Re: [Qemu-devel] [PATCH v2 0/2] Delay initialization of memory backends

[Maxime Coquelin <maxime.coquelin@redhat.com>]

On 09/06/2016 04:24 AM, Michael S. Tsirkin wrote:
> On Mon, Sep 05, 2016 at 04:01:30PM -0300, Eduardo Habkost wrote:
>> On Mon, Sep 05, 2016 at 02:26:30PM +0200, Paolo Bonzini wrote:
>>> On 02/09/2016 20:59, Eduardo Habkost wrote:
>>>> While trying to fix the original bug in v1, another bug was fixed
>>>> by accident: TCG initialization of dirty_log_mask was broken when
>>>> using memory backends.
>>>>
>>>> The fix, on the other hand, broke vhost-user-test because it
>>>> relied on TCG, even though TCG is incompatible with vhost.
>>>>
>>>> This new version changes vhost-user-test to not rely on TCG, so
>>>> we can finally fix the initialization ordering of memory
>>>> backends.
>>>>
>>>> Eduardo Habkost (2):
>>>>   vhost-user-test: Use libqos instead of pxe-virtio.rom
>>>>   vl: Delay initialization of memory backends
>>>>
>>>>  tests/Makefile.include  |  2 +-
>>>>  tests/vhost-user-test.c | 37 ++++++++++++++++++++++++++++++++++---
>>>>  vl.c                    | 13 +++++++++++++
>>>>  3 files changed, 48 insertions(+), 4 deletions(-)
>>>>
>>>
>>> Acked-by: Paolo Bonzini <pbonzini@redhat.com>
>>>
>>> for whoever wants to pick this up...
>>
>> I will commit it. Thanks!
>
> Feel free to commit if you didn't already.
>
> Belatedly:
>
> Acked-by: Michael S. Tsirkin <mst@redhat.com>

Michael, please note that not using TCG also fixes the hang observed
with patch:
28ed5ef ("vhost-user: Attempt to fix a race with set_mem_table.")
It confirms my analysis that it would only happen with relying on TCG.
Should we revert your revert once this series applied to mainline?

So, FWIW now that the patch may be already committed:
Tested-by: Maxime Coquelin <maxime.coquelin@redhat.com>

Thanks,
Maxime