Re: [PATCH] net:cpsw: Pass unhandled ioctl's on to generic phy ioctl

[Ben Hutchings <ben@decadent.org.uk>]

[-- Attachment #1: Type: text/plain, Size: 1574 bytes --]

On Wed, 2014-02-05 at 08:12 +0100, Richard Cochran wrote:
> On Tue, Feb 04, 2014 at 09:51:59PM +0000, Ben Hutchings wrote:
> > 
> > Right.  If all versions of CPSW include hardware timestamping then
> > bother with PHY timestamping at all?  And why make CONFIG_TI_CPTS
> > configurable?
> 
> On the one hand, PHY time stamping is more accurate and offers
> synchronization performance that is measurably better than MAC time
> stamping.

I suppose that depends on how much jitter there is in the PHY?

With SFC boards and 10GBASE-SR modules or DA cables, I would see
synchronisation to within about 10 nanoseconds using MAC timestamps.

> On the other hand, when using a MAC the CPU usually has much
> more direct access to the clock (for example, direct register access
> or PCIe, versus MDIO).
> 
> I once worked on a project in which it was planned to have both kinds
> of hardware in the design, in order to keep our options open in the
> face of fluid requirements. So I think you can expect to see such
> combinations in the wild, especially in the embedded area.
> 
> We cannot reasonably support both types in the kernel at the same
> time, and so it makes sense to have compile time options in MAC
> drivers to disable time stamping.

I would think that it should be a *run-time* option, as I don't like
forcing people to rebuild their kernel.  But perhaps this is esoteric
enough that the people who care will be doing that anyway.

Ben.

-- 
Ben Hutchings
Once a job is fouled up, anything done to improve it makes it worse.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 811 bytes --]