On Fri, May 29, 2015 at 10:30:23AM +0100, Alex Bennée wrote:
> This adds support for single-stepping the guest. To do this we need to
> manipulate the guests PSTATE.SS and MDSCR_EL1.SS bits which we do in the
> kvm_arm_setup/clear_debug() so we don't affect the apparent state of the
> guest. Additionally while the host is debugging the guest we suppress
> the ability of the guest to single-step itself.
I feel like there should be a slightly more elaborate explanation of
exactly what works and what doesn't work when the guest is single
stepping something and which choices we've made for supporting or not
supporting this.
>
> Signed-off-by: Alex Bennée <alex.bennee@xxxxxxxxxx>
>
> ---
> v2
> - Move pstate/mdscr manipulation into C
> - don't export guest_debug to assembly
> - add accessor for saved_debug regs
> - tweak save/restore of mdscr_el1
> v3
> - don't save PC in debug information struct
> - rename debug_saved_regs->guest_debug_state
> - save whole value, only use bits in restore
> - add save/restore_guest-debug_regs helper functions
> - simplify commit message for clarity
> - rm vcpu_debug_saved_reg access fn
> v4
> - added more comments based on suggestions
> - guest_debug_state->guest_debug_preserved
> - no point masking restore, we will trap out
> v5
> - more comments
> - don't bother preserving pstate.ss
it would have been good if there was some comment explaining the reason
for this change.
> ---
> arch/arm/kvm/arm.c | 4 ++-
> arch/arm64/include/asm/kvm_host.h | 11 ++++++++
> arch/arm64/kvm/debug.c | 58 ++++++++++++++++++++++++++++++++++++---
> arch/arm64/kvm/handle_exit.c | 2 ++
> 4 files changed, 70 insertions(+), 5 deletions(-)
>
> diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
> index 064c105..9b3ed6d 100644
> --- a/arch/arm/kvm/arm.c
> +++ b/arch/arm/kvm/arm.c
> @@ -302,7 +302,9 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
> kvm_arm_set_running_vcpu(NULL);
> }
>
> -#define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)
> +#define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \
> + KVM_GUESTDBG_USE_SW_BP | \
> + KVM_GUESTDBG_SINGLESTEP)
>
> /**
> * kvm_arch_vcpu_ioctl_set_guest_debug - set up guest debugging
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index 7cb99b5..e2db6a6 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -123,6 +123,17 @@ struct kvm_vcpu_arch {
> * here.
> */
>
> + /*
> + * Guest registers we preserve during guest debugging.
> + *
> + * These shadow registers are updated by the kvm_handle_sys_reg
> + * trap handler if the guest accesses or updates them while we
> + * are using guest debug.
> + */
> + struct {
> + u32 mdscr_el1;
> + } guest_debug_preserved;
> +
> /* Don't run the guest */
> bool pause;
>
> diff --git a/arch/arm64/kvm/debug.c b/arch/arm64/kvm/debug.c
> index 8d1bfa4..10a6baa 100644
> --- a/arch/arm64/kvm/debug.c
> +++ b/arch/arm64/kvm/debug.c
> @@ -19,11 +19,41 @@
>
> #include <linux/kvm_host.h>
>
> +#include <asm/debug-monitors.h>
> +#include <asm/kvm_asm.h>
> #include <asm/kvm_arm.h>
> +#include <asm/kvm_emulate.h>
> +
> +/* These are the bits of MDSCR_EL1 we may manipulate */
> +#define MDSCR_EL1_DEBUG_MASK (DBG_MDSCR_SS | \
> + DBG_MDSCR_KDE | \
> + DBG_MDSCR_MDE)
>
> static DEFINE_PER_CPU(u32, mdcr_el2);
>
> /**
> + * save/restore_guest_debug_regs
> + *
> + * For some debug operations we need to tweak some guest registers. As
> + * a result we need to save the state of those registers before we
> + * make those modifications. This does get confused if the guest
> + * attempts to control single step while being debugged. It will start
> + * working again once it is no longer being debugged by the host.
What gets confused and what starts working?
> + *
> + * Guest access to MDSCR_EL1 is trapped by the hypervisor and handled
> + * after we have restored the preserved value to the main context.
