[openib-general] Re: RFC: process_mad extension
Michael S. Tsirkin
mst at mellanox.co.il
Mon Jan 10 01:55:09 PST 2005
Hello!
Quoting r. Roland Dreier (roland at topspin.com) "Re: [openib-general] Re: RFC: process_mad extension":
> Thanks, but:
>
> - The patch has some bad spacing in places like:
>
> + wc->wr_id=send_wr->wr_id;
> + wc->status=IB_WC_SUCCESS;
Ugh, need fixing.
> as well as C++-style declarations:
>
> + struct ib_wc* in_wc,
> + struct ib_grh* in_grh,
what do you mean "C++-style"? what's unusual here?
> - Rather than changing IB_MAD_IGNORE_MKEY to IB_MAD_IGNORE_MKEY_BKEY,
> why not introduce new flags IB_MAD_IGNORE_BKEY and IB_MAD_IGNORE_ALL
> with IB_MAD_IGNORE_ALL set to IB_MAD_IGNORE_MKEY | IB_MAD_IGNORE_BKEY?
> This matches the behavior of the FW better and lets existing code
> such as sysfs.c remain unchanged.
I think these are mutually exclusive (no mad has both mkey and bkey).
And I believe the existing code wants to "ignore all" anyway, since
it sends a trusted MAD.
> - This last comment is just taste and maybe I'm wrong, but I don't
> see much advantage to defining a separate function build_smp_wc()
> that is only called from one place.
Maybe, I can arrange to call it from another place too :)
> Thanks,
> Roland
More information about the general
mailing list