On Tue, Nov 25, 2014 at 04:10:05PM +0000, Alex Bennée wrote:
> This adds support for userspace to control the HW debug registers for
> guest debug. We'll only copy the $ARCH defined number across as that's
> all that hyp.S will use anyway. I've moved some helper functions into
> the hw_breakpoint.h header for re-use.
>
> As with single step we need to tweak the guest registers to enable the
> exceptions but we don't want to overwrite the guest copy of these
> registers so this is done close to the guest entry.
>
> Two new capabilities have been added to the KVM_EXTENSION ioctl to allow
> userspace to query the number of hardware break and watch points
> available on the host hardware.
>
> Signed-off-by: Alex Bennée <alex.bennee@xxxxxxxxxx>
>
> diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
> index 9383359..5e8c673 100644
> --- a/Documentation/virtual/kvm/api.txt
> +++ b/Documentation/virtual/kvm/api.txt
> @@ -2593,7 +2593,7 @@ The top 16 bits of the control field are architecture specific control
> flags which can include the following:
>
> - KVM_GUESTDBG_USE_SW_BP: using software breakpoints [x86, arm64]
> - - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390]
> + - KVM_GUESTDBG_USE_HW_BP: using hardware breakpoints [x86, s390, arm64]
> - KVM_GUESTDBG_INJECT_DB: inject DB type exception [x86]
> - KVM_GUESTDBG_INJECT_BP: inject BP type exception [x86]
> - KVM_GUESTDBG_EXIT_PENDING: trigger an immediate guest exit [s390]
> @@ -2606,7 +2606,10 @@ we need to ensure the guest vCPUs architecture specific registers are
> updated to the correct (supplied) values.
>
> The second part of the structure is architecture specific and
> -typically contains a set of debug registers.
> +typically contains a set of debug registers. For arm64 the number of
new paragraph for the arch-specific stuff.
> +debug registers is implementation defined and can be determined by
> +querying the KVM_CAP_GUEST_DEBUG_HW_BPS and KVM_CAP_GUEST_DEBUG_HW_WPS
> +capabilities.
>
> When debug events exit the main run loop with the reason
> KVM_EXIT_DEBUG with the kvm_debug_exit_arch part of the kvm_run
> diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
> index a76daae..c8ec23a 100644
> --- a/arch/arm/kvm/arm.c
> +++ b/arch/arm/kvm/arm.c
> @@ -39,6 +39,7 @@
> #include <asm/cacheflush.h>
> #include <asm/virt.h>
> #include <asm/debug-monitors.h>
> +#include <asm/hw_breakpoint.h>
> #include <asm/kvm_arm.h>
> #include <asm/kvm_asm.h>
> #include <asm/kvm_mmu.h>
> @@ -341,8 +342,37 @@ int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
>
> /* Hardware assisted Break and Watch points */
> if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) {
> - kvm_info("HW BP support requested, not yet implemented\n");
> - return -EINVAL;
> + int i;
> + int nb = get_num_brps();
> + int nw = get_num_wrps();
> +
> + /* Copy across up to IMPDEF debug registers to our
coding style
> + * shadow copy in the vcpu structure. The hyp.S code
> + * will then set them up before we re-enter the guest.
