On Mon, Jan 12, 2015 at 09:18:52AM +0000, Andre Przywara wrote:
> Hi,
>
> On 11/01/15 18:42, Christoffer Dall wrote:
> > On Thu, Jan 08, 2015 at 01:33:35PM -0600, Wei Huang wrote:
> >> On 01/07/2015 02:27 PM, Wei Huang wrote:
> >>> arm64 uses its own copy of exit handler (arm64/kvm/handle_exit.c).
> >>> Currently this file doesn't hook up with any trace points. As a result
> >>> users might not see certain events (e.g. HVC & WFI) while using ftrace
> >>> with arm64 KVM. This patch fixes this issue by adding a new trace file
> >>> for arm64 and linking new trace points with related functions.
> >>>
> >>> Signed-off-by: Wei Huang <wei@xxxxxxxxxx>
> >>> ---
> >>> arch/arm64/include/asm/kvm_arm.h | 2 ++
> >>> arch/arm64/include/asm/kvm_emulate.h | 5 ++++
> >>> arch/arm64/kvm/handle_exit.c | 8 ++++++
> >>> arch/arm64/kvm/trace.h | 52 ++++++++++++++++++++++++++++++++++++
> >>> 4 files changed, 67 insertions(+)
> >>> create mode 100644 arch/arm64/kvm/trace.h
> >>>
> >>> diff --git a/arch/arm64/include/asm/kvm_arm.h b/arch/arm64/include/asm/kvm_arm.h
> >>> index 7fd3e27..e3e2350 100644
> >>> --- a/arch/arm64/include/asm/kvm_arm.h
> >>> +++ b/arch/arm64/include/asm/kvm_arm.h
> >>> @@ -256,4 +256,6 @@
> >>>
> >>> #define ESR_EL2_EC_WFI_ISS_WFE (1 << 0)
> >>>
> >>> +#define ESR_EL2_HVC_IMM_MASK ((1UL << 16) - 1)
> >>> +
> >>> #endif /* __ARM64_KVM_ARM_H__ */
> >>> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
> >>> index 8127e45..a6fa2d2 100644
> >>> --- a/arch/arm64/include/asm/kvm_emulate.h
> >>> +++ b/arch/arm64/include/asm/kvm_emulate.h
> >>> @@ -126,6 +126,11 @@ static inline phys_addr_t kvm_vcpu_get_fault_ipa(const struct kvm_vcpu *vcpu)
> >>> return ((phys_addr_t)vcpu->arch.fault.hpfar_el2 & HPFAR_MASK) << 8;
> >>> }
> >>>
> >>> +static inline u32 kvm_vcpu_hvc_get_imm(const struct kvm_vcpu *vcpu)
> >>> +{
> >>> + return kvm_vcpu_get_hsr(vcpu) & ESR_EL2_HVC_IMM_MASK;
> >>> +}
> >>> +
> >>> static inline bool kvm_vcpu_dabt_isvalid(const struct kvm_vcpu *vcpu)
> >>> {
> >>> return !!(kvm_vcpu_get_hsr(vcpu) & ESR_EL2_ISV);
> >>> diff --git a/arch/arm64/kvm/handle_exit.c b/arch/arm64/kvm/handle_exit.c
> >>> index 34b8bd0..4fdb907 100644
> >>> --- a/arch/arm64/kvm/handle_exit.c
> >>> +++ b/arch/arm64/kvm/handle_exit.c
> >>> @@ -26,12 +26,18 @@
> >>> #include <asm/kvm_mmu.h>
> >>> #include <asm/kvm_psci.h>
> >>>
> >>> +#define CREATE_TRACE_POINTS
> >>> +#include "trace.h"
> >>> +
> >>> typedef int (*exit_handle_fn)(struct kvm_vcpu *, struct kvm_run *);
> >>>
> >>> static int handle_hvc(struct kvm_vcpu *vcpu, struct kvm_run *run)
> >>> {
> >>> int ret;
> >>>
> >>> + trace_kvm_hvc_arm64(*vcpu_pc(vcpu), *vcpu_reg(vcpu, 0),
> >>> + kvm_vcpu_hvc_get_imm(vcpu));
> >>> +
> >>> ret = kvm_psci_call(vcpu);
> >>> if (ret < 0) {
> >>> kvm_inject_undefined(vcpu);
> >>> @@ -61,6 +67,8 @@ static int handle_smc(struct kvm_vcpu *vcpu, struct kvm_run *run)
> >>> */
> >>> static int kvm_handle_wfx(struct kvm_vcpu *vcpu, struct kvm_run *run)
> >>> {
> >>> + trace_kvm_wfi_arm64(*vcpu_pc(vcpu));
> >>> +
> >>> if (kvm_vcpu_get_hsr(vcpu) & ESR_EL2_EC_WFI_ISS_WFE)
> >>> kvm_vcpu_on_spin(vcpu);
> >>> else
> >>> diff --git a/arch/arm64/kvm/trace.h b/arch/arm64/kvm/trace.h
> >>> new file mode 100644
> >>> index 0000000..96c7de0
> >>> --- /dev/null
> >>> +++ b/arch/arm64/kvm/trace.h
> >>> @@ -0,0 +1,52 @@
> >>> +#if !defined(_TRACE_ARM64_KVM_H) || defined(TRACE_HEADER_MULTI_READ)
> >>> +#define _TRACE_ARM64_KVM_H
> >>> +
> >>> +#include <linux/tracepoint.h>
> >>> +
> >>> +#undef TRACE_SYSTEM
> >>> +#define TRACE_SYSTEM kvm
> >>> +
> >>> +TRACE_EVENT(kvm_wfi_arm64,
> >>> + TP_PROTO(unsigned long vcpu_pc),
> >>> + TP_ARGS(vcpu_pc),
> >>> +
> >>> + TP_STRUCT__entry(
> >>> + __field(unsigned long, vcpu_pc)
> >>> + ),
> >>> +
> >>> + TP_fast_assign(
> >>> + __entry->vcpu_pc = vcpu_pc;
> >>> + ),
> >>> +
> >>> + TP_printk("guest executed wfxx at: 0x%08lx", __entry->vcpu_pc)
> >>
> >> I realized that there is a typo in TP_printk. "guest executed wfxx" is
> >> supposed to be "guest executed wfi". Sorry. That is a debug string
> >> accidentally left behind. Please correct it (or I can send another patch
> >> if requested).
> >>
> >
> > it really should all be _wfx, not wfi, and we should identify if it's a
> > wfi or wfe we're seeing, now when we're at it. (It could be good for
> > counting stats too).
>
> I have a patch for this in my queue. I will rebase it on top of Wei's
> patch and send it out ASAP.
>
huh? I was thinking Wei just fixes up his patch and sends it with the
required changes, not sure if we need to introduce more patches. But
whatever...
-Christoffer
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
