[PATCH v3 00/68] accel: Preparatory cleanups for split-accel

[PATCH v3 00/68] accel: Preparatory cleanups for split-accel

[Philippe Mathieu-Daudé]

Since v2:
- Addressed Richard's review comments

Few changes needed before being able to add the
split acceleration:

- few method docstring added
- remove pointless stubs
- propagate soon required AccelState argument
- try to reduce current_accel() uses
- move declarations AccelClass <-> AccelOpsClass
- display model name in 'info cpus'
- add 'info accel' command to monitor
- rename CPUState @dirty -> @hwaccel_synchronized
- make accel_create_vcpu_thread() more generic
- introduce hwaccel_enabled()
- expose init/exec/destroy vcpu hooks in AccelOpsClass for re-use
- add SysemuCPUOps::can_accelerate() hook

Regards,

Phil.

Philippe Mathieu-Daudé (68):
  system/runstate: Document qemu_add_vm_change_state_handler()
  system/cpus: Defer memory layout changes until vCPUs are realized
  system/cpus: Assert interrupt handling is done with BQL locked
  accel: Keep reference to AccelOpsClass in AccelClass
  accel: Introduce AccelOpsClass::cpu_target_realize() hook
  accel/hvf: Add hvf_arch_cpu_realize() stubs
  accel/kvm: Remove kvm_init_cpu_signals() stub
  accel/kvm: Reduce kvm_create_vcpu() declaration scope
  accel: Propagate AccelState to AccelClass::init_machine()
  accel/kvm: Prefer local AccelState over global MachineState::accel
  accel/hvf: Re-use QOM allocated state
  accel/tcg: Prefer local AccelState over global current_accel()
  accel: Directly pass AccelState argument to AccelClass::has_memory()
  accel/kvm: Directly pass KVMState argument to do_kvm_create_vm()
  accel: Remove unused MachineState argument of AccelClass::setup_post()
  accel: Pass AccelState argument to gdbstub_supported_sstep_flags()
  accel: Move supports_guest_debug() declaration to AccelClass
  accel: Move cpus_are_resettable() declaration to AccelClass
  accel: Move cpu_common_[un]realize() declarations to AccelOpsClass
  accel/system: Introduce AccelClass::pre_resume_vm() handler
  accel/whpx: Convert ops::synchronize_pre_resume() -> pre_resume_vm()
  accel/system: Remove AccelOpsClass::synchronize_pre_resume()
  hw/core/machine: Display CPU model name in 'info cpus' command
  accel/system: Add 'info accel' on human monitor
  accel/tcg: Factor tcg_dump_flush_info() out
  accel/tcg: Implement get_[vcpu]_stats()
  accel/hvf: Implement get_vcpu_stats()
  accel/hvf: Report missing com.apple.security.hypervisor entitlement
  accel/hvf: Restrict internal declarations
  accel/hvf: Move per-cpu method declarations to hvf-accel-ops.c
  accel/hvf: Move generic method declarations to hvf-all.c
  cpus: Document CPUState::vcpu_dirty field
  accel/hvf: Replace @dirty field by generic CPUState::vcpu_dirty field
  accel/nvmm: Replace @dirty field by generic CPUState::vcpu_dirty field
  accel/whpx: Replace @dirty field by generic CPUState::vcpu_dirty field
  accel/kvm: Remove kvm_cpu_synchronize_state() stub
  accel/system: Document cpu_synchronize_state()
  accel/system: Document cpu_synchronize_state_post_init/reset()
  accel/nvmm: Expose nvmm_enabled() to common code
  accel/whpx: Expose whpx_enabled() to common code
  accel/system: Introduce hwaccel_enabled() helper
  target/arm: Use generic hwaccel_enabled() to check 'host' cpu type
  accel/dummy: Extract 'dummy-cpus.h' header from 'system/cpus.h'
  accel/dummy: Factor dummy_thread_precreate() out
  accel/tcg: Factor tcg_vcpu_thread_precreate() out
  accel: Factor accel_create_vcpu_thread() out
  accel: Introduce AccelOpsClass::cpu_thread_routine handler
  accel/dummy: Convert to AccelOpsClass::cpu_thread_routine
  accel/tcg: Convert to AccelOpsClass::cpu_thread_routine
  accel/hvf: Convert to AccelOpsClass::cpu_thread_routine
  accel/kvm: Convert to AccelOpsClass::cpu_thread_routine
  accel/nvmm: Convert to AccelOpsClass::cpu_thread_routine
  accel/whpx: Convert to AccelOpsClass::cpu_thread_routine
  accel: Factor accel_cpu_realize() out
  accel/tcg: Factor tcg_vcpu_init() out for re-use
  accel/tcg: Unregister the RCU before exiting RR thread
  accel/tcg: Expose vcpu_[un]register() for MTTCG
  accel/tcg: Factor mttcg_cpu_exec() out for re-use
  accel/tcg: Expose vcpu_[un]register() for RR
  accel/tcg: Factor rr_cpu_exec() out for re-use
  accel/tcg: Clear exit_request once in tcg_cpu_exec()
  accel/tcg: Add rr_vcpu_destroy() stub
  accel/system: Declare init/exec/destroy vcpu hooks
  accel/tcg: Register init/exec/destroy vcpu hooks
  accel/hvf: Register init/exec/destroy vcpu hooks
  accel/system: Pass old/new interrupt mask to handle_interrupt()
    handler
  accel: Have each accelerator implement the handle_interrupt() hook
  RFC system/memory: Restrict eventfd dispatch_write() to emulators

 qapi/machine.json                 |   3 +
 accel/accel-internal.h            |   2 +
 accel/dummy-cpus.h                |  16 ++
 accel/tcg/internal-common.h       |   2 +
 accel/tcg/tcg-accel-ops-icount.h  |   6 +-
 accel/tcg/tcg-accel-ops-mttcg.h   |   4 +-
 accel/tcg/tcg-accel-ops-rr.h      |   4 +
 accel/tcg/tcg-accel-ops.h         |   5 +-
 include/hw/core/cpu.h             |   5 +-
 include/qemu/accel.h              |  21 +-
 include/system/accel-ops.h        |  38 +++-
 include/system/cpus.h             |   5 -
 include/system/hvf.h              |  38 ----
 include/system/hvf_int.h          |  37 +++-
 include/system/hw_accel.h         |  34 ++-
 include/system/kvm.h              |   8 -
 include/system/nvmm.h             |  23 +-
 include/system/runstate.h         |  10 +
 include/system/whpx.h             |  27 +--
 target/i386/whpx/whpx-accel-ops.h |   1 -
 accel/accel-common.c              |  55 ++++-
 accel/accel-system.c              |  54 ++++-
 accel/dummy-cpus.c                |  18 +-
 accel/hvf/hvf-accel-ops.c         | 342 ++++++------------------------
 accel/hvf/hvf-all.c               | 276 ++++++++++++++++++++++--
 accel/kvm/kvm-accel-ops.c         |  20 +-
 accel/kvm/kvm-all.c               |  33 +--
 accel/qtest/qtest.c               |   7 +-
 accel/stubs/kvm-stub.c            |   9 -
 accel/stubs/nvmm-stub.c           |  12 ++
 accel/stubs/whpx-stub.c           |  12 ++
 accel/tcg/monitor.c               |  27 ++-
 accel/tcg/tcg-accel-ops-icount.c  |  16 +-
 accel/tcg/tcg-accel-ops-mttcg.c   |  58 +++--
 accel/tcg/tcg-accel-ops-rr.c      |  79 +++++--
 accel/tcg/tcg-accel-ops.c         |  44 ++--
 accel/tcg/tcg-all.c               |  20 +-
 accel/xen/xen-all.c               |   8 +-
 bsd-user/main.c                   |   2 +-
 gdbstub/system.c                  |   7 +-
 hw/core/machine-hmp-cmds.c        |   3 +-
 hw/core/machine-qmp-cmds.c        |   1 +
 linux-user/main.c                 |   2 +-
 system/cpus.c                     |  39 +---
 system/memory.c                   |  11 +-
 system/physmem.c                  |   8 +
 target/arm/arm-qmp-cmds.c         |   5 +-
 target/arm/cpu.c                  |   5 +-
 target/arm/hvf/hvf.c              |  11 +-
 target/i386/hvf/hvf.c             |  11 +-
 target/i386/hvf/x86hvf.c          |   2 +-
 target/i386/nvmm/nvmm-accel-ops.c |  14 +-
 target/i386/nvmm/nvmm-all.c       |  29 +--
 target/i386/whpx/whpx-accel-ops.c |  15 +-
 target/i386/whpx/whpx-all.c       |  33 ++-
 accel/stubs/meson.build           |   2 +
 hmp-commands-info.hx              |  12 ++
 57 files changed, 947 insertions(+), 644 deletions(-)
 create mode 100644 accel/dummy-cpus.h
 create mode 100644 accel/stubs/nvmm-stub.c
 create mode 100644 accel/stubs/whpx-stub.c

-- 
2.49.0

[PATCH v3 01/68] system/runstate: Document qemu_add_vm_change_state_handler()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/runstate.h | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/include/system/runstate.h b/include/system/runstate.h
index fdd5c4a5172..b6e8d6beab7 100644
--- a/include/system/runstate.h
+++ b/include/system/runstate.h
@@ -14,6 +14,16 @@ void runstate_replay_enable(void);
 typedef void VMChangeStateHandler(void *opaque, bool running, RunState state);
 typedef int VMChangeStateHandlerWithRet(void *opaque, bool running, RunState state);
 
+/**
+ * qemu_add_vm_change_state_handler:
+ * @cb: the callback to invoke
+ * @opaque: user data passed to the callback
+ *
+ * Register a callback function that is invoked when the vm starts or stops
+ * running.
+ *
+ * Returns: an entry to be freed using qemu_del_vm_change_state_handler()
+ */
 VMChangeStateEntry *qemu_add_vm_change_state_handler(VMChangeStateHandler *cb,
                                                      void *opaque);
 VMChangeStateEntry *qemu_add_vm_change_state_handler_prio(
-- 
2.49.0

[PATCH v3 02/68] system/cpus: Defer memory layout changes until vCPUs are realized

[Philippe Mathieu-Daudé]

vCPUs are not really usable until fully realized. Do not attempt
to commit memory changes in the middle of vCPU realization. Defer
until realization is completed and vCPU fully operational.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 system/physmem.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/system/physmem.c b/system/physmem.c
index ff0ca40222d..8b2be31fa7e 100644
--- a/system/physmem.c
+++ b/system/physmem.c
@@ -2792,6 +2792,14 @@ static void tcg_commit(MemoryListener *listener)
     cpuas = container_of(listener, CPUAddressSpace, tcg_as_listener);
     cpu = cpuas->cpu;
 
+    if (!qdev_is_realized(DEVICE(cpu))) {
+        /*
+         * The listener is also called during realize, before
+         * all of the tcg machinery for run-on is initialized.
+         */
+        return;
+    }
+
     /*
      * Defer changes to as->memory_dispatch until the cpu is quiescent.
      * Otherwise we race between (1) other cpu threads and (2) ongoing
-- 
2.49.0

Re: [PATCH v3 02/68] system/cpus: Defer memory layout changes until vCPUs are realized

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> vCPUs are not really usable until fully realized. Do not attempt
> to commit memory changes in the middle of vCPU realization. Defer
> until realization is completed and vCPU fully operational.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   system/physmem.c | 8 ++++++++
>   1 file changed, 8 insertions(+)


Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

[PATCH v3 03/68] system/cpus: Assert interrupt handling is done with BQL locked

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/tcg/tcg-accel-ops.c | 2 --
 system/cpus.c             | 2 ++
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index b24d6a75625..6116644d1c0 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -93,8 +93,6 @@ static void tcg_cpu_reset_hold(CPUState *cpu)
 /* mask must never be zero, except for A20 change call */
 void tcg_handle_interrupt(CPUState *cpu, int mask)
 {
-    g_assert(bql_locked());
-
     cpu->interrupt_request |= mask;
 
     /*
diff --git a/system/cpus.c b/system/cpus.c
index d16b0dff989..a43e0e4e796 100644
--- a/system/cpus.c
+++ b/system/cpus.c
@@ -265,6 +265,8 @@ static void generic_handle_interrupt(CPUState *cpu, int mask)
 
 void cpu_interrupt(CPUState *cpu, int mask)
 {
+    g_assert(bql_locked());
+
     if (cpus_accel->handle_interrupt) {
         cpus_accel->handle_interrupt(cpu, mask);
     } else {
-- 
2.49.0

[PATCH v3 04/68] accel: Keep reference to AccelOpsClass in AccelClass

[Philippe Mathieu-Daudé]

Allow dereferencing AccelOpsClass outside of accel/accel-system.c.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
---
 include/qemu/accel.h       | 3 +++
 include/system/accel-ops.h | 3 ++-
 accel/accel-common.c       | 1 +
 accel/accel-system.c       | 3 ++-
 accel/tcg/tcg-accel-ops.c  | 4 +++-
 5 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index fbd3d897fef..9dea3145429 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -37,6 +37,9 @@ typedef struct AccelClass {
     /*< public >*/
 
     const char *name;
+    /* Cached by accel_init_ops_interfaces() when created */
+    AccelOpsClass *ops;
+
     int (*init_machine)(MachineState *ms);
     bool (*cpu_common_realize)(CPUState *cpu, Error **errp);
     void (*cpu_common_unrealize)(CPUState *cpu);
diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index 4c99d25aeff..44b37592d02 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -10,6 +10,7 @@
 #ifndef ACCEL_OPS_H
 #define ACCEL_OPS_H
 
+#include "qemu/accel.h"
 #include "exec/vaddr.h"
 #include "qom/object.h"
 
@@ -31,7 +32,7 @@ struct AccelOpsClass {
     /*< public >*/
 
     /* initialization function called when accel is chosen */
-    void (*ops_init)(AccelOpsClass *ops);
+    void (*ops_init)(AccelClass *ac);
 
     bool (*cpus_are_resettable)(void);
     void (*cpu_reset_hold)(CPUState *cpu);
diff --git a/accel/accel-common.c b/accel/accel-common.c
index 4894b98d64a..56d88940f92 100644
--- a/accel/accel-common.c
+++ b/accel/accel-common.c
@@ -10,6 +10,7 @@
 #include "qemu/osdep.h"
 #include "qemu/accel.h"
 #include "qemu/target-info.h"
+#include "system/accel-ops.h"
 #include "accel/accel-cpu.h"
 #include "accel-internal.h"
 
diff --git a/accel/accel-system.c b/accel/accel-system.c
index a0f562ae9ff..64bc991b1ce 100644
--- a/accel/accel-system.c
+++ b/accel/accel-system.c
@@ -85,8 +85,9 @@ void accel_init_ops_interfaces(AccelClass *ac)
      * non-NULL create_vcpu_thread operation.
      */
     ops = ACCEL_OPS_CLASS(oc);
+    ac->ops = ops;
     if (ops->ops_init) {
-        ops->ops_init(ops);
+        ops->ops_init(ac);
     }
     cpus_register_accel(ops);
 }
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 6116644d1c0..37b4b21f882 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -196,8 +196,10 @@ static inline void tcg_remove_all_breakpoints(CPUState *cpu)
     cpu_watchpoint_remove_all(cpu, BP_GDB);
 }
 
-static void tcg_accel_ops_init(AccelOpsClass *ops)
+static void tcg_accel_ops_init(AccelClass *ac)
 {
+    AccelOpsClass *ops = ac->ops;
+
     if (qemu_tcg_mttcg_enabled()) {
         ops->create_vcpu_thread = mttcg_start_vcpu_thread;
         ops->kick_vcpu_thread = mttcg_kick_vcpu_thread;
-- 
2.49.0

[PATCH v3 05/68] accel: Introduce AccelOpsClass::cpu_target_realize() hook

[Philippe Mathieu-Daudé]

Allow accelerators to set vCPU properties before its realization.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/accel-ops.h | 1 +
 accel/accel-common.c       | 3 +++
 2 files changed, 4 insertions(+)

diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index 44b37592d02..a863fe59388 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -35,6 +35,7 @@ struct AccelOpsClass {
     void (*ops_init)(AccelClass *ac);
 
     bool (*cpus_are_resettable)(void);
+    bool (*cpu_target_realize)(CPUState *cpu, Error **errp);
     void (*cpu_reset_hold)(CPUState *cpu);
 
     void (*create_vcpu_thread)(CPUState *cpu); /* MANDATORY NON-NULL */
diff --git a/accel/accel-common.c b/accel/accel-common.c
index 56d88940f92..55d21b63a48 100644
--- a/accel/accel-common.c
+++ b/accel/accel-common.c
@@ -105,6 +105,9 @@ bool accel_cpu_common_realize(CPUState *cpu, Error **errp)
     if (acc->cpu_common_realize && !acc->cpu_common_realize(cpu, errp)) {
         return false;
     }
+    if (acc->ops->cpu_target_realize && !acc->ops->cpu_target_realize(cpu, errp)) {
+        return false;
+    }
 
     return true;
 }
-- 
2.49.0

[PATCH v3 06/68] accel/hvf: Add hvf_arch_cpu_realize() stubs

[Philippe Mathieu-Daudé]

Implement HVF AccelOpsClass::cpu_target_realize() hook as
empty stubs. Target implementations will come separately.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/hvf.h      | 3 +++
 accel/hvf/hvf-accel-ops.c | 2 ++
 target/arm/hvf/hvf.c      | 5 +++++
 target/i386/hvf/hvf.c     | 5 +++++
 4 files changed, 15 insertions(+)

diff --git a/include/system/hvf.h b/include/system/hvf.h
index a9a502f0c8f..8c4409a13f1 100644
--- a/include/system/hvf.h
+++ b/include/system/hvf.h
@@ -72,6 +72,9 @@ void hvf_arch_update_guest_debug(CPUState *cpu);
  * Return whether the guest supports debugging.
  */
 bool hvf_arch_supports_guest_debug(void);
+
+bool hvf_arch_cpu_realize(CPUState *cpu, Error **errp);
+
 #endif /* COMPILING_PER_TARGET */
 
 #endif
diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index d60446b85b8..b85284017c5 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -582,6 +582,8 @@ static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
 {
     AccelOpsClass *ops = ACCEL_OPS_CLASS(oc);
 
+    ops->cpu_target_realize = hvf_arch_cpu_realize;
+
     ops->create_vcpu_thread = hvf_start_vcpu_thread;
     ops->kick_vcpu_thread = hvf_kick_vcpu_thread;
 
diff --git a/target/arm/hvf/hvf.c b/target/arm/hvf/hvf.c
index 42258cc2d88..6551cf1bed1 100644
--- a/target/arm/hvf/hvf.c
+++ b/target/arm/hvf/hvf.c
@@ -1076,6 +1076,11 @@ int hvf_arch_init_vcpu(CPUState *cpu)
     return 0;
 }
 
+bool hvf_arch_cpu_realize(CPUState *cs, Error **errp)
+{
+    return true;
+}
+
 void hvf_kick_vcpu_thread(CPUState *cpu)
 {
     cpus_kick_thread(cpu);
diff --git a/target/i386/hvf/hvf.c b/target/i386/hvf/hvf.c
index 99e37a33e50..28484496710 100644
--- a/target/i386/hvf/hvf.c
+++ b/target/i386/hvf/hvf.c
@@ -367,6 +367,11 @@ int hvf_arch_init_vcpu(CPUState *cpu)
     return 0;
 }
 
+bool hvf_arch_cpu_realize(CPUState *cs, Error **errp)
+{
+    return true;
+}
+
 static void hvf_store_events(CPUState *cpu, uint32_t ins_len, uint64_t idtvec_info)
 {
     X86CPU *x86_cpu = X86_CPU(cpu);
-- 
2.49.0

[PATCH v3 07/68] accel/kvm: Remove kvm_init_cpu_signals() stub

[Philippe Mathieu-Daudé]

Since commit 57038a92bb0 ("cpus: extract out kvm-specific code
to accel/kvm") the kvm_init_cpu_signals() stub is not necessary.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/stubs/kvm-stub.c | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/accel/stubs/kvm-stub.c b/accel/stubs/kvm-stub.c
index ecfd7636f5f..b9b4427c919 100644
--- a/accel/stubs/kvm-stub.c
+++ b/accel/stubs/kvm-stub.c
@@ -105,11 +105,6 @@ unsigned int kvm_get_free_memslots(void)
     return 0;
 }
 
-void kvm_init_cpu_signals(CPUState *cpu)
-{
-    abort();
-}
-
 bool kvm_arm_supports_user_irq(void)
 {
     return false;
-- 
2.49.0

[PATCH v3 08/68] accel/kvm: Reduce kvm_create_vcpu() declaration scope

[Philippe Mathieu-Daudé]

kvm_create_vcpu() is only used within the same file unit.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/kvm.h | 8 --------
 accel/kvm/kvm-all.c  | 8 +++++++-
 2 files changed, 7 insertions(+), 9 deletions(-)

diff --git a/include/system/kvm.h b/include/system/kvm.h
index 7cc60d26f24..e943df2c09d 100644
--- a/include/system/kvm.h
+++ b/include/system/kvm.h
@@ -316,14 +316,6 @@ int kvm_create_device(KVMState *s, uint64_t type, bool test);
  */
 bool kvm_device_supported(int vmfd, uint64_t type);
 
-/**
- * kvm_create_vcpu - Gets a parked KVM vCPU or creates a KVM vCPU
- * @cpu: QOM CPUState object for which KVM vCPU has to be fetched/created.
- *
- * @returns: 0 when success, errno (<0) when failed.
- */
-int kvm_create_vcpu(CPUState *cpu);
-
 /**
  * kvm_park_vcpu - Park QEMU KVM vCPU context
  * @cpu: QOM CPUState object for which QEMU KVM vCPU context has to be parked.
diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index d095d1b98f8..17235f26464 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -453,7 +453,13 @@ static void kvm_reset_parked_vcpus(KVMState *s)
     }
 }
 
-int kvm_create_vcpu(CPUState *cpu)
+/**
+ * kvm_create_vcpu - Gets a parked KVM vCPU or creates a KVM vCPU
+ * @cpu: QOM CPUState object for which KVM vCPU has to be fetched/created.
+ *
+ * @returns: 0 when success, errno (<0) when failed.
+ */
+static int kvm_create_vcpu(CPUState *cpu)
 {
     unsigned long vcpu_id = kvm_arch_vcpu_id(cpu);
     KVMState *s = kvm_state;
-- 
2.49.0

[PATCH v3 09/68] accel: Propagate AccelState to AccelClass::init_machine()

[Philippe Mathieu-Daudé]

In order to avoid init_machine() to call current_accel(),
pass AccelState along.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
---
 include/qemu/accel.h        | 2 +-
 accel/accel-system.c        | 2 +-
 accel/hvf/hvf-accel-ops.c   | 2 +-
 accel/kvm/kvm-all.c         | 2 +-
 accel/qtest/qtest.c         | 2 +-
 accel/tcg/tcg-all.c         | 2 +-
 accel/xen/xen-all.c         | 2 +-
 bsd-user/main.c             | 2 +-
 linux-user/main.c           | 2 +-
 target/i386/nvmm/nvmm-all.c | 2 +-
 target/i386/whpx/whpx-all.c | 2 +-
 11 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index 9dea3145429..b9a9b3593d8 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -40,7 +40,7 @@ typedef struct AccelClass {
     /* Cached by accel_init_ops_interfaces() when created */
     AccelOpsClass *ops;
 
-    int (*init_machine)(MachineState *ms);
+    int (*init_machine)(AccelState *as, MachineState *ms);
     bool (*cpu_common_realize)(CPUState *cpu, Error **errp);
     void (*cpu_common_unrealize)(CPUState *cpu);
 
diff --git a/accel/accel-system.c b/accel/accel-system.c
index 64bc991b1ce..913b7155d77 100644
--- a/accel/accel-system.c
+++ b/accel/accel-system.c
@@ -37,7 +37,7 @@ int accel_init_machine(AccelState *accel, MachineState *ms)
     int ret;
     ms->accelerator = accel;
     *(acc->allowed) = true;
-    ret = acc->init_machine(ms);
+    ret = acc->init_machine(accel, ms);
     if (ret < 0) {
         ms->accelerator = NULL;
         *(acc->allowed) = false;
diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index b85284017c5..24c21c582f8 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -315,7 +315,7 @@ static void dummy_signal(int sig)
 
 bool hvf_allowed;
 
-static int hvf_accel_init(MachineState *ms)
+static int hvf_accel_init(AccelState *as, MachineState *ms)
 {
     int x;
     hv_return_t ret;
diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index 17235f26464..264f288dc64 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -2573,7 +2573,7 @@ static int kvm_setup_dirty_ring(KVMState *s)
     return 0;
 }
 
-static int kvm_init(MachineState *ms)
+static int kvm_init(AccelState *as, MachineState *ms)
 {
     MachineClass *mc = MACHINE_GET_CLASS(ms);
     static const char upgrade_note[] =
diff --git a/accel/qtest/qtest.c b/accel/qtest/qtest.c
index 92bed9264ce..8b109d4c03b 100644
--- a/accel/qtest/qtest.c
+++ b/accel/qtest/qtest.c
@@ -37,7 +37,7 @@ static void qtest_set_virtual_clock(int64_t count)
     qatomic_set_i64(&qtest_clock_counter, count);
 }
 
-static int qtest_init_accel(MachineState *ms)
+static int qtest_init_accel(AccelState *as, MachineState *ms)
 {
     return 0;
 }
diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index 6e5dc333d59..d68fbb23773 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -80,7 +80,7 @@ static void tcg_accel_instance_init(Object *obj)
 
 bool one_insn_per_tb;
 
-static int tcg_init_machine(MachineState *ms)
+static int tcg_init_machine(AccelState *as, MachineState *ms)
 {
     TCGState *s = TCG_STATE(current_accel());
     unsigned max_threads = 1;
diff --git a/accel/xen/xen-all.c b/accel/xen/xen-all.c
index de52a8f882a..1117f52bef0 100644
--- a/accel/xen/xen-all.c
+++ b/accel/xen/xen-all.c
@@ -76,7 +76,7 @@ static void xen_setup_post(MachineState *ms, AccelState *accel)
     }
 }
 
-static int xen_init(MachineState *ms)
+static int xen_init(AccelState *as, MachineState *ms)
 {
     MachineClass *mc = MACHINE_GET_CLASS(ms);
 
diff --git a/bsd-user/main.c b/bsd-user/main.c
index 7c0a059c3ba..d0cc8e0088f 100644
--- a/bsd-user/main.c
+++ b/bsd-user/main.c
@@ -474,7 +474,7 @@ int main(int argc, char **argv)
                                  opt_one_insn_per_tb, &error_abort);
         object_property_set_int(OBJECT(accel), "tb-size",
                                 opt_tb_size, &error_abort);
-        ac->init_machine(NULL);
+        ac->init_machine(accel, NULL);
     }
 
     /*
diff --git a/linux-user/main.c b/linux-user/main.c
index 5ac5b55dc65..a9142ee7268 100644
--- a/linux-user/main.c
+++ b/linux-user/main.c
@@ -820,7 +820,7 @@ int main(int argc, char **argv, char **envp)
                                  opt_one_insn_per_tb, &error_abort);
         object_property_set_int(OBJECT(accel), "tb-size",
                                 opt_tb_size, &error_abort);
-        ac->init_machine(NULL);
+        ac->init_machine(accel, NULL);
     }
 
     /*
diff --git a/target/i386/nvmm/nvmm-all.c b/target/i386/nvmm/nvmm-all.c
index f1c6120ccf1..eaae175aa5d 100644
--- a/target/i386/nvmm/nvmm-all.c
+++ b/target/i386/nvmm/nvmm-all.c
@@ -1153,7 +1153,7 @@ static struct RAMBlockNotifier nvmm_ram_notifier = {
 /* -------------------------------------------------------------------------- */
 
 static int
-nvmm_accel_init(MachineState *ms)
+nvmm_accel_init(AccelState *as, MachineState *ms)
 {
     int ret, err;
 
diff --git a/target/i386/whpx/whpx-all.c b/target/i386/whpx/whpx-all.c
index cf6d3e4cdd4..f0be840b7db 100644
--- a/target/i386/whpx/whpx-all.c
+++ b/target/i386/whpx/whpx-all.c
@@ -2505,7 +2505,7 @@ static void whpx_set_kernel_irqchip(Object *obj, Visitor *v,
  * Partition support
  */
 
-static int whpx_accel_init(MachineState *ms)
+static int whpx_accel_init(AccelState *as, MachineState *ms)
 {
     struct whpx_state *whpx;
     int ret;
-- 
2.49.0

[PATCH v3 10/68] accel/kvm: Prefer local AccelState over global MachineState::accel

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/kvm/kvm-all.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index 264f288dc64..72fba12d9fa 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -2588,15 +2588,13 @@ static int kvm_init(AccelState *as, MachineState *ms)
         { /* end of list */ }
     }, *nc = num_cpus;
     int soft_vcpus_limit, hard_vcpus_limit;
-    KVMState *s;
+    KVMState *s = KVM_STATE(as);
     const KVMCapabilityInfo *missing_cap;
     int ret;
     int type;
 
     qemu_mutex_init(&kml_slots_lock);
 
-    s = KVM_STATE(ms->accelerator);
-
     /*
      * On systems where the kernel can support different base page
      * sizes, host page size may be different from TARGET_PAGE_SIZE,
-- 
2.49.0

[PATCH v3 11/68] accel/hvf: Re-use QOM allocated state

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/hvf/hvf-accel-ops.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index 24c21c582f8..c24bc1e9c28 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -319,7 +319,7 @@ static int hvf_accel_init(AccelState *as, MachineState *ms)
 {
     int x;
     hv_return_t ret;
-    HVFState *s;
+    HVFState *s = HVF_STATE(as);
     int pa_range = 36;
     MachineClass *mc = MACHINE_GET_CLASS(ms);
 
@@ -333,8 +333,6 @@ static int hvf_accel_init(AccelState *as, MachineState *ms)
     ret = hvf_arch_vm_create(ms, (uint32_t)pa_range);
     assert_hvf_ok(ret);
 
-    s = g_new0(HVFState, 1);
-
     s->num_slots = ARRAY_SIZE(s->slots);
     for (x = 0; x < s->num_slots; ++x) {
         s->slots[x].size = 0;
-- 
2.49.0

[PATCH v3 12/68] accel/tcg: Prefer local AccelState over global current_accel()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/tcg/tcg-all.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index d68fbb23773..c674d5bcf78 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -82,7 +82,7 @@ bool one_insn_per_tb;
 
 static int tcg_init_machine(AccelState *as, MachineState *ms)
 {
-    TCGState *s = TCG_STATE(current_accel());
+    TCGState *s = TCG_STATE(as);
     unsigned max_threads = 1;
 
 #ifndef CONFIG_USER_ONLY
-- 
2.49.0

[PATCH v3 13/68] accel: Directly pass AccelState argument to AccelClass::has_memory()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 include/qemu/accel.h | 2 +-
 accel/kvm/kvm-all.c  | 4 ++--
 system/memory.c      | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index b9a9b3593d8..f327a71282c 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -46,7 +46,7 @@ typedef struct AccelClass {
 
