Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:7 as permitted sender) client-ip=2620:137:e000::3:7;
From:   Xu Lu <luxu.kernel@bytedance.com>
To:     paul.walmsley@sifive.com, palmer@dabbelt.com,
        aou@eecs.berkeley.edu, tglx@linutronix.de, maz@kernel.org,
        anup@brainfault.org, atishp@atishpatra.org
Cc:     dengliang.1214@bytedance.com, liyu.yukiteru@bytedance.com,
        sunjiadong.lff@bytedance.com, xieyongji@bytedance.com,
        lihangjing@bytedance.com, chaiwen.cc@bytedance.com,
        linux-kernel@vger.kernel.org, linux-riscv@lists.infradead.org,
        Xu Lu <luxu.kernel@bytedance.com>
Subject: [RFC 05/12] riscv: kvm: Switch back to CSR_STATUS masking when entering guest
Date:   Mon, 23 Oct 2023 16:29:04 +0800
Message-Id: <20231023082911.23242-6-luxu.kernel@bytedance.com>
In-Reply-To: <20231023082911.23242-1-luxu.kernel@bytedance.com>
References: <20231023082911.23242-1-luxu.kernel@bytedance.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

When kvm enters vcpu, it first disables local irqs before preparing vcpu
context and uses SRET instruction to enter guest mode after vcpu context
is ready, which automatically restores guest's irq status. However, after
we switch to CSR_IE masking for interrupt disabling, the SRET instruction
itself can not restore guest's irq status correctly as interrupts are
still masked by CSR_IE.

This commit handles this special case by switching to traditional
CSR_STATUS way to disable irqs before entering guest mode.

Signed-off-by: Xu Lu <luxu.kernel@bytedance.com>
---
 arch/riscv/include/asm/irqflags.h |  3 +++
 arch/riscv/kvm/vcpu.c             | 18 +++++++++++++-----
 2 files changed, 16 insertions(+), 5 deletions(-)

diff --git a/arch/riscv/include/asm/irqflags.h b/arch/riscv/include/asm/irqflags.h
index e0ff37315178..60c19f8b57f0 100644
--- a/arch/riscv/include/asm/irqflags.h
+++ b/arch/riscv/include/asm/irqflags.h
@@ -64,6 +64,9 @@ static inline void arch_local_irq_restore(unsigned long flags)
 	csr_write(CSR_IE, flags);
 }
 
+#define local_irq_enable_vcpu_run		local_irq_switch_on
+#define local_irq_disable_vcpu_run		local_irq_switch_off
+
 #else /* CONFIG_RISCV_PSEUDO_NMI */
 
 /* read interrupt enabled status */
diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
index 82229db1ce73..233408247da7 100644
--- a/arch/riscv/kvm/vcpu.c
+++ b/arch/riscv/kvm/vcpu.c
@@ -621,6 +621,14 @@ static void noinstr kvm_riscv_vcpu_enter_exit(struct kvm_vcpu *vcpu)
 	guest_state_exit_irqoff();
 }
 
+#ifndef local_irq_enable_vcpu_run
+#define local_irq_enable_vcpu_run		local_irq_enable
+#endif
+
+#ifndef local_irq_disable_vcpu_run
+#define local_irq_disable_vcpu_run		local_irq_disable
+#endif
+
 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
 {
 	int ret;
@@ -685,7 +693,7 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
 			continue;
 		}
 
-		local_irq_disable();
+		local_irq_disable_vcpu_run();
 
 		/*
 		 * Ensure we set mode to IN_GUEST_MODE after we disable
@@ -712,7 +720,7 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
 		    kvm_request_pending(vcpu) ||
 		    xfer_to_guest_mode_work_pending()) {
 			vcpu->mode = OUTSIDE_GUEST_MODE;
-			local_irq_enable();
+			local_irq_enable_vcpu_run();
 			preempt_enable();
 			kvm_vcpu_srcu_read_lock(vcpu);
 			continue;
@@ -757,12 +765,12 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
 		 * recognised, so we just hope that the CPU takes any pending
 		 * interrupts between the enable and disable.
 		 */
-		local_irq_enable();
-		local_irq_disable();
+		local_irq_enable_vcpu_run();
+		local_irq_disable_vcpu_run();
 
 		guest_timing_exit_irqoff();
 
-		local_irq_enable();
+		local_irq_enable_vcpu_run();
 
 		preempt_enable();
 
-- 
2.20.1