FW: PROBLEM: Devices behind PCI Express-to-PCI bridge not mapped

FW: PROBLEM: Devices behind PCI Express-to-PCI bridge not mapped

[Hodle, Brian]

-----Original Message-----
From: Hodle, Brian 
Sent: Tuesday, June 21, 2005 8:53 AM
To: 'Peter Buckingham'
Subject: RE: PROBLEM: Devices behind PCI Express-to-PCI bridge not
mapped


Peter,
	I am experiencing exactly the same problem. I am using an ASUS
K8N-DL MB with the x86_64 kernel. My PCIX devices are not allocated
correctly. I tried using the  'pci=routeirq' option to no avail. Disabling
ACPI in the BIOS does not help the situation either. X will not use my PCIX
for GLX since none of the  extra txture memory has been allocated! Anyone
have any ideas?

regards,

Brian

-----Original Message-----
From: Peter Buckingham [mailto:peter@pantasys.com]
Sent: Monday, June 20, 2005 4:31 PM
To: Ivan Kokshaysky
Cc: sean.bruno@dsl-only.net; koch@esa.informatik.tu-darmstadt.de;
torvalds@osdl.org; benh@kernel.crashing.org;
linux-pci@atrey.karlin.mff.cuni.cz; linux-kernel@vger.kernel.org;
gregkh@suse.de
Subject: Re: PROBLEM: Devices behind PCI Express-to-PCI bridge not
mapped


Hi Ivan,

I've just tried a recent pull from Linus post 2.6.12. It seems that the 
bar sizes are now (mostly) correct. However, there are still issues with 
the resources failing to be allocated and the bars being disabled. I've 
attached the latest dmesg and lspci -vvx to see whether there's any 
enlightenment out there...

thanks,

peter

Re: FW: PROBLEM: Devices behind PCI Express-to-PCI bridge not mapped

[Peter Buckingham]

Hi Brian,

Hodle, Brian wrote:
> 	I am experiencing exactly the same problem. I am using an ASUS
> K8N-DL MB with the x86_64 kernel. My PCIX devices are not allocated
> correctly. I tried using the  'pci=routeirq' option to no avail. Disabling
> ACPI in the BIOS does not help the situation either. X will not use my PCIX
> for GLX since none of the  extra txture memory has been allocated! Anyone
> have any ideas?

well my system is using PCIe instead so it's a little different. It 
seems that the PCIe fixups are enough to get the BAR regions assigned 
correctly. prior to loading the nvidia driver the BARs are listed as 
disabled in lspci -vvx, but after loading the BARs are not disabled and 
I am able run X on both GPUs fine. I haven't tried running any OpenGL 
type tests yet on the system.

pci=routeirq is a red herring, this is about the interrupts and 
shouldn't affect the bar allocation.

it might be useful for you to post a dmesg with PCI_DEBUG enabled and 
specify which kernel version you are using. My system only worked with 
the most recent 2.6.12 (but i think that's due to some of the pci 
express changes and some of the bridge handling since our system is a 
little unique...)

peter

Re: FW: PROBLEM: Devices behind PCI Express-to-PCI bridge not mapped

[Karim Yaghmour]

Hodle, Brian wrote:
> 	I am experiencing exactly the same problem. I am using an ASUS
> K8N-DL MB with the x86_64 kernel. My PCIX devices are not allocated
> correctly. I tried using the  'pci=routeirq' option to no avail. Disabling
> ACPI in the BIOS does not help the situation either. X will not use my PCIX
> for GLX since none of the  extra txture memory has been allocated! Anyone
> have any ideas?

I've got a K8N-DL here myself and it's got a bunch of issues, some of
which seem to be outside the realm of the OS. Like others have
mentioned, the NIC and USB are out of reach.

