On 12/05/2019 00:08, Mark Cave-Ayland wrote:
> Hi all,
>
> Whilst trying to investigate some issues with MacOS under KVM PR I noticed that when
> setting software breakpoints the KVM VCPU would stop as requested, but QEMU's gdbstub
> would hang indefinitely.
What are you trying to do exactly? Just breakpoints or single stepping?
Anyway, I am cc-ing Fabiano who is fixing single stepping and knows this
code well.
> I eventually traced it down to this code in QEMU's target/ppc/kvm.c:
>
>
> static int kvm_handle_debug(PowerPCCPU *cpu, struct kvm_run *run)
> {
> CPUState *cs = CPU(cpu);
> CPUPPCState *env = &cpu->env;
> struct kvm_debug_exit_arch *arch_info = &run->debug.arch;
>
> if (cs->singlestep_enabled) {
> return kvm_handle_singlestep();
> }
>
> if (arch_info->status) {
> return kvm_handle_hw_breakpoint(cs, arch_info);
> }
>
> if (kvm_find_sw_breakpoint(cs, arch_info->address)) {
> return kvm_handle_sw_breakpoint();
> }
>
>
> The problem here is that with Book3S PR on my Mac hardware, run->debug.arch.status !=
> 0 which causes QEMU to think that this is a hardware breakpoint and so the software
> breakpoint doesn't get handled correctly.
>
> For comparison both booke.c and e500_emulate.c set debug.arch.status = 0 for software
> breakpoints, whereas both book3s_hv.c and book3s_pr.c do not. Given that emulate.c
> contains shared code for handling software breakpoints, would the following simple
> patch suffice?
>
>
> diff --git a/arch/powerpc/kvm/emulate.c b/arch/powerpc/kvm/emulate.c
> index 9f5b8c01c4e1..e77becaad5dd 100644
> --- a/arch/powerpc/kvm/emulate.c
> +++ b/arch/powerpc/kvm/emulate.c
> @@ -282,6 +282,7 @@ int kvmppc_emulate_instruction(struct kvm_run *run, struct
> kvm_vcpu *vcpu)
> */
> if (inst == KVMPPC_INST_SW_BREAKPOINT) {
> run->exit_reason = KVM_EXIT_DEBUG;
> + run->debug.arch.status = 0;
> run->debug.arch.address = kvmppc_get_pc(vcpu);
> emulated = EMULATE_EXIT_USER;
> advance = 0;
>
>
> ATB,
>
> Mark.
>
--
Alexey
