On Fri, 2015-03-20 at 16:27 +0100, Hans de Goede wrote: > Hi, > > On 20-03-15 15:39, Ian Campbell wrote: > > On Fri, 2015-03-20 at 13:42 +0000, Ian Campbell wrote: > >> On Fri, 2015-03-20 at 13:14 +0100, Hans de Goede wrote: > >>> On 20-03-15 12:47, Ian Campbell wrote: > >> > >>>> I'm not really sure where to start looking. Perhaps CONFIG_GMAC_TX_DELAY > >>>> on the u-boot side might be relevant? > >>> > >>> Yes that is the first thing I was thinking of, the cubietruck is the only > >>> gbit phy using a20 design which does not need it, which is kinda fishy. > >> > >> Ah, I hadn't cottoned on to that little fact-let, it's certainly worth a > >> try then! > > > > Setting CONFIG_GMAC_TX_DELAY=3 (or actually, the moral equivalent on > > v2014.10, i.e removing the ifdef BANANPI) resulted in non-functional > > tftp from u-boot (luckily I tried it on a local board first!). > > Interesting ... > > > Using 0, 1, or 2 seems to work from u-boot at least. I've not tried > > anything more aggressive, my local board didn't have issues to start > > with though. > > > > I'm going to triple check my remote recovery procedures and then do some > > experiments on one of the problematic boards. > > I wonder if it could be that there are 2 production runs of the cubietruck, > with potentially different phy revisions.
FWIW the 4 which are remote were purchased in the same batch, which may not mean much but makes it a bit less likely they were different batches. If, say, tx delay == 1 was needed, then is it possible that the default setting of 0 might work on some systems perhaps depending on other factors (manufacturing imperfections, cable differences)? > Tom Cubie (added to the Cc), can you tell us if it is possible that there > are different ethernet phy revisions on some cubeitrucks? Has there been > a second production run ? I think Tom moved onto other things quite a long time ago. Ian. -- You received this message because you are subscribed to the Google Groups "linux-sunxi" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
