[coreboot] Using USB keyboard in payloads

[Dave Frodin]

Thanks to everyone who sent feedback. Here's where I've gotten.

First I got a current libpayload. That made the biggest difference.
Evidently something critical changed in the last month.

The USB keyboard specific stuff in my code is...
   usb_initialize();
   if (havekey())
     usb_key = getchar();

The havekey() function is still always returning 1. But now
the call to getchar() is returning valid keystrokes.

The first time havekey() is called the following is output to
the console...

   00:12.2 4396:1002.2 EHCI controller
   * found device (0x0403:0x6014, USB 2.0), class: unsupported class ff
   00:12.0 4397:1002.0 OHCI controller
   OHCI Version 1.0
   00:13.2 4396:1002.2 EHCI controller
   00:13.0 4397:1002.0 OHCI controller
   OHCI Version 1.0
   * found device (0x045e:0x009d, USB 2.0)
   NOTICE: This driver defaults to using the first interface.
   This might be the wrong choice and lead to limited functionality
   of the device. Please report such a case to coreboot at coreboot.org
   as you might be the first.
   , class: HID
     Keyboard layout 'us'

The mainboard I'm currently debugging this on doesn't have a PS/2 keyboard
controller. It just has a USB keyboard. I'm going to move over to another
mainboard that has both to see what happens there.

Thanks again,
Dave

----- Original Message -----
> From: "Nico Huber" <nico.huber at secunet.com>
> To: "Dave Frodin" <dave.frodin at se-eng.com>
> Cc: coreboot at coreboot.org
> Sent: Monday, November 19, 2012 6:14:42 AM
> Subject: Re: [coreboot] Using USB keyboard in payloads
> 
> > 
> > I'm unable to get the USB keyboard to work in payloads.
> > I'm using it with libpayload, and configured libpayload to enable
> > the USB_HID, USB_OHCI, USB_EHCI, USB_UHCI, USB_XHCI drivers.
> > 
> > I call usb_initialize() in the early part of my payload.
> > 
> > Calls to usbhid_havechar() just return 0.
> When calling usbhid_havechar() directly, one should also call
> usb_poll()
> regularly.
> > Calls to havechar() just return 0xFF.
> Suspicious, havechar is defined as havekey in libpayload.h, havekey()
> should return 0 or 1. However, havekey() calls usb_poll() and, thus,
> should work.
> 
> > 
> > The USB keyboard works fine with SeaBIOS, which I am using to
> > load my payload as a boot option. I see no difference if I load
> > my payload directly from coreboot.
> > 
> > Does anyone have any ideas as to what I'm missing?
> If you use an USB hub, you should also enable USB_HUB. If you have
> an OHCI controller, try the latest upstream version, OHCI was broken
> for some time.
> 
> > 
> > Any suggestions or examples?
> Well, the USB HID implementation is quite fragile when it comes to
> timing. usb_poll() should be called at least once every 200ms with no
> guaranty what happens when you fail to do so. Also, detaching devices
> (hot unplugging) is broken.
> 
> You should see some output from libpayload when it detects USB
> controllers / devices, did you?
> 
> With some modification (see below), libpayload's 'Hello World'
> example
> shows input from keyboards (works with UHCI / EHCI+hub for me).
> 
> 
> Nico
> 
> > 
> > Thanks in advance,
> > Dave
> > 
> > 
> 
> diff --git a/payloads/libpayload/sample/hello.c
> b/payloads/libpayload/sample/hello.c
> index 677a96f..86113af 100644
> --- a/payloads/libpayload/sample/hello.c
> +++ b/payloads/libpayload/sample/hello.c
> @@ -34,7 +34,13 @@
>  
>  int main(void)
>  {
> +	usb_initialize();
> +
>  	printf("Hello world!\n");
> +	while (1) {
> +		if (havechar())
> +			putchar(getchar());
> +	}
>  	halt();
>  	return 0;
>  }
>