[parisc-linux] ASP Parallel port driver broken on HP9000/730

[parisc-linux] ASP Parallel port driver broken on HP9000/730

[M. Grabert]

Hello there,

As I said before (in another mail), the ASP parallel port driver
is broken: the kernel hangs (lastest CVS, one hour ago).
Everything's fine just without ASP parallel port driver!

Unfortunately I don't know how to redirect the kernel output
to the serial console, so I can give you only some hints...



parport_init_chip: initializing bidirectional mode.

The kernel makes a memory dump from 0000 to 097f,
then:

High Priority Machine Check (HPMC): Code=1 regs=c034a080 (Addr=00000000)

Then follows a dump of the processor registers:

PSW: 00000000000001000000000000001111
r0-3   00000000 c02eb800 c0246328 c1f30788
r4-7   0000002c c1fe5220 f0824800 ffffffdf
r8-11  ffffffff 00000000 0004000e c037d8b0
r12-15 000a5a59 000ff000 f0000124 00000000
r16-19 00000001 50000059 0a000a00 0000000f
r20-23 000000ff 00000041 c1fdc000 0000000f
r24-27 ffffffff f0824800 f0824800 c0342000
r28-31 00000034 00000800 c1fe0980 c012a7dc
sr0-3  00000000 00000000 00000000 00000000
sr4-7  00000000 00000000 00000000 00000000

IASQ: 00000000 00000000 IAOQ: c026e458 c026e458
 IIR: e840c000     ISR: 00000000   IOR: c0246328
ORIG_R28: 00000000


NB: i typed all this stuff, no copy'n'paste; so perhaps
there are some errors ... 

BTW, no HIL connected, just serial console (but since
the kernel hangs before, i cannot see any output :))).
No PC parport selected, no support for foreign h/w,
no IEEE 1284 transfer modes ...)


hope that helps!


greetings max

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[Grant Grundler]

"M. Grabert" wrote:
> Unfortunately I don't know how to redirect the kernel output
> to the serial console, so I can give you only some hints...

Hints on using serial console on 715 (you can search the archive
for lots more 715 hints):

a) disable CONFIG_STI_CONSOLE
b) enabled CONFIG_SERIAL=y
           CONFIG_SERIAL_CONSOLE=y
           CONFIG_SERIAL_GSC=y

c) connect x86 serial to port 1 of 715 and run minicom with 9600 8N1
   serial line settings.
d) Disconnect HIL keyboard and Mouse.
   (*or* hold the TOC button for about 10 seconds while powering on the box.)
   (Either should force the PDC to use serial port 1)

> parport_init_chip: initializing bidirectional mode.
> 
> The kernel makes a memory dump from 0000 to 097f,
> then:
> 
> High Priority Machine Check (HPMC): Code=1 regs=c034a080 (Addr=00000000)
> 
> Then follows a dump of the processor registers:
> 
> PSW: 00000000000001000000000000001111
> r0-3   00000000 c02eb800 c0246328 c1f30788
> r4-7   0000002c c1fe5220 f0824800 ffffffdf
> r8-11  ffffffff 00000000 0004000e c037d8b0
> r12-15 000a5a59 000ff000 f0000124 00000000
> r16-19 00000001 50000059 0a000a00 0000000f
> r20-23 000000ff 00000041 c1fdc000 0000000f
> r24-27 ffffffff f0824800 f0824800 c0342000
> r28-31 00000034 00000800 c1fe0980 c012a7dc
> sr0-3  00000000 00000000 00000000 00000000
> sr4-7  00000000 00000000 00000000 00000000
> 
> IASQ: 00000000 00000000 IAOQ: c026e458 c026e458
>  IIR: e840c000     ISR: 00000000   IOR: c0246328
> ORIG_R28: 00000000

IOAQ is the address of the instruction that the CPU
was trying to execute when the HPMC occurred.
IIR is the instruction at IOAQ.
echo 0xe840c000=i | adb   says:
		BV              r0(r2)

GR02 is the most recent subroutine call return addr.
(Typically the caller).

So look up c026e458 and c0246328 in your System.map.
You can at least figure out about where it's crashing.
(Hint: cvs co build-tools and look at a.c)

> NB: i typed all this stuff, no copy'n'paste; so perhaps
> there are some errors ... 
> 
> BTW, no HIL connected, just serial console (but since
> the kernel hangs before, i cannot see any output :))).