     /* system related hooks */
     void (*setup_post)(MachineState *ms, AccelState *accel);
-    bool (*has_memory)(MachineState *ms, AddressSpace *as,
+    bool (*has_memory)(AccelState *accel, AddressSpace *as,
                        hwaddr start_addr, hwaddr size);
 
     /* gdbstub related hooks */
diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index 72fba12d9fa..d56ac2a058f 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -3789,10 +3789,10 @@ int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target)
     return r;
 }
 
-static bool kvm_accel_has_memory(MachineState *ms, AddressSpace *as,
+static bool kvm_accel_has_memory(AccelState *accel, AddressSpace *as,
                                  hwaddr start_addr, hwaddr size)
 {
-    KVMState *kvm = KVM_STATE(ms->accelerator);
+    KVMState *s = KVM_STATE(accel);
     int i;
 
     for (i = 0; i < kvm->nr_as; ++i) {
diff --git a/system/memory.c b/system/memory.c
index 76b44b8220f..e8d9b15b28f 100644
--- a/system/memory.c
+++ b/system/memory.c
@@ -3501,7 +3501,7 @@ static void mtree_print_flatview(gpointer key, gpointer value,
         if (fvi->ac) {
             for (i = 0; i < fv_address_spaces->len; ++i) {
                 as = g_array_index(fv_address_spaces, AddressSpace*, i);
-                if (fvi->ac->has_memory(current_machine, as,
+                if (fvi->ac->has_memory(current_machine->accelerator, as,
                                         int128_get64(range->addr.start),
                                         MR_SIZE(range->addr.size) + 1)) {
                     qemu_printf(" %s", fvi->ac->name);
-- 
2.49.0

Re: [PATCH v3 13/68] accel: Directly pass AccelState argument to AccelClass::has_memory()

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   include/qemu/accel.h | 2 +-
>   accel/kvm/kvm-all.c  | 4 ++--
>   system/memory.c      | 2 +-
>   3 files changed, 4 insertions(+), 4 deletions(-)
> 

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [PATCH v3 13/68] accel: Directly pass AccelState argument to AccelClass::has_memory()

[Philippe Mathieu-Daudé]

On 1/7/25 16:39, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   include/qemu/accel.h | 2 +-
>   accel/kvm/kvm-all.c  | 4 ++--
>   system/memory.c      | 2 +-
>   3 files changed, 4 insertions(+), 4 deletions(-)


> diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
> index 72fba12d9fa..d56ac2a058f 100644
> --- a/accel/kvm/kvm-all.c
> +++ b/accel/kvm/kvm-all.c
> @@ -3789,10 +3789,10 @@ int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target)
>       return r;
>   }
>   
> -static bool kvm_accel_has_memory(MachineState *ms, AddressSpace *as,
> +static bool kvm_accel_has_memory(AccelState *accel, AddressSpace *as,
>                                    hwaddr start_addr, hwaddr size)
>   {
> -    KVMState *kvm = KVM_STATE(ms->accelerator);
> +    KVMState *s = KVM_STATE(accel);

s/s/kvm/

[PATCH v3 14/68] accel/kvm: Directly pass KVMState argument to do_kvm_create_vm()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/kvm/kvm-all.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index d56ac2a058f..d5917c0a344 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -2470,13 +2470,10 @@ uint32_t kvm_dirty_ring_size(void)
     return kvm_state->kvm_dirty_ring_size;
 }
 
-static int do_kvm_create_vm(MachineState *ms, int type)
+static int do_kvm_create_vm(KVMState *s, int type)
 {
-    KVMState *s;
     int ret;
 
-    s = KVM_STATE(ms->accelerator);
-
     do {
         ret = kvm_ioctl(s, KVM_CREATE_VM, type);
     } while (ret == -EINTR);
@@ -2646,7 +2643,7 @@ static int kvm_init(AccelState *as, MachineState *ms)
         goto err;
     }
 
-    ret = do_kvm_create_vm(ms, type);
+    ret = do_kvm_create_vm(s, type);
     if (ret < 0) {
         goto err;
     }
-- 
2.49.0

Re: [PATCH v3 14/68] accel/kvm: Directly pass KVMState argument to do_kvm_create_vm()

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/kvm/kvm-all.c | 7 ++-----
>   1 file changed, 2 insertions(+), 5 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 15/68] accel: Remove unused MachineState argument of AccelClass::setup_post()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 include/qemu/accel.h | 2 +-
 accel/accel-system.c | 2 +-
 accel/xen/xen-all.c  | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index f327a71282c..a6a95ff0bcd 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -45,7 +45,7 @@ typedef struct AccelClass {
     void (*cpu_common_unrealize)(CPUState *cpu);
 
     /* system related hooks */
-    void (*setup_post)(MachineState *ms, AccelState *accel);
+    void (*setup_post)(AccelState *as);
     bool (*has_memory)(AccelState *accel, AddressSpace *as,
                        hwaddr start_addr, hwaddr size);
 
diff --git a/accel/accel-system.c b/accel/accel-system.c
index 913b7155d77..af713cc9024 100644
--- a/accel/accel-system.c
+++ b/accel/accel-system.c
@@ -58,7 +58,7 @@ void accel_setup_post(MachineState *ms)
     AccelState *accel = ms->accelerator;
     AccelClass *acc = ACCEL_GET_CLASS(accel);
     if (acc->setup_post) {
-        acc->setup_post(ms, accel);
+        acc->setup_post(accel);
     }
 }
 
diff --git a/accel/xen/xen-all.c b/accel/xen/xen-all.c
index 1117f52bef0..ba752bbe5de 100644
--- a/accel/xen/xen-all.c
+++ b/accel/xen/xen-all.c
@@ -63,7 +63,7 @@ static void xen_set_igd_gfx_passthru(Object *obj, bool value, Error **errp)
     xen_igd_gfx_pt_set(value, errp);
 }
 
-static void xen_setup_post(MachineState *ms, AccelState *accel)
+static void xen_setup_post(AccelState *as)
 {
     int rc;
 
-- 
2.49.0

Re: [PATCH v3 15/68] accel: Remove unused MachineState argument of AccelClass::setup_post()

[Philippe Mathieu-Daudé]

On 1/7/25 16:39, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   include/qemu/accel.h | 2 +-
>   accel/accel-system.c | 2 +-
>   accel/xen/xen-all.c  | 2 +-
>   3 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/include/qemu/accel.h b/include/qemu/accel.h
> index f327a71282c..a6a95ff0bcd 100644
> --- a/include/qemu/accel.h
> +++ b/include/qemu/accel.h
> @@ -45,7 +45,7 @@ typedef struct AccelClass {
>       void (*cpu_common_unrealize)(CPUState *cpu);
>   
>       /* system related hooks */
> -    void (*setup_post)(MachineState *ms, AccelState *accel);
> +    void (*setup_post)(AccelState *as);
>       bool (*has_memory)(AccelState *accel, AddressSpace *as,
>                          hwaddr start_addr, hwaddr size);
>   
> diff --git a/accel/accel-system.c b/accel/accel-system.c
> index 913b7155d77..af713cc9024 100644
> --- a/accel/accel-system.c
> +++ b/accel/accel-system.c
> @@ -58,7 +58,7 @@ void accel_setup_post(MachineState *ms)
>       AccelState *accel = ms->accelerator;
>       AccelClass *acc = ACCEL_GET_CLASS(accel);
>       if (acc->setup_post) {
> -        acc->setup_post(ms, accel);
> +        acc->setup_post(accel);
>       }
>   }
>   
> diff --git a/accel/xen/xen-all.c b/accel/xen/xen-all.c
> index 1117f52bef0..ba752bbe5de 100644
> --- a/accel/xen/xen-all.c
> +++ b/accel/xen/xen-all.c
> @@ -63,7 +63,7 @@ static void xen_set_igd_gfx_passthru(Object *obj, bool value, Error **errp)
>       xen_igd_gfx_pt_set(value, errp);
>   }
>   
> -static void xen_setup_post(MachineState *ms, AccelState *accel)
> +static void xen_setup_post(AccelState *as)
>   {

This method only accesses xen_domid/xen_domid_restrict, which are both
related to the 'accelerator', not the machine. Besides, xen_domid aims
to be in Xen AccelState and xen_domid_restrict a xen_domid_restrict
QOM property.

Regards,

Phil.

Re: [PATCH v3 15/68] accel: Remove unused MachineState argument of AccelClass::setup_post()

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   include/qemu/accel.h | 2 +-
>   accel/accel-system.c | 2 +-
>   accel/xen/xen-all.c  | 2 +-
>   3 files changed, 3 insertions(+), 3 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 16/68] accel: Pass AccelState argument to gdbstub_supported_sstep_flags()

[Philippe Mathieu-Daudé]

In order to have AccelClass methods instrospect their state,
we need to pass AccelState by argument.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/qemu/accel.h      | 2 +-
 accel/accel-common.c      | 2 +-
 accel/hvf/hvf-accel-ops.c | 2 +-
 accel/kvm/kvm-all.c       | 2 +-
 accel/tcg/tcg-all.c       | 2 +-
 5 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index a6a95ff0bcd..1c097ac4dfb 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -50,7 +50,7 @@ typedef struct AccelClass {
                        hwaddr start_addr, hwaddr size);
 
     /* gdbstub related hooks */
-    int (*gdbstub_supported_sstep_flags)(void);
+    int (*gdbstub_supported_sstep_flags)(AccelState *as);
 
     bool *allowed;
     /*
diff --git a/accel/accel-common.c b/accel/accel-common.c
index 55d21b63a48..1d04610f55e 100644
--- a/accel/accel-common.c
+++ b/accel/accel-common.c
@@ -128,7 +128,7 @@ int accel_supported_gdbstub_sstep_flags(void)
     AccelState *accel = current_accel();
     AccelClass *acc = ACCEL_GET_CLASS(accel);
     if (acc->gdbstub_supported_sstep_flags) {
-        return acc->gdbstub_supported_sstep_flags();
+        return acc->gdbstub_supported_sstep_flags(accel);
     }
     return 0;
 }
diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index c24bc1e9c28..b57a75f83e3 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -347,7 +347,7 @@ static int hvf_accel_init(AccelState *as, MachineState *ms)
     return hvf_arch_init();
 }
 
-static inline int hvf_gdbstub_sstep_flags(void)
+static inline int hvf_gdbstub_sstep_flags(AccelState *as)
 {
     return SSTEP_ENABLE | SSTEP_NOIRQ;
 }
diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index d5917c0a344..5cd38d536cf 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -3980,7 +3980,7 @@ static void kvm_accel_instance_init(Object *obj)
  * Returns: SSTEP_* flags that KVM supports for guest debug. The
  * support is probed during kvm_init()
  */
-static int kvm_gdbstub_sstep_flags(void)
+static int kvm_gdbstub_sstep_flags(AccelState *as)
 {
     return kvm_sstep_flags;
 }
diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index c674d5bcf78..5904582a68d 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -219,7 +219,7 @@ static void tcg_set_one_insn_per_tb(Object *obj, bool value, Error **errp)
     qatomic_set(&one_insn_per_tb, value);
 }
 
-static int tcg_gdbstub_supported_sstep_flags(void)
+static int tcg_gdbstub_supported_sstep_flags(AccelState *as)
 {
     /*
      * In replay mode all events will come from the log and can't be
-- 
2.49.0

[PATCH v3 17/68] accel: Move supports_guest_debug() declaration to AccelClass

[Philippe Mathieu-Daudé]

AccelOpsClass is for methods dealing with vCPUs.
When only dealing with AccelState, AccelClass is sufficient.

In order to have AccelClass methods instrospect their state,
we need to pass AccelState by argument.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/qemu/accel.h       | 1 +
 include/system/accel-ops.h | 1 -
 include/system/hvf.h       | 2 +-
 accel/hvf/hvf-accel-ops.c  | 2 +-
 accel/tcg/tcg-accel-ops.c  | 6 ------
 accel/tcg/tcg-all.c        | 6 ++++++
 gdbstub/system.c           | 7 ++++---
 target/arm/hvf/hvf.c       | 2 +-
 target/i386/hvf/hvf.c      | 2 +-
 9 files changed, 15 insertions(+), 14 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index 1c097ac4dfb..c6fe8dc3913 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -50,6 +50,7 @@ typedef struct AccelClass {
                        hwaddr start_addr, hwaddr size);
 
     /* gdbstub related hooks */
+    bool (*supports_guest_debug)(AccelState *as);
     int (*gdbstub_supported_sstep_flags)(AccelState *as);
 
     bool *allowed;
diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index a863fe59388..51faf47ac69 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -65,7 +65,6 @@ struct AccelOpsClass {
     int64_t (*get_elapsed_ticks)(void);
 
     /* gdbstub hooks */
-    bool (*supports_guest_debug)(void);
     int (*update_guest_debug)(CPUState *cpu);
     int (*insert_breakpoint)(CPUState *cpu, int type, vaddr addr, vaddr len);
     int (*remove_breakpoint)(CPUState *cpu, int type, vaddr addr, vaddr len);
diff --git a/include/system/hvf.h b/include/system/hvf.h
index 8c4409a13f1..7b9384d816c 100644
--- a/include/system/hvf.h
+++ b/include/system/hvf.h
@@ -71,7 +71,7 @@ void hvf_arch_update_guest_debug(CPUState *cpu);
 /*
  * Return whether the guest supports debugging.
  */
-bool hvf_arch_supports_guest_debug(void);
+bool hvf_arch_supports_guest_debug(AccelState *as);
 
 bool hvf_arch_cpu_realize(CPUState *cpu, Error **errp);
 
diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index b57a75f83e3..a2e0f890463 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -358,6 +358,7 @@ static void hvf_accel_class_init(ObjectClass *oc, const void *data)
     ac->name = "HVF";
     ac->init_machine = hvf_accel_init;
     ac->allowed = &hvf_allowed;
+    ac->supports_guest_debug = hvf_arch_supports_guest_debug;
     ac->gdbstub_supported_sstep_flags = hvf_gdbstub_sstep_flags;
 }
 
@@ -594,7 +595,6 @@ static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->remove_breakpoint = hvf_remove_breakpoint;
     ops->remove_all_breakpoints = hvf_remove_all_breakpoints;
     ops->update_guest_debug = hvf_update_guest_debug;
-    ops->supports_guest_debug = hvf_arch_supports_guest_debug;
 };
 static const TypeInfo hvf_accel_ops_type = {
     .name = ACCEL_OPS_NAME("hvf"),
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 37b4b21f882..07b1ec4ea50 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -106,11 +106,6 @@ void tcg_handle_interrupt(CPUState *cpu, int mask)
     }
 }
 
-static bool tcg_supports_guest_debug(void)
-{
-    return true;
-}
-
 /* Translate GDB watchpoint type to a flags value for cpu_watchpoint_* */
 static inline int xlat_gdb_type(CPUState *cpu, int gdbtype)
 {
@@ -218,7 +213,6 @@ static void tcg_accel_ops_init(AccelClass *ac)
     }
 
     ops->cpu_reset_hold = tcg_cpu_reset_hold;
-    ops->supports_guest_debug = tcg_supports_guest_debug;
     ops->insert_breakpoint = tcg_insert_breakpoint;
     ops->remove_breakpoint = tcg_remove_breakpoint;
     ops->remove_all_breakpoints = tcg_remove_all_breakpoints;
diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index 5904582a68d..93972bc0919 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -219,6 +219,11 @@ static void tcg_set_one_insn_per_tb(Object *obj, bool value, Error **errp)
     qatomic_set(&one_insn_per_tb, value);
 }
 
+static bool tcg_supports_guest_debug(AccelState *as)
+{
+    return true;
+}
+
 static int tcg_gdbstub_supported_sstep_flags(AccelState *as)
 {
     /*
@@ -242,6 +247,7 @@ static void tcg_accel_class_init(ObjectClass *oc, const void *data)
     ac->cpu_common_realize = tcg_exec_realizefn;
     ac->cpu_common_unrealize = tcg_exec_unrealizefn;
     ac->allowed = &tcg_allowed;
+    ac->supports_guest_debug = tcg_supports_guest_debug;
     ac->gdbstub_supported_sstep_flags = tcg_gdbstub_supported_sstep_flags;
 
     object_class_property_add_str(oc, "thread",
diff --git a/gdbstub/system.c b/gdbstub/system.c
index 8a32d8e1a1d..bced226fd94 100644
--- a/gdbstub/system.c
+++ b/gdbstub/system.c
@@ -634,9 +634,10 @@ int gdb_signal_to_target(int sig)
 
 bool gdb_supports_guest_debug(void)
 {
-    const AccelOpsClass *ops = cpus_get_accel();
-    if (ops->supports_guest_debug) {
-        return ops->supports_guest_debug();
+    AccelState *accel = current_accel();
+    AccelClass *acc = ACCEL_GET_CLASS(accel);
+    if (acc->supports_guest_debug) {
+        return acc->supports_guest_debug(accel);
     }
     return false;
 }
diff --git a/target/arm/hvf/hvf.c b/target/arm/hvf/hvf.c
index 6551cf1bed1..3fb0b49df8a 100644
--- a/target/arm/hvf/hvf.c
+++ b/target/arm/hvf/hvf.c
@@ -2343,7 +2343,7 @@ void hvf_arch_update_guest_debug(CPUState *cpu)
     hvf_arch_set_traps(cpu);
 }
 
-bool hvf_arch_supports_guest_debug(void)
+bool hvf_arch_supports_guest_debug(AccelState *as)
 {
     return true;
 }
diff --git a/target/i386/hvf/hvf.c b/target/i386/hvf/hvf.c
index 28484496710..bcf30662bec 100644
--- a/target/i386/hvf/hvf.c
+++ b/target/i386/hvf/hvf.c
@@ -1024,7 +1024,7 @@ void hvf_arch_update_guest_debug(CPUState *cpu)
 {
 }
 
-bool hvf_arch_supports_guest_debug(void)
+bool hvf_arch_supports_guest_debug(AccelState *as)
 {
     return false;
 }
-- 
2.49.0

[PATCH v3 18/68] accel: Move cpus_are_resettable() declaration to AccelClass

[Philippe Mathieu-Daudé]

AccelOpsClass is for methods dealing with vCPUs.
When only dealing with AccelState, AccelClass is sufficient.

Move cpus_are_resettable() declaration to accel/accel-system.c.

In order to have AccelClass methods instrospect their state,
we need to pass AccelState by argument.

Adapt KVM handler.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/qemu/accel.h       |  1 +
 include/system/accel-ops.h |  1 -
 accel/accel-system.c       | 10 ++++++++++
 accel/kvm/kvm-accel-ops.c  |  6 ------
 accel/kvm/kvm-all.c        |  6 ++++++
 system/cpus.c              |  8 --------
 6 files changed, 17 insertions(+), 15 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index c6fe8dc3913..3c6350d6d63 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -48,6 +48,7 @@ typedef struct AccelClass {
     void (*setup_post)(AccelState *as);
     bool (*has_memory)(AccelState *accel, AddressSpace *as,
                        hwaddr start_addr, hwaddr size);
+    bool (*cpus_are_resettable)(AccelState *as);
 
     /* gdbstub related hooks */
     bool (*supports_guest_debug)(AccelState *as);
diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index 51faf47ac69..d854b84a66a 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -34,7 +34,6 @@ struct AccelOpsClass {
     /* initialization function called when accel is chosen */
     void (*ops_init)(AccelClass *ac);
 
-    bool (*cpus_are_resettable)(void);
     bool (*cpu_target_realize)(CPUState *cpu, Error **errp);
     void (*cpu_reset_hold)(CPUState *cpu);
 
diff --git a/accel/accel-system.c b/accel/accel-system.c
index af713cc9024..637e2390f35 100644
--- a/accel/accel-system.c
+++ b/accel/accel-system.c
@@ -62,6 +62,16 @@ void accel_setup_post(MachineState *ms)
     }
 }
 
+bool cpus_are_resettable(void)
+{
+    AccelState *accel = current_accel();
+    AccelClass *acc = ACCEL_GET_CLASS(accel);
+    if (acc->cpus_are_resettable) {
+        return acc->cpus_are_resettable(accel);
+    }
+    return true;
+}
+
 /* initialize the arch-independent accel operation interfaces */
 void accel_init_ops_interfaces(AccelClass *ac)
 {
diff --git a/accel/kvm/kvm-accel-ops.c b/accel/kvm/kvm-accel-ops.c
index e5c15449aa6..be960bde5c4 100644
--- a/accel/kvm/kvm-accel-ops.c
+++ b/accel/kvm/kvm-accel-ops.c
@@ -78,11 +78,6 @@ static bool kvm_vcpu_thread_is_idle(CPUState *cpu)
     return !kvm_halt_in_kernel();
 }
 
-static bool kvm_cpus_are_resettable(void)
-{
-    return !kvm_enabled() || !kvm_state->guest_state_protected;
-}
-
 #ifdef TARGET_KVM_HAVE_GUEST_DEBUG
 static int kvm_update_guest_debug_ops(CPUState *cpu)
 {
@@ -96,7 +91,6 @@ static void kvm_accel_ops_class_init(ObjectClass *oc, const void *data)
 
     ops->create_vcpu_thread = kvm_start_vcpu_thread;
     ops->cpu_thread_is_idle = kvm_vcpu_thread_is_idle;
-    ops->cpus_are_resettable = kvm_cpus_are_resettable;
     ops->synchronize_post_reset = kvm_cpu_synchronize_post_reset;
     ops->synchronize_post_init = kvm_cpu_synchronize_post_init;
     ops->synchronize_state = kvm_cpu_synchronize_state;
diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index 5cd38d536cf..f29e5c4eed4 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -3974,6 +3974,11 @@ static void kvm_accel_instance_init(Object *obj)
     s->msr_energy.enable = false;
 }
 
+static bool kvm_cpus_are_resettable(AccelState *as)
+{
+    return !kvm_enabled() || !kvm_state->guest_state_protected;
+}
+
 /**
  * kvm_gdbstub_sstep_flags():
  *
@@ -3992,6 +3997,7 @@ static void kvm_accel_class_init(ObjectClass *oc, const void *data)
     ac->init_machine = kvm_init;
     ac->has_memory = kvm_accel_has_memory;
     ac->allowed = &kvm_allowed;
+    ac->cpus_are_resettable = kvm_cpus_are_resettable;
     ac->gdbstub_supported_sstep_flags = kvm_gdbstub_sstep_flags;
 
     object_class_property_add(oc, "kernel-irqchip", "on|off|split",
diff --git a/system/cpus.c b/system/cpus.c
index a43e0e4e796..4fb764ac880 100644
--- a/system/cpus.c
+++ b/system/cpus.c
@@ -195,14 +195,6 @@ void cpu_synchronize_pre_loadvm(CPUState *cpu)
     }
 }
 
-bool cpus_are_resettable(void)
-{
-    if (cpus_accel->cpus_are_resettable) {
-        return cpus_accel->cpus_are_resettable();
-    }
-    return true;
-}
-
 void cpu_exec_reset_hold(CPUState *cpu)
 {
     if (cpus_accel->cpu_reset_hold) {
-- 
2.49.0

[PATCH v3 19/68] accel: Move cpu_common_[un]realize() declarations to AccelOpsClass

[Philippe Mathieu-Daudé]

AccelClass is for methods dealing with AccelState.
When dealing with vCPUs, we want AccelOpsClass.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/qemu/accel.h       |  2 --
 include/system/accel-ops.h |  2 ++
 accel/accel-common.c       | 10 ++++++----
 accel/tcg/tcg-accel-ops.c  |  3 +++
 accel/tcg/tcg-all.c        |  2 --
 5 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index 3c6350d6d63..518c99ab643 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -41,8 +41,6 @@ typedef struct AccelClass {
     AccelOpsClass *ops;
 
     int (*init_machine)(AccelState *as, MachineState *ms);
-    bool (*cpu_common_realize)(CPUState *cpu, Error **errp);
-    void (*cpu_common_unrealize)(CPUState *cpu);
 
     /* system related hooks */
     void (*setup_post)(AccelState *as);
diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index d854b84a66a..fb199dc78f0 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -34,6 +34,8 @@ struct AccelOpsClass {
     /* initialization function called when accel is chosen */
     void (*ops_init)(AccelClass *ac);
 
+    bool (*cpu_common_realize)(CPUState *cpu, Error **errp);
+    void (*cpu_common_unrealize)(CPUState *cpu);
     bool (*cpu_target_realize)(CPUState *cpu, Error **errp);
     void (*cpu_reset_hold)(CPUState *cpu);
 
diff --git a/accel/accel-common.c b/accel/accel-common.c
index 1d04610f55e..d1a5f3ca3df 100644
--- a/accel/accel-common.c
+++ b/accel/accel-common.c
@@ -102,10 +102,12 @@ bool accel_cpu_common_realize(CPUState *cpu, Error **errp)
     }
 
     /* generic realization */
-    if (acc->cpu_common_realize && !acc->cpu_common_realize(cpu, errp)) {
+    if (acc->ops->cpu_common_realize
+        && !acc->ops->cpu_common_realize(cpu, errp)) {
         return false;
     }
-    if (acc->ops->cpu_target_realize && !acc->ops->cpu_target_realize(cpu, errp)) {
+    if (acc->ops->cpu_target_realize
+        && !acc->ops->cpu_target_realize(cpu, errp)) {
         return false;
     }
 
@@ -118,8 +120,8 @@ void accel_cpu_common_unrealize(CPUState *cpu)
     AccelClass *acc = ACCEL_GET_CLASS(accel);
 
     /* generic unrealization */
-    if (acc->cpu_common_unrealize) {
-        acc->cpu_common_unrealize(cpu);
+    if (acc->ops->cpu_common_unrealize) {
+        acc->ops->cpu_common_unrealize(cpu);
     }
 }
 
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 07b1ec4ea50..95ff451c148 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -46,6 +46,7 @@
 #include "tcg-accel-ops-mttcg.h"
 #include "tcg-accel-ops-rr.h"
 #include "tcg-accel-ops-icount.h"
+#include "internal-common.h"
 
 /* common functionality among all TCG variants */
 
@@ -212,6 +213,8 @@ static void tcg_accel_ops_init(AccelClass *ac)
         }
     }
 
+    ops->cpu_common_realize = tcg_exec_realizefn;
+    ops->cpu_common_unrealize = tcg_exec_unrealizefn;
     ops->cpu_reset_hold = tcg_cpu_reset_hold;
     ops->insert_breakpoint = tcg_insert_breakpoint;
     ops->remove_breakpoint = tcg_remove_breakpoint;
diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index 93972bc0919..ae83ca0bd10 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -244,8 +244,6 @@ static void tcg_accel_class_init(ObjectClass *oc, const void *data)
     AccelClass *ac = ACCEL_CLASS(oc);
     ac->name = "tcg";
     ac->init_machine = tcg_init_machine;
-    ac->cpu_common_realize = tcg_exec_realizefn;
-    ac->cpu_common_unrealize = tcg_exec_unrealizefn;
     ac->allowed = &tcg_allowed;
     ac->supports_guest_debug = tcg_supports_guest_debug;
     ac->gdbstub_supported_sstep_flags = tcg_gdbstub_supported_sstep_flags;
-- 
2.49.0

[PATCH v3 20/68] accel/system: Introduce AccelClass::pre_resume_vm() handler

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 include/qemu/accel.h | 3 +++
 accel/accel-system.c | 9 +++++++++
 system/cpus.c        | 1 +
 3 files changed, 13 insertions(+)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index 518c99ab643..065de80a87b 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -47,6 +47,7 @@ typedef struct AccelClass {
     bool (*has_memory)(AccelState *accel, AddressSpace *as,
                        hwaddr start_addr, hwaddr size);
     bool (*cpus_are_resettable)(AccelState *as);
+    void (*pre_resume_vm)(AccelState *as, bool step_pending);
 
     /* gdbstub related hooks */
     bool (*supports_guest_debug)(AccelState *as);
@@ -86,6 +87,8 @@ int accel_init_machine(AccelState *accel, MachineState *ms);
 /* Called just before os_setup_post (ie just before drop OS privs) */
 void accel_setup_post(MachineState *ms);
 
+void accel_pre_resume(MachineState *ms, bool step_pending);
+
 /**
  * accel_cpu_instance_init:
  * @cpu: The CPU that needs to do accel-specific object initializations.
diff --git a/accel/accel-system.c b/accel/accel-system.c
index 637e2390f35..11ba8e24d60 100644
--- a/accel/accel-system.c
+++ b/accel/accel-system.c
@@ -62,6 +62,15 @@ void accel_setup_post(MachineState *ms)
     }
 }
 
+void accel_pre_resume(MachineState *ms, bool step_pending)
+{
+    AccelState *accel = ms->accelerator;
+    AccelClass *acc = ACCEL_GET_CLASS(accel);
+    if (acc->pre_resume_vm) {
+        acc->pre_resume_vm(accel, step_pending);
+    }
+}
+
 bool cpus_are_resettable(void)
 {
     AccelState *accel = current_accel();
diff --git a/system/cpus.c b/system/cpus.c
index 4fb764ac880..98ae8b1e271 100644
--- a/system/cpus.c
+++ b/system/cpus.c
@@ -765,6 +765,7 @@ int vm_prepare_start(bool step_pending)
     if (cpus_accel->synchronize_pre_resume) {
         cpus_accel->synchronize_pre_resume(step_pending);
     }
+    accel_pre_resume(MACHINE(qdev_get_machine()), step_pending);
 
     /* We are sending this now, but the CPUs will be resumed shortly later */
     qapi_event_send_resume();
-- 
2.49.0

Re: [PATCH v3 20/68] accel/system: Introduce AccelClass::pre_resume_vm() handler

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   include/qemu/accel.h | 3 +++
>   accel/accel-system.c | 9 +++++++++
>   system/cpus.c        | 1 +
>   3 files changed, 13 insertions(+)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 21/68] accel/whpx: Convert ops::synchronize_pre_resume() -> pre_resume_vm()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 target/i386/whpx/whpx-accel-ops.h | 1 -
 target/i386/whpx/whpx-accel-ops.c | 1 -
 target/i386/whpx/whpx-all.c       | 3 ++-
 3 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/target/i386/whpx/whpx-accel-ops.h b/target/i386/whpx/whpx-accel-ops.h
index e6cf15511d4..54cfc25a147 100644
--- a/target/i386/whpx/whpx-accel-ops.h
+++ b/target/i386/whpx/whpx-accel-ops.h
@@ -21,7 +21,6 @@ void whpx_cpu_synchronize_state(CPUState *cpu);
 void whpx_cpu_synchronize_post_reset(CPUState *cpu);
 void whpx_cpu_synchronize_post_init(CPUState *cpu);
 void whpx_cpu_synchronize_pre_loadvm(CPUState *cpu);
-void whpx_cpu_synchronize_pre_resume(bool step_pending);
 
 /* state subset only touched by the VCPU itself during runtime */
 #define WHPX_SET_RUNTIME_STATE   1
diff --git a/target/i386/whpx/whpx-accel-ops.c b/target/i386/whpx/whpx-accel-ops.c
index b8bebe403c9..011810b5e50 100644
--- a/target/i386/whpx/whpx-accel-ops.c
+++ b/target/i386/whpx/whpx-accel-ops.c
@@ -95,7 +95,6 @@ static void whpx_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->synchronize_post_init = whpx_cpu_synchronize_post_init;
     ops->synchronize_state = whpx_cpu_synchronize_state;
     ops->synchronize_pre_loadvm = whpx_cpu_synchronize_pre_loadvm;
-    ops->synchronize_pre_resume = whpx_cpu_synchronize_pre_resume;
 }
 
 static const TypeInfo whpx_accel_ops_type = {
diff --git a/target/i386/whpx/whpx-all.c b/target/i386/whpx/whpx-all.c
index f0be840b7db..821167a2a77 100644
--- a/target/i386/whpx/whpx-all.c
+++ b/target/i386/whpx/whpx-all.c
@@ -2106,7 +2106,7 @@ void whpx_cpu_synchronize_pre_loadvm(CPUState *cpu)
     run_on_cpu(cpu, do_whpx_cpu_synchronize_pre_loadvm, RUN_ON_CPU_NULL);
 }
 
-void whpx_cpu_synchronize_pre_resume(bool step_pending)
+static void whpx_pre_resume_vm(AccelState *as, bool step_pending)
 {
     whpx_global.step_pending = step_pending;
 }
@@ -2703,6 +2703,7 @@ static void whpx_accel_class_init(ObjectClass *oc, const void *data)
     AccelClass *ac = ACCEL_CLASS(oc);
     ac->name = "WHPX";
     ac->init_machine = whpx_accel_init;
+    ac->pre_resume_vm = whpx_pre_resume_vm;
     ac->allowed = &whpx_allowed;
 
     object_class_property_add(oc, "kernel-irqchip", "on|off|split",
-- 
2.49.0

Re: [PATCH v3 21/68] accel/whpx: Convert ops::synchronize_pre_resume() -> pre_resume_vm()

[Philippe Mathieu-Daudé]

On 1/7/25 16:39, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   target/i386/whpx/whpx-accel-ops.h | 1 -
>   target/i386/whpx/whpx-accel-ops.c | 1 -
>   target/i386/whpx/whpx-all.c       | 3 ++-
>   3 files changed, 2 insertions(+), 3 deletions(-)

> diff --git a/target/i386/whpx/whpx-accel-ops.c b/target/i386/whpx/whpx-accel-ops.c
> index b8bebe403c9..011810b5e50 100644
> --- a/target/i386/whpx/whpx-accel-ops.c
> +++ b/target/i386/whpx/whpx-accel-ops.c
> @@ -95,7 +95,6 @@ static void whpx_accel_ops_class_init(ObjectClass *oc, const void *data)
>       ops->synchronize_post_init = whpx_cpu_synchronize_post_init;
>       ops->synchronize_state = whpx_cpu_synchronize_state;
>       ops->synchronize_pre_loadvm = whpx_cpu_synchronize_pre_loadvm;
> -    ops->synchronize_pre_resume = whpx_cpu_synchronize_pre_resume;
>   }

> @@ -2703,6 +2703,7 @@ static void whpx_accel_class_init(ObjectClass *oc, const void *data)
>       AccelClass *ac = ACCEL_CLASS(oc);
>       ac->name = "WHPX";
>       ac->init_machine = whpx_accel_init;
> +    ac->pre_resume_vm = whpx_pre_resume_vm;
>       ac->allowed = &whpx_allowed;
>   
>       object_class_property_add(oc, "kernel-irqchip", "on|off|split",

If preferred I can squash 20-22. The point here is this method does not
belong to AccelOpsClass (with vcpu argument) but AccelClass because only
requiring the (unique) AccelState argument.

