[ofa-general] Re: IB_SEND_INLINE in the kernel? (was [PATCH -2] ib core - checksum offload)

[Or Gerlitz]

Eli Cohen wrote:
> Index: linux-2.6.23-rc1/include/rdma/ib_verbs.h
> ===================================================================
> --- linux-2.6.23-rc1.orig/include/rdma/ib_verbs.h	2007-08-15 20:50:16.000000000 +0300
> +++ linux-2.6.23-rc1/include/rdma/ib_verbs.h	2007-08-15 20:50:28.000000000 +0300
> @@ -613,7 +616,9 @@ enum ib_send_flags {
>  	IB_SEND_FENCE		= 1,
>  	IB_SEND_SIGNALED	= (1<<1),
>  	IB_SEND_SOLICITED	= (1<<2),
> -	IB_SEND_INLINE		= (1<<3)

Roland,

My thinking was the inline sending is not supported for kernel level 
ULPs, eg from the reason that using CPU cycles in the kernel to copy 
data should be avoided when possible (eg when the device does HW DMA)

Grepping for IB_SEND_INLINE under drivers/infiniband shows that all the 
HW drivers just ignore it, can it be removed?

Or.