Wait. If PDC is printing to the serial console...I'm confused.

grant

Grant Grundler
parisc-linux {PCI|IOMMU|SMP} hacker
+1.408.447.7253

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[M. Grabert]

On Thu, 8 Feb 2001, Grant Grundler wrote:

> Hints on using serial console on 715 (you can search the archive
> for lots more 715 hints):
> 
> a) disable CONFIG_STI_CONSOLE
> b) enabled CONFIG_SERIAL=y
>            CONFIG_SERIAL_CONSOLE=y
>            CONFIG_SERIAL_GSC=y
> 
> c) connect x86 serial to port 1 of 715 and run minicom with 9600 8N1
>    serial line settings.
> d) Disconnect HIL keyboard and Mouse.
>    (*or* hold the TOC button for about 10 seconds while powering on the box.)
>    (Either should force the PDC to use serial port 1)

sorry about the confusing error report.
I'm actually using serial console, with the configuration above;
but the initial kernel messages are not redirected to the serial console.
not until the serial ports are found & initialized ...
... but since these kernel messages are printed before the time the
serial ports are recognized, I can't see them on the terminal, but on the
monitor.
My question was, whether there is a kernel parameter (i can pass to palo),
which redirects ALL kernel messages from the beginning to the serial port
(before they are regognized by the linux kernel). 

> GR02 is the most recent subroutine call return addr.
> (Typically the caller).
> 
> So look up c026e458 and c0246328 in your System.map.
> You can at least figure out about where it's crashing.
> (Hint: cvs co build-tools and look at a.c)

I'll do this tonight ...
 
> > NB: i typed all this stuff, no copy'n'paste; so perhaps
> > there are some errors ... 
> > 
> > BTW, no HIL connected, just serial console (but since
> > the kernel hangs before, i cannot see any output :))).

... on the serial console, hence the serial ports are not initialized,
but i can see the messages on the monitor! therefore I had to type
all the register dump by hand ...

> Wait. If PDC is printing to the serial console...I'm confused.

I hope you're not confused anymore ;)

greetings, max

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[Helge Deller]

On Friday 09 February 2001 14:45, M. Grabert wrote:
>
> sorry about the confusing error report.
> I'm actually using serial console, with the configuration above;
> but the initial kernel messages are not redirected to the serial console.
> not until the serial ports are found & initialized ...
> ... but since these kernel messages are printed before the time the
> serial ports are recognized, I can't see them on the terminal, but on the
> monitor.
> My question was, whether there is a kernel parameter (i can pass to palo),
> which redirects ALL kernel messages from the beginning to the serial port
> (before they are regognized by the linux kernel).

on the IPL screen type 

PATH CONSOLE RS232_1
*or*
PATH CONSOLE SERIAL_1

At the next boot PDC will redirect all PDC-messages to the first serial port 
and you will able to boot your PARISC through minicom.

Helge


BTW: Max, your domain doesn't seem to resolve 
	("M. Grabert" <xam@deathsdoor.com>)!
bash-2.04$ nslookup deathsdoor.com
Server:  devserv.devel.redhat.com
*** devserv.devel.redhat.com can't find deathsdoor.com: Non-existent 
host/domain

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[M. Grabert]

On Fri, 9 Feb 2001, Helge Deller wrote:

> on the IPL screen type 
> 
> PATH CONSOLE RS232_1
> *or*
> PATH CONSOLE SERIAL_1

thanks alot !!!

> BTW: Max, your domain doesn't seem to resolve 
> 	("M. Grabert" <xam@deathsdoor.com>)!
> bash-2.04$ nslookup deathsdoor.com
> Server:  devserv.devel.redhat.com
> *** devserv.devel.redhat.com can't find deathsdoor.com: Non-existent 
> host/domain

my nslookup says:

Name:   deathsdoor.com
Address: 206.96.221.142


well, deathsdoor.com is a email-forward servive from netforward.com
I'm using it since 1995/96 (about).
Since I'll finish my study in March, I'll lose my email address there,
therefore I use this email address ...

my real address is markus dot grabert at stud dot fh-regensburg dot de
:))) 

geeting max

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[Grant Grundler]

"M. Grabert" wrote:
> My question was, whether there is a kernel parameter (i can pass to palo),
> which redirects ALL kernel messages from the beginning to the serial port
> (before they are regognized by the linux kernel). 

No. (I'm being more blunt than Helge).

Read the directions I posted again and try them.

