If there are more than one events, wait_until_completed() might return
the 2nd event even if the 1st event is JOB_COMPLETED, since the for loop
will continue to run even if completed is set to True.

It never happened before, but it can be triggered when OOB is enabled
due to the RESUME startup message. Fix that up by removing the boolean
and make sure we return the correct event.

Signed-off-by: Peter Xu <peterx@redhat.com>
---
 tests/qemu-iotests/iotests.py | 20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)

diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py
index b5d7945af8..11704e6583 100644
--- a/tests/qemu-iotests/iotests.py
+++ b/tests/qemu-iotests/iotests.py
@@ -470,18 +470,14 @@ class QMPTestCase(unittest.TestCase):
 
     def wait_until_completed(self, drive='drive0', check_offset=True):
         '''Wait for a block job to finish, returning the event'''
-        completed = False
-        while not completed:
-            for event in self.vm.get_qmp_events(wait=True):
-                if event['event'] == 'BLOCK_JOB_COMPLETED':
-                    self.assert_qmp(event, 'data/device', drive)
-                    self.assert_qmp_absent(event, 'data/error')
-                    if check_offset:
-                        self.assert_qmp(event, 'data/offset', event['data']['len'])
-                    completed = True
-
-        self.assert_no_active_block_jobs()
-        return event
+        for event in self.vm.get_qmp_events(wait=True):
+            if event['event'] == 'BLOCK_JOB_COMPLETED':
+                self.assert_qmp(event, 'data/device', drive)
+                self.assert_qmp_absent(event, 'data/error')
+                if check_offset:
+                    self.assert_qmp(event, 'data/offset', event['data']['len'])
+                self.assert_no_active_block_jobs()
+                return event
 
     def wait_ready(self, drive='drive0'):
         '''Wait until a block job BLOCK_JOB_READY event'''
-- 
2.14.3

Am 26.03.2018 um 08:11 hat Peter Xu geschrieben:
> If there are more than one events, wait_until_completed() might return
> the 2nd event even if the 1st event is JOB_COMPLETED, since the for loop
> will continue to run even if completed is set to True.
> 
> It never happened before, but it can be triggered when OOB is enabled
> due to the RESUME startup message. Fix that up by removing the boolean
> and make sure we return the correct event.
> 
> Signed-off-by: Peter Xu <peterx@redhat.com>
> ---
>  tests/qemu-iotests/iotests.py | 20 ++++++++------------
>  1 file changed, 8 insertions(+), 12 deletions(-)
> 
> diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py
> index b5d7945af8..11704e6583 100644
> --- a/tests/qemu-iotests/iotests.py
> +++ b/tests/qemu-iotests/iotests.py
> @@ -470,18 +470,14 @@ class QMPTestCase(unittest.TestCase):
>  
>      def wait_until_completed(self, drive='drive0', check_offset=True):
>          '''Wait for a block job to finish, returning the event'''
> -        completed = False
> -        while not completed:
> -            for event in self.vm.get_qmp_events(wait=True):
> -                if event['event'] == 'BLOCK_JOB_COMPLETED':
> -                    self.assert_qmp(event, 'data/device', drive)
> -                    self.assert_qmp_absent(event, 'data/error')
> -                    if check_offset:
> -                        self.assert_qmp(event, 'data/offset', event['data']['len'])
> -                    completed = True
> -
> -        self.assert_no_active_block_jobs()
> -        return event
> +        for event in self.vm.get_qmp_events(wait=True):
> +            if event['event'] == 'BLOCK_JOB_COMPLETED':
> +                self.assert_qmp(event, 'data/device', drive)
> +                self.assert_qmp_absent(event, 'data/error')
> +                if check_offset:
> +                    self.assert_qmp(event, 'data/offset', event['data']['len'])
> +                self.assert_no_active_block_jobs()
> +                return event
>  
>      def wait_ready(self, drive='drive0'):
>          '''Wait until a block job BLOCK_JOB_READY event'''

If an event is pending, but it's not the expected event, won't we return
None now instead of waiting for the BLOCK_JOB_COMPLETED event?

Wouldn't it be much easier to just add a 'break'?

