Re: [PATCH v4 2/2] target/arm: kvm: Handle potential issue with dabt injection

[Beata Michalska <beata.michalska@xxxxxxxxxx>]

On Fri, 3 Apr 2020 at 09:44, Andrew Jones <drjones@xxxxxxxxxx> wrote:
>

> On Mon, Mar 23, 2020 at 11:32:27AM +0000, Beata Michalska wrote:
> > Injecting external data abort through KVM might trigger
> > an issue on kernels that do not get updated to include the KVM fix.
> > For those and aarch32 guests, the injected abort gets misconfigured
> > to be an implementation defined exception. This leads to the guest
> > repeatedly re-running the faulting instruction.
> >
> > Add support for handling that case.
> > [
> >   Fixed-by: 018f22f95e8a
> >       ('KVM: arm: Fix DFSR setting for non-LPAE aarch32 guests')
> >   Fixed-by: 21aecdbd7f3a
> >       ('KVM: arm: Make inject_abt32() inject an external abort instead')
> > ]
> >
> > Signed-off-by: Beata Michalska <beata.michalska@xxxxxxxxxx>
> > ---
> >  target/arm/cpu.h     |  1 +
> >  target/arm/kvm.c     | 30 +++++++++++++++++++++++++++++-
> >  target/arm/kvm32.c   | 25 +++++++++++++++++++++++++
> >  target/arm/kvm64.c   | 34 ++++++++++++++++++++++++++++++++++
> >  target/arm/kvm_arm.h | 10 ++++++++++
> >  5 files changed, 99 insertions(+), 1 deletion(-)
> >
> > diff --git a/target/arm/cpu.h b/target/arm/cpu.h
> > index 4f834c1..868afc6 100644
> > --- a/target/arm/cpu.h
> > +++ b/target/arm/cpu.h
> > @@ -561,6 +561,7 @@ typedef struct CPUARMState {
> >      } serror;
> >
> >      uint8_t ext_dabt_pending; /* Request for injecting ext DABT */
> > +    uint8_t ext_dabt_raised; /* Tracking/verifying injection of ext DABT */
> >
> >      /* State of our input IRQ/FIQ/VIRQ/VFIQ lines */
> >      uint32_t irq_line_state;
> > diff --git a/target/arm/kvm.c b/target/arm/kvm.c
> > index c088589..58ad734 100644
> > --- a/target/arm/kvm.c
> > +++ b/target/arm/kvm.c
> > @@ -721,7 +721,12 @@ int kvm_put_vcpu_events(ARMCPU *cpu)
> >      ret = kvm_vcpu_ioctl(CPU(cpu), KVM_SET_VCPU_EVENTS, &events);
> >      if (ret) {
> >          error_report("failed to put vcpu events");
> > -    } else {
> > +    } else if (env->ext_dabt_pending) {
> > +        /*
> > +         * Mark that the external DABT has been injected,
> > +         * if one has been requested
> > +         */
> > +        env->ext_dabt_raised = env->ext_dabt_pending;
> >          /* Clear instantly if the call was successful */
> >          env->ext_dabt_pending = 0;
> >      }
> > @@ -755,6 +760,29 @@ int kvm_get_vcpu_events(ARMCPU *cpu)
> >
> >  void kvm_arch_pre_run(CPUState *cs, struct kvm_run *run)
> >  {
> > +    ARMCPU *cpu = ARM_CPU(cs);
> > +    CPUARMState *env = &cpu->env;
> > +
> > +    if (unlikely(env->ext_dabt_raised)) {
> > +        /*
> > +         * Verifying that the ext DABT has been properly injected,
> > +         * otherwise risking indefinitely re-running the faulting instruction
> > +         * Covering a very narrow case for kernels 5.5..5.5.4
> > +         * when injected abort was misconfigured to be
> > +         * an IMPLEMENTATION DEFINED exception (for 32-bit EL1)
> > +         */
> > +        if (!arm_feature(env, ARM_FEATURE_AARCH64) &&
> > +            unlikely(!kvm_arm_verify_ext_dabt_pending(cs))) {
> > +
> > +            error_report("Data abort exception with no valid ISS generated by "
> > +                   "guest memory access. KVM unable to emulate faulting "
> > +                   "instruction. Failed to inject an external data abort "
> > +                   "into the guest.");
> > +            abort();
> > +       }
> > +       /* Clear the status */
> > +       env->ext_dabt_raised = 0;
> > +    }
> >  }
> >
> >  MemTxAttrs kvm_arch_post_run(CPUState *cs, struct kvm_run *run)
> > diff --git a/target/arm/kvm32.c b/target/arm/kvm32.c
> > index f271181..86c4fe7 100644
> > --- a/target/arm/kvm32.c
> > +++ b/target/arm/kvm32.c