Re: [PATCH v3 21/68] accel/whpx: Convert ops::synchronize_pre_resume() -> pre_resume_vm()

[Richard Henderson]

On 7/2/25 01:20, Philippe Mathieu-Daudé wrote:
> On 1/7/25 16:39, Philippe Mathieu-Daudé wrote:
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   target/i386/whpx/whpx-accel-ops.h | 1 -
>>   target/i386/whpx/whpx-accel-ops.c | 1 -
>>   target/i386/whpx/whpx-all.c       | 3 ++-
>>   3 files changed, 2 insertions(+), 3 deletions(-)
> 
>> diff --git a/target/i386/whpx/whpx-accel-ops.c b/target/i386/whpx/whpx-accel-ops.c
>> index b8bebe403c9..011810b5e50 100644
>> --- a/target/i386/whpx/whpx-accel-ops.c
>> +++ b/target/i386/whpx/whpx-accel-ops.c
>> @@ -95,7 +95,6 @@ static void whpx_accel_ops_class_init(ObjectClass *oc, const void *data)
>>       ops->synchronize_post_init = whpx_cpu_synchronize_post_init;
>>       ops->synchronize_state = whpx_cpu_synchronize_state;
>>       ops->synchronize_pre_loadvm = whpx_cpu_synchronize_pre_loadvm;
>> -    ops->synchronize_pre_resume = whpx_cpu_synchronize_pre_resume;
>>   }
> 
>> @@ -2703,6 +2703,7 @@ static void whpx_accel_class_init(ObjectClass *oc, const void *data)
>>       AccelClass *ac = ACCEL_CLASS(oc);
>>       ac->name = "WHPX";
>>       ac->init_machine = whpx_accel_init;
>> +    ac->pre_resume_vm = whpx_pre_resume_vm;
>>       ac->allowed = &whpx_allowed;
>>       object_class_property_add(oc, "kernel-irqchip", "on|off|split",
> 
> If preferred I can squash 20-22. The point here is this method does not
> belong to AccelOpsClass (with vcpu argument) but AccelClass because only
> requiring the (unique) AccelState argument.

That might be nicer.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

[PATCH v3 22/68] accel/system: Remove AccelOpsClass::synchronize_pre_resume()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 include/system/accel-ops.h | 1 -
 system/cpus.c              | 3 ---
 2 files changed, 4 deletions(-)

diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index fb199dc78f0..af54302409c 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -47,7 +47,6 @@ struct AccelOpsClass {
     void (*synchronize_post_init)(CPUState *cpu);
     void (*synchronize_state)(CPUState *cpu);
     void (*synchronize_pre_loadvm)(CPUState *cpu);
-    void (*synchronize_pre_resume)(bool step_pending);
 
     void (*handle_interrupt)(CPUState *cpu, int mask);
 
diff --git a/system/cpus.c b/system/cpus.c
index 98ae8b1e271..2c3759ea9be 100644
--- a/system/cpus.c
+++ b/system/cpus.c
@@ -762,9 +762,6 @@ int vm_prepare_start(bool step_pending)
      * WHPX accelerator needs to know whether we are going to step
      * any CPUs, before starting the first one.
      */
-    if (cpus_accel->synchronize_pre_resume) {
-        cpus_accel->synchronize_pre_resume(step_pending);
-    }
     accel_pre_resume(MACHINE(qdev_get_machine()), step_pending);
 
     /* We are sending this now, but the CPUs will be resumed shortly later */
-- 
2.49.0

[PATCH v3 23/68] hw/core/machine: Display CPU model name in 'info cpus' command

[Philippe Mathieu-Daudé]

Display the CPU model in 'info cpus'. Example before:

 $ qemu-system-aarch64 -M xlnx-versal-virt -S -monitor stdio
 QEMU 10.0.0 monitor - type 'help' for more information
 (qemu) info cpus
 * CPU #0: thread_id=42924
   CPU #1: thread_id=42924
   CPU #2: thread_id=42924
   CPU #3: thread_id=42924
 (qemu) q

and after:

 $ qemu-system-aarch64 -M xlnx-versal-virt -S -monitor stdio
 QEMU 10.0.50 monitor - type 'help' for more information
 (qemu) info cpus
 * CPU #0: thread_id=42916 (cortex-a72)
   CPU #1: thread_id=42916 (cortex-a72)
   CPU #2: thread_id=42916 (cortex-r5f)
   CPU #3: thread_id=42916 (cortex-r5f)
 (qemu)

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 qapi/machine.json          | 3 +++
 hw/core/machine-hmp-cmds.c | 3 ++-
 hw/core/machine-qmp-cmds.c | 1 +
 3 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/qapi/machine.json b/qapi/machine.json
index 0650b8de71a..d5bbb5e367e 100644
--- a/qapi/machine.json
+++ b/qapi/machine.json
@@ -80,6 +80,8 @@
 #
 # @thread-id: ID of the underlying host thread
 #
+# @model: CPU model name (since 10.1)
+#
 # @props: properties associated with a virtual CPU, e.g. the socket id
 #
 # @target: the QEMU system emulation target, which determines which
@@ -91,6 +93,7 @@
   'base'          : { 'cpu-index'    : 'int',
                       'qom-path'     : 'str',
                       'thread-id'    : 'int',
+                      'model'        : 'str',
                       '*props'       : 'CpuInstanceProperties',
                       'target'       : 'SysEmuTarget' },
   'discriminator' : 'target',
diff --git a/hw/core/machine-hmp-cmds.c b/hw/core/machine-hmp-cmds.c
index c6325cdcaaa..65eeb5e9cc2 100644
--- a/hw/core/machine-hmp-cmds.c
+++ b/hw/core/machine-hmp-cmds.c
@@ -40,7 +40,8 @@ void hmp_info_cpus(Monitor *mon, const QDict *qdict)
 
         monitor_printf(mon, "%c CPU #%" PRId64 ":", active,
                        cpu->value->cpu_index);
-        monitor_printf(mon, " thread_id=%" PRId64 "\n", cpu->value->thread_id);
+        monitor_printf(mon, " thread_id=%" PRId64 " (%s)\n",
+                       cpu->value->thread_id, cpu->value->model);
     }
 
     qapi_free_CpuInfoFastList(cpu_list);
diff --git a/hw/core/machine-qmp-cmds.c b/hw/core/machine-qmp-cmds.c
index d82043e1c68..ab4fd1ec08a 100644
--- a/hw/core/machine-qmp-cmds.c
+++ b/hw/core/machine-qmp-cmds.c
@@ -47,6 +47,7 @@ CpuInfoFastList *qmp_query_cpus_fast(Error **errp)
         value->cpu_index = cpu->cpu_index;
         value->qom_path = object_get_canonical_path(OBJECT(cpu));
         value->thread_id = cpu->thread_id;
+        value->model = cpu_model_from_type(object_get_typename(OBJECT(cpu)));
 
         if (mc->cpu_index_to_instance_props) {
             CpuInstanceProperties *props;
-- 
2.49.0

Re: [PATCH v3 23/68] hw/core/machine: Display CPU model name in 'info cpus' command

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> Display the CPU model in 'info cpus'. Example before:
> 
>   $ qemu-system-aarch64 -M xlnx-versal-virt -S -monitor stdio
>   QEMU 10.0.0 monitor - type 'help' for more information
>   (qemu) info cpus
>   * CPU #0: thread_id=42924
>     CPU #1: thread_id=42924
>     CPU #2: thread_id=42924
>     CPU #3: thread_id=42924
>   (qemu) q
> 
> and after:
> 
>   $ qemu-system-aarch64 -M xlnx-versal-virt -S -monitor stdio
>   QEMU 10.0.50 monitor - type 'help' for more information
>   (qemu) info cpus
>   * CPU #0: thread_id=42916 (cortex-a72)
>     CPU #1: thread_id=42916 (cortex-a72)
>     CPU #2: thread_id=42916 (cortex-r5f)
>     CPU #3: thread_id=42916 (cortex-r5f)
>   (qemu)
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   qapi/machine.json          | 3 +++
>   hw/core/machine-hmp-cmds.c | 3 ++-
>   hw/core/machine-qmp-cmds.c | 1 +
>   3 files changed, 6 insertions(+), 1 deletion(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 24/68] accel/system: Add 'info accel' on human monitor

[Philippe Mathieu-Daudé]

'info accel' dispatches to the AccelOpsClass::get_stats()
and get_vcpu_stats() handlers.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Acked-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/qemu/accel.h       |  1 +
 include/system/accel-ops.h |  2 ++
 accel/accel-system.c       | 28 ++++++++++++++++++++++++++++
 hmp-commands-info.hx       | 12 ++++++++++++
 4 files changed, 43 insertions(+)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index 065de80a87b..80bfe3c4d0f 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -41,6 +41,7 @@ typedef struct AccelClass {
     AccelOpsClass *ops;
 
     int (*init_machine)(AccelState *as, MachineState *ms);
+    void (*get_stats)(AccelState *as, GString *buf);
 
     /* system related hooks */
     void (*setup_post)(AccelState *as);
diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index af54302409c..106ff56d880 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -50,6 +50,8 @@ struct AccelOpsClass {
 
     void (*handle_interrupt)(CPUState *cpu, int mask);
 
+    void (*get_vcpu_stats)(CPUState *cpu, GString *buf);
+
     /**
      * @get_virtual_clock: fetch virtual clock
      * @set_virtual_clock: set virtual clock
diff --git a/accel/accel-system.c b/accel/accel-system.c
index 11ba8e24d60..918900a0a8a 100644
--- a/accel/accel-system.c
+++ b/accel/accel-system.c
@@ -25,6 +25,8 @@
 
 #include "qemu/osdep.h"
 #include "qemu/accel.h"
+#include "qapi/type-helpers.h"
+#include "monitor/monitor.h"
 #include "hw/boards.h"
 #include "system/accel-ops.h"
 #include "system/cpus.h"
@@ -81,6 +83,26 @@ bool cpus_are_resettable(void)
     return true;
 }
 
+static HumanReadableText *hmp_info_accel(Error **errp)
+{
+    AccelState *accel = current_accel();
+    AccelClass *acc = ACCEL_GET_CLASS(accel);
+    g_autoptr(GString) buf = g_string_new("");
+
+    if (acc->get_stats) {
+        acc->get_stats(accel, buf);
+    }
+    if (acc->ops->get_vcpu_stats) {
+        CPUState *cpu;
+
+        CPU_FOREACH(cpu) {
+            acc->ops->get_vcpu_stats(cpu, buf);
+        }
+    }
+
+    return human_readable_text_from_str(buf);
+}
+
 /* initialize the arch-independent accel operation interfaces */
 void accel_init_ops_interfaces(AccelClass *ac)
 {
@@ -111,11 +133,17 @@ void accel_init_ops_interfaces(AccelClass *ac)
     cpus_register_accel(ops);
 }
 
+static void accel_ops_class_init(ObjectClass *oc, const void *data)
+{
+    monitor_register_hmp_info_hrt("accel", hmp_info_accel);
+}
+
 static const TypeInfo accel_ops_type_info = {
     .name = TYPE_ACCEL_OPS,
     .parent = TYPE_OBJECT,
     .abstract = true,
     .class_size = sizeof(AccelOpsClass),
+    .class_init = accel_ops_class_init,
 };
 
 static void accel_system_register_types(void)
diff --git a/hmp-commands-info.hx b/hmp-commands-info.hx
index 639a450ee51..0496be6abfb 100644
--- a/hmp-commands-info.hx
+++ b/hmp-commands-info.hx
@@ -281,6 +281,18 @@ ERST
         .cmd        = hmp_info_sync_profile,
     },
 
+    {
+        .name       = "accel",
+        .args_type  = "",
+        .params     = "",
+        .help       = "show accelerator info",
+    },
+
+SRST
+  ``info accel``
+    Show accelerator info.
+ERST
+
 SRST
   ``info sync-profile [-m|-n]`` [*max*]
     Show synchronization profiling info, up to *max* entries (default: 10),
-- 
2.49.0

Re: [PATCH v3 24/68] accel/system: Add 'info accel' on human monitor

[Markus Armbruster]

Philippe Mathieu-Daudé <philmd@linaro.org> writes:

> 'info accel' dispatches to the AccelOpsClass::get_stats()
> and get_vcpu_stats() handlers.
>
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> Acked-by: Richard Henderson <richard.henderson@linaro.org>

Standard question for new HMP commands that don't wrap around QMP
commands: why is the functionality not useful in QMP?

Re: [PATCH v3 24/68] accel/system: Add 'info accel' on human monitor

[Philippe Mathieu-Daudé]

On 2/7/25 06:58, Markus Armbruster wrote:
> Philippe Mathieu-Daudé <philmd@linaro.org> writes:
> 
>> 'info accel' dispatches to the AccelOpsClass::get_stats()
>> and get_vcpu_stats() handlers.
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> Acked-by: Richard Henderson <richard.henderson@linaro.org>
> 
> Standard question for new HMP commands that don't wrap around QMP
> commands: why is the functionality not useful in QMP?

So far the sole use of this command is to prove the 'split accel'
works by using it in a test:
https://lore.kernel.org/qemu-devel/20250620172751.94231-43-philmd@linaro.org/

Is it worth overloading QAPI and its documentation, and overload
the QMP command set (even if prefixing with experimental / hidden 'x-')?

If so, I don't mind implementing yet another "embedded plain HMP string
to QMP command" in v4, or directly export each debug statistical value
via QAPI.

Regards,

Phil.

Re: [PATCH v3 24/68] accel/system: Add 'info accel' on human monitor

[Markus Armbruster]

Philippe Mathieu-Daudé <philmd@linaro.org> writes:

> On 2/7/25 06:58, Markus Armbruster wrote:
>> Philippe Mathieu-Daudé <philmd@linaro.org> writes:
>> 
>>> 'info accel' dispatches to the AccelOpsClass::get_stats()
>>> and get_vcpu_stats() handlers.
>>>
>>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>>> Acked-by: Richard Henderson <richard.henderson@linaro.org>
>> 
>> Standard question for new HMP commands that don't wrap around QMP
>> commands: why is the functionality not useful in QMP?
>
> So far the sole use of this command is to prove the 'split accel'
> works by using it in a test:
> https://lore.kernel.org/qemu-devel/20250620172751.94231-43-philmd@linaro.org/

Different series?  Best to add new interfaces together with their uses.

> Is it worth overloading QAPI and its documentation, and overload
> the QMP command set (even if prefixing with experimental / hidden 'x-')?
>
> If so, I don't mind implementing yet another "embedded plain HMP string
> to QMP command" in v4, or directly export each debug statistical value
> via QAPI.

Whatever the reasons for implementing something in HMP only may be, the
commit message should state them.

"Just for testing" is a valid reason for HMP-only.  It's not obvious
from the command documentation or code, though.

Re: [PATCH v3 24/68] accel/system: Add 'info accel' on human monitor

[Daniel P. Berrangé]

On Wed, Jul 02, 2025 at 02:00:01PM +0200, Markus Armbruster wrote:
> Philippe Mathieu-Daudé <philmd@linaro.org> writes:
> 
> > On 2/7/25 06:58, Markus Armbruster wrote:
> >> Philippe Mathieu-Daudé <philmd@linaro.org> writes:
> >> 
> >>> 'info accel' dispatches to the AccelOpsClass::get_stats()
> >>> and get_vcpu_stats() handlers.
> >>>
> >>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> >>> Acked-by: Richard Henderson <richard.henderson@linaro.org>
> >> 
> >> Standard question for new HMP commands that don't wrap around QMP
> >> commands: why is the functionality not useful in QMP?
> >
> > So far the sole use of this command is to prove the 'split accel'
> > works by using it in a test:
> > https://lore.kernel.org/qemu-devel/20250620172751.94231-43-philmd@linaro.org/
> 
> Different series?  Best to add new interfaces together with their uses.
> 
> > Is it worth overloading QAPI and its documentation, and overload
> > the QMP command set (even if prefixing with experimental / hidden 'x-')?
> >
> > If so, I don't mind implementing yet another "embedded plain HMP string
> > to QMP command" in v4, or directly export each debug statistical value
> > via QAPI.
> 
> Whatever the reasons for implementing something in HMP only may be, the
> commit message should state them.
> 
> "Just for testing" is a valid reason for HMP-only.  It's not obvious
> from the command documentation or code, though.

IMHO there is no valid reason for continuing to add HMP-only commands
which are not backed by QMP.

A test case can easily run QMP commands, and we accept arbitrary QMP
commands with the 'x-' prefix and experimental tag, to serve as the
basis for the HMP command printing human readable text. So if anything
I'd suggest exposing a QMP command and not bothering with HMP at all.

With regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|

Re: [PATCH v3 24/68] accel/system: Add 'info accel' on human monitor

[Philippe Mathieu-Daudé]

On 2/7/25 14:13, Daniel P. Berrangé wrote:
> On Wed, Jul 02, 2025 at 02:00:01PM +0200, Markus Armbruster wrote:
>> Philippe Mathieu-Daudé <philmd@linaro.org> writes:
>>
>>> On 2/7/25 06:58, Markus Armbruster wrote:

>>>> Standard question for new HMP commands that don't wrap around QMP
>>>> commands: why is the functionality not useful in QMP?
>>>
>>> So far the sole use of this command is to prove the 'split accel'
>>> works by using it in a test:
>>> https://lore.kernel.org/qemu-devel/20250620172751.94231-43-philmd@linaro.org/
>>
>> Different series?  Best to add new interfaces together with their uses.
>>
>>> Is it worth overloading QAPI and its documentation, and overload
>>> the QMP command set (even if prefixing with experimental / hidden 'x-')?
>>>
>>> If so, I don't mind implementing yet another "embedded plain HMP string
>>> to QMP command" in v4, or directly export each debug statistical value
>>> via QAPI.
>>
>> Whatever the reasons for implementing something in HMP only may be, the
>> commit message should state them.
>>
>> "Just for testing" is a valid reason for HMP-only.  It's not obvious
>> from the command documentation or code, though.
> 
> IMHO there is no valid reason for continuing to add HMP-only commands
> which are not backed by QMP.
> 
> A test case can easily run QMP commands, and we accept arbitrary QMP
> commands with the 'x-' prefix and experimental tag, to serve as the
> basis for the HMP command printing human readable text. So if anything
> I'd suggest exposing a QMP command and not bothering with HMP at all.

Fine.

[PATCH v3 25/68] accel/tcg: Factor tcg_dump_flush_info() out

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/tcg/internal-common.h |  2 ++
 accel/tcg/monitor.c         | 27 +++++++++++++++++----------
 2 files changed, 19 insertions(+), 10 deletions(-)

diff --git a/accel/tcg/internal-common.h b/accel/tcg/internal-common.h
index 1dbc45dd955..fb265d0cefa 100644
--- a/accel/tcg/internal-common.h
+++ b/accel/tcg/internal-common.h
@@ -139,4 +139,6 @@ G_NORETURN void cpu_io_recompile(CPUState *cpu, uintptr_t retaddr);
 void tb_phys_invalidate(TranslationBlock *tb, tb_page_addr_t page_addr);
 void tb_set_jmp_target(TranslationBlock *tb, int n, uintptr_t addr);
 
+void tcg_dump_flush_info(GString *buf);
+
 #endif
diff --git a/accel/tcg/monitor.c b/accel/tcg/monitor.c
index 1c182b6bfb5..5f74881f2a3 100644
--- a/accel/tcg/monitor.c
+++ b/accel/tcg/monitor.c
@@ -146,11 +146,26 @@ static void tcg_dump_info(GString *buf)
     g_string_append_printf(buf, "[TCG profiler not compiled]\n");
 }
 
+void tcg_dump_flush_info(GString *buf)
+{
+    size_t flush_full, flush_part, flush_elide;
+
+    g_string_append_printf(buf, "TB flush count      %u\n",
+                           qatomic_read(&tb_ctx.tb_flush_count));
+    g_string_append_printf(buf, "TB invalidate count %u\n",
+                           qatomic_read(&tb_ctx.tb_phys_invalidate_count));
+
+    tlb_flush_counts(&flush_full, &flush_part, &flush_elide);
+    g_string_append_printf(buf, "TLB full flushes    %zu\n", flush_full);
+    g_string_append_printf(buf, "TLB partial flushes %zu\n", flush_part);
+    g_string_append_printf(buf, "TLB elided flushes  %zu\n", flush_elide);
+}
+
 static void dump_exec_info(GString *buf)
 {
     struct tb_tree_stats tst = {};
     struct qht_stats hst;
-    size_t nb_tbs, flush_full, flush_part, flush_elide;
+    size_t nb_tbs;
 
     tcg_tb_foreach(tb_tree_stats_iter, &tst);
     nb_tbs = tst.nb_tbs;
@@ -187,15 +202,7 @@ static void dump_exec_info(GString *buf)
     qht_statistics_destroy(&hst);
 
     g_string_append_printf(buf, "\nStatistics:\n");
-    g_string_append_printf(buf, "TB flush count      %u\n",
-                           qatomic_read(&tb_ctx.tb_flush_count));
-    g_string_append_printf(buf, "TB invalidate count %u\n",
-                           qatomic_read(&tb_ctx.tb_phys_invalidate_count));
-
-    tlb_flush_counts(&flush_full, &flush_part, &flush_elide);
-    g_string_append_printf(buf, "TLB full flushes    %zu\n", flush_full);
-    g_string_append_printf(buf, "TLB partial flushes %zu\n", flush_part);
-    g_string_append_printf(buf, "TLB elided flushes  %zu\n", flush_elide);
+    tcg_dump_flush_info(buf);
     tcg_dump_info(buf);
 }
 
-- 
2.49.0

[PATCH v3 26/68] accel/tcg: Implement get_[vcpu]_stats()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/tcg/tcg-all.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index ae83ca0bd10..d49d2b3b0fa 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -239,11 +239,17 @@ static int tcg_gdbstub_supported_sstep_flags(AccelState *as)
     }
 }
 
+static void tcg_get_stats(AccelState *as, GString *buf)
+{
+    tcg_dump_flush_info(buf);
+}
+
 static void tcg_accel_class_init(ObjectClass *oc, const void *data)
 {
     AccelClass *ac = ACCEL_CLASS(oc);
     ac->name = "tcg";
     ac->init_machine = tcg_init_machine;
+    ac->get_stats = tcg_get_stats;
     ac->allowed = &tcg_allowed;
     ac->supports_guest_debug = tcg_supports_guest_debug;
     ac->gdbstub_supported_sstep_flags = tcg_gdbstub_supported_sstep_flags;
-- 
2.49.0

Re: [PATCH v3 26/68] accel/tcg: Implement get_[vcpu]_stats()

[Zhao Liu]

Hi Philippe,

On Tue, Jul 01, 2025 at 04:39:34PM +0200, Philippe Mathieu-Daudé wrote:
> Date: Tue,  1 Jul 2025 16:39:34 +0200
> From: Philippe Mathieu-Daudé <philmd@linaro.org>
> Subject: [PATCH v3 26/68] accel/tcg: Implement get_[vcpu]_stats()
> X-Mailer: git-send-email 2.49.0
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>  accel/tcg/tcg-all.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
> index ae83ca0bd10..d49d2b3b0fa 100644
> --- a/accel/tcg/tcg-all.c
> +++ b/accel/tcg/tcg-all.c
> @@ -239,11 +239,17 @@ static int tcg_gdbstub_supported_sstep_flags(AccelState *as)
>      }
>  }
>  
> +static void tcg_get_stats(AccelState *as, GString *buf)
> +{
> +    tcg_dump_flush_info(buf);
> +}

With:

./configure
make -j

I met this error:

/usr/bin/ld: libuser.a.p/accel_tcg_tcg-all.c.o: in function `tcg_get_stats':
/qemu/build/../accel/tcg/tcg-all.c:244: undefined reference to `tcg_dump_flush_info'

It seems tcg_dump_flush_info() needs a stub?

Thanks,
Zhao

>  static void tcg_accel_class_init(ObjectClass *oc, const void *data)
>  {
>      AccelClass *ac = ACCEL_CLASS(oc);
>      ac->name = "tcg";
>      ac->init_machine = tcg_init_machine;
> +    ac->get_stats = tcg_get_stats;
>      ac->allowed = &tcg_allowed;
>      ac->supports_guest_debug = tcg_supports_guest_debug;
>      ac->gdbstub_supported_sstep_flags = tcg_gdbstub_supported_sstep_flags;
> -- 
> 2.49.0
> 
> 

Re: [PATCH v3 26/68] accel/tcg: Implement get_[vcpu]_stats()

[Philippe Mathieu-Daudé]

On 3/7/25 09:14, Zhao Liu wrote:
> Hi Philippe,
> 
> On Tue, Jul 01, 2025 at 04:39:34PM +0200, Philippe Mathieu-Daudé wrote:
>> Date: Tue,  1 Jul 2025 16:39:34 +0200
>> From: Philippe Mathieu-Daudé <philmd@linaro.org>
>> Subject: [PATCH v3 26/68] accel/tcg: Implement get_[vcpu]_stats()
>> X-Mailer: git-send-email 2.49.0
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
>> ---
>>   accel/tcg/tcg-all.c | 6 ++++++
>>   1 file changed, 6 insertions(+)


> With:
> 
> ./configure
> make -j
> 
> I met this error:
> 
> /usr/bin/ld: libuser.a.p/accel_tcg_tcg-all.c.o: in function `tcg_get_stats':
> /qemu/build/../accel/tcg/tcg-all.c:244: undefined reference to `tcg_dump_flush_info'
> 
> It seems tcg_dump_flush_info() needs a stub?

Fixed in v5! :)

[PATCH v3 27/68] accel/hvf: Implement get_vcpu_stats()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/hvf/hvf-accel-ops.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index a2e0f890463..ca85922356b 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -352,6 +352,12 @@ static inline int hvf_gdbstub_sstep_flags(AccelState *as)
     return SSTEP_ENABLE | SSTEP_NOIRQ;
 }
 
+static void do_hvf_get_vcpu_exec_time(CPUState *cpu, run_on_cpu_data arg)
+{
+    int r = hv_vcpu_get_exec_time(cpu->accel->fd, arg.host_ptr);
+    assert_hvf_ok(r);
+}
+
 static void hvf_accel_class_init(ObjectClass *oc, const void *data)
 {
     AccelClass *ac = ACCEL_CLASS(oc);
@@ -577,6 +583,16 @@ static void hvf_remove_all_breakpoints(CPUState *cpu)
     }
 }
 
+static void hvf_get_vcpu_stats(CPUState *cpu, GString *buf)
+{
+    uint64_t time_us; /* units of mach_absolute_time() */
+
+    run_on_cpu(cpu, do_hvf_get_vcpu_exec_time, RUN_ON_CPU_HOST_PTR(&time_us));
+
+    g_string_append_printf(buf, "HVF cumulative execution time: %llu.%.3llus\n",
+                                time_us / 1000000, (time_us % 1000000) / 1000);
+}
+
 static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
 {
     AccelOpsClass *ops = ACCEL_OPS_CLASS(oc);
@@ -595,7 +611,10 @@ static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->remove_breakpoint = hvf_remove_breakpoint;
     ops->remove_all_breakpoints = hvf_remove_all_breakpoints;
     ops->update_guest_debug = hvf_update_guest_debug;
+
+    ops->get_vcpu_stats = hvf_get_vcpu_stats;
 };
+
 static const TypeInfo hvf_accel_ops_type = {
     .name = ACCEL_OPS_NAME("hvf"),
 
-- 
2.49.0

[PATCH v3 28/68] accel/hvf: Report missing com.apple.security.hypervisor entitlement

[Philippe Mathieu-Daudé]

We need the QEMU binary signed to be able to use HVF.
Improve the following:

  $ ./qemu-system-aarch64-unsigned -M virt -accel hvf
  qemu-system-aarch64-unsigned: -accel hvf: Error: ret = HV_DENIED (0xfae94007, at ../../accel/hvf/hvf-accel-ops.c:339)
  Abort trap: 6

to:

  $ ./qemu-system-aarch64-unsigned -M virt -accel hvf
  qemu-system-aarch64-unsigned: -accel hvf: Could not access HVF. Is the executable signed with com.apple.security.hypervisor entitlement?

