[PATCH 0/3] accel/mshv: add gdbstub guest debugging support

[PATCH 0/3] accel/mshv: add gdbstub guest debugging support

[Doru Blânzeanu]

This series brings guest debugging to the MSHV accelerator by
implementing software breakpoints and single-stepping, so that gdb can
attach to an MSHV guest through QEMU's gdbstub.

It follows the KVM gdbstub implementation closely, and borrows the WHPX
approach where the MSHV interface differs from KVM:

  - it installs a partition-wide intercept for the #DB exception
    (HVCALL_INSTALL_INTERCEPT). #BP (INT3) is deliberately left to the
    guest's own IDT and is never intercepted, as WHPX does.

  - only software breakpoints are supported. A breakpoint is an INT1
    (opcode 0xf1) patched into guest memory, which raises a #DB when
    executed. Hardware breakpoints and watchpoints are not supported.

  - it never touches the guest debug registers (DR0-DR7), so guest state
    is not clobbered. The only guest register used is RFLAGS.TF, toggled
    around the vCPU run for single-stepping and kept out of env->eflags
    so it cannot be read back and re-armed.

  - it re-injects the #DB back into the guest (via
    HV_REGISTER_PENDING_EVENT0) when the faulting RIP does not match a
    debugger breakpoint, so a guest-owned #DB is delivered through the
    guest's own IDT.

When a vCPU is created we register the accel ops that gdb drives once it
is attached:

  - mshv_update_guest_debug
  - mshv_insert_gdbstub_breakpoint
  - mshv_remove_gdbstub_breakpoint
  - mshv_remove_all_gdbstub_breakpoints

For each insert request from gdb the ops save the original byte at the
target address and patch in 0xf1; on a remove request the saved byte is
restored, after which gdb single-steps over the original instruction and
re-inserts the breakpoint. When the vCPU executes the 0xf1 byte a #DB
vmexits to QEMU. mshv_handle_debug() then checks whether the RIP where
the vCPU stopped matches a breakpoint installed by gdb, or whether
single-stepping is active, and if so reports the stop to gdb. If the RIP
does not match any debugger breakpoint the #DB is assumed to be
guest-owned and is re-injected so the guest can handle it.

Known limitations:

  - Once installed, the #DB intercept stays for the lifetime
    of the VM and every #DB - including those the guest raises itself -
    vmexits to QEMU.

  - Because attribution relies on matching the faulting RIP against the
    debugger's breakpoint list, a guest that legitimately executes INT1 or
    raises its own #DB at an address that also holds a debugger breakpoint
    cannot be disambiguated; such overlaps are not expected in practice.

The series is organised as:

  1. include/hw/hyperv: add ABI for exception intercepts and pending
     events
  2. accel/mshv: add gdbstub software breakpoint support
  3. target/i386/mshv: support single-stepping

I have tested this patch series by using both a guest that doesn't use debugging
and one that triggers the #DB and #BP exceptions (#DB is re-injected and #BP is
not intercepted at all).

Doru Blânzeanu (3):
  include/hw/hyperv: add ABI for exception intercepts and pending events
  accel/mshv: add gdbstub software breakpoint support
  target/i386/mshv: support single-stepping

 accel/mshv/mshv-all.c          | 165 ++++++++++++++++++++++++++++++++-
 include/hw/hyperv/hvgdk_mini.h |  56 +++++++++++
 include/system/mshv_int.h      |  14 +++
 target/i386/mshv/mshv-cpu.c    | 136 +++++++++++++++++++++++++++
 4 files changed, 370 insertions(+), 1 deletion(-)

-- 
2.53.0

[PATCH 1/3] include/hw/hyperv: add ABI for exception intercepts and pending events

[Doru Blânzeanu]

Import the hypervisor definitions needed to intercept guest exceptions
and to inject events back into a guest, in preparation for gdbstub
debugging support in the MSHV accelerator.