Setting the CONSOLE path (as suggested by Helge) will result in
the same thing. But later you must set the path back to a graphics
head once you (and parisc-linux) are ready for STI/graphical console.

grant

Grant Grundler
parisc-linux {PCI|IOMMU|SMP} hacker
+1.408.447.7253

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[M. Grabert]

On Thu, 8 Feb 2001, Grant Grundler wrote:

> echo 0xe840c000=i | adb   says:
> 		BV              r0(r2)
> 
> GR02 is the most recent subroutine call return addr.
> (Typically the caller).
> 
> So look up c026e458 and c0246328 in your System.map.
> You can at least figure out about where it's crashing.

done that (I also checked that is is the correct System.map),
but there is no entry for c026e458 and c0246328 !!!
Also build a new kernel (and a new System.map), but
the same result :(

> (Hint: cvs co build-tools and look at a.c)

I did not tried this yet ...
 

greetings max

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[Matthew Wilcox]

On Sun, Feb 11, 2001 at 09:16:39PM +0100, M. Grabert wrote:
> On Thu, 8 Feb 2001, Grant Grundler wrote:
> 
> > echo 0xe840c000=i | adb   says:
> > 		BV              r0(r2)
> > 
> > GR02 is the most recent subroutine call return addr.
> > (Typically the caller).
> > 
> > So look up c026e458 and c0246328 in your System.map.
> > You can at least figure out about where it's crashing.
> 
> done that (I also checked that is is the correct System.map),
> but there is no entry for c026e458 and c0246328 !!!

no, there wouldn't be.  the addresses in System.map are the start of
each function.  so the address below c026e458 will be the address of
the function which made the call.

Matthew `Give a man a match and you warm him for an hour; dose him in
petrol first and you warm him for the rest of his life'

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[M. Grabert]

On Mon, 12 Feb 2001, Matthew Wilcox wrote:

> > On Thu, 8 Feb 2001, Grant Grundler wrote:
> > 
> > > echo 0xe840c000=i | adb   says:
> > > 		BV              r0(r2)
> > > 
> > > GR02 is the most recent subroutine call return addr.
> > > (Typically the caller).
> > > 
> > > So look up c026e458 and c0246328 in your System.map.
> > > You can at least figure out about where it's crashing.

just a copy'n'paste from my System.map:

c026e44c T _gsc_readb
c026e460 T _gsc_readw
c026e474 T _gsc_readl

therefore c026e458 is in _gsc_readb, and:

c02462c4 t parport_PS2_supported
c02463c8 T parport_gsc_probe_port
c024677c T parport_gsc_write_data

 so c0246328 is in parport_PS2_supported!


greetings max

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[Richard Hirst]

I can reproduce this on a 715/75.  When it crashes I get:

  ASP version 1 at 0xf0800000 found.
  LED (ASP-style) display at f0800020 registered
  Found HIL at 0xf0821000, IRQ 94
  HIL: keyboard found at id 3
  parport_init_chip: initialize bidirectional-mode.
  Dumping Stack from c5ff0000 to c5ff0980:


IAOQ = _gsc_readb + 0x0c
r2 = parport_PS2_supported + 0x64
r26 = 0xf0824800

r26 corresponds to

  Scorpio Sr. Core Centronics (10) at 0xf0824000, versions 0x19, 0x0, 0x74, 0x0, 0x0

[parisc-linux] "M. Grabert" <xam@deathsdoor.com> delivery problems

[Richard Hirst]

<xam@deathsdoor.com>: host Mail.deathsdoor.com[206.96.221.110] said: 550 5.1.1
    <xam@deathsdoor.com>... User unknown

Richard

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[Helge Deller]

On Wednesday 14 February 2001 17:49, Richard Hirst wrote:
> I can reproduce this on a 715/75.  When it crashes I get:
>
>   ASP version 1 at 0xf0800000 found.
>   LED (ASP-style) display at f0800020 registered
>   Found HIL at 0xf0821000, IRQ 94
>   HIL: keyboard found at id 3
>   parport_init_chip: initialize bidirectional-mode.
>   Dumping Stack from c5ff0000 to c5ff0980:
>
>
> IAOQ = _gsc_readb + 0x0c
> r2 = parport_PS2_supported + 0x64
> r26 = 0xf0824800
>
> r26 corresponds to
>
>   Scorpio Sr. Core Centronics (10) at 0xf0824000, versions 0x19, 0x0, 0x74,
> 0x0, 0x0
>
>
> From drivers/parport/parport_gsc.c:
>
>         port = ((unsigned long) d->hpa) + PARPORT_GSC_OFFSET;
>
>         /*
>             some older machines with ASP-chip don't support the enhanced
> parport modes */
>         if (!pdc_add_valid( (void *)(port+4))) {
>             /* Initialize bidirectional-mode (0x10) & data-tranfer-mode #1
> (0x20) */ printk("%s: initialize bidirectional-mode.\n", __FUNCTION__);
> parport_writeb ( (0x10 + 0x20), port + 4);
>         } else {
>             printk("%s: enhanced parport-modes not supported.\n",
> __FUNCTION__); }
>
> So, I guess it shouldn't have tried to initialise bidirectional-mode
> on my old 715/75 with ASP.  If I force that 'if' to fail, then on reboot
> I get:
>
>   ASP version 1 at 0xf0800000 found.
>   LED (ASP-style) display at f0800020 registered
>   Found HIL at 0xf0821000, IRQ 94
>   HIL: keyboard found at id 3
>   parport_init_chip: enhanced parport-modes not supported.
>   parport0: PC-style at 0xf0824800, irq 88 [PCSPP]

On my 715/64 I get with bidirectional mode enabled:
parport0: PC-style at 0xf0102800, irq 88 [PCSPP,TRISTATE]
and this (TRISTATE) is why I wanted to enable this mode. 

