On 3/24/21 12:07 PM, Jason Gunthorpe wrote:
> On Wed, Mar 24, 2021 at 11:52:19AM -0500, Bob Pearson wrote:
>>>> +struct rxe_mw {
>>>> + struct rxe_pool_entry pelem;
>>>> + struct ib_mw ibmw;
>>>> + struct rxe_qp *qp; /* type 2B only */
>>>> + struct rxe_mr *mr;
>>>> + spinlock_t lock;
>>>> + enum rxe_mw_state state;
>>>> + u32 access;
>>>> + u64 addr;
>>>> + u64 length;
>>>> +};
>>>
>>> struct rxe_qp *qp; /* type 2B only */
>>> struct rxe_mr *mr;
>>> spinlock_t lock;
>>> enum rxe_mw_state state;
>>> u32 access;
>>> u64 addr;
>>> u64 length;
>>>
>>> The above member variables are not used in your commit. Why keep them
>>> in this struct rxe_mw?
>>>
>>> Zhu Yanjun
>>>
>>
>> There is more to come. The goal here is to implement MW and peeking ahead
>> MWs need each of those fields. As soon as this change gets accepted I will start
>> adding code to implement the MW verbs APIs.
>
> The requirement is to add things when you need them, so if these are
> unused here they should move to the patch that requires them
>
> Jason
>
OK can do. I need to wait another day to see if Zhu is ready to accept the whole idea
of renaming these things. There are two other nits in this patch that I could change.
There was one whitespace change that could come separately (n spaces -> tab) and Leon
moved the MW into core since the first time I sent this in which requires reversing
the order of the ibmw atruct and the pelem struct in mw. It has to change before the
mw struct can actually be used. What do you think?
bob