Add:
  - union hv_intercept_parameters, struct hv_input_install_intercept and
    HVCALL_INSTALL_INTERCEPT, used to install an exception intercept on
    the partition;
  - struct hv_x64_exception_intercept_message, the message delivered on
    an intercepted exception, carrying the faulting RIP, the exception
    vector and the instruction bytes;
  - HV_X64_PENDING_EVENT_EXCEPTION, the pending-event type for an
    exception.

These definitions mirror the hypervisor headers and have no functional
effect on their own.

Signed-off-by: Doru Blânzeanu <dblanzeanu@linux.microsoft.com>
---
 include/hw/hyperv/hvgdk_mini.h | 56 ++++++++++++++++++++++++++++++++++
 1 file changed, 56 insertions(+)

diff --git a/include/hw/hyperv/hvgdk_mini.h b/include/hw/hyperv/hvgdk_mini.h
index f8838a31bb..c81928d0c2 100644
--- a/include/hw/hyperv/hvgdk_mini.h
+++ b/include/hw/hyperv/hvgdk_mini.h
@@ -22,6 +22,9 @@
 #define HV_X64_MSR_TSC_FREQUENCY    0x40000022
 #define HV_X64_MSR_APIC_FREQUENCY   0x40000023
 
+/* event_type values for hv_x64_pending_exception_event */
+#define HV_X64_PENDING_EVENT_EXCEPTION 0
+
 typedef enum hv_register_name {
     /* Pending Interruption Register */
     HV_REGISTER_PENDING_INTERRUPTION = 0x00010002,
@@ -236,6 +239,29 @@ enum hv_intercept_type {
     HV_INTERCEPT_TYPE_INVALID               = 0XFFFFFFFF,
 };
 
+union hv_intercept_parameters {
+    /* HV_INTERCEPT_PARAMETERS is defined to be an 8-byte field. */
+    uint64_t as_uint64;
+    /* HV_INTERCEPT_TYPE_X64_IO_PORT */
+    uint16_t io_port;
+    /* HV_INTERCEPT_TYPE_X64_CPUID */
+    uint32_t cpuid_index;
+    /* HV_INTERCEPT_TYPE_X64_APIC_WRITE */
+    uint32_t apic_write_mask;
+    /* HV_INTERCEPT_TYPE_EXCEPTION */
+    uint16_t exception_vector;
+    /* HV_INTERCEPT_TYPE_X64_MSR_INDEX */
+    uint32_t msr_index;
+    /* N.B. Other intercept types do not have any parameters. */
+};
+
+struct hv_input_install_intercept {
+    uint64_t partition_id;
+    uint32_t access_type; /* mask */
+    uint32_t intercept_type; /* enum hv_intercept_type */
+    union hv_intercept_parameters intercept_parameter;
+} QEMU_PACKED;
+
 struct hv_u128 {
     uint64_t low_part;
     uint64_t high_part;
@@ -836,6 +862,35 @@ struct hv_x64_memory_intercept_message {
     uint8_t instruction_bytes[16];
 } QEMU_PACKED;
 
+struct hv_x64_exception_intercept_message {
+    struct hv_x64_intercept_message_header header;
+    uint16_t exception_vector;
+    uint8_t exception_info;
+    uint8_t instruction_byte_count;
+    uint32_t error_code;
+    uint64_t exception_parameter; /* DR6 for #DB, CR2 for #PF */
+    uint64_t reserved;
+    uint8_t instruction_bytes[16];
+    struct hv_x64_segment_register ds_segment;
+    struct hv_x64_segment_register ss_segment;
+    uint64_t rax;
+    uint64_t rcx;
+    uint64_t rdx;
+    uint64_t rbx;
+    uint64_t rsp;
+    uint64_t rbp;
+    uint64_t rsi;
+    uint64_t rdi;
+    uint64_t r8;
+    uint64_t r9;
+    uint64_t r10;
+    uint64_t r11;
+    uint64_t r12;
+    uint64_t r13;
+    uint64_t r14;
+    uint64_t r15;
+} QEMU_PACKED;
+
 union hv_message_flags {
     uint8_t asu8;
     struct {
@@ -943,6 +998,7 @@ struct hv_cpuid {
 
