[parisc-linux] Problem find devices on the E55 I/O interfaces

[parisc-linux] Problem find devices on the E55 I/O interfaces

[Christoph Plattner]

Hello PARISC hackers !

I have problems in the device probing on the E55, and I need some
hints here. I am at the very starting point of the SCSI driver
development, and other (Rayn) will also need this for MUX (serial),
etc....

The E55 is a machine of the type "SYSTEM MAP TYPE".
I instrumented the code to read out the values and elements at the
inventory probing. I will append the log of my "debug helps", they
are quite self explaining.

The `walk_central_bus()' calls seems to do correct work. It finds the 
basic elements on my E55:

	On the "56" the bus converter
	On "62" and "63" components on the system (main) board
	- "62" is the CPU, "63" the Memory

	The search was started at the "60".

My problem now. The `walk_lower_bus()' under the device "56" cannot
find any further devices. Seen on PDC or HP-UX, under "56/52" the
SCSI controller is present, further the parallel interface and the 
serial MUX also should be found.

Do we use the wrong search address here, taken from io_io_low of
the `struct bc_module'. Do we access it in a wrong way ?
Is THIS structure documented in any of the PARISC documents, other-
wise, where does this structure come from ?

I hope you have some hints here ...

Best regards
Christoph P.

-- 
-------------------------------------------------------
private:	christoph.plattner@gmx.at
company:	christoph.plattner@alcatel.at

Re: [parisc-linux] Problem find devices on the E55 I/O interfaces

[Grant Grundler]

Christoph Plattner wrote:
> Do we use the wrong search address here, taken from io_io_low of
> the `struct bc_module'. Do we access it in a wrong way ?
> Is THIS structure documented in any of the PARISC documents, other-
> wise, where does this structure come from ?

Ryan Bradetich has a pretty good handle on how to walk bus converters.
I expect the bus converter register layout to be documented in
the PA IO ACD (Arch Control Document). Sorry - I don't have a URL
handy for the IO ACD.

grant

Re: [parisc-linux] Problem find devices on the E55 I/O interfaces

[Ryan Bradetich]

Hello Christoph,

Sorry for the late reply, had a great time at OLs.


On Wed, 2002-06-26 at 15:51, Christoph Plattner wrote:
> Hello PARISC hackers !
> 
> I have problems in the device probing on the E55, and I need some
> hints here. I am at the very starting point of the SCSI driver
> development, and other (Rayn) will also need this for MUX (serial),
> etc....
> 
> The E55 is a machine of the type "SYSTEM MAP TYPE".
> I instrumented the code to read out the values and elements at the
> inventory probing. I will append the log of my "debug helps", they
> are quite self explaining.
> 
> The `walk_central_bus()' calls seems to do correct work. It finds the 
> basic elements on my E55:
> 
> 	On the "56" the bus converter
> 	On "62" and "63" components on the system (main) board
> 	- "62" is the CPU, "63" the Memory
> 
> 	The search was started at the "60".
> 
> My problem now. The `walk_lower_bus()' under the device "56" cannot
> find any further devices. Seen on PDC or HP-UX, under "56/52" the
> SCSI controller is present, further the parallel interface and the 
> serial MUX also should be found.

This make sense.....

What type of bus is at 56?  The walk_native_bus (as the comments
and function name indicates) only walks native busses.  It does
look like this function should be called and work for this bus
since it is a bus converter instead of a bus adapter.

let me know what type of your central bus (I'm assuming GSC), and
the what is the type of your lower bus and I'll see what I can
find for you.

> Do we use the wrong search address here, taken from io_io_low of
> the `struct bc_module'. Do we access it in a wrong way ?
> Is THIS structure documented in any of the PARISC documents, other-
> wise, where does this structure come from ?

This structure is a specialized version of the io_module structure.
I have not found any external documentation describing it, I have
mainly figure out it out from scattered internal documentation and
HP-UX source.

Not many hints, but I'll work with you to get this figured out.

