[Qemu-devel] [PATCH 0/2] dataplane: fix bogus coverity warnings

[Qemu-devel] [PATCH 0/2] dataplane: fix bogus coverity warnings

[Stefan Hajnoczi]

Coverity detects when variable are accessed under a mutex most of the time.  It
warns when they are not accessed under the mutex.  I initialized variables
before the mutex and threads that access them even exist - Coverity doesn't
like that.  Fix the code.

Stefan Hajnoczi (2):
  iothread: fix bogus coverity warning
  rfifolock: fix bogus coverity warning

 iothread.c       | 5 ++++-
 util/rfifolock.c | 4 +++-
 2 files changed, 7 insertions(+), 2 deletions(-)

-- 
1.8.5.3

[Qemu-devel] [PATCH 1/2] iothread: fix bogus coverity warning

[Stefan Hajnoczi]

Coverity warns about initializing variables that will later be accessed
under a mutex.  There is no problem with the code itself but let's avoid
accumulating Coverity warnings.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 iothread.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/iothread.c b/iothread.c
index cb5986b..9e25af9 100644
--- a/iothread.c
+++ b/iothread.c
@@ -75,11 +75,14 @@ static void iothread_complete(UserCreatable *obj, Error **errp)
 
     iothread->stopping = false;
     iothread->ctx = aio_context_new();
-    iothread->thread_id = -1;
 
     qemu_mutex_init(&iothread->init_done_lock);
     qemu_cond_init(&iothread->init_done_cond);
 
+    qemu_mutex_lock(&iothread->init_done_lock);
+    iothread->thread_id = -1;
+    qemu_mutex_unlock(&iothread->init_done_lock);
+
     /* This assumes we are called from a thread with useful CPU affinity for us
      * to inherit.
      */
-- 
1.8.5.3

[Qemu-devel] [PATCH 2/2] rfifolock: fix bogus coverity warning

[Stefan Hajnoczi]

Coverity warns about initializing variables that will later be accessed
under a mutex.  There is no problem with the code itself but let's avoid
accumulating Coverity warnings.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 util/rfifolock.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/util/rfifolock.c b/util/rfifolock.c
index afbf748..f42214d 100644
--- a/util/rfifolock.c
+++ b/util/rfifolock.c
@@ -17,12 +17,14 @@
 void rfifolock_init(RFifoLock *r, void (*cb)(void *), void *opaque)
 {
     qemu_mutex_init(&r->lock);
+    qemu_cond_init(&r->cond);
+    qemu_mutex_lock(&r->lock);
     r->head = 0;
     r->tail = 0;
-    qemu_cond_init(&r->cond);
     r->nesting = 0;
     r->cb = cb;
     r->cb_opaque = opaque;
+    qemu_mutex_unlock(&r->lock);
 }
 
 void rfifolock_destroy(RFifoLock *r)
-- 
1.8.5.3

Re: [Qemu-devel] [PATCH 0/2] dataplane: fix bogus coverity warnings

[Paolo Bonzini]

Il 14/03/2014 16:14, Stefan Hajnoczi ha scritto:
> Coverity detects when variable are accessed under a mutex most of the time.  It
> warns when they are not accessed under the mutex.  I initialized variables
> before the mutex and threads that access them even exist - Coverity doesn't
> like that.  Fix the code.
>
> Stefan Hajnoczi (2):
>   iothread: fix bogus coverity warning
>   rfifolock: fix bogus coverity warning
>
>  iothread.c       | 5 ++++-
>  util/rfifolock.c | 4 +++-
>  2 files changed, 7 insertions(+), 2 deletions(-)
>

Nah, I think Coverity is wrong.  It should detect initialization of the 
mutex, and treat surrounding code as single-threaded.  I silenced the 
defects in the report, like others before.

Paolo

Re: [Qemu-devel] [PATCH 0/2] dataplane: fix bogus coverity warnings

[Stefan Hajnoczi]

On Fri, Mar 14, 2014 at 06:25:34PM +0100, Paolo Bonzini wrote:
> Il 14/03/2014 16:14, Stefan Hajnoczi ha scritto:
> >Coverity detects when variable are accessed under a mutex most of the time.  It
> >warns when they are not accessed under the mutex.  I initialized variables
> >before the mutex and threads that access them even exist - Coverity doesn't
> >like that.  Fix the code.
> >
> >Stefan Hajnoczi (2):
> >  iothread: fix bogus coverity warning
> >  rfifolock: fix bogus coverity warning
> >
> > iothread.c       | 5 ++++-
> > util/rfifolock.c | 4 +++-
> > 2 files changed, 7 insertions(+), 2 deletions(-)
> >
> 
> Nah, I think Coverity is wrong.  It should detect initialization of
> the mutex, and treat surrounding code as single-threaded.  I
> silenced the defects in the report, like others before.

Okay, let's drop these patches.

Stefan