 #define HVCALL_GET_PARTITION_PROPERTY    0x0044
 #define HVCALL_SET_PARTITION_PROPERTY    0x0045
+#define HVCALL_INSTALL_INTERCEPT         0x004d
 #define HVCALL_GET_VP_REGISTERS          0x0050
 #define HVCALL_SET_VP_REGISTERS          0x0051
 #define HVCALL_TRANSLATE_VIRTUAL_ADDRESS 0x0052
-- 
2.53.0

[PATCH 2/3] accel/mshv: add gdbstub software breakpoint support

[Doru Blânzeanu]

Implement software breakpoint based guest debugging for the MSHV
accelerator, modeled on the WHPX backend. The guest-visible debug
registers are not used: a breakpoint is an INT1 (opcode 0xf1)
patched into guest memory, which raises a #DB when executed.

A partition-wide intercept on #DB is installed the first time a
breakpoint (or single step) is requested. The intercept is not uninstalled,
so it stays for the lifetime of the VM and every vmexit mshv_handle_debug()
decides whether the #DB belongs to gdb (it hit one of our INT1 breakpoints) or
to the guest; a guest-owned #DB is handed back through a pending exception
event so the guest takes it through its own IDT.
Guest INT3 (#BP) is deliberately left to the guest's own IDT and is never
intercepted.

Signed-off-by: Doru Blânzeanu <dblanzeanu@linux.microsoft.com>
---
 accel/mshv/mshv-all.c       | 165 +++++++++++++++++++++++++++++++++++-
 include/system/mshv_int.h   |  14 +++
 target/i386/mshv/mshv-cpu.c |  86 +++++++++++++++++++
 3 files changed, 264 insertions(+), 1 deletion(-)

diff --git a/accel/mshv/mshv-all.c b/accel/mshv/mshv-all.c
index 72721d0f0d..afcb87b3b8 100644
--- a/accel/mshv/mshv-all.c
+++ b/accel/mshv/mshv-all.c
@@ -29,6 +29,7 @@
 
 #include "qemu/accel.h"
 #include "qemu/guest-random.h"
+#include "gdbstub/enums.h"
 #include "accel/accel-ops.h"
 #include "accel/accel-cpu-ops.h"
 #include "exec/cpu-common.h"
@@ -583,6 +584,10 @@ static int mshv_init(AccelState *as, MachineState *ms)
     s->nr_as = 1;
     s->as = g_new0(MshvAddressSpace, s->nr_as);
 
+    QTAILQ_INIT(&s->sw_breakpoints);
+
+    as->gdbstub.sstep_flags = SSTEP_ENABLE;
+
     mshv_state = s;
 
     mshv_init_irq_routing(s);
@@ -607,6 +612,153 @@ static int mshv_destroy_vcpu(CPUState *cpu)
     return 0;
 }
 
