PCI video card hangs during system boot

PCI video card hangs during system boot

[Simone Mannori]

Dear All,

looking for a solution for a presistent issue (lack of any hardware
acceleration support for X11 because error during the loading of the
"radeon" or "nouveau" kernel module at boot time) I have bought a PCI
video card: ATI RADEON 9300 128M PCI. This test has been suggested by
many guys on several mailing list.

NOTA BENE: make sure that the PCI video card is 3.3V type or is an
"universal" (3.3V and 5.0V) model like mine.

I have put this video card leaving the ATI FireCL X1 installed; I have
removed the AGP video card; I have moved the PCI video card on
different positions.

No matter what I do, the c8000 is unable to boot.

Using the serial console, I have his message

<Cpu0> 0e000e7a00e00000  600601005d441002  CC_IODISC_PCI_DEVICE_CONFIG
pci_bus_walk line 2878 - bridge aperture too big
<Cpu0> 160012cb00e00000  0000000000000000  CC_MPS_SLAV.......
.......
Then the system boot process STOPs:
<Cpu1> e000108401e00000  0000000000000000  CC_BOOT_BOOT_FAILURE
<Cpu1> 030010d501e00000  0000000000000000  CC_CPU_STOP

This problem lie in the internal c8000 "system bios/diagnostic" code,
Linux boot process, kernel and modules are not involved.

Finally, the people that have suggested this test have not tried to make it.

Question: "Someone has tried with a NVDIA or MATROX PCI video cards?"

Thanks in advance for your help.

   Simone Mannori - ITALY

Re: PCI video card hangs during system boot

[Rolf Eike Beer]

[-- Attachment #1: Type: text/plain, Size: 1120 bytes --]

Am Dienstag, 12. April 2016, 10:20:17 schrieb Simone Mannori:
> Dear All,
> 
> looking for a solution for a presistent issue (lack of any hardware
> acceleration support for X11 because error during the loading of the
> "radeon" or "nouveau" kernel module at boot time) I have bought a PCI
> video card: ATI RADEON 9300 128M PCI. This test has been suggested by
> many guys on several mailing list.
> 
> NOTA BENE: make sure that the PCI video card is 3.3V type or is an
> "universal" (3.3V and 5.0V) model like mine.
> 
> I have put this video card leaving the ATI FireCL X1 installed; I have
> removed the AGP video card; I have moved the PCI video card on
> different positions.
> 
> No matter what I do, the c8000 is unable to boot.
> 
> Using the serial console, I have his message
> 
> <Cpu0> 0e000e7a00e00000  600601005d441002  CC_IODISC_PCI_DEVICE_CONFIG
> pci_bus_walk line 2878 - bridge aperture too big

This probably means the PCI BAR of the card is too big, i.e. it has too much 
RAM. Looks like the firmware can't handle that. I once had the same thing on a 
C3600, IIRC with a 128MB BAR.

Greetings,

Eike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: PCI video card hangs during system boot

[John David Anglin]

On 2016-04-12 12:55 PM, Rolf Eike Beer wrote:
> Am Dienstag, 12. April 2016, 10:20:17 schrieb Simone Mannori:
>> >Dear All,
>> >
>> >looking for a solution for a presistent issue (lack of any hardware
>> >acceleration support for X11 because error during the loading of the
>> >"radeon" or "nouveau" kernel module at boot time) I have bought a PCI
>> >video card: ATI RADEON 9300 128M PCI. This test has been suggested by
>> >many guys on several mailing list.
>> >
>> >NOTA BENE: make sure that the PCI video card is 3.3V type or is an
>> >"universal" (3.3V and 5.0V) model like mine.
>> >
>> >I have put this video card leaving the ATI FireCL X1 installed; I have
>> >removed the AGP video card; I have moved the PCI video card on
>> >different positions.
>> >
>> >No matter what I do, the c8000 is unable to boot.
>> >
>> >Using the serial console, I have his message
>> >
>> ><Cpu0> 0e000e7a00e00000  600601005d441002  CC_IODISC_PCI_DEVICE_CONFIG
>> >pci_bus_walk line 2878 - bridge aperture too big
> This probably means the PCI BAR of the card is too big, i.e. it has too much
> RAM. Looks like the firmware can't handle that. I once had the same thing on a
> C3600, IIRC with a 128MB BAR.
Maybe pulling FireGL X1 will help:
https://lists.freedesktop.org/archives/dri-users/2015-February/000034.html

