[PATCH-for-10.1 0/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[PATCH-for-10.1 0/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Philippe Mathieu-Daudé]

In this series we replace the TARGET_SUPPORTS_MTTCG (Makefile)
definition by a 'mttcg_supported' field in TCGCPUOps.

Based-on: <20250321125737.72839-1-philmd@linaro.org>

Philippe Mathieu-Daudé (4):
  target/riscv: Restrict RV128 MTTCG check on system emulation
  tcg: Move qemu_tcg_mttcg_enabled() to 'system/tcg.h'
  tcg: Convert TCGState::mttcg_enabled to TriState
  tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

 docs/devel/multi-thread-tcg.rst          |  2 +-
 configs/targets/aarch64-softmmu.mak      |  1 -
 configs/targets/alpha-softmmu.mak        |  1 -
 configs/targets/arm-softmmu.mak          |  1 -
 configs/targets/hppa-softmmu.mak         |  1 -
 configs/targets/i386-softmmu.mak         |  1 -
 configs/targets/loongarch64-softmmu.mak  |  1 -
 configs/targets/microblaze-softmmu.mak   |  1 -
 configs/targets/microblazeel-softmmu.mak |  1 -
 configs/targets/mips-softmmu.mak         |  1 -
 configs/targets/mipsel-softmmu.mak       |  1 -
 configs/targets/or1k-softmmu.mak         |  1 -
 configs/targets/ppc64-softmmu.mak        |  1 -
 configs/targets/riscv32-softmmu.mak      |  1 -
 configs/targets/riscv64-softmmu.mak      |  1 -
 configs/targets/s390x-softmmu.mak        |  1 -
 configs/targets/sparc-softmmu.mak        |  1 -
 configs/targets/sparc64-softmmu.mak      |  1 -
 configs/targets/x86_64-softmmu.mak       |  1 -
 configs/targets/xtensa-softmmu.mak       |  1 -
 configs/targets/xtensaeb-softmmu.mak     |  1 -
 include/accel/tcg/cpu-ops.h              |  8 +++
 include/exec/poison.h                    |  1 -
 include/hw/core/cpu.h                    |  9 ----
 include/system/tcg.h                     |  8 +++
 accel/tcg/tcg-all.c                      | 68 +++++++++++++-----------
 target/alpha/cpu.c                       |  1 +
 target/arm/cpu.c                         |  1 +
 target/arm/tcg/cpu-v7m.c                 |  1 +
 target/avr/cpu.c                         |  1 +
 target/hexagon/cpu.c                     |  1 +
 target/hppa/cpu.c                        |  1 +
 target/i386/tcg/tcg-cpu.c                |  1 +
 target/loongarch/cpu.c                   |  1 +
 target/m68k/cpu.c                        |  1 +
 target/microblaze/cpu.c                  |  1 +
 target/mips/cpu.c                        |  1 +
 target/openrisc/cpu.c                    |  1 +
 target/ppc/cpu_init.c                    |  1 +
 target/riscv/tcg/tcg-cpu.c               |  4 +-
 target/rx/cpu.c                          |  1 +
 target/s390x/cpu.c                       |  1 +
 target/sh4/cpu.c                         |  1 +
 target/sparc/cpu.c                       |  1 +
 target/tricore/cpu.c                     |  1 +
 target/xtensa/cpu.c                      |  1 +
 tcg/region.c                             |  4 +-
 47 files changed, 78 insertions(+), 65 deletions(-)

-- 
2.47.1

[PATCH-for-10.1 1/4] target/riscv: Restrict RV128 MTTCG check on system emulation

[Philippe Mathieu-Daudé]

Multi-threaded TCG only concerns system emulation.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 target/riscv/tcg/tcg-cpu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
index fb903992faa..60a26acc503 100644
--- a/target/riscv/tcg/tcg-cpu.c
+++ b/target/riscv/tcg/tcg-cpu.c
@@ -1050,6 +1050,7 @@ static bool riscv_tcg_cpu_realize(CPUState *cs, Error **errp)
         return false;
     }
 
+#ifndef CONFIG_USER_ONLY
     if (mcc->misa_mxl_max >= MXL_RV128 && qemu_tcg_mttcg_enabled()) {
         /* Missing 128-bit aligned atomics */
         error_setg(errp,
@@ -1058,7 +1059,6 @@ static bool riscv_tcg_cpu_realize(CPUState *cs, Error **errp)
         return false;
     }
 
-#ifndef CONFIG_USER_ONLY
     CPURISCVState *env = &cpu->env;
 
     tcg_cflags_set(CPU(cs), CF_PCREL);
-- 
2.47.1

Re: [PATCH-for-10.1 1/4] target/riscv: Restrict RV128 MTTCG check on system emulation

[Pierrick Bouvier]

On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> Multi-threaded TCG only concerns system emulation.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   target/riscv/tcg/tcg-cpu.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
> index fb903992faa..60a26acc503 100644
> --- a/target/riscv/tcg/tcg-cpu.c
> +++ b/target/riscv/tcg/tcg-cpu.c
> @@ -1050,6 +1050,7 @@ static bool riscv_tcg_cpu_realize(CPUState *cs, Error **errp)
>           return false;
>       }
>   
> +#ifndef CONFIG_USER_ONLY
>       if (mcc->misa_mxl_max >= MXL_RV128 && qemu_tcg_mttcg_enabled()) {
>           /* Missing 128-bit aligned atomics */
>           error_setg(errp,
> @@ -1058,7 +1059,6 @@ static bool riscv_tcg_cpu_realize(CPUState *cs, Error **errp)
>           return false;
>       }
>   
> -#ifndef CONFIG_USER_ONLY
>       CPURISCVState *env = &cpu->env;
>   
>       tcg_cflags_set(CPU(cs), CF_PCREL);

Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>

Re: [PATCH-for-10.1 1/4] target/riscv: Restrict RV128 MTTCG check on system emulation

[Richard Henderson]

On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> Multi-threaded TCG only concerns system emulation.

That's not really true.  User emulation simply has no option to
run in a single-threaded context.

I really don't think we should allow RV128 in user-mode at all.
Certainly not until there's a kernel abi for it.


r~

Re: [PATCH-for-10.1 1/4] target/riscv: Restrict RV128 MTTCG check on system emulation

[Philippe Mathieu-Daudé]

On 23/3/25 19:08, Richard Henderson wrote:
> On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
>> Multi-threaded TCG only concerns system emulation.
> 
> That's not really true.  User emulation simply has no option to
> run in a single-threaded context.
> 
> I really don't think we should allow RV128 in user-mode at all.
> Certainly not until there's a kernel abi for it.

It seems to be safe since commit 905b9fcde1f ("target/riscv: Replace
is_32bit with get_xl/get_xlen"):

  #ifdef TARGET_RISCV32
  #define get_xl(ctx)    MXL_RV32
  #elif defined(CONFIG_USER_ONLY)
  #define get_xl(ctx)    MXL_RV64
  #else
  #define get_xl(ctx)    ((ctx)->xl)
  #endif

Should we undefine MXL_RV128 on user-mode?

Re: [PATCH-for-10.1 1/4] target/riscv: Restrict RV128 MTTCG check on system emulation

[Philippe Mathieu-Daudé]

On 2/4/25 16:25, Philippe Mathieu-Daudé wrote:
> On 23/3/25 19:08, Richard Henderson wrote:
>> On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
>>> Multi-threaded TCG only concerns system emulation.
>>
>> That's not really true.  User emulation simply has no option to
>> run in a single-threaded context.
>>
>> I really don't think we should allow RV128 in user-mode at all.
>> Certainly not until there's a kernel abi for it.
> 
> It seems to be safe since commit 905b9fcde1f ("target/riscv: Replace
> is_32bit with get_xl/get_xlen"):
> 
>   #ifdef TARGET_RISCV32
>   #define get_xl(ctx)    MXL_RV32
>   #elif defined(CONFIG_USER_ONLY)
>   #define get_xl(ctx)    MXL_RV64
>   #else
>   #define get_xl(ctx)    ((ctx)->xl)
>   #endif
> 
> Should we undefine MXL_RV128 on user-mode?

Indeed the CPU is exposed on user-mode...

$ qemu-riscv64 -cpu help
Available CPUs:
   max
   rv64
   rv64e
   rv64i
   rva22s64
   rva22u64
   rva23s64
   rva23u64
   shakti-c
   sifive-e51
   sifive-u54
   thead-c906
   tt-ascalon
   veyron-v1
   x-rv128           <---------
   xiangshan-nanhu

Per commit 6df3747a274 ("riscv: Introduce satp mode hw
capabilities") I wonder if this is expected.

Anyhow, I'll post a patch disabling it as:

-- >8 --
diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 430b02d2a58..33abcef0073 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -699,3 +699,3 @@ static void rv64_xiangshan_nanhu_cpu_init(Object *obj)

-#ifdef CONFIG_TCG
+#if defined(CONFIG_TCG) && !defined(CONFIG_USER_ONLY)
  static void rv128_base_cpu_init(Object *obj)
@@ -710,7 +710,6 @@ static void rv128_base_cpu_init(Object *obj)
      env->priv_ver = PRIV_VERSION_LATEST;
-#ifndef CONFIG_USER_ONLY
+
      set_satp_mode_max_supported(RISCV_CPU(obj), VM_1_10_SV57);
-#endif
  }
-#endif /* CONFIG_TCG */
+#endif /* CONFIG_TCG && !CONFIG_USER_ONLY */

@@ -3257,3 +3256,3 @@ static const TypeInfo riscv_cpu_type_infos[] = {
                                                   MXL_RV64, 
rv64_xiangshan_nanhu_cpu_init),
-#ifdef CONFIG_TCG
+#if defined(CONFIG_TCG) && !defined(CONFIG_USER_ONLY)
      DEFINE_DYNAMIC_CPU(TYPE_RISCV_CPU_BASE128,   MXL_RV128, 
rv128_base_cpu_init),

---

Re: [PATCH-for-10.1 1/4] target/riscv: Restrict RV128 MTTCG check on system emulation

[Alistair Francis]

On Thu, Apr 3, 2025 at 12:41 AM Philippe Mathieu-Daudé
<philmd@linaro.org> wrote:
>
> On 2/4/25 16:25, Philippe Mathieu-Daudé wrote:
> > On 23/3/25 19:08, Richard Henderson wrote:
> >> On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> >>> Multi-threaded TCG only concerns system emulation.
> >>
> >> That's not really true.  User emulation simply has no option to
> >> run in a single-threaded context.
> >>
> >> I really don't think we should allow RV128 in user-mode at all.
> >> Certainly not until there's a kernel abi for it.
> >
> > It seems to be safe since commit 905b9fcde1f ("target/riscv: Replace
> > is_32bit with get_xl/get_xlen"):
> >
> >   #ifdef TARGET_RISCV32
> >   #define get_xl(ctx)    MXL_RV32
> >   #elif defined(CONFIG_USER_ONLY)
> >   #define get_xl(ctx)    MXL_RV64
> >   #else
> >   #define get_xl(ctx)    ((ctx)->xl)
> >   #endif
> >
> > Should we undefine MXL_RV128 on user-mode?
>
> Indeed the CPU is exposed on user-mode...
>
> $ qemu-riscv64 -cpu help
> Available CPUs:
>    max
>    rv64
>    rv64e
>    rv64i
>    rva22s64
>    rva22u64
>    rva23s64
>    rva23u64
>    shakti-c
>    sifive-e51
>    sifive-u54
>    thead-c906
>    tt-ascalon
>    veyron-v1
>    x-rv128           <---------
>    xiangshan-nanhu
>
> Per commit 6df3747a274 ("riscv: Introduce satp mode hw
> capabilities") I wonder if this is expected.

We probably didn't really think about it at the time.

I agree that we don't need it

>
> Anyhow, I'll post a patch disabling it as:

Thanks

Alistair

>
> -- >8 --
> diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
> index 430b02d2a58..33abcef0073 100644
> --- a/target/riscv/cpu.c
> +++ b/target/riscv/cpu.c
> @@ -699,3 +699,3 @@ static void rv64_xiangshan_nanhu_cpu_init(Object *obj)
>
> -#ifdef CONFIG_TCG
> +#if defined(CONFIG_TCG) && !defined(CONFIG_USER_ONLY)
>   static void rv128_base_cpu_init(Object *obj)
> @@ -710,7 +710,6 @@ static void rv128_base_cpu_init(Object *obj)
>       env->priv_ver = PRIV_VERSION_LATEST;
> -#ifndef CONFIG_USER_ONLY
> +
>       set_satp_mode_max_supported(RISCV_CPU(obj), VM_1_10_SV57);
> -#endif
>   }
> -#endif /* CONFIG_TCG */
> +#endif /* CONFIG_TCG && !CONFIG_USER_ONLY */
>
> @@ -3257,3 +3256,3 @@ static const TypeInfo riscv_cpu_type_infos[] = {
>                                                    MXL_RV64,
> rv64_xiangshan_nanhu_cpu_init),
> -#ifdef CONFIG_TCG
> +#if defined(CONFIG_TCG) && !defined(CONFIG_USER_ONLY)
>       DEFINE_DYNAMIC_CPU(TYPE_RISCV_CPU_BASE128,   MXL_RV128,
> rv128_base_cpu_init),
>
> ---
>

[PATCH-for-10.1 2/4] tcg: Move qemu_tcg_mttcg_enabled() to 'system/tcg.h'

[Philippe Mathieu-Daudé]

qemu_tcg_mttcg_enabled() is specific to 1/ TCG and
2/ system emulation. Move the prototype declaration
to "system/tcg.h", reducing 'mttcg_enabled' variable
scope.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 include/hw/core/cpu.h      |  9 ---------
 include/system/tcg.h       |  8 ++++++++
 accel/tcg/tcg-all.c        | 11 ++++++++++-
 target/riscv/tcg/tcg-cpu.c |  1 +
 tcg/region.c               |  4 +++-
 5 files changed, 22 insertions(+), 11 deletions(-)

diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 5d11d26556a..54570d21aea 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -597,15 +597,6 @@ extern CPUTailQ cpus_queue;
 
 extern __thread CPUState *current_cpu;
 
-/**
- * qemu_tcg_mttcg_enabled:
- * Check whether we are running MultiThread TCG or not.
- *
- * Returns: %true if we are in MTTCG mode %false otherwise.
- */
-extern bool mttcg_enabled;
-#define qemu_tcg_mttcg_enabled() (mttcg_enabled)
-
 /**
  * cpu_paging_enabled:
  * @cpu: The CPU whose state is to be inspected.
diff --git a/include/system/tcg.h b/include/system/tcg.h
index 73229648c63..7622dcea302 100644
--- a/include/system/tcg.h
+++ b/include/system/tcg.h
@@ -17,4 +17,12 @@ extern bool tcg_allowed;
 #define tcg_enabled() 0
 #endif
 
+/**
+ * qemu_tcg_mttcg_enabled:
+ * Check whether we are running MultiThread TCG or not.
+ *
+ * Returns: %true if we are in MTTCG mode %false otherwise.
+ */
+bool qemu_tcg_mttcg_enabled(void);
+
 #endif
diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index cb632cc8cc7..d75ecf531b6 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -38,6 +38,7 @@
 #include "hw/qdev-core.h"
 #else
 #include "hw/boards.h"
+#include "system/tcg.h"
 #endif
 #include "internal-common.h"
 #include "cpu-param.h"
@@ -58,6 +59,15 @@ typedef struct TCGState TCGState;
 DECLARE_INSTANCE_CHECKER(TCGState, TCG_STATE,
                          TYPE_TCG_ACCEL)
 
+static bool mttcg_enabled;
+
+#ifndef CONFIG_USER_ONLY
+bool qemu_tcg_mttcg_enabled(void)
+{
+    return mttcg_enabled;
+}
+#endif
+
 /*
  * We default to false if we know other options have been enabled
  * which are currently incompatible with MTTCG. Otherwise when each
@@ -97,7 +107,6 @@ static void tcg_accel_instance_init(Object *obj)
 #endif
 }
 
-bool mttcg_enabled;
 bool one_insn_per_tb;
 
 static int tcg_init_machine(MachineState *ms)
diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
index 60a26acc503..5ccd7b9dc07 100644
--- a/target/riscv/tcg/tcg-cpu.c
+++ b/target/riscv/tcg/tcg-cpu.c
@@ -35,6 +35,7 @@
 #include "tcg/tcg.h"
 #ifndef CONFIG_USER_ONLY
 #include "hw/boards.h"
+#include "system/tcg.h"
 #endif
 
 /* Hash that stores user set extensions */
diff --git a/tcg/region.c b/tcg/region.c
index 478ec051c4b..56d2e988719 100644
--- a/tcg/region.c
+++ b/tcg/region.c
@@ -34,7 +34,9 @@
 #include "exec/translation-block.h"
 #include "tcg-internal.h"
 #include "host/cpuinfo.h"
-
+#ifndef CONFIG_USER_ONLY
+#include "system/tcg.h"
+#endif
 
 /*
  * Local source-level compatibility with Unix.
-- 
2.47.1

Re: [PATCH-for-10.1 2/4] tcg: Move qemu_tcg_mttcg_enabled() to 'system/tcg.h'

[Pierrick Bouvier]

On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> qemu_tcg_mttcg_enabled() is specific to 1/ TCG and
> 2/ system emulation. Move the prototype declaration
> to "system/tcg.h", reducing 'mttcg_enabled' variable
> scope.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   include/hw/core/cpu.h      |  9 ---------
>   include/system/tcg.h       |  8 ++++++++
>   accel/tcg/tcg-all.c        | 11 ++++++++++-
>   target/riscv/tcg/tcg-cpu.c |  1 +
>   tcg/region.c               |  4 +++-
>   5 files changed, 22 insertions(+), 11 deletions(-)
> 
> diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
> index 5d11d26556a..54570d21aea 100644
> --- a/include/hw/core/cpu.h
> +++ b/include/hw/core/cpu.h
> @@ -597,15 +597,6 @@ extern CPUTailQ cpus_queue;
>   
>   extern __thread CPUState *current_cpu;
>   
> -/**
> - * qemu_tcg_mttcg_enabled:
> - * Check whether we are running MultiThread TCG or not.
> - *
> - * Returns: %true if we are in MTTCG mode %false otherwise.
> - */
> -extern bool mttcg_enabled;
> -#define qemu_tcg_mttcg_enabled() (mttcg_enabled)
> -
>   /**
>    * cpu_paging_enabled:
>    * @cpu: The CPU whose state is to be inspected.
> diff --git a/include/system/tcg.h b/include/system/tcg.h
> index 73229648c63..7622dcea302 100644
> --- a/include/system/tcg.h
> +++ b/include/system/tcg.h
> @@ -17,4 +17,12 @@ extern bool tcg_allowed;
>   #define tcg_enabled() 0
>   #endif
>   
> +/**
> + * qemu_tcg_mttcg_enabled:
> + * Check whether we are running MultiThread TCG or not.
> + *
> + * Returns: %true if we are in MTTCG mode %false otherwise.
> + */
> +bool qemu_tcg_mttcg_enabled(void);
> +
>   #endif
> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
> index cb632cc8cc7..d75ecf531b6 100644
> --- a/accel/tcg/tcg-all.c
> +++ b/accel/tcg/tcg-all.c
> @@ -38,6 +38,7 @@
>   #include "hw/qdev-core.h"
>   #else
>   #include "hw/boards.h"
> +#include "system/tcg.h"
>   #endif
>   #include "internal-common.h"
>   #include "cpu-param.h"
> @@ -58,6 +59,15 @@ typedef struct TCGState TCGState;
>   DECLARE_INSTANCE_CHECKER(TCGState, TCG_STATE,
>                            TYPE_TCG_ACCEL)
>   
> +static bool mttcg_enabled;
> +
> +#ifndef CONFIG_USER_ONLY
> +bool qemu_tcg_mttcg_enabled(void)
> +{
> +    return mttcg_enabled;
> +}
> +#endif
> +
>   /*
>    * We default to false if we know other options have been enabled
>    * which are currently incompatible with MTTCG. Otherwise when each
> @@ -97,7 +107,6 @@ static void tcg_accel_instance_init(Object *obj)
>   #endif
>   }
>   
> -bool mttcg_enabled;
>   bool one_insn_per_tb;
>   
>   static int tcg_init_machine(MachineState *ms)
> diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
> index 60a26acc503..5ccd7b9dc07 100644
> --- a/target/riscv/tcg/tcg-cpu.c
> +++ b/target/riscv/tcg/tcg-cpu.c
> @@ -35,6 +35,7 @@
>   #include "tcg/tcg.h"
>   #ifndef CONFIG_USER_ONLY
>   #include "hw/boards.h"
> +#include "system/tcg.h"
>   #endif
>   
>   /* Hash that stores user set extensions */
> diff --git a/tcg/region.c b/tcg/region.c
> index 478ec051c4b..56d2e988719 100644
> --- a/tcg/region.c
> +++ b/tcg/region.c
> @@ -34,7 +34,9 @@
>   #include "exec/translation-block.h"
>   #include "tcg-internal.h"
>   #include "host/cpuinfo.h"
> -
> +#ifndef CONFIG_USER_ONLY
> +#include "system/tcg.h"
> +#endif
>   
>   /*
>    * Local source-level compatibility with Unix.

Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>

Re: [PATCH-for-10.1 2/4] tcg: Move qemu_tcg_mttcg_enabled() to 'system/tcg.h'

[Richard Henderson]

On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> qemu_tcg_mttcg_enabled() is specific to 1/ TCG and
> 2/ system emulation. Move the prototype declaration
> to "system/tcg.h", reducing 'mttcg_enabled' variable
> scope.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   include/hw/core/cpu.h      |  9 ---------
>   include/system/tcg.h       |  8 ++++++++
>   accel/tcg/tcg-all.c        | 11 ++++++++++-
>   target/riscv/tcg/tcg-cpu.c |  1 +
>   tcg/region.c               |  4 +++-
>   5 files changed, 22 insertions(+), 11 deletions(-)
> 
> diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
> index 5d11d26556a..54570d21aea 100644
> --- a/include/hw/core/cpu.h
> +++ b/include/hw/core/cpu.h
> @@ -597,15 +597,6 @@ extern CPUTailQ cpus_queue;
>   
>   extern __thread CPUState *current_cpu;
>   
> -/**
> - * qemu_tcg_mttcg_enabled:
> - * Check whether we are running MultiThread TCG or not.
> - *
> - * Returns: %true if we are in MTTCG mode %false otherwise.
> - */
> -extern bool mttcg_enabled;
> -#define qemu_tcg_mttcg_enabled() (mttcg_enabled)
> -
>   /**
>    * cpu_paging_enabled:
>    * @cpu: The CPU whose state is to be inspected.
> diff --git a/include/system/tcg.h b/include/system/tcg.h
> index 73229648c63..7622dcea302 100644
> --- a/include/system/tcg.h
> +++ b/include/system/tcg.h
> @@ -17,4 +17,12 @@ extern bool tcg_allowed;
>   #define tcg_enabled() 0
>   #endif
>   
> +/**
> + * qemu_tcg_mttcg_enabled:
> + * Check whether we are running MultiThread TCG or not.
> + *
> + * Returns: %true if we are in MTTCG mode %false otherwise.
> + */
> +bool qemu_tcg_mttcg_enabled(void);
> +
>   #endif
> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
> index cb632cc8cc7..d75ecf531b6 100644
> --- a/accel/tcg/tcg-all.c
> +++ b/accel/tcg/tcg-all.c
> @@ -38,6 +38,7 @@
>   #include "hw/qdev-core.h"
>   #else
>   #include "hw/boards.h"
> +#include "system/tcg.h"
>   #endif
>   #include "internal-common.h"
>   #include "cpu-param.h"
> @@ -58,6 +59,15 @@ typedef struct TCGState TCGState;
>   DECLARE_INSTANCE_CHECKER(TCGState, TCG_STATE,
>                            TYPE_TCG_ACCEL)
>   
> +static bool mttcg_enabled;
> +
> +#ifndef CONFIG_USER_ONLY
> +bool qemu_tcg_mttcg_enabled(void)
> +{
> +    return mttcg_enabled;
> +}
> +#endif

It would appear (and it makes sense) that mttcg_enabled is only set for user-mode.  I 
think you should be conditionalizing the variable for system mode as well.


r~

[PATCH-for-10.1 3/4] tcg: Convert TCGState::mttcg_enabled to TriState

[Philippe Mathieu-Daudé]

Use the OnOffAuto type as 3-state.

Since the TCGState instance is zero-initialized, the
mttcg_enabled is initialzed as AUTO (ON_OFF_AUTO_AUTO).

In tcg_init_machine(), if mttcg_enabled is still AUTO,
set a default value (effectively inlining the
default_mttcg_enabled() method content).

Instead of emiting a warning when the 'thread' property
is set in tcg_set_thread(), emit it in tcg_init_machine()
where it is consumed.

In the tcg_get_thread() getter, consider AUTO / OFF states
as "single", otherwise ON is "multi".

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 accel/tcg/tcg-all.c | 68 ++++++++++++++++++++++-----------------------
 1 file changed, 33 insertions(+), 35 deletions(-)

diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index d75ecf531b6..2b7f89eaa20 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -32,6 +32,7 @@
 #include "qemu/error-report.h"
 #include "qemu/accel.h"
 #include "qemu/atomic.h"
+#include "qapi/qapi-types-common.h"
 #include "qapi/qapi-builtin-visit.h"
 #include "qemu/units.h"
 #if defined(CONFIG_USER_ONLY)
@@ -47,7 +48,7 @@
 struct TCGState {
     AccelState parent_obj;
 
-    bool mttcg_enabled;
+    OnOffAuto mttcg_enabled;
     bool one_insn_per_tb;
     int splitwx_enabled;
     unsigned long tb_size;
@@ -68,37 +69,10 @@ bool qemu_tcg_mttcg_enabled(void)
 }
 #endif
 
-/*
- * We default to false if we know other options have been enabled
- * which are currently incompatible with MTTCG. Otherwise when each
- * guest (target) has been updated to support:
- *   - atomic instructions
- *   - memory ordering primitives (barriers)
- * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
- *
- * Once a guest architecture has been converted to the new primitives
- * there is one remaining limitation to check:
- *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
- */
-
-static bool default_mttcg_enabled(void)
-{
-    if (icount_enabled()) {
-        return false;
-    }
-#ifdef TARGET_SUPPORTS_MTTCG
-    return true;
-#else
-    return false;
-#endif
-}
-
 static void tcg_accel_instance_init(Object *obj)
 {
     TCGState *s = TCG_STATE(obj);
 
-    s->mttcg_enabled = default_mttcg_enabled();
-
     /* If debugging enabled, default "auto on", otherwise off. */
 #if defined(CONFIG_DEBUG_TCG) && !defined(CONFIG_USER_ONLY)
     s->splitwx_enabled = -1;
@@ -117,9 +91,37 @@ static int tcg_init_machine(MachineState *ms)
 #else
     unsigned max_cpus = ms->smp.max_cpus;
 #endif
+#ifdef TARGET_SUPPORTS_MTTCG
+    bool mttcg_supported = true;
+#else
+    bool mttcg_supported = false;
+#endif
 
     tcg_allowed = true;
     mttcg_enabled = s->mttcg_enabled;
+    if (mttcg_enabled == ON_OFF_AUTO_AUTO) {
+        /*
+         * We default to false if we know other options have been enabled
+         * which are currently incompatible with MTTCG. Otherwise when each
+         * guest (target) has been updated to support:
+         *   - atomic instructions
+         *   - memory ordering primitives (barriers)
+         * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
+         *
+         * Once a guest architecture has been converted to the new primitives
+         * there is one remaining limitation to check:
+         *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
+         */
+        if (icount_enabled()) {
+            mttcg_enabled = ON_OFF_AUTO_OFF;
+        } else {
+            mttcg_enabled = mttcg_supported;
+        }
+    }
+    if (mttcg_enabled == ON_OFF_AUTO_ON && !mttcg_supported) {
+        warn_report("Guest not yet converted to MTTCG - "
+                    "you may get unexpected results");
+    }
 
     page_init();
     tb_htable_init();
@@ -144,7 +146,7 @@ static char *tcg_get_thread(Object *obj, Error **errp)
 {
     TCGState *s = TCG_STATE(obj);
 
-    return g_strdup(s->mttcg_enabled ? "multi" : "single");
+    return g_strdup(s->mttcg_enabled == ON_OFF_AUTO_ON ? "multi" : "single");
 }
 
 static void tcg_set_thread(Object *obj, const char *value, Error **errp)
@@ -155,14 +157,10 @@ static void tcg_set_thread(Object *obj, const char *value, Error **errp)
         if (icount_enabled()) {
             error_setg(errp, "No MTTCG when icount is enabled");
         } else {
-#ifndef TARGET_SUPPORTS_MTTCG
-            warn_report("Guest not yet converted to MTTCG - "
-                        "you may get unexpected results");
-#endif
-            s->mttcg_enabled = true;
+            s->mttcg_enabled = ON_OFF_AUTO_ON;
         }
     } else if (strcmp(value, "single") == 0) {
-        s->mttcg_enabled = false;
+        s->mttcg_enabled = ON_OFF_AUTO_OFF;
     } else {
         error_setg(errp, "Invalid 'thread' setting %s", value);
     }
-- 
2.47.1

Re: [PATCH-for-10.1 3/4] tcg: Convert TCGState::mttcg_enabled to TriState

[Anton Johansson via]

On 21/03/25, Philippe Mathieu-Daudé wrote:
> Use the OnOffAuto type as 3-state.
> 
> Since the TCGState instance is zero-initialized, the
> mttcg_enabled is initialzed as AUTO (ON_OFF_AUTO_AUTO).
> 
> In tcg_init_machine(), if mttcg_enabled is still AUTO,
> set a default value (effectively inlining the
> default_mttcg_enabled() method content).
> 
> Instead of emiting a warning when the 'thread' property
> is set in tcg_set_thread(), emit it in tcg_init_machine()
> where it is consumed.
> 
> In the tcg_get_thread() getter, consider AUTO / OFF states
> as "single", otherwise ON is "multi".
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>  accel/tcg/tcg-all.c | 68 ++++++++++++++++++++++-----------------------
>  1 file changed, 33 insertions(+), 35 deletions(-)
> 
> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
> index d75ecf531b6..2b7f89eaa20 100644
> --- a/accel/tcg/tcg-all.c
> +++ b/accel/tcg/tcg-all.c
> @@ -32,6 +32,7 @@
>  #include "qemu/error-report.h"
>  #include "qemu/accel.h"
>  #include "qemu/atomic.h"
> +#include "qapi/qapi-types-common.h"
>  #include "qapi/qapi-builtin-visit.h"
>  #include "qemu/units.h"
>  #if defined(CONFIG_USER_ONLY)
> @@ -47,7 +48,7 @@
>  struct TCGState {
>      AccelState parent_obj;
>  
> -    bool mttcg_enabled;
> +    OnOffAuto mttcg_enabled;
>      bool one_insn_per_tb;
>      int splitwx_enabled;
>      unsigned long tb_size;
> @@ -68,37 +69,10 @@ bool qemu_tcg_mttcg_enabled(void)
>  }
>  #endif
>  
> -/*
> - * We default to false if we know other options have been enabled
> - * which are currently incompatible with MTTCG. Otherwise when each
> - * guest (target) has been updated to support:
> - *   - atomic instructions
> - *   - memory ordering primitives (barriers)
> - * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
> - *
> - * Once a guest architecture has been converted to the new primitives
> - * there is one remaining limitation to check:
> - *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
> - */
> -
> -static bool default_mttcg_enabled(void)
> -{
> -    if (icount_enabled()) {
> -        return false;
> -    }
> -#ifdef TARGET_SUPPORTS_MTTCG
> -    return true;
> -#else
> -    return false;
> -#endif
> -}
> -
>  static void tcg_accel_instance_init(Object *obj)
>  {
>      TCGState *s = TCG_STATE(obj);
>  
> -    s->mttcg_enabled = default_mttcg_enabled();
> -
>      /* If debugging enabled, default "auto on", otherwise off. */
>  #if defined(CONFIG_DEBUG_TCG) && !defined(CONFIG_USER_ONLY)
>      s->splitwx_enabled = -1;
> @@ -117,9 +91,37 @@ static int tcg_init_machine(MachineState *ms)
>  #else
>      unsigned max_cpus = ms->smp.max_cpus;
>  #endif
> +#ifdef TARGET_SUPPORTS_MTTCG
> +    bool mttcg_supported = true;
> +#else
> +    bool mttcg_supported = false;
> +#endif
>  
>      tcg_allowed = true;
>      mttcg_enabled = s->mttcg_enabled;
> +    if (mttcg_enabled == ON_OFF_AUTO_AUTO) {
> +        /*
> +         * We default to false if we know other options have been enabled
> +         * which are currently incompatible with MTTCG. Otherwise when each
> +         * guest (target) has been updated to support:
> +         *   - atomic instructions
> +         *   - memory ordering primitives (barriers)
> +         * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
> +         *
> +         * Once a guest architecture has been converted to the new primitives
> +         * there is one remaining limitation to check:
> +         *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
> +         */
> +        if (icount_enabled()) {
> +            mttcg_enabled = ON_OFF_AUTO_OFF;
> +        } else {
> +            mttcg_enabled = mttcg_supported;
> +        }
> +    }
> +    if (mttcg_enabled == ON_OFF_AUTO_ON && !mttcg_supported) {
> +        warn_report("Guest not yet converted to MTTCG - "
> +                    "you may get unexpected results");
> +    }

[...]

>  
>  static void tcg_set_thread(Object *obj, const char *value, Error **errp)
> @@ -155,14 +157,10 @@ static void tcg_set_thread(Object *obj, const char *value, Error **errp)
>          if (icount_enabled()) {
>              error_setg(errp, "No MTTCG when icount is enabled");
>          } else {
> -#ifndef TARGET_SUPPORTS_MTTCG
> -            warn_report("Guest not yet converted to MTTCG - "
> -                        "you may get unexpected results");
> -#endif

Patch itself looks good!  My only concern is moving the warning, would
it be worthwhile to have a mttcg_supported field in TCGState as well?
I.e in a heterogeneous setup it would correspond to wether or not all
targets support mttcg.

Otherwise:

Reviewed-by: Anton Johansson <anjo@rev.ng>

Re: [PATCH-for-10.1 3/4] tcg: Convert TCGState::mttcg_enabled to TriState

[Philippe Mathieu-Daudé]

On 21/3/25 18:01, Anton Johansson wrote:
> On 21/03/25, Philippe Mathieu-Daudé wrote:
>> Use the OnOffAuto type as 3-state.
>>
>> Since the TCGState instance is zero-initialized, the
>> mttcg_enabled is initialzed as AUTO (ON_OFF_AUTO_AUTO).
>>
>> In tcg_init_machine(), if mttcg_enabled is still AUTO,
>> set a default value (effectively inlining the
>> default_mttcg_enabled() method content).
>>
>> Instead of emiting a warning when the 'thread' property
>> is set in tcg_set_thread(), emit it in tcg_init_machine()
>> where it is consumed.
>>
>> In the tcg_get_thread() getter, consider AUTO / OFF states
>> as "single", otherwise ON is "multi".
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   accel/tcg/tcg-all.c | 68 ++++++++++++++++++++++-----------------------
>>   1 file changed, 33 insertions(+), 35 deletions(-)
>>
>> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
>> index d75ecf531b6..2b7f89eaa20 100644
>> --- a/accel/tcg/tcg-all.c
>> +++ b/accel/tcg/tcg-all.c
>> @@ -32,6 +32,7 @@
>>   #include "qemu/error-report.h"
>>   #include "qemu/accel.h"
>>   #include "qemu/atomic.h"
>> +#include "qapi/qapi-types-common.h"
>>   #include "qapi/qapi-builtin-visit.h"
>>   #include "qemu/units.h"
>>   #if defined(CONFIG_USER_ONLY)
>> @@ -47,7 +48,7 @@
>>   struct TCGState {
>>       AccelState parent_obj;
>>   
>> -    bool mttcg_enabled;
>> +    OnOffAuto mttcg_enabled;
>>       bool one_insn_per_tb;
>>       int splitwx_enabled;
>>       unsigned long tb_size;


>> -/*
>> - * We default to false if we know other options have been enabled
>> - * which are currently incompatible with MTTCG. Otherwise when each
>> - * guest (target) has been updated to support:
>> - *   - atomic instructions
>> - *   - memory ordering primitives (barriers)
>> - * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
>> - *
>> - * Once a guest architecture has been converted to the new primitives
>> - * there is one remaining limitation to check:
>> - *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
>> - */
>> -
>> -static bool default_mttcg_enabled(void)
>> -{
>> -    if (icount_enabled()) {
>> -        return false;
>> -    }
>> -#ifdef TARGET_SUPPORTS_MTTCG
>> -    return true;
>> -#else
>> -    return false;
>> -#endif
>> -}
>> -
>>   static void tcg_accel_instance_init(Object *obj)
>>   {
>>       TCGState *s = TCG_STATE(obj);
>>   
>> -    s->mttcg_enabled = default_mttcg_enabled();
>> -
>>       /* If debugging enabled, default "auto on", otherwise off. */
>>   #if defined(CONFIG_DEBUG_TCG) && !defined(CONFIG_USER_ONLY)
>>       s->splitwx_enabled = -1;
>> @@ -117,9 +91,37 @@ static int tcg_init_machine(MachineState *ms)
>>   #else
>>       unsigned max_cpus = ms->smp.max_cpus;
>>   #endif
>> +#ifdef TARGET_SUPPORTS_MTTCG
>> +    bool mttcg_supported = true;
>> +#else
>> +    bool mttcg_supported = false;
>> +#endif
>>   
>>       tcg_allowed = true;
>>       mttcg_enabled = s->mttcg_enabled;
>> +    if (mttcg_enabled == ON_OFF_AUTO_AUTO) {
>> +        /*
>> +         * We default to false if we know other options have been enabled
>> +         * which are currently incompatible with MTTCG. Otherwise when each
>> +         * guest (target) has been updated to support:
>> +         *   - atomic instructions
>> +         *   - memory ordering primitives (barriers)
>> +         * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
>> +         *
>> +         * Once a guest architecture has been converted to the new primitives
>> +         * there is one remaining limitation to check:
>> +         *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
>> +         */
>> +        if (icount_enabled()) {
>> +            mttcg_enabled = ON_OFF_AUTO_OFF;
>> +        } else {
>> +            mttcg_enabled = mttcg_supported;
>> +        }
>> +    }
>> +    if (mttcg_enabled == ON_OFF_AUTO_ON && !mttcg_supported) {
>> +        warn_report("Guest not yet converted to MTTCG - "
>> +                    "you may get unexpected results");
>> +    }
> 
> [...]
> 
>>   
>>   static void tcg_set_thread(Object *obj, const char *value, Error **errp)
>> @@ -155,14 +157,10 @@ static void tcg_set_thread(Object *obj, const char *value, Error **errp)
>>           if (icount_enabled()) {
>>               error_setg(errp, "No MTTCG when icount is enabled");
>>           } else {
>> -#ifndef TARGET_SUPPORTS_MTTCG
>> -            warn_report("Guest not yet converted to MTTCG - "
>> -                        "you may get unexpected results");
>> -#endif
> 
> Patch itself looks good!  My only concern is moving the warning, would
> it be worthwhile to have a mttcg_supported field in TCGState as well?
> I.e in a heterogeneous setup it would correspond to wether or not all
> targets support mttcg.