+struct MshvSwBreakpoint *mshv_find_sw_breakpoint(CPUState *cpu, vaddr pc)
+{
+    struct MshvSwBreakpoint *bp;
+
+    QTAILQ_FOREACH(bp, &mshv_state->sw_breakpoints, entry) {
+        if (bp->pc == pc) {
+            return bp;
+        }
+    }
+    return NULL;
+}
+
+static int mshv_install_exception_intercept(int vm_fd, uint16_t vector)
+{
+    struct hv_input_install_intercept in = {0};
+    struct mshv_root_hvcall args = {0};
+    int ret;
+
+    in.access_type = 1 << HV_X64_INTERCEPT_ACCESS_TYPE_EXECUTE;
+    in.intercept_type = HV_INTERCEPT_TYPE_EXCEPTION;
+    in.intercept_parameter.as_uint64 = vector;
+
+    args.code = HVCALL_INSTALL_INTERCEPT;
+    args.in_sz = sizeof(in);
+    args.in_ptr = (uint64_t)&in;
+
+    ret = mshv_hvcall(vm_fd, &args);
+    if (ret < 0) {
+        error_report("Failed to install exception intercept for vector %u",
+                     vector);
+    }
+    return ret;
+}
+
+/*
+ * Install the #DB intercept. This is a permanent, partition-wide latch: it is
+ * installed once on the first debug use and stays for the lifetime of the VM.
+ */
+static int mshv_init_exception_intercept(void)
+{
+    int ret;
+
+    if (mshv_state->exception_intercepts_installed) {
+        return 0;
+    }
+
+    /* Only #DB is needed: gdb breakpoints are INT1 (0xf1) */
+    ret = mshv_install_exception_intercept(mshv_state->vm, 1); /* #DB */
+    if (ret < 0) {
+        return ret;
+    }
+
+    mshv_state->exception_intercepts_installed = true;
+    return 0;
+}
+
+static int mshv_update_guest_debug(CPUState *cpu)
+{
+    if (cpu_single_stepping(cpu)) {
+        /* Installs exception intercept only on the first call */
+        return mshv_init_exception_intercept();
+    }
+    return 0;
+}
+
+static int mshv_insert_gdbstub_breakpoint(CPUState *cpu, GdbBreakpointType type,
+                                          vaddr addr, vaddr len)
+{
+    struct MshvSwBreakpoint *bp;
+    int err;
+
+    /* We don't use the guest's debug registers. */
+    if (type != GDB_BREAKPOINT_SW) {
+        return -ENOSYS;
+    }
+
+    bp = mshv_find_sw_breakpoint(cpu, addr);
+    if (bp) {
+        bp->use_count++;
+        return 0;
+    }
+
+    bp = g_new(struct MshvSwBreakpoint, 1);
+    bp->pc = addr;
+    bp->use_count = 1;
+    err = mshv_arch_insert_sw_breakpoint(cpu, bp);
+    if (err) {
+        g_free(bp);
+        return err;
+    }
+
+    QTAILQ_INSERT_HEAD(&mshv_state->sw_breakpoints, bp, entry);
+
+    /* Installs exception intercept only on the first call */
+    return mshv_init_exception_intercept();
+}
+
+static int mshv_remove_gdbstub_breakpoint(CPUState *cpu, GdbBreakpointType type,
+                                          vaddr addr, vaddr len)
+{
+    struct MshvSwBreakpoint *bp;
+    int err;
+
+    if (type != GDB_BREAKPOINT_SW) {
+        return -ENOSYS;
+    }
+
+    bp = mshv_find_sw_breakpoint(cpu, addr);
+    if (!bp) {
+        return -ENOENT;
+    }
+
+    if (bp->use_count > 1) {
+        bp->use_count--;
+        return 0;
+    }
+
+    err = mshv_arch_remove_sw_breakpoint(cpu, bp);
+    if (err) {
+        return err;
+    }
+
+    QTAILQ_REMOVE(&mshv_state->sw_breakpoints, bp, entry);
+    g_free(bp);
+
+    return 0;
+}
+
+static void mshv_remove_all_gdbstub_breakpoints(CPUState *cpu)
+{
+    struct MshvSwBreakpoint *bp, *next;
+    CPUState *tmpcpu;
+
+    QTAILQ_FOREACH_SAFE(bp, &mshv_state->sw_breakpoints, entry, next) {
+        if (mshv_arch_remove_sw_breakpoint(cpu, bp) != 0) {
+            /* Fall back to whichever CPU still has it mapped. */
+            CPU_FOREACH(tmpcpu) {
+                if (mshv_arch_remove_sw_breakpoint(tmpcpu, bp) == 0) {
+                    break;
+                }
+            }
+        }
+        QTAILQ_REMOVE(&mshv_state->sw_breakpoints, bp, entry);
+        g_free(bp);
+    }
+}
+
 static int mshv_cpu_exec(CPUState *cpu)
 {
     hv_message mshv_msg;
@@ -637,6 +789,9 @@ static int mshv_cpu_exec(CPUState *cpu)
         switch (exit_reason) {
         case MshvVmExitIgnore:
             break;
+        case MshvVmExitDebug:
+            ret = EXCP_DEBUG;
+            break;
         default:
             ret = EXCP_INTERRUPT;
             break;
@@ -708,7 +863,10 @@ static void *mshv_vcpu_thread(void *arg)
     do {
         qemu_process_cpu_events(cpu);
         if (cpu_can_run(cpu)) {
-            mshv_cpu_exec(cpu);
+            ret = mshv_cpu_exec(cpu);
+            if (ret == EXCP_DEBUG) {
+                cpu_handle_guest_debug(cpu);
+            }
         }
     } while (!cpu->unplug || cpu_can_run(cpu));
 