Suggested-by: Shatyuka <shatyuka@qq.com>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2800
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/hvf/hvf-accel-ops.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index ca85922356b..8876a026120 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -331,6 +331,11 @@ static int hvf_accel_init(AccelState *as, MachineState *ms)
     }
 
     ret = hvf_arch_vm_create(ms, (uint32_t)pa_range);
+    if (ret == HV_DENIED) {
+        error_report("Could not access HVF. Is the executable signed"
+                     " with com.apple.security.hypervisor entitlement?");
+        exit(1);
+    }
     assert_hvf_ok(ret);
 
     s->num_slots = ARRAY_SIZE(s->slots);
-- 
2.49.0

Re: [PATCH v3 28/68] accel/hvf: Report missing com.apple.security.hypervisor entitlement

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> We need the QEMU binary signed to be able to use HVF.
> Improve the following:
> 
>    $ ./qemu-system-aarch64-unsigned -M virt -accel hvf
>    qemu-system-aarch64-unsigned: -accel hvf: Error: ret = HV_DENIED (0xfae94007, at ../../accel/hvf/hvf-accel-ops.c:339)
>    Abort trap: 6
> 
> to:
> 
>    $ ./qemu-system-aarch64-unsigned -M virt -accel hvf
>    qemu-system-aarch64-unsigned: -accel hvf: Could not access HVF. Is the executable signed with com.apple.security.hypervisor entitlement?
> 
> Suggested-by: Shatyuka <shatyuka@qq.com>
> Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2800
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/hvf/hvf-accel-ops.c | 5 +++++
>   1 file changed, 5 insertions(+)

Acked-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 29/68] accel/hvf: Restrict internal declarations

[Philippe Mathieu-Daudé]

Common code only needs to know whether HVF is enabled and
the QOM type. Move the rest to "hvf_int.h", removing the
need for COMPILING_PER_TARGET #ifdef'ry.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 include/system/hvf.h     | 41 ----------------------------------------
 include/system/hvf_int.h | 36 +++++++++++++++++++++++++++++++++++
 2 files changed, 36 insertions(+), 41 deletions(-)

diff --git a/include/system/hvf.h b/include/system/hvf.h
index 7b9384d816c..d3dcf088b3f 100644
--- a/include/system/hvf.h
+++ b/include/system/hvf.h
@@ -14,10 +14,6 @@
 #define HVF_H
 
 #include "qemu/accel.h"
-#include "qemu/queue.h"
-#include "exec/vaddr.h"
-#include "qom/object.h"
-#include "exec/vaddr.h"
 
 #ifdef COMPILING_PER_TARGET
 # ifdef CONFIG_HVF
@@ -40,41 +36,4 @@ typedef struct HVFState HVFState;
 DECLARE_INSTANCE_CHECKER(HVFState, HVF_STATE,
                          TYPE_HVF_ACCEL)
 
-#ifdef COMPILING_PER_TARGET
-struct hvf_sw_breakpoint {
-    vaddr pc;
-    vaddr saved_insn;
-    int use_count;
-    QTAILQ_ENTRY(hvf_sw_breakpoint) entry;
-};
-
-struct hvf_sw_breakpoint *hvf_find_sw_breakpoint(CPUState *cpu,
-                                                 vaddr pc);
-int hvf_sw_breakpoints_active(CPUState *cpu);
-
-int hvf_arch_insert_sw_breakpoint(CPUState *cpu, struct hvf_sw_breakpoint *bp);
-int hvf_arch_remove_sw_breakpoint(CPUState *cpu, struct hvf_sw_breakpoint *bp);
-int hvf_arch_insert_hw_breakpoint(vaddr addr, vaddr len, int type);
-int hvf_arch_remove_hw_breakpoint(vaddr addr, vaddr len, int type);
-void hvf_arch_remove_all_hw_breakpoints(void);
-
-/*
- * hvf_update_guest_debug:
- * @cs: CPUState for the CPU to update
- *
- * Update guest to enable or disable debugging. Per-arch specifics will be
- * handled by calling down to hvf_arch_update_guest_debug.
- */
-int hvf_update_guest_debug(CPUState *cpu);
-void hvf_arch_update_guest_debug(CPUState *cpu);
-
-/*
- * Return whether the guest supports debugging.
- */
-bool hvf_arch_supports_guest_debug(AccelState *as);
-
-bool hvf_arch_cpu_realize(CPUState *cpu, Error **errp);
-
-#endif /* COMPILING_PER_TARGET */
-
 #endif
diff --git a/include/system/hvf_int.h b/include/system/hvf_int.h
index d774e58df91..ea6730f255d 100644
--- a/include/system/hvf_int.h
+++ b/include/system/hvf_int.h
@@ -12,6 +12,8 @@
 #define HVF_INT_H
 
 #include "qemu/queue.h"
+#include "exec/vaddr.h"
+#include "qom/object.h"
 
 #ifdef __aarch64__
 #include <Hypervisor/Hypervisor.h>
@@ -77,4 +79,38 @@ int hvf_put_registers(CPUState *);
 int hvf_get_registers(CPUState *);
 void hvf_kick_vcpu_thread(CPUState *cpu);
 
+struct hvf_sw_breakpoint {
+    vaddr pc;
+    vaddr saved_insn;
+    int use_count;
+    QTAILQ_ENTRY(hvf_sw_breakpoint) entry;
+};
+
+struct hvf_sw_breakpoint *hvf_find_sw_breakpoint(CPUState *cpu,
+                                                 vaddr pc);
+int hvf_sw_breakpoints_active(CPUState *cpu);
+
+int hvf_arch_insert_sw_breakpoint(CPUState *cpu, struct hvf_sw_breakpoint *bp);
+int hvf_arch_remove_sw_breakpoint(CPUState *cpu, struct hvf_sw_breakpoint *bp);
+int hvf_arch_insert_hw_breakpoint(vaddr addr, vaddr len, int type);
+int hvf_arch_remove_hw_breakpoint(vaddr addr, vaddr len, int type);
+void hvf_arch_remove_all_hw_breakpoints(void);
+
+/*
+ * hvf_update_guest_debug:
+ * @cs: CPUState for the CPU to update
+ *
+ * Update guest to enable or disable debugging. Per-arch specifics will be
+ * handled by calling down to hvf_arch_update_guest_debug.
+ */
+int hvf_update_guest_debug(CPUState *cpu);
+void hvf_arch_update_guest_debug(CPUState *cpu);
+
+/*
+ * Return whether the guest supports debugging.
+ */
+bool hvf_arch_supports_guest_debug(AccelState *as);
+
+bool hvf_arch_cpu_realize(CPUState *cpu, Error **errp);
+
 #endif
-- 
2.49.0

Re: [PATCH v3 29/68] accel/hvf: Restrict internal declarations

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> Common code only needs to know whether HVF is enabled and
> the QOM type. Move the rest to "hvf_int.h", removing the
> need for COMPILING_PER_TARGET #ifdef'ry.
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   include/system/hvf.h     | 41 ----------------------------------------
>   include/system/hvf_int.h | 36 +++++++++++++++++++++++++++++++++++
>   2 files changed, 36 insertions(+), 41 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 30/68] accel/hvf: Move per-cpu method declarations to hvf-accel-ops.c

[Philippe Mathieu-Daudé]

hvf-all.c aims to contain the generic accel methods (TYPE_ACCEL),
while hvf-accel-ops.c the per-vcpu methods (TYPE_ACCEL_OPS).

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/hvf/hvf-accel-ops.c | 29 +++++++++++++++++++++++++++++
 accel/hvf/hvf-all.c       | 29 -----------------------------
 2 files changed, 29 insertions(+), 29 deletions(-)

diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index 8876a026120..8242a78c0c0 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -50,6 +50,7 @@
 #include "qemu/osdep.h"
 #include "qemu/error-report.h"
 #include "qemu/main-loop.h"
+#include "qemu/queue.h"
 #include "system/address-spaces.h"
 #include "gdbstub/enums.h"
 #include "hw/boards.h"
@@ -486,6 +487,34 @@ static void hvf_start_vcpu_thread(CPUState *cpu)
                        cpu, QEMU_THREAD_JOINABLE);
 }
 
+struct hvf_sw_breakpoint *hvf_find_sw_breakpoint(CPUState *cpu, vaddr pc)
+{
+    struct hvf_sw_breakpoint *bp;
+
+    QTAILQ_FOREACH(bp, &hvf_state->hvf_sw_breakpoints, entry) {
+        if (bp->pc == pc) {
+            return bp;
+        }
+    }
+    return NULL;
+}
+
+int hvf_sw_breakpoints_active(CPUState *cpu)
+{
+    return !QTAILQ_EMPTY(&hvf_state->hvf_sw_breakpoints);
+}
+
+static void do_hvf_update_guest_debug(CPUState *cpu, run_on_cpu_data arg)
+{
+    hvf_arch_update_guest_debug(cpu);
+}
+
+int hvf_update_guest_debug(CPUState *cpu)
+{
+    run_on_cpu(cpu, do_hvf_update_guest_debug, RUN_ON_CPU_NULL);
+    return 0;
+}
+
 static int hvf_insert_breakpoint(CPUState *cpu, int type, vaddr addr, vaddr len)
 {
     struct hvf_sw_breakpoint *bp;
diff --git a/accel/hvf/hvf-all.c b/accel/hvf/hvf-all.c
index 8c387fda24d..481d7dece57 100644
--- a/accel/hvf/hvf-all.c
+++ b/accel/hvf/hvf-all.c
@@ -12,7 +12,6 @@
 #include "qemu/error-report.h"
 #include "system/hvf.h"
 #include "system/hvf_int.h"
-#include "hw/core/cpu.h"
 
 const char *hvf_return_string(hv_return_t ret)
 {
@@ -41,31 +40,3 @@ void assert_hvf_ok_impl(hv_return_t ret, const char *file, unsigned int line,
 
     abort();
 }
-
-struct hvf_sw_breakpoint *hvf_find_sw_breakpoint(CPUState *cpu, vaddr pc)
-{
-    struct hvf_sw_breakpoint *bp;
-
-    QTAILQ_FOREACH(bp, &hvf_state->hvf_sw_breakpoints, entry) {
-        if (bp->pc == pc) {
-            return bp;
-        }
-    }
-    return NULL;
-}
-
-int hvf_sw_breakpoints_active(CPUState *cpu)
-{
-    return !QTAILQ_EMPTY(&hvf_state->hvf_sw_breakpoints);
-}
-
-static void do_hvf_update_guest_debug(CPUState *cpu, run_on_cpu_data arg)
-{
-    hvf_arch_update_guest_debug(cpu);
-}
-
-int hvf_update_guest_debug(CPUState *cpu)
-{
-    run_on_cpu(cpu, do_hvf_update_guest_debug, RUN_ON_CPU_NULL);
-    return 0;
-}
-- 
2.49.0

Re: [PATCH v3 30/68] accel/hvf: Move per-cpu method declarations to hvf-accel-ops.c

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> hvf-all.c aims to contain the generic accel methods (TYPE_ACCEL),
> while hvf-accel-ops.c the per-vcpu methods (TYPE_ACCEL_OPS).
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/hvf/hvf-accel-ops.c | 29 +++++++++++++++++++++++++++++
>   accel/hvf/hvf-all.c       | 29 -----------------------------
>   2 files changed, 29 insertions(+), 29 deletions(-)
> 


Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

[PATCH v3 31/68] accel/hvf: Move generic method declarations to hvf-all.c

[Philippe Mathieu-Daudé]

hvf-all.c aims to contain the generic accel methods (TYPE_ACCEL),
while hvf-accel-ops.c the per-vcpu methods (TYPE_ACCEL_OPS).

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/hvf/hvf-accel-ops.c | 272 +-------------------------------------
 accel/hvf/hvf-all.c       | 267 +++++++++++++++++++++++++++++++++++++
 2 files changed, 272 insertions(+), 267 deletions(-)

diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index 8242a78c0c0..319c30f703c 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -48,18 +48,16 @@
  */
 
 #include "qemu/osdep.h"
-#include "qemu/error-report.h"
+#include "qemu/guest-random.h"
 #include "qemu/main-loop.h"
 #include "qemu/queue.h"
-#include "system/address-spaces.h"
 #include "gdbstub/enums.h"
-#include "hw/boards.h"
+#include "exec/cpu-common.h"
 #include "system/accel-ops.h"
 #include "system/cpus.h"
 #include "system/hvf.h"
 #include "system/hvf_int.h"
-#include "system/runstate.h"
-#include "qemu/guest-random.h"
+#include "hw/core/cpu.h"
 
 HVFState *hvf_state;
 
@@ -79,127 +77,6 @@ hvf_slot *hvf_find_overlap_slot(uint64_t start, uint64_t size)
     return NULL;
 }
 
-struct mac_slot {
-    int present;
-    uint64_t size;
-    uint64_t gpa_start;
-    uint64_t gva;
-};
-
-struct mac_slot mac_slots[32];
-
-static int do_hvf_set_memory(hvf_slot *slot, hv_memory_flags_t flags)
-{
-    struct mac_slot *macslot;
-    hv_return_t ret;
-
-    macslot = &mac_slots[slot->slot_id];
-
-    if (macslot->present) {
-        if (macslot->size != slot->size) {
-            macslot->present = 0;
-            ret = hv_vm_unmap(macslot->gpa_start, macslot->size);
-            assert_hvf_ok(ret);
-        }
-    }
-
-    if (!slot->size) {
-        return 0;
-    }
-
-    macslot->present = 1;
-    macslot->gpa_start = slot->start;
-    macslot->size = slot->size;
-    ret = hv_vm_map(slot->mem, slot->start, slot->size, flags);
-    assert_hvf_ok(ret);
-    return 0;
-}
-
-static void hvf_set_phys_mem(MemoryRegionSection *section, bool add)
-{
-    hvf_slot *mem;
-    MemoryRegion *area = section->mr;
-    bool writable = !area->readonly && !area->rom_device;
-    hv_memory_flags_t flags;
-    uint64_t page_size = qemu_real_host_page_size();
-
-    if (!memory_region_is_ram(area)) {
-        if (writable) {
-            return;
-        } else if (!memory_region_is_romd(area)) {
-            /*
-             * If the memory device is not in romd_mode, then we actually want
-             * to remove the hvf memory slot so all accesses will trap.
-             */
-             add = false;
-        }
-    }
-
-    if (!QEMU_IS_ALIGNED(int128_get64(section->size), page_size) ||
-        !QEMU_IS_ALIGNED(section->offset_within_address_space, page_size)) {
-        /* Not page aligned, so we can not map as RAM */
-        add = false;
-    }
-
-    mem = hvf_find_overlap_slot(
-            section->offset_within_address_space,
-            int128_get64(section->size));
-
-    if (mem && add) {
-        if (mem->size == int128_get64(section->size) &&
-            mem->start == section->offset_within_address_space &&
-            mem->mem == (memory_region_get_ram_ptr(area) +
-            section->offset_within_region)) {
-            return; /* Same region was attempted to register, go away. */
-        }
-    }
-
-    /* Region needs to be reset. set the size to 0 and remap it. */
-    if (mem) {
-        mem->size = 0;
-        if (do_hvf_set_memory(mem, 0)) {
-            error_report("Failed to reset overlapping slot");
-            abort();
-        }
-    }
-
-    if (!add) {
-        return;
-    }
-
-    if (area->readonly ||
-        (!memory_region_is_ram(area) && memory_region_is_romd(area))) {
-        flags = HV_MEMORY_READ | HV_MEMORY_EXEC;
-    } else {
-        flags = HV_MEMORY_READ | HV_MEMORY_WRITE | HV_MEMORY_EXEC;
-    }
-
-    /* Now make a new slot. */
-    int x;
-
-    for (x = 0; x < hvf_state->num_slots; ++x) {
-        mem = &hvf_state->slots[x];
-        if (!mem->size) {
-            break;
-        }
-    }
-
-    if (x == hvf_state->num_slots) {
-        error_report("No free slots");
-        abort();
-    }
-
-    mem->size = int128_get64(section->size);
-    mem->mem = memory_region_get_ram_ptr(area) + section->offset_within_region;
-    mem->start = section->offset_within_address_space;
-    mem->region = area;
-
-    if (do_hvf_set_memory(mem, flags)) {
-        error_report("Error registering new memory slot");
-        abort();
-    }
-}
-
 static void do_hvf_cpu_synchronize_state(CPUState *cpu, run_on_cpu_data arg)
 {
     if (!cpu->accel->dirty) {
@@ -237,157 +114,16 @@ static void hvf_cpu_synchronize_pre_loadvm(CPUState *cpu)
     run_on_cpu(cpu, do_hvf_cpu_synchronize_set_dirty, RUN_ON_CPU_NULL);
 }
 
-static void hvf_set_dirty_tracking(MemoryRegionSection *section, bool on)
-{
-    hvf_slot *slot;
-
-    slot = hvf_find_overlap_slot(
-            section->offset_within_address_space,
-            int128_get64(section->size));
-
-    /* protect region against writes; begin tracking it */
-    if (on) {
-        slot->flags |= HVF_SLOT_LOG;
-        hv_vm_protect((uintptr_t)slot->start, (size_t)slot->size,
-                      HV_MEMORY_READ | HV_MEMORY_EXEC);
-    /* stop tracking region*/
-    } else {
-        slot->flags &= ~HVF_SLOT_LOG;
-        hv_vm_protect((uintptr_t)slot->start, (size_t)slot->size,
-                      HV_MEMORY_READ | HV_MEMORY_WRITE | HV_MEMORY_EXEC);
-    }
-}
-
-static void hvf_log_start(MemoryListener *listener,
-                          MemoryRegionSection *section, int old, int new)
-{
-    if (old != 0) {
-        return;
-    }
-
-    hvf_set_dirty_tracking(section, 1);
-}
-
-static void hvf_log_stop(MemoryListener *listener,
-                         MemoryRegionSection *section, int old, int new)
-{
-    if (new != 0) {
-        return;
-    }
-
-    hvf_set_dirty_tracking(section, 0);
-}
-
-static void hvf_log_sync(MemoryListener *listener,
-                         MemoryRegionSection *section)
-{
-    /*
-     * sync of dirty pages is handled elsewhere; just make sure we keep
-     * tracking the region.
-     */
-    hvf_set_dirty_tracking(section, 1);
-}
-
-static void hvf_region_add(MemoryListener *listener,
-                           MemoryRegionSection *section)
-{
-    hvf_set_phys_mem(section, true);
-}
-
-static void hvf_region_del(MemoryListener *listener,
-                           MemoryRegionSection *section)
-{
-    hvf_set_phys_mem(section, false);
-}
-
-static MemoryListener hvf_memory_listener = {
-    .name = "hvf",
-    .priority = MEMORY_LISTENER_PRIORITY_ACCEL,
-    .region_add = hvf_region_add,
-    .region_del = hvf_region_del,
-    .log_start = hvf_log_start,
-    .log_stop = hvf_log_stop,
-    .log_sync = hvf_log_sync,
-};
-
 static void dummy_signal(int sig)
 {
 }
 
-bool hvf_allowed;
-
-static int hvf_accel_init(AccelState *as, MachineState *ms)
-{
-    int x;
-    hv_return_t ret;
-    HVFState *s = HVF_STATE(as);
-    int pa_range = 36;
-    MachineClass *mc = MACHINE_GET_CLASS(ms);
-
-    if (mc->hvf_get_physical_address_range) {
-        pa_range = mc->hvf_get_physical_address_range(ms);
-        if (pa_range < 0) {
-            return -EINVAL;
-        }
-    }
-
-    ret = hvf_arch_vm_create(ms, (uint32_t)pa_range);
-    if (ret == HV_DENIED) {
-        error_report("Could not access HVF. Is the executable signed"
-                     " with com.apple.security.hypervisor entitlement?");
-        exit(1);
-    }
-    assert_hvf_ok(ret);
-
-    s->num_slots = ARRAY_SIZE(s->slots);
-    for (x = 0; x < s->num_slots; ++x) {
-        s->slots[x].size = 0;
-        s->slots[x].slot_id = x;
-    }
-
-    QTAILQ_INIT(&s->hvf_sw_breakpoints);
-
-    hvf_state = s;
-    memory_listener_register(&hvf_memory_listener, &address_space_memory);
-
-    return hvf_arch_init();
-}
-
-static inline int hvf_gdbstub_sstep_flags(AccelState *as)
-{
-    return SSTEP_ENABLE | SSTEP_NOIRQ;
-}
-
 static void do_hvf_get_vcpu_exec_time(CPUState *cpu, run_on_cpu_data arg)
 {
     int r = hv_vcpu_get_exec_time(cpu->accel->fd, arg.host_ptr);
     assert_hvf_ok(r);
 }
 
-static void hvf_accel_class_init(ObjectClass *oc, const void *data)
-{
-    AccelClass *ac = ACCEL_CLASS(oc);
-    ac->name = "HVF";
-    ac->init_machine = hvf_accel_init;
-    ac->allowed = &hvf_allowed;
-    ac->supports_guest_debug = hvf_arch_supports_guest_debug;
-    ac->gdbstub_supported_sstep_flags = hvf_gdbstub_sstep_flags;
-}
-
-static const TypeInfo hvf_accel_type = {
-    .name = TYPE_HVF_ACCEL,
-    .parent = TYPE_ACCEL,
-    .instance_size = sizeof(HVFState),
-    .class_init = hvf_accel_class_init,
-};
-
-static void hvf_type_init(void)
-{
-    type_register_static(&hvf_accel_type);
-}
-
-type_init(hvf_type_init);
-
 static void hvf_vcpu_destroy(CPUState *cpu)
 {
     hv_return_t ret = hv_vcpu_destroy(cpu->accel->fd);
@@ -656,8 +392,10 @@ static const TypeInfo hvf_accel_ops_type = {
     .class_init = hvf_accel_ops_class_init,
     .abstract = true,
 };
+
 static void hvf_accel_ops_register_types(void)
 {
     type_register_static(&hvf_accel_ops_type);
 }
+
 type_init(hvf_accel_ops_register_types);
diff --git a/accel/hvf/hvf-all.c b/accel/hvf/hvf-all.c
index 481d7dece57..66a4a201c34 100644
--- a/accel/hvf/hvf-all.c
+++ b/accel/hvf/hvf-all.c
@@ -10,8 +10,24 @@
 
 #include "qemu/osdep.h"
 #include "qemu/error-report.h"
+#include "system/address-spaces.h"
+#include "system/memory.h"
 #include "system/hvf.h"
 #include "system/hvf_int.h"
+#include "hw/core/cpu.h"
+#include "hw/boards.h"
+#include "trace.h"
+
+bool hvf_allowed;
+
+struct mac_slot {
+    int present;
+    uint64_t size;
+    uint64_t gpa_start;
+    uint64_t gva;
+};
+
+struct mac_slot mac_slots[32];
 
 const char *hvf_return_string(hv_return_t ret)
 {
@@ -40,3 +56,254 @@ void assert_hvf_ok_impl(hv_return_t ret, const char *file, unsigned int line,
 
     abort();
 }
+
+static int do_hvf_set_memory(hvf_slot *slot, hv_memory_flags_t flags)
+{
+    struct mac_slot *macslot;
+    hv_return_t ret;
+
+    macslot = &mac_slots[slot->slot_id];
+
+    if (macslot->present) {
+        if (macslot->size != slot->size) {
+            macslot->present = 0;
+            ret = hv_vm_unmap(macslot->gpa_start, macslot->size);
+            assert_hvf_ok(ret);
+        }
+    }
+
+    if (!slot->size) {
+        return 0;
+    }
+
+    macslot->present = 1;
+    macslot->gpa_start = slot->start;
+    macslot->size = slot->size;
+    ret = hv_vm_map(slot->mem, slot->start, slot->size, flags);
+    assert_hvf_ok(ret);
+    return 0;
+}
+
+static void hvf_set_phys_mem(MemoryRegionSection *section, bool add)
+{
+    hvf_slot *mem;
+    MemoryRegion *area = section->mr;
+    bool writable = !area->readonly && !area->rom_device;
+    hv_memory_flags_t flags;
+    uint64_t page_size = qemu_real_host_page_size();
+
+    if (!memory_region_is_ram(area)) {
+        if (writable) {
+            return;
+        } else if (!memory_region_is_romd(area)) {
+            /*
+             * If the memory device is not in romd_mode, then we actually want
+             * to remove the hvf memory slot so all accesses will trap.
+             */
+             add = false;
+        }
+    }
+
+    if (!QEMU_IS_ALIGNED(int128_get64(section->size), page_size) ||
+        !QEMU_IS_ALIGNED(section->offset_within_address_space, page_size)) {
+        /* Not page aligned, so we can not map as RAM */
+        add = false;
+    }
+
+    mem = hvf_find_overlap_slot(
+            section->offset_within_address_space,
+            int128_get64(section->size));
+
+    if (mem && add) {
+        if (mem->size == int128_get64(section->size) &&
+            mem->start == section->offset_within_address_space &&
+            mem->mem == (memory_region_get_ram_ptr(area) +
+            section->offset_within_region)) {
+            return; /* Same region was attempted to register, go away. */
+        }
+    }
+
+    /* Region needs to be reset. set the size to 0 and remap it. */
+    if (mem) {
+        mem->size = 0;
+        if (do_hvf_set_memory(mem, 0)) {
+            error_report("Failed to reset overlapping slot");
+            abort();
+        }
+    }
+
+    if (!add) {
+        return;
+    }
+
+    if (area->readonly ||
+        (!memory_region_is_ram(area) && memory_region_is_romd(area))) {
+        flags = HV_MEMORY_READ | HV_MEMORY_EXEC;
+    } else {
+        flags = HV_MEMORY_READ | HV_MEMORY_WRITE | HV_MEMORY_EXEC;
+    }
+
+    /* Now make a new slot. */
+    int x;
+
+    for (x = 0; x < hvf_state->num_slots; ++x) {
+        mem = &hvf_state->slots[x];
+        if (!mem->size) {
+            break;
+        }
+    }
+
+    if (x == hvf_state->num_slots) {
+        error_report("No free slots");
+        abort();
+    }
+
+    mem->size = int128_get64(section->size);
+    mem->mem = memory_region_get_ram_ptr(area) + section->offset_within_region;
+    mem->start = section->offset_within_address_space;
+    mem->region = area;
+
+    if (do_hvf_set_memory(mem, flags)) {
+        error_report("Error registering new memory slot");
+        abort();
+    }
+}
+
+static void hvf_set_dirty_tracking(MemoryRegionSection *section, bool on)
+{
+    hvf_slot *slot;
+
+    slot = hvf_find_overlap_slot(
+            section->offset_within_address_space,
+            int128_get64(section->size));
+
+    /* protect region against writes; begin tracking it */
+    if (on) {
+        slot->flags |= HVF_SLOT_LOG;
+        hv_vm_protect((uintptr_t)slot->start, (size_t)slot->size,
+                      HV_MEMORY_READ | HV_MEMORY_EXEC);
+    /* stop tracking region*/
+    } else {
+        slot->flags &= ~HVF_SLOT_LOG;
+        hv_vm_protect((uintptr_t)slot->start, (size_t)slot->size,
+                      HV_MEMORY_READ | HV_MEMORY_WRITE | HV_MEMORY_EXEC);
+    }
+}
+
+static void hvf_log_start(MemoryListener *listener,
+                          MemoryRegionSection *section, int old, int new)
+{
+    if (old != 0) {
+        return;
+    }
+
+    hvf_set_dirty_tracking(section, 1);
+}
+
+static void hvf_log_stop(MemoryListener *listener,
+                         MemoryRegionSection *section, int old, int new)
+{
+    if (new != 0) {
+        return;
+    }
+
+    hvf_set_dirty_tracking(section, 0);
+}
+
+static void hvf_log_sync(MemoryListener *listener,
+                         MemoryRegionSection *section)
+{
+    /*
+     * sync of dirty pages is handled elsewhere; just make sure we keep
+     * tracking the region.
+     */
+    hvf_set_dirty_tracking(section, 1);
+}
+
+static void hvf_region_add(MemoryListener *listener,
+                           MemoryRegionSection *section)
+{
+    hvf_set_phys_mem(section, true);
+}
+
+static void hvf_region_del(MemoryListener *listener,
+                           MemoryRegionSection *section)
+{
+    hvf_set_phys_mem(section, false);
+}
+
+static MemoryListener hvf_memory_listener = {
+    .name = "hvf",
+    .priority = MEMORY_LISTENER_PRIORITY_ACCEL,
+    .region_add = hvf_region_add,
+    .region_del = hvf_region_del,
+    .log_start = hvf_log_start,
+    .log_stop = hvf_log_stop,
+    .log_sync = hvf_log_sync,
+};
+
+static int hvf_accel_init(AccelState *as, MachineState *ms)
+{
+    int x;
+    hv_return_t ret;
+    HVFState *s = HVF_STATE(as);
+    int pa_range = 36;
+    MachineClass *mc = MACHINE_GET_CLASS(ms);
+
+    if (mc->hvf_get_physical_address_range) {
+        pa_range = mc->hvf_get_physical_address_range(ms);
+        if (pa_range < 0) {
+            return -EINVAL;
+        }
+    }
+
+    ret = hvf_arch_vm_create(ms, (uint32_t)pa_range);
+    if (ret == HV_DENIED) {
+        error_report("Could not access HVF. Is the executable signed"
+                     " with com.apple.security.hypervisor entitlement?");
+        exit(1);
+    }
+    assert_hvf_ok(ret);
+
+    s->num_slots = ARRAY_SIZE(s->slots);
+    for (x = 0; x < s->num_slots; ++x) {
+        s->slots[x].size = 0;
+        s->slots[x].slot_id = x;
+    }
+
+    QTAILQ_INIT(&s->hvf_sw_breakpoints);
+
+    hvf_state = s;
+    memory_listener_register(&hvf_memory_listener, &address_space_memory);
+
+    return hvf_arch_init();
+}
+
+static inline int hvf_gdbstub_sstep_flags(AccelState *as)
+{
+    return SSTEP_ENABLE | SSTEP_NOIRQ;
+}
+
+static void hvf_accel_class_init(ObjectClass *oc, const void *data)
+{
+    AccelClass *ac = ACCEL_CLASS(oc);
+    ac->name = "HVF";
+    ac->init_machine = hvf_accel_init;
+    ac->allowed = &hvf_allowed;
+    ac->supports_guest_debug = hvf_arch_supports_guest_debug;
+    ac->gdbstub_supported_sstep_flags = hvf_gdbstub_sstep_flags;
+}
+
+static const TypeInfo hvf_accel_type = {
+    .name = TYPE_HVF_ACCEL,
+    .parent = TYPE_ACCEL,
+    .instance_size = sizeof(HVFState),
+    .class_init = hvf_accel_class_init,
+};
+
+static void hvf_type_init(void)
+{
+    type_register_static(&hvf_accel_type);
+}
+
+type_init(hvf_type_init);
-- 
2.49.0