OK, I'll update s->mttcg_enabled then.

> 
> Otherwise:
> 
> Reviewed-by: Anton Johansson <anjo@rev.ng>

Thanks!

Re: [PATCH-for-10.1 3/4] tcg: Convert TCGState::mttcg_enabled to TriState

[Pierrick Bouvier]

On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> Use the OnOffAuto type as 3-state.
> 
> Since the TCGState instance is zero-initialized, the
> mttcg_enabled is initialzed as AUTO (ON_OFF_AUTO_AUTO).
> 
> In tcg_init_machine(), if mttcg_enabled is still AUTO,
> set a default value (effectively inlining the
> default_mttcg_enabled() method content).
> 
> Instead of emiting a warning when the 'thread' property
> is set in tcg_set_thread(), emit it in tcg_init_machine()
> where it is consumed.
> 
> In the tcg_get_thread() getter, consider AUTO / OFF states
> as "single", otherwise ON is "multi".
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   accel/tcg/tcg-all.c | 68 ++++++++++++++++++++++-----------------------
>   1 file changed, 33 insertions(+), 35 deletions(-)
> 
> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
> index d75ecf531b6..2b7f89eaa20 100644
> --- a/accel/tcg/tcg-all.c
> +++ b/accel/tcg/tcg-all.c
> @@ -32,6 +32,7 @@
>   #include "qemu/error-report.h"
>   #include "qemu/accel.h"
>   #include "qemu/atomic.h"
> +#include "qapi/qapi-types-common.h"
>   #include "qapi/qapi-builtin-visit.h"
>   #include "qemu/units.h"
>   #if defined(CONFIG_USER_ONLY)
> @@ -47,7 +48,7 @@
>   struct TCGState {
>       AccelState parent_obj;
>   
> -    bool mttcg_enabled;
> +    OnOffAuto mttcg_enabled;
>       bool one_insn_per_tb;
>       int splitwx_enabled;
>       unsigned long tb_size;
> @@ -68,37 +69,10 @@ bool qemu_tcg_mttcg_enabled(void)
>   }
>   #endif
>   
> -/*
> - * We default to false if we know other options have been enabled
> - * which are currently incompatible with MTTCG. Otherwise when each
> - * guest (target) has been updated to support:
> - *   - atomic instructions
> - *   - memory ordering primitives (barriers)
> - * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
> - *
> - * Once a guest architecture has been converted to the new primitives
> - * there is one remaining limitation to check:
> - *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
> - */
> -
> -static bool default_mttcg_enabled(void)
> -{
> -    if (icount_enabled()) {
> -        return false;
> -    }
> -#ifdef TARGET_SUPPORTS_MTTCG
> -    return true;
> -#else
> -    return false;
> -#endif
> -}
> -
>   static void tcg_accel_instance_init(Object *obj)
>   {
>       TCGState *s = TCG_STATE(obj);
>   
> -    s->mttcg_enabled = default_mttcg_enabled();
> -
>       /* If debugging enabled, default "auto on", otherwise off. */
>   #if defined(CONFIG_DEBUG_TCG) && !defined(CONFIG_USER_ONLY)
>       s->splitwx_enabled = -1;
> @@ -117,9 +91,37 @@ static int tcg_init_machine(MachineState *ms)
>   #else
>       unsigned max_cpus = ms->smp.max_cpus;
>   #endif
> +#ifdef TARGET_SUPPORTS_MTTCG
> +    bool mttcg_supported = true;
> +#else
> +    bool mttcg_supported = false;
> +#endif
>   
>       tcg_allowed = true;
>       mttcg_enabled = s->mttcg_enabled;
> +    if (mttcg_enabled == ON_OFF_AUTO_AUTO) {
> +        /*
> +         * We default to false if we know other options have been enabled
> +         * which are currently incompatible with MTTCG. Otherwise when each
> +         * guest (target) has been updated to support:
> +         *   - atomic instructions
> +         *   - memory ordering primitives (barriers)
> +         * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
> +         *
> +         * Once a guest architecture has been converted to the new primitives
> +         * there is one remaining limitation to check:
> +         *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
> +         */
> +        if (icount_enabled()) {
> +            mttcg_enabled = ON_OFF_AUTO_OFF;
> +        } else {
> +            mttcg_enabled = mttcg_supported;
> +        }
> +    }
> +    if (mttcg_enabled == ON_OFF_AUTO_ON && !mttcg_supported) {
> +        warn_report("Guest not yet converted to MTTCG - "
> +                    "you may get unexpected results");
> +    }
>   
>       page_init();
>       tb_htable_init();
> @@ -144,7 +146,7 @@ static char *tcg_get_thread(Object *obj, Error **errp)
>   {
>       TCGState *s = TCG_STATE(obj);
>   
> -    return g_strdup(s->mttcg_enabled ? "multi" : "single");
> +    return g_strdup(s->mttcg_enabled == ON_OFF_AUTO_ON ? "multi" : "single");
>   }
>   
>   static void tcg_set_thread(Object *obj, const char *value, Error **errp)
> @@ -155,14 +157,10 @@ static void tcg_set_thread(Object *obj, const char *value, Error **errp)
>           if (icount_enabled()) {
>               error_setg(errp, "No MTTCG when icount is enabled");
>           } else {
> -#ifndef TARGET_SUPPORTS_MTTCG
> -            warn_report("Guest not yet converted to MTTCG - "
> -                        "you may get unexpected results");
> -#endif
> -            s->mttcg_enabled = true;
> +            s->mttcg_enabled = ON_OFF_AUTO_ON;
>           }
>       } else if (strcmp(value, "single") == 0) {
> -        s->mttcg_enabled = false;
> +        s->mttcg_enabled = ON_OFF_AUTO_OFF;
>       } else {
>           error_setg(errp, "Invalid 'thread' setting %s", value);
>       }

What are we gaining by moving this warning?
Are there cases where it was not reported before?

Re: [PATCH-for-10.1 3/4] tcg: Convert TCGState::mttcg_enabled to TriState

[Philippe Mathieu-Daudé]

On 21/3/25 18:36, Pierrick Bouvier wrote:
> On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
>> Use the OnOffAuto type as 3-state.
>>
>> Since the TCGState instance is zero-initialized, the
>> mttcg_enabled is initialzed as AUTO (ON_OFF_AUTO_AUTO).
>>
>> In tcg_init_machine(), if mttcg_enabled is still AUTO,
>> set a default value (effectively inlining the
>> default_mttcg_enabled() method content).
>>
>> Instead of emiting a warning when the 'thread' property
>> is set in tcg_set_thread(), emit it in tcg_init_machine()
>> where it is consumed.
>>
>> In the tcg_get_thread() getter, consider AUTO / OFF states
>> as "single", otherwise ON is "multi".
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   accel/tcg/tcg-all.c | 68 ++++++++++++++++++++++-----------------------
>>   1 file changed, 33 insertions(+), 35 deletions(-)
>>
>> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
>> index d75ecf531b6..2b7f89eaa20 100644
>> --- a/accel/tcg/tcg-all.c
>> +++ b/accel/tcg/tcg-all.c
>> @@ -32,6 +32,7 @@
>>   #include "qemu/error-report.h"
>>   #include "qemu/accel.h"
>>   #include "qemu/atomic.h"
>> +#include "qapi/qapi-types-common.h"
>>   #include "qapi/qapi-builtin-visit.h"
>>   #include "qemu/units.h"
>>   #if defined(CONFIG_USER_ONLY)
>> @@ -47,7 +48,7 @@
>>   struct TCGState {
>>       AccelState parent_obj;
>> -    bool mttcg_enabled;
>> +    OnOffAuto mttcg_enabled;
>>       bool one_insn_per_tb;
>>       int splitwx_enabled;
>>       unsigned long tb_size;
>> @@ -68,37 +69,10 @@ bool qemu_tcg_mttcg_enabled(void)
>>   }
>>   #endif
>> -/*
>> - * We default to false if we know other options have been enabled
>> - * which are currently incompatible with MTTCG. Otherwise when each
>> - * guest (target) has been updated to support:
>> - *   - atomic instructions
>> - *   - memory ordering primitives (barriers)
>> - * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
>> - *
>> - * Once a guest architecture has been converted to the new primitives
>> - * there is one remaining limitation to check:
>> - *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
>> - */
>> -
>> -static bool default_mttcg_enabled(void)
>> -{
>> -    if (icount_enabled()) {
>> -        return false;
>> -    }
>> -#ifdef TARGET_SUPPORTS_MTTCG
>> -    return true;
>> -#else
>> -    return false;
>> -#endif
>> -}
>> -
>>   static void tcg_accel_instance_init(Object *obj)
>>   {
>>       TCGState *s = TCG_STATE(obj);
>> -    s->mttcg_enabled = default_mttcg_enabled();
>> -
>>       /* If debugging enabled, default "auto on", otherwise off. */
>>   #if defined(CONFIG_DEBUG_TCG) && !defined(CONFIG_USER_ONLY)
>>       s->splitwx_enabled = -1;
>> @@ -117,9 +91,37 @@ static int tcg_init_machine(MachineState *ms)
>>   #else
>>       unsigned max_cpus = ms->smp.max_cpus;
>>   #endif
>> +#ifdef TARGET_SUPPORTS_MTTCG
>> +    bool mttcg_supported = true;
>> +#else
>> +    bool mttcg_supported = false;
>> +#endif
>>       tcg_allowed = true;
>>       mttcg_enabled = s->mttcg_enabled;
>> +    if (mttcg_enabled == ON_OFF_AUTO_AUTO) {
>> +        /*
>> +         * We default to false if we know other options have been 
>> enabled
>> +         * which are currently incompatible with MTTCG. Otherwise 
>> when each
>> +         * guest (target) has been updated to support:
>> +         *   - atomic instructions
>> +         *   - memory ordering primitives (barriers)
>> +         * they can set the appropriate CONFIG flags in ${target}- 
>> softmmu.mak
>> +         *
>> +         * Once a guest architecture has been converted to the new 
>> primitives
>> +         * there is one remaining limitation to check:
>> +         *   - The guest can't be oversized (e.g. 64 bit guest on 32 
>> bit host)
>> +         */
>> +        if (icount_enabled()) {
>> +            mttcg_enabled = ON_OFF_AUTO_OFF;
>> +        } else {
>> +            mttcg_enabled = mttcg_supported;
>> +        }
>> +    }
>> +    if (mttcg_enabled == ON_OFF_AUTO_ON && !mttcg_supported) {
>> +        warn_report("Guest not yet converted to MTTCG - "
>> +                    "you may get unexpected results");
>> +    }
>>       page_init();
>>       tb_htable_init();
>> @@ -144,7 +146,7 @@ static char *tcg_get_thread(Object *obj, Error 
>> **errp)
>>   {
>>       TCGState *s = TCG_STATE(obj);
>> -    return g_strdup(s->mttcg_enabled ? "multi" : "single");
>> +    return g_strdup(s->mttcg_enabled == ON_OFF_AUTO_ON ? "multi" : 
>> "single");
>>   }
>>   static void tcg_set_thread(Object *obj, const char *value, Error 
>> **errp)
>> @@ -155,14 +157,10 @@ static void tcg_set_thread(Object *obj, const 
>> char *value, Error **errp)
>>           if (icount_enabled()) {
>>               error_setg(errp, "No MTTCG when icount is enabled");
>>           } else {
>> -#ifndef TARGET_SUPPORTS_MTTCG
>> -            warn_report("Guest not yet converted to MTTCG - "
>> -                        "you may get unexpected results");
>> -#endif
>> -            s->mttcg_enabled = true;
>> +            s->mttcg_enabled = ON_OFF_AUTO_ON;
>>           }
>>       } else if (strcmp(value, "single") == 0) {
>> -        s->mttcg_enabled = false;
>> +        s->mttcg_enabled = ON_OFF_AUTO_OFF;
>>       } else {
>>           error_setg(errp, "Invalid 'thread' setting %s", value);
>>       }
> 
> What are we gaining by moving this warning?

Before: default_mttcg_enabled() was directly using TARGET_SUPPORTS_MTTCG 
definition, so we could call it to initialize a default value in
tcg_accel_instance_init().

After: considering the next patch where we need a CPUState initialized
to get TCGCPUOps::mttcg_supported, we can not use instance_init(), so
we defer to tcg_init_machine() where we access &first_cpu.

> Are there cases where it was not reported before?

No idea, I'll let Alex answer that.

Re: [PATCH-for-10.1 3/4] tcg: Convert TCGState::mttcg_enabled to TriState

[Pierrick Bouvier]

On 3/21/25 11:26, Philippe Mathieu-Daudé wrote:
> On 21/3/25 18:36, Pierrick Bouvier wrote:
>> On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
>>> Use the OnOffAuto type as 3-state.
>>>
>>> Since the TCGState instance is zero-initialized, the
>>> mttcg_enabled is initialzed as AUTO (ON_OFF_AUTO_AUTO).
>>>
>>> In tcg_init_machine(), if mttcg_enabled is still AUTO,
>>> set a default value (effectively inlining the
>>> default_mttcg_enabled() method content).
>>>
>>> Instead of emiting a warning when the 'thread' property
>>> is set in tcg_set_thread(), emit it in tcg_init_machine()
>>> where it is consumed.
>>>
>>> In the tcg_get_thread() getter, consider AUTO / OFF states
>>> as "single", otherwise ON is "multi".
>>>
>>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>>> ---
>>>    accel/tcg/tcg-all.c | 68 ++++++++++++++++++++++-----------------------
>>>    1 file changed, 33 insertions(+), 35 deletions(-)
>>>
>>> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
>>> index d75ecf531b6..2b7f89eaa20 100644
>>> --- a/accel/tcg/tcg-all.c
>>> +++ b/accel/tcg/tcg-all.c
>>> @@ -32,6 +32,7 @@
>>>    #include "qemu/error-report.h"
>>>    #include "qemu/accel.h"
>>>    #include "qemu/atomic.h"
>>> +#include "qapi/qapi-types-common.h"
>>>    #include "qapi/qapi-builtin-visit.h"
>>>    #include "qemu/units.h"
>>>    #if defined(CONFIG_USER_ONLY)
>>> @@ -47,7 +48,7 @@
>>>    struct TCGState {
>>>        AccelState parent_obj;
>>> -    bool mttcg_enabled;
>>> +    OnOffAuto mttcg_enabled;
>>>        bool one_insn_per_tb;
>>>        int splitwx_enabled;
>>>        unsigned long tb_size;
>>> @@ -68,37 +69,10 @@ bool qemu_tcg_mttcg_enabled(void)
>>>    }
>>>    #endif
>>> -/*
>>> - * We default to false if we know other options have been enabled
>>> - * which are currently incompatible with MTTCG. Otherwise when each
>>> - * guest (target) has been updated to support:
>>> - *   - atomic instructions
>>> - *   - memory ordering primitives (barriers)
>>> - * they can set the appropriate CONFIG flags in ${target}-softmmu.mak
>>> - *
>>> - * Once a guest architecture has been converted to the new primitives
>>> - * there is one remaining limitation to check:
>>> - *   - The guest can't be oversized (e.g. 64 bit guest on 32 bit host)
>>> - */
>>> -
>>> -static bool default_mttcg_enabled(void)
>>> -{
>>> -    if (icount_enabled()) {
>>> -        return false;
>>> -    }
>>> -#ifdef TARGET_SUPPORTS_MTTCG
>>> -    return true;
>>> -#else
>>> -    return false;
>>> -#endif
>>> -}
>>> -
>>>    static void tcg_accel_instance_init(Object *obj)
>>>    {
>>>        TCGState *s = TCG_STATE(obj);
>>> -    s->mttcg_enabled = default_mttcg_enabled();
>>> -
>>>        /* If debugging enabled, default "auto on", otherwise off. */
>>>    #if defined(CONFIG_DEBUG_TCG) && !defined(CONFIG_USER_ONLY)
>>>        s->splitwx_enabled = -1;
>>> @@ -117,9 +91,37 @@ static int tcg_init_machine(MachineState *ms)
>>>    #else
>>>        unsigned max_cpus = ms->smp.max_cpus;
>>>    #endif
>>> +#ifdef TARGET_SUPPORTS_MTTCG
>>> +    bool mttcg_supported = true;
>>> +#else
>>> +    bool mttcg_supported = false;
>>> +#endif
>>>        tcg_allowed = true;
>>>        mttcg_enabled = s->mttcg_enabled;
>>> +    if (mttcg_enabled == ON_OFF_AUTO_AUTO) {
>>> +        /*
>>> +         * We default to false if we know other options have been
>>> enabled
>>> +         * which are currently incompatible with MTTCG. Otherwise
>>> when each
>>> +         * guest (target) has been updated to support:
>>> +         *   - atomic instructions
>>> +         *   - memory ordering primitives (barriers)
>>> +         * they can set the appropriate CONFIG flags in ${target}-
>>> softmmu.mak
>>> +         *
>>> +         * Once a guest architecture has been converted to the new
>>> primitives
>>> +         * there is one remaining limitation to check:
>>> +         *   - The guest can't be oversized (e.g. 64 bit guest on 32
>>> bit host)
>>> +         */
>>> +        if (icount_enabled()) {
>>> +            mttcg_enabled = ON_OFF_AUTO_OFF;
>>> +        } else {
>>> +            mttcg_enabled = mttcg_supported;
>>> +        }
>>> +    }
>>> +    if (mttcg_enabled == ON_OFF_AUTO_ON && !mttcg_supported) {
>>> +        warn_report("Guest not yet converted to MTTCG - "
>>> +                    "you may get unexpected results");
>>> +    }
>>>        page_init();
>>>        tb_htable_init();
>>> @@ -144,7 +146,7 @@ static char *tcg_get_thread(Object *obj, Error
>>> **errp)
>>>    {
>>>        TCGState *s = TCG_STATE(obj);
>>> -    return g_strdup(s->mttcg_enabled ? "multi" : "single");
>>> +    return g_strdup(s->mttcg_enabled == ON_OFF_AUTO_ON ? "multi" :
>>> "single");
>>>    }
>>>    static void tcg_set_thread(Object *obj, const char *value, Error
>>> **errp)
>>> @@ -155,14 +157,10 @@ static void tcg_set_thread(Object *obj, const
>>> char *value, Error **errp)
>>>            if (icount_enabled()) {
>>>                error_setg(errp, "No MTTCG when icount is enabled");
>>>            } else {
>>> -#ifndef TARGET_SUPPORTS_MTTCG
>>> -            warn_report("Guest not yet converted to MTTCG - "
>>> -                        "you may get unexpected results");
>>> -#endif
>>> -            s->mttcg_enabled = true;
>>> +            s->mttcg_enabled = ON_OFF_AUTO_ON;
>>>            }
>>>        } else if (strcmp(value, "single") == 0) {
>>> -        s->mttcg_enabled = false;
>>> +        s->mttcg_enabled = ON_OFF_AUTO_OFF;
>>>        } else {
>>>            error_setg(errp, "Invalid 'thread' setting %s", value);
>>>        }
>>
>> What are we gaining by moving this warning?
> 
> Before: default_mttcg_enabled() was directly using TARGET_SUPPORTS_MTTCG
> definition, so we could call it to initialize a default value in
> tcg_accel_instance_init().
> 
> After: considering the next patch where we need a CPUState initialized
> to get TCGCPUOps::mttcg_supported, we can not use instance_init(), so
> we defer to tcg_init_machine() where we access &first_cpu.
> 

Ok, makes sense.
It might be worth to mention in the commit message, as it seems 
initially this change is more a "personal" preference, than something 
needed.

