[Qemu-devel] small memory leak due to MachineClass rework

[Qemu-devel] small memory leak due to MachineClass rework

[Christian Borntraeger]

Marcel,

after 

commit 261747f176f6 (vl: Use MachineClass instead of global QEMUMachine list) valgrind complains about the following:

==54082== 57 bytes in 3 blocks are definitely lost in loss record 365 of 729
==54082==    at 0x4031AFE: malloc (vg_replace_malloc.c:292)
==54082==    by 0x4145569: g_malloc (in /usr/lib64/libglib-2.0.so.0.3400.2)
==54082==    by 0x415F9E9: g_strconcat (in /usr/lib64/libglib-2.0.so.0.3400.2)
==54082==    by 0x80157FE7: qemu_register_machine (vl.c:1597)
==54082==    by 0x80208E6B: module_call_init (module.c:105)
==54082==    by 0x80013B91: main (vl.c:3000)

Turns out that valgrind is right. We simply forget the memory that g_strconcat has allocated.
This fixes the small leak, but I have to cast away the constness of .name.
Any better ideas?


diff --git a/vl.c b/vl.c
index 03488b9..b06d186 100644
--- a/vl.c
+++ b/vl.c
@@ -1594,13+1594,14 @@ static void machine_class_init(ObjectClass *oc, void *data)
 int qemu_register_machine(QEMUMachine *m)
 {
     TypeInfo ti = {
         .name       = g_strconcat(m->name, TYPE_MACHINE_SUFFIX, NULL),
         .parent     = TYPE_MACHINE,
         .class_init = machine_class_init,
         .class_data = (void *)m,
     };
 
     type_register(&ti);
+    g_free((gpointer) ti.name);
 
     return 0;
 }

Re: [Qemu-devel] small memory leak due to MachineClass rework

[Peter Maydell]

On 17 March 2014 20:56, Christian Borntraeger <borntraeger@de.ibm.com> wrote:
> Turns out that valgrind is right. We simply forget the memory that g_strconcat has allocated.
> This fixes the small leak, but I have to cast away the constness of .name.
> Any better ideas?

It's how cpu_register() in target-arm/cpu.c does the same thing
(though we use void* rather than gpointer as the cast).
If you really dislike the const I guess you could use:
    char *name = g_strconcat(...);
    TypeInfo ti = {
          .name = name,
          [...]
    };
    [...]
    g_free(name);

thanks
-- PMM

Re: [Qemu-devel] small memory leak due to MachineClass rework

[Christian Borntraeger]

On 17/03/14 22:25, Peter Maydell wrote:
> On 17 March 2014 20:56, Christian Borntraeger <borntraeger@de.ibm.com> wrote:
>> Turns out that valgrind is right. We simply forget the memory that g_strconcat has allocated.
>> This fixes the small leak, but I have to cast away the constness of .name.
>> Any better ideas?
> 
> It's how cpu_register() in target-arm/cpu.c does the same thing
> (though we use void* rather than gpointer as the cast).

Ok, if you dont have a problem with that approach, I can submit a proper
patch with signoff.  Any preference regarding void * vs gpointer?

Christian


> If you really dislike the const I guess you could use:
>     char *name = g_strconcat(...);
>     TypeInfo ti = {
>           .name = name,
>           [...]
>     };
>     [...]
>     g_free(name);
> 

Re: [Qemu-devel] small memory leak due to MachineClass rework

[Andreas Färber]

Am 17.03.2014 22:29, schrieb Christian Borntraeger:
> On 17/03/14 22:25, Peter Maydell wrote:
>> On 17 March 2014 20:56, Christian Borntraeger <borntraeger@de.ibm.com> wrote:
>>> Turns out that valgrind is right. We simply forget the memory that g_strconcat has allocated.
>>> This fixes the small leak, but I have to cast away the constness of .name.
>>> Any better ideas?
>>
>> It's how cpu_register() in target-arm/cpu.c does the same thing
>> (though we use void* rather than gpointer as the cast).
> 
> Ok, if you dont have a problem with that approach, I can submit a proper
> patch with signoff.  Any preference regarding void * vs gpointer?

void *. We try to avoid g* types where possible/sensible.

