On 11-Dec-18 19:03, Jason Gunthorpe wrote:
> On Tue, Dec 11, 2018 at 06:31:09PM +0200, Gal Pressman wrote:
>> On 11-Dec-18 18:22, Jason Gunthorpe wrote:
>>> On Tue, Dec 11, 2018 at 10:59:25AM +0200, Gal Pressman wrote:
>>>
>>>> diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
>>>> index 85021451eee0..499d96302262 100644
>>>> +++ b/include/rdma/ib_verbs.h
>>>> @@ -2374,6 +2374,7 @@ struct ib_device {
>>>> int (*dealloc_pd)(struct ib_pd *pd);
>>>> struct ib_ah * (*create_ah)(struct ib_pd *pd,
>>>> struct rdma_ah_attr *ah_attr,
>>>> + bool sleepable,
>>>> struct ib_udata *udata);
>>>
>>> I think the Linux style consensus here is that these should be a flag with a
>>> enum name not a bool
>>
>> Will change.
>> I'm familiar with the no bool fields in structs
>> (https://lkml.org/lkml/2017/11/21/384), good to know it applies to function
>> parameters as well.
>
> I think it is under the rational that
>
> foo(true)
>
> tells the user nothing, while
>
> foo(DO_SOMETHING)
>
> Should be more informative
>
> Same argument applies to 'int' pretending to be bool
>
> At least a struct member has a name..
>
> Jason
>
That makes sense, thanks for the explanation.
