On Thu, 10 Sep 2015 19:21:11 +0200 Sascha Hauer <s.hauer@xxxxxxxxxxxxxx> wrote: > On Thu, Sep 10, 2015 at 06:51:33PM +0300, Peter Mamonov wrote: > > On Thu, 10 Sep 2015 09:38:15 +0200 > > Sascha Hauer <s.hauer@xxxxxxxxxxxxxx> wrote: > > > > > Excellent! This makes the framebuffer console support complete :) > > > > I'm sorry to dissapoint you, but this implementation of the usb > > keyboard driver seems to be wrong. The driver employs the get_report > > request to poll the keyboard state via the control endpoint, however > > the "Device Class Definition for Human Interface Devices (HID)" > > states the following: > > "This request is not intended to be used for polling the device > > state on a regular basis. [...] The Interrupt In pipe should be > > used for recurring Input reports." > > > > I guess this is the source of the problems with some keyboards: 2 > > of 4 keyboards I've tested changed their state only on key down, > > but not on key up. > > Yes, same here. The first keyboard I tested ended up with this > behaviour, the second worked. > > > > > So, according the standard, we need to poll a keyboard via interrupt > > endpoint. However, current implementation of EHCI driver doesn't > > support interrupt transactions: submit_int_msg() in ehci-hcd.c is a > > stub. I'll try to implement it. > > U-Boot has an implementation for it, should be possible to adopt the > corresponding code. I've ported missing functionality from the u-boot and switched to polling via interrupt endpoint. After that all keyboards work fine. Will clean up the code and post the patches. Peter > > Sascha > _______________________________________________ barebox mailing list barebox@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/barebox