Re: [PATCH v3 31/68] accel/hvf: Move generic method declarations to hvf-all.c

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> hvf-all.c aims to contain the generic accel methods (TYPE_ACCEL),
> while hvf-accel-ops.c the per-vcpu methods (TYPE_ACCEL_OPS).
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   accel/hvf/hvf-accel-ops.c | 272 +-------------------------------------
>   accel/hvf/hvf-all.c       | 267 +++++++++++++++++++++++++++++++++++++
>   2 files changed, 272 insertions(+), 267 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 32/68] cpus: Document CPUState::vcpu_dirty field

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/hw/core/cpu.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 33296a1c080..726427449da 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -441,6 +441,7 @@ struct qemu_work_item;
  * @opaque: User data.
  * @mem_io_pc: Host Program Counter at which the memory was accessed.
  * @accel: Pointer to accelerator specific state.
+ * @vcpu_dirty: Hardware accelerator is not synchronized with QEMU state
  * @kvm_fd: vCPU file descriptor for KVM.
  * @work_mutex: Lock to prevent multiple access to @work_list.
  * @work_list: List of pending asynchronous work.
@@ -537,7 +538,6 @@ struct CPUState {
     uint32_t kvm_fetch_index;
     uint64_t dirty_pages;
     int kvm_vcpu_stats_fd;
-    bool vcpu_dirty;
 
     /* Use by accel-block: CPU is executing an ioctl() */
     QemuLockCnt in_ioctl_lock;
@@ -553,6 +553,7 @@ struct CPUState {
     uint32_t halted;
     int32_t exception_index;
 
+    bool vcpu_dirty;
     AccelCPUState *accel;
 
     /* Used to keep track of an outstanding cpu throttle thread for migration
-- 
2.49.0

[PATCH v3 33/68] accel/hvf: Replace @dirty field by generic CPUState::vcpu_dirty field

[Philippe Mathieu-Daudé]

No need for accel-specific @dirty field when we have
a generic one in CPUState.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 include/system/hvf_int.h  |  1 -
 accel/hvf/hvf-accel-ops.c | 10 +++++-----
 target/arm/hvf/hvf.c      |  4 ++--
 target/i386/hvf/hvf.c     |  4 ++--
 target/i386/hvf/x86hvf.c  |  2 +-
 5 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/include/system/hvf_int.h b/include/system/hvf_int.h
index ea6730f255d..a8ee7c7dae6 100644
--- a/include/system/hvf_int.h
+++ b/include/system/hvf_int.h
@@ -62,7 +62,6 @@ struct AccelCPUState {
     bool vtimer_masked;
     sigset_t unblock_ipi_mask;
     bool guest_debug_enabled;
-    bool dirty;
 };
 
 void assert_hvf_ok_impl(hv_return_t ret, const char *file, unsigned int line,
diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index 319c30f703c..c91e18bc3dd 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -79,15 +79,15 @@ hvf_slot *hvf_find_overlap_slot(uint64_t start, uint64_t size)
 
 static void do_hvf_cpu_synchronize_state(CPUState *cpu, run_on_cpu_data arg)
 {
-    if (!cpu->accel->dirty) {
+    if (!cpu->vcpu_dirty) {
         hvf_get_registers(cpu);
-        cpu->accel->dirty = true;
+        cpu->vcpu_dirty = true;
     }
 }
 
 static void hvf_cpu_synchronize_state(CPUState *cpu)
 {
-    if (!cpu->accel->dirty) {
+    if (!cpu->vcpu_dirty) {
         run_on_cpu(cpu, do_hvf_cpu_synchronize_state, RUN_ON_CPU_NULL);
     }
 }
@@ -96,7 +96,7 @@ static void do_hvf_cpu_synchronize_set_dirty(CPUState *cpu,
                                              run_on_cpu_data arg)
 {
     /* QEMU state is the reference, push it to HVF now and on next entry */
-    cpu->accel->dirty = true;
+    cpu->vcpu_dirty = true;
 }
 
 static void hvf_cpu_synchronize_post_reset(CPUState *cpu)
@@ -156,8 +156,8 @@ static int hvf_init_vcpu(CPUState *cpu)
 #else
     r = hv_vcpu_create(&cpu->accel->fd, HV_VCPU_DEFAULT);
 #endif
-    cpu->accel->dirty = true;
     assert_hvf_ok(r);
+    cpu->vcpu_dirty = true;
 
     cpu->accel->guest_debug_enabled = false;
 
diff --git a/target/arm/hvf/hvf.c b/target/arm/hvf/hvf.c
index 3fb0b49df8a..c3876cead18 100644
--- a/target/arm/hvf/hvf.c
+++ b/target/arm/hvf/hvf.c
@@ -812,9 +812,9 @@ int hvf_put_registers(CPUState *cpu)
 
 static void flush_cpu_state(CPUState *cpu)
 {
-    if (cpu->accel->dirty) {
+    if (cpu->vcpu_dirty) {
         hvf_put_registers(cpu);
-        cpu->accel->dirty = false;
+        cpu->vcpu_dirty = false;
     }
 }
 
diff --git a/target/i386/hvf/hvf.c b/target/i386/hvf/hvf.c
index bcf30662bec..c893aaac1b0 100644
--- a/target/i386/hvf/hvf.c
+++ b/target/i386/hvf/hvf.c
@@ -738,9 +738,9 @@ int hvf_vcpu_exec(CPUState *cpu)
     }
 
     do {
-        if (cpu->accel->dirty) {
+        if (cpu->vcpu_dirty) {
             hvf_put_registers(cpu);
-            cpu->accel->dirty = false;
+            cpu->vcpu_dirty = false;
         }
 
         if (hvf_inject_interrupts(cpu)) {
diff --git a/target/i386/hvf/x86hvf.c b/target/i386/hvf/x86hvf.c
index 2057314892a..17fce1d3cdd 100644
--- a/target/i386/hvf/x86hvf.c
+++ b/target/i386/hvf/x86hvf.c
@@ -427,7 +427,7 @@ int hvf_process_events(CPUState *cs)
     X86CPU *cpu = X86_CPU(cs);
     CPUX86State *env = &cpu->env;
 
-    if (!cs->accel->dirty) {
+    if (!cs->vcpu_dirty) {
         /* light weight sync for CPU_INTERRUPT_HARD and IF_MASK */
         env->eflags = rreg(cs->accel->fd, HV_X86_RFLAGS);
     }
-- 
2.49.0

Re: [PATCH v3 33/68] accel/hvf: Replace @dirty field by generic CPUState::vcpu_dirty field

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> No need for accel-specific @dirty field when we have
> a generic one in CPUState.
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   include/system/hvf_int.h  |  1 -
>   accel/hvf/hvf-accel-ops.c | 10 +++++-----
>   target/arm/hvf/hvf.c      |  4 ++--
>   target/i386/hvf/hvf.c     |  4 ++--
>   target/i386/hvf/x86hvf.c  |  2 +-
>   5 files changed, 10 insertions(+), 11 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 34/68] accel/nvmm: Replace @dirty field by generic CPUState::vcpu_dirty field

[Philippe Mathieu-Daudé]

No need for accel-specific @dirty field when we have
a generic one in CPUState.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 target/i386/nvmm/nvmm-all.c | 21 ++++++++++-----------
 1 file changed, 10 insertions(+), 11 deletions(-)

diff --git a/target/i386/nvmm/nvmm-all.c b/target/i386/nvmm/nvmm-all.c
index eaae175aa5d..f521c36dc53 100644
--- a/target/i386/nvmm/nvmm-all.c
+++ b/target/i386/nvmm/nvmm-all.c
@@ -30,7 +30,6 @@ struct AccelCPUState {
     struct nvmm_vcpu vcpu;
     uint8_t tpr;
     bool stop;
-    bool dirty;
 
     /* Window-exiting for INTs/NMIs. */
     bool int_window_exit;
@@ -508,7 +507,7 @@ nvmm_io_callback(struct nvmm_io *io)
     }
 
     /* Needed, otherwise infinite loop. */
-    current_cpu->accel->dirty = false;
+    current_cpu->vcpu_dirty = false;
 }
 
 static void
@@ -517,7 +516,7 @@ nvmm_mem_callback(struct nvmm_mem *mem)
     cpu_physical_memory_rw(mem->gpa, mem->data, mem->size, mem->write);
 
     /* Needed, otherwise infinite loop. */
-    current_cpu->accel->dirty = false;
+    current_cpu->vcpu_dirty = false;
 }
 
 static struct nvmm_assist_callbacks nvmm_callbacks = {
@@ -727,9 +726,9 @@ nvmm_vcpu_loop(CPUState *cpu)
      * Inner VCPU loop.
      */
     do {
-        if (cpu->accel->dirty) {
+        if (cpu->vcpu_dirty) {
             nvmm_set_registers(cpu);
-            cpu->accel->dirty = false;
+            cpu->vcpu_dirty = false;
         }
 
         if (qcpu->stop) {
@@ -827,32 +826,32 @@ static void
 do_nvmm_cpu_synchronize_state(CPUState *cpu, run_on_cpu_data arg)
 {
     nvmm_get_registers(cpu);
-    cpu->accel->dirty = true;
+    cpu->vcpu_dirty = true;
 }
 
 static void
 do_nvmm_cpu_synchronize_post_reset(CPUState *cpu, run_on_cpu_data arg)
 {
     nvmm_set_registers(cpu);
-    cpu->accel->dirty = false;
+    cpu->vcpu_dirty = false;
 }
 
 static void
 do_nvmm_cpu_synchronize_post_init(CPUState *cpu, run_on_cpu_data arg)
 {
     nvmm_set_registers(cpu);
-    cpu->accel->dirty = false;
+    cpu->vcpu_dirty = false;
 }
 
 static void
 do_nvmm_cpu_synchronize_pre_loadvm(CPUState *cpu, run_on_cpu_data arg)
 {
-    cpu->accel->dirty = true;
+    cpu->vcpu_dirty = true;
 }
 
 void nvmm_cpu_synchronize_state(CPUState *cpu)
 {
-    if (!cpu->accel->dirty) {
+    if (!cpu->vcpu_dirty) {
         run_on_cpu(cpu, do_nvmm_cpu_synchronize_state, RUN_ON_CPU_NULL);
     }
 }
@@ -982,7 +981,7 @@ nvmm_init_vcpu(CPUState *cpu)
         }
     }
 
-    qcpu->dirty = true;
+    qcpu->vcpu_dirty = true;
     cpu->accel = qcpu;
 
     return 0;
-- 
2.49.0

Re: [PATCH v3 34/68] accel/nvmm: Replace @dirty field by generic CPUState::vcpu_dirty field

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> No need for accel-specific @dirty field when we have
> a generic one in CPUState.
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   target/i386/nvmm/nvmm-all.c | 21 ++++++++++-----------
>   1 file changed, 10 insertions(+), 11 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 35/68] accel/whpx: Replace @dirty field by generic CPUState::vcpu_dirty field

[Philippe Mathieu-Daudé]

No need for accel-specific @dirty field when we have
a generic one in CPUState.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 target/i386/whpx/whpx-all.c | 23 +++++++++++------------
 1 file changed, 11 insertions(+), 12 deletions(-)

diff --git a/target/i386/whpx/whpx-all.c b/target/i386/whpx/whpx-all.c
index 821167a2a77..f812aa36c46 100644
--- a/target/i386/whpx/whpx-all.c
+++ b/target/i386/whpx/whpx-all.c
@@ -237,7 +237,6 @@ struct AccelCPUState {
     uint64_t tpr;
     uint64_t apic_base;
     bool interruption_pending;
-    bool dirty;
 
     /* Must be the last field as it may have a tail */
     WHV_RUN_VP_EXIT_CONTEXT exit_ctx;
@@ -836,7 +835,7 @@ static HRESULT CALLBACK whpx_emu_setreg_callback(
      * The emulator just successfully wrote the register state. We clear the
      * dirty state so we avoid the double write on resume of the VP.
      */
-    cpu->accel->dirty = false;
+    cpu->vcpu_dirty = false;
 
     return hr;
 }
@@ -1391,7 +1390,7 @@ static int whpx_last_vcpu_stopping(CPUState *cpu)
 /* Returns the address of the next instruction that is about to be executed. */
 static vaddr whpx_vcpu_get_pc(CPUState *cpu, bool exit_context_valid)
 {
-    if (cpu->accel->dirty) {
+    if (cpu->vcpu_dirty) {
         /* The CPU registers have been modified by other parts of QEMU. */
         return cpu_env(cpu)->eip;
     } else if (exit_context_valid) {
@@ -1704,9 +1703,9 @@ static int whpx_vcpu_run(CPUState *cpu)
     }
 
     do {
-        if (cpu->accel->dirty) {
+        if (cpu->vcpu_dirty) {
             whpx_set_registers(cpu, WHPX_SET_RUNTIME_STATE);
-            cpu->accel->dirty = false;
+            cpu->vcpu_dirty = false;
         }
 
         if (exclusive_step_mode == WHPX_STEP_NONE) {
@@ -2054,9 +2053,9 @@ static int whpx_vcpu_run(CPUState *cpu)
 
 static void do_whpx_cpu_synchronize_state(CPUState *cpu, run_on_cpu_data arg)
 {
-    if (!cpu->accel->dirty) {
+    if (!cpu->vcpu_dirty) {
         whpx_get_registers(cpu);
-        cpu->accel->dirty = true;
+        cpu->vcpu_dirty = true;
     }
 }
 
@@ -2064,20 +2063,20 @@ static void do_whpx_cpu_synchronize_post_reset(CPUState *cpu,
                                                run_on_cpu_data arg)
 {
     whpx_set_registers(cpu, WHPX_SET_RESET_STATE);
-    cpu->accel->dirty = false;
+    cpu->vcpu_dirty = false;
 }
 
 static void do_whpx_cpu_synchronize_post_init(CPUState *cpu,
                                               run_on_cpu_data arg)
 {
     whpx_set_registers(cpu, WHPX_SET_FULL_STATE);
-    cpu->accel->dirty = false;
+    cpu->vcpu_dirty = false;
 }
 
 static void do_whpx_cpu_synchronize_pre_loadvm(CPUState *cpu,
                                                run_on_cpu_data arg)
 {
-    cpu->accel->dirty = true;
+    cpu->vcpu_dirty = true;
 }
 
 /*
@@ -2086,7 +2085,7 @@ static void do_whpx_cpu_synchronize_pre_loadvm(CPUState *cpu,
 
 void whpx_cpu_synchronize_state(CPUState *cpu)
 {
-    if (!cpu->accel->dirty) {
+    if (!cpu->vcpu_dirty) {
         run_on_cpu(cpu, do_whpx_cpu_synchronize_state, RUN_ON_CPU_NULL);
     }
 }
@@ -2226,7 +2225,7 @@ int whpx_init_vcpu(CPUState *cpu)
     }
 
     vcpu->interruptable = true;
-    vcpu->dirty = true;
+    vcpu->vcpu_dirty = true;
     cpu->accel = vcpu;
     max_vcpu_index = max(max_vcpu_index, cpu->cpu_index);
     qemu_add_vm_change_state_handler(whpx_cpu_update_state, env);
-- 
2.49.0

Re: [PATCH v3 35/68] accel/whpx: Replace @dirty field by generic CPUState::vcpu_dirty field

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> No need for accel-specific @dirty field when we have
> a generic one in CPUState.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   target/i386/whpx/whpx-all.c | 23 +++++++++++------------
>   1 file changed, 11 insertions(+), 12 deletions(-)
...
> @@ -2226,7 +2225,7 @@ int whpx_init_vcpu(CPUState *cpu)
>       }
>   
>       vcpu->interruptable = true;
> -    vcpu->dirty = true;
> +    vcpu->vcpu_dirty = true;

cpu->vcpu_dirty

Otherwise,
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [PATCH v3 35/68] accel/whpx: Replace @dirty field by generic CPUState::vcpu_dirty field

[Philippe Mathieu-Daudé]

On 2/7/25 17:11, Richard Henderson wrote:
> On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
>> No need for accel-specific @dirty field when we have
>> a generic one in CPUState.
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   target/i386/whpx/whpx-all.c | 23 +++++++++++------------
>>   1 file changed, 11 insertions(+), 12 deletions(-)
> ...
>> @@ -2226,7 +2225,7 @@ int whpx_init_vcpu(CPUState *cpu)
>>       }
>>       vcpu->interruptable = true;
>> -    vcpu->dirty = true;
>> +    vcpu->vcpu_dirty = true;
> 
> cpu->vcpu_dirty

Nice catch!

> 
> Otherwise,
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> 
> 
> r~

[PATCH v3 36/68] accel/kvm: Remove kvm_cpu_synchronize_state() stub

[Philippe Mathieu-Daudé]

Since commit 57038a92bb0 ("cpus: extract out kvm-specific code
to accel/kvm") the kvm_cpu_synchronize_state() stub is not
necessary.

Fixes: e0715f6abce ("kvm: remove kvm specific functions from global includes")
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/stubs/kvm-stub.c | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/accel/stubs/kvm-stub.c b/accel/stubs/kvm-stub.c
index b9b4427c919..68cd33ba973 100644
--- a/accel/stubs/kvm-stub.c
+++ b/accel/stubs/kvm-stub.c
@@ -29,10 +29,6 @@ void kvm_flush_coalesced_mmio_buffer(void)
 {
 }
 
-void kvm_cpu_synchronize_state(CPUState *cpu)
-{
-}
-
 bool kvm_has_sync_mmu(void)
 {
     return false;
-- 
2.49.0

[PATCH v3 37/68] accel/system: Document cpu_synchronize_state()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/accel-ops.h |  8 ++++++++
 include/system/hw_accel.h  | 13 +++++++++++--
 2 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index 106ff56d880..6f9aebbee8f 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -45,6 +45,14 @@ struct AccelOpsClass {
 
     void (*synchronize_post_reset)(CPUState *cpu);
     void (*synchronize_post_init)(CPUState *cpu);
+    /**
+     * synchronize_state:
+     * synchronize_pre_loadvm:
+     * @cpu: The vCPU to synchronize.
+     *
+     * Request to synchronize QEMU vCPU registers from the hardware accelerator
+     * (the hardware accelerator is the reference).
+     */
     void (*synchronize_state)(CPUState *cpu);
     void (*synchronize_pre_loadvm)(CPUState *cpu);
 
diff --git a/include/system/hw_accel.h b/include/system/hw_accel.h
index 380e9e640b6..574c9738408 100644
--- a/include/system/hw_accel.h
+++ b/include/system/hw_accel.h
@@ -17,9 +17,18 @@
 #include "system/whpx.h"
 #include "system/nvmm.h"
 
+/**
+ * cpu_synchronize_state:
+ * cpu_synchronize_pre_loadvm:
+ * @cpu: The vCPU to synchronize.
+ *
+ * Request to synchronize QEMU vCPU registers from the hardware accelerator
+ * (the hardware accelerator is the reference).
+ */
 void cpu_synchronize_state(CPUState *cpu);
-void cpu_synchronize_post_reset(CPUState *cpu);
-void cpu_synchronize_post_init(CPUState *cpu);
 void cpu_synchronize_pre_loadvm(CPUState *cpu);
 
+void cpu_synchronize_post_reset(CPUState *cpu);
+void cpu_synchronize_post_init(CPUState *cpu);
+
 #endif /* QEMU_HW_ACCEL_H */
-- 
2.49.0

[PATCH v3 38/68] accel/system: Document cpu_synchronize_state_post_init/reset()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/accel-ops.h | 8 ++++++++
 include/system/hw_accel.h  | 8 ++++++++
 2 files changed, 16 insertions(+)

diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index 6f9aebbee8f..14113bb5c10 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -43,6 +43,14 @@ struct AccelOpsClass {
     void (*kick_vcpu_thread)(CPUState *cpu);
     bool (*cpu_thread_is_idle)(CPUState *cpu);
 
+    /**
+     * synchronize_post_reset:
+     * synchronize_post_init:
+     * @cpu: The vCPU to synchronize.
+     *
+     * Request to synchronize QEMU vCPU registers to the hardware accelerator
+     * (QEMU is the reference).
+     */
     void (*synchronize_post_reset)(CPUState *cpu);
     void (*synchronize_post_init)(CPUState *cpu);
     /**
diff --git a/include/system/hw_accel.h b/include/system/hw_accel.h
index 574c9738408..fa9228d5d2d 100644
--- a/include/system/hw_accel.h
+++ b/include/system/hw_accel.h
@@ -28,6 +28,14 @@
 void cpu_synchronize_state(CPUState *cpu);
 void cpu_synchronize_pre_loadvm(CPUState *cpu);
 
+/**
+ * cpu_synchronize_post_reset:
+ * cpu_synchronize_post_init:
+ * @cpu: The vCPU to synchronize.
+ *
+ * Request to synchronize QEMU vCPU registers to the hardware accelerator
+ * (QEMU is the reference).
+ */
 void cpu_synchronize_post_reset(CPUState *cpu);
 void cpu_synchronize_post_init(CPUState *cpu);
 
-- 
2.49.0

[PATCH v3 39/68] accel/nvmm: Expose nvmm_enabled() to common code

[Philippe Mathieu-Daudé]

Currently nvmm_enabled() is restricted to target-specific code.
By defining CONFIG_NVMM_IS_POSSIBLE we allow its use anywhere.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/nvmm.h       | 23 ++++++++++++-----------
 accel/stubs/nvmm-stub.c     | 12 ++++++++++++
 target/i386/nvmm/nvmm-all.c |  6 ------
 accel/stubs/meson.build     |  1 +
 4 files changed, 25 insertions(+), 17 deletions(-)
 create mode 100644 accel/stubs/nvmm-stub.c

diff --git a/include/system/nvmm.h b/include/system/nvmm.h
index 6971ddb3a5a..7390def9adb 100644
--- a/include/system/nvmm.h
+++ b/include/system/nvmm.h
@@ -13,17 +13,18 @@
 #define QEMU_NVMM_H
 
 #ifdef COMPILING_PER_TARGET
-
-#ifdef CONFIG_NVMM
-
-int nvmm_enabled(void);
-
-#else /* CONFIG_NVMM */
-
-#define nvmm_enabled() (0)
-
-#endif /* CONFIG_NVMM */
-
+# ifdef CONFIG_NVMM
+#  define CONFIG_NVMM_IS_POSSIBLE
+# endif /* !CONFIG_NVMM */
+#else
+# define CONFIG_NVMM_IS_POSSIBLE
 #endif /* COMPILING_PER_TARGET */
 
+#ifdef CONFIG_NVMM_IS_POSSIBLE
+extern bool nvmm_allowed;
+#define nvmm_enabled() (nvmm_allowed)
+#else /* !CONFIG_NVMM_IS_POSSIBLE */
+#define nvmm_enabled() 0
+#endif /* !CONFIG_NVMM_IS_POSSIBLE */
+
 #endif /* QEMU_NVMM_H */
diff --git a/accel/stubs/nvmm-stub.c b/accel/stubs/nvmm-stub.c
new file mode 100644
index 00000000000..cc58114ceb3
--- /dev/null
+++ b/accel/stubs/nvmm-stub.c
@@ -0,0 +1,12 @@
+/*
+ * NVMM stubs for QEMU
+ *
+ *  Copyright (c) Linaro
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include "qemu/osdep.h"
+#include "system/hvf.h"
+
+bool nvmm_allowed;
diff --git a/target/i386/nvmm/nvmm-all.c b/target/i386/nvmm/nvmm-all.c
index f521c36dc53..a392d3fc232 100644
--- a/target/i386/nvmm/nvmm-all.c
+++ b/target/i386/nvmm/nvmm-all.c
@@ -1192,12 +1192,6 @@ nvmm_accel_init(AccelState *as, MachineState *ms)
     return 0;
 }
 
-int
-nvmm_enabled(void)
-{
-    return nvmm_allowed;
-}
-
 static void
 nvmm_accel_class_init(ObjectClass *oc, const void *data)
 {
diff --git a/accel/stubs/meson.build b/accel/stubs/meson.build
index 8ca1a4529e2..4c34287215f 100644
--- a/accel/stubs/meson.build
+++ b/accel/stubs/meson.build
@@ -3,5 +3,6 @@ system_stubs_ss.add(when: 'CONFIG_XEN', if_false: files('xen-stub.c'))
 system_stubs_ss.add(when: 'CONFIG_KVM', if_false: files('kvm-stub.c'))
 system_stubs_ss.add(when: 'CONFIG_TCG', if_false: files('tcg-stub.c'))
 system_stubs_ss.add(when: 'CONFIG_HVF', if_false: files('hvf-stub.c'))
+system_stubs_ss.add(when: 'CONFIG_NVMM', if_false: files('nvmm-stub.c'))
 
 specific_ss.add_all(when: ['CONFIG_SYSTEM_ONLY'], if_true: system_stubs_ss)
-- 
2.49.0

[PATCH v3 40/68] accel/whpx: Expose whpx_enabled() to common code

[Philippe Mathieu-Daudé]

Currently whpx_enabled() is restricted to target-specific code.
By defining CONFIG_WHPX_IS_POSSIBLE we allow its use anywhere.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/whpx.h       | 27 ++++++++++++++-------------
 accel/stubs/whpx-stub.c     | 12 ++++++++++++
 target/i386/whpx/whpx-all.c |  5 -----
 accel/stubs/meson.build     |  1 +
 4 files changed, 27 insertions(+), 18 deletions(-)
 create mode 100644 accel/stubs/whpx-stub.c

diff --git a/include/system/whpx.h b/include/system/whpx.h
index 00ff409b682..00f6a3e5236 100644
--- a/include/system/whpx.h
+++ b/include/system/whpx.h
@@ -16,19 +16,20 @@
 #define QEMU_WHPX_H
 
 #ifdef COMPILING_PER_TARGET
-
-#ifdef CONFIG_WHPX
-
-int whpx_enabled(void);
-bool whpx_apic_in_platform(void);
-
-#else /* CONFIG_WHPX */
-
-#define whpx_enabled() (0)
-#define whpx_apic_in_platform() (0)
-
-#endif /* CONFIG_WHPX */
-
+# ifdef CONFIG_WHPX
+#  define CONFIG_WHPX_IS_POSSIBLE
+# endif /* !CONFIG_WHPX */
+#else
+# define CONFIG_WHPX_IS_POSSIBLE
 #endif /* COMPILING_PER_TARGET */
 
+#ifdef CONFIG_WHPX_IS_POSSIBLE
+extern bool whpx_allowed;
+#define whpx_enabled() (whpx_allowed)
+bool whpx_apic_in_platform(void);
+#else /* !CONFIG_WHPX_IS_POSSIBLE */
+#define whpx_enabled() 0
+#define whpx_apic_in_platform() (0)
+#endif /* !CONFIG_WHPX_IS_POSSIBLE */
+
 #endif /* QEMU_WHPX_H */
diff --git a/accel/stubs/whpx-stub.c b/accel/stubs/whpx-stub.c
new file mode 100644
index 00000000000..c564c89fd0b
--- /dev/null
+++ b/accel/stubs/whpx-stub.c
@@ -0,0 +1,12 @@
+/*
+ * WHPX stubs for QEMU
+ *
+ *  Copyright (c) Linaro
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include "qemu/osdep.h"
+#include "system/whpx.h"
+
+bool whpx_allowed;
diff --git a/target/i386/whpx/whpx-all.c b/target/i386/whpx/whpx-all.c
index f812aa36c46..4052fadb97c 100644
--- a/target/i386/whpx/whpx-all.c
+++ b/target/i386/whpx/whpx-all.c
@@ -2688,11 +2688,6 @@ error:
     return ret;
 }
 
-int whpx_enabled(void)
-{
-    return whpx_allowed;
-}
-
 bool whpx_apic_in_platform(void) {
     return whpx_global.apic_in_platform;
 }
diff --git a/accel/stubs/meson.build b/accel/stubs/meson.build
index 4c34287215f..9dfc4f9ddaf 100644
--- a/accel/stubs/meson.build
+++ b/accel/stubs/meson.build
@@ -4,5 +4,6 @@ system_stubs_ss.add(when: 'CONFIG_KVM', if_false: files('kvm-stub.c'))
 system_stubs_ss.add(when: 'CONFIG_TCG', if_false: files('tcg-stub.c'))
 system_stubs_ss.add(when: 'CONFIG_HVF', if_false: files('hvf-stub.c'))
 system_stubs_ss.add(when: 'CONFIG_NVMM', if_false: files('nvmm-stub.c'))
+system_stubs_ss.add(when: 'CONFIG_WHPX', if_false: files('whpx-stub.c'))
 
 specific_ss.add_all(when: ['CONFIG_SYSTEM_ONLY'], if_true: system_stubs_ss)
-- 
2.49.0

[PATCH v3 41/68] accel/system: Introduce hwaccel_enabled() helper

[Philippe Mathieu-Daudé]

hwaccel_enabled() return whether any hardware accelerator
is available.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/hw_accel.h | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/include/system/hw_accel.h b/include/system/hw_accel.h
index fa9228d5d2d..49556b026e0 100644
--- a/include/system/hw_accel.h
+++ b/include/system/hw_accel.h
@@ -39,4 +39,17 @@ void cpu_synchronize_pre_loadvm(CPUState *cpu);
 void cpu_synchronize_post_reset(CPUState *cpu);
 void cpu_synchronize_post_init(CPUState *cpu);
 
+/**
+ * hwaccel_enabled:
+ *
+ * Returns: %true if a hardware accelerator is enabled, %false otherwise.
+ */
+static inline bool hwaccel_enabled(void)
+{
+    return hvf_enabled()
+        || kvm_enabled()
+        || nvmm_enabled()
+        || whpx_enabled();
+}
+
 #endif /* QEMU_HW_ACCEL_H */
-- 
2.49.0

[PATCH v3 42/68] target/arm: Use generic hwaccel_enabled() to check 'host' cpu type

[Philippe Mathieu-Daudé]

We should be able to use the 'host' CPU with any hardware accelerator.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 target/arm/arm-qmp-cmds.c | 5 +++--
 target/arm/cpu.c          | 5 +++--
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/target/arm/arm-qmp-cmds.c b/target/arm/arm-qmp-cmds.c
index cefd2352638..ee5eb1bac9f 100644
--- a/target/arm/arm-qmp-cmds.c
+++ b/target/arm/arm-qmp-cmds.c
@@ -30,6 +30,7 @@
 #include "qapi/qapi-commands-misc-arm.h"
 #include "qobject/qdict.h"
 #include "qom/qom-qobject.h"
+#include "system/hw_accel.h"
 #include "cpu.h"
 
 static GICCapability *gic_cap_new(int version)
@@ -116,8 +117,8 @@ CpuModelExpansionInfo *qmp_query_cpu_model_expansion(CpuModelExpansionType type,
         return NULL;
     }
 
-    if (!kvm_enabled() && !strcmp(model->name, "host")) {
-        error_setg(errp, "The CPU type '%s' requires KVM", model->name);
+    if (!hwaccel_enabled() && !strcmp(model->name, "host")) {
+        error_setg(errp, "The CPU type 'host' requires hardware accelerator");
         return NULL;
     }
 
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index e025e241eda..f3bde82b3a6 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -1976,8 +1976,9 @@ static void arm_cpu_realizefn(DeviceState *dev, Error **errp)
      * this is the first point where we can report it.
      */
     if (cpu->host_cpu_probe_failed) {
-        if (!kvm_enabled() && !hvf_enabled()) {
-            error_setg(errp, "The 'host' CPU type can only be used with KVM or HVF");
+        if (!hwaccel_enabled()) {
+            error_setg(errp, "The 'host' CPU type can only be used with "
+                             "hardware accelator such KVM/HVF");
         } else {
             error_setg(errp, "Failed to retrieve host CPU features");
         }
-- 
2.49.0

Re: [PATCH v3 42/68] target/arm: Use generic hwaccel_enabled() to check 'host' cpu type

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> We should be able to use the 'host' CPU with any hardware accelerator.
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   target/arm/arm-qmp-cmds.c | 5 +++--
>   target/arm/cpu.c          | 5 +++--
>   2 files changed, 6 insertions(+), 4 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 43/68] accel/dummy: Extract 'dummy-cpus.h' header from 'system/cpus.h'

[Philippe Mathieu-Daudé]

'dummy' helpers are specific to accelerator implementations,
no need to expose them via "system/cpus.h".

