Re: [PATCH 2/4] x86/amd_nb: add support for newer PCI topologies

[Borislav Petkov]

On Mon, Nov 05, 2018 at 08:33:34PM +0000, Woods, Brian wrote:
> I think having them togeter is cleaner. If you aren't finding any
> misc IDs, I highly doubt you'll find any root IDs.  There shouldn't
> be much of a difference in how fast the function exits, either way.
> If you want it the other way though, I don't mind changing it.

Yes please. Because this is the usual kernel coding style of calling a
function (or a loop which has some result in this case) and testing that
result immediately after the function call.

> Would
> 
> 		/*
> 		 * If there are more PCI root devices than data fabric/
> 		 * system management network interfaces, then the (N)
> 		 * PCI roots per DF/SMN interface are functionally the
> 		 * same (for DF/SMN access) and N-1 are redundant.  The
> 		 * N-1 PCI roots should be skipped per DF/SMN interface
> 		 * so the DF/SMN interfaces get mapped to the correct
> 		 * PCI root.

You say "correct" as there is a special one. But the text before it says
they're "functionally the same" wrt DF/SMN access so it sounds to me
like we wanna map the first one we find and ignore the others.

I.e., we wanna say

"... so the DF/SMN interfaces get mapped to the *first* PCI root and the
others N-1 ignored."

Or am I misreading this?

Thx.

-- 
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.