Re: [PATCH rdma-next 2/6] IB/uverbs: Enable QP creation which is associated to underlay QP

[Jiri Pirko <jiri-rHqAuBHg3fBzbRFIqnYvSA@public.gmane.org>]

Wed, May 31, 2017 at 06:20:31AM CEST, leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org wrote:
>On Tue, May 30, 2017 at 08:03:58PM +0200, Jiri Pirko wrote:
>> Tue, May 30, 2017 at 07:22:59PM CEST, leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org wrote:
>> >On Tue, May 30, 2017 at 10:15:14AM +0200, Jiri Pirko wrote:
>> >> Tue, May 30, 2017 at 09:58:45AM CEST, leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org wrote:
>> >> >On Tue, May 30, 2017 at 09:33:52AM +0200, Jiri Pirko wrote:
>> >> >> Tue, May 30, 2017 at 09:15:58AM CEST, leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org wrote:
>> >> >> >From: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
>> >> >> >
>> >> >> >Enable QP creation which is associated to underlay QP. This comes as a
>> >> >> >pre-patch for downstream patches in this series to enable flow steering
>> >> >> >from user space application on the underlay IPoIB QP.
>> >> >> >
>> >> >> >Signed-off-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
>> >> >> >Reviewed-by: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org
>> >> >> >Signed-off-by: Leon Romanovsky <leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
>> >> >> >---
>> >> >>
>> >> >> [...]
>> >> >>
>> >> >>
>> >> >> >diff --git a/include/uapi/rdma/ib_user_verbs.h b/include/uapi/rdma/ib_user_verbs.h
>> >> >> >index 477d629f539d..422b20456975 100644
>> >> >> >--- a/include/uapi/rdma/ib_user_verbs.h
>> >> >> >+++ b/include/uapi/rdma/ib_user_verbs.h
>> >> >> >@@ -578,7 +578,7 @@ struct ib_uverbs_ex_create_qp {
>> >> >> > 	__u32 comp_mask;
>> >> >> > 	__u32 create_flags;
>> >> >> > 	__u32 rwq_ind_tbl_handle;
>> >> >> >-	__u32  reserved1;
>> >> >> >+	__u32  associated_qpn;
>> >> >>
>> >> >> This breaks uapi...
>> >> >
>> >> >Not really, we are using comp_mask scheme to signal changes/predecessor
>> >> >of these structs. Users of this file and structs check comp_mask and
>> >> >access new fields only if they are exposed there.
>> >> >
>> >> >In this case, there are no users of reserved1 and hence it can be
>> >> >replace to the new field with corresponding comp_mask flag (IB_QP_CREATE_ASSOC_QPN)
>> >> >without worry of breakage.
>> >>
>> >> I, as an app could include ib_user_verbs.h and do:
>> >>
>> >> struct ib_uverbs_ex_create_qp something;
>> >>
>> >> something.reserved1 = 0;
>> >>
>> >> Than I won't compile, would I?
>> >
>> >You are right, the good thing that in IB world, there is no direct usage
>> >of these files and they are needed for rdma-core (libibverbs) library.
>>
>> Unfortunatelly, that does not make any difference. UAPI is UAPI...
>
>Yes, it makes. Changes in UAPI are allowed as long they don't break user
>space applications. In this case, there are no applications which uses
>this reserved fields directly.

How do you know? I might have one in my drawer :) 
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html