> And the machine boots ok.  I added debug just to check, (port+4) is
> 0xf0824804.  So is pdc_add_valid() broken, or is that not the right
> address to be testing?

I'm still sure, that this is the correct address to test, but maybe we should 
just disable the initialisation of bidirectional mode for machines which
- have an ASP chip, or
- have their parport-chip at address 0xf0824800 (which is the only region 
where the ASP has it's parport), or
- on machines, where "parisc_get_cpu_type(cpu_data[0]) < pcxt_"  (from 
hardware.h and processor.h) [but I'm not sure if pcxt_ is correct. You'll 
need to verify]

This are just some possible solutions to work around the problem for now, but 
I'm sure there may be some others too.

> Oh, and a simple 'ls / > /dev/lp0' worked too.
Great.

> Richard

Greetings,

	Helge

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[Richard Hirst]

On Thu, Feb 15, 2001 at 01:01:35AM +0100, Helge Deller wrote:
> I'm still sure, that this is the correct address to test, but maybe we should 
> just disable the initialisation of bidirectional mode for machines which
> - have an ASP chip, or
> - have their parport-chip at address 0xf0824800 (which is the only region 
> where the ASP has it's parport), or
> - on machines, where "parisc_get_cpu_type(cpu_data[0]) < pcxt_"  (from 
> hardware.h and processor.h) [but I'm not sure if pcxt_ is correct. You'll 
> need to verify]

That parisc_get_cpu_type() bit isn't quite right, how about:

diff -u -r1.7 parport_gsc.c
--- parport_gsc.c       2001/01/25 00:01:32     1.7
+++ parport_gsc.c       2001/02/15 09:42:08
@@ -482,7 +482,7 @@
        /* 
            some older machines with ASP-chip don't support the enhanced parport modes 
        */
-       if (!pdc_add_valid( (void *)(port+4))) {
+       if (boot_cpu_data.cpu_type > pcxt && !pdc_add_valid( (void *)(port+4))) {
            /* Initialize bidirectional-mode (0x10) & data-tranfer-mode #1 (0x20) */
            printk("%s: initialize bidirectional-mode.\n", __FUNCTION__);
            parport_writeb ( (0x10 + 0x20), port + 4);


Works on my 715/75, where boot_cpu_data.cpu_type = 2 = pcxt.

Richard

Re: [parisc-linux] ASP Parallel port driver broken on HP9000/730

[Helge Deller]

On Thursday 15 February 2001 10:54, Richard Hirst wrote:
> On Thu, Feb 15, 2001 at 01:01:35AM +0100, Helge Deller wrote:
> > I'm still sure, that this is the correct address to test, but maybe we
> > should just disable the initialisation of bidirectional mode for machines
> > which - have an ASP chip, or
> > - have their parport-chip at address 0xf0824800 (which is the only region
> > where the ASP has it's parport), or
> > - on machines, where "parisc_get_cpu_type(cpu_data[0]) < pcxt_"  (from
> > hardware.h and processor.h) [but I'm not sure if pcxt_ is correct. You'll
> > need to verify]
>
> That parisc_get_cpu_type() bit isn't quite right, how about:
>
> diff -u -r1.7 parport_gsc.c
> --- parport_gsc.c       2001/01/25 00:01:32     1.7
> +++ parport_gsc.c       2001/02/15 09:42:08
> @@ -482,7 +482,7 @@
>         /*
>             some older machines with ASP-chip don't support the enhanced
> parport modes */
> -       if (!pdc_add_valid( (void *)(port+4))) {
> +       if (boot_cpu_data.cpu_type > pcxt && !pdc_add_valid( (void
> *)(port+4))) { /* Initialize bidirectional-mode (0x10) & data-tranfer-mode
> #1 (0x20) */ printk("%s: initialize bidirectional-mode.\n", __FUNCTION__);
> parport_writeb ( (0x10 + 0x20), port + 4);

Looks good. 
Could you commit ?

> Works on my 715/75, where boot_cpu_data.cpu_type = 2 = pcxt.
> Richard

Greetings,

	Helge