Dave

-- 
John David Anglin  dave.anglin@bell.net

Re: PCI video card hangs during system boot

[Simone Mannori]

> Maybe pulling FireGL X1 will help:
> https://lists.freedesktop.org/archives/dri-users/2015-February/000034.html

Already done Dave: same error message during system boot. Just to be
clear, the error message is also indipendent by the PCI slot position.
Thanks for the hint, anyway.

My conclusion is that the "vanilla-PC" ATI RADEON 9200 128M PCI is not
"BIOS compatible" with the c8000.

Simone Mannori - ITALY

Re: PCI video card hangs during system boot

[John David Anglin]

On 2016-04-13 3:06 AM, Simone Mannori wrote:
>> Maybe pulling FireGL X1 will help:
>> https://lists.freedesktop.org/archives/dri-users/2015-February/000034.html
> Already done Dave: same error message during system boot. Just to be
> clear, the error message is also indipendent by the PCI slot position.
> Thanks for the hint, anyway.
Just one further thought regarding FireGL X1.  The OLS article by Grant 
referenced in the above
link indicated that a common problem in porting drivers to parisc was 
timing issues with I/O writes.
This was fixed in a couple of drivers by doing a read after doing a 
write.  Possibly, the radeon driver
could be hacked to always read after write as a test.

I also read somewhere ia64 linux using zx1 chipset did work at one 
time.  This link gives some history
on ia64 and might give some hints about other things to try:
http://comments.gmane.org/gmane.linux.debian.ports.ia64/3031

>
> My conclusion is that the "vanilla-PC" ATI RADEON 9200 128M PCI is not
> "BIOS compatible" with the c8000.
Agreed.  There's no way to change the c8000 BIOS.   I'd hoped that 
removing the X1 might free up
I/O space and allow 9200 to work.

Dave

-- 
John David Anglin  dave.anglin@bell.net

Re: PCI video card hangs during system boot

[Graham Gower]

On 14 April 2016 at 00:32, John David Anglin <dave.anglin@bell.net> wrote:
> On 2016-04-13 3:06 AM, Simone Mannori wrote:
>>>
>>> Maybe pulling FireGL X1 will help:
>>>
>>> https://lists.freedesktop.org/archives/dri-users/2015-February/000034.html
>>
>> Already done Dave: same error message during system boot. Just to be
>> clear, the error message is also indipendent by the PCI slot position.
>> Thanks for the hint, anyway.
>
> Just one further thought regarding FireGL X1.  The OLS article by Grant
> referenced in the above
> link indicated that a common problem in porting drivers to parisc was timing
> issues with I/O writes.
> This was fixed in a couple of drivers by doing a read after doing a write.
> Possibly, the radeon driver
> could be hacked to always read after write as a test.

I've tried this. It doesn't change the failure mode.

Re: PCI video card hangs during system boot

[John David Anglin]

On 2016-04-13, at 8:18 PM, Graham Gower wrote:

> On 14 April 2016 at 00:32, John David Anglin <dave.anglin@bell.net> wrote:
>> On 2016-04-13 3:06 AM, Simone Mannori wrote:
>>>> 
>>>> Maybe pulling FireGL X1 will help:
>>>> 
>>>> https://lists.freedesktop.org/archives/dri-users/2015-February/000034.html
>>> 
>>> Already done Dave: same error message during system boot. Just to be
>>> clear, the error message is also indipendent by the PCI slot position.
>>> Thanks for the hint, anyway.
>> 
>> Just one further thought regarding FireGL X1.  The OLS article by Grant
>> referenced in the above
>> link indicated that a common problem in porting drivers to parisc was timing
>> issues with I/O writes.
>> This was fixed in a couple of drivers by doing a read after doing a write.
>> Possibly, the radeon driver
>> could be hacked to always read after write as a test.
> 
> I've tried this. It doesn't change the failure mode.

