timeout during PS/2 keyboard init

[Scott]

Hello,

While testing on AMD simnow I encounter a timeout in keyboard.c line 246:

/* All is well - enable keyboard interface */
if (!kbc_input_buffer_empty()) return;
outb(0x60, KBD_COMMAND);
if (!kbc_input_buffer_empty()) return;
outb(0x61, KBD_DATA);	/* send cmd: enable keyboard and IRQ 1 */
if (kbc_output_buffer_full()) {
    printk(BIOS_ERR, "Timeout during final keyboard enable\n"); <=======
    return;
}

It seems like line 245 should call kbc_input_buffer_empty() instead of
kbc_output_buffer_full() because the previous I/O does not cause the
keyboard to generate any data.

On simnow, this causes a boot delay of a minute or so. On real hardware,
It appears it could cause a boot delay of 400 ms. Does anyone testing 
with real hardware ever notice the "Timeout during final keyboard enable"
message when logging is enabled?

Thanks,
Scott

[Myles Watson]

On Tue, Aug 31, 2010 at 11:16 PM, Scott  wrote:
> Hello,
>
> While testing on AMD simnow I encounter a timeout in keyboard.c line 246:
>
> /* All is well - enable keyboard interface */
> if (!kbc_input_buffer_empty()) return;
> outb(0x60, KBD_COMMAND);
> if (!kbc_input_buffer_empty()) return;
> outb(0x61, KBD_DATA);   /* send cmd: enable keyboard and IRQ 1 */
> if (kbc_output_buffer_full()) {
>        printk(BIOS_ERR, "Timeout during final keyboard enable\n"); <=======
>        return;
> }
>
> It seems like line 245 should call kbc_input_buffer_empty() instead of
> kbc_output_buffer_full() because the previous I/O does not cause the
> keyboard to generate any data.
>
> On simnow, this causes a boot delay of a minute or so. On real hardware,
> It appears it could cause a boot delay of 400 ms. Does anyone testing
> with real hardware ever notice the "Timeout during final keyboard enable"
> message when logging is enabled?

if (!timeout) {
                printk(BIOS_INFO, "Keyboard controller output buffer
result timeout\n");
        }

I see this one in hardware.

Thanks,
Myles

[Marc Jones]

On Tue, Aug 31, 2010 at 11:16 PM, Scott  wrote:
> Hello,
>
> While testing on AMD simnow I encounter a timeout in keyboard.c line 246:
>
> /* All is well - enable keyboard interface */
> if (!kbc_input_buffer_empty()) return;
> outb(0x60, KBD_COMMAND);
> if (!kbc_input_buffer_empty()) return;
> outb(0x61, KBD_DATA);   /* send cmd: enable keyboard and IRQ 1 */
> if (kbc_output_buffer_full()) {
>        printk(BIOS_ERR, "Timeout during final keyboard enable\n"); <=======
>        return;
> }
>
> It seems like line 245 should call kbc_input_buffer_empty() instead of
> kbc_output_buffer_full() because the previous I/O does not cause the
> keyboard to generate any data.
>
> On simnow, this causes a boot delay of a minute or so. On real hardware,
> It appears it could cause a boot delay of 400 ms. Does anyone testing
> with real hardware ever notice the "Timeout during final keyboard enable"
> message when logging is enabled?
>
> Thanks,
> Scott

Yes, That should be a input buffer empty check since there isn't an
ack to that command. Send a qucik patch and signed-off-by, I will ack
it.

Marc