Kevin

On Mon, Mar 26, 2018 at 12:47:39PM +0200, Kevin Wolf wrote:
> Am 26.03.2018 um 08:11 hat Peter Xu geschrieben:
> > If there are more than one events, wait_until_completed() might return
> > the 2nd event even if the 1st event is JOB_COMPLETED, since the for loop
> > will continue to run even if completed is set to True.
> > 
> > It never happened before, but it can be triggered when OOB is enabled
> > due to the RESUME startup message. Fix that up by removing the boolean
> > and make sure we return the correct event.
> > 
> > Signed-off-by: Peter Xu <peterx@redhat.com>
> > ---
> >  tests/qemu-iotests/iotests.py | 20 ++++++++------------
> >  1 file changed, 8 insertions(+), 12 deletions(-)
> > 
> > diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py
> > index b5d7945af8..11704e6583 100644
> > --- a/tests/qemu-iotests/iotests.py
> > +++ b/tests/qemu-iotests/iotests.py
> > @@ -470,18 +470,14 @@ class QMPTestCase(unittest.TestCase):
> >  
> >      def wait_until_completed(self, drive='drive0', check_offset=True):
> >          '''Wait for a block job to finish, returning the event'''
> > -        completed = False
> > -        while not completed:
> > -            for event in self.vm.get_qmp_events(wait=True):
> > -                if event['event'] == 'BLOCK_JOB_COMPLETED':
> > -                    self.assert_qmp(event, 'data/device', drive)
> > -                    self.assert_qmp_absent(event, 'data/error')
> > -                    if check_offset:
> > -                        self.assert_qmp(event, 'data/offset', event['data']['len'])
> > -                    completed = True
> > -
> > -        self.assert_no_active_block_jobs()
> > -        return event
> > +        for event in self.vm.get_qmp_events(wait=True):
> > +            if event['event'] == 'BLOCK_JOB_COMPLETED':
> > +                self.assert_qmp(event, 'data/device', drive)
> > +                self.assert_qmp_absent(event, 'data/error')
> > +                if check_offset:
> > +                    self.assert_qmp(event, 'data/offset', event['data']['len'])
> > +                self.assert_no_active_block_jobs()
> > +                return event
> >  
> >      def wait_ready(self, drive='drive0'):
> >          '''Wait until a block job BLOCK_JOB_READY event'''
> 
> If an event is pending, but it's not the expected event, won't we return
> None now instead of waiting for the BLOCK_JOB_COMPLETED event?

If so, we'll return none.  The patch fixes the other case when there
are two events: one JOB_COMPLETED plus another (e.g., RESUME) event.
When that happens, logically we should return one JOB_COMPLETED event,
but the old code will return the other event (e.g., RESUME).

> 
> Wouldn't it be much easier to just add a 'break'?

Yes, it's the same.  But IMHO those logics (e.g., the completed
variable) are not really needed at all.  This one is simpler.

Or do you want me to post the oneliner that you prefer?

Thanks,

-- 
Peter Xu

On Tue, Mar 27, 2018 at 10:21:55AM +0800, Peter Xu wrote:
> On Mon, Mar 26, 2018 at 12:47:39PM +0200, Kevin Wolf wrote:
> > Am 26.03.2018 um 08:11 hat Peter Xu geschrieben:
> > > If there are more than one events, wait_until_completed() might return
> > > the 2nd event even if the 1st event is JOB_COMPLETED, since the for loop
> > > will continue to run even if completed is set to True.
> > > 
> > > It never happened before, but it can be triggered when OOB is enabled
> > > due to the RESUME startup message. Fix that up by removing the boolean
> > > and make sure we return the correct event.
> > > 
> > > Signed-off-by: Peter Xu <peterx@redhat.com>
> > > ---
> > >  tests/qemu-iotests/iotests.py | 20 ++++++++------------
> > >  1 file changed, 8 insertions(+), 12 deletions(-)
> > > 
> > > diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py
> > > index b5d7945af8..11704e6583 100644
> > > --- a/tests/qemu-iotests/iotests.py
> > > +++ b/tests/qemu-iotests/iotests.py
> > > @@ -470,18 +470,14 @@ class QMPTestCase(unittest.TestCase):
> > >  
> > >      def wait_until_completed(self, drive='drive0', check_offset=True):
> > >          '''Wait for a block job to finish, returning the event'''
> > > -        completed = False
> > > -        while not completed:
> > > -            for event in self.vm.get_qmp_events(wait=True):
> > > -                if event['event'] == 'BLOCK_JOB_COMPLETED':
> > > -                    self.assert_qmp(event, 'data/device', drive)
> > > -                    self.assert_qmp_absent(event, 'data/error')
> > > -                    if check_offset:
> > > -                        self.assert_qmp(event, 'data/offset', event['data']['len'])
> > > -                    completed = True
> > > -
> > > -        self.assert_no_active_block_jobs()
> > > -        return event
> > > +        for event in self.vm.get_qmp_events(wait=True):
> > > +            if event['event'] == 'BLOCK_JOB_COMPLETED':
> > > +                self.assert_qmp(event, 'data/device', drive)
> > > +                self.assert_qmp_absent(event, 'data/error')
> > > +                if check_offset:
> > > +                    self.assert_qmp(event, 'data/offset', event['data']['len'])
> > > +                self.assert_no_active_block_jobs()
> > > +                return event
> > >  
> > >      def wait_ready(self, drive='drive0'):
> > >          '''Wait until a block job BLOCK_JOB_READY event'''
> > 
> > If an event is pending, but it's not the expected event, won't we return
> > None now instead of waiting for the BLOCK_JOB_COMPLETED event?
> 
> If so, we'll return none.

