Re: [PATCH liburing] io-cancel: add check for -ECANCELED

[Hao Xu <haoxu@linux.alibaba.com>]

在 2021/10/29 下午9:57, Pavel Begunkov 写道:
> On 10/29/21 13:27, Hao Xu wrote:
>> ping this one since test/io-cancel will be broken
>> if the async hybrid logic merges to 5.16
>> 在 2021/10/19 下午5:23, Hao Xu 写道:
>>> The req to be async cancelled will most likely return -ECANCELED after
>>> cancellation with the new async bybrid optimization applied. And -EINTR
>>> is impossible to be returned anymore since we won't be in INTERRUPTABLE
>>> sleep when reading, so remove it.
>>>
>>> Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
>>> ---
>>>   test/io-cancel.c | 2 +-
>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/test/io-cancel.c b/test/io-cancel.c
>>> index b5b443dc467b..c761e126be0c 100644
>>> --- a/test/io-cancel.c
>>> +++ b/test/io-cancel.c
>>> @@ -341,7 +341,7 @@ static int test_cancel_req_across_fork(void)
>>>                   fprintf(stderr, "wait_cqe=%d\n", ret);
>>>                   return 1;
>>>               }
>>> -            if ((cqe->user_data == 1 && cqe->res != -EINTR) ||
>>> +            if ((cqe->user_data == 1 && cqe->res != -ECANCELED) ||
> 
> cqe->res != -ECANCELED && cqe->res != -EINTR?
> 
> First backward compatibility, and in case internals or the test
> changes.
Ah, I see. Thanks Pavel.
> 
>>>                   (cqe->user_data == 2 && cqe->res != -EALREADY && 
>>> cqe->res)) {
>>>                   fprintf(stderr, "%i %i\n", (int)cqe->user_data, 
>>> cqe->res);
>>>                   exit(1);
>>>
>>
>