Re: Re: IDE in Mode 1 (Sandpoint) how ?

Re: Re: IDE in Mode 1 (Sandpoint) how ?

[Sarnath Kannan]

Hi,
> Sounds like you have the switches in the wrong position
> and the interrupts
> are not routed correctly.  The serial port will
> eventually time out when it
> doesn't receive interrupts, giving the perception of a
> very slow running system.
>
> Just keep in mind the troubles are local to your
> configuration, as this software
> operates fine on a properly configured board.  Just
> move switches until it works,
> then leave them alone :-).

  Ur suggestion is acceptable. But whatss troubling
my head is, the kernel boots up fine until the NFS mount.
Had the switch settings been wrong, the serial output
would be slow from early bootup, right ?  But thats
not the case. Why should it be slow after INIT starts
up ?
 Meanwhile, I am also trying the same with different
switch settings.

Thanks
Sarnath


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: IDE in Mode 1 (Sandpoint) how ?

[Dan Malek]

Sarnath Kannan wrote:


> ..... But whatss troubling
> my head is, the kernel boots up fine until the NFS mount.
> Had the switch settings been wrong, the serial output
> would be slow from early bootup, right ?

Nope.  All of the kernel print out is polled I/O.

> .....  But thats
> not the case. Why should it be slow after INIT starts
> up ?

Because it uses interrupt driven serial I/O.

Believe me, all of us using Sandpoints have seen this before.
Your interrupt/arbitration routing switches on the board are wrong.


	-- Dan


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Re: IDE in Mode 1 (Sandpoint) how ?

[Sarnath Kannan]

hi Dan,

> Nope.  All of the kernel print out is polled I/O.
>
  yeah ! Thats right.

>me: .....  But thats
>me: not the case. Why should it be slow after INIT
>me:starts
>me: up ?
>
> Because it uses interrupt driven serial I/O.

  I'am currently investigating that code path, Dan.
  We have also come to the same conclusion as u.
  But believe me, I had tried with all possible
  Switch settings, S3,S4,S5,S6. :-( Its tantalising !
  This means that all ISA interrupts are not
  being routed properly. One more, I haven't
  initialized the WinBond, coz reset values seem
  meaninful. Could that be the reason for
  this ?
>
> Believe me, all of us using Sandpoints have seen this
> before.
> Your interrupt/arbitration routing switches on the
> board are wrong.

  Any suggestion from experienced people is welcome.


Is there any other switch, other than the S3,S4,S5,S6
I need to worry about.

sarnath


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Re: IDE in Mode 1 (Sandpoint) how ?

[Sarnath Kannan]

>
> I forgot, is this an X2 or X3 (or X3b) version of the
> Sandpoint?
>
  Dan, This is an X2 board. I am able to
telnet to the machine and work cooly..But
serial is dead slow and the reason is as u said,
ISA interrupts.. They are not coming in.. Its
very clear. After a small investigation, I found
out that the serial chip times out on the
interrupt and "rs_timer" prints output in polled
mode. By changing the "Timeout value", I was
able to get faster output on the serial line.
  But the problem of ISA interrupts not being
routed is still there. This is very irritating.

Dan, I further noticed that "i8259_init" disables
all 8259 interrupts during initialization.
I thot this could be a reason and enabled all
8259 interrupts in a seperate loop. Even that
dint pay much except for a gush of ISA interrupts
immediately after enabling.

But what I am wondering is "How ISA interrupts
tick" for mvista kernel, when all 8259 interrupts
are disabled !! Am i missing sthg ?

Thanks,
Sincerely
Sarnath


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/