Re: [PATCH] qemu-kvm: Fix test for I/O thread

[Avi Kivity <avi@qumranet.com>]

Jan Kiszka wrote:
> Avi Kivity wrote:
>   
>> Jan Kiszka wrote:
>>     
>>> Looks to me like this was rather intended. But given that
>>> 4e8b8a6d92c5ece048e65be3a3980d24f065b32b claims to actually fix a bug in
>>> its original broken form, please have a careful look.
>>>
>>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>>>
>>> diff --git a/qemu/qemu-kvm.c b/qemu/qemu-kvm.c
>>> index 431e26d..c36f60f 100644
>>> --- a/qemu/qemu-kvm.c
>>> +++ b/qemu/qemu-kvm.c
>>> @@ -89,7 +89,7 @@ static void qemu_cond_wait(pthread_cond_t *cond)
>>>      pthread_cond_timedwait(cond, &qemu_mutex, &ts);
>>>      /* If we're the I/O thread, some other thread may be waiting for aio
>>>       * completion */
>>> -    if (!vcpu_info)
>>> +    if (!env)
>>>          qemu_aio_poll();
>>>      cpu_single_env = env;
>>>  }
>>>   
>>>       
>> Aren't the two lines equivalent?  vcpu_info is a thread-local-storage
>> variable, and is unset for the iothread.
>>
>>     
>
> Then you probably wanted to type 'vcpu', don't you? :)
>
>   

Doh.

> However, I would stick with what the function already uses, ie. 'env'.
>   

env might be null even in a non-aio-thread (during initialization, 
perhaps?).  I changed it to be vcpu since that's a stronger indicator.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.