Thanks,

- Ryan


> I hope you have some hints here ...
> 
> Best regards
> Christoph P.
> 
> -- 
> -------------------------------------------------------
> private:	christoph.plattner@gmx.at
> company:	christoph.plattner@alcatel.at
> 
> 
> _______________________________________________
> parisc-linux mailing list
> parisc-linux@lists.parisc-linux.org
> http://lists.parisc-linux.org/mailman/listinfo/parisc-linux
> 

Re: [parisc-linux] Problem find devices on the E55 I/O interfaces

[Christoph Plattner]

Hello Ryan !

Really nice to here from you !

Did you do some hacks on the MUX (of your type of machine) ?

In one of my mails I sent a log with my instrumented debug output,
so you may be able to read out something concerning the question
of bus type.

Here is one question: Although I a have hardware knowledge, I am
not fix HP terminology here.
Can you give me some hints here with the kewords "main bus", 
"lower bus", etc...

Is the "main bus" the CPU bus which has connected the on board 
devices and has a bridge to the the backpanel, or without bridge 
to the backpanel ?
Or is the "main bus" the bus behind the bridge running on the back-
panel ?
Is the lower bus a bus on the interface card behind the bridge to 
the backpanel bus, or ... ?
Is the bus converter itself a bridge ?
Is it transperent bridge or has it to be initialized and is special
drivers needed, etc...

A ASCII block diagram would be helpful, only very simple like

   +-+                    +-+
   | |----- main bus -----| |----- backpanel bus ------------------
   +-+                    +-+    |                 |           |
  CPU (core)           Bridge    I/F 1            I/F 2       I/F 3
				+-+
				| |
				+-+  Bus Converter
				 |
				 +-------------------------- I/O bus
					|          |	
					MUX 	 SCSI

This is not true for the E55, it is only an EXAMPLE !!!

Together we will solve it, or ?

Christoph




Ryan Bradetich wrote:
> 
> Hello Christoph,
> 
> Sorry for the late reply, had a great time at OLs.
> 
> On Wed, 2002-06-26 at 15:51, Christoph Plattner wrote:
> > Hello PARISC hackers !
> >
> > I have problems in the device probing on the E55, and I need some
> > hints here. I am at the very starting point of the SCSI driver
> > development, and other (Rayn) will also need this for MUX (serial),
> > etc....
> >
> > The E55 is a machine of the type "SYSTEM MAP TYPE".
> > I instrumented the code to read out the values and elements at the
> > inventory probing. I will append the log of my "debug helps", they
> > are quite self explaining.
> >
> > The `walk_central_bus()' calls seems to do correct work. It finds the
> > basic elements on my E55:
> >
> >       On the "56" the bus converter
> >       On "62" and "63" components on the system (main) board
> >       - "62" is the CPU, "63" the Memory
> >
> >       The search was started at the "60".
> >
> > My problem now. The `walk_lower_bus()' under the device "56" cannot
> > find any further devices. Seen on PDC or HP-UX, under "56/52" the
> > SCSI controller is present, further the parallel interface and the
> > serial MUX also should be found.
> 
> This make sense.....
> 
> What type of bus is at 56?  The walk_native_bus (as the comments
> and function name indicates) only walks native busses.  It does
> look like this function should be called and work for this bus
> since it is a bus converter instead of a bus adapter.
> 
> let me know what type of your central bus (I'm assuming GSC), and
> the what is the type of your lower bus and I'll see what I can
> find for you.
> 
> > Do we use the wrong search address here, taken from io_io_low of
> > the `struct bc_module'. Do we access it in a wrong way ?
> > Is THIS structure documented in any of the PARISC documents, other-
> > wise, where does this structure come from ?
> 
> This structure is a specialized version of the io_module structure.
> I have not found any external documentation describing it, I have
> mainly figure out it out from scattered internal documentation and
> HP-UX source.
> 
> Not many hints, but I'll work with you to get this figured out.
> 
> Thanks,
> 
> - Ryan
> 
> > I hope you have some hints here ...
> >
> > Best regards
> > Christoph P.
> >
> > --
> > -------------------------------------------------------
> > private:      christoph.plattner@gmx.at
> > company:      christoph.plattner@alcatel.at
> >
> >
> > _______________________________________________
> > parisc-linux mailing list
> > parisc-linux@lists.parisc-linux.org
> > http://lists.parisc-linux.org/mailman/listinfo/parisc-linux
> >
> 
> _______________________________________________
> parisc-linux mailing list
> parisc-linux@lists.parisc-linux.org
> http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