@@ -852,6 +1010,11 @@ static void mshv_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->synchronize_pre_loadvm = mshv_cpu_synchronize_pre_loadvm;
     ops->cpus_are_resettable = mshv_cpus_are_resettable;
     ops->handle_interrupt = generic_handle_interrupt;
+
+    ops->update_guest_debug = mshv_update_guest_debug;
+    ops->insert_gdbstub_breakpoint = mshv_insert_gdbstub_breakpoint;
+    ops->remove_gdbstub_breakpoint = mshv_remove_gdbstub_breakpoint;
+    ops->remove_all_gdbstub_breakpoints = mshv_remove_all_gdbstub_breakpoints;
 }
 
 static const TypeInfo mshv_accel_ops_type = {
diff --git a/include/system/mshv_int.h b/include/system/mshv_int.h
index b91c4d661a..9244915eb6 100644
--- a/include/system/mshv_int.h
+++ b/include/system/mshv_int.h
@@ -15,6 +15,7 @@
 #define QEMU_MSHV_INT_H
 
 #include "hw/hyperv/hvhdk.h"
+#include "exec/vaddr.h"
 
 #define MSHV_MSR_ENTRIES_COUNT 64
 
@@ -56,6 +57,13 @@ typedef struct MshvAddressSpace {
     AddressSpace *as;
 } MshvAddressSpace;
 
+struct MshvSwBreakpoint {
+    vaddr pc;
+    vaddr saved_insn;
+    int use_count;
+    QTAILQ_ENTRY(MshvSwBreakpoint) entry;
+};
+
 struct MshvState {
     AccelState parent_obj;
     int vm;
@@ -70,6 +78,8 @@ struct MshvState {
     unsigned long *used_gsi_bitmap;
     unsigned int gsi_count;
     union hv_partition_processor_features processor_features;
+    QTAILQ_HEAD(, MshvSwBreakpoint) sw_breakpoints;
+    bool exception_intercepts_installed;
 };
 
 typedef struct MshvMsiControl {
@@ -85,6 +95,7 @@ typedef enum MshvVmExit {
     MshvVmExitIgnore   = 0,
     MshvVmExitShutdown = 1,
     MshvVmExitSpecial  = 2,
+    MshvVmExitDebug    = 3,
 } MshvVmExit;
 
 void mshv_init_mmio_emu(void);
@@ -98,6 +109,9 @@ int mshv_get_generic_regs(CPUState *cpu, hv_register_assoc *assocs,
                           size_t n_regs);
 int mshv_arch_store_vcpu_state(const CPUState *cpu);
 int mshv_arch_load_vcpu_state(CPUState *cpu);
+int mshv_arch_insert_sw_breakpoint(CPUState *cpu, struct MshvSwBreakpoint *bp);
+int mshv_arch_remove_sw_breakpoint(CPUState *cpu, struct MshvSwBreakpoint *bp);
+struct MshvSwBreakpoint *mshv_find_sw_breakpoint(CPUState *cpu, vaddr pc);
 void mshv_arch_init_vcpu(CPUState *cpu);
 void mshv_arch_destroy_vcpu(CPUState *cpu);
 void mshv_arch_amend_proc_features(
diff --git a/target/i386/mshv/mshv-cpu.c b/target/i386/mshv/mshv-cpu.c
index 1c433c408c..2333d3304a 100644
--- a/target/i386/mshv/mshv-cpu.c
+++ b/target/i386/mshv/mshv-cpu.c
@@ -12,6 +12,7 @@
 
 #include "qemu/osdep.h"
 #include "qemu/error-report.h"
+#include "qemu/main-loop.h"
 #include "qemu/memalign.h"
 