Kevin is pointing out that this patch is broken.  Previously the
function waited for BLOCK_JOB_COMPLETED, even when other events were
pending when we entered the function.  Now it returns None and does not
wait for BLOCK_JOB_COMPLETED!

> The patch fixes the other case when there
> are two events: one JOB_COMPLETED plus another (e.g., RESUME) event.
> When that happens, logically we should return one JOB_COMPLETED event,
> but the old code will return the other event (e.g., RESUME).
> 
> > 
> > Wouldn't it be much easier to just add a 'break'?
> 
> Yes, it's the same.  But IMHO those logics (e.g., the completed
> variable) are not really needed at all.  This one is simpler.

No, the outer loop is needed so that the function waits until
BLOCK_JOB_COMPLETED is received.  It's not possible to do it with a
single for loop.

Stefan

On Tue, Apr 03, 2018 at 01:59:18PM +0100, Stefan Hajnoczi wrote:
> On Tue, Mar 27, 2018 at 10:21:55AM +0800, Peter Xu wrote:
> > On Mon, Mar 26, 2018 at 12:47:39PM +0200, Kevin Wolf wrote:
> > > Am 26.03.2018 um 08:11 hat Peter Xu geschrieben:
> > > > If there are more than one events, wait_until_completed() might return
> > > > the 2nd event even if the 1st event is JOB_COMPLETED, since the for loop
> > > > will continue to run even if completed is set to True.
> > > > 
> > > > It never happened before, but it can be triggered when OOB is enabled
> > > > due to the RESUME startup message. Fix that up by removing the boolean
> > > > and make sure we return the correct event.
> > > > 
> > > > Signed-off-by: Peter Xu <peterx@redhat.com>
> > > > ---
> > > >  tests/qemu-iotests/iotests.py | 20 ++++++++------------
> > > >  1 file changed, 8 insertions(+), 12 deletions(-)
> > > > 
> > > > diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py
> > > > index b5d7945af8..11704e6583 100644
> > > > --- a/tests/qemu-iotests/iotests.py
> > > > +++ b/tests/qemu-iotests/iotests.py
> > > > @@ -470,18 +470,14 @@ class QMPTestCase(unittest.TestCase):
> > > >  
> > > >      def wait_until_completed(self, drive='drive0', check_offset=True):
> > > >          '''Wait for a block job to finish, returning the event'''
> > > > -        completed = False
> > > > -        while not completed:
> > > > -            for event in self.vm.get_qmp_events(wait=True):
> > > > -                if event['event'] == 'BLOCK_JOB_COMPLETED':
> > > > -                    self.assert_qmp(event, 'data/device', drive)
> > > > -                    self.assert_qmp_absent(event, 'data/error')
> > > > -                    if check_offset:
> > > > -                        self.assert_qmp(event, 'data/offset', event['data']['len'])
> > > > -                    completed = True
> > > > -
> > > > -        self.assert_no_active_block_jobs()
> > > > -        return event
> > > > +        for event in self.vm.get_qmp_events(wait=True):
> > > > +            if event['event'] == 'BLOCK_JOB_COMPLETED':
> > > > +                self.assert_qmp(event, 'data/device', drive)
> > > > +                self.assert_qmp_absent(event, 'data/error')
> > > > +                if check_offset:
> > > > +                    self.assert_qmp(event, 'data/offset', event['data']['len'])
> > > > +                self.assert_no_active_block_jobs()
> > > > +                return event
> > > >  
> > > >      def wait_ready(self, drive='drive0'):
> > > >          '''Wait until a block job BLOCK_JOB_READY event'''
> > > 
> > > If an event is pending, but it's not the expected event, won't we return
> > > None now instead of waiting for the BLOCK_JOB_COMPLETED event?
> > 
> > If so, we'll return none.
> 
> Kevin is pointing out that this patch is broken.  Previously the
> function waited for BLOCK_JOB_COMPLETED, even when other events were
> pending when we entered the function.  Now it returns None and does not
> wait for BLOCK_JOB_COMPLETED!

