Bug 116364

Please supply the YaST logfiles as explained in:
 
http://www.opensuse.org/Bug_Reporting_FAQ#YaST

Please attach as well a hwinfo output from the 10.0 system.

I'm trying to install RC1 (32 bits) on my Dell PowerEdge 800, which is an EM64T.
It does not recognize the SATA devices in the machine. Attaching y2logs and hwinfo.

Created attachment 49557 [details]
y2logs

Created attachment 49558 [details]
hwinfo

Created attachment 49570 [details]
hwinfo output

Created attachment 49572 [details]
save_y2log output

Thomas...?

fehr, arvin: have a look at it, please.

2005-09-11 22:09:51 <1> linux(2974) [YCP] StorageControllers.ycp:520 Initialize
controller $["bus":"PCI", "bus_hwcfg":"pci", "class_id":1, "device":"82801FR/FRW
(ICH6R/ICH6RW) SATA Controller", "device_id":75346, "driver":"ahci",
"drivers":[$["active":true, "modprobe":true, "modules":[["ata_piix", ""]]],
$["active":true, "modprobe":true, "modules":[["ahci", ""]]], $["active":false,
"modprobe":true, "modules":[["generic", ""]]]], "func_id":2, "model":"Intel
82801FR/FRW (ICH6R/ICH6RW) SATA Controller",
"old_unique_key":"+9wv.qzt_27n5wPB", "prog_if":143,
"resource":$["io":[$["active":true, "length":8, "mode":"rw", "start":56560],
$["active":true, "length":4, "mode":"rw", "start":56552], $["active":true,
"length":8, "mode":"rw", "start":56408], $["active":true, "length":4,
"mode":"rw", "start":56544], $["active":true, "length":16, "mode":"rw",
"start":56384]], "irq":[$["count":100000, "enabled":true, "irq":3]],
"mem":[$["active":true, "length":1024, "start":3757834240]]], "rev":"3",
"slot_id":31, "sub_class_id":1, "sub_device_id":75346, "sub_vendor":"Intel
Corporation", "sub_vendor_id":98438, "sysfs_bus_id":"0000:00:1f.2",
"udi":"/org/freedesktop/Hal/devices/pci_8086_2652",
"unique_key":"w7Y8.qzt_27n5wPB", "vendor":"Intel Corporation", "vendor_id":98438]

StorageDevices.ycp:534 localProbe: all_disks []

Steffen SCR::Read(.probe.disk) does not return any disks but "piix" and "ata_piix"
seem to be proberly loaded.

In comment 4 the drivers don't work; instead it gives:   
    
<3>irq 3: nobody cared (try booting with the "irqpoll" option)   
   
(Maybe related to bug 105984?)   
   
Comment 5 is from a running system, not from installation time. It does  
not help.  
  
Comment 6 is completely broken, btw.  
  
I'd point my finger at ahci/ata_piix. 

The BIOS doesn't provide and irq information for the SATA controller.  
 
  <7>libata version 1.12 loaded. 
  <7>ahci version 1.01 
  <4>    ACPI-0402: *** Warning: Unable to derive IRQ for device 0000:00:1f.2 
  <4>ACPI: PCI Interrupt 0000:00:1f.2[C]: no GSI - using IRQ 3 
  <4>PCI: setting IRQ 3 as level-triggered 
  <7>PCI: Setting latency timer of device 0000:00:1f.2 to 64 
  <6>ahci(0000:00:1f.2) AHCI 0001.0000 32 slots 4 ports 1.5 Gbps 0xf impl IDE 
mode 
  <6>ahci(0000:00:1f.2) flags: 64bit stag led pmp  
  <6>ata1: SATA max UDMA/133 cmd 0xF883C100 ctl 0x0 bmdma 0x0 irq 3 
  <6>ata2: SATA max UDMA/133 cmd 0xF883C180 ctl 0x0 bmdma 0x0 irq 3 
  <6>ata3: SATA max UDMA/133 cmd 0xF883C200 ctl 0x0 bmdma 0x0 irq 3 
  <6>ata4: SATA max UDMA/133 cmd 0xF883C280 ctl 0x0 bmdma 0x0 irq 3 
  <3>irq 3: nobody cared (try booting with the "irqpoll" option) 

So this could be yet another ACPI problem...

