On Tuesday 24 March 2009, Martin Fuzzey wrote: > The i.MX21 has a built in root hub and 3 ports, however depending on > the board layout, only some of the ports may be accessible. > When only some ports are available should GetHubDescriptor still > indicate 3? [I presume so as other drivers do this] > But in this case I just shouldn't activate the inaccessible ports - correct? It's simplest to return "3" rather than try any kind of remapping ("only port 3 works, remap it as port 1") or even just shrinking "3" ("only port 1 works, report it as only having 1 port"). If something happens on the other ports, you'd have to cope with it anyway. What you do with those ports is somewhat SOC-specific. If "activating" has any side effects, don't; but if you can avoid such special-casing, save the effort. > Furthermore there are two host only ports and one port that can be > configured as host / function / OTG. > Will OTG support be required before merging the hcd driver? (once all > the host only problems are fixed obviously) No. In fact you *really* really want to get host support working well before you touch OTG support. > I currently know next to nothing about OTG in general and the linux > implementation in particular. I've looked at drivers/usb/otg but the > isp1301 tranceiver (which my board uses) code is mixed up with omap. Yes. Somewhere in my mailbox I have patch from someone at TI to split that driver in the middle, following some kind of CEA-standardized version of the ISP1301 model. Right now, that's the most functional OTG support in Linux, and it's a bit specific to certain chip versions. As in, OMAP1611b ... but not OMAP1710, which updated the OTG controller. > If possible I would prefer to get the hcd driver merged before looking > at OTG. Are the any special things to be aware of to make later > addition of OTG go smoothly? Just be prepared for a bit of black magic needing to get stirred into the mix. :) The framework level interfaces are sort of OK, but a lot of the interactions between drivers are more on the custom-to-this-hardware level. So far. That's obviously not the best way to be, long term. - Dave -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html