Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Adrian Bunk]

On Wed, Nov 08, 2006 at 01:44:29PM +0100, Olivier Nicolas wrote:

> Hi,

Hi Olivier,

> 2.6.19-rc5 does not boot properly, I have tried pci=routeirq, irpoll 
> without success.
> 
> Full details (.config, dmesg, /proc/interrupts) are in 
> http://olivn.trollprod.org/2.6.19-rc5-irq.tar.gz

thanks for your report!

I might be wrong, but looking at the dmesg:
- irq 22 is the hda_intel IRQ
- the "irq 22: nobody cared" is immediately before the
  "hda_intel: No response from codec, disabling MSI..."
- in the routeirq case, the hda_intel IRQ as well as the
  IRQ in the error message change to 21

So it might be related to the hda_intel MSI check.

If this was a wrong guess, other interesting messages in the dmesg are:

<--  snip  -->

...
PCI: Using MMCONFIG at f0000000
PCI: No mmconfig possible on device 00:18
...
PCI: Setting latency timer of device 0000:00:03.0 to 64
pcie_portdrv_probe->Dev[02fd:10de] has invalid IRQ. Check vendor BIOS
assign_interrupt_mode Found MSI capability
Allocate Port Service[0000:00:03.0:pcie00]
Allocate Port Service[0000:00:03.0:pcie03]
PCI: Setting latency timer of device 0000:00:04.0 to 64
pcie_portdrv_probe->Dev[02fb:10de] has invalid IRQ. Check vendor BIOS
assign_interrupt_mode Found MSI capability
Allocate Port Service[0000:00:04.0:pcie00]
Allocate Port Service[0000:00:04.0:pcie03]
PCI: Setting latency timer of device 0000:00:12.0 to 64
pcie_portdrv_probe->Dev[0376:10de] has invalid IRQ. Check vendor BIOS
assign_interrupt_mode Found MSI capability
Allocate Port Service[0000:00:12.0:pcie00]
Allocate Port Service[0000:00:12.0:pcie03]
PCI: Setting latency timer of device 0000:00:14.0 to 64
pcie_portdrv_probe->Dev[0374:10de] has invalid IRQ. Check vendor BIOS
assign_interrupt_mode Found MSI capability
Allocate Port Service[0000:00:14.0:pcie00]
Allocate Port Service[0000:00:14.0:pcie03]
PCI: Setting latency timer of device 0000:00:16.0 to 64
pcie_portdrv_probe->Dev[0375:10de] has invalid IRQ. Check vendor BIOS
assign_interrupt_mode Found MSI capability
Allocate Port Service[0000:00:16.0:pcie00]
Allocate Port Service[0000:00:16.0:pcie03]
PCI: Setting latency timer of device 0000:00:17.0 to 64
pcie_portdrv_probe->Dev[0377:10de] has invalid IRQ. Check vendor BIOS
assign_interrupt_mode Found MSI capability
Allocate Port Service[0000:00:17.0:pcie00]
Allocate Port Service[0000:00:17.0:pcie03]
...