Do you know if c8000 can setup any non HP card (e.g., ATI Rage XL)?  It was reported that
some old cards would work with PDC firmware in unaccelerated fb mode, but c8000 has
x86 emulator so it might be able to enable accelerated 2d/3d.

Dave
--
John David Anglin	dave.anglin@bell.net

Re: PCI video card hangs during system boot

[Graham Gower]

On 14 April 2016 at 09:59, John David Anglin <dave.anglin@bell.net> wrote:
> On 2016-04-13, at 8:18 PM, Graham Gower wrote:
>
>> On 14 April 2016 at 00:32, John David Anglin <dave.anglin@bell.net> wrote:
>>> On 2016-04-13 3:06 AM, Simone Mannori wrote:
>>>>>
>>>>> Maybe pulling FireGL X1 will help:
>>>>>
>>>>> https://lists.freedesktop.org/archives/dri-users/2015-February/000034.html
>>>>
>>>> Already done Dave: same error message during system boot. Just to be
>>>> clear, the error message is also indipendent by the PCI slot position.
>>>> Thanks for the hint, anyway.
>>>
>>> Just one further thought regarding FireGL X1.  The OLS article by Grant
>>> referenced in the above
>>> link indicated that a common problem in porting drivers to parisc was timing
>>> issues with I/O writes.
>>> This was fixed in a couple of drivers by doing a read after doing a write.
>>> Possibly, the radeon driver
>>> could be hacked to always read after write as a test.
>>
>> I've tried this. It doesn't change the failure mode.
>
> Do you know if c8000 can setup any non HP card (e.g., ATI Rage XL)?  It was reported that
> some old cards would work with PDC firmware in unaccelerated fb mode, but c8000 has
> x86 emulator so it might be able to enable accelerated 2d/3d.
>


I have not tested this. I vaguely recall getting the same "bridge
aperture too big" message once with a 128Mb radeon, but never obtained
older cards which might be compatible.

Re: PCI video card hangs during system boot

[Simone Mannori]

> Do you know if c8000 can setup any non HP card (e.g., ATI Rage XL)?  It was reported that
> some old cards would work with PDC firmware in unaccelerated fb mode, but c8000 has
> x86 emulator so it might be able to enable accelerated 2d/3d.

Good hint. I have several "vintage" PCI video cards to try (including
ATI RAGE, Vodoo, etc.) but I can't because the have 5V PCI interface.
HP c8000 have several PCI slots BUT PCI 3.3 volt ONLY. Find
"universal" 5V/3V PCI card is possible but the only cards that I have
find (the ATI 9200 128Mega) fails at "BIOS" time (see the first
message of this thread).

I'm accumulating several indirect evidences that the primary source of
the problem is not the video driver (both "radeon" and "nouveau" fail)
or the bus interface (degrading APG to PCI using video driver options
at Linux kernel boot time does not change the final results; the only
PCI video card that I was able to find, if fails at BIOS time).

Indirect evidences point the ZX1 chipset. For instance, on my c8000
with FireGL X1, if I add more than 8G DRAM, Linux boot fails (kernel
panic) during "radeon" loading. If I remove the video card and I use
the workstation as "server" via serial port or Ethernet, I can upgrade
to 16 Giga DRAM (8 x 2GB modules) with zero issue. I will open another
specific thread with the details soon.

Few days ago, I have mismatched the memory modules; the c8000 was able
to start Linux boot but I had a kernel panic during "radeon" loading.

As Helge has already point out, the only debug option is to recompile
the kernel with some "printk" in stategic point, hoping to be able to
find the primary source of this _very_nasty_ issue.