Thanks for catching this!

Cheers,
Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg

Re: [Qemu-devel] small memory leak due to MachineClass rework

[Marcel Apfelbaum]

On Mon, 2014-03-17 at 22:29 +0100, Christian Borntraeger wrote:
> On 17/03/14 22:25, Peter Maydell wrote:
> > On 17 March 2014 20:56, Christian Borntraeger <borntraeger@de.ibm.com> wrote:
> >> Turns out that valgrind is right. We simply forget the memory that g_strconcat has allocated.
> >> This fixes the small leak, but I have to cast away the constness of .name.
> >> Any better ideas?
> > 
> > It's how cpu_register() in target-arm/cpu.c does the same thing
> > (though we use void* rather than gpointer as the cast).
> 
> Ok, if you dont have a problem with that approach, I can submit a proper
> patch with signoff.  Any preference regarding void * vs gpointer?
Hi Cristian,

Thank you for catching this! (I also would go with void *)
Marcel

> 
> Christian
> 
> 
> > If you really dislike the const I guess you could use:
> >     char *name = g_strconcat(...);
> >     TypeInfo ti = {
> >           .name = name,
> >           [...]
> >     };
> >     [...]
> >     g_free(name);
> > 
> 
> 

Re: [Qemu-devel] small memory leak due to MachineClass rework

[Michael S. Tsirkin]

On Mon, Mar 17, 2014 at 09:56:45PM +0100, Christian Borntraeger wrote:
> Marcel,
> 
> after 
> 
> commit 261747f176f6 (vl: Use MachineClass instead of global QEMUMachine list) valgrind complains about the following:
> 
> ==54082== 57 bytes in 3 blocks are definitely lost in loss record 365 of 729
> ==54082==    at 0x4031AFE: malloc (vg_replace_malloc.c:292)
> ==54082==    by 0x4145569: g_malloc (in /usr/lib64/libglib-2.0.so.0.3400.2)
> ==54082==    by 0x415F9E9: g_strconcat (in /usr/lib64/libglib-2.0.so.0.3400.2)
> ==54082==    by 0x80157FE7: qemu_register_machine (vl.c:1597)
> ==54082==    by 0x80208E6B: module_call_init (module.c:105)
> ==54082==    by 0x80013B91: main (vl.c:3000)
> 
> Turns out that valgrind is right. We simply forget the memory that g_strconcat has allocated.
> This fixes the small leak, but I have to cast away the constness of .name.
> Any better ideas?
> 
> 
> diff --git a/vl.c b/vl.c
> index 03488b9..b06d186 100644
> --- a/vl.c
> +++ b/vl.c
> @@ -1594,13+1594,14 @@ static void machine_class_init(ObjectClass *oc, void *data)
>  int qemu_register_machine(QEMUMachine *m)
>  {
>      TypeInfo ti = {
>          .name       = g_strconcat(m->name, TYPE_MACHINE_SUFFIX, NULL),
>          .parent     = TYPE_MACHINE,
>          .class_init = machine_class_init,
>          .class_data = (void *)m,
>      };
>  
>      type_register(&ti);
> +    g_free((gpointer) ti.name);
>  
>      return 0;
>  }


How about:

 int qemu_register_machine(QEMUMachine *m)
 {
+  char *name = g_strconcat(m->name, TYPE_MACHINE_SUFFIX, NULL);
   TypeInfo ti = {
-        .name       = g_strconcat(m->name, TYPE_MACHINE_SUFFIX, NULL),
+        .name       = name,
         .parent     = TYPE_MACHINE,
         .class_init = machine_class_init,
         .class_data = (void *)m,
     };
 
     type_register(&ti);
+    g_free(name);
 
     return 0;
 }