>> Are there cases where it was not reported before?
> 
> No idea, I'll let Alex answer that.

Re: [PATCH-for-10.1 3/4] tcg: Convert TCGState::mttcg_enabled to TriState

[Richard Henderson]

On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> Use the OnOffAuto type as 3-state.
> 
> Since the TCGState instance is zero-initialized, the
> mttcg_enabled is initialzed as AUTO (ON_OFF_AUTO_AUTO).
> 
> In tcg_init_machine(), if mttcg_enabled is still AUTO,
> set a default value (effectively inlining the
> default_mttcg_enabled() method content).
> 
> Instead of emiting a warning when the 'thread' property
> is set in tcg_set_thread(), emit it in tcg_init_machine()
> where it is consumed.
> 
> In the tcg_get_thread() getter, consider AUTO / OFF states
> as "single", otherwise ON is "multi".
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

I think it would be better to leave the warning where it is for this patch.
Move the warning when required by other changes.

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


r~

Re: [PATCH-for-10.1 3/4] tcg: Convert TCGState::mttcg_enabled to TriState

[Philippe Mathieu-Daudé]

On 23/3/25 19:43, Richard Henderson wrote:
> On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
>> Use the OnOffAuto type as 3-state.
>>
>> Since the TCGState instance is zero-initialized, the
>> mttcg_enabled is initialzed as AUTO (ON_OFF_AUTO_AUTO).
>>
>> In tcg_init_machine(), if mttcg_enabled is still AUTO,
>> set a default value (effectively inlining the
>> default_mttcg_enabled() method content).
>>
>> Instead of emiting a warning when the 'thread' property
>> is set in tcg_set_thread(), emit it in tcg_init_machine()
>> where it is consumed.
>>
>> In the tcg_get_thread() getter, consider AUTO / OFF states
>> as "single", otherwise ON is "multi".
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> 
> I think it would be better to leave the warning where it is for this patch.
> Move the warning when required by other changes.

Doh, I forgot to address this comment. Do you mean, defer moving the
warning to the next patch, where we remove the TARGET_SUPPORTS_MTTCG
definition?

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

[PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Philippe Mathieu-Daudé]

Instead of having a compile-time TARGET_SUPPORTS_MTTCG definition,
have each target set the 'mttcg_supported' field in the TCGCPUOps
structure.

Since so far we only emulate one target architecture at a time,
tcg_init_machine() gets whether MTTCG is supported via the
&first_cpu global.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 docs/devel/multi-thread-tcg.rst          | 2 +-
 configs/targets/aarch64-softmmu.mak      | 1 -
 configs/targets/alpha-softmmu.mak        | 1 -
 configs/targets/arm-softmmu.mak          | 1 -
 configs/targets/hppa-softmmu.mak         | 1 -
 configs/targets/i386-softmmu.mak         | 1 -
 configs/targets/loongarch64-softmmu.mak  | 1 -
 configs/targets/microblaze-softmmu.mak   | 1 -
 configs/targets/microblazeel-softmmu.mak | 1 -
 configs/targets/mips-softmmu.mak         | 1 -
 configs/targets/mipsel-softmmu.mak       | 1 -
 configs/targets/or1k-softmmu.mak         | 1 -
 configs/targets/ppc64-softmmu.mak        | 1 -
 configs/targets/riscv32-softmmu.mak      | 1 -
 configs/targets/riscv64-softmmu.mak      | 1 -
 configs/targets/s390x-softmmu.mak        | 1 -
 configs/targets/sparc-softmmu.mak        | 1 -
 configs/targets/sparc64-softmmu.mak      | 1 -
 configs/targets/x86_64-softmmu.mak       | 1 -
 configs/targets/xtensa-softmmu.mak       | 1 -
 configs/targets/xtensaeb-softmmu.mak     | 1 -
 include/accel/tcg/cpu-ops.h              | 8 ++++++++
 include/exec/poison.h                    | 1 -
 accel/tcg/tcg-all.c                      | 7 ++-----
 target/alpha/cpu.c                       | 1 +
 target/arm/cpu.c                         | 1 +
 target/arm/tcg/cpu-v7m.c                 | 1 +
 target/avr/cpu.c                         | 1 +
 target/hexagon/cpu.c                     | 1 +
 target/hppa/cpu.c                        | 1 +
 target/i386/tcg/tcg-cpu.c                | 1 +
 target/loongarch/cpu.c                   | 1 +
 target/m68k/cpu.c                        | 1 +
 target/microblaze/cpu.c                  | 1 +
 target/mips/cpu.c                        | 1 +
 target/openrisc/cpu.c                    | 1 +
 target/ppc/cpu_init.c                    | 1 +
 target/riscv/tcg/tcg-cpu.c               | 1 +
 target/rx/cpu.c                          | 1 +
 target/s390x/cpu.c                       | 1 +
 target/sh4/cpu.c                         | 1 +
 target/sparc/cpu.c                       | 1 +
 target/tricore/cpu.c                     | 1 +
 target/xtensa/cpu.c                      | 1 +
 44 files changed, 31 insertions(+), 27 deletions(-)

diff --git a/docs/devel/multi-thread-tcg.rst b/docs/devel/multi-thread-tcg.rst
index 14a2a9dc7b5..da9a1530c9f 100644
--- a/docs/devel/multi-thread-tcg.rst
+++ b/docs/devel/multi-thread-tcg.rst
@@ -30,7 +30,7 @@ user-space thread. This is enabled by default for all FE/BE
 combinations where the host memory model is able to accommodate the
 guest (TCGCPUOps::guest_default_memory_order & ~TCG_TARGET_DEFAULT_MO is zero)
 and the guest has had the required work done to support this safely
-(TARGET_SUPPORTS_MTTCG).
+(TCGCPUOps::mttcg_supported).
 
 System emulation will fall back to the original round robin approach
 if:
diff --git a/configs/targets/aarch64-softmmu.mak b/configs/targets/aarch64-softmmu.mak
index 82cb72cb83d..5dfeb35af90 100644
--- a/configs/targets/aarch64-softmmu.mak
+++ b/configs/targets/aarch64-softmmu.mak
@@ -1,6 +1,5 @@
 TARGET_ARCH=aarch64
 TARGET_BASE_ARCH=arm
-TARGET_SUPPORTS_MTTCG=y
 TARGET_KVM_HAVE_GUEST_DEBUG=y
 TARGET_XML_FILES= gdb-xml/aarch64-core.xml gdb-xml/aarch64-fpu.xml gdb-xml/arm-core.xml gdb-xml/arm-vfp.xml gdb-xml/arm-vfp3.xml gdb-xml/arm-vfp-sysregs.xml gdb-xml/arm-neon.xml gdb-xml/arm-m-profile.xml gdb-xml/arm-m-profile-mve.xml gdb-xml/aarch64-pauth.xml
 # needed by boot.c
diff --git a/configs/targets/alpha-softmmu.mak b/configs/targets/alpha-softmmu.mak
index 89f3517aca0..5275076e50d 100644
--- a/configs/targets/alpha-softmmu.mak
+++ b/configs/targets/alpha-softmmu.mak
@@ -1,3 +1,2 @@
 TARGET_ARCH=alpha
-TARGET_SUPPORTS_MTTCG=y
 TARGET_LONG_BITS=64
diff --git a/configs/targets/arm-softmmu.mak b/configs/targets/arm-softmmu.mak
index afc64f5927b..6a5a8eda949 100644
--- a/configs/targets/arm-softmmu.mak
+++ b/configs/targets/arm-softmmu.mak
@@ -1,5 +1,4 @@
 TARGET_ARCH=arm
-TARGET_SUPPORTS_MTTCG=y
 TARGET_XML_FILES= gdb-xml/arm-core.xml gdb-xml/arm-vfp.xml gdb-xml/arm-vfp3.xml gdb-xml/arm-vfp-sysregs.xml gdb-xml/arm-neon.xml gdb-xml/arm-m-profile.xml gdb-xml/arm-m-profile-mve.xml
 # needed by boot.c
 TARGET_NEED_FDT=y
diff --git a/configs/targets/hppa-softmmu.mak b/configs/targets/hppa-softmmu.mak
index 63ca74ed5e6..ea331107a08 100644
--- a/configs/targets/hppa-softmmu.mak
+++ b/configs/targets/hppa-softmmu.mak
@@ -1,4 +1,3 @@
 TARGET_ARCH=hppa
 TARGET_BIG_ENDIAN=y
-TARGET_SUPPORTS_MTTCG=y
 TARGET_LONG_BITS=64
diff --git a/configs/targets/i386-softmmu.mak b/configs/targets/i386-softmmu.mak
index 5dd89217560..e9d89e8ab41 100644
--- a/configs/targets/i386-softmmu.mak
+++ b/configs/targets/i386-softmmu.mak
@@ -1,5 +1,4 @@
 TARGET_ARCH=i386
-TARGET_SUPPORTS_MTTCG=y
 TARGET_KVM_HAVE_GUEST_DEBUG=y
 TARGET_KVM_HAVE_RESET_PARKED_VCPU=y
 TARGET_XML_FILES= gdb-xml/i386-32bit.xml
diff --git a/configs/targets/loongarch64-softmmu.mak b/configs/targets/loongarch64-softmmu.mak
index 351341132f6..fc44c54233d 100644
--- a/configs/targets/loongarch64-softmmu.mak
+++ b/configs/targets/loongarch64-softmmu.mak
@@ -1,7 +1,6 @@
 TARGET_ARCH=loongarch64
 TARGET_BASE_ARCH=loongarch
 TARGET_KVM_HAVE_GUEST_DEBUG=y
-TARGET_SUPPORTS_MTTCG=y
 TARGET_XML_FILES= gdb-xml/loongarch-base32.xml gdb-xml/loongarch-base64.xml gdb-xml/loongarch-fpu.xml gdb-xml/loongarch-lsx.xml gdb-xml/loongarch-lasx.xml
 # all boards require libfdt
 TARGET_NEED_FDT=y
diff --git a/configs/targets/microblaze-softmmu.mak b/configs/targets/microblaze-softmmu.mak
index 99a33ed44a8..23457d0ae65 100644
--- a/configs/targets/microblaze-softmmu.mak
+++ b/configs/targets/microblaze-softmmu.mak
@@ -1,6 +1,5 @@
 TARGET_ARCH=microblaze
 TARGET_BIG_ENDIAN=y
-TARGET_SUPPORTS_MTTCG=y
 # needed by boot.c
 TARGET_NEED_FDT=y
 TARGET_XML_FILES=gdb-xml/microblaze-core.xml gdb-xml/microblaze-stack-protect.xml
diff --git a/configs/targets/microblazeel-softmmu.mak b/configs/targets/microblazeel-softmmu.mak
index 52cdeae1a28..c82c509623d 100644
--- a/configs/targets/microblazeel-softmmu.mak
+++ b/configs/targets/microblazeel-softmmu.mak
@@ -1,5 +1,4 @@
 TARGET_ARCH=microblaze
-TARGET_SUPPORTS_MTTCG=y
 # needed by boot.c
 TARGET_NEED_FDT=y
 TARGET_XML_FILES=gdb-xml/microblaze-core.xml gdb-xml/microblaze-stack-protect.xml
diff --git a/configs/targets/mips-softmmu.mak b/configs/targets/mips-softmmu.mak
index b62a0882499..c9588066b8d 100644
--- a/configs/targets/mips-softmmu.mak
+++ b/configs/targets/mips-softmmu.mak
@@ -1,4 +1,3 @@
 TARGET_ARCH=mips
 TARGET_BIG_ENDIAN=y
-TARGET_SUPPORTS_MTTCG=y
 TARGET_LONG_BITS=32
diff --git a/configs/targets/mipsel-softmmu.mak b/configs/targets/mipsel-softmmu.mak
index 620ec681785..90e09bdc3e5 100644
--- a/configs/targets/mipsel-softmmu.mak
+++ b/configs/targets/mipsel-softmmu.mak
@@ -1,3 +1,2 @@
 TARGET_ARCH=mips
-TARGET_SUPPORTS_MTTCG=y
 TARGET_LONG_BITS=32
diff --git a/configs/targets/or1k-softmmu.mak b/configs/targets/or1k-softmmu.mak
index adfddb1a8ac..0e47d9878b0 100644
--- a/configs/targets/or1k-softmmu.mak
+++ b/configs/targets/or1k-softmmu.mak
@@ -1,5 +1,4 @@
 TARGET_ARCH=openrisc
-TARGET_SUPPORTS_MTTCG=y
 TARGET_BIG_ENDIAN=y
 # needed by boot.c and all boards
 TARGET_NEED_FDT=y
diff --git a/configs/targets/ppc64-softmmu.mak b/configs/targets/ppc64-softmmu.mak
index 7cee0e97f43..74572864b36 100644
--- a/configs/targets/ppc64-softmmu.mak
+++ b/configs/targets/ppc64-softmmu.mak
@@ -1,7 +1,6 @@
 TARGET_ARCH=ppc64
 TARGET_BASE_ARCH=ppc
 TARGET_BIG_ENDIAN=y
-TARGET_SUPPORTS_MTTCG=y
 TARGET_KVM_HAVE_GUEST_DEBUG=y
 TARGET_XML_FILES= gdb-xml/power64-core.xml gdb-xml/power-fpu.xml gdb-xml/power-altivec.xml gdb-xml/power-spe.xml gdb-xml/power-vsx.xml
 # all boards require libfdt
diff --git a/configs/targets/riscv32-softmmu.mak b/configs/targets/riscv32-softmmu.mak
index c828066ce6b..db55275b868 100644
--- a/configs/targets/riscv32-softmmu.mak
+++ b/configs/targets/riscv32-softmmu.mak
@@ -1,6 +1,5 @@
 TARGET_ARCH=riscv32
 TARGET_BASE_ARCH=riscv
-TARGET_SUPPORTS_MTTCG=y
 TARGET_XML_FILES= gdb-xml/riscv-32bit-cpu.xml gdb-xml/riscv-32bit-fpu.xml gdb-xml/riscv-64bit-fpu.xml gdb-xml/riscv-32bit-virtual.xml
 # needed by boot.c
 TARGET_NEED_FDT=y
diff --git a/configs/targets/riscv64-softmmu.mak b/configs/targets/riscv64-softmmu.mak
index 09f613d24a0..2bdd4a62cd2 100644
--- a/configs/targets/riscv64-softmmu.mak
+++ b/configs/targets/riscv64-softmmu.mak
@@ -1,6 +1,5 @@
 TARGET_ARCH=riscv64
 TARGET_BASE_ARCH=riscv
-TARGET_SUPPORTS_MTTCG=y
 TARGET_KVM_HAVE_GUEST_DEBUG=y
 TARGET_XML_FILES= gdb-xml/riscv-64bit-cpu.xml gdb-xml/riscv-32bit-fpu.xml gdb-xml/riscv-64bit-fpu.xml gdb-xml/riscv-64bit-virtual.xml gdb-xml/riscv-32bit-cpu.xml gdb-xml/riscv-32bit-virtual.xml
 # needed by boot.c
diff --git a/configs/targets/s390x-softmmu.mak b/configs/targets/s390x-softmmu.mak
index 5242ebe7c2e..76dd5de6584 100644
--- a/configs/targets/s390x-softmmu.mak
+++ b/configs/targets/s390x-softmmu.mak
@@ -1,6 +1,5 @@
 TARGET_ARCH=s390x
 TARGET_BIG_ENDIAN=y
-TARGET_SUPPORTS_MTTCG=y
 TARGET_KVM_HAVE_GUEST_DEBUG=y
 TARGET_XML_FILES= gdb-xml/s390x-core64.xml gdb-xml/s390-acr.xml gdb-xml/s390-fpr.xml gdb-xml/s390-vx.xml gdb-xml/s390-cr.xml gdb-xml/s390-virt.xml gdb-xml/s390-virt-kvm.xml gdb-xml/s390-gs.xml
 TARGET_LONG_BITS=64
diff --git a/configs/targets/sparc-softmmu.mak b/configs/targets/sparc-softmmu.mak
index 78c2e25bd13..57801faf1fc 100644
--- a/configs/targets/sparc-softmmu.mak
+++ b/configs/targets/sparc-softmmu.mak
@@ -1,4 +1,3 @@
 TARGET_ARCH=sparc
 TARGET_BIG_ENDIAN=y
-TARGET_SUPPORTS_MTTCG=y
 TARGET_LONG_BITS=32
diff --git a/configs/targets/sparc64-softmmu.mak b/configs/targets/sparc64-softmmu.mak
index f7bab97a002..2504e31ae33 100644
--- a/configs/targets/sparc64-softmmu.mak
+++ b/configs/targets/sparc64-softmmu.mak
@@ -1,5 +1,4 @@
 TARGET_ARCH=sparc64
 TARGET_BASE_ARCH=sparc
 TARGET_BIG_ENDIAN=y
-TARGET_SUPPORTS_MTTCG=y
 TARGET_LONG_BITS=64
diff --git a/configs/targets/x86_64-softmmu.mak b/configs/targets/x86_64-softmmu.mak
index 1ceefde1313..5619b2bc686 100644
--- a/configs/targets/x86_64-softmmu.mak
+++ b/configs/targets/x86_64-softmmu.mak
@@ -1,6 +1,5 @@
 TARGET_ARCH=x86_64
 TARGET_BASE_ARCH=i386
-TARGET_SUPPORTS_MTTCG=y
 TARGET_KVM_HAVE_GUEST_DEBUG=y
 TARGET_KVM_HAVE_RESET_PARKED_VCPU=y
 TARGET_XML_FILES= gdb-xml/i386-64bit.xml
diff --git a/configs/targets/xtensa-softmmu.mak b/configs/targets/xtensa-softmmu.mak
index 65845df4ffa..2a9797338a6 100644
--- a/configs/targets/xtensa-softmmu.mak
+++ b/configs/targets/xtensa-softmmu.mak
@@ -1,3 +1,2 @@
 TARGET_ARCH=xtensa
-TARGET_SUPPORTS_MTTCG=y
 TARGET_LONG_BITS=32
diff --git a/configs/targets/xtensaeb-softmmu.mak b/configs/targets/xtensaeb-softmmu.mak
index f1f789d6971..5204729af8b 100644
--- a/configs/targets/xtensaeb-softmmu.mak
+++ b/configs/targets/xtensaeb-softmmu.mak
@@ -1,4 +1,3 @@
 TARGET_ARCH=xtensa
 TARGET_BIG_ENDIAN=y
-TARGET_SUPPORTS_MTTCG=y
 TARGET_LONG_BITS=32
diff --git a/include/accel/tcg/cpu-ops.h b/include/accel/tcg/cpu-ops.h
index 5fd299cefb6..31e63b8a663 100644
--- a/include/accel/tcg/cpu-ops.h
+++ b/include/accel/tcg/cpu-ops.h
@@ -19,6 +19,14 @@
 #include "tcg/tcg-mo.h"
 