 #include "system/mshv.h"
@@ -28,6 +29,7 @@
 #include "emulate/x86_decode.h"
 #include "emulate/x86_emu.h"
 #include "emulate/x86_flags.h"
+#include "gdbstub/enums.h"
 
 #include "accel/accel-cpu-target.h"
 
@@ -1932,6 +1934,46 @@ static int handle_pio(CPUState *cpu, const struct hyperv_message *msg)
     return handle_pio_non_str(cpu, &info);
 }
 
+/* Re-inject a guest-owned #DB via a pending event */
+static int reinject_exception(CPUState *cpu,
+    struct hv_x64_exception_intercept_message *info)
+{
+    hv_register_assoc assoc = { .name = HV_REGISTER_PENDING_EVENT0 };
+
+    assoc.value.pending_exception_event.event_pending = 1;
+    assoc.value.pending_exception_event.event_type =
+        HV_X64_PENDING_EVENT_EXCEPTION;
+    assoc.value.pending_exception_event.exception_parameter =
+        info->exception_parameter;
+    assoc.value.pending_exception_event.vector = info->exception_vector;
+
+    return mshv_set_generic_regs(cpu, &assoc, 1);
+}
+
+/* Check if the intercepted #DB has been set by gdb. */
+static int handle_debug(CPUState *cpu, hv_message *msg)
+{
+    struct hv_x64_exception_intercept_message *info = (void *)msg->payload;
+
+    /* Only #DB is intercepted */
+    if (info->exception_vector != EXCP01_DB) {
+        return 0;
+    }
+
+    /* INT1 hit: RIP points at the breakpoint */
+    if (mshv_find_sw_breakpoint(cpu, info->header.rip) != NULL) {
+        return EXCP_DEBUG;
+    }
+
+    /* Check if single stepping */
+    if (cpu_single_stepping(cpu)) {
+        return EXCP_DEBUG;
+    }
+
+    /* The guest's own #DB; caller re-injects it. */
+    return 0;
+}
+
 int mshv_run_vcpu(int vm_fd, CPUState *cpu, hv_message *msg, MshvVmExit *exit)
 {
     int ret;
@@ -1960,6 +2002,24 @@ int mshv_run_vcpu(int vm_fd, CPUState *cpu, hv_message *msg, MshvVmExit *exit)
             return MshvVmExitSpecial;
         }
         return MshvVmExitIgnore;
+    case HVMSG_X64_EXCEPTION_INTERCEPT:
+        bql_lock();
+        if (handle_debug(cpu, msg) == EXCP_DEBUG) {
+            *exit = MshvVmExitDebug;
+            bql_unlock();
+            return 0;
+        }
+        /* Not ours - hand the #DB back to the guest and keep running. */
+        ret = reinject_exception(cpu,
+            (struct hv_x64_exception_intercept_message *)msg->payload);
+        bql_unlock();
+        if (ret < 0) {
+            error_report("failed to reinject exception on vcpu %d",
+                         cpu->cpu_index);
+            return -1;
+        }
+        *exit = MshvVmExitIgnore;
+        return 0;
     default:
         break;
     }
@@ -1973,6 +2033,32 @@ void mshv_remove_vcpu(int vm_fd, int cpu_fd)
     close(cpu_fd);
 }
 
+int mshv_arch_insert_sw_breakpoint(CPUState *cpu, struct MshvSwBreakpoint *bp)
+{
+    static const uint8_t int1 = 0xf1;
+
+    if (cpu_memory_rw_debug(cpu, bp->pc, (uint8_t *)&bp->saved_insn, 1, 0) ||
+        cpu_memory_rw_debug(cpu, bp->pc, (uint8_t *)&int1, 1, 1)) {
+        return -EINVAL;
+    }
+    return 0;
+}
+
+int mshv_arch_remove_sw_breakpoint(CPUState *cpu, struct MshvSwBreakpoint *bp)
+{
+    uint8_t int1;
+
+    if (cpu_memory_rw_debug(cpu, bp->pc, &int1, 1, 0)) {
+        return -EINVAL;
+    }
+    if (int1 != 0xf1) {
+        return 0;
+    }
+    if (cpu_memory_rw_debug(cpu, bp->pc, (uint8_t *)&bp->saved_insn, 1, 1)) {
+        return -EINVAL;
+    }
+    return 0;
+}
 
 int mshv_create_vcpu(int vm_fd, uint8_t vp_index, int *cpu_fd)
 {
-- 
2.53.0

[PATCH 3/3] target/i386/mshv: support single-stepping

[Doru Blânzeanu]

