keyboard timeout

keyboard timeout

[linux-os]

If Linux 2.6.9 is booted on a 40 MHz `486 with the standard
ISA clock of 14.3 MHz (yes that's the standard), the kernel
will complain about a keyboard timeout for every key touched!

Somebody apparently didn't test their changes sometime in
the past I've seen this also on 2.4.26, but not so often.

Will somebody please double the time-out or tell me where
to fix it! The timeout was only supposed to handle the
fact that a keyboard could get (or be) disconnected. It
can be a long timeout without ever affecting ordinary
use.

Cheers,
Dick Johnson
Penguin : Linux version 2.6.9 on an i686 machine (5537.79 BogoMips).
  Notice : All mail here is now cached for review by John Ashcroft.
                  98.36% of all statistics are fiction.

Re: keyboard timeout

[Dave Dillow]

On Wed, 2004-12-01 at 17:29, linux-os wrote:
> If Linux 2.6.9 is booted on a 40 MHz `486 with the standard
> ISA clock of 14.3 MHz (yes that's the standard), the kernel
> will complain about a keyboard timeout for every key touched!

Umm, no.

The BCLK signal, which is the system bus clock, is 4.77, 8.0, or 8.33
MHz, depending on platform. There may be other non-standard frequencies
in use. Since everything is referenced to this signal, I'd call it the
ISA clock.

You may be thinking of the OSC signal, which is 14.33MHz, and was used
to generate the composite video signal on early CGA cards. It is not
synchronized to any other signal. It is (was) also often divided by 12
and fed to the 8254 PIT.

As for your keyboard problem, I dunno. I just wanted to play "The find
the subtle misconception in Dick's post" game.

My reference: "ISA & EISA Theory & Operation" by Edward Solari.
-- 
Dave Dillow <dave@thedillows.org>

Re: keyboard timeout

[Alan Cox]

On Mer, 2004-12-01 at 22:29, linux-os wrote:
> If Linux 2.6.9 is booted on a 40 MHz `486 with the standard
> ISA clock of 14.3 MHz (yes that's the standard), the kernel
> will complain about a keyboard timeout for every key touched!

8.33Mhz. The delays should be correct but given that just about all
hardware under 15 years old doesn't care (I think the last thing to care
was the digital hi-note laptop) it is possible that the new input code
has a tiny missing delay somewhere. Having said that I have specifically
audited the input keyboard driver for such problems in 2.6.5 or so and
found only one (which is fixed)

Nor should the ISA bus speed matter - the uController chugs along at
about 2Mhz and the delays it needs are a bit longer than just ISA
cycles.

Re: keyboard timeout

[linux-os]

On Wed, 1 Dec 2004, Alan Cox wrote:

> On Mer, 2004-12-01 at 22:29, linux-os wrote:
>> If Linux 2.6.9 is booted on a 40 MHz `486 with the standard
>> ISA clock of 14.3 MHz (yes that's the standard), the kernel
>> will complain about a keyboard timeout for every key touched!
>
> 8.33Mhz. The delays should be correct but given that just about all
> hardware under 15 years old doesn't care (I think the last thing to care
> was the digital hi-note laptop) it is possible that the new input code
> has a tiny missing delay somewhere. Having said that I have specifically
> audited the input keyboard driver for such problems in 2.6.5 or so and
> found only one (which is fixed)
>
> Nor should the ISA bus speed matter - the uController chugs along at
> about 2Mhz and the delays it needs are a bit longer than just ISA
> cycles.

>From original IBM specification...
"...8284A Clock generator clock generator supplies the multiphase
clock signals that are needed to drive the microprocessor and
the peripherals. Its base frequency is 14.31818 MHz....."

 	14.31818 / 3 =  4.7727... MHz (the original CPU clock frequency)
 	14.31818 / 12 = 1.1931... frequency fed to 8253 PIT and
 		keyboard controller.

 	... etc...

Low power clones, designed to use very low power __still__ use
these low frequencies. These frequencies were chosen by ME
because they can derive from the 3.579545 MHz color sugarier
frequency that was used in color television. The early PCs
were expected to be connected to color TV.

FYI 3.579545 (NTSC color sub-carrier frequency) is 14.31818 / 4.
I know all these numbers by heart because I had to defend them
over several murderous design reviews in Boca Raton. Modern
motherboards generate the PIT and keyboard frequencies (also
the UART input) using other methods, but these low frequencies
are still in use. And, yes the keyboard MUST have a longer timeout
regardless of your "audit". Thanks for the help.

Cheers,
Dick Johnson
Penguin : Linux version 2.6.9 on an i686 machine (5537.79 BogoMips).
  Notice : All mail here is now cached for review by John Ashcroft.
                  98.36% of all statistics are fiction.

Re: keyboard timeout

[linux-os]

On Wed, 1 Dec 2004, Dave Dillow wrote:

> On Wed, 2004-12-01 at 17:29, linux-os wrote:
>> If Linux 2.6.9 is booted on a 40 MHz `486 with the standard
>> ISA clock of 14.3 MHz (yes that's the standard), the kernel
>> will complain about a keyboard timeout for every key touched!
>
> Umm, no.
>
Bullshit. Read my post to Alan. Learn something.

Cheers,
Dick Johnson
Penguin : Linux version 2.6.9 on an i686 machine (5537.79 BogoMips).
  Notice : All mail here is now cached for review by John Ashcroft.
                  98.36% of all statistics are fiction.

Re: keyboard timeout

[Alan Cox]

On Iau, 2004-12-02 at 13:11, linux-os wrote:
> >From original IBM specification...
> "...8284A Clock generator clock generator supplies the multiphase
> clock signals that are needed to drive the microprocessor and
> the peripherals. Its base frequency is 14.31818 MHz....."

And the ISA bus side is 8.33Mhz (4.77Mhz on original PC, 6Mhz on
original
XT, originally 10MHz put rapidly fixed back to 8 on some others).

The 1Mhz clock to the keyboard controller is where the required delays
for talking to it come from. Similarly for the other low frequency parts
hacked into the PC (DMA etc).

Our keyboard code appears correct in all respects so you probably want
to adjust the various pc_kbd delays or look to see if you can find one
missing that matters to your PC as I doubt anyone using a post 1990
computer can actually help debug it because their hardware simply won't
care.

Alan

Re: keyboard timeout

[Dave Dillow]

On Thu, 2004-12-02 at 08:13, linux-os wrote:
> On Wed, 1 Dec 2004, Dave Dillow wrote:
> 
> > On Wed, 2004-12-01 at 17:29, linux-os wrote:
> >> If Linux 2.6.9 is booted on a 40 MHz `486 with the standard
> >> ISA clock of 14.3 MHz (yes that's the standard), the kernel
> >> will complain about a keyboard timeout for every key touched!
> >
> > Umm, no.
> >
> Bullshit. Read my post to Alan. Learn something.

And you should've read the rest of mine. Just because the clock
generator is 14.3MHz doesn't mean that's what the bus runs at -- all
signals a relative to BCLK, which is 4.77MHz, 8Mhz, or 8.33Mhz (or 6, or
10, by Alan's last post).

By your logic, the PCI bus in my PIII machine runs at 100MHz because
it's divided down from the 100MHz Front Side Bus.
-- 
Dave Dillow <dave@thedillows.org>

Re: keyboard timeout

[Alan Cox]

On Iau, 2004-12-02 at 13:13, linux-os wrote:
> Bullshit. Read my post to Alan. Learn something.

Yes I've read your post, I've learned you are an idiot.

Now back to something useful