 struct TCGCPUOps {
+    /**
+     * mttcg_supported: multi-threaded TCG is supported
+     *
+     * Target (TCG frontend) supports:
+     *   - atomic instructions
+     *   - memory ordering primitives (barriers)
+     */
+    bool mttcg_supported;
 
     /**
      * @guest_default_memory_order: default barrier that is required
diff --git a/include/exec/poison.h b/include/exec/poison.h
index 8ed04b31083..6df627868c6 100644
--- a/include/exec/poison.h
+++ b/include/exec/poison.h
@@ -35,7 +35,6 @@
 
 #pragma GCC poison TARGET_HAS_BFLT
 #pragma GCC poison TARGET_NAME
-#pragma GCC poison TARGET_SUPPORTS_MTTCG
 #pragma GCC poison TARGET_BIG_ENDIAN
 #pragma GCC poison BSWAP_NEEDED
 
diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
index 2b7f89eaa20..50b3fd75c32 100644
--- a/accel/tcg/tcg-all.c
+++ b/accel/tcg/tcg-all.c
@@ -41,6 +41,7 @@
 #include "hw/boards.h"
 #include "system/tcg.h"
 #endif
+#include "accel/tcg/cpu-ops.h"
 #include "internal-common.h"
 #include "cpu-param.h"
 
@@ -91,11 +92,7 @@ static int tcg_init_machine(MachineState *ms)
 #else
     unsigned max_cpus = ms->smp.max_cpus;
 #endif
-#ifdef TARGET_SUPPORTS_MTTCG
-    bool mttcg_supported = true;
-#else
-    bool mttcg_supported = false;
-#endif
+    bool mttcg_supported = first_cpu->cc->tcg_ops->mttcg_supported; /* FIXME avoid first_cpu */
 
     tcg_allowed = true;
     mttcg_enabled = s->mttcg_enabled;
diff --git a/target/alpha/cpu.c b/target/alpha/cpu.c
index e5e14976f51..2bad59a48a4 100644
--- a/target/alpha/cpu.c
+++ b/target/alpha/cpu.c
@@ -239,6 +239,7 @@ static const TCGCPUOps alpha_tcg_ops = {
     .synchronize_from_tb = alpha_cpu_synchronize_from_tb,
     .restore_state_to_opc = alpha_restore_state_to_opc,
 
+    .mttcg_supported = true,
     /* Alpha processors have a weak memory model */
     .guest_default_memory_order = 0,
 
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index 8b9f2acf82b..a332d9a5148 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -2675,6 +2675,7 @@ static const TCGCPUOps arm_tcg_ops = {
     .debug_excp_handler = arm_debug_excp_handler,
     .restore_state_to_opc = arm_restore_state_to_opc,
 
+    .mttcg_supported = true,
     /* ARM processors have a weak memory model */
     .guest_default_memory_order = 0,
 
diff --git a/target/arm/tcg/cpu-v7m.c b/target/arm/tcg/cpu-v7m.c
index df6b7198944..7069a7a2ca8 100644
--- a/target/arm/tcg/cpu-v7m.c
+++ b/target/arm/tcg/cpu-v7m.c
@@ -238,6 +238,7 @@ static const TCGCPUOps arm_v7m_tcg_ops = {
     .debug_excp_handler = arm_debug_excp_handler,
     .restore_state_to_opc = arm_restore_state_to_opc,
 
+    .mttcg_supported = true,
     /* ARM processors have a weak memory model */
     .guest_default_memory_order = 0,
 
diff --git a/target/avr/cpu.c b/target/avr/cpu.c
index 24e52e28f44..9bc921c1cb2 100644
--- a/target/avr/cpu.c
+++ b/target/avr/cpu.c
@@ -216,6 +216,7 @@ static const TCGCPUOps avr_tcg_ops = {
     .cpu_exec_halt = avr_cpu_has_work,
     .tlb_fill = avr_cpu_tlb_fill,
     .do_interrupt = avr_cpu_do_interrupt,
+    .mttcg_supported = false,
     .guest_default_memory_order = 0,
 };
 
diff --git a/target/hexagon/cpu.c b/target/hexagon/cpu.c
index 34734b0edb0..7bcf297998f 100644
--- a/target/hexagon/cpu.c
+++ b/target/hexagon/cpu.c
@@ -324,6 +324,7 @@ static const TCGCPUOps hexagon_tcg_ops = {
     .translate_code = hexagon_translate_code,
     .synchronize_from_tb = hexagon_cpu_synchronize_from_tb,
     .restore_state_to_opc = hexagon_restore_state_to_opc,
+    .mttcg_supported = false,
     /* MTTCG not yet supported: require strict ordering */
     .guest_default_memory_order = TCG_MO_ALL,
 };
diff --git a/target/hppa/cpu.c b/target/hppa/cpu.c
index 997bd69db19..31f29a95ac0 100644
--- a/target/hppa/cpu.c
+++ b/target/hppa/cpu.c
@@ -257,6 +257,7 @@ static const TCGCPUOps hppa_tcg_ops = {
     .synchronize_from_tb = hppa_cpu_synchronize_from_tb,
     .restore_state_to_opc = hppa_restore_state_to_opc,
 
+    .mttcg_supported = true,
     /* PA-RISC 1.x processors have a strong memory model.  */
     /*
      * ??? While we do not yet implement PA-RISC 2.0, those processors have
diff --git a/target/i386/tcg/tcg-cpu.c b/target/i386/tcg/tcg-cpu.c
index 4a76c475971..62d36880ceb 100644
--- a/target/i386/tcg/tcg-cpu.c
+++ b/target/i386/tcg/tcg-cpu.c
@@ -129,6 +129,7 @@ static const TCGCPUOps x86_tcg_ops = {
     .need_replay_interrupt = x86_need_replay_interrupt,
 #endif /* !CONFIG_USER_ONLY */
 
+    .mttcg_supported = true,
     /*
      * The x86 has a strong memory model with some store-after-load re-ordering
      */
diff --git a/target/loongarch/cpu.c b/target/loongarch/cpu.c
index c39ff056157..460d05b1ee5 100644
--- a/target/loongarch/cpu.c
+++ b/target/loongarch/cpu.c
@@ -869,6 +869,7 @@ static const TCGCPUOps loongarch_tcg_ops = {
     .synchronize_from_tb = loongarch_cpu_synchronize_from_tb,
     .restore_state_to_opc = loongarch_restore_state_to_opc,
 
+    .mttcg_supported = true,
     .guest_default_memory_order = 0,
 
 #ifndef CONFIG_USER_ONLY
diff --git a/target/m68k/cpu.c b/target/m68k/cpu.c
index e96b379e266..4c94f031b13 100644
--- a/target/m68k/cpu.c
+++ b/target/m68k/cpu.c
@@ -593,6 +593,7 @@ static const TCGCPUOps m68k_tcg_ops = {
     .translate_code = m68k_translate_code,
     .restore_state_to_opc = m68k_restore_state_to_opc,
 
+    .mttcg_supported = false,
     /* MTTCG not yet supported: require strict ordering */
     .guest_default_memory_order = TCG_MO_ALL,
 
diff --git a/target/microblaze/cpu.c b/target/microblaze/cpu.c
index 4b9ef6e52c4..5604c2a34a0 100644
--- a/target/microblaze/cpu.c
+++ b/target/microblaze/cpu.c
@@ -432,6 +432,7 @@ static const TCGCPUOps mb_tcg_ops = {
     .synchronize_from_tb = mb_cpu_synchronize_from_tb,
     .restore_state_to_opc = mb_restore_state_to_opc,
 
+    .mttcg_supported = true,
     /* MicroBlaze is always in-order. */
     .guest_default_memory_order = TCG_MO_ALL,
 
diff --git a/target/mips/cpu.c b/target/mips/cpu.c
index 5ddc9bbb829..eee604e9254 100644
--- a/target/mips/cpu.c
+++ b/target/mips/cpu.c
@@ -554,6 +554,7 @@ static const TCGCPUOps mips_tcg_ops = {
     .synchronize_from_tb = mips_cpu_synchronize_from_tb,
     .restore_state_to_opc = mips_restore_state_to_opc,
 
+    .mttcg_supported = TARGET_LONG_BITS == 32,
     .guest_default_memory_order = 0,
 
 #if !defined(CONFIG_USER_ONLY)
diff --git a/target/openrisc/cpu.c b/target/openrisc/cpu.c
index 6a878aaadd8..1a6ca405a0c 100644
--- a/target/openrisc/cpu.c
+++ b/target/openrisc/cpu.c
@@ -248,6 +248,7 @@ static const TCGCPUOps openrisc_tcg_ops = {
     .synchronize_from_tb = openrisc_cpu_synchronize_from_tb,
     .restore_state_to_opc = openrisc_restore_state_to_opc,
 
+    .mttcg_supported = true,
     .guest_default_memory_order = 0,
 
 #ifndef CONFIG_USER_ONLY
diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
index 28fbbb8d3c1..ed79cc1a6b7 100644
--- a/target/ppc/cpu_init.c
+++ b/target/ppc/cpu_init.c
@@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
   .translate_code = ppc_translate_code,
   .restore_state_to_opc = ppc_restore_state_to_opc,
 
+  .mttcg_supported = TARGET_LONG_BITS == 64,
   .guest_default_memory_order = 0,
 
 #ifdef CONFIG_USER_ONLY
diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
index 5ccd7b9dc07..f26cfae6cdb 100644
--- a/target/riscv/tcg/tcg-cpu.c
+++ b/target/riscv/tcg/tcg-cpu.c
@@ -140,6 +140,7 @@ static const TCGCPUOps riscv_tcg_ops = {
     .synchronize_from_tb = riscv_cpu_synchronize_from_tb,
     .restore_state_to_opc = riscv_restore_state_to_opc,
 
+    .mttcg_supported = true,
     .guest_default_memory_order = 0,
 
 #ifndef CONFIG_USER_ONLY
diff --git a/target/rx/cpu.c b/target/rx/cpu.c
index 6a24e7e9136..4bcbc22d3ad 100644
--- a/target/rx/cpu.c
+++ b/target/rx/cpu.c
@@ -213,6 +213,7 @@ static const TCGCPUOps rx_tcg_ops = {
     .cpu_exec_halt = rx_cpu_has_work,
     .do_interrupt = rx_cpu_do_interrupt,
 
+    .mttcg_supported = false,
     /* MTTCG not yet supported: require strict ordering */
     .guest_default_memory_order = TCG_MO_ALL,
 };
diff --git a/target/s390x/cpu.c b/target/s390x/cpu.c
index 12fd853c00a..7ca34002d37 100644
--- a/target/s390x/cpu.c
+++ b/target/s390x/cpu.c
@@ -361,6 +361,7 @@ static const TCGCPUOps s390_tcg_ops = {
     .do_unaligned_access = s390x_cpu_do_unaligned_access,
 #endif /* !CONFIG_USER_ONLY */
 
+    .mttcg_supported = true,
     /*
      * The z/Architecture has a strong memory model with some
      * store-after-load re-ordering.
diff --git a/target/sh4/cpu.c b/target/sh4/cpu.c
index ce9ed75107a..b796bac8afd 100644
--- a/target/sh4/cpu.c
+++ b/target/sh4/cpu.c
@@ -267,6 +267,7 @@ static const TCGCPUOps superh_tcg_ops = {
     .synchronize_from_tb = superh_cpu_synchronize_from_tb,
     .restore_state_to_opc = superh_restore_state_to_opc,
 
+    .mttcg_supported = false,
     /* MTTCG not yet supported: require strict ordering */
     .guest_default_memory_order = TCG_MO_ALL,
 
diff --git a/target/sparc/cpu.c b/target/sparc/cpu.c
index 39bd0c42855..2ae7173c0bc 100644
--- a/target/sparc/cpu.c
+++ b/target/sparc/cpu.c
@@ -1005,6 +1005,7 @@ static const TCGCPUOps sparc_tcg_ops = {
     .synchronize_from_tb = sparc_cpu_synchronize_from_tb,
     .restore_state_to_opc = sparc_restore_state_to_opc,
 
+    .mttcg_supported = true,
     /*
      * From Oracle SPARC Architecture 2015:
      *
diff --git a/target/tricore/cpu.c b/target/tricore/cpu.c
index e0a48065948..aea1f211516 100644
--- a/target/tricore/cpu.c
+++ b/target/tricore/cpu.c
@@ -179,6 +179,7 @@ static const TCGCPUOps tricore_tcg_ops = {
     .tlb_fill = tricore_cpu_tlb_fill,
     .cpu_exec_interrupt = tricore_cpu_exec_interrupt,
     .cpu_exec_halt = tricore_cpu_has_work,
+    .mttcg_supported = false,
     /* MTTCG not yet supported: require strict ordering */
     .guest_default_memory_order = TCG_MO_ALL,
 };
diff --git a/target/xtensa/cpu.c b/target/xtensa/cpu.c
index dd9061ba469..8cc987f8408 100644
--- a/target/xtensa/cpu.c
+++ b/target/xtensa/cpu.c
@@ -237,6 +237,7 @@ static const TCGCPUOps xtensa_tcg_ops = {
     .debug_excp_handler = xtensa_breakpoint_handler,
     .restore_state_to_opc = xtensa_restore_state_to_opc,
 
+    .mttcg_supported = true,
     /* Xtensa processors have a weak memory model */
     .guest_default_memory_order = 0,
 
-- 
2.47.1

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Anton Johansson via]

On 21/03/25, Philippe Mathieu-Daudé wrote:
> Instead of having a compile-time TARGET_SUPPORTS_MTTCG definition,
> have each target set the 'mttcg_supported' field in the TCGCPUOps
> structure.
> 
> Since so far we only emulate one target architecture at a time,
> tcg_init_machine() gets whether MTTCG is supported via the
> &first_cpu global.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>  docs/devel/multi-thread-tcg.rst          | 2 +-
>  configs/targets/aarch64-softmmu.mak      | 1 -
>  configs/targets/alpha-softmmu.mak        | 1 -
>  configs/targets/arm-softmmu.mak          | 1 -
>  configs/targets/hppa-softmmu.mak         | 1 -
>  configs/targets/i386-softmmu.mak         | 1 -
>  configs/targets/loongarch64-softmmu.mak  | 1 -
>  configs/targets/microblaze-softmmu.mak   | 1 -
>  configs/targets/microblazeel-softmmu.mak | 1 -
>  configs/targets/mips-softmmu.mak         | 1 -
>  configs/targets/mipsel-softmmu.mak       | 1 -
>  configs/targets/or1k-softmmu.mak         | 1 -
>  configs/targets/ppc64-softmmu.mak        | 1 -
>  configs/targets/riscv32-softmmu.mak      | 1 -
>  configs/targets/riscv64-softmmu.mak      | 1 -
>  configs/targets/s390x-softmmu.mak        | 1 -
>  configs/targets/sparc-softmmu.mak        | 1 -
>  configs/targets/sparc64-softmmu.mak      | 1 -
>  configs/targets/x86_64-softmmu.mak       | 1 -
>  configs/targets/xtensa-softmmu.mak       | 1 -
>  configs/targets/xtensaeb-softmmu.mak     | 1 -
>  include/accel/tcg/cpu-ops.h              | 8 ++++++++
>  include/exec/poison.h                    | 1 -
>  accel/tcg/tcg-all.c                      | 7 ++-----
>  target/alpha/cpu.c                       | 1 +
>  target/arm/cpu.c                         | 1 +
>  target/arm/tcg/cpu-v7m.c                 | 1 +
>  target/avr/cpu.c                         | 1 +
>  target/hexagon/cpu.c                     | 1 +
>  target/hppa/cpu.c                        | 1 +
>  target/i386/tcg/tcg-cpu.c                | 1 +
>  target/loongarch/cpu.c                   | 1 +
>  target/m68k/cpu.c                        | 1 +
>  target/microblaze/cpu.c                  | 1 +
>  target/mips/cpu.c                        | 1 +
>  target/openrisc/cpu.c                    | 1 +
>  target/ppc/cpu_init.c                    | 1 +
>  target/riscv/tcg/tcg-cpu.c               | 1 +
>  target/rx/cpu.c                          | 1 +
>  target/s390x/cpu.c                       | 1 +
>  target/sh4/cpu.c                         | 1 +
>  target/sparc/cpu.c                       | 1 +
>  target/tricore/cpu.c                     | 1 +
>  target/xtensa/cpu.c                      | 1 +
>  44 files changed, 31 insertions(+), 27 deletions(-)

Nice!