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/dummy-cpus.h    | 14 ++++++++++++++
 include/system/cpus.h |  5 -----
 accel/dummy-cpus.c    |  1 +
 accel/qtest/qtest.c   |  1 +
 4 files changed, 16 insertions(+), 5 deletions(-)
 create mode 100644 accel/dummy-cpus.h

diff --git a/accel/dummy-cpus.h b/accel/dummy-cpus.h
new file mode 100644
index 00000000000..d18dd0fdc51
--- /dev/null
+++ b/accel/dummy-cpus.h
@@ -0,0 +1,14 @@
+/*
+ * Dummy cpu thread code
+ *
+ * Copyright IBM, Corp. 2011
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#ifndef ACCEL_DUMMY_CPUS_H
+#define ACCEL_DUMMY_CPUS_H
+
+void dummy_start_vcpu_thread(CPUState *cpu);
+
+#endif
diff --git a/include/system/cpus.h b/include/system/cpus.h
index 3226c765d01..69be6a77a75 100644
--- a/include/system/cpus.h
+++ b/include/system/cpus.h
@@ -7,11 +7,6 @@ void cpus_register_accel(const AccelOpsClass *i);
 /* return registers ops */
 const AccelOpsClass *cpus_get_accel(void);
 
-/* accel/dummy-cpus.c */
-
-/* Create a dummy vcpu for AccelOpsClass->create_vcpu_thread */
-void dummy_start_vcpu_thread(CPUState *);
-
 /* interface available for cpus accelerator threads */
 
 /* For temporary buffers for forming a name */
diff --git a/accel/dummy-cpus.c b/accel/dummy-cpus.c
index 867276144fa..03cfc0fa01e 100644
--- a/accel/dummy-cpus.c
+++ b/accel/dummy-cpus.c
@@ -17,6 +17,7 @@
 #include "qemu/guest-random.h"
 #include "qemu/main-loop.h"
 #include "hw/core/cpu.h"
+#include "accel/dummy-cpus.h"
 
 static void *dummy_cpu_thread_fn(void *arg)
 {
diff --git a/accel/qtest/qtest.c b/accel/qtest/qtest.c
index 8b109d4c03b..2606fe97b49 100644
--- a/accel/qtest/qtest.c
+++ b/accel/qtest/qtest.c
@@ -24,6 +24,7 @@
 #include "qemu/guest-random.h"
 #include "qemu/main-loop.h"
 #include "hw/core/cpu.h"
+#include "accel/dummy-cpus.h"
 
 static int64_t qtest_clock_counter;
 
-- 
2.49.0

Re: [PATCH v3 43/68] accel/dummy: Extract 'dummy-cpus.h' header from 'system/cpus.h'

[Richard Henderson]

On 7/1/25 08:39, Philippe Mathieu-Daudé wrote:
> 'dummy' helpers are specific to accelerator implementations,
> no need to expose them via "system/cpus.h".
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   accel/dummy-cpus.h    | 14 ++++++++++++++
>   include/system/cpus.h |  5 -----
>   accel/dummy-cpus.c    |  1 +
>   accel/qtest/qtest.c   |  1 +
>   4 files changed, 16 insertions(+), 5 deletions(-)
>   create mode 100644 accel/dummy-cpus.h

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 44/68] accel/dummy: Factor dummy_thread_precreate() out

[Philippe Mathieu-Daudé]

Initialize the semaphore before creating the thread,
factor out as dummy_thread_precreate().

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/dummy-cpus.h |  1 +
 accel/dummy-cpus.c | 12 +++++++++---
 2 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/accel/dummy-cpus.h b/accel/dummy-cpus.h
index d18dd0fdc51..c3af710ee8c 100644
--- a/accel/dummy-cpus.h
+++ b/accel/dummy-cpus.h
@@ -9,6 +9,7 @@
 #ifndef ACCEL_DUMMY_CPUS_H
 #define ACCEL_DUMMY_CPUS_H
 
+void dummy_thread_precreate(CPUState *cpu);
 void dummy_start_vcpu_thread(CPUState *cpu);
 
 #endif
diff --git a/accel/dummy-cpus.c b/accel/dummy-cpus.c
index 03cfc0fa01e..2cbc3fecc93 100644
--- a/accel/dummy-cpus.c
+++ b/accel/dummy-cpus.c
@@ -65,15 +65,21 @@ static void *dummy_cpu_thread_fn(void *arg)
     return NULL;
 }
 
+void dummy_thread_precreate(CPUState *cpu)
+{
+#ifdef _WIN32
+    qemu_sem_init(&cpu->sem, 0);
+#endif
+}
+
 void dummy_start_vcpu_thread(CPUState *cpu)
 {
     char thread_name[VCPU_THREAD_NAME_SIZE];
 
+    dummy_thread_precreate(cpu);
+
     snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/DUMMY",
              cpu->cpu_index);
     qemu_thread_create(cpu->thread, thread_name, dummy_cpu_thread_fn, cpu,
                        QEMU_THREAD_JOINABLE);
-#ifdef _WIN32
-    qemu_sem_init(&cpu->sem, 0);
-#endif
 }
-- 
2.49.0

[PATCH v3 45/68] accel/tcg: Factor tcg_vcpu_thread_precreate() out

[Philippe Mathieu-Daudé]

Factor tcg_vcpu_thread_precreate() out for re-use.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/tcg/tcg-accel-ops.h       | 1 +
 accel/tcg/tcg-accel-ops-mttcg.c | 3 +--
 accel/tcg/tcg-accel-ops-rr.c    | 3 +--
 accel/tcg/tcg-accel-ops.c       | 7 +++++++
 4 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops.h b/accel/tcg/tcg-accel-ops.h
index 6feeb3f3e9b..129af89c3e7 100644
--- a/accel/tcg/tcg-accel-ops.h
+++ b/accel/tcg/tcg-accel-ops.h
@@ -14,6 +14,7 @@
 
 #include "system/cpus.h"
 
+void tcg_vcpu_thread_precreate(CPUState *cpu);
 void tcg_cpu_destroy(CPUState *cpu);
 int tcg_cpu_exec(CPUState *cpu);
 void tcg_handle_interrupt(CPUState *cpu, int mask);
diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
index dfcee30947e..462be7596b9 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.c
+++ b/accel/tcg/tcg-accel-ops-mttcg.c
@@ -133,8 +133,7 @@ void mttcg_start_vcpu_thread(CPUState *cpu)
 {
     char thread_name[VCPU_THREAD_NAME_SIZE];
 
-    g_assert(tcg_enabled());
-    tcg_cpu_init_cflags(cpu, current_machine->smp.max_cpus > 1);
+    tcg_vcpu_thread_precreate(cpu);
 
     /* create a thread per vCPU with TCG (MTTCG) */
     snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/TCG",
diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
index 6eec5c9eee9..fc33a13e4e8 100644
--- a/accel/tcg/tcg-accel-ops-rr.c
+++ b/accel/tcg/tcg-accel-ops-rr.c
@@ -311,8 +311,7 @@ void rr_start_vcpu_thread(CPUState *cpu)
     static QemuCond *single_tcg_halt_cond;
     static QemuThread *single_tcg_cpu_thread;
 
-    g_assert(tcg_enabled());
-    tcg_cpu_init_cflags(cpu, false);
+    tcg_vcpu_thread_precreate(cpu);
 
     if (!single_tcg_cpu_thread) {
         single_tcg_halt_cond = cpu->halt_cond;
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 95ff451c148..861996649b7 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -41,6 +41,7 @@
 #include "gdbstub/enums.h"
 
 #include "hw/core/cpu.h"
+#include "hw/boards.h"
 
 #include "tcg-accel-ops.h"
 #include "tcg-accel-ops-mttcg.h"
@@ -69,6 +70,12 @@ void tcg_cpu_init_cflags(CPUState *cpu, bool parallel)
     tcg_cflags_set(cpu, cflags);
 }
 
+void tcg_vcpu_thread_precreate(CPUState *cpu)
+{
+    g_assert(tcg_enabled());
+    tcg_cpu_init_cflags(cpu, current_machine->smp.max_cpus > 1);
+}
+
 void tcg_cpu_destroy(CPUState *cpu)
 {
     cpu_thread_signal_destroyed(cpu);
-- 
2.49.0

[PATCH v3 46/68] accel: Factor accel_create_vcpu_thread() out

[Philippe Mathieu-Daudé]

Factor accel_create_vcpu_thread() out of system/cpus.c
to be able to access accel/ internal definitions.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/qemu/accel.h |  2 ++
 accel/accel-common.c | 19 +++++++++++++++++++
 system/cpus.c        |  4 +---
 3 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/include/qemu/accel.h b/include/qemu/accel.h
index 80bfe3c4d0f..be94b138346 100644
--- a/include/qemu/accel.h
+++ b/include/qemu/accel.h
@@ -96,6 +96,8 @@ void accel_pre_resume(MachineState *ms, bool step_pending);
  */
 void accel_cpu_instance_init(CPUState *cpu);
 
+void accel_create_vcpu_thread(AccelState *accel, CPUState *cpu);
+
 /**
  * accel_cpu_common_realize:
  * @cpu: The CPU that needs to call accel-specific cpu realization.
diff --git a/accel/accel-common.c b/accel/accel-common.c
index d1a5f3ca3df..d719917063e 100644
--- a/accel/accel-common.c
+++ b/accel/accel-common.c
@@ -89,6 +89,25 @@ void accel_cpu_instance_init(CPUState *cpu)
     }
 }
 
+void accel_create_vcpu_thread(AccelState *accel, CPUState *cpu)
+{
+    AccelClass *ac;
+
+    if (!accel) {
+        accel = current_accel();
+    }
+    ac = ACCEL_GET_CLASS(accel);
+
+    /* accelerators all implement the AccelOpsClass */
+    g_assert(ac->ops);
+
+    if (ac->ops->create_vcpu_thread != NULL) {
+        ac->ops->create_vcpu_thread(cpu);
+    } else {
+        g_assert_not_reached();
+    }
+}
+
 bool accel_cpu_common_realize(CPUState *cpu, Error **errp)
 {
     AccelState *accel = current_accel();
diff --git a/system/cpus.c b/system/cpus.c
index 2c3759ea9be..6055f7c1c5f 100644
--- a/system/cpus.c
+++ b/system/cpus.c
@@ -698,9 +698,7 @@ void qemu_init_vcpu(CPUState *cpu)
         cpu_address_space_init(cpu, 0, "cpu-memory", cpu->memory);
     }
 
-    /* accelerators all implement the AccelOpsClass */
-    g_assert(cpus_accel != NULL && cpus_accel->create_vcpu_thread != NULL);
-    cpus_accel->create_vcpu_thread(cpu);
+    accel_create_vcpu_thread(NULL, cpu);
 
     while (!cpu->created) {
         qemu_cond_wait(&qemu_cpu_cond, &bql);
-- 
2.49.0

[PATCH v3 47/68] accel: Introduce AccelOpsClass::cpu_thread_routine handler

[Philippe Mathieu-Daudé]

In order to have a generic function creating threads,
introduce the thread_precreate() and cpu_thread_routine()
handlers.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/system/accel-ops.h |  5 ++++-
 accel/accel-common.c       | 16 +++++++++++++++-
 system/cpus.c              |  2 +-
 3 files changed, 20 insertions(+), 3 deletions(-)

diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index 14113bb5c10..28e29cfa06d 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -39,7 +39,10 @@ struct AccelOpsClass {
     bool (*cpu_target_realize)(CPUState *cpu, Error **errp);
     void (*cpu_reset_hold)(CPUState *cpu);
 
-    void (*create_vcpu_thread)(CPUState *cpu); /* MANDATORY NON-NULL */
+    /* Either cpu_thread_routine() or create_vcpu_thread() is mandatory */
+    void *(*cpu_thread_routine)(void *);
+    void (*thread_precreate)(CPUState *cpu);
+    void (*create_vcpu_thread)(CPUState *cpu);
     void (*kick_vcpu_thread)(CPUState *cpu);
     bool (*cpu_thread_is_idle)(CPUState *cpu);
 
diff --git a/accel/accel-common.c b/accel/accel-common.c
index d719917063e..24038acf4aa 100644
--- a/accel/accel-common.c
+++ b/accel/accel-common.c
@@ -11,6 +11,7 @@
 #include "qemu/accel.h"
 #include "qemu/target-info.h"
 #include "system/accel-ops.h"
+#include "system/cpus.h"
 #include "accel/accel-cpu.h"
 #include "accel-internal.h"
 
@@ -104,7 +105,20 @@ void accel_create_vcpu_thread(AccelState *accel, CPUState *cpu)
     if (ac->ops->create_vcpu_thread != NULL) {
         ac->ops->create_vcpu_thread(cpu);
     } else {
-        g_assert_not_reached();
+        char thread_name[VCPU_THREAD_NAME_SIZE];
+
+        assert(ac->name);
+        assert(ac->ops->cpu_thread_routine);
+
+        if (ac->ops->thread_precreate) {
+            ac->ops->thread_precreate(cpu);
+        }
+
+        snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/%s",
+                 cpu->cpu_index, ac->name);
+        qemu_thread_create(cpu->thread, thread_name,
+                           ac->ops->cpu_thread_routine,
+                           cpu, QEMU_THREAD_JOINABLE);
     }
 }
 
diff --git a/system/cpus.c b/system/cpus.c
index 6055f7c1c5f..c2ad640980c 100644
--- a/system/cpus.c
+++ b/system/cpus.c
@@ -671,7 +671,7 @@ void cpu_remove_sync(CPUState *cpu)
 void cpus_register_accel(const AccelOpsClass *ops)
 {
     assert(ops != NULL);
-    assert(ops->create_vcpu_thread != NULL); /* mandatory */
+    assert(ops->create_vcpu_thread || ops->cpu_thread_routine);
     cpus_accel = ops;
 }
 
-- 
2.49.0

[PATCH v3 48/68] accel/dummy: Convert to AccelOpsClass::cpu_thread_routine

[Philippe Mathieu-Daudé]

By converting to AccelOpsClass::cpu_thread_routine we can
let the common accel_create_vcpu_thread() create the thread.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/dummy-cpus.h  |  2 +-
 accel/dummy-cpus.c  | 14 +-------------
 accel/qtest/qtest.c |  3 ++-
 accel/xen/xen-all.c |  3 ++-
 4 files changed, 6 insertions(+), 16 deletions(-)

diff --git a/accel/dummy-cpus.h b/accel/dummy-cpus.h
index c3af710ee8c..c2f9fee164c 100644
--- a/accel/dummy-cpus.h
+++ b/accel/dummy-cpus.h
@@ -10,6 +10,6 @@
 #define ACCEL_DUMMY_CPUS_H
 
 void dummy_thread_precreate(CPUState *cpu);
-void dummy_start_vcpu_thread(CPUState *cpu);
+void *dummy_cpu_thread_routine(void *arg);
 
 #endif
diff --git a/accel/dummy-cpus.c b/accel/dummy-cpus.c
index 2cbc3fecc93..f637ab05e32 100644
--- a/accel/dummy-cpus.c
+++ b/accel/dummy-cpus.c
@@ -19,7 +19,7 @@
 #include "hw/core/cpu.h"
 #include "accel/dummy-cpus.h"
 
-static void *dummy_cpu_thread_fn(void *arg)
+void *dummy_cpu_thread_routine(void *arg)
 {
     CPUState *cpu = arg;
 
@@ -71,15 +71,3 @@ void dummy_thread_precreate(CPUState *cpu)
     qemu_sem_init(&cpu->sem, 0);
 #endif
 }
-
-void dummy_start_vcpu_thread(CPUState *cpu)
-{
-    char thread_name[VCPU_THREAD_NAME_SIZE];
-
-    dummy_thread_precreate(cpu);
-
-    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/DUMMY",
-             cpu->cpu_index);
-    qemu_thread_create(cpu->thread, thread_name, dummy_cpu_thread_fn, cpu,
-                       QEMU_THREAD_JOINABLE);
-}
diff --git a/accel/qtest/qtest.c b/accel/qtest/qtest.c
index 2606fe97b49..9f30098d133 100644
--- a/accel/qtest/qtest.c
+++ b/accel/qtest/qtest.c
@@ -64,7 +64,8 @@ static void qtest_accel_ops_class_init(ObjectClass *oc, const void *data)
 {
     AccelOpsClass *ops = ACCEL_OPS_CLASS(oc);
 
-    ops->create_vcpu_thread = dummy_start_vcpu_thread;
+    ops->thread_precreate = dummy_thread_precreate;
+    ops->cpu_thread_routine = dummy_cpu_thread_routine;
     ops->get_virtual_clock = qtest_get_virtual_clock;
     ops->set_virtual_clock = qtest_set_virtual_clock;
 };
diff --git a/accel/xen/xen-all.c b/accel/xen/xen-all.c
index ba752bbe5de..5ff72d9532c 100644
--- a/accel/xen/xen-all.c
+++ b/accel/xen/xen-all.c
@@ -151,7 +151,8 @@ static void xen_accel_ops_class_init(ObjectClass *oc, const void *data)
 {
     AccelOpsClass *ops = ACCEL_OPS_CLASS(oc);
 
-    ops->create_vcpu_thread = dummy_start_vcpu_thread;
+    ops->thread_precreate = dummy_thread_precreate;
+    ops->cpu_thread_routine = dummy_cpu_thread_routine;
 }
 
 static const TypeInfo xen_accel_ops_type = {
-- 
2.49.0

[PATCH v3 49/68] accel/tcg: Convert to AccelOpsClass::cpu_thread_routine

[Philippe Mathieu-Daudé]

By converting to AccelOpsClass::cpu_thread_routine we can
let the common accel_create_vcpu_thread() create the thread.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/tcg/tcg-accel-ops-mttcg.h |  3 +--
 accel/tcg/tcg-accel-ops-mttcg.c | 16 +---------------
 accel/tcg/tcg-accel-ops.c       |  3 ++-
 3 files changed, 4 insertions(+), 18 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops-mttcg.h b/accel/tcg/tcg-accel-ops-mttcg.h
index 8ffa7a9a9fe..8bf2452c886 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.h
+++ b/accel/tcg/tcg-accel-ops-mttcg.h
@@ -13,7 +13,6 @@
 /* kick MTTCG vCPU thread */
 void mttcg_kick_vcpu_thread(CPUState *cpu);
 
-/* start an mttcg vCPU thread */
-void mttcg_start_vcpu_thread(CPUState *cpu);
+void *mttcg_cpu_thread_routine(void *arg);
 
 #endif /* TCG_ACCEL_OPS_MTTCG_H */
diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
index 462be7596b9..96ce065eb59 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.c
+++ b/accel/tcg/tcg-accel-ops-mttcg.c
@@ -61,7 +61,7 @@ static void mttcg_force_rcu(Notifier *notify, void *data)
  * current CPUState for a given thread.
  */
 
-static void *mttcg_cpu_thread_fn(void *arg)
+void *mttcg_cpu_thread_routine(void *arg)
 {
     MttcgForceRcuNotifier force_rcu;
     CPUState *cpu = arg;
@@ -128,17 +128,3 @@ void mttcg_kick_vcpu_thread(CPUState *cpu)
 {
     cpu_exit(cpu);
 }
-
-void mttcg_start_vcpu_thread(CPUState *cpu)
-{
-    char thread_name[VCPU_THREAD_NAME_SIZE];
-
-    tcg_vcpu_thread_precreate(cpu);
-
-    /* create a thread per vCPU with TCG (MTTCG) */
-    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/TCG",
-             cpu->cpu_index);
-
-    qemu_thread_create(cpu->thread, thread_name, mttcg_cpu_thread_fn,
-                       cpu, QEMU_THREAD_JOINABLE);
-}
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 861996649b7..4931e536beb 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -204,7 +204,7 @@ static void tcg_accel_ops_init(AccelClass *ac)
     AccelOpsClass *ops = ac->ops;
 
     if (qemu_tcg_mttcg_enabled()) {
-        ops->create_vcpu_thread = mttcg_start_vcpu_thread;
+        ops->cpu_thread_routine = mttcg_cpu_thread_routine;
         ops->kick_vcpu_thread = mttcg_kick_vcpu_thread;
         ops->handle_interrupt = tcg_handle_interrupt;
     } else {
@@ -222,6 +222,7 @@ static void tcg_accel_ops_init(AccelClass *ac)
 
     ops->cpu_common_realize = tcg_exec_realizefn;
     ops->cpu_common_unrealize = tcg_exec_unrealizefn;
+    ops->thread_precreate = tcg_vcpu_thread_precreate;
     ops->cpu_reset_hold = tcg_cpu_reset_hold;
     ops->insert_breakpoint = tcg_insert_breakpoint;
     ops->remove_breakpoint = tcg_remove_breakpoint;
-- 
2.49.0

[PATCH v3 50/68] accel/hvf: Convert to AccelOpsClass::cpu_thread_routine

[Philippe Mathieu-Daudé]

By converting to AccelOpsClass::cpu_thread_routine we can
let the common accel_create_vcpu_thread() create the thread.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/hvf/hvf-accel-ops.c | 18 +-----------------
 1 file changed, 1 insertion(+), 17 deletions(-)

diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index c91e18bc3dd..b61f08330f1 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -207,22 +207,6 @@ static void *hvf_cpu_thread_fn(void *arg)
     return NULL;
 }
 
-static void hvf_start_vcpu_thread(CPUState *cpu)
-{
-    char thread_name[VCPU_THREAD_NAME_SIZE];
-
-    /*
-     * HVF currently does not support TCG, and only runs in
-     * unrestricted-guest mode.
-     */
-    assert(hvf_enabled());
-
-    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/HVF",
-             cpu->cpu_index);
-    qemu_thread_create(cpu->thread, thread_name, hvf_cpu_thread_fn,
-                       cpu, QEMU_THREAD_JOINABLE);
-}
-
 struct hvf_sw_breakpoint *hvf_find_sw_breakpoint(CPUState *cpu, vaddr pc)
 {
     struct hvf_sw_breakpoint *bp;
@@ -369,7 +353,7 @@ static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
 
     ops->cpu_target_realize = hvf_arch_cpu_realize;
 
-    ops->create_vcpu_thread = hvf_start_vcpu_thread;
+    ops->cpu_thread_routine = hvf_cpu_thread_fn,
     ops->kick_vcpu_thread = hvf_kick_vcpu_thread;
 
     ops->synchronize_post_reset = hvf_cpu_synchronize_post_reset;
-- 
2.49.0

[PATCH v3 51/68] accel/kvm: Convert to AccelOpsClass::cpu_thread_routine

[Philippe Mathieu-Daudé]

By converting to AccelOpsClass::cpu_thread_routine we can
let the common accel_create_vcpu_thread() create the thread.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 accel/kvm/kvm-accel-ops.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)

diff --git a/accel/kvm/kvm-accel-ops.c b/accel/kvm/kvm-accel-ops.c
index be960bde5c4..21ff3af306f 100644
--- a/accel/kvm/kvm-accel-ops.c
+++ b/accel/kvm/kvm-accel-ops.c
@@ -63,16 +63,6 @@ static void *kvm_vcpu_thread_fn(void *arg)
     return NULL;
 }
 
-static void kvm_start_vcpu_thread(CPUState *cpu)
-{
-    char thread_name[VCPU_THREAD_NAME_SIZE];
-
-    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/KVM",
-             cpu->cpu_index);
-    qemu_thread_create(cpu->thread, thread_name, kvm_vcpu_thread_fn,
-                       cpu, QEMU_THREAD_JOINABLE);
-}
-
 static bool kvm_vcpu_thread_is_idle(CPUState *cpu)
 {
     return !kvm_halt_in_kernel();
@@ -89,7 +79,7 @@ static void kvm_accel_ops_class_init(ObjectClass *oc, const void *data)
 {
     AccelOpsClass *ops = ACCEL_OPS_CLASS(oc);
 
-    ops->create_vcpu_thread = kvm_start_vcpu_thread;
+    ops->cpu_thread_routine = kvm_vcpu_thread_fn;
     ops->cpu_thread_is_idle = kvm_vcpu_thread_is_idle;
     ops->synchronize_post_reset = kvm_cpu_synchronize_post_reset;
     ops->synchronize_post_init = kvm_cpu_synchronize_post_init;
-- 
2.49.0

[PATCH v3 52/68] accel/nvmm: Convert to AccelOpsClass::cpu_thread_routine

[Philippe Mathieu-Daudé]

By converting to AccelOpsClass::cpu_thread_routine we can
let the common accel_create_vcpu_thread() create the thread.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/i386/nvmm/nvmm-accel-ops.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)

diff --git a/target/i386/nvmm/nvmm-accel-ops.c b/target/i386/nvmm/nvmm-accel-ops.c
index 21443078b72..bef6f61b776 100644
--- a/target/i386/nvmm/nvmm-accel-ops.c
+++ b/target/i386/nvmm/nvmm-accel-ops.c
@@ -61,16 +61,6 @@ static void *qemu_nvmm_cpu_thread_fn(void *arg)
     return NULL;
 }
 
-static void nvmm_start_vcpu_thread(CPUState *cpu)
-{
-    char thread_name[VCPU_THREAD_NAME_SIZE];
-
-    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/NVMM",
-             cpu->cpu_index);
-    qemu_thread_create(cpu->thread, thread_name, qemu_nvmm_cpu_thread_fn,
-                       cpu, QEMU_THREAD_JOINABLE);
-}
-
 /*
  * Abort the call to run the virtual processor by another thread, and to
  * return the control to that thread.
@@ -85,7 +75,7 @@ static void nvmm_accel_ops_class_init(ObjectClass *oc, const void *data)
 {
     AccelOpsClass *ops = ACCEL_OPS_CLASS(oc);
 
-    ops->create_vcpu_thread = nvmm_start_vcpu_thread;
+    ops->cpu_thread_routine = qemu_nvmm_cpu_thread_fn;
     ops->kick_vcpu_thread = nvmm_kick_vcpu_thread;
 
     ops->synchronize_post_reset = nvmm_cpu_synchronize_post_reset;
-- 
2.49.0

[PATCH v3 53/68] accel/whpx: Convert to AccelOpsClass::cpu_thread_routine

[Philippe Mathieu-Daudé]

By converting to AccelOpsClass::cpu_thread_routine we can
let the common accel_create_vcpu_thread() create the thread.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/i386/whpx/whpx-accel-ops.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)

diff --git a/target/i386/whpx/whpx-accel-ops.c b/target/i386/whpx/whpx-accel-ops.c
index 011810b5e50..8cbc6f4e2d8 100644
--- a/target/i386/whpx/whpx-accel-ops.c
+++ b/target/i386/whpx/whpx-accel-ops.c
@@ -61,16 +61,6 @@ static void *whpx_cpu_thread_fn(void *arg)
     return NULL;
 }
 
-static void whpx_start_vcpu_thread(CPUState *cpu)
-{
-    char thread_name[VCPU_THREAD_NAME_SIZE];
-
-    snprintf(thread_name, VCPU_THREAD_NAME_SIZE, "CPU %d/WHPX",
-             cpu->cpu_index);
-    qemu_thread_create(cpu->thread, thread_name, whpx_cpu_thread_fn,
-                       cpu, QEMU_THREAD_JOINABLE);
-}
-
 static void whpx_kick_vcpu_thread(CPUState *cpu)
 {
     if (!qemu_cpu_is_self(cpu)) {
@@ -87,7 +77,7 @@ static void whpx_accel_ops_class_init(ObjectClass *oc, const void *data)
 {
     AccelOpsClass *ops = ACCEL_OPS_CLASS(oc);
 
-    ops->create_vcpu_thread = whpx_start_vcpu_thread;
+    ops->cpu_thread_routine = whpx_cpu_thread_fn;
     ops->kick_vcpu_thread = whpx_kick_vcpu_thread;
     ops->cpu_thread_is_idle = whpx_vcpu_thread_is_idle;
 
-- 
2.49.0

[PATCH v3 54/68] accel: Factor accel_cpu_realize() out

[Philippe Mathieu-Daudé]

Factor accel_cpu_realize() out of accel_cpu_common_realize()
for re-use.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/accel-internal.h | 2 ++
 accel/accel-common.c   | 8 ++++++--
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/accel/accel-internal.h b/accel/accel-internal.h
index d3a4422cbf7..b541377c349 100644
--- a/accel/accel-internal.h
+++ b/accel/accel-internal.h
@@ -14,4 +14,6 @@
 
 void accel_init_ops_interfaces(AccelClass *ac);
 
+bool accel_cpu_realize(AccelState *accel, CPUState *cpu, Error **errp);
+
 #endif /* ACCEL_SYSTEM_H */
diff --git a/accel/accel-common.c b/accel/accel-common.c
index 24038acf4aa..de010adb484 100644
--- a/accel/accel-common.c
+++ b/accel/accel-common.c
@@ -122,9 +122,8 @@ void accel_create_vcpu_thread(AccelState *accel, CPUState *cpu)
     }
 }
 
-bool accel_cpu_common_realize(CPUState *cpu, Error **errp)
+bool accel_cpu_realize(AccelState *accel, CPUState *cpu, Error **errp)
 {
-    AccelState *accel = current_accel();
     AccelClass *acc = ACCEL_GET_CLASS(accel);
 
     /* target specific realization */
@@ -147,6 +146,11 @@ bool accel_cpu_common_realize(CPUState *cpu, Error **errp)
     return true;
 }
 
+bool accel_cpu_common_realize(CPUState *cpu, Error **errp)
+{
+    return accel_cpu_realize(current_accel(), cpu, errp);
+}
+
 void accel_cpu_common_unrealize(CPUState *cpu)
 {
     AccelState *accel = current_accel();
-- 
2.49.0

Re: [PATCH v3 54/68] accel: Factor accel_cpu_realize() out

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Factor accel_cpu_realize() out of accel_cpu_common_realize()
> for re-use.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/accel-internal.h | 2 ++
>   accel/accel-common.c   | 8 ++++++--
>   2 files changed, 8 insertions(+), 2 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 55/68] accel/tcg: Factor tcg_vcpu_init() out for re-use

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops.h       | 2 ++
 accel/tcg/tcg-accel-ops-mttcg.c | 4 +++-
 accel/tcg/tcg-accel-ops-rr.c    | 4 +++-
 accel/tcg/tcg-accel-ops.c       | 7 +++++++
 4 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops.h b/accel/tcg/tcg-accel-ops.h
index 129af89c3e7..1263a666774 100644
--- a/accel/tcg/tcg-accel-ops.h
+++ b/accel/tcg/tcg-accel-ops.h
@@ -20,4 +20,6 @@ int tcg_cpu_exec(CPUState *cpu);
 void tcg_handle_interrupt(CPUState *cpu, int mask);
 void tcg_cpu_init_cflags(CPUState *cpu, bool parallel);
 
+int tcg_vcpu_init(CPUState *cpu);
+
 #endif /* TCG_ACCEL_OPS_H */
diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
index 96ce065eb59..4de506a80ca 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.c
+++ b/accel/tcg/tcg-accel-ops-mttcg.c
@@ -79,8 +79,10 @@ void *mttcg_cpu_thread_routine(void *arg)
     qemu_thread_get_self(cpu->thread);
 
     cpu->thread_id = qemu_get_thread_id();
-    cpu->neg.can_do_io = true;
     current_cpu = cpu;
+
+    tcg_vcpu_init(cpu);
+
     cpu_thread_signal_created(cpu);
     qemu_guest_random_seed_thread_part2(cpu->random_seed);
 
diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
index fc33a13e4e8..9578bc639cb 100644
--- a/accel/tcg/tcg-accel-ops-rr.c
+++ b/accel/tcg/tcg-accel-ops-rr.c
@@ -192,7 +192,9 @@ static void *rr_cpu_thread_fn(void *arg)
     qemu_thread_get_self(cpu->thread);
 
     cpu->thread_id = qemu_get_thread_id();
-    cpu->neg.can_do_io = true;
+
+    tcg_vcpu_init(cpu);
+
     cpu_thread_signal_created(cpu);
     qemu_guest_random_seed_thread_part2(cpu->random_seed);
 
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 4931e536beb..83fb2d1362c 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -76,6 +76,13 @@ void tcg_vcpu_thread_precreate(CPUState *cpu)
     tcg_cpu_init_cflags(cpu, current_machine->smp.max_cpus > 1);
 }
 
+int tcg_vcpu_init(CPUState *cpu)
+{
+    cpu->neg.can_do_io = true;
+
+    return 0;
+}
+
 void tcg_cpu_destroy(CPUState *cpu)
 {
     cpu_thread_signal_destroyed(cpu);
-- 
2.49.0

Re: [PATCH v3 55/68] accel/tcg: Factor tcg_vcpu_init() out for re-use

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops.h       | 2 ++
>   accel/tcg/tcg-accel-ops-mttcg.c | 4 +++-
>   accel/tcg/tcg-accel-ops-rr.c    | 4 +++-
>   accel/tcg/tcg-accel-ops.c       | 7 +++++++
>   4 files changed, 15 insertions(+), 2 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 56/68] accel/tcg: Unregister the RCU before exiting RR thread

[Philippe Mathieu-Daudé]

Although unreachable, still unregister the RCU before exiting
the thread, as documented in "qemu/rcu.h":

 /*
  * Important !
  *
  * Each thread containing read-side critical sections must be registered
  * with rcu_register_thread() before calling rcu_read_lock().
  * rcu_unregister_thread() should be called before the thread exits.
  */

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops-rr.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
index 9578bc639cb..57a4bcab203 100644
--- a/accel/tcg/tcg-accel-ops-rr.c
+++ b/accel/tcg/tcg-accel-ops-rr.c
@@ -304,6 +304,8 @@ static void *rr_cpu_thread_fn(void *arg)
         rr_deal_with_unplugged_cpus();
     }
 
+    rcu_unregister_thread();
+
     g_assert_not_reached();
 }
 
-- 
2.49.0

Re: [PATCH v3 56/68] accel/tcg: Unregister the RCU before exiting RR thread

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Although unreachable, still unregister the RCU before exiting
> the thread, as documented in "qemu/rcu.h":
> 
>   /*
>    * Important !
>    *
>    * Each thread containing read-side critical sections must be registered
>    * with rcu_register_thread() before calling rcu_read_lock().
>    * rcu_unregister_thread() should be called before the thread exits.
>    */
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops-rr.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
> index 9578bc639cb..57a4bcab203 100644
> --- a/accel/tcg/tcg-accel-ops-rr.c
> +++ b/accel/tcg/tcg-accel-ops-rr.c
> @@ -304,6 +304,8 @@ static void *rr_cpu_thread_fn(void *arg)
>           rr_deal_with_unplugged_cpus();
>       }
>   
> +    rcu_unregister_thread();
> +
>       g_assert_not_reached();
>   }
>   

I suppose.  This isn't like some clean exit or anything... we're about to abort.