  Simone Mannori - ITALY

Re: PCI video card hangs during system boot

[John David Anglin]

On 2016-04-14 4:14 AM, Simone Mannori wrote:
>> Do you know if c8000 can setup any non HP card (e.g., ATI Rage XL)?  It was reported that
>> >some old cards would work with PDC firmware in unaccelerated fb mode, but c8000 has
>> >x86 emulator so it might be able to enable accelerated 2d/3d.
> Good hint. I have several "vintage" PCI video cards to try (including
> ATI RAGE, Vodoo, etc.) but I can't because the have 5V PCI interface.
> HP c8000 have several PCI slots BUT PCI 3.3 volt ONLY. Find
> "universal" 5V/3V PCI card is possible but the only cards that I have
> find (the ATI 9200 128Mega) fails at "BIOS" time (see the first
> message of this thread).
It seems to me that an AGP card might be a better option.  The C8000 
manual indicates that
it has one AGP 8X Pro 110W 32-bit slot.  It looks from connector pinout 
in the Technical Reference
Guide that the C8000 has a AGP 1.5V motherboard (pins 9, 16, 25 and 28 
are VCC 1.5).

The HP "supported" FireGL cards are 1.5V, 0.8V:
http://www.synthmind.com/737-18872ATIAGPGraphicsCards-AGPComplianceDetails.pdf

So, I think one needs to look for a 1.5V or Universal AGP card.  The 
first ATI AGP card to
support 1.%V appears to be RAGE 128 PRO.

These articles give info about AGP slots and what works electrically:
http://www.playtool.com/pages/agpcompat/agp.html
https://en.wikipedia.org/wiki/Accelerated_Graphics_Port

Dave

-- 
John David Anglin  dave.anglin@bell.net

Re: PCI video card hangs during system boot

[Simone Mannori]

>
> It seems to me that an AGP card might be a better option.  The C8000 manual
> indicates that
> it has one AGP 8X Pro 110W 32-bit slot.  It looks from connector pinout in
> the Technical Reference
> Guide that the C8000 has a AGP 1.5V motherboard (pins 9, 16, 25 and 28 are
> VCC 1.5).
>
> The HP "supported" FireGL cards are 1.5V, 0.8V:
> http://www.synthmind.com/737-18872ATIAGPGraphicsCards-AGPComplianceDetails.pdf
>
> So, I think one needs to look for a 1.5V or Universal AGP card.  The first
> ATI AGP card to
> support 1.%V appears to be RAGE 128 PRO.
>
> These articles give info about AGP slots and what works electrically:
> http://www.playtool.com/pages/agpcompat/agp.html
> https://en.wikipedia.org/wiki/Accelerated_Graphics_Port

Very interesting and useful info, but they don't help us to fix the basic issue.

We have several AGP cards able to work correctly BUT in passive frame
buffer only mode, therefore the c8000, a quite fast quad core
workstation , with X11, looks like a bradypus instead that a
velociraptor.

This is the moment to recompile a custom kernel with some printk.

     Simone Mannori - ITALY

Re: PCI video card hangs during system boot

[John David Anglin]

On 2016-04-14 11:54 AM, Simone Mannori wrote:
> Very interesting and useful info, but they don't help us to fix the basic issue.
>
> We have several AGP cards able to work correctly BUT in passive frame
> buffer only mode, therefore the c8000, a quite fast quad core
> workstation , with X11, looks like a bradypus instead that a
> velociraptor.
>
> This is the moment to recompile a custom kernel with some printk.
>
I tried this a year or so ago.  I pretty much convinced myself that the 
firmware loaded
correctly, and that read and write operations seemed to work. However, I 
wasn't able
to find why the the ring buffer test failed.  I also checked that the 
firmware was identical
to that used on hpux.

There is an ATI document somewhere on the register interface.

I understand your frustration.

Dave

-- 
John David Anglin  dave.anglin@bell.net