Reviewed-by: Anton Johansson <anjo@rev.ng>

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Pierrick Bouvier]

On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> Instead of having a compile-time TARGET_SUPPORTS_MTTCG definition,
> have each target set the 'mttcg_supported' field in the TCGCPUOps
> structure.
> 
> Since so far we only emulate one target architecture at a time,
> tcg_init_machine() gets whether MTTCG is supported via the
> &first_cpu global.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   docs/devel/multi-thread-tcg.rst          | 2 +-
>   configs/targets/aarch64-softmmu.mak      | 1 -
>   configs/targets/alpha-softmmu.mak        | 1 -
>   configs/targets/arm-softmmu.mak          | 1 -
>   configs/targets/hppa-softmmu.mak         | 1 -
>   configs/targets/i386-softmmu.mak         | 1 -
>   configs/targets/loongarch64-softmmu.mak  | 1 -
>   configs/targets/microblaze-softmmu.mak   | 1 -
>   configs/targets/microblazeel-softmmu.mak | 1 -
>   configs/targets/mips-softmmu.mak         | 1 -
>   configs/targets/mipsel-softmmu.mak       | 1 -
>   configs/targets/or1k-softmmu.mak         | 1 -
>   configs/targets/ppc64-softmmu.mak        | 1 -
>   configs/targets/riscv32-softmmu.mak      | 1 -
>   configs/targets/riscv64-softmmu.mak      | 1 -
>   configs/targets/s390x-softmmu.mak        | 1 -
>   configs/targets/sparc-softmmu.mak        | 1 -
>   configs/targets/sparc64-softmmu.mak      | 1 -
>   configs/targets/x86_64-softmmu.mak       | 1 -
>   configs/targets/xtensa-softmmu.mak       | 1 -
>   configs/targets/xtensaeb-softmmu.mak     | 1 -
>   include/accel/tcg/cpu-ops.h              | 8 ++++++++
>   include/exec/poison.h                    | 1 -
>   accel/tcg/tcg-all.c                      | 7 ++-----
>   target/alpha/cpu.c                       | 1 +
>   target/arm/cpu.c                         | 1 +
>   target/arm/tcg/cpu-v7m.c                 | 1 +
>   target/avr/cpu.c                         | 1 +
>   target/hexagon/cpu.c                     | 1 +
>   target/hppa/cpu.c                        | 1 +
>   target/i386/tcg/tcg-cpu.c                | 1 +
>   target/loongarch/cpu.c                   | 1 +
>   target/m68k/cpu.c                        | 1 +
>   target/microblaze/cpu.c                  | 1 +
>   target/mips/cpu.c                        | 1 +
>   target/openrisc/cpu.c                    | 1 +
>   target/ppc/cpu_init.c                    | 1 +
>   target/riscv/tcg/tcg-cpu.c               | 1 +
>   target/rx/cpu.c                          | 1 +
>   target/s390x/cpu.c                       | 1 +
>   target/sh4/cpu.c                         | 1 +
>   target/sparc/cpu.c                       | 1 +
>   target/tricore/cpu.c                     | 1 +
>   target/xtensa/cpu.c                      | 1 +
>   44 files changed, 31 insertions(+), 27 deletions(-)
> 
> diff --git a/docs/devel/multi-thread-tcg.rst b/docs/devel/multi-thread-tcg.rst
> index 14a2a9dc7b5..da9a1530c9f 100644
> --- a/docs/devel/multi-thread-tcg.rst
> +++ b/docs/devel/multi-thread-tcg.rst
> @@ -30,7 +30,7 @@ user-space thread. This is enabled by default for all FE/BE
>   combinations where the host memory model is able to accommodate the
>   guest (TCGCPUOps::guest_default_memory_order & ~TCG_TARGET_DEFAULT_MO is zero)
>   and the guest has had the required work done to support this safely
> -(TARGET_SUPPORTS_MTTCG).
> +(TCGCPUOps::mttcg_supported).
>   
>   System emulation will fall back to the original round robin approach
>   if:
> diff --git a/configs/targets/aarch64-softmmu.mak b/configs/targets/aarch64-softmmu.mak
> index 82cb72cb83d..5dfeb35af90 100644
> --- a/configs/targets/aarch64-softmmu.mak
> +++ b/configs/targets/aarch64-softmmu.mak
> @@ -1,6 +1,5 @@
>   TARGET_ARCH=aarch64
>   TARGET_BASE_ARCH=arm
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_KVM_HAVE_GUEST_DEBUG=y
>   TARGET_XML_FILES= gdb-xml/aarch64-core.xml gdb-xml/aarch64-fpu.xml gdb-xml/arm-core.xml gdb-xml/arm-vfp.xml gdb-xml/arm-vfp3.xml gdb-xml/arm-vfp-sysregs.xml gdb-xml/arm-neon.xml gdb-xml/arm-m-profile.xml gdb-xml/arm-m-profile-mve.xml gdb-xml/aarch64-pauth.xml
>   # needed by boot.c
> diff --git a/configs/targets/alpha-softmmu.mak b/configs/targets/alpha-softmmu.mak
> index 89f3517aca0..5275076e50d 100644
> --- a/configs/targets/alpha-softmmu.mak
> +++ b/configs/targets/alpha-softmmu.mak
> @@ -1,3 +1,2 @@
>   TARGET_ARCH=alpha
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_LONG_BITS=64
> diff --git a/configs/targets/arm-softmmu.mak b/configs/targets/arm-softmmu.mak
> index afc64f5927b..6a5a8eda949 100644
> --- a/configs/targets/arm-softmmu.mak
> +++ b/configs/targets/arm-softmmu.mak
> @@ -1,5 +1,4 @@
>   TARGET_ARCH=arm
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_XML_FILES= gdb-xml/arm-core.xml gdb-xml/arm-vfp.xml gdb-xml/arm-vfp3.xml gdb-xml/arm-vfp-sysregs.xml gdb-xml/arm-neon.xml gdb-xml/arm-m-profile.xml gdb-xml/arm-m-profile-mve.xml
>   # needed by boot.c
>   TARGET_NEED_FDT=y
> diff --git a/configs/targets/hppa-softmmu.mak b/configs/targets/hppa-softmmu.mak
> index 63ca74ed5e6..ea331107a08 100644
> --- a/configs/targets/hppa-softmmu.mak
> +++ b/configs/targets/hppa-softmmu.mak
> @@ -1,4 +1,3 @@
>   TARGET_ARCH=hppa
>   TARGET_BIG_ENDIAN=y
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_LONG_BITS=64
> diff --git a/configs/targets/i386-softmmu.mak b/configs/targets/i386-softmmu.mak
> index 5dd89217560..e9d89e8ab41 100644
> --- a/configs/targets/i386-softmmu.mak
> +++ b/configs/targets/i386-softmmu.mak
> @@ -1,5 +1,4 @@
>   TARGET_ARCH=i386
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_KVM_HAVE_GUEST_DEBUG=y
>   TARGET_KVM_HAVE_RESET_PARKED_VCPU=y
>   TARGET_XML_FILES= gdb-xml/i386-32bit.xml
> diff --git a/configs/targets/loongarch64-softmmu.mak b/configs/targets/loongarch64-softmmu.mak
> index 351341132f6..fc44c54233d 100644
> --- a/configs/targets/loongarch64-softmmu.mak
> +++ b/configs/targets/loongarch64-softmmu.mak
> @@ -1,7 +1,6 @@
>   TARGET_ARCH=loongarch64
>   TARGET_BASE_ARCH=loongarch
>   TARGET_KVM_HAVE_GUEST_DEBUG=y
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_XML_FILES= gdb-xml/loongarch-base32.xml gdb-xml/loongarch-base64.xml gdb-xml/loongarch-fpu.xml gdb-xml/loongarch-lsx.xml gdb-xml/loongarch-lasx.xml
>   # all boards require libfdt
>   TARGET_NEED_FDT=y
> diff --git a/configs/targets/microblaze-softmmu.mak b/configs/targets/microblaze-softmmu.mak
> index 99a33ed44a8..23457d0ae65 100644
> --- a/configs/targets/microblaze-softmmu.mak
> +++ b/configs/targets/microblaze-softmmu.mak
> @@ -1,6 +1,5 @@
>   TARGET_ARCH=microblaze
>   TARGET_BIG_ENDIAN=y
> -TARGET_SUPPORTS_MTTCG=y
>   # needed by boot.c
>   TARGET_NEED_FDT=y
>   TARGET_XML_FILES=gdb-xml/microblaze-core.xml gdb-xml/microblaze-stack-protect.xml
> diff --git a/configs/targets/microblazeel-softmmu.mak b/configs/targets/microblazeel-softmmu.mak
> index 52cdeae1a28..c82c509623d 100644
> --- a/configs/targets/microblazeel-softmmu.mak
> +++ b/configs/targets/microblazeel-softmmu.mak
> @@ -1,5 +1,4 @@
>   TARGET_ARCH=microblaze
> -TARGET_SUPPORTS_MTTCG=y
>   # needed by boot.c
>   TARGET_NEED_FDT=y
>   TARGET_XML_FILES=gdb-xml/microblaze-core.xml gdb-xml/microblaze-stack-protect.xml
> diff --git a/configs/targets/mips-softmmu.mak b/configs/targets/mips-softmmu.mak
> index b62a0882499..c9588066b8d 100644
> --- a/configs/targets/mips-softmmu.mak
> +++ b/configs/targets/mips-softmmu.mak
> @@ -1,4 +1,3 @@
>   TARGET_ARCH=mips
>   TARGET_BIG_ENDIAN=y
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_LONG_BITS=32
> diff --git a/configs/targets/mipsel-softmmu.mak b/configs/targets/mipsel-softmmu.mak
> index 620ec681785..90e09bdc3e5 100644
> --- a/configs/targets/mipsel-softmmu.mak
> +++ b/configs/targets/mipsel-softmmu.mak
> @@ -1,3 +1,2 @@
>   TARGET_ARCH=mips
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_LONG_BITS=32
> diff --git a/configs/targets/or1k-softmmu.mak b/configs/targets/or1k-softmmu.mak
> index adfddb1a8ac..0e47d9878b0 100644
> --- a/configs/targets/or1k-softmmu.mak
> +++ b/configs/targets/or1k-softmmu.mak
> @@ -1,5 +1,4 @@
>   TARGET_ARCH=openrisc
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_BIG_ENDIAN=y
>   # needed by boot.c and all boards
>   TARGET_NEED_FDT=y
> diff --git a/configs/targets/ppc64-softmmu.mak b/configs/targets/ppc64-softmmu.mak
> index 7cee0e97f43..74572864b36 100644
> --- a/configs/targets/ppc64-softmmu.mak
> +++ b/configs/targets/ppc64-softmmu.mak
> @@ -1,7 +1,6 @@
>   TARGET_ARCH=ppc64
>   TARGET_BASE_ARCH=ppc
>   TARGET_BIG_ENDIAN=y
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_KVM_HAVE_GUEST_DEBUG=y
>   TARGET_XML_FILES= gdb-xml/power64-core.xml gdb-xml/power-fpu.xml gdb-xml/power-altivec.xml gdb-xml/power-spe.xml gdb-xml/power-vsx.xml
>   # all boards require libfdt
> diff --git a/configs/targets/riscv32-softmmu.mak b/configs/targets/riscv32-softmmu.mak
> index c828066ce6b..db55275b868 100644
> --- a/configs/targets/riscv32-softmmu.mak
> +++ b/configs/targets/riscv32-softmmu.mak
> @@ -1,6 +1,5 @@
>   TARGET_ARCH=riscv32
>   TARGET_BASE_ARCH=riscv
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_XML_FILES= gdb-xml/riscv-32bit-cpu.xml gdb-xml/riscv-32bit-fpu.xml gdb-xml/riscv-64bit-fpu.xml gdb-xml/riscv-32bit-virtual.xml
>   # needed by boot.c
>   TARGET_NEED_FDT=y
> diff --git a/configs/targets/riscv64-softmmu.mak b/configs/targets/riscv64-softmmu.mak
> index 09f613d24a0..2bdd4a62cd2 100644
> --- a/configs/targets/riscv64-softmmu.mak
> +++ b/configs/targets/riscv64-softmmu.mak
> @@ -1,6 +1,5 @@
>   TARGET_ARCH=riscv64
>   TARGET_BASE_ARCH=riscv
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_KVM_HAVE_GUEST_DEBUG=y
>   TARGET_XML_FILES= gdb-xml/riscv-64bit-cpu.xml gdb-xml/riscv-32bit-fpu.xml gdb-xml/riscv-64bit-fpu.xml gdb-xml/riscv-64bit-virtual.xml gdb-xml/riscv-32bit-cpu.xml gdb-xml/riscv-32bit-virtual.xml
>   # needed by boot.c
> diff --git a/configs/targets/s390x-softmmu.mak b/configs/targets/s390x-softmmu.mak
> index 5242ebe7c2e..76dd5de6584 100644
> --- a/configs/targets/s390x-softmmu.mak
> +++ b/configs/targets/s390x-softmmu.mak
> @@ -1,6 +1,5 @@
>   TARGET_ARCH=s390x
>   TARGET_BIG_ENDIAN=y
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_KVM_HAVE_GUEST_DEBUG=y
>   TARGET_XML_FILES= gdb-xml/s390x-core64.xml gdb-xml/s390-acr.xml gdb-xml/s390-fpr.xml gdb-xml/s390-vx.xml gdb-xml/s390-cr.xml gdb-xml/s390-virt.xml gdb-xml/s390-virt-kvm.xml gdb-xml/s390-gs.xml
>   TARGET_LONG_BITS=64
> diff --git a/configs/targets/sparc-softmmu.mak b/configs/targets/sparc-softmmu.mak
> index 78c2e25bd13..57801faf1fc 100644
> --- a/configs/targets/sparc-softmmu.mak
> +++ b/configs/targets/sparc-softmmu.mak
> @@ -1,4 +1,3 @@
>   TARGET_ARCH=sparc
>   TARGET_BIG_ENDIAN=y
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_LONG_BITS=32
> diff --git a/configs/targets/sparc64-softmmu.mak b/configs/targets/sparc64-softmmu.mak
> index f7bab97a002..2504e31ae33 100644
> --- a/configs/targets/sparc64-softmmu.mak
> +++ b/configs/targets/sparc64-softmmu.mak
> @@ -1,5 +1,4 @@
>   TARGET_ARCH=sparc64
>   TARGET_BASE_ARCH=sparc
>   TARGET_BIG_ENDIAN=y
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_LONG_BITS=64
> diff --git a/configs/targets/x86_64-softmmu.mak b/configs/targets/x86_64-softmmu.mak
> index 1ceefde1313..5619b2bc686 100644
> --- a/configs/targets/x86_64-softmmu.mak
> +++ b/configs/targets/x86_64-softmmu.mak
> @@ -1,6 +1,5 @@
>   TARGET_ARCH=x86_64
>   TARGET_BASE_ARCH=i386
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_KVM_HAVE_GUEST_DEBUG=y
>   TARGET_KVM_HAVE_RESET_PARKED_VCPU=y
>   TARGET_XML_FILES= gdb-xml/i386-64bit.xml
> diff --git a/configs/targets/xtensa-softmmu.mak b/configs/targets/xtensa-softmmu.mak
> index 65845df4ffa..2a9797338a6 100644
> --- a/configs/targets/xtensa-softmmu.mak
> +++ b/configs/targets/xtensa-softmmu.mak
> @@ -1,3 +1,2 @@
>   TARGET_ARCH=xtensa
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_LONG_BITS=32
> diff --git a/configs/targets/xtensaeb-softmmu.mak b/configs/targets/xtensaeb-softmmu.mak
> index f1f789d6971..5204729af8b 100644
> --- a/configs/targets/xtensaeb-softmmu.mak
> +++ b/configs/targets/xtensaeb-softmmu.mak
> @@ -1,4 +1,3 @@
>   TARGET_ARCH=xtensa
>   TARGET_BIG_ENDIAN=y
> -TARGET_SUPPORTS_MTTCG=y
>   TARGET_LONG_BITS=32
> diff --git a/include/accel/tcg/cpu-ops.h b/include/accel/tcg/cpu-ops.h
> index 5fd299cefb6..31e63b8a663 100644
> --- a/include/accel/tcg/cpu-ops.h
> +++ b/include/accel/tcg/cpu-ops.h
> @@ -19,6 +19,14 @@
>   #include "tcg/tcg-mo.h"
>   
>   struct TCGCPUOps {
> +    /**
> +     * mttcg_supported: multi-threaded TCG is supported
> +     *
> +     * Target (TCG frontend) supports:
> +     *   - atomic instructions
> +     *   - memory ordering primitives (barriers)
> +     */
> +    bool mttcg_supported;
>   
>       /**
>        * @guest_default_memory_order: default barrier that is required
> diff --git a/include/exec/poison.h b/include/exec/poison.h
> index 8ed04b31083..6df627868c6 100644
> --- a/include/exec/poison.h
> +++ b/include/exec/poison.h
> @@ -35,7 +35,6 @@
>   
>   #pragma GCC poison TARGET_HAS_BFLT
>   #pragma GCC poison TARGET_NAME
> -#pragma GCC poison TARGET_SUPPORTS_MTTCG
>   #pragma GCC poison TARGET_BIG_ENDIAN
>   #pragma GCC poison BSWAP_NEEDED
>   
> diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c
> index 2b7f89eaa20..50b3fd75c32 100644
> --- a/accel/tcg/tcg-all.c
> +++ b/accel/tcg/tcg-all.c
> @@ -41,6 +41,7 @@
>   #include "hw/boards.h"
>   #include "system/tcg.h"
>   #endif
> +#include "accel/tcg/cpu-ops.h"
>   #include "internal-common.h"
>   #include "cpu-param.h"
>   
> @@ -91,11 +92,7 @@ static int tcg_init_machine(MachineState *ms)
>   #else
>       unsigned max_cpus = ms->smp.max_cpus;
>   #endif
> -#ifdef TARGET_SUPPORTS_MTTCG
> -    bool mttcg_supported = true;
> -#else
> -    bool mttcg_supported = false;
> -#endif
> +    bool mttcg_supported = first_cpu->cc->tcg_ops->mttcg_supported; /* FIXME avoid first_cpu */
>   
>       tcg_allowed = true;
>       mttcg_enabled = s->mttcg_enabled;
> diff --git a/target/alpha/cpu.c b/target/alpha/cpu.c
> index e5e14976f51..2bad59a48a4 100644
> --- a/target/alpha/cpu.c
> +++ b/target/alpha/cpu.c
> @@ -239,6 +239,7 @@ static const TCGCPUOps alpha_tcg_ops = {
>       .synchronize_from_tb = alpha_cpu_synchronize_from_tb,
>       .restore_state_to_opc = alpha_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       /* Alpha processors have a weak memory model */
>       .guest_default_memory_order = 0,
>   
> diff --git a/target/arm/cpu.c b/target/arm/cpu.c
> index 8b9f2acf82b..a332d9a5148 100644
> --- a/target/arm/cpu.c
> +++ b/target/arm/cpu.c
> @@ -2675,6 +2675,7 @@ static const TCGCPUOps arm_tcg_ops = {
>       .debug_excp_handler = arm_debug_excp_handler,
>       .restore_state_to_opc = arm_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       /* ARM processors have a weak memory model */
>       .guest_default_memory_order = 0,
>   
> diff --git a/target/arm/tcg/cpu-v7m.c b/target/arm/tcg/cpu-v7m.c
> index df6b7198944..7069a7a2ca8 100644
> --- a/target/arm/tcg/cpu-v7m.c
> +++ b/target/arm/tcg/cpu-v7m.c
> @@ -238,6 +238,7 @@ static const TCGCPUOps arm_v7m_tcg_ops = {
>       .debug_excp_handler = arm_debug_excp_handler,
>       .restore_state_to_opc = arm_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       /* ARM processors have a weak memory model */
>       .guest_default_memory_order = 0,
>   
> diff --git a/target/avr/cpu.c b/target/avr/cpu.c
> index 24e52e28f44..9bc921c1cb2 100644
> --- a/target/avr/cpu.c
> +++ b/target/avr/cpu.c
> @@ -216,6 +216,7 @@ static const TCGCPUOps avr_tcg_ops = {
>       .cpu_exec_halt = avr_cpu_has_work,
>       .tlb_fill = avr_cpu_tlb_fill,
>       .do_interrupt = avr_cpu_do_interrupt,
> +    .mttcg_supported = false,
>       .guest_default_memory_order = 0,
>   };
>   
> diff --git a/target/hexagon/cpu.c b/target/hexagon/cpu.c
> index 34734b0edb0..7bcf297998f 100644
> --- a/target/hexagon/cpu.c
> +++ b/target/hexagon/cpu.c
> @@ -324,6 +324,7 @@ static const TCGCPUOps hexagon_tcg_ops = {
>       .translate_code = hexagon_translate_code,
>       .synchronize_from_tb = hexagon_cpu_synchronize_from_tb,
>       .restore_state_to_opc = hexagon_restore_state_to_opc,
> +    .mttcg_supported = false,
>       /* MTTCG not yet supported: require strict ordering */
>       .guest_default_memory_order = TCG_MO_ALL,
>   };
> diff --git a/target/hppa/cpu.c b/target/hppa/cpu.c
> index 997bd69db19..31f29a95ac0 100644
> --- a/target/hppa/cpu.c
> +++ b/target/hppa/cpu.c
> @@ -257,6 +257,7 @@ static const TCGCPUOps hppa_tcg_ops = {
>       .synchronize_from_tb = hppa_cpu_synchronize_from_tb,
>       .restore_state_to_opc = hppa_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       /* PA-RISC 1.x processors have a strong memory model.  */
>       /*
>        * ??? While we do not yet implement PA-RISC 2.0, those processors have
> diff --git a/target/i386/tcg/tcg-cpu.c b/target/i386/tcg/tcg-cpu.c
> index 4a76c475971..62d36880ceb 100644
> --- a/target/i386/tcg/tcg-cpu.c
> +++ b/target/i386/tcg/tcg-cpu.c
> @@ -129,6 +129,7 @@ static const TCGCPUOps x86_tcg_ops = {
>       .need_replay_interrupt = x86_need_replay_interrupt,
>   #endif /* !CONFIG_USER_ONLY */
>   
> +    .mttcg_supported = true,
>       /*
>        * The x86 has a strong memory model with some store-after-load re-ordering
>        */
> diff --git a/target/loongarch/cpu.c b/target/loongarch/cpu.c
> index c39ff056157..460d05b1ee5 100644
> --- a/target/loongarch/cpu.c
> +++ b/target/loongarch/cpu.c
> @@ -869,6 +869,7 @@ static const TCGCPUOps loongarch_tcg_ops = {
>       .synchronize_from_tb = loongarch_cpu_synchronize_from_tb,
>       .restore_state_to_opc = loongarch_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       .guest_default_memory_order = 0,
>   
>   #ifndef CONFIG_USER_ONLY
> diff --git a/target/m68k/cpu.c b/target/m68k/cpu.c
> index e96b379e266..4c94f031b13 100644
> --- a/target/m68k/cpu.c
> +++ b/target/m68k/cpu.c
> @@ -593,6 +593,7 @@ static const TCGCPUOps m68k_tcg_ops = {
>       .translate_code = m68k_translate_code,
>       .restore_state_to_opc = m68k_restore_state_to_opc,
>   
> +    .mttcg_supported = false,
>       /* MTTCG not yet supported: require strict ordering */
>       .guest_default_memory_order = TCG_MO_ALL,
>   
> diff --git a/target/microblaze/cpu.c b/target/microblaze/cpu.c
> index 4b9ef6e52c4..5604c2a34a0 100644
> --- a/target/microblaze/cpu.c
> +++ b/target/microblaze/cpu.c
> @@ -432,6 +432,7 @@ static const TCGCPUOps mb_tcg_ops = {
>       .synchronize_from_tb = mb_cpu_synchronize_from_tb,
>       .restore_state_to_opc = mb_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       /* MicroBlaze is always in-order. */
>       .guest_default_memory_order = TCG_MO_ALL,
>   
> diff --git a/target/mips/cpu.c b/target/mips/cpu.c
> index 5ddc9bbb829..eee604e9254 100644
> --- a/target/mips/cpu.c
> +++ b/target/mips/cpu.c
> @@ -554,6 +554,7 @@ static const TCGCPUOps mips_tcg_ops = {
>       .synchronize_from_tb = mips_cpu_synchronize_from_tb,
>       .restore_state_to_opc = mips_restore_state_to_opc,
>   
> +    .mttcg_supported = TARGET_LONG_BITS == 32,
>       .guest_default_memory_order = 0,
>   
>   #if !defined(CONFIG_USER_ONLY)
> diff --git a/target/openrisc/cpu.c b/target/openrisc/cpu.c
> index 6a878aaadd8..1a6ca405a0c 100644
> --- a/target/openrisc/cpu.c
> +++ b/target/openrisc/cpu.c
> @@ -248,6 +248,7 @@ static const TCGCPUOps openrisc_tcg_ops = {
>       .synchronize_from_tb = openrisc_cpu_synchronize_from_tb,
>       .restore_state_to_opc = openrisc_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       .guest_default_memory_order = 0,
>   
>   #ifndef CONFIG_USER_ONLY
> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
> index 28fbbb8d3c1..ed79cc1a6b7 100644
> --- a/target/ppc/cpu_init.c
> +++ b/target/ppc/cpu_init.c
> @@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
>     .translate_code = ppc_translate_code,
>     .restore_state_to_opc = ppc_restore_state_to_opc,
>   
> +  .mttcg_supported = TARGET_LONG_BITS == 64,
>     .guest_default_memory_order = 0,
>   
>   #ifdef CONFIG_USER_ONLY
> diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c
> index 5ccd7b9dc07..f26cfae6cdb 100644
> --- a/target/riscv/tcg/tcg-cpu.c
> +++ b/target/riscv/tcg/tcg-cpu.c
> @@ -140,6 +140,7 @@ static const TCGCPUOps riscv_tcg_ops = {
>       .synchronize_from_tb = riscv_cpu_synchronize_from_tb,
>       .restore_state_to_opc = riscv_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       .guest_default_memory_order = 0,
>   
>   #ifndef CONFIG_USER_ONLY
> diff --git a/target/rx/cpu.c b/target/rx/cpu.c
> index 6a24e7e9136..4bcbc22d3ad 100644
> --- a/target/rx/cpu.c
> +++ b/target/rx/cpu.c
> @@ -213,6 +213,7 @@ static const TCGCPUOps rx_tcg_ops = {
>       .cpu_exec_halt = rx_cpu_has_work,
>       .do_interrupt = rx_cpu_do_interrupt,
>   
> +    .mttcg_supported = false,
>       /* MTTCG not yet supported: require strict ordering */
>       .guest_default_memory_order = TCG_MO_ALL,
>   };
> diff --git a/target/s390x/cpu.c b/target/s390x/cpu.c
> index 12fd853c00a..7ca34002d37 100644
> --- a/target/s390x/cpu.c
> +++ b/target/s390x/cpu.c
> @@ -361,6 +361,7 @@ static const TCGCPUOps s390_tcg_ops = {
>       .do_unaligned_access = s390x_cpu_do_unaligned_access,
>   #endif /* !CONFIG_USER_ONLY */
>   
> +    .mttcg_supported = true,
>       /*
>        * The z/Architecture has a strong memory model with some
>        * store-after-load re-ordering.
> diff --git a/target/sh4/cpu.c b/target/sh4/cpu.c
> index ce9ed75107a..b796bac8afd 100644
> --- a/target/sh4/cpu.c
> +++ b/target/sh4/cpu.c
> @@ -267,6 +267,7 @@ static const TCGCPUOps superh_tcg_ops = {
>       .synchronize_from_tb = superh_cpu_synchronize_from_tb,
>       .restore_state_to_opc = superh_restore_state_to_opc,
>   
> +    .mttcg_supported = false,
>       /* MTTCG not yet supported: require strict ordering */
>       .guest_default_memory_order = TCG_MO_ALL,
>   
> diff --git a/target/sparc/cpu.c b/target/sparc/cpu.c
> index 39bd0c42855..2ae7173c0bc 100644
> --- a/target/sparc/cpu.c
> +++ b/target/sparc/cpu.c
> @@ -1005,6 +1005,7 @@ static const TCGCPUOps sparc_tcg_ops = {
>       .synchronize_from_tb = sparc_cpu_synchronize_from_tb,
>       .restore_state_to_opc = sparc_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       /*
>        * From Oracle SPARC Architecture 2015:
>        *
> diff --git a/target/tricore/cpu.c b/target/tricore/cpu.c
> index e0a48065948..aea1f211516 100644
> --- a/target/tricore/cpu.c
> +++ b/target/tricore/cpu.c
> @@ -179,6 +179,7 @@ static const TCGCPUOps tricore_tcg_ops = {
>       .tlb_fill = tricore_cpu_tlb_fill,
>       .cpu_exec_interrupt = tricore_cpu_exec_interrupt,
>       .cpu_exec_halt = tricore_cpu_has_work,
> +    .mttcg_supported = false,
>       /* MTTCG not yet supported: require strict ordering */
>       .guest_default_memory_order = TCG_MO_ALL,
>   };
> diff --git a/target/xtensa/cpu.c b/target/xtensa/cpu.c
> index dd9061ba469..8cc987f8408 100644
> --- a/target/xtensa/cpu.c
> +++ b/target/xtensa/cpu.c
> @@ -237,6 +237,7 @@ static const TCGCPUOps xtensa_tcg_ops = {
>       .debug_excp_handler = xtensa_breakpoint_handler,
>       .restore_state_to_opc = xtensa_restore_state_to_opc,
>   
> +    .mttcg_supported = true,
>       /* Xtensa processors have a weak memory model */
>       .guest_default_memory_order = 0,
>   

