On 2021/8/20 7:58, Jason Gunthorpe wrote:
> On Thu, Jul 29, 2021 at 10:19:19AM +0800, Wenpeng Liang wrote:
>
>> DECLARE_UVERBS_NAMED_OBJECT(HNS_IB_OBJECT_DCA_MEM,
>> UVERBS_TYPE_ALLOC_IDR(dca_cleanup),
>> &UVERBS_METHOD(HNS_IB_METHOD_DCA_MEM_REG),
>> &UVERBS_METHOD(HNS_IB_METHOD_DCA_MEM_DEREG),
>> &UVERBS_METHOD(HNS_IB_METHOD_DCA_MEM_SHRINK),
>> &UVERBS_METHOD(HNS_IB_METHOD_DCA_MEM_ATTACH),
>> - &UVERBS_METHOD(HNS_IB_METHOD_DCA_MEM_DETACH));
>> + &UVERBS_METHOD(HNS_IB_METHOD_DCA_MEM_DETACH),
>> + &UVERBS_METHOD(HNS_IB_METHOD_DCA_MEM_QUERY));
>
> These lists should be kept sorted in each patch
>
DECLARE_UVERBS_NAMED_METHOD() will be put together in order.
>> +struct hns_dca_query_resp {
>> + u64 mem_key;
>> + u32 mem_ofs;
>> + u32 page_count;
>> +};
>
> This is strange, why in a public header, why have the query_dca_mem
> function at all? Just inline it and remove the struct entirely.
>
> Jason
> .
>
There is indeed no need to form a separate query_dca_mem function,
will inline it.
Thanks,
Wenpeng.
