kvm_riscv_vcpu_aia_imsic_update() assumes that the vCPU IMSIC state has
already been initialized and unconditionally accesses imsic->vsfile_lock.
However, in fuzzed ioctl sequences, the AIA device may be initialized at
the VM level while the per-vCPU IMSIC state is still NULL.

This leads to invalid access when entering the vCPU run loop before
IMSIC initialization has completed.

The crash manifests as:
  Unable to handle kernel paging request at virtual address
  dfffffff00000006
  ...
  kvm_riscv_vcpu_aia_imsic_update arch/riscv/kvm/aia_imsic.c:801
  kvm_riscv_vcpu_aia_update arch/riscv/kvm/aia_device.c:493
  kvm_arch_vcpu_ioctl_run arch/riscv/kvm/vcpu.c:927
  ...

Add a guard to skip the IMSIC update path when imsic_state is NULL. This
allows the vCPU run loop to continue safely.

This issue was discovered during fuzzing of RISC-V KVM code.

Fixes: db8b7e97d6137a ("RISC-V: KVM: Add in-kernel virtualization of AIA IMSIC")
Signed-off-by: Jiakai Xu <xujiakai2025@iscas.ac.cn>
Signed-off-by: Jiakai Xu <jiakaiPeanut@gmail.com>
---
 arch/riscv/kvm/aia_imsic.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/arch/riscv/kvm/aia_imsic.c b/arch/riscv/kvm/aia_imsic.c
index e597e86491c3b..a7d387e280d49 100644
--- a/arch/riscv/kvm/aia_imsic.c
+++ b/arch/riscv/kvm/aia_imsic.c
@@ -797,6 +797,10 @@ int kvm_riscv_vcpu_aia_imsic_update(struct kvm_vcpu *vcpu)
 	if (kvm->arch.aia.mode == KVM_DEV_RISCV_AIA_MODE_EMUL)
 		return 1;
 
+	/* IMSIC vCPU state may not be initialized yet */
+	if (!imsic)
+		return 1;
+
 	/* Read old IMSIC VS-file details */
 	read_lock_irqsave(&imsic->vsfile_lock, flags);
 	old_vsfile_hgei = imsic->vsfile_hgei;
-- 
2.34.1

On Tue, Jan 27, 2026 at 2:13 PM Jiakai Xu <jiakaipeanut@gmail.com> wrote:
>
> kvm_riscv_vcpu_aia_imsic_update() assumes that the vCPU IMSIC state has
> already been initialized and unconditionally accesses imsic->vsfile_lock.
> However, in fuzzed ioctl sequences, the AIA device may be initialized at
> the VM level while the per-vCPU IMSIC state is still NULL.
>
> This leads to invalid access when entering the vCPU run loop before
> IMSIC initialization has completed.
>
> The crash manifests as:
>   Unable to handle kernel paging request at virtual address
>   dfffffff00000006
>   ...
>   kvm_riscv_vcpu_aia_imsic_update arch/riscv/kvm/aia_imsic.c:801
>   kvm_riscv_vcpu_aia_update arch/riscv/kvm/aia_device.c:493
>   kvm_arch_vcpu_ioctl_run arch/riscv/kvm/vcpu.c:927
>   ...
>
> Add a guard to skip the IMSIC update path when imsic_state is NULL. This
> allows the vCPU run loop to continue safely.
>
> This issue was discovered during fuzzing of RISC-V KVM code.
>
> Fixes: db8b7e97d6137a ("RISC-V: KVM: Add in-kernel virtualization of AIA IMSIC")
> Signed-off-by: Jiakai Xu <xujiakai2025@iscas.ac.cn>
> Signed-off-by: Jiakai Xu <jiakaiPeanut@gmail.com>

LGTM.

Reviewed-by: Anup Patel <anup@brainfault.org>

Queued this patch for Linux-6.20

Thanks,
Anup


> ---
>  arch/riscv/kvm/aia_imsic.c | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/arch/riscv/kvm/aia_imsic.c b/arch/riscv/kvm/aia_imsic.c
> index e597e86491c3b..a7d387e280d49 100644
> --- a/arch/riscv/kvm/aia_imsic.c
> +++ b/arch/riscv/kvm/aia_imsic.c
> @@ -797,6 +797,10 @@ int kvm_riscv_vcpu_aia_imsic_update(struct kvm_vcpu *vcpu)
>         if (kvm->arch.aia.mode == KVM_DEV_RISCV_AIA_MODE_EMUL)
>                 return 1;
>
> +       /* IMSIC vCPU state may not be initialized yet */
> +       if (!imsic)
> +               return 1;
> +
>         /* Read old IMSIC VS-file details */
>         read_lock_irqsave(&imsic->vsfile_lock, flags);
>         old_vsfile_hgei = imsic->vsfile_hgei;
> --
> 2.34.1
>