diff mbox

[v7,02/13] exec: Remove cpu from cpus list during cpu_exec_exit()

Message ID 1453960195-15181-3-git-send-email-bharata@linux.vnet.ibm.com (mailing list archive)
State New, archived
Headers show

Commit Message

Bharata B Rao Jan. 28, 2016, 5:49 a.m. UTC
CPUState *cpu gets added to the cpus list during cpu_exec_init(). It
should be removed from cpu_exec_exit().

cpu_exec_init() is called from generic CPU::instance_finalize and some
archs like PowerPC call it from CPU unrealizefn. So ensure that we
dequeue the cpu only once.

Now -1 value for cpu->cpu_index indicates that we have already dequeued
the cpu for CONFIG_USER_ONLY case also.

Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
---
 exec.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

Comments

Eduardo Habkost Jan. 28, 2016, 7:19 p.m. UTC | #1
On Thu, Jan 28, 2016 at 11:19:44AM +0530, Bharata B Rao wrote:
> CPUState *cpu gets added to the cpus list during cpu_exec_init(). It
> should be removed from cpu_exec_exit().
> 
> cpu_exec_init() is called from generic CPU::instance_finalize and some
> archs like PowerPC call it from CPU unrealizefn. So ensure that we
> dequeue the cpu only once.
> 
> Now -1 value for cpu->cpu_index indicates that we have already dequeued
> the cpu for CONFIG_USER_ONLY case also.
> 
> Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
> Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
> ---
>  exec.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/exec.c b/exec.c
> index 7115403..c8da9d4 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -596,6 +596,7 @@ void cpu_exec_exit(CPUState *cpu)
>          return;
>      }
>  
> +    QTAILQ_REMOVE(&cpus, cpu, node);
>      bitmap_clear(cpu_index_map, cpu->cpu_index, 1);
>      cpu->cpu_index = -1;
>  }
> @@ -614,6 +615,15 @@ static int cpu_get_free_index(Error **errp)
>  
>  void cpu_exec_exit(CPUState *cpu)
>  {
> +    cpu_list_lock();
> +    if (cpu->cpu_index == -1) {
> +        cpu_list_unlock();
> +        return;
> +    }
> +
> +    QTAILQ_REMOVE(&cpus, cpu, node);
> +    cpu->cpu_index = -1;
> +    cpu_list_unlock();

With this, the only differences between the two cpu_exec_exit()
implementations are:

* cpu_list_lock()/cpu_list_unlock() functions.
  * We can add !CONFIG_USER_ONLY stubs for them.
* The bitmap_clear() call.
  * It can be abstracted away in a cpu_release_index() function,
    just like we already have a CONFIG_USER_ONLY version of
    cpu_get_free_index().
    * I was going to suggest using cpu_index_map on
      CONFIG_USER_ONLY too, but I assume we don't want to limit
      the number of threads in *-user to MAX_CPUMASK_BITS.

This way we won't need to duplicate cpu_exec_exit() code.
Bharata B Rao Jan. 29, 2016, 6:14 a.m. UTC | #2
On Thu, Jan 28, 2016 at 05:19:33PM -0200, Eduardo Habkost wrote:
> On Thu, Jan 28, 2016 at 11:19:44AM +0530, Bharata B Rao wrote:
> > CPUState *cpu gets added to the cpus list during cpu_exec_init(). It
> > should be removed from cpu_exec_exit().
> > 
> > cpu_exec_init() is called from generic CPU::instance_finalize and some
> > archs like PowerPC call it from CPU unrealizefn. So ensure that we
> > dequeue the cpu only once.
> > 
> > Now -1 value for cpu->cpu_index indicates that we have already dequeued
> > the cpu for CONFIG_USER_ONLY case also.
> > 
> > Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
> > Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
> > ---
> >  exec.c | 10 ++++++++++
> >  1 file changed, 10 insertions(+)
> > 
> > diff --git a/exec.c b/exec.c
> > index 7115403..c8da9d4 100644
> > --- a/exec.c
> > +++ b/exec.c
> > @@ -596,6 +596,7 @@ void cpu_exec_exit(CPUState *cpu)
> >          return;
> >      }
> >  
> > +    QTAILQ_REMOVE(&cpus, cpu, node);
> >      bitmap_clear(cpu_index_map, cpu->cpu_index, 1);
> >      cpu->cpu_index = -1;
> >  }
> > @@ -614,6 +615,15 @@ static int cpu_get_free_index(Error **errp)
> >  
> >  void cpu_exec_exit(CPUState *cpu)
> >  {
> > +    cpu_list_lock();
> > +    if (cpu->cpu_index == -1) {
> > +        cpu_list_unlock();
> > +        return;
> > +    }
> > +
> > +    QTAILQ_REMOVE(&cpus, cpu, node);
> > +    cpu->cpu_index = -1;
> > +    cpu_list_unlock();
> 
> With this, the only differences between the two cpu_exec_exit()
> implementations are:
> 
> * cpu_list_lock()/cpu_list_unlock() functions.
>   * We can add !CONFIG_USER_ONLY stubs for them.
> * The bitmap_clear() call.
>   * It can be abstracted away in a cpu_release_index() function,
>     just like we already have a CONFIG_USER_ONLY version of
>     cpu_get_free_index().

Ok, made those changes so that cpu_exec_exit() will be a common routine
with some CONFIG_USER_ONLY ifdefs in between.

Regards,
Bharata.
diff mbox

Patch

diff --git a/exec.c b/exec.c
index 7115403..c8da9d4 100644
--- a/exec.c
+++ b/exec.c
@@ -596,6 +596,7 @@  void cpu_exec_exit(CPUState *cpu)
         return;
     }
 
+    QTAILQ_REMOVE(&cpus, cpu, node);
     bitmap_clear(cpu_index_map, cpu->cpu_index, 1);
     cpu->cpu_index = -1;
 }
@@ -614,6 +615,15 @@  static int cpu_get_free_index(Error **errp)
 
 void cpu_exec_exit(CPUState *cpu)
 {
+    cpu_list_lock();
+    if (cpu->cpu_index == -1) {
+        cpu_list_unlock();
+        return;
+    }
+
+    QTAILQ_REMOVE(&cpus, cpu, node);
+    cpu->cpu_index = -1;
+    cpu_list_unlock();
 }
 #endif