Acked-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 57/68] accel/tcg: Expose vcpu_[un]register() for MTTCG

[Philippe Mathieu-Daudé]

Allocate ForceRcuNotifier on the Heap.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops-mttcg.c | 30 ++++++++++++++++++++++++------
 1 file changed, 24 insertions(+), 6 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
index 4de506a80ca..2d31b00ee59 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.c
+++ b/accel/tcg/tcg-accel-ops-mttcg.c
@@ -55,6 +55,27 @@ static void mttcg_force_rcu(Notifier *notify, void *data)
     async_run_on_cpu(cpu, do_nothing, RUN_ON_CPU_NULL);
 }
 
+static void *mttcg_vcpu_register(CPUState *cpu)
+{
+    MttcgForceRcuNotifier *force_rcu = g_new(MttcgForceRcuNotifier, 1);
+
+    force_rcu->notifier.notify = mttcg_force_rcu;
+    force_rcu->cpu = cpu;
+    rcu_add_force_rcu_notifier(&force_rcu->notifier);
+    tcg_register_thread();
+
+    return force_rcu;
+}
+
+static void mttcg_vcpu_unregister(CPUState *cpu, void *opaque)
+{
+    MttcgForceRcuNotifier *force_rcu = opaque;
+
+    rcu_remove_force_rcu_notifier(&force_rcu->notifier);
+
+    g_free(force_rcu);
+}
+
 /*
  * In the multi-threaded case each vCPU has its own thread. The TLS
  * variable current_cpu can be used deep in the code to find the
@@ -63,17 +84,14 @@ static void mttcg_force_rcu(Notifier *notify, void *data)
 
 void *mttcg_cpu_thread_routine(void *arg)
 {
-    MttcgForceRcuNotifier force_rcu;
+    MttcgForceRcuNotifier *force_rcu;
     CPUState *cpu = arg;
 
     assert(tcg_enabled());
     g_assert(!icount_enabled());
 
     rcu_register_thread();
-    force_rcu.notifier.notify = mttcg_force_rcu;
-    force_rcu.cpu = cpu;
-    rcu_add_force_rcu_notifier(&force_rcu.notifier);
-    tcg_register_thread();
+    force_rcu = mttcg_vcpu_register(cpu);
 
     bql_lock();
     qemu_thread_get_self(cpu->thread);
@@ -121,7 +139,7 @@ void *mttcg_cpu_thread_routine(void *arg)
 
     tcg_cpu_destroy(cpu);
     bql_unlock();
-    rcu_remove_force_rcu_notifier(&force_rcu.notifier);
+    mttcg_vcpu_unregister(cpu, force_rcu);
     rcu_unregister_thread();
     return NULL;
 }
-- 
2.49.0

Re: [PATCH v3 57/68] accel/tcg: Expose vcpu_[un]register() for MTTCG

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Allocate ForceRcuNotifier on the Heap.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops-mttcg.c | 30 ++++++++++++++++++++++++------
>   1 file changed, 24 insertions(+), 6 deletions(-)

Please document the motivation.

r~

> 
> diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
> index 4de506a80ca..2d31b00ee59 100644
> --- a/accel/tcg/tcg-accel-ops-mttcg.c
> +++ b/accel/tcg/tcg-accel-ops-mttcg.c
> @@ -55,6 +55,27 @@ static void mttcg_force_rcu(Notifier *notify, void *data)
>       async_run_on_cpu(cpu, do_nothing, RUN_ON_CPU_NULL);
>   }
>   
> +static void *mttcg_vcpu_register(CPUState *cpu)
> +{
> +    MttcgForceRcuNotifier *force_rcu = g_new(MttcgForceRcuNotifier, 1);
> +
> +    force_rcu->notifier.notify = mttcg_force_rcu;
> +    force_rcu->cpu = cpu;
> +    rcu_add_force_rcu_notifier(&force_rcu->notifier);
> +    tcg_register_thread();
> +
> +    return force_rcu;
> +}
> +
> +static void mttcg_vcpu_unregister(CPUState *cpu, void *opaque)
> +{
> +    MttcgForceRcuNotifier *force_rcu = opaque;
> +
> +    rcu_remove_force_rcu_notifier(&force_rcu->notifier);
> +
> +    g_free(force_rcu);
> +}
> +
>   /*
>    * In the multi-threaded case each vCPU has its own thread. The TLS
>    * variable current_cpu can be used deep in the code to find the
> @@ -63,17 +84,14 @@ static void mttcg_force_rcu(Notifier *notify, void *data)
>   
>   void *mttcg_cpu_thread_routine(void *arg)
>   {
> -    MttcgForceRcuNotifier force_rcu;
> +    MttcgForceRcuNotifier *force_rcu;
>       CPUState *cpu = arg;
>   
>       assert(tcg_enabled());
>       g_assert(!icount_enabled());
>   
>       rcu_register_thread();
> -    force_rcu.notifier.notify = mttcg_force_rcu;
> -    force_rcu.cpu = cpu;
> -    rcu_add_force_rcu_notifier(&force_rcu.notifier);
> -    tcg_register_thread();
> +    force_rcu = mttcg_vcpu_register(cpu);
>   
>       bql_lock();
>       qemu_thread_get_self(cpu->thread);
> @@ -121,7 +139,7 @@ void *mttcg_cpu_thread_routine(void *arg)
>   
>       tcg_cpu_destroy(cpu);
>       bql_unlock();
> -    rcu_remove_force_rcu_notifier(&force_rcu.notifier);
> +    mttcg_vcpu_unregister(cpu, force_rcu);
>       rcu_unregister_thread();
>       return NULL;
>   }

Re: [PATCH v3 57/68] accel/tcg: Expose vcpu_[un]register() for MTTCG

[Philippe Mathieu-Daudé]

On 2/7/25 17:19, Richard Henderson wrote:
> On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
>> Allocate ForceRcuNotifier on the Heap.
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   accel/tcg/tcg-accel-ops-mttcg.c | 30 ++++++++++++++++++++++++------
>>   1 file changed, 24 insertions(+), 6 deletions(-)
> 
> Please document the motivation.

 > [...] the structure is still not accessible from anywhere outside
 > of the function, and has the same lifetime as the function.

We need this to register the MTTCG thread in split_cpu_thread_routine():
https://lore.kernel.org/qemu-devel/20250620172751.94231-12-philmd@linaro.org/

Better to have AccelOpsClass::[un]register_thread_rcu() hooks?

>>
>> diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel- 
>> ops-mttcg.c
>> index 4de506a80ca..2d31b00ee59 100644
>> --- a/accel/tcg/tcg-accel-ops-mttcg.c
>> +++ b/accel/tcg/tcg-accel-ops-mttcg.c
>> @@ -55,6 +55,27 @@ static void mttcg_force_rcu(Notifier *notify, void 
>> *data)
>>       async_run_on_cpu(cpu, do_nothing, RUN_ON_CPU_NULL);
>>   }
>> +static void *mttcg_vcpu_register(CPUState *cpu)
>> +{
>> +    MttcgForceRcuNotifier *force_rcu = g_new(MttcgForceRcuNotifier, 1);
>> +
>> +    force_rcu->notifier.notify = mttcg_force_rcu;
>> +    force_rcu->cpu = cpu;
>> +    rcu_add_force_rcu_notifier(&force_rcu->notifier);
>> +    tcg_register_thread();
>> +
>> +    return force_rcu;
>> +}
>> +
>> +static void mttcg_vcpu_unregister(CPUState *cpu, void *opaque)
>> +{
>> +    MttcgForceRcuNotifier *force_rcu = opaque;
>> +
>> +    rcu_remove_force_rcu_notifier(&force_rcu->notifier);
>> +
>> +    g_free(force_rcu);
>> +}
>> +
>>   /*
>>    * In the multi-threaded case each vCPU has its own thread. The TLS
>>    * variable current_cpu can be used deep in the code to find the
>> @@ -63,17 +84,14 @@ static void mttcg_force_rcu(Notifier *notify, void 
>> *data)
>>   void *mttcg_cpu_thread_routine(void *arg)
>>   {
>> -    MttcgForceRcuNotifier force_rcu;
>> +    MttcgForceRcuNotifier *force_rcu;
>>       CPUState *cpu = arg;
>>       assert(tcg_enabled());
>>       g_assert(!icount_enabled());
>>       rcu_register_thread();
>> -    force_rcu.notifier.notify = mttcg_force_rcu;
>> -    force_rcu.cpu = cpu;
>> -    rcu_add_force_rcu_notifier(&force_rcu.notifier);
>> -    tcg_register_thread();
>> +    force_rcu = mttcg_vcpu_register(cpu);
>>       bql_lock();
>>       qemu_thread_get_self(cpu->thread);
>> @@ -121,7 +139,7 @@ void *mttcg_cpu_thread_routine(void *arg)
>>       tcg_cpu_destroy(cpu);
>>       bql_unlock();
>> -    rcu_remove_force_rcu_notifier(&force_rcu.notifier);
>> +    mttcg_vcpu_unregister(cpu, force_rcu);
>>       rcu_unregister_thread();
>>       return NULL;
>>   }
> 

Re: [PATCH v3 57/68] accel/tcg: Expose vcpu_[un]register() for MTTCG

[Richard Henderson]

On 7/2/25 12:25, Philippe Mathieu-Daudé wrote:
> On 2/7/25 17:19, Richard Henderson wrote:
>> On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
>>> Allocate ForceRcuNotifier on the Heap.
>>>
>>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>>> ---
>>>   accel/tcg/tcg-accel-ops-mttcg.c | 30 ++++++++++++++++++++++++------
>>>   1 file changed, 24 insertions(+), 6 deletions(-)
>>
>> Please document the motivation.
> 
>  > [...] the structure is still not accessible from anywhere outside
>  > of the function, and has the same lifetime as the function.
> 
> We need this to register the MTTCG thread in split_cpu_thread_routine():
> https://lore.kernel.org/qemu-devel/20250620172751.94231-12-philmd@linaro.org/
> 
> Better to have AccelOpsClass::[un]register_thread_rcu() hooks?

I think this is more complex than it needs to be.  I think we don't actually need 
different implementations between mttcg and rr: we just need to make use of current_cpu.

I'll experiment with this.


r~

[PATCH v3 58/68] accel/tcg: Factor mttcg_cpu_exec() out for re-use

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops-mttcg.h |  1 +
 accel/tcg/tcg-accel-ops-mttcg.c | 16 ++++++++++++----
 2 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops-mttcg.h b/accel/tcg/tcg-accel-ops-mttcg.h
index 8bf2452c886..72eb1a71d61 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.h
+++ b/accel/tcg/tcg-accel-ops-mttcg.h
@@ -14,5 +14,6 @@
 void mttcg_kick_vcpu_thread(CPUState *cpu);
 
 void *mttcg_cpu_thread_routine(void *arg);
+int mttcg_cpu_exec(CPUState *cpu);
 
 #endif /* TCG_ACCEL_OPS_MTTCG_H */
diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
index 2d31b00ee59..8a0295e2410 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.c
+++ b/accel/tcg/tcg-accel-ops-mttcg.c
@@ -109,10 +109,7 @@ void *mttcg_cpu_thread_routine(void *arg)
 
     do {
         if (cpu_can_run(cpu)) {
-            int r;
-            bql_unlock();
-            r = tcg_cpu_exec(cpu);
-            bql_lock();
+            int r = mttcg_cpu_exec(cpu);
             switch (r) {
             case EXCP_DEBUG:
                 cpu_handle_guest_debug(cpu);
@@ -148,3 +145,14 @@ void mttcg_kick_vcpu_thread(CPUState *cpu)
 {
     cpu_exit(cpu);
 }
+
+int mttcg_cpu_exec(CPUState *cpu)
+{
+    int ret;
+
+    bql_unlock();
+    ret = tcg_cpu_exec(cpu);
+    bql_lock();
+
+    return ret;
+}
-- 
2.49.0

Re: [PATCH v3 58/68] accel/tcg: Factor mttcg_cpu_exec() out for re-use

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops-mttcg.h |  1 +
>   accel/tcg/tcg-accel-ops-mttcg.c | 16 ++++++++++++----
>   2 files changed, 13 insertions(+), 4 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 59/68] accel/tcg: Expose vcpu_[un]register() for RR

[Philippe Mathieu-Daudé]

Allocate ForceRcuNotifier on the Heap.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops-icount.h |  4 ++--
 include/hw/core/cpu.h            |  2 ++
 accel/tcg/tcg-accel-ops-icount.c |  8 +++----
 accel/tcg/tcg-accel-ops-rr.c     | 36 +++++++++++++++++++++++---------
 4 files changed, 34 insertions(+), 16 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops-icount.h b/accel/tcg/tcg-accel-ops-icount.h
index 16a301b6dc0..5f3ebea50ff 100644
--- a/accel/tcg/tcg-accel-ops-icount.h
+++ b/accel/tcg/tcg-accel-ops-icount.h
@@ -11,8 +11,8 @@
 #define TCG_ACCEL_OPS_ICOUNT_H
 
 void icount_handle_deadline(void);
-void icount_prepare_for_run(CPUState *cpu, int64_t cpu_budget);
-int64_t icount_percpu_budget(int cpu_count);
+void icount_prepare_for_run(CPUState *cpu);
+void icount_update_percpu_budget(CPUState *cpu, int cpu_count);
 void icount_process_data(CPUState *cpu);
 
 void icount_handle_interrupt(CPUState *cpu, int mask);
diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 726427449da..952e44587b3 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -498,6 +498,8 @@ struct CPUState {
     int singlestep_enabled;
     int64_t icount_budget;
     int64_t icount_extra;
+    int64_t cpu_budget; /* FIXME TCG specific */
+
     uint64_t random_seed;
     sigjmp_buf jmp_env;
 
diff --git a/accel/tcg/tcg-accel-ops-icount.c b/accel/tcg/tcg-accel-ops-icount.c
index d0f7b410fab..ae1297ff7f3 100644
--- a/accel/tcg/tcg-accel-ops-icount.c
+++ b/accel/tcg/tcg-accel-ops-icount.c
@@ -90,7 +90,7 @@ void icount_handle_deadline(void)
 }
 
 /* Distribute the budget evenly across all CPUs */
-int64_t icount_percpu_budget(int cpu_count)
+void icount_update_percpu_budget(CPUState *cpu, int cpu_count)
 {
     int64_t limit = icount_get_limit();
     int64_t timeslice = limit / cpu_count;
@@ -99,10 +99,10 @@ int64_t icount_percpu_budget(int cpu_count)
         timeslice = limit;
     }
 
-    return timeslice;
+    cpu->cpu_budget = timeslice;
 }
 
-void icount_prepare_for_run(CPUState *cpu, int64_t cpu_budget)
+void icount_prepare_for_run(CPUState *cpu)
 {
     int insns_left;
 
@@ -116,7 +116,7 @@ void icount_prepare_for_run(CPUState *cpu, int64_t cpu_budget)
 
     replay_mutex_lock();
 
-    cpu->icount_budget = MIN(icount_get_limit(), cpu_budget);
+    cpu->icount_budget = MIN(icount_get_limit(), cpu->cpu_budget);
     insns_left = MIN(0xffff, cpu->icount_budget);
     cpu->neg.icount_decr.u16.low = insns_left;
     cpu->icount_extra = cpu->icount_budget - insns_left;
diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
index 57a4bcab203..f5af7818d51 100644
--- a/accel/tcg/tcg-accel-ops-rr.c
+++ b/accel/tcg/tcg-accel-ops-rr.c
@@ -169,6 +169,27 @@ static int rr_cpu_count(void)
     return cpu_count;
 }
 
+static void *rr_vcpu_register(CPUState *cpu)
+{
+    Notifier *force_rcu = g_new(Notifier, 1);
+
+    assert(tcg_enabled());
+    force_rcu->notify = rr_force_rcu;
+    rcu_add_force_rcu_notifier(force_rcu);
+    tcg_register_thread();
+
+    return force_rcu;
+}
+
+static void rr_vcpu_unregister(CPUState *cpu, void *opaque)
+{
+    Notifier *force_rcu = opaque;
+
+    rcu_remove_force_rcu_notifier(force_rcu);
+
+    g_free(force_rcu);
+}
+
 /*
  * In the single-threaded case each vCPU is simulated in turn. If
  * there is more than a single vCPU we create a simple timer to kick
@@ -179,14 +200,11 @@ static int rr_cpu_count(void)
 
 static void *rr_cpu_thread_fn(void *arg)
 {
-    Notifier force_rcu;
+    Notifier *force_rcu;
     CPUState *cpu = arg;
 
-    assert(tcg_enabled());
     rcu_register_thread();
-    force_rcu.notify = rr_force_rcu;
-    rcu_add_force_rcu_notifier(&force_rcu);
-    tcg_register_thread();
+    force_rcu = rr_vcpu_register(cpu);
 
     bql_lock();
     qemu_thread_get_self(cpu->thread);
@@ -217,9 +235,6 @@ static void *rr_cpu_thread_fn(void *arg)
     cpu->exit_request = 1;
 
     while (1) {
-        /* Only used for icount_enabled() */
-        int64_t cpu_budget = 0;
-
         bql_unlock();
         replay_mutex_lock();
         bql_lock();
@@ -235,7 +250,7 @@ static void *rr_cpu_thread_fn(void *arg)
              */
             icount_handle_deadline();
 
-            cpu_budget = icount_percpu_budget(cpu_count);
+            icount_update_percpu_budget(cpu, cpu_count);
         }
 
         replay_mutex_unlock();
@@ -258,7 +273,7 @@ static void *rr_cpu_thread_fn(void *arg)
 
                 bql_unlock();
                 if (icount_enabled()) {
-                    icount_prepare_for_run(cpu, cpu_budget);
+                    icount_prepare_for_run(cpu);
                 }
                 r = tcg_cpu_exec(cpu);
                 if (icount_enabled()) {
@@ -304,6 +319,7 @@ static void *rr_cpu_thread_fn(void *arg)
         rr_deal_with_unplugged_cpus();
     }
 
+    rr_vcpu_unregister(cpu, force_rcu);
     rcu_unregister_thread();
 
     g_assert_not_reached();
-- 
2.49.0

Re: [PATCH v3 59/68] accel/tcg: Expose vcpu_[un]register() for RR

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Allocate ForceRcuNotifier on the Heap.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops-icount.h |  4 ++--
>   include/hw/core/cpu.h            |  2 ++
>   accel/tcg/tcg-accel-ops-icount.c |  8 +++----
>   accel/tcg/tcg-accel-ops-rr.c     | 36 +++++++++++++++++++++++---------
>   4 files changed, 34 insertions(+), 16 deletions(-)

Again, motivation?

As before, the structure is still not accessible from anywhere outside of the function, 
and has the same lifetime as the function.

Also, this is doing two things:

> +++ b/include/hw/core/cpu.h
> @@ -498,6 +498,8 @@ struct CPUState {
>      int singlestep_enabled;
>      int64_t icount_budget;
>      int64_t icount_extra;
> +    int64_t cpu_budget; /* FIXME TCG specific */

also moving cpu_budget.


r~

[PATCH v3 60/68] accel/tcg: Factor rr_cpu_exec() out for re-use

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops-rr.h |  2 ++
 accel/tcg/tcg-accel-ops-rr.c | 31 ++++++++++++++++++++-----------
 2 files changed, 22 insertions(+), 11 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops-rr.h b/accel/tcg/tcg-accel-ops-rr.h
index 2a76a296127..a73fe5b94a6 100644
--- a/accel/tcg/tcg-accel-ops-rr.h
+++ b/accel/tcg/tcg-accel-ops-rr.h
@@ -18,4 +18,6 @@ void rr_kick_vcpu_thread(CPUState *unused);
 /* start the round robin vcpu thread */
 void rr_start_vcpu_thread(CPUState *cpu);
 
+int rr_cpu_exec(CPUState *cpu);
+
 #endif /* TCG_ACCEL_OPS_RR_H */
diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
index f5af7818d51..a9150802369 100644
--- a/accel/tcg/tcg-accel-ops-rr.c
+++ b/accel/tcg/tcg-accel-ops-rr.c
@@ -169,6 +169,25 @@ static int rr_cpu_count(void)
     return cpu_count;
 }
 
+int rr_cpu_exec(CPUState *cpu)
+{
+    int ret;
+
+    bql_unlock();
+    if (icount_enabled()) {
+        icount_prepare_for_run(cpu);
+    }
+
+    ret = tcg_cpu_exec(cpu);
+
+    if (icount_enabled()) {
+        icount_process_data(cpu);
+    }
+    bql_lock();
+
+    return ret;
+}
+
 static void *rr_vcpu_register(CPUState *cpu)
 {
     Notifier *force_rcu = g_new(Notifier, 1);
@@ -269,17 +288,7 @@ static void *rr_cpu_thread_fn(void *arg)
                               (cpu->singlestep_enabled & SSTEP_NOTIMER) == 0);
 
             if (cpu_can_run(cpu)) {
-                int r;
-
-                bql_unlock();
-                if (icount_enabled()) {
-                    icount_prepare_for_run(cpu);
-                }
-                r = tcg_cpu_exec(cpu);
-                if (icount_enabled()) {
-                    icount_process_data(cpu);
-                }
-                bql_lock();
+                int r = rr_cpu_exec(cpu);
 
                 if (r == EXCP_DEBUG) {
                     cpu_handle_guest_debug(cpu);
-- 
2.49.0

Re: [PATCH v3 60/68] accel/tcg: Factor rr_cpu_exec() out for re-use

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops-rr.h |  2 ++
>   accel/tcg/tcg-accel-ops-rr.c | 31 ++++++++++++++++++++-----------
>   2 files changed, 22 insertions(+), 11 deletions(-)

I guess this is just completeness in matching mttcg, not that rr_cpu_exec will *ever* be 
called from outside tcg-accel-ops-rr.c.

Acked-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [PATCH v3 60/68] accel/tcg: Factor rr_cpu_exec() out for re-use

[Richard Henderson]

On 7/2/25 09:37, Richard Henderson wrote:
> On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   accel/tcg/tcg-accel-ops-rr.h |  2 ++
>>   accel/tcg/tcg-accel-ops-rr.c | 31 ++++++++++++++++++++-----------
>>   2 files changed, 22 insertions(+), 11 deletions(-)
> 
> I guess this is just completeness in matching mttcg, not that rr_cpu_exec will *ever* be 
> called from outside tcg-accel-ops-rr.c.
> 
> Acked-by: Richard Henderson <richard.henderson@linaro.org>

Alternately, split out the function but make it static; do *not* register it in patch 64.


r~

[PATCH v3 61/68] accel/tcg: Clear exit_request once in tcg_cpu_exec()

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops-mttcg.c | 1 -
 accel/tcg/tcg-accel-ops.c       | 3 +++
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
index 8a0295e2410..2fb6ced2572 100644
--- a/accel/tcg/tcg-accel-ops-mttcg.c
+++ b/accel/tcg/tcg-accel-ops-mttcg.c
@@ -130,7 +130,6 @@ void *mttcg_cpu_thread_routine(void *arg)
             }
         }
 
-        qatomic_set_mb(&cpu->exit_request, 0);
         qemu_wait_io_event(cpu);
     } while (!cpu->unplug || cpu_can_run(cpu));
 
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 83fb2d1362c..9b5caf9c4f5 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -95,6 +95,9 @@ int tcg_cpu_exec(CPUState *cpu)
     cpu_exec_start(cpu);
     ret = cpu_exec(cpu);
     cpu_exec_end(cpu);
+
+    qatomic_set_mb(&cpu->exit_request, 0);
+
     return ret;
 }
 
-- 
2.49.0

Re: [PATCH v3 61/68] accel/tcg: Clear exit_request once in tcg_cpu_exec()

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops-mttcg.c | 1 -
>   accel/tcg/tcg-accel-ops.c       | 3 +++
>   2 files changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/accel/tcg/tcg-accel-ops-mttcg.c b/accel/tcg/tcg-accel-ops-mttcg.c
> index 8a0295e2410..2fb6ced2572 100644
> --- a/accel/tcg/tcg-accel-ops-mttcg.c
> +++ b/accel/tcg/tcg-accel-ops-mttcg.c
> @@ -130,7 +130,6 @@ void *mttcg_cpu_thread_routine(void *arg)
>               }
>           }
>   
> -        qatomic_set_mb(&cpu->exit_request, 0);
>           qemu_wait_io_event(cpu);
>       } while (!cpu->unplug || cpu_can_run(cpu));
>   
> diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
> index 83fb2d1362c..9b5caf9c4f5 100644
> --- a/accel/tcg/tcg-accel-ops.c
> +++ b/accel/tcg/tcg-accel-ops.c
> @@ -95,6 +95,9 @@ int tcg_cpu_exec(CPUState *cpu)
>       cpu_exec_start(cpu);
>       ret = cpu_exec(cpu);
>       cpu_exec_end(cpu);
> +
> +    qatomic_set_mb(&cpu->exit_request, 0);
> +
>       return ret;
>   }
>   

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 62/68] accel/tcg: Add rr_vcpu_destroy() stub

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops-rr.h | 2 ++
 accel/tcg/tcg-accel-ops-rr.c | 5 +++++
 2 files changed, 7 insertions(+)

diff --git a/accel/tcg/tcg-accel-ops-rr.h b/accel/tcg/tcg-accel-ops-rr.h
index a73fe5b94a6..4234ef2f706 100644
--- a/accel/tcg/tcg-accel-ops-rr.h
+++ b/accel/tcg/tcg-accel-ops-rr.h
@@ -20,4 +20,6 @@ void rr_start_vcpu_thread(CPUState *cpu);
 
 int rr_cpu_exec(CPUState *cpu);
 
+void rr_vcpu_destroy(CPUState *cpu);
+
 #endif /* TCG_ACCEL_OPS_RR_H */
diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
index a9150802369..95c7d3a3172 100644
--- a/accel/tcg/tcg-accel-ops-rr.c
+++ b/accel/tcg/tcg-accel-ops-rr.c
@@ -137,6 +137,11 @@ static void rr_deal_with_unplugged_cpus(void)
     }
 }
 
+void rr_vcpu_destroy(CPUState *cpu)
+{
+    /* Already dealt with in rr_deal_with_unplugged_cpus() */
+}
+
 static void rr_force_rcu(Notifier *notify, void *data)
 {
     rr_kick_next_cpu();
-- 
2.49.0

Re: [PATCH v3 62/68] accel/tcg: Add rr_vcpu_destroy() stub

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops-rr.h | 2 ++
>   accel/tcg/tcg-accel-ops-rr.c | 5 +++++
>   2 files changed, 7 insertions(+)
> 
> diff --git a/accel/tcg/tcg-accel-ops-rr.h b/accel/tcg/tcg-accel-ops-rr.h
> index a73fe5b94a6..4234ef2f706 100644
> --- a/accel/tcg/tcg-accel-ops-rr.h
> +++ b/accel/tcg/tcg-accel-ops-rr.h
> @@ -20,4 +20,6 @@ void rr_start_vcpu_thread(CPUState *cpu);
>   
>   int rr_cpu_exec(CPUState *cpu);
>   
> +void rr_vcpu_destroy(CPUState *cpu);
> +
>   #endif /* TCG_ACCEL_OPS_RR_H */
> diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
> index a9150802369..95c7d3a3172 100644
> --- a/accel/tcg/tcg-accel-ops-rr.c
> +++ b/accel/tcg/tcg-accel-ops-rr.c
> @@ -137,6 +137,11 @@ static void rr_deal_with_unplugged_cpus(void)
>       }
>   }
>   
> +void rr_vcpu_destroy(CPUState *cpu)
> +{
> +    /* Already dealt with in rr_deal_with_unplugged_cpus() */
> +}
> +
>   static void rr_force_rcu(Notifier *notify, void *data)
>   {
>       rr_kick_next_cpu();

Why?  Is this hook mandatory?

r~

Re: [PATCH v3 62/68] accel/tcg: Add rr_vcpu_destroy() stub

[Philippe Mathieu-Daudé]

On 2/7/25 17:34, Richard Henderson wrote:
> On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   accel/tcg/tcg-accel-ops-rr.h | 2 ++
>>   accel/tcg/tcg-accel-ops-rr.c | 5 +++++
>>   2 files changed, 7 insertions(+)
>>
>> diff --git a/accel/tcg/tcg-accel-ops-rr.h b/accel/tcg/tcg-accel-ops-rr.h
>> index a73fe5b94a6..4234ef2f706 100644
>> --- a/accel/tcg/tcg-accel-ops-rr.h
>> +++ b/accel/tcg/tcg-accel-ops-rr.h
>> @@ -20,4 +20,6 @@ void rr_start_vcpu_thread(CPUState *cpu);
>>   int rr_cpu_exec(CPUState *cpu);
>> +void rr_vcpu_destroy(CPUState *cpu);
>> +
>>   #endif /* TCG_ACCEL_OPS_RR_H */
>> diff --git a/accel/tcg/tcg-accel-ops-rr.c b/accel/tcg/tcg-accel-ops-rr.c
>> index a9150802369..95c7d3a3172 100644
>> --- a/accel/tcg/tcg-accel-ops-rr.c
>> +++ b/accel/tcg/tcg-accel-ops-rr.c
>> @@ -137,6 +137,11 @@ static void rr_deal_with_unplugged_cpus(void)
>>       }
>>   }
>> +void rr_vcpu_destroy(CPUState *cpu)
>> +{
>> +    /* Already dealt with in rr_deal_with_unplugged_cpus() */
>> +}
>> +
>>   static void rr_force_rcu(Notifier *notify, void *data)
>>   {
>>       rr_kick_next_cpu();
> 
> Why?  Is this hook mandatory?

