On Wed, Nov 24, 2010 at 6:38 PM, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > On Wed, 24 Nov 2010, Alex Sinchenko wrote: > >> Hello. >> >> I'm trying to deal with OHCI issues on AMD Geode CS5536. I use >> PCEngines Alix alix2d3 mainboard, that have one 1.1 and one 2.0 port. > > Are you sure that each port isn't connected to both USB host > controllers? That's how mainboards are usually designed. You are right, but I can see these ports in Linux (and equipment connected to them) only if I load "proper" module - ehci-hcd for first port and ohci-hcd for second. > >> On each of ports we connect USB hub tusb2046, with two CM108 USB >> audiocards and two CP2102 usb-serial converters on each, all >> externally-powered. > > That's the way they are connected now? Yes >> We used to have four sound cards on one USB host and four converters >> on other, and all worked fine. > > Do you mean that you used to have all four sound cards plugged into one > of the hubs and all four serial converters plugged into the other hub? > How do you know that the two hubs used different host controllers? Well, I was wrong with it. There is only one host controller in CS5536 that provides independent mode for each port. > >> But on 12 we was not be able to use all > > What do you mean by "on 12"? > Sorry. I mean 12 mbit/s. Probably it is not enough to use four sound cards connected to the same hub. >> 4 sound cards in parallel, getting "no bandwidth" error. >> After we rearranged devices, problems with device detection arised. I > > Did you go back to the old arrangement to see whether the detection > problems went away? > Yes. > What are the exact differences between the working and non-working > arrangements? Difference is only in devices connected to each port on board. Combination "4 soundcards on one port and 4 serial converter on other" works. Combination "2 soundcards and 2 converters on port" do not work. > >> tried different versions of the kernel (specifically, Voyade Linux >> kernels 2.6.26 and 2.6.30 optimized for Alix boards, and vanilla >> 2.6.35.1). >> I tried all possible combinations with ehci-hcd and ohci-hcd loading >> order, tried to not load ehci-hcd to see if usb 1.1 port will work >> correctly.. No success. > > It would be best to run the tests with ehci-hcd not loaded. Ok, but for test only - without ehci-hcd one USB port on mainboard is invisible in Linux at all > >> I turned on USB Debug, see dmesg output in attach. > > Try again, without ehci-hcd and with only one hub plugged in, no other > USB devices. Let's see what happens then. I will try that ASAP. > >> Interestinly, if switch off and on power on devices and hubs when >> Linux is loaded, devices are detected correctly. > > Please be more precise. Do you switch the devices and hubs off > _before_ booting, _during_ booting, or _after_ booting? And at what > time do you switch them back on? After Linux fully booted, I switched device and hub power off and instantly switched it on. > >> Just removing and >> loading again ohci-hcd and ehci-hcd does not help. > > What happens if you boot with all the devices and hubs unplugged, and > then plug them in one at a time after the system is running? All works correctly. Obviously, it's the same situation as I said about switching off and on. > > Alan Stern > > Thank you very much, Alan. -- WBR, Alex Sinchenko -- 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
