Messages in this thread | | | Date | Thu, 27 Jul 2023 06:30:22 -0700 | From | Richard Cochran <> | Subject | Re: [PATCH v2] net: stmmac: correct MAC propagation delay |
| |
On Thu, Jul 27, 2023 at 08:40:51AM +0200, Johannes Zink wrote: > Hi Richard, > > On 7/26/23 17:34, Richard Cochran wrote: > > That is great, until they change the data sheet. Really, this happens. > > I think I don't get your point here. > > That's true for literally any register of any peripheral in a datasheet. > I think we can just stop doing driver development if we wait for a final > revision that is not changed any more. Datasheets change, and if they do we > update the driver.
This is different than normal registers, because the values are a guess as to what the latency in the hardware design is.
Here is how it works in practice: Vendor first asks a summer intern to measure the latency. Intern does some kind of random measurement, and that goes into silicon. One year later, customers discover that the values are bogus. Vendor doesn't spin a new silicon revision just for that. If vendor is honest, a footnote appears in the errata that the corrections are wrong.
Thanks, Richard
| |