No, I'll drop.

[PATCH v3 63/68] accel/system: Declare init/exec/destroy vcpu hooks

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 include/system/accel-ops.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index 28e29cfa06d..f98a1c9b662 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -43,8 +43,11 @@ struct AccelOpsClass {
     void *(*cpu_thread_routine)(void *);
     void (*thread_precreate)(CPUState *cpu);
     void (*create_vcpu_thread)(CPUState *cpu);
+    int (*init_vcpu_thread)(CPUState *cpu);
     void (*kick_vcpu_thread)(CPUState *cpu);
     bool (*cpu_thread_is_idle)(CPUState *cpu);
+    int (*exec_vcpu_thread)(CPUState *cpu);
+    void (*destroy_vcpu_thread)(CPUState *cpu);
 
     /**
      * synchronize_post_reset:
-- 
2.49.0

Re: [PATCH v3 63/68] accel/system: Declare init/exec/destroy vcpu hooks

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   include/system/accel-ops.h | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
> index 28e29cfa06d..f98a1c9b662 100644
> --- a/include/system/accel-ops.h
> +++ b/include/system/accel-ops.h
> @@ -43,8 +43,11 @@ struct AccelOpsClass {
>       void *(*cpu_thread_routine)(void *);
>       void (*thread_precreate)(CPUState *cpu);
>       void (*create_vcpu_thread)(CPUState *cpu);
> +    int (*init_vcpu_thread)(CPUState *cpu);
>       void (*kick_vcpu_thread)(CPUState *cpu);
>       bool (*cpu_thread_is_idle)(CPUState *cpu);
> +    int (*exec_vcpu_thread)(CPUState *cpu);
> +    void (*destroy_vcpu_thread)(CPUState *cpu);
>   
>       /**
>        * synchronize_post_reset:

These could use some documentation, distinguishing them from e.g. cpu_thread_routine.


r~

[PATCH v3 64/68] accel/tcg: Register init/exec/destroy vcpu hooks

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index 9b5caf9c4f5..c6b5a567f9d 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -216,10 +216,14 @@ static void tcg_accel_ops_init(AccelClass *ac)
     if (qemu_tcg_mttcg_enabled()) {
         ops->cpu_thread_routine = mttcg_cpu_thread_routine;
         ops->kick_vcpu_thread = mttcg_kick_vcpu_thread;
+        ops->exec_vcpu_thread = mttcg_cpu_exec;
+        ops->destroy_vcpu_thread = tcg_cpu_destroy;
         ops->handle_interrupt = tcg_handle_interrupt;
     } else {
         ops->create_vcpu_thread = rr_start_vcpu_thread;
         ops->kick_vcpu_thread = rr_kick_vcpu_thread;
+        ops->exec_vcpu_thread = rr_cpu_exec;
+        ops->destroy_vcpu_thread = rr_vcpu_destroy;
 
         if (icount_enabled()) {
             ops->handle_interrupt = icount_handle_interrupt;
@@ -233,6 +237,7 @@ static void tcg_accel_ops_init(AccelClass *ac)
     ops->cpu_common_realize = tcg_exec_realizefn;
     ops->cpu_common_unrealize = tcg_exec_unrealizefn;
     ops->thread_precreate = tcg_vcpu_thread_precreate;
+    ops->init_vcpu_thread = tcg_vcpu_init,
     ops->cpu_reset_hold = tcg_cpu_reset_hold;
     ops->insert_breakpoint = tcg_insert_breakpoint;
     ops->remove_breakpoint = tcg_remove_breakpoint;
-- 
2.49.0

[PATCH v3 65/68] accel/hvf: Register init/exec/destroy vcpu hooks

[Philippe Mathieu-Daudé]

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/hvf/hvf-accel-ops.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index b61f08330f1..d2b3217f145 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -354,7 +354,10 @@ static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->cpu_target_realize = hvf_arch_cpu_realize;
 
     ops->cpu_thread_routine = hvf_cpu_thread_fn,
+    ops->init_vcpu_thread = hvf_init_vcpu,
     ops->kick_vcpu_thread = hvf_kick_vcpu_thread;
+    ops->exec_vcpu_thread = hvf_vcpu_exec;
+    ops->destroy_vcpu_thread = hvf_vcpu_destroy;
 
     ops->synchronize_post_reset = hvf_cpu_synchronize_post_reset;
     ops->synchronize_post_init = hvf_cpu_synchronize_post_init;
-- 
2.49.0

Re: [PATCH v3 65/68] accel/hvf: Register init/exec/destroy vcpu hooks

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/hvf/hvf-accel-ops.c | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
> index b61f08330f1..d2b3217f145 100644
> --- a/accel/hvf/hvf-accel-ops.c
> +++ b/accel/hvf/hvf-accel-ops.c
> @@ -354,7 +354,10 @@ static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
>       ops->cpu_target_realize = hvf_arch_cpu_realize;
>   
>       ops->cpu_thread_routine = hvf_cpu_thread_fn,
> +    ops->init_vcpu_thread = hvf_init_vcpu,
>       ops->kick_vcpu_thread = hvf_kick_vcpu_thread;
> +    ops->exec_vcpu_thread = hvf_vcpu_exec;
> +    ops->destroy_vcpu_thread = hvf_vcpu_destroy;
>   
>       ops->synchronize_post_reset = hvf_cpu_synchronize_post_reset;
>       ops->synchronize_post_init = hvf_cpu_synchronize_post_init;

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v3 66/68] accel/system: Pass old/new interrupt mask to handle_interrupt() handler

[Philippe Mathieu-Daudé]

Update CPUState::interrupt_request once in cpu_interrupt().
Pass the old and new masks along.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-accel-ops-icount.h |  2 +-
 accel/tcg/tcg-accel-ops.h        |  2 +-
 include/system/accel-ops.h       |  2 +-
 accel/tcg/tcg-accel-ops-icount.c |  8 +++-----
 accel/tcg/tcg-accel-ops.c        |  4 +---
 system/cpus.c                    | 12 +++++++-----
 6 files changed, 14 insertions(+), 16 deletions(-)

diff --git a/accel/tcg/tcg-accel-ops-icount.h b/accel/tcg/tcg-accel-ops-icount.h
index 5f3ebea50ff..ddd53cc1e4f 100644
--- a/accel/tcg/tcg-accel-ops-icount.h
+++ b/accel/tcg/tcg-accel-ops-icount.h
@@ -15,6 +15,6 @@ void icount_prepare_for_run(CPUState *cpu);
 void icount_update_percpu_budget(CPUState *cpu, int cpu_count);
 void icount_process_data(CPUState *cpu);
 
-void icount_handle_interrupt(CPUState *cpu, int mask);
+void icount_handle_interrupt(CPUState *cpu, int old_mask, int new_mask);
 
 #endif /* TCG_ACCEL_OPS_ICOUNT_H */
diff --git a/accel/tcg/tcg-accel-ops.h b/accel/tcg/tcg-accel-ops.h
index 1263a666774..a95d97fca29 100644
--- a/accel/tcg/tcg-accel-ops.h
+++ b/accel/tcg/tcg-accel-ops.h
@@ -17,7 +17,7 @@
 void tcg_vcpu_thread_precreate(CPUState *cpu);
 void tcg_cpu_destroy(CPUState *cpu);
 int tcg_cpu_exec(CPUState *cpu);
-void tcg_handle_interrupt(CPUState *cpu, int mask);
+void tcg_handle_interrupt(CPUState *cpu, int old_mask, int new_mask);
 void tcg_cpu_init_cflags(CPUState *cpu, bool parallel);
 
 int tcg_vcpu_init(CPUState *cpu);
diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index f98a1c9b662..9d2577fe67f 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -70,7 +70,7 @@ struct AccelOpsClass {
     void (*synchronize_state)(CPUState *cpu);
     void (*synchronize_pre_loadvm)(CPUState *cpu);
 
-    void (*handle_interrupt)(CPUState *cpu, int mask);
+    void (*handle_interrupt)(CPUState *cpu, int old_mask, int new_mask);
 
     void (*get_vcpu_stats)(CPUState *cpu, GString *buf);
 
diff --git a/accel/tcg/tcg-accel-ops-icount.c b/accel/tcg/tcg-accel-ops-icount.c
index ae1297ff7f3..d02b319951c 100644
--- a/accel/tcg/tcg-accel-ops-icount.c
+++ b/accel/tcg/tcg-accel-ops-icount.c
@@ -147,14 +147,12 @@ void icount_process_data(CPUState *cpu)
     replay_mutex_unlock();
 }
 
-void icount_handle_interrupt(CPUState *cpu, int mask)
+void icount_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
 {
-    int old_mask = cpu->interrupt_request;
-
-    tcg_handle_interrupt(cpu, mask);
+    tcg_handle_interrupt(cpu, old_mask, new_mask);
     if (qemu_cpu_is_self(cpu) &&
         !cpu->neg.can_do_io
-        && (mask & ~old_mask) != 0) {
+        && (new_mask & ~old_mask) != 0) {
         cpu_abort(cpu, "Raised interrupt while not in I/O function");
     }
 }
diff --git a/accel/tcg/tcg-accel-ops.c b/accel/tcg/tcg-accel-ops.c
index c6b5a567f9d..e7716dbc8da 100644
--- a/accel/tcg/tcg-accel-ops.c
+++ b/accel/tcg/tcg-accel-ops.c
@@ -109,10 +109,8 @@ static void tcg_cpu_reset_hold(CPUState *cpu)
 }
 
 /* mask must never be zero, except for A20 change call */
-void tcg_handle_interrupt(CPUState *cpu, int mask)
+void tcg_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
 {
-    cpu->interrupt_request |= mask;
-
     /*
      * If called from iothread context, wake the target cpu in
      * case its halted.
diff --git a/system/cpus.c b/system/cpus.c
index c2ad640980c..8c2647f5f19 100644
--- a/system/cpus.c
+++ b/system/cpus.c
@@ -246,10 +246,8 @@ int64_t cpus_get_elapsed_ticks(void)
     return cpu_get_ticks();
 }
 
-static void generic_handle_interrupt(CPUState *cpu, int mask)
+static void generic_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
 {
-    cpu->interrupt_request |= mask;
-
     if (!qemu_cpu_is_self(cpu)) {
         qemu_cpu_kick(cpu);
     }
@@ -257,12 +255,16 @@ static void generic_handle_interrupt(CPUState *cpu, int mask)
 
 void cpu_interrupt(CPUState *cpu, int mask)
 {
+    int old_mask = cpu->interrupt_request;
+
     g_assert(bql_locked());
 
+    cpu->interrupt_request |= mask;
+
     if (cpus_accel->handle_interrupt) {
-        cpus_accel->handle_interrupt(cpu, mask);
+        cpus_accel->handle_interrupt(cpu, old_mask, cpu->interrupt_request);
     } else {
-        generic_handle_interrupt(cpu, mask);
+        generic_handle_interrupt(cpu, old_mask, cpu->interrupt_request);
     }
 }
 
-- 
2.49.0

Re: [PATCH v3 66/68] accel/system: Pass old/new interrupt mask to handle_interrupt() handler

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Update CPUState::interrupt_request once in cpu_interrupt().
> Pass the old and new masks along.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-accel-ops-icount.h |  2 +-
>   accel/tcg/tcg-accel-ops.h        |  2 +-
>   include/system/accel-ops.h       |  2 +-
>   accel/tcg/tcg-accel-ops-icount.c |  8 +++-----
>   accel/tcg/tcg-accel-ops.c        |  4 +---
>   system/cpus.c                    | 12 +++++++-----
>   6 files changed, 14 insertions(+), 16 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

[PATCH v3 67/68] accel: Have each accelerator implement the handle_interrupt() hook

[Philippe Mathieu-Daudé]

Better to make the interrupt handler an explicit implementation
rather than depending on some generic handler.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/dummy-cpus.h                |  1 +
 include/system/accel-ops.h        |  1 +
 accel/dummy-cpus.c                |  7 +++++++
 accel/hvf/hvf-accel-ops.c         |  8 ++++++++
 accel/kvm/kvm-accel-ops.c         |  8 ++++++++
 accel/qtest/qtest.c               |  1 +
 accel/xen/xen-all.c               |  1 +
 system/cpus.c                     | 15 ++-------------
 target/i386/nvmm/nvmm-accel-ops.c |  8 ++++++++
 target/i386/whpx/whpx-accel-ops.c |  8 ++++++++
 10 files changed, 45 insertions(+), 13 deletions(-)

diff --git a/accel/dummy-cpus.h b/accel/dummy-cpus.h
index c2f9fee164c..98a1a30f9ca 100644
--- a/accel/dummy-cpus.h
+++ b/accel/dummy-cpus.h
@@ -11,5 +11,6 @@
 
 void dummy_thread_precreate(CPUState *cpu);
 void *dummy_cpu_thread_routine(void *arg);
+void dummy_handle_interrupt(CPUState *cpu, int old_mask, int new_mask);
 
 #endif
diff --git a/include/system/accel-ops.h b/include/system/accel-ops.h
index 9d2577fe67f..14861eae60c 100644
--- a/include/system/accel-ops.h
+++ b/include/system/accel-ops.h
@@ -70,6 +70,7 @@ struct AccelOpsClass {
     void (*synchronize_state)(CPUState *cpu);
     void (*synchronize_pre_loadvm)(CPUState *cpu);
 
+    /* handle_interrupt is mandatory. */
     void (*handle_interrupt)(CPUState *cpu, int old_mask, int new_mask);
 
     void (*get_vcpu_stats)(CPUState *cpu, GString *buf);
diff --git a/accel/dummy-cpus.c b/accel/dummy-cpus.c
index f637ab05e32..e9076851c58 100644
--- a/accel/dummy-cpus.c
+++ b/accel/dummy-cpus.c
@@ -71,3 +71,10 @@ void dummy_thread_precreate(CPUState *cpu)
     qemu_sem_init(&cpu->sem, 0);
 #endif
 }
+
+void dummy_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
+{
+    if (!qemu_cpu_is_self(cpu)) {
+        qemu_cpu_kick(cpu);
+    }
+}
diff --git a/accel/hvf/hvf-accel-ops.c b/accel/hvf/hvf-accel-ops.c
index d2b3217f145..a9ed93fe8eb 100644
--- a/accel/hvf/hvf-accel-ops.c
+++ b/accel/hvf/hvf-accel-ops.c
@@ -207,6 +207,13 @@ static void *hvf_cpu_thread_fn(void *arg)
     return NULL;
 }
 
+static void hvf_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
+{
+    if (!qemu_cpu_is_self(cpu)) {
+        qemu_cpu_kick(cpu);
+    }
+}
+
 struct hvf_sw_breakpoint *hvf_find_sw_breakpoint(CPUState *cpu, vaddr pc)
 {
     struct hvf_sw_breakpoint *bp;
@@ -358,6 +365,7 @@ static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->kick_vcpu_thread = hvf_kick_vcpu_thread;
     ops->exec_vcpu_thread = hvf_vcpu_exec;
     ops->destroy_vcpu_thread = hvf_vcpu_destroy;
+    ops->handle_interrupt = hvf_handle_interrupt;
 
     ops->synchronize_post_reset = hvf_cpu_synchronize_post_reset;
     ops->synchronize_post_init = hvf_cpu_synchronize_post_init;
diff --git a/accel/kvm/kvm-accel-ops.c b/accel/kvm/kvm-accel-ops.c
index 21ff3af306f..749c4f244a1 100644
--- a/accel/kvm/kvm-accel-ops.c
+++ b/accel/kvm/kvm-accel-ops.c
@@ -68,6 +68,13 @@ static bool kvm_vcpu_thread_is_idle(CPUState *cpu)
     return !kvm_halt_in_kernel();
 }
 
+static void kvm_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
+{
+    if (!qemu_cpu_is_self(cpu)) {
+        qemu_cpu_kick(cpu);
+    }
+}
+
 #ifdef TARGET_KVM_HAVE_GUEST_DEBUG
 static int kvm_update_guest_debug_ops(CPUState *cpu)
 {
@@ -85,6 +92,7 @@ static void kvm_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->synchronize_post_init = kvm_cpu_synchronize_post_init;
     ops->synchronize_state = kvm_cpu_synchronize_state;
     ops->synchronize_pre_loadvm = kvm_cpu_synchronize_pre_loadvm;
+    ops->handle_interrupt = kvm_handle_interrupt;
 
 #ifdef TARGET_KVM_HAVE_GUEST_DEBUG
     ops->update_guest_debug = kvm_update_guest_debug_ops;
diff --git a/accel/qtest/qtest.c b/accel/qtest/qtest.c
index 9f30098d133..19eea8d8daa 100644
--- a/accel/qtest/qtest.c
+++ b/accel/qtest/qtest.c
@@ -68,6 +68,7 @@ static void qtest_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->cpu_thread_routine = dummy_cpu_thread_routine;
     ops->get_virtual_clock = qtest_get_virtual_clock;
     ops->set_virtual_clock = qtest_set_virtual_clock;
+    ops->handle_interrupt = dummy_handle_interrupt;
 };
 
 static const TypeInfo qtest_accel_ops_type = {
diff --git a/accel/xen/xen-all.c b/accel/xen/xen-all.c
index 5ff72d9532c..6a967a8c63d 100644
--- a/accel/xen/xen-all.c
+++ b/accel/xen/xen-all.c
@@ -153,6 +153,7 @@ static void xen_accel_ops_class_init(ObjectClass *oc, const void *data)
 
     ops->thread_precreate = dummy_thread_precreate;
     ops->cpu_thread_routine = dummy_cpu_thread_routine;
+    ops->handle_interrupt = dummy_handle_interrupt;
 }
 
 static const TypeInfo xen_accel_ops_type = {
diff --git a/system/cpus.c b/system/cpus.c
index 8c2647f5f19..e217e42ba03 100644
--- a/system/cpus.c
+++ b/system/cpus.c
@@ -246,13 +246,6 @@ int64_t cpus_get_elapsed_ticks(void)
     return cpu_get_ticks();
 }
 
-static void generic_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
-{
-    if (!qemu_cpu_is_self(cpu)) {
-        qemu_cpu_kick(cpu);
-    }
-}
-
 void cpu_interrupt(CPUState *cpu, int mask)
 {
     int old_mask = cpu->interrupt_request;
@@ -260,12 +253,7 @@ void cpu_interrupt(CPUState *cpu, int mask)
     g_assert(bql_locked());
 
     cpu->interrupt_request |= mask;
-
-    if (cpus_accel->handle_interrupt) {
-        cpus_accel->handle_interrupt(cpu, old_mask, cpu->interrupt_request);
-    } else {
-        generic_handle_interrupt(cpu, old_mask, cpu->interrupt_request);
-    }
+    cpus_accel->handle_interrupt(cpu, old_mask, cpu->interrupt_request);
 }
 
 /*
@@ -674,6 +662,7 @@ void cpus_register_accel(const AccelOpsClass *ops)
 {
     assert(ops != NULL);
     assert(ops->create_vcpu_thread || ops->cpu_thread_routine);
+    assert(ops->handle_interrupt);
     cpus_accel = ops;
 }
 
diff --git a/target/i386/nvmm/nvmm-accel-ops.c b/target/i386/nvmm/nvmm-accel-ops.c
index bef6f61b776..62fc6438c60 100644
--- a/target/i386/nvmm/nvmm-accel-ops.c
+++ b/target/i386/nvmm/nvmm-accel-ops.c
@@ -61,6 +61,13 @@ static void *qemu_nvmm_cpu_thread_fn(void *arg)
     return NULL;
 }
 
+static void nvmm_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
+{
+    if (!qemu_cpu_is_self(cpu)) {
+        qemu_cpu_kick(cpu);
+    }
+}
+
 /*
  * Abort the call to run the virtual processor by another thread, and to
  * return the control to that thread.
@@ -77,6 +84,7 @@ static void nvmm_accel_ops_class_init(ObjectClass *oc, const void *data)
 
     ops->cpu_thread_routine = qemu_nvmm_cpu_thread_fn;
     ops->kick_vcpu_thread = nvmm_kick_vcpu_thread;
+    ops->handle_interrupt = nvmm_handle_interrupt;
 
     ops->synchronize_post_reset = nvmm_cpu_synchronize_post_reset;
     ops->synchronize_post_init = nvmm_cpu_synchronize_post_init;
diff --git a/target/i386/whpx/whpx-accel-ops.c b/target/i386/whpx/whpx-accel-ops.c
index 8cbc6f4e2d8..e9969ef2cf3 100644
--- a/target/i386/whpx/whpx-accel-ops.c
+++ b/target/i386/whpx/whpx-accel-ops.c
@@ -61,6 +61,13 @@ static void *whpx_cpu_thread_fn(void *arg)
     return NULL;
 }
 
+static void whpx_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
+{
+    if (!qemu_cpu_is_self(cpu)) {
+        qemu_cpu_kick(cpu);
+    }
+}
+
 static void whpx_kick_vcpu_thread(CPUState *cpu)
 {
     if (!qemu_cpu_is_self(cpu)) {
@@ -80,6 +87,7 @@ static void whpx_accel_ops_class_init(ObjectClass *oc, const void *data)
     ops->cpu_thread_routine = whpx_cpu_thread_fn;
     ops->kick_vcpu_thread = whpx_kick_vcpu_thread;
     ops->cpu_thread_is_idle = whpx_vcpu_thread_is_idle;
+    ops->handle_interrupt = whpx_handle_interrupt;
 
     ops->synchronize_post_reset = whpx_cpu_synchronize_post_reset;
     ops->synchronize_post_init = whpx_cpu_synchronize_post_init;
-- 
2.49.0

Re: [PATCH v3 67/68] accel: Have each accelerator implement the handle_interrupt() hook

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Better to make the interrupt handler an explicit implementation
> rather than depending on some generic handler.
> 
> Suggested-by: Richard Henderson <richard.henderson@linaro.org>
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/dummy-cpus.h                |  1 +
>   include/system/accel-ops.h        |  1 +
>   accel/dummy-cpus.c                |  7 +++++++
>   accel/hvf/hvf-accel-ops.c         |  8 ++++++++
>   accel/kvm/kvm-accel-ops.c         |  8 ++++++++
>   accel/qtest/qtest.c               |  1 +
>   accel/xen/xen-all.c               |  1 +
>   system/cpus.c                     | 15 ++-------------
>   target/i386/nvmm/nvmm-accel-ops.c |  8 ++++++++
>   target/i386/whpx/whpx-accel-ops.c |  8 ++++++++
>   10 files changed, 45 insertions(+), 13 deletions(-)

I had suggested exporting generic_handle_interrupt

> @@ -358,6 +365,7 @@ static void hvf_accel_ops_class_init(ObjectClass *oc, const void *data)
>       ops->kick_vcpu_thread = hvf_kick_vcpu_thread;
>       ops->exec_vcpu_thread = hvf_vcpu_exec;
>       ops->destroy_vcpu_thread = hvf_vcpu_destroy;
> +    ops->handle_interrupt = hvf_handle_interrupt;

... and then registering it in each accelerator,

> diff --git a/system/cpus.c b/system/cpus.c
> index 8c2647f5f19..e217e42ba03 100644
> --- a/system/cpus.c
> +++ b/system/cpus.c
> @@ -246,13 +246,6 @@ int64_t cpus_get_elapsed_ticks(void)
>       return cpu_get_ticks();
>   }
>   
> -static void generic_handle_interrupt(CPUState *cpu, int old_mask, int new_mask)
> -{
> -    if (!qemu_cpu_is_self(cpu)) {
> -        qemu_cpu_kick(cpu);
> -    }
> -}

... rather than removing the common implementation.


r~

Re: [PATCH v3 67/68] accel: Have each accelerator implement the handle_interrupt() hook

[Philippe Mathieu-Daudé]

On 2/7/25 17:30, Richard Henderson wrote:
> On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
>> Better to make the interrupt handler an explicit implementation
>> rather than depending on some generic handler.
>>
>> Suggested-by: Richard Henderson <richard.henderson@linaro.org>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   accel/dummy-cpus.h                |  1 +
>>   include/system/accel-ops.h        |  1 +
>>   accel/dummy-cpus.c                |  7 +++++++
>>   accel/hvf/hvf-accel-ops.c         |  8 ++++++++
>>   accel/kvm/kvm-accel-ops.c         |  8 ++++++++
>>   accel/qtest/qtest.c               |  1 +
>>   accel/xen/xen-all.c               |  1 +
>>   system/cpus.c                     | 15 ++-------------
>>   target/i386/nvmm/nvmm-accel-ops.c |  8 ++++++++
>>   target/i386/whpx/whpx-accel-ops.c |  8 ++++++++
>>   10 files changed, 45 insertions(+), 13 deletions(-)
> 
> I had suggested exporting generic_handle_interrupt
> 
>> @@ -358,6 +365,7 @@ static void hvf_accel_ops_class_init(ObjectClass 
>> *oc, const void *data)
>>       ops->kick_vcpu_thread = hvf_kick_vcpu_thread;
>>       ops->exec_vcpu_thread = hvf_vcpu_exec;
>>       ops->destroy_vcpu_thread = hvf_vcpu_destroy;
>> +    ops->handle_interrupt = hvf_handle_interrupt;
> 
> ... and then registering it in each accelerator,
> 
>> diff --git a/system/cpus.c b/system/cpus.c
>> index 8c2647f5f19..e217e42ba03 100644
>> --- a/system/cpus.c
>> +++ b/system/cpus.c
>> @@ -246,13 +246,6 @@ int64_t cpus_get_elapsed_ticks(void)
>>       return cpu_get_ticks();
>>   }
>> -static void generic_handle_interrupt(CPUState *cpu, int old_mask, int 
>> new_mask)
>> -{
>> -    if (!qemu_cpu_is_self(cpu)) {
>> -        qemu_cpu_kick(cpu);
>> -    }
>> -}
> 
> ... rather than removing the common implementation.

Oh I misunderstood then.

[RFC PATCH v3 68/68] system/memory: Restrict eventfd dispatch_write() to emulators

[Philippe Mathieu-Daudé]

Commit 8c56c1a592b ("memory: emulate ioeventfd") added a !KVM
check because the only accelerator available back then were TCG,
QTest and KVM. Then commit 126e7f78036 ("kvm: require
KVM_CAP_IOEVENTFD and KVM_CAP_IOEVENTFD_ANY_LENGTH") suggested
'!KVM' check should be '(TCG || QTest)'. Later more accelerator
were added. Implement the suggestion as a safety measure, not
dispatching to eventfd when hardware accelerator is used.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 system/memory.c | 9 ++-------
 1 file changed, 2 insertions(+), 7 deletions(-)

diff --git a/system/memory.c b/system/memory.c
index e8d9b15b28f..b072a6bef83 100644
--- a/system/memory.c
+++ b/system/memory.c
@@ -25,7 +25,7 @@
 #include "qom/object.h"
 #include "trace.h"
 #include "system/ram_addr.h"
-#include "system/kvm.h"
+#include "system/qtest.h"
 #include "system/runstate.h"
 #include "system/tcg.h"
 #include "qemu/accel.h"
@@ -1530,12 +1530,7 @@ MemTxResult memory_region_dispatch_write(MemoryRegion *mr,
 
     adjust_endianness(mr, &data, op);
 
-    /*
-     * FIXME: it's not clear why under KVM the write would be processed
-     * directly, instead of going through eventfd.  This probably should
-     * test "tcg_enabled() || qtest_enabled()", or should just go away.
-     */
-    if (!kvm_enabled() &&
+    if ((tcg_enabled() || qtest_enabled()) &&
         memory_region_dispatch_write_eventfds(mr, addr, data, size, attrs)) {
         return MEMTX_OK;
     }
-- 
2.49.0

Re: [RFC PATCH v3 68/68] system/memory: Restrict eventfd dispatch_write() to emulators

[Richard Henderson]

On 7/1/25 08:40, Philippe Mathieu-Daudé wrote:
> Commit 8c56c1a592b ("memory: emulate ioeventfd") added a !KVM
> check because the only accelerator available back then were TCG,
> QTest and KVM. Then commit 126e7f78036 ("kvm: require
> KVM_CAP_IOEVENTFD and KVM_CAP_IOEVENTFD_ANY_LENGTH") suggested
> '!KVM' check should be '(TCG || QTest)'. Later more accelerator
> were added. Implement the suggestion as a safety measure, not
> dispatching to eventfd when hardware accelerator is used.
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   system/memory.c | 9 ++-------
>   1 file changed, 2 insertions(+), 7 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~