-- 
-------------------------------------------------------
private:	christoph.plattner@gmx.at
company:	christoph.plattner@alcatel.at

Re: [parisc-linux] Problem find devices on the E55 I/O interfaces

[Grant Grundler]

Christoph Plattner wrote:
> Here is one question: Although I a have hardware knowledge, I am
> not fix HP terminology here.
> Can you give me some hints here with the kewords "main bus", 
> "lower bus", etc...

"central bus" = bus which connects all CPUs with all Memory.
On N-class this would be the Merced Bus.
On C3000 it's the runway bus.
On 712 is the GSC bus.

> Is the lower bus a bus on the interface card behind the bridge to 
> the backpanel bus, or ... ?

"lower bus" is usually the secondary bus below a bus adapter or bus converter.

Bus adapter and Bus converter differ in that one is transperent
to SW and the other is not. Eg. PCI Host Bus Adapter.
Both are forms of bridges.

hth,
grant

Re: [parisc-linux] Problem find devices on the E55 I/O interfaces

[Matthew Wilcox]

On Tue, Jul 02, 2002 at 03:25:01PM -0600, Grant Grundler wrote:
> Christoph Plattner wrote:
> > Here is one question: Although I a have hardware knowledge, I am
> > not fix HP terminology here.
> > Can you give me some hints here with the kewords "main bus", 
> > "lower bus", etc...
> 
> "central bus" = bus which connects all CPUs with all Memory.
> On N-class this would be the Merced Bus.
> On C3000 it's the runway bus.
> On 712 is the GSC bus.

E class is probably different somehow.  It has a PCX-L chip in it, so the
only external bus from that chip is the GSC bus.  Look at the ioscan output
from the hwdb:

:pa:F:F:F:-1:-1:4294967295:bc:root:::0:root:root:CLAIMED:BUS_NEXUS::0
pa:pa:F:F:F:-1:-1:4294967295:bc:bus_converter:56:80 48 0 71 0 0 12 0 0 0 0 0 0 0 0 0 :1:root.bc:bc:CLAIMED:BUS_NEXUS:Bus Converter:1
pa:sio:F:F:F:-1:-1:4294967295:ext_bus:scsi3:56/40:0 64 128 132 0 0 59 128 6 0 0 0 43 6 0 2 :0:root.bc.scsi3:scsi3:CLAIMED:INTERFACE:HP 28696A - Wide SCSI ID=7:0
pa:sio:F:F:F:-1:-1:4294967295:ext_bus:scsi1:56/52:4 64 128 132 0 0 57 128 4 0 0 0 21 136 0 3 :1:root.bc.scsi1:scsi1:CLAIMED:INTERFACE:HP 28655A - SE SCSI ID=7:1pa:sio:F:T:F:-1:76:131072:ext_bus:cent0:56/53:0 64 0 5 0 0 58 128 0 0 0 0 0 0 0 0 :2:root.bc.lpr2:lpr2:CLAIMED:INTERFACE:HP 28655A - Parallel Interface:2
pa:sio:F:T:F:-1:193:0:tty:mux2:56/56:4 64 128 133 16 0 13 0 3 0 0 0 80 252 0 2 :0:root.bc.mux2:mux2:CLAIMED:INTERFACE:MUX:0
pa:core:F:F:F:-1:-1:4294967295:ba:core_adapter:60:1 17 128 11 0 0 129 0 0 0 0 0 0 0 0 0 :0:root.bus_adapter:bus_adapter:CLAIMED:BUS_NEXUS:Core I/O Adapter:0
core:wsio:F:T:F:-1:52:0:lan:lan2:60/6:1 17 128 138 0 0 138 0 2 0 0 0 219 151 0 2 :0:root.bus_adapter.lan2:lan2:CLAIMED:INTERFACE:Built-in LAN:0
pa:pa:F:F:F:-1:-1:4294967295:processor:processor:62:72 49 0 0 0 0 4 129 0 0 0 0 0 0 0 0 :0:root.processor:processor:CLAIMED:PROCESSOR:Processor:0
pa:pa:F:F:F:-1:-1:4294967295:memory:memory:63:2 80 31 1 0 0 9 0 0 0 0 0 0 0 0 0 :0:root.memory:memory:CLAIMED:MEMORY:Memory:0

