information passing is different. For the ERETU, information
is passed by the shared PVCS structure, and for the ERETS, information
is passed by the current guest stack.

Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
Signed-off-by: Hou Wenlong <houwenlong.hwl@antgroup.com>
---
 arch/x86/kvm/pvm/pvm.c | 74 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 74 insertions(+)

diff --git a/arch/x86/kvm/pvm/pvm.c b/arch/x86/kvm/pvm/pvm.c
index c6fd01c19c3e..514f0573f70f 100644
--- a/arch/x86/kvm/pvm/pvm.c
+++ b/arch/x86/kvm/pvm/pvm.c
@@ -1153,12 +1153,86 @@ static void pvm_setup_mce(struct kvm_vcpu *vcpu)
 {
 }
 
+static int handle_synthetic_instruction_return_user(struct kvm_vcpu *vcpu)
+{
+	struct vcpu_pvm *pvm = to_pvm(vcpu);
+	struct pvm_vcpu_struct *pvcs;
+
+	// instruction to return user means nmi allowed.
+	pvm->nmi_mask = false;
+
+	/*
+	 * switch to user mode before kvm_set_rflags() to avoid PVM_EVENT_FLAGS_IF
+	 * to be set.
+	 */
+	switch_to_umod(vcpu);
+
+	pvcs = pvm_get_vcpu_struct(pvm);
+	if (!pvcs) {
+		kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
+		return 1;
+	}
+
+	/*
+	 * pvm_set_rflags() doesn't clear PVM_EVENT_FLAGS_IP
+	 * for user mode, so clear it here.
+	 */
+	if (pvcs->event_flags & PVM_EVENT_FLAGS_IP) {
+		pvcs->event_flags &= ~PVM_EVENT_FLAGS_IP;
+		kvm_make_request(KVM_REQ_EVENT, vcpu);
+	}
+
+	pvm->hw_cs = pvcs->user_cs | USER_RPL;
+	pvm->hw_ss = pvcs->user_ss | USER_RPL;
+
+	pvm_write_guest_gs_base(pvm, pvcs->user_gsbase);
+	kvm_set_rflags(vcpu, pvcs->eflags | X86_EFLAGS_IF);
+	kvm_rip_write(vcpu, pvcs->rip);
+	kvm_rsp_write(vcpu, pvcs->rsp);
+	kvm_rcx_write(vcpu, pvcs->rcx);
+	kvm_r11_write(vcpu, pvcs->r11);
+
+	pvm_put_vcpu_struct(pvm, false);
+
+	return 1;
+}
+
+static int handle_synthetic_instruction_return_supervisor(struct kvm_vcpu *vcpu)
+{
+	struct vcpu_pvm *pvm = to_pvm(vcpu);
+	unsigned long rsp = kvm_rsp_read(vcpu);
+	struct pvm_supervisor_event frame;
+	struct x86_exception e;
+
+	if (kvm_read_guest_virt(vcpu, rsp, &frame, sizeof(frame), &e)) {
+		kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
+		return 1;
+	}
+
+	// instruction to return supervisor means nmi allowed.
+	pvm->nmi_mask = false;
+
+	kvm_set_rflags(vcpu, frame.rflags);
+	kvm_rip_write(vcpu, frame.rip);
+	kvm_rsp_write(vcpu, frame.rsp);
+	kvm_rcx_write(vcpu, frame.rcx);
+	kvm_r11_write(vcpu, frame.r11);
+
+	return 1;
+}
+
 static int handle_exit_syscall(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_pvm *pvm = to_pvm(vcpu);
+	unsigned long rip = kvm_rip_read(vcpu);
 
 	if (!is_smod(pvm))
 		return do_pvm_user_event(vcpu, PVM_SYSCALL_VECTOR, false, 0);
+
+	if (rip == pvm->msr_retu_rip_plus2)
+		return handle_synthetic_instruction_return_user(vcpu);
+	if (rip == pvm->msr_rets_rip_plus2)
+		return handle_synthetic_instruction_return_supervisor(vcpu);
 	return 1;
 }
 
-- 
2.19.1.6.gb485710b