On Mon, Oct 17, 2016 at 09:58:26AM +0200, Jacek Anaszewski wrote: > On 10/15/2016 02:00 PM, Matt Ranostay wrote: > > On Fri, Oct 14, 2016 at 7:20 AM, Tony Lindgren <tony@xxxxxxxxxxx> wrote: > > > * Jacek Anaszewski <j.anaszewski@xxxxxxxxxxx> [161013 23:37]: > > > > On 10/13/2016 04:20 PM, Matt Ranostay wrote: > > > > > On Thu, Oct 13, 2016 at 4:05 PM, Jacek Anaszewski > > > > > <j.anaszewski@xxxxxxxxxxx> wrote: > > > > > > Why DT property? Is it somehow dependent on the board configuration? > > > > > > How this period-scale value is calculated? Is it inferred empirically? > > > > > > > > > > > > > > > > We empirically discovered and verified this with an logic analyzer on > > > > > multiple batches of this part. > > > > > Reason for the DT entry is we aren't 100% sure that it is always going > > > > > to be the same with different board revs. > > > > > > > > > > Could be that parts clock acts differently with supply voltage. This > > > > > has been calculated by setting it an expected value, and measuring the > > > > > actual result with the logic analyzer. > > > > > > > > I'd like to have DT maintainer's ack for this. > > > > > > > > Cc Rob and Mark. > > > > > > How about do this based on the compatible property instead? If there > > > are multiple manufacturers for this part and only a certain > > > parts have this issue we should have multiple compatible properties. > > > > > > > I could only find that NXP as the manufacturer of that part. It is > > possible since the clock is internal to the chipset that the vdd of > > 2.5V is doing something undefined. > > > > > Then if it turns out all of them need this scaling there's no need > > > to update the binding. > > > > Understandable. > > Since at present we can't guarantee that all produced devices > are affected, then we should strive to avoid breaking any existing > users of the possible non-affected devices. > > In view of that the addition of a new "compatible" proposed by Tony > seems most reasonable. > > Still, DT maintainer's opinion is required. Seems like a quirk of this board, so I think the added property is fine. It could be existing users just didn't notice the rate being off. 30% is probably not all that noticeable to the human eye. Rob -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html