linux-kernel - [PATCH v3] RISC-V: KVM: Fix guest page fault within HLV* instructions

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [day] [month] [year] [list]

Message-Id: <20251121133543.46822-1-fangyu.yu@linux.alibaba.com>
Date: Fri, 21 Nov 2025 21:35:43 +0800
From: fangyu.yu@...ux.alibaba.com
To: anup@...infault.org,
	atish.patra@...ux.dev,
	pjw@...nel.org,
	palmer@...belt.com,
	aou@...s.berkeley.edu,
	alex@...ti.fr
Cc: guoren@...nel.org,
	ajones@...tanamicro.com,
	kvm@...r.kernel.org,
	kvm-riscv@...ts.infradead.org,
	linux-riscv@...ts.infradead.org,
	linux-kernel@...r.kernel.org,
	Fangyu Yu <fangyu.yu@...ux.alibaba.com>
Subject: [PATCH v3] RISC-V: KVM: Fix guest page fault within HLV* instructions

From: Fangyu Yu <fangyu.yu@...ux.alibaba.com>

When executing HLV* instructions at the HS mode, a guest page fault
may occur when a g-stage page table migration between triggering the
virtual instruction exception and executing the HLV* instruction.

This may be a corner case, and one simpler way to handle this is to
re-execute the instruction where the virtual  instruction exception
occurred, and the guest page fault will be automatically handled.

Fixes: b91f0e4cb8a3 ("RISC-V: KVM: Factor-out instruction emulation into separate sources")
Signed-off-by: Fangyu Yu <fangyu.yu@...ux.alibaba.com>

---
Changes in v3:
- Add a helper function to avoid repeating the same paragraph(suggested by drew)
- Link to v2: https://lore.kernel.org/linux-riscv/20251111135506.8526-1-fangyu.yu@linux.alibaba.com/

Changes in v2:
- Remove unnecessary modifications and add comments(suggested by Anup)
- Update Fixes tag
- Link to v1: https://lore.kernel.org/linux-riscv/20250912134332.22053-1-fangyu.yu@linux.alibaba.com/
---
 arch/riscv/kvm/vcpu_insn.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/arch/riscv/kvm/vcpu_insn.c b/arch/riscv/kvm/vcpu_insn.c
index de1f96ea6225..4d89b94128ae 100644
--- a/arch/riscv/kvm/vcpu_insn.c
+++ b/arch/riscv/kvm/vcpu_insn.c
@@ -298,6 +298,22 @@ static int system_opcode_insn(struct kvm_vcpu *vcpu, struct kvm_run *run,
 	return (rc <= 0) ? rc : 1;
 }
 
+static bool is_load_guest_page_fault(unsigned long scause)
+{
+	/**
+	 * If a g-stage page fault occurs, the direct approach
+	 * is to let the g-stage page fault handler handle it
+	 * naturally, however, calling the g-stage page fault
+	 * handler here seems rather strange.
+	 * Considering this is a corner case, we can directly
+	 * return to the guest and re-execute the same PC, this
+	 * will trigger a g-stage page fault again and then the
+	 * regular g-stage page fault handler will populate
+	 * g-stage page table.
+	 */
+	return (scause == EXC_LOAD_GUEST_PAGE_FAULT);
+}
+
 /**
  * kvm_riscv_vcpu_virtual_insn -- Handle virtual instruction trap
  *
@@ -323,6 +339,8 @@ int kvm_riscv_vcpu_virtual_insn(struct kvm_vcpu *vcpu, struct kvm_run *run,
 							  ct->sepc,
 							  &utrap);
 			if (utrap.scause) {
+				if (is_load_guest_page_fault(utrap.scause))
+					return 1;
 				utrap.sepc = ct->sepc;
 				kvm_riscv_vcpu_trap_redirect(vcpu, &utrap);
 				return 1;
@@ -378,6 +396,8 @@ int kvm_riscv_vcpu_mmio_load(struct kvm_vcpu *vcpu, struct kvm_run *run,
 		insn = kvm_riscv_vcpu_unpriv_read(vcpu, true, ct->sepc,
 						  &utrap);
 		if (utrap.scause) {
+			if (is_load_guest_page_fault(utrap.scause))
+				return 1;
 			/* Redirect trap if we failed to read instruction */
 			utrap.sepc = ct->sepc;
 			kvm_riscv_vcpu_trap_redirect(vcpu, &utrap);
@@ -504,6 +524,8 @@ int kvm_riscv_vcpu_mmio_store(struct kvm_vcpu *vcpu, struct kvm_run *run,
 		insn = kvm_riscv_vcpu_unpriv_read(vcpu, true, ct->sepc,
 						  &utrap);
 		if (utrap.scause) {
+			if (is_load_guest_page_fault(utrap.scause))
+				return 1;
 			/* Redirect trap if we failed to read instruction */
 			utrap.sepc = ct->sepc;
 			kvm_riscv_vcpu_trap_redirect(vcpu, &utrap);
-- 
2.50.1