On Thu, Dec 22, 2016 at 04:29:27AM +1000, Nicholas Piggin wrote:
> 64-bit Book3S exception handlers must find the dynamic kernel base
> to add to the target address when branching beyond __end_interrupts,
> in order to support kernel running at non-0 physical address.
>
> Support this in KVM by branching with CTR, similarly to regular
> interrupt handlers. The guest CTR saved in HSTATE_SCRATCH1 and
> restored after the branch.
>
> Without this, the host kernel hangs and crashes randomly when it is
> running at a non-0 address and a KVM guest is started.
>
> Signed-off-by: Nicholas Piggin <npig...@gmail.com>
Unfortunately, this patch causes guests to crash in various ways when
running under a CONFIG_RELOCATABLE kernel. See below...
> diff --git a/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> index 11882aac8216..c18ce740452b 100644
> --- a/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> +++ b/arch/powerpc/kvm/book3s_hv_rmhandlers.S
> @@ -1060,15 +1060,16 @@ kvmppc_interrupt_hv:
> * R12 = (guest CR << 32) | interrupt vector
> * R13 = PACA
> * guest R12 saved in shadow VCPU SCRATCH0
> + * guest CTR saved in shadow VCPU SCRATCH1 if RELOCATABLE
> * guest R13 saved in SPRN_SCRATCH0
> */
> - std r9, HSTATE_SCRATCH1(r13)
> + std r9, HSTATE_SCRATCH2(r13)
> lbz r9, HSTATE_IN_GUEST(r13)
> cmpwi r9, KVM_GUEST_MODE_HOST_HV
> beq kvmppc_bad_host_intr
> #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE
> cmpwi r9, KVM_GUEST_MODE_GUEST
> - ld r9, HSTATE_SCRATCH1(r13)
> + ld r9, HSTATE_SCRATCH2(r13)
> beq kvmppc_interrupt_pr
> #endif
> /* We're now back in the host but in guest MMU context */
> @@ -1088,7 +1089,7 @@ kvmppc_interrupt_hv:
> std r6, VCPU_GPR(R6)(r9)
> std r7, VCPU_GPR(R7)(r9)
> std r8, VCPU_GPR(R8)(r9)
> - ld r0, HSTATE_SCRATCH1(r13)
> + ld r0, HSTATE_SCRATCH2(r13)
> std r0, VCPU_GPR(R9)(r9)
> std r10, VCPU_GPR(R10)(r9)
> std r11, VCPU_GPR(R11)(r9)
> @@ -1151,7 +1152,11 @@ END_FTR_SECTION_IFSET(CPU_FTR_HAS_PPR)
> 11: stw r3,VCPU_HEIR(r9)
>
> /* these are volatile across C function calls */
> +#ifdef CONFIG_RELOCATABLE
> + ld r3, HSTATE_SCRATCH1(r13)
The problem is here. Depending on what sort of interrupt it is, we
can end up going back to the guest via fast_guest_return, which
doesn't reload CTR. Adding
mtctr r3
at this point makes things work.
Paul.