Just for reference:
- hda = DVD
- hdc = regular IDE drive
- sda = SATA drive
- video = ATI Radeon X300 (MSI-branded)

For quite some time, I struggled trying to get any FCx to install, but
they would all fail trying to mount the DVD on secondary IDE master.
Then I swapped a regular IDE drive I have on primary IDE master with
the DVD and then the install went through. Nevertheless, Linux still
can't deal with the secondary IDE, the HD produces DMA timeouts and
eventually the kernel just shuts down the DMA on hdc. It's worth
pointing out that winxp64 which used to boot when DVD was hdc, now
simply continues showing the progress bar forever (with the IDE drive
as hdc). It too doesn't like the secondary IDE. For now, the only way
to get a clean boot in either OS is to for me to disable secondary IDE
altogether.

In all cases, Linux just can't deal with the main SATA drive if it's
connected to the ck804. It used to just freeze loading sata_nv, now
it just stays there when printing info about the disk during boot.
The machine is still responding (ctrl-alt-del does get to the kernel),
but it can't finish the bootup process. Eventually, I had to plug
the sata to the Silicon Image controller to get it to work with
Linux. Unfortunately, winxp64 has got no problem with booting off a
SATA drive connected to the same ck804 :(

When vendor installed winxp64, they reported having to use an nvidia
video card instead of the ATI one, otherwise they couldn't install
the drivers for the chipset (ck804) for "some obscure reason", as
they said. I'm guessing they were just getting the same problems I
was seing with Linux. This could actually be a hint of a conflict
between the PCIe video card and the secondary IDE. Others I've
spoken to who have this board don't report problems on the secondary
IDE, but they don't have the same card on the PCIe.

I'm guessing this board is in need of a BIOS upgrade, but I've got
the latest one from ASUS and, as others mentioned, their tech
support is so-so.

... just thought I'd write this down for others who may be having
similar problems ...

Karim
-- 
Author, Speaker, Developer, Consultant
Pushing Embedded and Real-Time Linux Systems Beyond the Limits
http://www.opersys.com || karim@opersys.com || 1-866-677-4546

Re: FW: PROBLEM: Devices behind PCI Express-to-PCI bridge not mapped

[Sean Bruno]

On Sat, 2005-06-25 at 13:02 -0700, Sean Bruno wrote:
> On Sat, 2005-06-25 at 15:40 -0400, Karim Yaghmour wrote:
> > I've got a K8N-DL here myself and it's got a bunch of issues, some of
> > which seem to be outside the realm of the OS. Like others have
> > mentioned, the NIC and USB are out of reach.
> > 
> > Just for reference:
> > - hda = DVD
> > - hdc = regular IDE drive
> > - sda = SATA drive
> > - video = ATI Radeon X300 (MSI-branded)
> > 
> > For quite some time, I struggled trying to get any FCx to install, but
> > they would all fail trying to mount the DVD on secondary IDE master.
> > Then I swapped a regular IDE drive I have on primary IDE master with
> > the DVD and then the install went through. Nevertheless, Linux still
> > can't deal with the secondary IDE, the HD produces DMA timeouts and
> > eventually the kernel just shuts down the DMA on hdc. It's worth
> > pointing out that winxp64 which used to boot when DVD was hdc, now
> > simply continues showing the progress bar forever (with the IDE drive
> > as hdc). It too doesn't like the secondary IDE. For now, the only way
> > to get a clean boot in either OS is to for me to disable secondary IDE
> > altogether.
> 
> > In all cases, Linux just can't deal with the main SATA drive if it's
> > connected to the ck804. It used to just freeze loading sata_nv, now
> > it just stays there when printing info about the disk during boot.
> > The machine is still responding (ctrl-alt-del does get to the kernel),
> > but it can't finish the bootup process. Eventually, I had to plug
> > the sata to the Silicon Image controller to get it to work with
> > Linux. Unfortunately, winxp64 has got no problem with booting off a
> > SATA drive connected to the same ck804 :(
> > 
> 
I got FC4 to install using the Silicon Image SATA controller and an IDE
CDROM.  I also inserted a PCI 10/100 card to get network support.  After
installing FC4(or any distro for that matter), edit your inittab before 
you reboot and set it to boot into run level 3.  Once you have rebooted,
you can run an update(via yum) to get the latest kernel and Xorg stuff.
This allowed me to boot the machine into level 5 and my video card did
function(NV 6200).
> 
> > 
> > When vendor installed winxp64, they reported having to use an nvidia
> > video card instead of the ATI one, otherwise they couldn't install
> > the drivers for the chipset (ck804) for "some obscure reason", as
> > they said. I'm guessing they were just getting the same problems I
> > was seing with Linux. This could actually be a hint of a conflict
> > between the PCIe video card and the secondary IDE. Others I've
> > spoken to who have this board don't report problems on the secondary
> > IDE, but they don't have the same card on the PCIe.
> > 
> > I'm guessing this board is in need of a BIOS upgrade, but I've got
> > the latest one from ASUS and, as others mentioned, their tech
> > support is so-so.
> > 
> > ... just thought I'd write this down for others who may be having
> > similar problems ...
> > 
> > Karim