> + */
> + memcpy(vcpu->arch.guest_debug_regs.dbg_bcr,
> + dbg->arch.dbg_bcr, sizeof(__u64)*nb);
> + memcpy(vcpu->arch.guest_debug_regs.dbg_bvr,
> + dbg->arch.dbg_bvr, sizeof(__u64)*nb);
> + memcpy(vcpu->arch.guest_debug_regs.dbg_wcr,
> + dbg->arch.dbg_wcr, sizeof(__u64)*nw);
> + memcpy(vcpu->arch.guest_debug_regs.dbg_wvr,
> + dbg->arch.dbg_wvr, sizeof(__u64)*nw);
> +
> + kvm_info("HW BP support requested\n");
> + for (i = 0; i < nb; i++) {
> + kvm_info("%d: dbg_bcr=0x%llx dbg_bvr=0x%llx\n",
> + i,
> + vcpu->arch.guest_debug_regs.dbg_bcr[i],
> + vcpu->arch.guest_debug_regs.dbg_bvr[i]);
> + }
> + for (i = 0; i < nw; i++) {
> + kvm_info("%d: dbg_wcr=0x%llx dbg_wvr=0x%llx\n",
> + i,
> + vcpu->arch.guest_debug_regs.dbg_wcr[i],
> + vcpu->arch.guest_debug_regs.dbg_wvr[i]);
> + }
> + route_el2 = true;
> }
>
> /* If we are going to handle any debug exceptions we need to
> diff --git a/arch/arm64/include/asm/hw_breakpoint.h b/arch/arm64/include/asm/hw_breakpoint.h
> index 52b484b..c450552 100644
> --- a/arch/arm64/include/asm/hw_breakpoint.h
> +++ b/arch/arm64/include/asm/hw_breakpoint.h
> @@ -130,6 +130,18 @@ static inline void ptrace_hw_copy_thread(struct task_struct *task)
> }
> #endif
>
> +/* Determine number of BRP registers available. */
> +static inline int get_num_brps(void)
> +{
> + return ((read_cpuid(ID_AA64DFR0_EL1) >> 12) & 0xf) + 1;
> +}
> +
> +/* Determine number of WRP registers available. */
> +static inline int get_num_wrps(void)
> +{
> + return ((read_cpuid(ID_AA64DFR0_EL1) >> 20) & 0xf) + 1;
> +}
> +
> extern struct pmu perf_ops_bp;
>
> #endif /* __KERNEL__ */
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index 38b0f07..e386bf4 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -103,8 +103,9 @@ struct kvm_vcpu_arch {
> /* Exception Information */
> struct kvm_vcpu_fault_info fault;
>
> - /* Debug state */
> + /* Guest debug state */
> u64 debug_flags;
> + struct kvm_guest_debug_arch guest_debug_regs;
>
> /* Pointer to host CPU context */
> kvm_cpu_context_t *host_cpu_context;
> diff --git a/arch/arm64/kernel/asm-offsets.c b/arch/arm64/kernel/asm-offsets.c
> index 78e5ae1..c9ecfd3 100644
> --- a/arch/arm64/kernel/asm-offsets.c
> +++ b/arch/arm64/kernel/asm-offsets.c
> @@ -122,6 +122,10 @@ int main(void)
> DEFINE(VCPU_HPFAR_EL2, offsetof(struct kvm_vcpu, arch.fault.hpfar_el2));
> DEFINE(VCPU_DEBUG_FLAGS, offsetof(struct kvm_vcpu, arch.debug_flags));
> DEFINE(GUEST_DEBUG, offsetof(struct kvm_vcpu, guest_debug));
> + DEFINE(GUEST_DEBUG_BCR, offsetof(struct kvm_vcpu, arch.guest_debug_regs.dbg_bcr));
