On 1/14/19 8:18 AM, Peter Xu wrote:
> On Mon, Jan 14, 2019 at 08:02:23AM +0100, Markus Armbruster wrote:
>> Peter Xu <pet...@redhat.com> writes:
>>
>>> On Sun, Jan 13, 2019 at 10:36:41PM +0800, Fei Li wrote:
>>>> From: Fei Li <shirley17...@gmail.com>
>>>>
>>>> Let's supplement the msi_uninit() when failing to realize
>>>> the pci edu device.
>>>>
>>>> Cc: Markus Armbruster <arm...@redhat.com>
>>>> Cc: Peter Xu <pet...@redhat.com>
>>>> Cc: Michael S. Tsirkin <m...@redhat.com>
>>>> Cc: Marcel Apfelbaum <marcel.apfelb...@gmail.com>
>>>> Signed-off-by: Fei Li <shirley17...@gmail.com>
>>>> ---
>>>> hw/misc/edu.c | 1 +
>>>> 1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/hw/misc/edu.c b/hw/misc/edu.c
>>>> index cdcf550dd7..4feb7503de 100644
>>>> --- a/hw/misc/edu.c
>>>> +++ b/hw/misc/edu.c
>>>> @@ -367,6 +367,7 @@ static void pci_edu_uninit(PCIDevice *pdev)
>>>> {
>>>> EduState *edu = EDU(pdev);
>>>>
>>>> + msi_uninit(pdev);
>>>
>>> It would be cleaner to me to call this after the join() since
>>> edu_fact_thread() could potentially use msi_*() helpers then the
>>> destructions follow the reverse order of init.
>>
>> Destruction in reverse creation order is good practice.
>>
>> This being the "QEMU educational PCI device", good practice is even more
>> desirable.
>
> Very persuasive. With that, I'd like to withdraw my excuse of
> "msi_uninit() is optional" too. :)
>
> Fei, please feel free to pick my r-b if you want to repost, and IMHO
> you can also add:
>
> Reported-by: Markus Armbruster <arm...@redhat.com>
>
> Thanks,
>
Moving msi_uninit() after timer_del():
Reviewed-by: Philippe Mathieu-Daudé <phi...@redhat.com>
