| Message ID | 20240807154943.150540-2-ajones@ventanamicro.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | RISC-V: KVM: Fix sbiret init before forwarding to userspace | expand |
On Wed, Aug 7, 2024 at 9:19 PM Andrew Jones <ajones@ventanamicro.com> wrote: > > When forwarding SBI calls to userspace ensure sbiret.error is > initialized to SBI_ERR_NOT_SUPPORTED first, in case userspace > neglects to set it to anything. If userspace neglects it then we > can't be sure it did anything else either, so we just report it > didn't do or try anything. Just init sbiret.value to zero, which is > the preferred value to return when nothing special is specified. > > KVM was already initializing both sbiret.error and sbiret.value, but > the values used appear to come from a copy+paste of the __sbi_ecall() > implementation, i.e. a0 and a1, which don't apply prior to the call > being executed, nor at all when forwarding to userspace. > > Fixes: dea8ee31a039 ("RISC-V: KVM: Add SBI v0.1 support") > Signed-off-by: Andrew Jones <ajones@ventanamicro.com> Queued this patch for Linux-6.11 fixes. Thanks, Anup > --- > arch/riscv/kvm/vcpu_sbi.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/arch/riscv/kvm/vcpu_sbi.c b/arch/riscv/kvm/vcpu_sbi.c > index 62f409d4176e..7de128be8db9 100644 > --- a/arch/riscv/kvm/vcpu_sbi.c > +++ b/arch/riscv/kvm/vcpu_sbi.c > @@ -127,8 +127,8 @@ void kvm_riscv_vcpu_sbi_forward(struct kvm_vcpu *vcpu, struct kvm_run *run) > run->riscv_sbi.args[3] = cp->a3; > run->riscv_sbi.args[4] = cp->a4; > run->riscv_sbi.args[5] = cp->a5; > - run->riscv_sbi.ret[0] = cp->a0; > - run->riscv_sbi.ret[1] = cp->a1; > + run->riscv_sbi.ret[0] = SBI_ERR_NOT_SUPPORTED; > + run->riscv_sbi.ret[1] = 0; > } > > void kvm_riscv_vcpu_sbi_system_reset(struct kvm_vcpu *vcpu, > -- > 2.45.2 > >
diff --git a/arch/riscv/kvm/vcpu_sbi.c b/arch/riscv/kvm/vcpu_sbi.c index 62f409d4176e..7de128be8db9 100644 --- a/arch/riscv/kvm/vcpu_sbi.c +++ b/arch/riscv/kvm/vcpu_sbi.c @@ -127,8 +127,8 @@ void kvm_riscv_vcpu_sbi_forward(struct kvm_vcpu *vcpu, struct kvm_run *run) run->riscv_sbi.args[3] = cp->a3; run->riscv_sbi.args[4] = cp->a4; run->riscv_sbi.args[5] = cp->a5; - run->riscv_sbi.ret[0] = cp->a0; - run->riscv_sbi.ret[1] = cp->a1; + run->riscv_sbi.ret[0] = SBI_ERR_NOT_SUPPORTED; + run->riscv_sbi.ret[1] = 0; } void kvm_riscv_vcpu_sbi_system_reset(struct kvm_vcpu *vcpu,
When forwarding SBI calls to userspace ensure sbiret.error is initialized to SBI_ERR_NOT_SUPPORTED first, in case userspace neglects to set it to anything. If userspace neglects it then we can't be sure it did anything else either, so we just report it didn't do or try anything. Just init sbiret.value to zero, which is the preferred value to return when nothing special is specified. KVM was already initializing both sbiret.error and sbiret.value, but the values used appear to come from a copy+paste of the __sbi_ecall() implementation, i.e. a0 and a1, which don't apply prior to the call being executed, nor at all when forwarding to userspace. Fixes: dea8ee31a039 ("RISC-V: KVM: Add SBI v0.1 support") Signed-off-by: Andrew Jones <ajones@ventanamicro.com> --- arch/riscv/kvm/vcpu_sbi.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)