> + DEFINE(GUEST_DEBUG_BVR, offsetof(struct kvm_vcpu, arch.guest_debug_regs.dbg_bvr));
> + DEFINE(GUEST_DEBUG_WCR, offsetof(struct kvm_vcpu, arch.guest_debug_regs.dbg_wcr));
> + DEFINE(GUEST_DEBUG_WVR, offsetof(struct kvm_vcpu, arch.guest_debug_regs.dbg_wvr));
> DEFINE(VCPU_HCR_EL2, offsetof(struct kvm_vcpu, arch.hcr_el2));
> DEFINE(VCPU_MDCR_EL2, offsetof(struct kvm_vcpu, arch.mdcr_el2));
> DEFINE(VCPU_IRQ_LINES, offsetof(struct kvm_vcpu, arch.irq_lines));
> diff --git a/arch/arm64/kernel/hw_breakpoint.c b/arch/arm64/kernel/hw_breakpoint.c
> index df1cf15..45dcc6f 100644
> --- a/arch/arm64/kernel/hw_breakpoint.c
> +++ b/arch/arm64/kernel/hw_breakpoint.c
> @@ -49,18 +49,6 @@ static DEFINE_PER_CPU(int, stepping_kernel_bp);
> static int core_num_brps;
> static int core_num_wrps;
>
> -/* Determine number of BRP registers available. */
> -static int get_num_brps(void)
> -{
> - return ((read_cpuid(ID_AA64DFR0_EL1) >> 12) & 0xf) + 1;
> -}
> -
> -/* Determine number of WRP registers available. */
> -static int get_num_wrps(void)
> -{
> - return ((read_cpuid(ID_AA64DFR0_EL1) >> 20) & 0xf) + 1;
> -}
> -
> int hw_breakpoint_slots(int type)
> {
> /*
> diff --git a/arch/arm64/kvm/handle_exit.c b/arch/arm64/kvm/handle_exit.c
> index 6def054..d024e47 100644
> --- a/arch/arm64/kvm/handle_exit.c
> +++ b/arch/arm64/kvm/handle_exit.c
> @@ -110,6 +110,42 @@ static int kvm_handle_ss(struct kvm_vcpu *vcpu, struct kvm_run *run)
> return 0;
> }
>
> +/**
> + * kvm_handle_hw_bp - handle HW assisted break point
> + *
> + * @vcpu: the vcpu pointer
> + *
> + */
> +static int kvm_handle_hw_bp(struct kvm_vcpu *vcpu, struct kvm_run *run)
> +{
> + WARN_ON(!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP));
> +
> + run->exit_reason = KVM_EXIT_DEBUG;
> + run->debug.arch.exit_type = KVM_DEBUG_EXIT_HW_BKPT;
> + run->debug.arch.address = *vcpu_pc(vcpu);
> + return 0;
> +}
> +
> +/**
> + * kvm_handle_watch - handle HW assisted watch point
> + *
> + * @vcpu: the vcpu pointer
> + *
> + * These are basically the same as breakpoints (and indeed may use the
> + * breakpoint in a linked fashion). However they generate a specific
> + * exception so we trap it here for reporting to the guest.
> + *
> + */
> +static int kvm_handle_watch(struct kvm_vcpu *vcpu, struct kvm_run *run)
> +{
> + WARN_ON(!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP));
> +
> + run->exit_reason = KVM_EXIT_DEBUG;
> + run->debug.arch.exit_type = KVM_DEBUG_EXIT_HW_WTPT;
> + run->debug.arch.address = *vcpu_pc(vcpu);
> + return 0;
> +}
the usual comments about the @run documentation and let's get rid of all
these WARN_ON() statements.