> > @@ -564,3 +564,28 @@ void kvm_arm_pmu_init(CPUState *cs)
> >  {
> >      qemu_log_mask(LOG_UNIMP, "%s: not implemented\n", __func__);
> >  }
> > +
> > +#define ARM_REG_DFSR  ARM_CP15_REG32(0, 5, 0, 0)
> > +#define ARM_REG_TTBCR ARM_CP15_REG32(0, 2, 0, 2)
> > +
> > +#define DFSR_FSC(v)   (((v) >> 6 | (v)) & 0x1F)
> > +#define DFSC_EXTABT(lpae) (lpae) ? 0x10 : 0x08
>
> We should put () around the whole ?: _expression_ when it's in a macro
>
> > +
> > +bool kvm_arm_verify_ext_dabt_pending(CPUState *cs)
> > +{
> > +    uint32_t dfsr_val;
> > +
> > +    if (!kvm_get_one_reg(cs, ARM_REG_DFSR, &dfsr_val)) {
> > +        ARMCPU *cpu = ARM_CPU(cs);
> > +        CPUARMState *env = &cpu->env;
> > +        uint32_t ttbcr;
> > +        int lpae = 0;
> > +
> > +        if (!kvm_get_one_reg(cs, ARM_REG_TTBCR, &ttbcr)) {
> > +            lpae = arm_feature(env, ARM_FEATURE_LPAE) && (ttbcr & TTBCR_EAE);
> > +        }
> > +        return !(DFSR_FSC(dfsr_val) != DFSC_EXTABT(lpae));
>
>  !(a != b) is a convoluted way to write a == b
>
> > +    }
> > +    return false;
> > +}
> > +
> > diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c
> > index be5b31c..18594e9 100644
> > --- a/target/arm/kvm64.c
> > +++ b/target/arm/kvm64.c
> > @@ -1430,3 +1430,37 @@ bool kvm_arm_handle_debug(CPUState *cs, struct kvm_debug_exit_arch *debug_exit)
> >
> >      return false;
> >  }
> > +
> > +#define ARM64_REG_ESR_EL1 ARM64_SYS_REG(3, 0, 5, 2, 0)
> > +#define ARM64_REG_TCR_EL1 ARM64_SYS_REG(3, 0, 2, 0, 2)
> > +
> > +#define ESR_DFSC(aarch64, v)    \
> > +    ((aarch64) ? ((v) & 0x3F)   \
> > +               : (((v) >> 6 | (v)) & 0x1F))
> > +
> > +#define ESR_DFSC_EXTABT(aarch64, lpae) \
> > +    ((aarch64) ? 0x10 : (lpae) ? 0x10 : 0x8)
> > +
> > +bool kvm_arm_verify_ext_dabt_pending(CPUState *cs)
> > +{
> > +    uint64_t dfsr_val;
> > +
> > +    if (!kvm_get_one_reg(cs, ARM64_REG_ESR_EL1, &dfsr_val)) {
> > +        ARMCPU *cpu = ARM_CPU(cs);
> > +        CPUARMState *env = &cpu->env;
> > +        int aarch64_mode = arm_feature(env, ARM_FEATURE_AARCH64);
> > +        int lpae = 0;
> > +
> > +        if (!aarch64_mode) {
> > +            uint64_t ttbcr;
> > +
> > +            if (!kvm_get_one_reg(cs, ARM64_REG_TCR_EL1, &ttbcr)) {
> > +                lpae = arm_feature(env, ARM_FEATURE_LPAE)
> > +                        && (ttbcr & TTBCR_EAE);
> > +            }
> > +        }
> > +        return !(ESR_DFSC(aarch64_mode, dfsr_val) !=
> > +                ESR_DFSC_EXTABT(aarch64_mode, lpae));
>
> a == b, please
>
> > +    }
> > +    return false;
> > +}
> > diff --git a/target/arm/kvm_arm.h b/target/arm/kvm_arm.h
> > index 39472d5..f2dc6a2 100644
> > --- a/target/arm/kvm_arm.h
> > +++ b/target/arm/kvm_arm.h
> > @@ -461,6 +461,16 @@ void kvm_arm_copy_hw_debug_data(struct kvm_guest_debug_arch *ptr);
> >  int kvm_arm_handle_dabt_nisv(CPUState *cs, uint64_t esr_iss,
> >                               uint64_t fault_ipa);
> >  /**
> > + * kvm_arm_verify_ext_dabt_pending:
> > + * @cs: CPUState
> > + *
> > + * Verify the fault status code wrt the Ext DABT injection
> > + *
> > + * Returns: true if the fault status code is as expected, false otherwise
> > + */
> > +bool kvm_arm_verify_ext_dabt_pending(CPUState *cs);
> > +
> > +/**
> >   * its_class_name:
> >   *
> >   * Return the ITS class name to use depending on whether KVM acceleration
> > --
> > 2.7.4
> >
> >
>
> I'll leave the decision to take this KVM bug workaround patch at all to Peter,
> and I didn't actually review whether or not kvm_arm_verify_ext_dabt_pending
> is doing what it claims it's doing, so I'm reluctant to give an r-b on
> this patch. But, as far as the code goes, besides the comments above, it
> looks fine to me.
>

Thanks for the feedback.
Will apply the changes for the next version.

BR
Beata
> Thanks,
> drew
>

_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm