[9fans] too fast key repeat in pcpae 2cpu mode

[9fans] too fast key repeat in pcpae 2cpu mode

[kokamoto]

Eric,

I reported I had no problem to use pcpae kernel so far.
This time, I must report a problem.

I switched from 1CPU mode to 2CPU mode, and have
problem of too much high speed of keyboard input,
like aaaaaaaaaaa when I typed only one 'a'.

Please check this, I want to use pcpae kernel for
CPU server soon.

Kenji

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[erik quanstrom]

On Sat Apr  5 03:58:14 EDT 2014, kokamoto@hera.eonet.ne.jp wrote:
> Eric,
>
> I reported I had no problem to use pcpae kernel so far.
> This time, I must report a problem.
>
> I switched from 1CPU mode to 2CPU mode, and have
> problem of too much high speed of keyboard input,
> like aaaaaaaaaaa when I typed only one 'a'.
>
> Please check this, I want to use pcpae kernel for
> CPU server soon.

i assume this is a usb keyboard, and happens on
any key?  and i'll assume that you're using 9atom's
usb/usbd built recently.  is this what you're doing?

can you run sos/sos and copy or email the output?
sos itself gives instructions.

also, check with a watch that sleep 10 or sleep 60
tracks correctly.

i'm fairly confident in the usb/kb repeat process itself.
i rewrote it because i was frustrated with this very
problem.  you are using usb and 9atom's usb/kb, right?

the repeat mechanism is in /sys/src/cmd/usb/kb/kb.c:^/repeatproc
and :^/repeattimerproc.  the initial repeat delay
is 500msec, and the subsequent delay is 30ms.
which should be 33 repeats/sec.

sources uses 160 and 25 (!) and warns

	 * Rewrite instead of debug, if needed.

you can verify the repeat rate tracks correctly with the
system clock by running clock -s in one window and
typing

	echo <tick>| freq

hold down the 'a'  where <tick> is for 10s exactly.
we would expect

	1 + 9.5*(1000/30)  = 317 'a's

on my machine, i get about 270-280.  likely because
kbin is unbuffered.

- erik

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[kokamoto]

> can you run sos/sos and copy or email the output?
> sos itself gives instructions.

I'll email you soon.

However, one point I should report here is that it may not be
concerned only keyboard.   I experienced the one window
for fscons did not open, and some error messages ( I forgot)
were there.   I suppose this is not related to keyboard, right?

Yes, I'm using your new usb/kb (not mine) with usb Japanese
keyboard.

Kenji

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[erik quanstrom]

> However, one point I should report here is that it may not be
> concerned only keyboard.   I experienced the one window
> for fscons did not open, and some error messages ( I forgot)
> were there.   I suppose this is not related to keyboard, right?

i'm going to start off with the assumption that the console
error messages are related.

- erik

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[erik quanstrom]

these errors

	init: starting /bin/rc
	i8042: fe returned to the f5 command
	i8042: fe returned to the f6 command

usually happen when mouseport=ps2 or mouseport=ps2intellimouse
when there is only a a usb mouse.  set mouseport=usb.

if you tried all the other debugging suggestions, you could also type
f7, which will print all the usb kbd input on the console.

i don't see any issues in the sos output.

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[kokamoto]

> if you tried all the other debugging suggestions, you could also type
> f7, which will print all the usb kbd input on the console.
>
> i don't see any issues in the sos output.

My 2cpu machine goes 10 times faster than 1cpu.
This happens on also the pc kernel (not pcpae).

Kenji

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[erik quanstrom]

On Tue Apr  8 22:58:29 EDT 2014, kokamoto@hera.eonet.ne.jp wrote:
> > if you tried all the other debugging suggestions, you could also type
> > f7, which will print all the usb kbd input on the console.
> >
> > i don't see any issues in the sos output.
>
> My 2cpu machine goes 10 times faster than 1cpu.
> This happens on also the pc kernel (not pcpae).

so basically the timing loop is messed up.  if you could run the amd64
kernel on this machine, this will not happen.  i forgot if that's possible.

if it's not possible, i'll try to take a look tomorrow.

- erik

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[kokamoto]

> so basically the timing loop is messed up.  if you could run the amd64
> kernel on this machine, this will not happen.

Before I do this, I tried the 'sources' pc kernel for 2 cpu.
It recognizes 2 cpus and has no problem.

Kenji

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[erik quanstrom]

On Tue Apr  8 23:12:24 EDT 2014, kokamoto@hera.eonet.ne.jp wrote:
> > so basically the timing loop is messed up.  if you could run the amd64
> > kernel on this machine, this will not happen.
>
> Before I do this, I tried the 'sources' pc kernel for 2 cpu.
> It recognizes 2 cpus and has no problem.

you can try using the pc i8253.c for pcpae, and see if that solves
the issue.

- erik

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[kokamoto]

> you can try using the pc i8253.c for pcpae, and see if that solves
> the issue.

As you know, it's not so easy.
For example, pc i8253.c kernel wants watchdog(), and
different Mach structure which you changed, etc.

I know I lost about 800MB of my memory.☺
However, stability is the most important factor for
me now.

Kenji

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[erik quanstrom]

> > you can try using the pc i8253.c for pcpae, and see if that solves
> > the issue.
> 
> As you know, it's not so easy.
> For example, pc i8253.c kernel wants watchdog(), and
> different Mach structure which you changed, etc.

i ment the i8253.c from the atom pc kernel.  i thought
that was understood.

> I know I lost about 800MB of my memory.☺
> However, stability is the most important factor for
> me now.

not nearly as much fun to work on stuff if there's no
possibility for resolution.  i have not seen this issue
on the ~40 machines i've run the pae kernel on.

- erik

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[kokamoto]

>> > you can try using the pc i8253.c for pcpae, and see if that solves
>> > the issue.
>>
>> As you know, it's not so easy.
>> For example, pc i8253.c kernel wants watchdog(), and
>> different Mach structure which you changed, etc.
>
> i ment the i8253.c from the atom pc kernel.  i thought
> that was understood.

As I wrote before, the atom pc kernel has the same problem,
runs ten times faster.

Kenji

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[kokamoto]

I forgot.
This mail is using 'sources' pc kernel for 2 cpu, and
nupas mail system.

Kenji

Re: [9fans] too fast key repeat in pcpae 2cpu mode

[Bruce Ellis]

[-- Attachment #1: Type: text/plain, Size: 328 bytes --]

Fixed!
On 09/04/2014 1:13 PM, <kokamoto@hera.eonet.ne.jp> wrote:

> > so basically the timing loop is messed up.  if you could run the amd64
> > kernel on this machine, this will not happen.
>
> Before I do this, I tried the 'sources' pc kernel for 2 cpu.
> It recognizes 2 cpus and has no problem.
>
> Kenji
>
>
>

[-- Attachment #2: Type: text/html, Size: 602 bytes --]