This patch enables multithread TCG for OpenRISC.  Since the or1k shared
syncrhonized timer can be updated from each vCPU via helpers we use a
mutex to synchronize updates.

Signed-off-by: Stafford Horne <shorne@gmail.com>
---
Since v2:
 - Removed cpu_openrisc_timer_has_advanced lock optimization, measuring revealed
   it did not help much.

 configs/targets/or1k-softmmu.mak | 1 +
 target/openrisc/cpu.h            | 2 ++
 target/openrisc/sys_helper.c     | 7 ++++++-
 3 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/configs/targets/or1k-softmmu.mak b/configs/targets/or1k-softmmu.mak
index 263e970870..432f855a30 100644
--- a/configs/targets/or1k-softmmu.mak
+++ b/configs/targets/or1k-softmmu.mak
@@ -1,3 +1,4 @@
 TARGET_ARCH=openrisc
+TARGET_SUPPORTS_MTTCG=y
 TARGET_BIG_ENDIAN=y
 TARGET_NEED_FDT=y
diff --git a/target/openrisc/cpu.h b/target/openrisc/cpu.h
index b9584f10d4..1d5efa5ca2 100644
--- a/target/openrisc/cpu.h
+++ b/target/openrisc/cpu.h
@@ -25,6 +25,8 @@
 #include "hw/core/cpu.h"
 #include "qom/object.h"
 
+#define TCG_GUEST_DEFAULT_MO (0)
+
 #define TYPE_OPENRISC_CPU "or1k-cpu"
 
 OBJECT_DECLARE_CPU_TYPE(OpenRISCCPU, OpenRISCCPUClass, OPENRISC_CPU)
diff --git a/target/openrisc/sys_helper.c b/target/openrisc/sys_helper.c
index 48674231e7..da88ad9e77 100644
--- a/target/openrisc/sys_helper.c
+++ b/target/openrisc/sys_helper.c
@@ -145,6 +145,7 @@ void HELPER(mtspr)(CPUOpenRISCState *env, target_ulong spr, target_ulong rb)
         break;
     case TO_SPR(10, 0): /* TTMR */
         {
+            qemu_mutex_lock_iothread();
             if ((env->ttmr & TTMR_M) ^ (rb & TTMR_M)) {
                 switch (rb & TTMR_M) {
                 case TIMER_NONE:
@@ -168,14 +169,16 @@ void HELPER(mtspr)(CPUOpenRISCState *env, target_ulong spr, target_ulong rb)
                 env->ttmr = rb & ~TTMR_IP;
                 cs->interrupt_request &= ~CPU_INTERRUPT_TIMER;
             }
-
             cpu_openrisc_timer_update(cpu);
+            qemu_mutex_unlock_iothread();
         }
         break;
 
     case TO_SPR(10, 1): /* TTCR */
+        qemu_mutex_lock_iothread();
         cpu_openrisc_count_set(cpu, rb);
         cpu_openrisc_timer_update(cpu);
+        qemu_mutex_unlock_iothread();
         break;
 #endif
 
@@ -303,7 +306,9 @@ target_ulong HELPER(mfspr)(CPUOpenRISCState *env, target_ulong rd,
         return env->ttmr;
 
     case TO_SPR(10, 1): /* TTCR */
+        qemu_mutex_lock_iothread();
         cpu_openrisc_count_update(cpu);
+        qemu_mutex_unlock_iothread();
         return cpu_openrisc_count_get(cpu);
 #endif
 
-- 
2.37.1

On 7/29/22 16:01, Stafford Horne wrote:
> This patch enables multithread TCG for OpenRISC.  Since the or1k shared
> syncrhonized timer can be updated from each vCPU via helpers we use a
> mutex to synchronize updates.
> 
> Signed-off-by: Stafford Horne<shorne@gmail.com>
> ---
> Since v2:
>   - Removed cpu_openrisc_timer_has_advanced lock optimization, measuring revealed
>     it did not help much.
> 
>   configs/targets/or1k-softmmu.mak | 1 +
>   target/openrisc/cpu.h            | 2 ++
>   target/openrisc/sys_helper.c     | 7 ++++++-
>   3 files changed, 9 insertions(+), 1 deletion(-)

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

r~

On Fri, Jul 29, 2022 at 04:42:54PM -0700, Richard Henderson wrote:
> On 7/29/22 16:01, Stafford Horne wrote:
> > This patch enables multithread TCG for OpenRISC.  Since the or1k shared
> > syncrhonized timer can be updated from each vCPU via helpers we use a
> > mutex to synchronize updates.
> > 
> > Signed-off-by: Stafford Horne<shorne@gmail.com>
> > ---
> > Since v2:
> >   - Removed cpu_openrisc_timer_has_advanced lock optimization, measuring revealed
> >     it did not help much.
> > 
> >   configs/targets/or1k-softmmu.mak | 1 +
> >   target/openrisc/cpu.h            | 2 ++
> >   target/openrisc/sys_helper.c     | 7 ++++++-
> >   3 files changed, 9 insertions(+), 1 deletion(-)
> 
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

Thank you,

I guess this whole series is a bit late for 7.1.0 now.

I will post the PR after 7.1.0 is released and target it for 7.2.0.

-Stafford