> + */
> +static void save_guest_debug_regs(struct kvm_vcpu *vcpu)
> +{
> + vcpu->arch.guest_debug_preserved.mdscr_el1 = vcpu_sys_reg(vcpu, MDSCR_EL1);
> +}
> +
> +static void restore_guest_debug_regs(struct kvm_vcpu *vcpu)
> +{
> + vcpu_sys_reg(vcpu, MDSCR_EL1) = vcpu->arch.guest_debug_preserved.mdscr_el1;
> +}
> +
> +/**
> * kvm_arm_init_debug - grab what we need for debug
> *
> * Currently the sole task of this function is to retrieve the initial
> @@ -38,7 +68,6 @@ void kvm_arm_init_debug(void)
> __this_cpu_write(mdcr_el2, kvm_call_hyp(__kvm_get_mdcr_el2));
> }
>
> -
> /**
> * kvm_arm_setup_debug - set up debug related stuff
> *
> @@ -73,12 +102,33 @@ void kvm_arm_setup_debug(struct kvm_vcpu *vcpu)
> if (trap_debug)
> vcpu->arch.mdcr_el2 |= MDCR_EL2_TDA;
>
> - /* Trap breakpoints? */
> - if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP)
> + /* Is Guest debugging in effect? */
> + if (vcpu->guest_debug) {
> vcpu->arch.mdcr_el2 |= MDCR_EL2_TDE;
> +
> + /* Save guest debug state */
> + save_guest_debug_regs(vcpu);
> +
> + /*
> + * Single Step (ARM ARM D2.12.3 The software step state
> + * machine)
> + *
> + * If we are doing Single Step we need to manipulate
> + * MDSCR_EL1.SS and PSTATE.SS. If not we need to
> + * suppress the guests ability to trigger single step
> + * exceptions as otherwise the host will deal with them.
is this because if you are doing any kind of guest debugging, we trap
all debug exceptions to EL2 and therefore single-stepping in the guest
won't work anyway and the host doesn't know what to do with such
exceptions?
I would feel slightly better if the comment assured me this doesn't
outright break something, but ok...
> + */
> + if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) {
> + *vcpu_cpsr(vcpu) |= DBG_SPSR_SS;
> + vcpu_sys_reg(vcpu, MDSCR_EL1) |= DBG_MDSCR_SS;
> + } else {
> + vcpu_sys_reg(vcpu, MDSCR_EL1) &= ~DBG_MDSCR_SS;
> + }
> + }
> }
>
> void kvm_arm_clear_debug(struct kvm_vcpu *vcpu)
> {
> - /* Nothing to do yet */
> + if (vcpu->guest_debug)
> + restore_guest_debug_regs(vcpu);
> }
> diff --git a/arch/arm64/kvm/handle_exit.c b/arch/arm64/kvm/handle_exit.c
> index 27f38a9..e9de13e 100644
> --- a/arch/arm64/kvm/handle_exit.c
> +++ b/arch/arm64/kvm/handle_exit.c
> @@ -103,6 +103,7 @@ static int kvm_handle_guest_debug(struct kvm_vcpu *vcpu, struct kvm_run *run)
> run->debug.arch.hsr = hsr;
>
> switch (hsr >> ESR_ELx_EC_SHIFT) {
> + case ESR_ELx_EC_SOFTSTP_LOW:
> case ESR_ELx_EC_BKPT32:
> case ESR_ELx_EC_BRK64:
> break;
> @@ -130,6 +131,7 @@ static exit_handle_fn arm_exit_handlers[] = {
> [ESR_ELx_EC_SYS64] = kvm_handle_sys_reg,
> [ESR_ELx_EC_IABT_LOW] = kvm_handle_guest_abort,
> [ESR_ELx_EC_DABT_LOW] = kvm_handle_guest_abort,
> + [ESR_ELx_EC_SOFTSTP_LOW]= kvm_handle_guest_debug,
> [ESR_ELx_EC_BKPT32] = kvm_handle_guest_debug,
> [ESR_ELx_EC_BRK64] = kvm_handle_guest_debug,
> };
> --
> 2.4.1
>
As for the code of this patch:
Reviewed-by: Christoffer Dall <christoffer.dall@xxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