Indeed, ASUS is kind of "sucking" in the tech support realm on this
issue.  I have been poking around with "dmidecode" and "lshw" and found
some really interesting stuff.

Handle 0x0023
        DMI type 9, 13 bytes.
        System Slot Information
                Designation: PCIEX1_1
                Type: <OUT OF SPEC><OUT OF SPEC>
                Current Usage: In Use
                Length: Other
                Characteristics:
                        5.0 V is provided
                        PME signal is supported
Handle 0x0024
        DMI type 9, 13 bytes.
        System Slot Information
                Designation: PCIEX16_1
                Type: <OUT OF SPEC><OUT OF SPEC>
                Current Usage: Available
                Length: Other
                Characteristics:
                        5.0 V is provided
                        PME signal is supported

I am assuming that this is causing some kind of weirdness that disables
the use of the NV SATA, USB, Sound and the Broadcom ethernet device.
Any ideas what the "OUT OF SPEC" warning actually means?

Sean

Re: FW: PROBLEM: Devices behind PCI Express-to-PCI bridge not mapped

[Karim Yaghmour]

Sean Bruno wrote:
> I got FC4 to install using the Silicon Image SATA controller and an IDE
> CDROM.  I also inserted a PCI 10/100 card to get network support.  After
> installing FC4(or any distro for that matter), edit your inittab before 
> you reboot and set it to boot into run level 3.  Once you have rebooted,
> you can run an update(via yum) to get the latest kernel and Xorg stuff.
> This allowed me to boot the machine into level 5 and my video card did
> function(NV 6200).

I actually never had problems with the video working, it always did
work. Given the information I have at hand, though, the video card
I'm using is apparently conflicting with the secondary IDE.

> I am assuming that this is causing some kind of weirdness that disables
> the use of the NV SATA, USB, Sound and the Broadcom ethernet device.
> Any ideas what the "OUT OF SPEC" warning actually means?

Well I don't know whether this is what is causing the weirdness that's
creating the ck804 problems, but I can confirm that I do get those exact
same "OUT OF SPEC" messages when running dmidecode.

Here's what google turns up:
https://bugzilla.redhat.com/bugzilla/attachment.cgi?id=114778

Looking at the dmidecode sources, it seems that it reports "OUT OF SPEC"
for anything it doesn't recognize. There's not much more about it in the
sources or the docs, from what I can see that is.

Karim
-- 
Author, Speaker, Developer, Consultant
Pushing Embedded and Real-Time Linux Systems Beyond the Limits
http://www.opersys.com || karim@opersys.com || 1-866-677-4546

Re: FW: PROBLEM: Devices behind PCI Express-to-PCI bridge not mapped

[Sean Bruno]

> 
> Here's what google turns up:
> https://bugzilla.redhat.com/bugzilla/attachment.cgi?id=114778

lol...look who the bug submitter for that issue is...  :)

Sean