[PATCH] tests/unit: remove block layer code from test-nested-aio-poll

Paolo Bonzini posted 1 patch 2 months, 3 weeks ago
tests/unit/meson.build | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
[PATCH] tests/unit: remove block layer code from test-nested-aio-poll
Posted by Paolo Bonzini 2 months, 3 weeks ago
GCC is reporting a NULL pointer dereference when compiling aio_wait_kick()
with LTO.

The issue is that test-nested-aio-poll.c does not call qemu_init_main_loop().
It doesn't _need_ to because it never calls AIO_WAIT_WHILE(), but it seems
that LTO does not do enough dead-code elimination to catch that.

Fortunately aio_wait_kick() is only used in few places, and only in block
layer or system emulation code; and this test only needs the core event loop
functionality.  It does not even need iothreads.   So remove everything that
calls aio_wait_kick(), which is nice for coverage compared to adding the call
to qemu_init_main_loop().

Reported-by: Richard W.M. Jones <rjones@redhat.com>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2434
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 tests/unit/meson.build | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/unit/meson.build b/tests/unit/meson.build
index 490ab8182dc..6c377917636 100644
--- a/tests/unit/meson.build
+++ b/tests/unit/meson.build
@@ -115,7 +115,7 @@ if have_block
   if host_os != 'windows'
     tests += {
       'test-image-locking': [testblock],
-      'test-nested-aio-poll': [testblock],
+      'test-nested-aio-poll': [],
     }
   endif
   if config_host_data.get('CONFIG_REPLICATION')
-- 
2.46.0
Re: [PATCH] tests/unit: remove block layer code from test-nested-aio-poll
Posted by Richard W.M. Jones 2 months, 3 weeks ago
On Mon, Sep 02, 2024 at 02:03:43AM +0200, Paolo Bonzini wrote:
> GCC is reporting a NULL pointer dereference when compiling aio_wait_kick()
> with LTO.
> 
> The issue is that test-nested-aio-poll.c does not call qemu_init_main_loop().
> It doesn't _need_ to because it never calls AIO_WAIT_WHILE(), but it seems
> that LTO does not do enough dead-code elimination to catch that.
> 
> Fortunately aio_wait_kick() is only used in few places, and only in block
> layer or system emulation code; and this test only needs the core event loop
> functionality.  It does not even need iothreads.   So remove everything that
> calls aio_wait_kick(), which is nice for coverage compared to adding the call
> to qemu_init_main_loop().
> 
> Reported-by: Richard W.M. Jones <rjones@redhat.com>
> Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2434
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  tests/unit/meson.build | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tests/unit/meson.build b/tests/unit/meson.build
> index 490ab8182dc..6c377917636 100644
> --- a/tests/unit/meson.build
> +++ b/tests/unit/meson.build
> @@ -115,7 +115,7 @@ if have_block
>    if host_os != 'windows'
>      tests += {
>        'test-image-locking': [testblock],
> -      'test-nested-aio-poll': [testblock],
> +      'test-nested-aio-poll': [],
>      }
>    endif
>    if config_host_data.get('CONFIG_REPLICATION')

I tested this patch & it fixes the build problem here, thanks!

Tested-by: Richard W.M. Jones <rjones@redhat.com>

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
Fedora Windows cross-compiler. Compile Windows programs, test, and
build Windows installers. Over 100 libraries supported.
http://fedoraproject.org/wiki/MinGW