On 5/13/08, <b class="gmail_sendername">Roland Dreier</b> <<a href="mailto:rdreier@cisco.com">rdreier@cisco.com</a>> wrote:<div><span class="gmail_quote"></span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
> +enum rdma_ha_mode {<br> > + RDMA_ALIGN_WITH_NETDEVICE = 1<br> > +};<br> <br> > +int rdma_set_high_availability_mode(struct rdma_cm_id *id, enum rdma_ha_mode mode)<br> <br> this seems like overengineering to me... given there are no other modes,<br>
you are adding an elaborate NOP. (Nothing looks at ha_mode)</blockquote><div><br>
First, this patch would be later extended for the rdma_ucm part
(exposing the ha_mode to user space). Second, indeed nothing looks on
ha_mode in this patch, but the next one (4/4) uses it. I was thinking
its better to decompose the changes this way such that patches are not
too small and not too big both in size and the change they carry in
their content.<br>
</div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Do you have plans for other modes?</blockquote><div><br>
down the road someone might want to add APM support for the rdma-cm, or more modes that I can't think of now. </div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
> u8 srq;<br> > u8 tos;<br> > + enum rdma_ha_mode ha_mode;<br> <br>Side note -- you're wasting two bytes here because of alignment.</blockquote>
<div><br>
</div></div>What would be the easy way to avoid it?<br>
<br>
Or.<br>
<br>