(I trimmed out the SCSI devices.)

The device at iopath 60 is clearly Lasi.  The device at iopath 56 is
_reported as_ Java BC GSC+ Port which is obviously wrong.  Ditto the E45
(claims to have a Uturn-IOA BC GSC+ Port), E35 (U2-IOA BC GSC+ Port)
and E25 (GeckoBOA BC GSC+ Port).  Anyone on this list know what device
was _actually_ used in the E class machines?  And do they have docs
for it?  ;-)

I expect that the bus below 56/ is actually an HP-PB bus because of the
devies it claims to have on it.

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] Problem find devices on the E55 I/O interfaces

[Christoph Plattner]

So, the devices below may be the HP-PB bus, as you mentioned.
The resulting question: is this HP-PB bus now transperent ?
Why are the devices not found ?

The point confusing me, is that the PDC boot monitor can display the
hardware paths below the "56", but Linux calling PDC calls cannot see
those devices, independent, if the device type is really know or only
found as unknown.

What is the reason here.

So I thought the interpretion of the data structure is wrong for those
machines ....

Christoph


Matthew Wilcox wrote:
> 
> On Tue, Jul 02, 2002 at 03:25:01PM -0600, Grant Grundler wrote:
> > Christoph Plattner wrote:
> > > Here is one question: Although I a have hardware knowledge, I am
> > > not fix HP terminology here.
> > > Can you give me some hints here with the kewords "main bus",
> > > "lower bus", etc...
> >
> > "central bus" = bus which connects all CPUs with all Memory.
> > On N-class this would be the Merced Bus.
> > On C3000 it's the runway bus.
> > On 712 is the GSC bus.
> 
> E class is probably different somehow.  It has a PCX-L chip in it, so the
> only external bus from that chip is the GSC bus.  Look at the ioscan output
> >from the hwdb:
> 
> :pa:F:F:F:-1:-1:4294967295:bc:root:::0:root:root:CLAIMED:BUS_NEXUS::0
> pa:pa:F:F:F:-1:-1:4294967295:bc:bus_converter:56:80 48 0 71 0 0 12 0 0 0 0 0 0 0 0 0 :1:root.bc:bc:CLAIMED:BUS_NEXUS:Bus Converter:1
> pa:sio:F:F:F:-1:-1:4294967295:ext_bus:scsi3:56/40:0 64 128 132 0 0 59 128 6 0 0 0 43 6 0 2 :0:root.bc.scsi3:scsi3:CLAIMED:INTERFACE:HP 28696A - Wide SCSI ID=7:0
> pa:sio:F:F:F:-1:-1:4294967295:ext_bus:scsi1:56/52:4 64 128 132 0 0 57 128 4 0 0 0 21 136 0 3 :1:root.bc.scsi1:scsi1:CLAIMED:INTERFACE:HP 28655A - SE SCSI ID=7:1pa:sio:F:T:F:-1:76:131072:ext_bus:cent0:56/53:0 64 0 5 0 0 58 128 0 0 0 0 0 0 0 0 :2:root.bc.lpr2:lpr2:CLAIMED:INTERFACE:HP 28655A - Parallel Interface:2
> pa:sio:F:T:F:-1:193:0:tty:mux2:56/56:4 64 128 133 16 0 13 0 3 0 0 0 80 252 0 2 :0:root.bc.mux2:mux2:CLAIMED:INTERFACE:MUX:0
> pa:core:F:F:F:-1:-1:4294967295:ba:core_adapter:60:1 17 128 11 0 0 129 0 0 0 0 0 0 0 0 0 :0:root.bus_adapter:bus_adapter:CLAIMED:BUS_NEXUS:Core I/O Adapter:0
> core:wsio:F:T:F:-1:52:0:lan:lan2:60/6:1 17 128 138 0 0 138 0 2 0 0 0 219 151 0 2 :0:root.bus_adapter.lan2:lan2:CLAIMED:INTERFACE:Built-in LAN:0