Sounds like a good place to store this information, vs makefiles/headers.

Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Richard Henderson]

On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
> Instead of having a compile-time TARGET_SUPPORTS_MTTCG definition,
> have each target set the 'mttcg_supported' field in the TCGCPUOps
> structure.
> 
> Since so far we only emulate one target architecture at a time,
> tcg_init_machine() gets whether MTTCG is supported via the
> &first_cpu global.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>   docs/devel/multi-thread-tcg.rst          | 2 +-
>   configs/targets/aarch64-softmmu.mak      | 1 -
>   configs/targets/alpha-softmmu.mak        | 1 -
>   configs/targets/arm-softmmu.mak          | 1 -
>   configs/targets/hppa-softmmu.mak         | 1 -
>   configs/targets/i386-softmmu.mak         | 1 -
>   configs/targets/loongarch64-softmmu.mak  | 1 -
>   configs/targets/microblaze-softmmu.mak   | 1 -
>   configs/targets/microblazeel-softmmu.mak | 1 -
>   configs/targets/mips-softmmu.mak         | 1 -
>   configs/targets/mipsel-softmmu.mak       | 1 -
>   configs/targets/or1k-softmmu.mak         | 1 -
>   configs/targets/ppc64-softmmu.mak        | 1 -
>   configs/targets/riscv32-softmmu.mak      | 1 -
>   configs/targets/riscv64-softmmu.mak      | 1 -
>   configs/targets/s390x-softmmu.mak        | 1 -
>   configs/targets/sparc-softmmu.mak        | 1 -
>   configs/targets/sparc64-softmmu.mak      | 1 -
>   configs/targets/x86_64-softmmu.mak       | 1 -
>   configs/targets/xtensa-softmmu.mak       | 1 -
>   configs/targets/xtensaeb-softmmu.mak     | 1 -
>   include/accel/tcg/cpu-ops.h              | 8 ++++++++
>   include/exec/poison.h                    | 1 -
>   accel/tcg/tcg-all.c                      | 7 ++-----
>   target/alpha/cpu.c                       | 1 +
>   target/arm/cpu.c                         | 1 +
>   target/arm/tcg/cpu-v7m.c                 | 1 +
>   target/avr/cpu.c                         | 1 +
>   target/hexagon/cpu.c                     | 1 +
>   target/hppa/cpu.c                        | 1 +
>   target/i386/tcg/tcg-cpu.c                | 1 +
>   target/loongarch/cpu.c                   | 1 +
>   target/m68k/cpu.c                        | 1 +
>   target/microblaze/cpu.c                  | 1 +
>   target/mips/cpu.c                        | 1 +
>   target/openrisc/cpu.c                    | 1 +
>   target/ppc/cpu_init.c                    | 1 +
>   target/riscv/tcg/tcg-cpu.c               | 1 +
>   target/rx/cpu.c                          | 1 +
>   target/s390x/cpu.c                       | 1 +
>   target/sh4/cpu.c                         | 1 +
>   target/sparc/cpu.c                       | 1 +
>   target/tricore/cpu.c                     | 1 +
>   target/xtensa/cpu.c                      | 1 +
>   44 files changed, 31 insertions(+), 27 deletions(-)

As far as this goes,
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

> @@ -91,11 +92,7 @@ static int tcg_init_machine(MachineState *ms)
>   #else
>       unsigned max_cpus = ms->smp.max_cpus;
>   #endif
> -#ifdef TARGET_SUPPORTS_MTTCG
> -    bool mttcg_supported = true;
> -#else
> -    bool mttcg_supported = false;
> -#endif
> +    bool mttcg_supported = first_cpu->cc->tcg_ops->mttcg_supported; /* FIXME avoid first_cpu */

Intersection of CPU_FOREACH?

> diff --git a/target/avr/cpu.c b/target/avr/cpu.c
> index 24e52e28f44..9bc921c1cb2 100644
> --- a/target/avr/cpu.c
> +++ b/target/avr/cpu.c
> @@ -216,6 +216,7 @@ static const TCGCPUOps avr_tcg_ops = {
>       .cpu_exec_halt = avr_cpu_has_work,
>       .tlb_fill = avr_cpu_tlb_fill,
>       .do_interrupt = avr_cpu_do_interrupt,
> +    .mttcg_supported = false,
>       .guest_default_memory_order = 0,
>   };

This may well be the only true not supported.
I'm not sure what needs to happen to make that true.

> diff --git a/target/hexagon/cpu.c b/target/hexagon/cpu.c
> index 34734b0edb0..7bcf297998f 100644
> --- a/target/hexagon/cpu.c
> +++ b/target/hexagon/cpu.c
> @@ -324,6 +324,7 @@ static const TCGCPUOps hexagon_tcg_ops = {
>       .translate_code = hexagon_translate_code,
>       .synchronize_from_tb = hexagon_cpu_synchronize_from_tb,
>       .restore_state_to_opc = hexagon_restore_state_to_opc,
> +    .mttcg_supported = false,
>       /* MTTCG not yet supported: require strict ordering */
>       .guest_default_memory_order = TCG_MO_ALL,

We now have strict ordering, right there with TCG_MO_ALL.
Likewise for m68k, sh4, tricore.
I think we simply forgot to update.

> diff --git a/target/mips/cpu.c b/target/mips/cpu.c
> index 5ddc9bbb829..eee604e9254 100644
> --- a/target/mips/cpu.c
> +++ b/target/mips/cpu.c
> @@ -554,6 +554,7 @@ static const TCGCPUOps mips_tcg_ops = {
>       .synchronize_from_tb = mips_cpu_synchronize_from_tb,
>       .restore_state_to_opc = mips_restore_state_to_opc,
>   
> +    .mttcg_supported = TARGET_LONG_BITS == 32,
>       .guest_default_memory_order = 0,

What is missing from mip64?

> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
> index 28fbbb8d3c1..ed79cc1a6b7 100644
> --- a/target/ppc/cpu_init.c
> +++ b/target/ppc/cpu_init.c
> @@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
>     .translate_code = ppc_translate_code,
>     .restore_state_to_opc = ppc_restore_state_to_opc,
>   
> +  .mttcg_supported = TARGET_LONG_BITS == 64,
>     .guest_default_memory_order = 0,

Similarly.  I'd be surprised if ppc32 can't use mttcg, really.


r~

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Philippe Mathieu-Daudé]

On 23/3/25 20:07, Richard Henderson wrote:
> On 3/21/25 08:59, Philippe Mathieu-Daudé wrote:
>> Instead of having a compile-time TARGET_SUPPORTS_MTTCG definition,
>> have each target set the 'mttcg_supported' field in the TCGCPUOps
>> structure.
>>
>> Since so far we only emulate one target architecture at a time,
>> tcg_init_machine() gets whether MTTCG is supported via the
>> &first_cpu global.
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   docs/devel/multi-thread-tcg.rst          | 2 +-
>>   configs/targets/aarch64-softmmu.mak      | 1 -
>>   configs/targets/alpha-softmmu.mak        | 1 -
>>   configs/targets/arm-softmmu.mak          | 1 -
>>   configs/targets/hppa-softmmu.mak         | 1 -
>>   configs/targets/i386-softmmu.mak         | 1 -
>>   configs/targets/loongarch64-softmmu.mak  | 1 -
>>   configs/targets/microblaze-softmmu.mak   | 1 -
>>   configs/targets/microblazeel-softmmu.mak | 1 -
>>   configs/targets/mips-softmmu.mak         | 1 -
>>   configs/targets/mipsel-softmmu.mak       | 1 -
>>   configs/targets/or1k-softmmu.mak         | 1 -
>>   configs/targets/ppc64-softmmu.mak        | 1 -
>>   configs/targets/riscv32-softmmu.mak      | 1 -
>>   configs/targets/riscv64-softmmu.mak      | 1 -
>>   configs/targets/s390x-softmmu.mak        | 1 -
>>   configs/targets/sparc-softmmu.mak        | 1 -
>>   configs/targets/sparc64-softmmu.mak      | 1 -
>>   configs/targets/x86_64-softmmu.mak       | 1 -
>>   configs/targets/xtensa-softmmu.mak       | 1 -
>>   configs/targets/xtensaeb-softmmu.mak     | 1 -
>>   include/accel/tcg/cpu-ops.h              | 8 ++++++++
>>   include/exec/poison.h                    | 1 -
>>   accel/tcg/tcg-all.c                      | 7 ++-----
>>   target/alpha/cpu.c                       | 1 +
>>   target/arm/cpu.c                         | 1 +
>>   target/arm/tcg/cpu-v7m.c                 | 1 +
>>   target/avr/cpu.c                         | 1 +
>>   target/hexagon/cpu.c                     | 1 +
>>   target/hppa/cpu.c                        | 1 +
>>   target/i386/tcg/tcg-cpu.c                | 1 +
>>   target/loongarch/cpu.c                   | 1 +
>>   target/m68k/cpu.c                        | 1 +
>>   target/microblaze/cpu.c                  | 1 +
>>   target/mips/cpu.c                        | 1 +
>>   target/openrisc/cpu.c                    | 1 +
>>   target/ppc/cpu_init.c                    | 1 +
>>   target/riscv/tcg/tcg-cpu.c               | 1 +
>>   target/rx/cpu.c                          | 1 +
>>   target/s390x/cpu.c                       | 1 +
>>   target/sh4/cpu.c                         | 1 +
>>   target/sparc/cpu.c                       | 1 +
>>   target/tricore/cpu.c                     | 1 +
>>   target/xtensa/cpu.c                      | 1 +
>>   44 files changed, 31 insertions(+), 27 deletions(-)
> 
> As far as this goes,
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> 
>> @@ -91,11 +92,7 @@ static int tcg_init_machine(MachineState *ms)
>>   #else
>>       unsigned max_cpus = ms->smp.max_cpus;
>>   #endif
>> -#ifdef TARGET_SUPPORTS_MTTCG
>> -    bool mttcg_supported = true;
>> -#else
>> -    bool mttcg_supported = false;
>> -#endif
>> +    bool mttcg_supported = first_cpu->cc->tcg_ops->mttcg_supported; / 
>> * FIXME avoid first_cpu */
> 
> Intersection of CPU_FOREACH?

Short term I used CPU_RESOLVING_TYPE:

     CPUClass *cc = CPU_CLASS(object_class_by_name(CPU_RESOLVING_TYPE));
     mttcg_supported = cc->tcg_ops->mttcg_supported;

I'll move to some helper.

Long term we need the intersection. I'll add a comment for now.

>> diff --git a/target/avr/cpu.c b/target/avr/cpu.c
>> index 24e52e28f44..9bc921c1cb2 100644
>> --- a/target/avr/cpu.c
>> +++ b/target/avr/cpu.c
>> @@ -216,6 +216,7 @@ static const TCGCPUOps avr_tcg_ops = {
>>       .cpu_exec_halt = avr_cpu_has_work,
>>       .tlb_fill = avr_cpu_tlb_fill,
>>       .do_interrupt = avr_cpu_do_interrupt,
>> +    .mttcg_supported = false,
>>       .guest_default_memory_order = 0,
>>   };
> 
> This may well be the only true not supported.
> I'm not sure what needs to happen to make that true.

I'd postpone that for later ;)

>> diff --git a/target/hexagon/cpu.c b/target/hexagon/cpu.c
>> index 34734b0edb0..7bcf297998f 100644
>> --- a/target/hexagon/cpu.c
>> +++ b/target/hexagon/cpu.c
>> @@ -324,6 +324,7 @@ static const TCGCPUOps hexagon_tcg_ops = {
>>       .translate_code = hexagon_translate_code,
>>       .synchronize_from_tb = hexagon_cpu_synchronize_from_tb,
>>       .restore_state_to_opc = hexagon_restore_state_to_opc,
>> +    .mttcg_supported = false,
>>       /* MTTCG not yet supported: require strict ordering */
>>       .guest_default_memory_order = TCG_MO_ALL,
> 
> We now have strict ordering, right there with TCG_MO_ALL.
> Likewise for m68k, sh4, tricore.
> I think we simply forgot to update.

Doh... I'll clean in a preliminary patch, thanks.

>> diff --git a/target/mips/cpu.c b/target/mips/cpu.c
>> index 5ddc9bbb829..eee604e9254 100644
>> --- a/target/mips/cpu.c
>> +++ b/target/mips/cpu.c
>> @@ -554,6 +554,7 @@ static const TCGCPUOps mips_tcg_ops = {
>>       .synchronize_from_tb = mips_cpu_synchronize_from_tb,
>>       .restore_state_to_opc = mips_restore_state_to_opc,
>> +    .mttcg_supported = TARGET_LONG_BITS == 32,
>>       .guest_default_memory_order = 0,
> 
> What is missing from mip64?

See:

commit a092a95547710bcbb8c168e5ec03f55fb4ab1ad7
Author: Alex Bennée <alex.bennee@linaro.org>
Date:   Mon Mar 23 16:15:09 2020 +0000

     configure: disable MTTCG for MIPS guests

     While debugging check-acceptance failures I found an instability in
     the mips64el test case. Briefly the test case:

       retry.py -n 100 -c -- ./mips64el-softmmu/qemu-system-mips64el \
         -display none -vga none -serial mon:stdio \
         -machine malta -kernel ./vmlinux-4.7.0-rc1.I6400 \
         -cpu I6400 -smp 8 -vga std \
         -append "printk.time=0 clocksource=GIC console=tty0 " \
                 "console=ttyS0 panic=-1" \
         --no-reboot

     Reports about a 9% failure rate:

       Results summary:
       0: 91 times (91.00%), avg time 5.547 (0.45 varience/0.67 deviation)
       -6: 9 times (9.00%), avg time 3.394 (0.02 varience/0.13 deviation)
       Ran command 100 times, 91 passes

     When re-run with "--accel tcg,thread=single" the instability goes
     away.

       Results summary:
       0: 100 times (100.00%), avg time 17.318 (249.76 varience/15.80 
deviation)
       Ran command 100 times, 100 passes

     Which seems to indicate there is some aspect of the MIPS MTTCG fixes
     that has been missed. Ideally we would fix that but I'm afraid I
     don't have time to investigate and am not super familiar with the
     architecture anyway. In lieu of someone tracking down the failure
     lets disable it for now.

> 
>> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
>> index 28fbbb8d3c1..ed79cc1a6b7 100644
>> --- a/target/ppc/cpu_init.c
>> +++ b/target/ppc/cpu_init.c
>> @@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
>>     .translate_code = ppc_translate_code,
>>     .restore_state_to_opc = ppc_restore_state_to_opc,
>> +  .mttcg_supported = TARGET_LONG_BITS == 64,
>>     .guest_default_memory_order = 0,
> 
> Similarly.  I'd be surprised if ppc32 can't use mttcg, really.

Per Cédric on IRC our ppc32 implementations are single core,
so never tested for mttcg.

Regards,

Phil.

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[BALATON Zoltan]

[-- Attachment #1: Type: text/plain, Size: 1268 bytes --]

On Sun, 23 Mar 2025, Philippe Mathieu-Daudé wrote:
> On 23/3/25 20:07, Richard Henderson wrote:
>>> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
>>> index 28fbbb8d3c1..ed79cc1a6b7 100644
>>> --- a/target/ppc/cpu_init.c
>>> +++ b/target/ppc/cpu_init.c
>>> @@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
>>>     .translate_code = ppc_translate_code,
>>>     .restore_state_to_opc = ppc_restore_state_to_opc,
>>> +  .mttcg_supported = TARGET_LONG_BITS == 64,
>>>     .guest_default_memory_order = 0,
>> 
>> Similarly.  I'd be surprised if ppc32 can't use mttcg, really.
>
> Per Cédric on IRC our ppc32 implementations are single core,
> so never tested for mttcg.

We're just trying to implement more CPUs for G4 mac99 and tested it with 
-smp 2 and it seems to generally work but we need to implement more of the 
Mac hardware for this but we could already run Mac OS X with two CPUs and 
mttcg and Linux too if we add a delay on boot when second CPU is started 
(it may need implementing timebase control as done on the G4 Mac to avoid 
that but if it gets past that it seems to work). So please leave this 
enabled. Not tested does not mean it does not work so only disable if 
there's a known problem.

Regards,
BALATON Zoltan

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Richard Henderson]

On 3/23/25 15:13, BALATON Zoltan wrote:
> On Sun, 23 Mar 2025, Philippe Mathieu-Daudé wrote:
>> On 23/3/25 20:07, Richard Henderson wrote:
>>>> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
>>>> index 28fbbb8d3c1..ed79cc1a6b7 100644
>>>> --- a/target/ppc/cpu_init.c
>>>> +++ b/target/ppc/cpu_init.c
>>>> @@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
>>>>     .translate_code = ppc_translate_code,
>>>>     .restore_state_to_opc = ppc_restore_state_to_opc,
>>>> +  .mttcg_supported = TARGET_LONG_BITS == 64,
>>>>     .guest_default_memory_order = 0,
>>>
>>> Similarly.  I'd be surprised if ppc32 can't use mttcg, really.
>>
>> Per Cédric on IRC our ppc32 implementations are single core,
>> so never tested for mttcg.
> 
> We're just trying to implement more CPUs for G4 mac99 and tested it with -smp 2 and it 
> seems to generally work but we need to implement more of the Mac hardware for this but we 
> could already run Mac OS X with two CPUs and mttcg and Linux too if we add a delay on boot 
> when second CPU is started (it may need implementing timebase control as done on the G4 
> Mac to avoid that but if it gets past that it seems to work). So please leave this 
> enabled. Not tested does not mean it does not work so only disable if there's a known 
> problem.

MTTCG is not enabled for qemu-system-ppc right now.
Phil's patch does not current change behaviour.


r~

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[BALATON Zoltan]

[-- Attachment #1: Type: text/plain, Size: 2261 bytes --]

On Sun, 23 Mar 2025, Richard Henderson wrote:
> On 3/23/25 15:13, BALATON Zoltan wrote:
>> On Sun, 23 Mar 2025, Philippe Mathieu-Daudé wrote:
>>> On 23/3/25 20:07, Richard Henderson wrote:
>>>>> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
>>>>> index 28fbbb8d3c1..ed79cc1a6b7 100644
>>>>> --- a/target/ppc/cpu_init.c
>>>>> +++ b/target/ppc/cpu_init.c
>>>>> @@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
>>>>>     .translate_code = ppc_translate_code,
>>>>>     .restore_state_to_opc = ppc_restore_state_to_opc,
>>>>> +  .mttcg_supported = TARGET_LONG_BITS == 64,
>>>>>     .guest_default_memory_order = 0,
>>>> 
>>>> Similarly.  I'd be surprised if ppc32 can't use mttcg, really.
>>> 
>>> Per Cédric on IRC our ppc32 implementations are single core,
>>> so never tested for mttcg.
>> 
>> We're just trying to implement more CPUs for G4 mac99 and tested it with 
>> -smp 2 and it seems to generally work but we need to implement more of the 
>> Mac hardware for this but we could already run Mac OS X with two CPUs and 
>> mttcg and Linux too if we add a delay on boot when second CPU is started 
>> (it may need implementing timebase control as done on the G4 Mac to avoid 
>> that but if it gets past that it seems to work). So please leave this 
>> enabled. Not tested does not mean it does not work so only disable if 
>> there's a known problem.
>
> MTTCG is not enabled for qemu-system-ppc right now.

I could use -smp 2 with qemu-system-ppc after this:
https://lists.nongnu.org/archive/html/qemu-ppc/2025-02/msg00377.html
It complains about not tested for the board but works nevertheless. I 
don't remember if I've enabled mttcg or not so maybe it only ran with 
single thread which was enough for testing but others did test mttcg:
https://lists.nongnu.org/archive/html/qemu-ppc/2025-02/msg00455.html
although could be that with qemu-system-ppc64 which also includes 32 bit 
machines. But then it should work the same with qemu-system-ppc so it 
could be enabled there as well.

> Phil's patch does not current change behaviour.

Then what does the line do at all? In any case we may want to enable mttcg 
in qemu-system-ppc in the future so maybe there's no need to restrict it 
here.

Regards,
BALATON Zoltan

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Philippe Mathieu-Daudé]

On 23/3/25 23:49, BALATON Zoltan wrote:
> On Sun, 23 Mar 2025, Richard Henderson wrote:
>> On 3/23/25 15:13, BALATON Zoltan wrote:
>>> On Sun, 23 Mar 2025, Philippe Mathieu-Daudé wrote:
>>>> On 23/3/25 20:07, Richard Henderson wrote:
>>>>>> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
>>>>>> index 28fbbb8d3c1..ed79cc1a6b7 100644
>>>>>> --- a/target/ppc/cpu_init.c
>>>>>> +++ b/target/ppc/cpu_init.c
>>>>>> @@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
>>>>>>     .translate_code = ppc_translate_code,
>>>>>>     .restore_state_to_opc = ppc_restore_state_to_opc,
>>>>>> +  .mttcg_supported = TARGET_LONG_BITS == 64,
>>>>>>     .guest_default_memory_order = 0,
>>>>>
>>>>> Similarly.  I'd be surprised if ppc32 can't use mttcg, really.
>>>>
>>>> Per Cédric on IRC our ppc32 implementations are single core,
>>>> so never tested for mttcg.
>>>
>>> We're just trying to implement more CPUs for G4 mac99 and tested it 
>>> with -smp 2 and it seems to generally work but we need to implement 
>>> more of the Mac hardware for this but we could already run Mac OS X 
>>> with two CPUs and mttcg and Linux too if we add a delay on boot when 
>>> second CPU is started (it may need implementing timebase control as 
>>> done on the G4 Mac to avoid that but if it gets past that it seems to 
>>> work). So please leave this enabled. Not tested does not mean it does 
>>> not work so only disable if there's a known problem.
>>
>> MTTCG is not enabled for qemu-system-ppc right now.
> 
> I could use -smp 2 with qemu-system-ppc after this:
> https://lists.nongnu.org/archive/html/qemu-ppc/2025-02/msg00377.html
> It complains about not tested for the board but works nevertheless. I 
> don't remember if I've enabled mttcg or not so maybe it only ran with 
> single thread which was enough for testing but others did test mttcg:
> https://lists.nongnu.org/archive/html/qemu-ppc/2025-02/msg00455.html
> although could be that with qemu-system-ppc64 which also includes 32 bit 
> machines. But then it should work the same with qemu-system-ppc so it 
> could be enabled there as well.
> 
>> Phil's patch does not current change behaviour.
> 
> Then what does the line do at all? In any case we may want to enable 
> mttcg in qemu-system-ppc in the future so maybe there's no need to 
> restrict it here.

In the future you can enable it posting this patch:

-- >8 --
diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
@@ -7492,6 +7492,7 @@ static const TCGCPUOps ppc_tcg_ops = {
-  .mttcg_supported = TARGET_LONG_BITS == 64,
+  .mttcg_supported = true,
---

which is equivalent of today's:

-- >8 --
diff --git a/configs/targets/ppc-softmmu.mak 
b/configs/targets/ppc-softmmu.mak
@@ -5 +5,2 @@ TARGET_XML_FILES= gdb-xml/power-core.xml 
gdb-xml/power-fpu.xml gdb-xml/power-alt
  TARGET_LONG_BITS=32
+TARGET_SUPPORTS_MTTCG=y
---

I'll let you write the commit description.

Regards,

Phil.

Re: [PATCH-for-10.1 4/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Philippe Mathieu-Daudé]

On 23/3/25 23:13, BALATON Zoltan wrote:
> On Sun, 23 Mar 2025, Philippe Mathieu-Daudé wrote:
>> On 23/3/25 20:07, Richard Henderson wrote:
>>>> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
>>>> index 28fbbb8d3c1..ed79cc1a6b7 100644
>>>> --- a/target/ppc/cpu_init.c
>>>> +++ b/target/ppc/cpu_init.c
>>>> @@ -7490,6 +7490,7 @@ static const TCGCPUOps ppc_tcg_ops = {
>>>>     .translate_code = ppc_translate_code,
>>>>     .restore_state_to_opc = ppc_restore_state_to_opc,
>>>> +  .mttcg_supported = TARGET_LONG_BITS == 64,
>>>>     .guest_default_memory_order = 0,
>>>
>>> Similarly.  I'd be surprised if ppc32 can't use mttcg, really.
>>
>> Per Cédric on IRC our ppc32 implementations are single core,
>> so never tested for mttcg.
> 
> We're just trying to implement more CPUs for G4 mac99 and tested it with 
> -smp 2 and it seems to generally work but we need to implement more of 
> the Mac hardware for this but we could already run Mac OS X with two 
> CPUs and mttcg and Linux too if we add a delay on boot when second CPU 
> is started (it may need implementing timebase control as done on the G4 
> Mac to avoid that but if it gets past that it seems to work). So please 
> leave this enabled. Not tested does not mean it does not work so only 
> disable if there's a known problem.

This patch aims to be a no-op (no logical change), so does not disable
it since it is already disabled:

$ git grep TARGET_SUPPORTS_MTTCG master -- configs/targets/ppc-softmmu.mak
$

Enabling would come as a distinct patch by someone with good
knowledge of the architecture, so not me :)

Re: [PATCH-for-10.1 0/4] tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field

[Philippe Mathieu-Daudé]

(forgot to Cc Alex!)

On 21/3/25 16:59, Philippe Mathieu-Daudé wrote:
> In this series we replace the TARGET_SUPPORTS_MTTCG (Makefile)
> definition by a 'mttcg_supported' field in TCGCPUOps.
> 
> Based-on: <20250321125737.72839-1-philmd@linaro.org>
> 
> Philippe Mathieu-Daudé (4):
>    target/riscv: Restrict RV128 MTTCG check on system emulation
>    tcg: Move qemu_tcg_mttcg_enabled() to 'system/tcg.h'
>    tcg: Convert TCGState::mttcg_enabled to TriState
>    tcg: Convert TARGET_SUPPORTS_MTTCG to TCGCPUOps::mttcg_supported field
> 
>   docs/devel/multi-thread-tcg.rst          |  2 +-