Has anyone out there had any success using lots of USB 1.1 devices through USB 2.0 hub transaction translators ?
Probably not, given the known restriction in the interrupt scheduling code: one interrupt transaction per frame. (A handy implementation shortcut, avoiding the need to maintain a schedule tree.)
I am building up to a fan-out including maybe 120 USB 1.1 devices. My initial test setup has 2 USB 2.0 hubs (cypress, with 4 x TT each), and each of the 8 ports has a USB 1.1 hub attached, to each of which is attached 2 USB 1.1 devices. So, this is 2 USB 2.0 devices, 8 x 1.1 hubs, and 16 x 1.1 non-hub devices, but the devices don't show up reliably under linux-2.5.75, problems with set_address, and hub connect debounce.
Has anyone had success with this many or more devices ? Does this improve under 2.6 ?
It ought to improve with that test6 patch I just sent.
If you can use such a 122-device monster tree to help debug that patch, that'd be a Good Thing ... heck, even just those 2 dozen devices should give it a good workover!
For example, I was unable to test with a TT-per-port hub; that needs specific testing. I could imagine that some of the problems you saw were because that kind of hub wasn't handled right somewhere.
- Dave
------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