> pa:pa:F:F:F:-1:-1:4294967295:processor:processor:62:72 49 0 0 0 0 4 129 0 0 0 0 0 0 0 0 :0:root.processor:processor:CLAIMED:PROCESSOR:Processor:0
> pa:pa:F:F:F:-1:-1:4294967295:memory:memory:63:2 80 31 1 0 0 9 0 0 0 0 0 0 0 0 0 :0:root.memory:memory:CLAIMED:MEMORY:Memory:0
> 
> (I trimmed out the SCSI devices.)
> 
> The device at iopath 60 is clearly Lasi.  The device at iopath 56 is
> _reported as_ Java BC GSC+ Port which is obviously wrong.  Ditto the E45
> (claims to have a Uturn-IOA BC GSC+ Port), E35 (U2-IOA BC GSC+ Port)
> and E25 (GeckoBOA BC GSC+ Port).  Anyone on this list know what device
> was _actually_ used in the E class machines?  And do they have docs
> for it?  ;-)
> 
> I expect that the bus below 56/ is actually an HP-PB bus because of the
> devies it claims to have on it.
> 
> --
> Revolutions do not require corporate support.
> 
> _______________________________________________
> parisc-linux mailing list
> parisc-linux@lists.parisc-linux.org
> http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

-- 
-------------------------------------------------------
private:	christoph.plattner@gmx.at
company:	christoph.plattner@alcatel.at

Re: [parisc-linux] Problem find devices on the E55 I/O interfaces

[Matthew Wilcox]

On Wed, Jul 03, 2002 at 12:26:53AM +0200, Christoph Plattner wrote:
> So, the devices below may be the HP-PB bus, as you mentioned.
> The resulting question: is this HP-PB bus now transperent ?
> Why are the devices not found ?
> 
> The point confusing me, is that the PDC boot monitor can display the
> hardware paths below the "56", but Linux calling PDC calls cannot see
> those devices, independent, if the device type is really know or only
> found as unknown.

Ah!  Now I understand why you were so confused ;-)

Yes, PDC tells us about some devices, but not all.  We have to walk
the native busses to find the remaining devices.  What's a native bus?
Uhh.. can't remember, ask grant.  The central bus will be native and
I'm sure GSC is also native.  I believe HP-PB must be native too, since
I seem to remember rbrad finding more devices once he walked the HP-PB
busses on his K class.

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] Problem find devices on the E55 I/O interfaces

[Grant Grundler]

Christoph Plattner wrote:
> So, the devices below may be the HP-PB bus, as you mentioned.
> The resulting question: is this HP-PB bus now transperent ?

The bus converter is "transparent", not the bus.
ie everything we need to know about the secondary bus, we should be able
to read from the bus converter.

> Why are the devices not found ?
> 
> The point confusing me, is that the PDC boot monitor can display the
> hardware paths below the "56", but Linux calling PDC calls cannot see
> those devices, independent, if the device type is really know or only
> found as unknown.

Sounds like the bus walk code is looking at the wrong address
for walking the lower BC port.

I'll point Ryan at the IO ACD (ioacd_v096.pdf only seems to be
available inside HP *sigh*) and he can sort it out.

grant