[PATCH v1 6/8] linux-user: properly "unrealize" vCPU object

Alex Bennée posted 8 patches 5 years, 9 months ago
[PATCH v1 6/8] linux-user: properly "unrealize" vCPU object
Posted by Alex Bennée 5 years, 9 months ago
We shouldn't be messing around with the CPU list in linux-user save
for the very special case of do_fork(). When threads end we need to
properly follow QOM object lifetime handling and allow the eventual
cpu_common_unrealizefn to both remove the CPU and ensure any clean-up
actions are taken place, for example calling plugin exit hooks.

There is still a race condition to avoid so use the linux-user
specific clone_lock instead of the cpu_list_lock to avoid it.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Cc: Nikolay Igotti <igotti@gmail.com>
---
 linux-user/syscall.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 05f03919ff0..7f6700c54e3 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -7635,30 +7635,33 @@ static abi_long do_syscall1(void *cpu_env, int num, abi_long arg1,
             return -TARGET_ERESTARTSYS;
         }
 
-        cpu_list_lock();
+        pthread_mutex_lock(&clone_lock);
 
         if (CPU_NEXT(first_cpu)) {
-            TaskState *ts;
+            TaskState *ts = cpu->opaque;
 
-            /* Remove the CPU from the list.  */
-            QTAILQ_REMOVE_RCU(&cpus, cpu, node);
+            object_property_set_bool(OBJECT(cpu), false, "realized", NULL);
+            object_unref(OBJECT(cpu));
+            /*
+             * At this point the CPU should be unrealized and removed
+             * from cpu lists. We can clean-up the rest of the thread
+             * data without the lock held.
+             */
 
-            cpu_list_unlock();
+            pthread_mutex_unlock(&clone_lock);
 
-            ts = cpu->opaque;
             if (ts->child_tidptr) {
                 put_user_u32(0, ts->child_tidptr);
                 do_sys_futex(g2h(ts->child_tidptr), FUTEX_WAKE, INT_MAX,
                           NULL, NULL, 0);
             }
             thread_cpu = NULL;
-            object_unref(OBJECT(cpu));
             g_free(ts);
             rcu_unregister_thread();
             pthread_exit(NULL);
         }
 
-        cpu_list_unlock();
+        pthread_mutex_unlock(&clone_lock);
         preexit_cleanup(cpu_env, arg1);
         _exit(arg1);
         return 0; /* avoid warning */
-- 
2.20.1


Re: [PATCH v1 6/8] linux-user: properly "unrealize" vCPU object
Posted by Philippe Mathieu-Daudé 5 years, 8 months ago
+Paolo

On 5/13/20 7:31 PM, Alex Bennée wrote:
> We shouldn't be messing around with the CPU list in linux-user save
> for the very special case of do_fork(). When threads end we need to
> properly follow QOM object lifetime handling and allow the eventual
> cpu_common_unrealizefn to both remove the CPU and ensure any clean-up
> actions are taken place, for example calling plugin exit hooks.
> 
> There is still a race condition to avoid so use the linux-user
> specific clone_lock instead of the cpu_list_lock to avoid it.
> 
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> Cc: Nikolay Igotti <igotti@gmail.com>

I dare to add:
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>

But I'd rather see someone else reviewing this too.

> ---
>   linux-user/syscall.c | 19 +++++++++++--------
>   1 file changed, 11 insertions(+), 8 deletions(-)
> 
> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
> index 05f03919ff0..7f6700c54e3 100644
> --- a/linux-user/syscall.c
> +++ b/linux-user/syscall.c
> @@ -7635,30 +7635,33 @@ static abi_long do_syscall1(void *cpu_env, int num, abi_long arg1,
>               return -TARGET_ERESTARTSYS;
>           }
>   
> -        cpu_list_lock();
> +        pthread_mutex_lock(&clone_lock);
>   
>           if (CPU_NEXT(first_cpu)) {
> -            TaskState *ts;
> +            TaskState *ts = cpu->opaque;
>   
> -            /* Remove the CPU from the list.  */
> -            QTAILQ_REMOVE_RCU(&cpus, cpu, node);
> +            object_property_set_bool(OBJECT(cpu), false, "realized", NULL);
> +            object_unref(OBJECT(cpu));
> +            /*
> +             * At this point the CPU should be unrealized and removed
> +             * from cpu lists. We can clean-up the rest of the thread
> +             * data without the lock held.
> +             */
>   
> -            cpu_list_unlock();
> +            pthread_mutex_unlock(&clone_lock);
>   
> -            ts = cpu->opaque;
>               if (ts->child_tidptr) {
>                   put_user_u32(0, ts->child_tidptr);
>                   do_sys_futex(g2h(ts->child_tidptr), FUTEX_WAKE, INT_MAX,
>                             NULL, NULL, 0);
>               }
>               thread_cpu = NULL;
> -            object_unref(OBJECT(cpu));
>               g_free(ts);
>               rcu_unregister_thread();
>               pthread_exit(NULL);
>           }
>   
> -        cpu_list_unlock();
> +        pthread_mutex_unlock(&clone_lock);
>           preexit_cleanup(cpu_env, arg1);
>           _exit(arg1);
>           return 0; /* avoid warning */
>