> +
> static exit_handle_fn arm_exit_handlers[] = {
> [ESR_EL2_EC_WFI] = kvm_handle_wfx,
> [ESR_EL2_EC_CP15_32] = kvm_handle_cp15_32,
> @@ -125,6 +161,8 @@ static exit_handle_fn arm_exit_handlers[] = {
> [ESR_EL2_EC_IABT] = kvm_handle_guest_abort,
> [ESR_EL2_EC_DABT] = kvm_handle_guest_abort,
> [ESR_EL2_EC_SOFTSTP] = kvm_handle_ss,
> + [ESR_EL2_EC_WATCHPT] = kvm_handle_watch,
> + [ESR_EL2_EC_BREAKPT] = kvm_handle_hw_bp,
> [ESR_EL2_EC_BKPT32] = kvm_handle_bkpt,
> [ESR_EL2_EC_BRK64] = kvm_handle_bkpt,
> };
> diff --git a/arch/arm64/kvm/hyp.S b/arch/arm64/kvm/hyp.S
> index b38ce3d..96f71ab 100644
> --- a/arch/arm64/kvm/hyp.S
> +++ b/arch/arm64/kvm/hyp.S
> @@ -18,6 +18,7 @@
> #include <linux/linkage.h>
> #include <linux/kvm.h>
>
> +#include <uapi/asm/kvm.h>
> #include <asm/assembler.h>
> #include <asm/memory.h>
> #include <asm/asm-offsets.h>
> @@ -174,6 +175,7 @@
> ldr x3, [x0, #GUEST_DEBUG]
> tbz x3, #KVM_GUESTDBG_ENABLE_SHIFT, 2f // No guest debug
>
> + // Both Step and HW BP/WP ops need to modify spsr_el2 and mdscr_el1
> // x0 - preserved as VCPU ptr
> // x1 - spsr
> // x2 - mdscr
> @@ -191,6 +193,11 @@
> eor x1, x1, #DBG_SPSR_SS
> eor x2, x2, #DBG_MDSCR_SS
> 1:
> + // If we are doing HW BP/WP - set MDSCR_EL1.KDE/MDE
> + tbz x3, #KVM_GUESTDBG_USE_HW_BP_SHIFT, 3f
> + orr x2, x2, #DBG_MDSCR_KDE
> + orr x2, x2, #DBG_MDSCR_MDE
> +3:
> msr spsr_el2, x1
> msr mdscr_el1, x2
> 2:
> @@ -815,6 +822,33 @@ __restore_debug:
>
> ret
>
> +/* Setup debug state for debug of guest */
> +__setup_debug:
> + // x0: vcpu base address
> + // x3: ptr to guest registers passed to setup_debug_registers
> + // x5..x20/x26: trashed
> +
> + mrs x26, id_aa64dfr0_el1
> + ubfx x24, x26, #12, #4 // Extract BRPs
> + ubfx x25, x26, #20, #4 // Extract WRPs
> + mov w26, #15
> + sub w24, w26, w24 // How many BPs to skip
> + sub w25, w26, w25 // How many WPs to skip
> +
> + mov x4, x24
> + add x3, x0, #GUEST_DEBUG_BCR
> + setup_debug_registers dbgbcr
> + add x3, x0, #GUEST_DEBUG_BVR
> + setup_debug_registers dbgbvr
> +
> + mov x4, x25
> + add x3, x0, #GUEST_DEBUG_WCR
> + setup_debug_registers dbgwcr
> + add x3, x0, #GUEST_DEBUG_WVR
> + setup_debug_registers dbgwvr
> +
> + ret
> +
> __save_fpsimd:
> save_fpsimd
> ret
> @@ -861,6 +895,13 @@ ENTRY(__kvm_vcpu_run)
> bl __restore_sysregs
> bl __restore_fpsimd
>
> + // Now is the time to set-up the debug registers if we
> + // are debugging the guest
> + ldr x3, [x0, #GUEST_DEBUG]
> + tbz x3, #KVM_GUESTDBG_USE_HW_BP_SHIFT, 2f
> + bl __setup_debug
> + b 1f
> +2:
> skip_debug_state x3, 1f
> bl __restore_debug
> 1:
> @@ -881,6 +922,11 @@ __kvm_vcpu_return:
> bl __save_fpsimd
> bl __save_sysregs
>
> + // If we are debugging the guest don't save debug registers
> + // otherwise we'll be trashing are only good copy we have.
> + ldr x3, [x0, #GUEST_DEBUG]
> + tbnz x3, #KVM_GUESTDBG_USE_HW_BP_SHIFT, 1f
> +
we're introducing an awful lot of conditionals in the assembly code with
these patches, can you re-consider if there's a cleaner abstraction that
allows us to deal with some of this stuff in C-code?
-Christoffer
--
To unsubscribe from this list: send the line "unsubscribe linux-api" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