I have tried it, unfortunately it did not work
(In reply to comment #11)
> The BIOS doesn't provide and irq information for the SATA controller.  
>  

>   <3>irq 3: nobody cared (try booting with the "irqpoll" option) 

Created attachment 49590 [details]
hwinfo output within RC1(installation)

Created attachment 49591 [details]
y2logs within RC1 (installation)

Could you also provide acpidmp output, please. I haven't had a look at irq
resources yet, but maybe I find something...
A dmesg output whether you have ACPI interpreter warnings/errors could also be
useful (boot with acpi_dbg_level=0xffffffff).

Created attachment 49608 [details]
acpidmp output

Created attachment 49609 [details]
dmesg output

In reply to comment #10, irqpoll works on the PowerEdge EM64T machine.

Downgrading since irqpoll works.  Not a blocker anymore.

Created attachment 50889 [details]
Disable MSI interrupt assignment when LAPIC/IOAPICs are not used.

Using irqpoll just covers up the PCI-MSI routing error.  MSI interrupts require
Local APIC + IO APIC support according to drivers/pci/Kconfig, so if "nosmp"
or "noapic" or "maxcpus=1" etc. are used on the kernel command line, prevent
MSI interrupt vectors from being assigned.

It looks like the 10.0 RC1 installer kernel is built with CONFIG_PCI_MSI=y but
without LOCAL_APIC or IO_APIC support.  At least that's what /proc/config.gz says.
How and why?  That seems to be what is causing this SATA installation problem,
with libata using a PCI-MSI interrupt.

Randy, what happened with this patch after the discussion on lkml?

Andrew merged it into the -mm patchset for testing.
But I don't see how CONFIG_PCI_MSI=y was enabled in the installer kernel.
These are separate but related issues.

Sorry, what's the status with this bug? I've tried to install with the pre-alpha3-release here and either the HDD are not found or the CD-Rom, depending on the mode ('AHCI' or 'compatibility'), but I am not sure if this is related to that bug.

I have just tried to install from a Kiosk-DVD (10.0): It still does not work

I have reported the same issue for 10.1 at
https://bugzilla.novell.com/show_bug.cgi?id=176249

Why is this on yast again?

Thomas, sorry for giving to you again, I hope you know who to reassign to (Steffen/kernel/...)

Could you please post whole dmesg (of a current kernel) output, if you can catch it over serial console or copy it away via network. apic=verbose or apic=debug kernel params might reveal why the local apic is not activated on this machine.

Moving product to 10.1, this should be fixed/verified in newer kernels first.
I compiled a kernel with randy's patch from comment #21. You can find it here:
ftp.suse.com/pub/people/trenn/disable_msi_without_ioapic_x86_64

Created attachment 83861 [details]
randy's patch rebased against our current CVS head kernel

Created attachment 83905 [details]
dmesg from 10.1 instllation.

Is this wright or do you need demsg from a working system, such as Mandrake, Ubuntu, Fedora or SuSE 9.3?

Created attachment 83907 [details]
boot.msg from 10.1 installation

We should probably also merge Randy's patch to allow disabling msi
on the command line. Can you reattach it?

Created attachment 83925 [details]
kernel boot parameter to disable MSI interrupt usage

This is at least partially merged already in
2.6.17-rc4.

We already seem to have the patch in #35 in the tree

As this bug/feature is relevant (it's the same hardware, the same problem) with the following...
https://bugzilla.novell.com/show_bug.cgi?id=176249#c13

So does it boot with pci=nomsi ? 

It boots with and without pci=nomsi but the HDDs are not recognized.
As far as I have understood from  https://bugzilla.novell.com/show_bug.cgi?id=176249#c13
the problem is the libata (read comment #13). 
I have ATA HDDs on SATA and this is supported in SuSE 9.3 by a libata devel branch  which is not in the rest versions.

Something for Jens to figure out then.

We can't help you for 10.1, unfortunately.

*** This bug has been marked as a duplicate of 176249 ***