Aha!  Surely I missed that. :)

> 
> > The patch fixes the other case when there
> > are two events: one JOB_COMPLETED plus another (e.g., RESUME) event.
> > When that happens, logically we should return one JOB_COMPLETED event,
> > but the old code will return the other event (e.g., RESUME).
> > 
> > > 
> > > Wouldn't it be much easier to just add a 'break'?
> > 
> > Yes, it's the same.  But IMHO those logics (e.g., the completed
> > variable) are not really needed at all.  This one is simpler.
> 
> No, the outer loop is needed so that the function waits until
> BLOCK_JOB_COMPLETED is received.  It's not possible to do it with a
> single for loop.

Indeed.  But then I would still slightly prefer removing the
"completed" var:

    def wait_until_completed(self, drive='drive0', check_offset=True):
        '''Wait for a block job to finish, returning the event'''
        while True:
            for event in self.vm.get_qmp_events(wait=True):
                if event['event'] == 'BLOCK_JOB_COMPLETED':
                    self.assert_qmp(event, 'data/device', drive)
                    self.assert_qmp_absent(event, 'data/error')
                    if check_offset:
                        self.assert_qmp(event, 'data/offset', event['data']['len'])
                    self.assert_no_active_block_jobs()
                    return event

Or a single break would work too.  Do either of you have any
preference?  I can repost in either way.  Thanks,

-- 
Peter Xu

On Wed, Apr 04, 2018 at 09:42:41AM +0800, Peter Xu wrote:
> On Tue, Apr 03, 2018 at 01:59:18PM +0100, Stefan Hajnoczi wrote:
> > On Tue, Mar 27, 2018 at 10:21:55AM +0800, Peter Xu wrote:
> > > On Mon, Mar 26, 2018 at 12:47:39PM +0200, Kevin Wolf wrote:
> > > > Am 26.03.2018 um 08:11 hat Peter Xu geschrieben:
> > > The patch fixes the other case when there
> > > are two events: one JOB_COMPLETED plus another (e.g., RESUME) event.
> > > When that happens, logically we should return one JOB_COMPLETED event,
> > > but the old code will return the other event (e.g., RESUME).
> > > 
> > > > 
> > > > Wouldn't it be much easier to just add a 'break'?
> > > 
> > > Yes, it's the same.  But IMHO those logics (e.g., the completed
> > > variable) are not really needed at all.  This one is simpler.
> > 
> > No, the outer loop is needed so that the function waits until
> > BLOCK_JOB_COMPLETED is received.  It's not possible to do it with a
> > single for loop.
> 
> Indeed.  But then I would still slightly prefer removing the
> "completed" var:
> 
>     def wait_until_completed(self, drive='drive0', check_offset=True):
>         '''Wait for a block job to finish, returning the event'''
>         while True:
>             for event in self.vm.get_qmp_events(wait=True):
>                 if event['event'] == 'BLOCK_JOB_COMPLETED':
>                     self.assert_qmp(event, 'data/device', drive)
>                     self.assert_qmp_absent(event, 'data/error')
>                     if check_offset:
>                         self.assert_qmp(event, 'data/offset', event['data']['len'])
>                     self.assert_no_active_block_jobs()
>                     return event
> 
> Or a single break would work too.  Do either of you have any
> preference?  I can repost in either way.  Thanks,

Looks good to me!

Stefan