diff mbox

R: qemu-nbd segmentation fault

Message ID 20130919100558.GD22814@stefanha-thinkpad.redhat.com
State New
Headers show

Commit Message

Stefan Hajnoczi Sept. 19, 2013, 10:05 a.m. UTC
On Wed, Sep 18, 2013 at 03:35:12PM +0200, Mario DE CHENNO wrote:
> This time crashed just issuing a mount command:
> 
> Starting program: /vmstore/vmtools/qemu-nbd-from1.6.0 /vmstore/playground/archivioweb.img
> [Thread debugging using libthread_db enabled]
> [New Thread 0x7ffff55e3700 (LWP 7626)]
> [New Thread 0x7ffff47dd700 (LWP 7645)]
> [New Thread 0x7ffff3fdc700 (LWP 7646)]
> [New Thread 0x7ffff37db700 (LWP 7647)]
> 
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 0x7ffff37db700 (LWP 7647)]
> 0x00007ffff7717249 in ?? () from /usr/lib64/libglib-2.0.so.0
> (gdb) bt
> #0  0x00007ffff7717249 in ?? () from /usr/lib64/libglib-2.0.so.0
> #1  0x00007ffff771759c in ?? () from /usr/lib64/libglib-2.0.so.0
> #2  0x00007ffff7718958 in g_slice_free1 () from /usr/lib64/libglib-2.0.so.0
> #3  0x0000555555591c2d in aio_worker (arg=0x555555c2f6c0) at block/raw-posix.c:776
> #4  0x00005555555da094 in worker_thread (opaque=0x555555c31a40) at thread-pool.c:109
> #5  0x00007ffff6c2ed6b in start_thread () from /lib64/libpthread.so.0
> #6  0x00007ffff6966abd in clone () from /lib64/libc.so.6

Please try this patch and let us know if it worked.

Older glib versions required the application to call g_thread_init().
Failure to do so results in the single-threaded code path being used.
In a multi-threaded application that means race conditions and I've seen
crashes similar to the backtrace you've posted.

Comments

Stefan Hajnoczi Oct. 2, 2013, 8:32 a.m. UTC | #1
On Thu, Sep 19, 2013 at 12:05 PM, Stefan Hajnoczi <stefanha@gmail.com> wrote:
> On Wed, Sep 18, 2013 at 03:35:12PM +0200, Mario DE CHENNO wrote:
>> This time crashed just issuing a mount command:
>>
>> Starting program: /vmstore/vmtools/qemu-nbd-from1.6.0 /vmstore/playground/archivioweb.img
>> [Thread debugging using libthread_db enabled]
>> [New Thread 0x7ffff55e3700 (LWP 7626)]
>> [New Thread 0x7ffff47dd700 (LWP 7645)]
>> [New Thread 0x7ffff3fdc700 (LWP 7646)]
>> [New Thread 0x7ffff37db700 (LWP 7647)]
>>
>> Program received signal SIGSEGV, Segmentation fault.
>> [Switching to Thread 0x7ffff37db700 (LWP 7647)]
>> 0x00007ffff7717249 in ?? () from /usr/lib64/libglib-2.0.so.0
>> (gdb) bt
>> #0  0x00007ffff7717249 in ?? () from /usr/lib64/libglib-2.0.so.0
>> #1  0x00007ffff771759c in ?? () from /usr/lib64/libglib-2.0.so.0
>> #2  0x00007ffff7718958 in g_slice_free1 () from /usr/lib64/libglib-2.0.so.0
>> #3  0x0000555555591c2d in aio_worker (arg=0x555555c2f6c0) at block/raw-posix.c:776
>> #4  0x00005555555da094 in worker_thread (opaque=0x555555c31a40) at thread-pool.c:109
>> #5  0x00007ffff6c2ed6b in start_thread () from /lib64/libpthread.so.0
>> #6  0x00007ffff6966abd in clone () from /lib64/libc.so.6
>
> Please try this patch and let us know if it worked.
>
> Older glib versions required the application to call g_thread_init().
> Failure to do so results in the single-threaded code path being used.
> In a multi-threaded application that means race conditions and I've seen
> crashes similar to the backtrace you've posted.
>
> diff --git a/qemu-nbd.c b/qemu-nbd.c
> index c26c98e..8ae3868 100644
> --- a/qemu-nbd.c
> +++ b/qemu-nbd.c
> @@ -357,6 +357,15 @@ int main(int argc, char **argv)
>      const char *fmt = NULL;
>      Error *local_err = NULL;
>
> +    if (!g_thread_supported()) {
> +#if !GLIB_CHECK_VERSION(2, 31, 0)
> +        g_thread_init(NULL);
> +#else
> +        fprintf(stderr, "glib threading failed to initialize.\n");
> +        exit(1);
> +#endif
> +    }
> +
>      /* The client thread uses SIGTERM to interrupt the server.  A signal
>       * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
>       */

Hi Mario,
Have you had time to try out this patch?

Stefan
diff mbox

Patch

diff --git a/qemu-nbd.c b/qemu-nbd.c
index c26c98e..8ae3868 100644
--- a/qemu-nbd.c
+++ b/qemu-nbd.c
@@ -357,6 +357,15 @@  int main(int argc, char **argv)
     const char *fmt = NULL;
     Error *local_err = NULL;
 
+    if (!g_thread_supported()) {
+#if !GLIB_CHECK_VERSION(2, 31, 0)
+        g_thread_init(NULL);
+#else
+        fprintf(stderr, "glib threading failed to initialize.\n");
+        exit(1);
+#endif
+    }
+
     /* The client thread uses SIGTERM to interrupt the server.  A signal
      * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
      */