[parisc-linux] Re: ide controller

[parisc-linux] Re: ide controller

[Grant Grundler]

Moritz Sinn wrote:
> 
> hi,
> 
> i've got an promise PDC20268 ide controller, which i would like to use
> in my hp c360 workstation. i'm using kernel 2.4.18 (couldn't compile
> 2.4.19 and 2.4.20-pre7).
> if i compile with "Generic PCI bus-master DMA support", the kernel
> crashes when i insmod the 'ide-mod' module. it doesn't print useable
> error messages, it just doesn't react anymore.

Are you using the "apt-get source kernel-image-2.4.18-hppa"
or from kernel.org?

If the latter, that won't work.

> when i compile without dma support, i can insmod the 'ide-mod' and
> 'ide-disk' module, but it crashes when i want to insmod the
> 'ide-probe-mod' module.
> 
> would be great if someone could help me.

1) see cvs.parisc-linux.org to get the latest bits
   (or ftp.parisc-linux.org/cvs/)

2) define the following to 1 in drivers/ide/pdc202xx.c:

#define PDC202XX_DEBUG_DRIVE_INFO               0
#define PDC202XX_DECODE_REGISTER_INFO           0

  and rebuild the module with that.

3) post the console log and "ser pim" output at PDC prompt.

grant

Re: [parisc-linux] ide controller

[j.steindlberger]

Hi Grant,

together with Moritz I/we tried to get some new results. We took the latest
kernel (-pa18) and did the changes to the PDC202XX-definitions in pdc202xx.c.
But I don't believe that the kernel gets the chance to recognice this
change. I think the kernel crashes earlier. We reduced the kernel by the specific
drivers. We disabled the following:
[ ]     Generic PCI bus-master DMA support
After compiling wie had these three modules:
ide-mod.o
ide-probe-mod.o
ide-disk.o
An insmod ide-mod.o was possible. But after insmod ide-probe-mod the kernel
crashes. In some cases we got an output with about 50 pages of memory
addresses. At the beginning we got some messages told about the SYM-SCSI-device.
Could that be a conflict? We did not succeed to reproduce that and log it.

Here my questions:
Did You here about anyone to succeed in using an IDE-controller in a
HP-RISC-architecture?
Do You see a chance to use IDE-disks in future (with some none
HP-controller)
Do You know about more people who are interrested in what we tried to do?
That would be an argument for me to continue experimenting.

Thank You
Joerg

> #define PDC202XX_DEBUG_DRIVE_INFO               0
> #define PDC202XX_DECODE_REGISTER_INFO           0
> 
>   and rebuild the module with that.
> 
> 3) post the console log and "ser pim" output at PDC prompt.

-- 
GMX - Die Kommunikationsplattform im Internet.
http://www.gmx.net

Re: [parisc-linux] ide controller

[Alan Cox]

On Thu, 2002-09-19 at 09:41, j.steindlberger@gmx.de wrote:
> together with Moritz I/we tried to get some new results. We took the latest
> kernel (-pa18) and did the changes to the PDC202XX-definitions in pdc202xx.c.
> But I don't believe that the kernel gets the chance to recognice this
> change. I think the kernel crashes earlier. We reduced the kernel by the specific
> drivers. We disabled the following:
> [ ]     Generic PCI bus-master DMA support
> After compiling wie had these three modules:
> ide-mod.o
> ide-probe-mod.o
> ide-disk.o
> An insmod ide-mod.o was possible. But after insmod ide-probe-mod the kernel
> crashes. In some cases we got an output with about 50 pages of memory
> addresses. At the beginning we got some messages told about the SYM-SCSI-device.
> Could that be a conflict? We did not succeed to reproduce that and log it.

Our IDE code has been run on non x86 platforms. It does require that the
defaults are set right for the platform so that it doesnt do stuff like
blindly probe I/O 0x170. 

I don't know if the Promise card needs its BIOS to have run, I would
suspect not but can't be sure.

Re: [parisc-linux] ide controller

[J.Steindlberger]

On Thursday 19 September 2002 12:06, you wrote:
> Our IDE code has been run on non x86 platforms. It does require that the
> defaults are set right for the platform so that it doesnt do stuff like
> blindly probe I/O 0x170.
>
> I don't know if the Promise card needs its BIOS to have run, I would
> suspect not but can't be sure.

Hi Alan,

we tried a different controller with originally an eprom bios. We simply put 
that eeprom away. (The Promise has a chip-integrated flash bios - putting 
that away seems not to be possible.) The kernel crashed alike. Do You have 
any ideas to get more debug output? I could set up my second C-Class with 
parisc-linux and test it there. (The first one is now running some essential 
services - continued crashing would not be good.)

Then I will have a deep look into the linux-code (oh, I'll have to learn C) 
and search for that stupid (sorry) 0x170 probe.

Thanks a lot! You will read more...
Joerg

Re: [parisc-linux] ide controller

[Grant Grundler]

"J.Steindlberger" wrote:
> The kernel crashed alike. Do You have any ideas to get more debug output?

If the crash was an "HPMC", then "ser pim" at PDC prompt will
tell you machine state and why it crashed. If the crash
was a regular panic, the console output will tell the reason, where,
and give a register dump. Post either or both of those here
and we can try to decode it for you.

grant

Re: [parisc-linux] ide controller

[Grant Grundler]

j.steindlberger@gmx.de wrote:
> drivers. We disabled the following:
> [ ]     Generic PCI bus-master DMA support

hmm....ok.

> After compiling wie had these three modules:
> ide-mod.o
> ide-probe-mod.o
> ide-disk.o

No pdc202xx.o?

> An insmod ide-mod.o was possible. But after insmod ide-probe-mod the kernel
> crashes. In some cases we got an output with about 50 pages of memory
> addresses. At the beginning we got some messages told about the SYM-SCSI-devi
>   ce.
> Could that be a conflict?

I don't think so.

> We did not succeed to reproduce that and log it.
> 
> Here my questions:
> Did You here about anyone to succeed in using an IDE-controller in a
> HP-RISC-architecture?

Yes. My C3000 has IDE CD-ROM hanging off the "SuckyIO" chip.
That works using PIO mode thanks to Randolph Chung/Thomas Bogendoerfer.

> Do You see a chance to use IDE-disks in future (with some none
> HP-controller)

yes - but I don't expect it to "just work".
IDE is a PoS with lots of buggy HW.
PARISC isn't very tolerant of many things the x86 PCs tolerate.
(eg access to disabled devices, PCI parity errors)

> Do You know about more people who are interrested in what we tried to do?

I don't offhand. If someone is, I ask they contact you directly.
Please continue to post status here and I'll offer advice when I can.

> That would be an argument for me to continue experimenting.

If you get it working, I'd be happy to help put any changes back
into our source tree and ultimately back to Alan Cox/Linus.

...
> > #define PDC202XX_DEBUG_DRIVE_INFO               0
> > #define PDC202XX_DECODE_REGISTER_INFO           0
> > 
> >   and rebuild the module with that.
> > 
> > 3) post the console log and "ser pim" output at PDC prompt.

Using a serial console makes it easier to grab console logs in
the case were the machine is crashing.

Generally, I can help decode HPMCs and explain what might be wrong.
That's what the "ser pim" output is for.

I really need both to debug most IO problems.


grant