Single-step by toggling RFLAGS.TF around the vCPU run, as WHPX does.
TF is set only on the live register, never in env->eflags, so it is not read
back and re-applied by a later register store.
The resulting #DB is reported to gdb by mshv_handle_debug() whenever the vCPU
is single-stepping.

Signed-off-by: Doru Blânzeanu <dblanzeanu@linux.microsoft.com>
---
 target/i386/mshv/mshv-cpu.c | 50 +++++++++++++++++++++++++++++++++++++
 1 file changed, 50 insertions(+)

diff --git a/target/i386/mshv/mshv-cpu.c b/target/i386/mshv/mshv-cpu.c
index 2333d3304a..036b142113 100644
--- a/target/i386/mshv/mshv-cpu.c
+++ b/target/i386/mshv/mshv-cpu.c
@@ -1974,17 +1974,67 @@ static int handle_debug(CPUState *cpu, hv_message *msg)
     return 0;
 }
 
+/*
+ * Flip RFLAGS.TF like WHPX. Set it only on the live register, not env->eflags,
+ * so a later store won't put it back.
+ */
+static int arch_set_single_step(CPUState *cpu, bool enable)
+{
+    X86CPU *x86cpu = X86_CPU(cpu);
+    CPUX86State *env = &x86cpu->env;
+    hv_register_assoc assoc = { .name = HV_X64_REGISTER_RFLAGS };
+    uint64_t rflags;
+    int ret;
+
+    if (env->regs_page && env->regs_page->isvalid != 0) {
+        rflags = env->regs_page->rflags;
+        rflags = enable ? (rflags | TF_MASK) : (rflags & ~TF_MASK);
+        env->regs_page->rflags = rflags;
+        env->regs_page->dirty |= (1u << HV_X64_REGISTER_CLASS_FLAGS);
+        return 0;
+    }
+
+    ret = mshv_get_generic_regs(cpu, &assoc, 1);
+    if (ret < 0) {
+        return ret;
+    }
+    rflags = assoc.value.reg64;
+    rflags = enable ? (rflags | TF_MASK) : (rflags & ~TF_MASK);
+    assoc.value.reg64 = rflags;
+    return mshv_set_generic_regs(cpu, &assoc, 1);
+}
+
 int mshv_run_vcpu(int vm_fd, CPUState *cpu, hv_message *msg, MshvVmExit *exit)
 {
     int ret;
     enum MshvVmExit exit_reason;
     int cpu_fd = mshv_vcpufd(cpu);
+    bool single_step;
+
+    /* enable single stepping by flipping RFLAGS.TF */
+    single_step = cpu_single_stepping(cpu);
+    if (single_step) {
+        ret = arch_set_single_step(cpu, true);
+        if (ret < 0) {
+            error_report("Failed to arm single-step (TF) on vcpu %d: %s",
+                         cpu->cpu_index, strerror(-ret));
+            *exit = MshvVmExitShutdown;
+            return -1;
+        }
+    }
 
     ret = ioctl(cpu_fd, MSHV_RUN_VP, msg);
     if (ret < 0) {
         return MshvVmExitShutdown;
     }
 
+    /* disable single stepping by flipping RFLAGS.TF */
+    if (single_step && arch_set_single_step(cpu, false) < 0) {
+        error_report("Failed to clear single-step (TF) on vcpu %d",
+                     cpu->cpu_index);
+        return -1;
+    }
+
     switch (msg->header.message_type) {
     case HVMSG_UNRECOVERABLE_EXCEPTION:
         return MshvVmExitShutdown;
-- 
2.53.0