<--  snip  -->

> Olivier
> 
> System: ASUS M2N32-SLI, AMD64 X2 Dual Core 4600
> 
> 
> kernel 2.6.9-rc5
> 
> irq 22: nobody cared (try booting with the "irqpoll" option)
> 
> Call Trace:
>  <IRQ>  [<ffffffff80259055>] __report_bad_irq+0x35/0x90
>  [<ffffffff802592d3>] note_interrupt+0x223/0x280
>  [<ffffffff80259d41>] handle_fasteoi_irq+0xb1/0xf0
>  [<ffffffff8020b17c>] call_softirq+0x1c/0x30
>  [<ffffffff8020d1ba>] do_IRQ+0x8a/0xe0
>  [<ffffffff8020a571>] ret_from_intr+0x0/0xa
>  <EOI>
> handlers:
> [<ffffffff8807f150>] (nv_generic_interrupt+0x0/0xc0 [sata_nv])
> Disabling IRQ #22
> 
> 
>            CPU0       CPU1
>   0:        223      45540   IO-APIC-edge      timer
>   1:          1        402   IO-APIC-edge      i8042
>   6:          1          4   IO-APIC-edge      floppy
>   8:          0          0   IO-APIC-edge      rtc
>   9:          0          0   IO-APIC-fasteoi   acpi
>  12:          0        103   IO-APIC-edge      i8042
>  14:          7        140   IO-APIC-edge      ide0
>  16:          0          0   IO-APIC-fasteoi   libata
>  17:          0         10   IO-APIC-fasteoi   bttv0
>  20:          0         20   IO-APIC-fasteoi   ehci_hcd:usb1
>  21:          0          0   IO-APIC-fasteoi   libata
>  22:        214      99786   IO-APIC-fasteoi   libata, HDA Intel
>  23:         76       6230   IO-APIC-fasteoi   libata, ohci_hcd:usb2
> 308:          5       3169   PCI-MSI-edge      eth1
> 309:          0         10   PCI-MSI-edge      eth1
> 310:          0         44   PCI-MSI-edge      eth1
> 311:          1       3193   PCI-MSI-edge      eth0
> 312:          0          0   PCI-MSI-edge      eth0
> 313:          0          1   PCI-MSI-edge      eth0
> NMI:         64         68
> LOC:      45716      45691
> ERR:          0
> 
> 
> 
> kernel 2.6.19-rc5 with pci=routeirq
> 
> irq 21: nobody cared (try booting with the "irqpoll" option)
> 
> Call Trace:
>  <IRQ>  [<ffffffff80259055>] __report_bad_irq+0x35/0x90
>  [<ffffffff802592d3>] note_interrupt+0x223/0x280
>  [<ffffffff80259d41>] handle_fasteoi_irq+0xb1/0xf0
>  [<ffffffff8020b17c>] call_softirq+0x1c/0x30
>  [<ffffffff8020d1ba>] do_IRQ+0x8a/0xe0
>  [<ffffffff802092f0>] default_idle+0x0/0x50
>  [<ffffffff8020a571>] ret_from_intr+0x0/0xa
>  <EOI>  [<ffffffff80209319>] default_idle+0x29/0x50
>  [<ffffffff8020939b>] cpu_idle+0x5b/0x80
>  [<ffffffff8050039c>] start_secondary+0x50c/0x520
> 
> handlers:
> [<ffffffff880e6fd0>] (usb_hcd_irq+0x0/0x60 [usbcore])
> Disabling IRQ #21
> 
>            CPU0       CPU1
>   0:        214      24856   IO-APIC-edge      timer
>   1:          0        359   IO-APIC-edge      i8042
>   6:          0          5   IO-APIC-edge      floppy
>   8:          0          0   IO-APIC-edge      rtc
>   9:          0          0   IO-APIC-fasteoi   acpi
>  12:          0        103   IO-APIC-edge      i8042
>  14:          0        128   IO-APIC-edge      ide0
>  16:          0          0   IO-APIC-fasteoi   libata
>  17:          1          2   IO-APIC-fasteoi   bttv0
>  20:         22       6469   IO-APIC-fasteoi   libata
>  21:         11      99989   IO-APIC-fasteoi   ehci_hcd:usb2, HDA Intel
>  22:          0          1   IO-APIC-fasteoi   libata, ohci_hcd:usb1
>  23:          0          0   IO-APIC-fasteoi   libata
> 308:          8       2378   PCI-MSI-edge      eth1
> 309:          0          9   PCI-MSI-edge      eth1
> 310:          0          9   PCI-MSI-edge      eth1
> 311:          4       2401   PCI-MSI-edge      eth0
> 312:          0          0   PCI-MSI-edge      eth0
> 313:          0          1   PCI-MSI-edge      eth0
> NMI:         74         47
> LOC:      25024      24991
> ERR:          0
> 
> 
> 
> kernel 2.6.19-rc5 with irqpoll
> 
> Hang during boot
> (See screenshot in http://olivn.trollprod.org/2.6.19-rc5-irq.tar.gz)
> 
> 
> 
> kernel 2.6.18 (works without any problem)
> 
>            CPU0       CPU1
>   0:       1590     998089    IO-APIC-edge  timer
>   1:          2        657    IO-APIC-edge  i8042
>   6:          0          5    IO-APIC-edge  floppy
>   8:          0          0    IO-APIC-edge  rtc
>   9:          0          0   IO-APIC-level  acpi
>  12:        194      59353    IO-APIC-edge  i8042
>  14:         13       6381    IO-APIC-edge  ide0
>  50:          0          0   IO-APIC-level  libata
>  58:          0          0   IO-APIC-level  libata
>  66:        109     181425   IO-APIC-level  libata, nvidia
>  74:         41        963   IO-APIC-level  ehci_hcd:usb1, HDA Intel
>  82:          4          6   IO-APIC-level  bttv0
>  90:          0          0       PCI-MSI-X  eth0
>  98:          3          0       PCI-MSI-X  eth0
> 106:     212234          0       PCI-MSI-X  eth0
> 114:        532          0       PCI-MSI-X  eth1
> 122:        445          0       PCI-MSI-X  eth1
> 130:     212217          0       PCI-MSI-X  eth1
> 233:         73      23009   IO-APIC-level  libata, ohci_hcd:usb2
> NMI:         99        104
> LOC:     999684     999664
> ERR:          0
> MIS:          0

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Yinghai Lu]

olivier,

lspci -vvxxx please.

it seems usb and audio share the interrtupts by ioapic.

YH

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Olivier Nicolas]

Yinghai Lu wrote:
> olivier,
> 
> lspci -vvxxx please.
> 
> it seems usb and audio share the interrtupts by ioapic.
> 
> YH

lspci -vvxxx is available at http://olivn.trollprod.org/lspci.gz

2.6.19-rc5 can be booted properly on my system if "notsc" parameter is used.

Could it be related to http://lkml.org/lkml/2006/10/27/141 as I'm usage 
a dual core AMD64 ?



Olivier

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Takashi Iwai]

At Thu, 9 Nov 2006 07:49:56 +0100,
Adrian Bunk wrote:
> 
> On Wed, Nov 08, 2006 at 01:44:29PM +0100, Olivier Nicolas wrote:
> 
> > Hi,
> 
> Hi Olivier,
> 
> > 2.6.19-rc5 does not boot properly, I have tried pci=routeirq, irpoll 
> > without success.
> > 
> > Full details (.config, dmesg, /proc/interrupts) are in 
> > http://olivn.trollprod.org/2.6.19-rc5-irq.tar.gz
> 
> thanks for your report!
> 
> I might be wrong, but looking at the dmesg:
> - irq 22 is the hda_intel IRQ
> - the "irq 22: nobody cared" is immediately before the
>   "hda_intel: No response from codec, disabling MSI..."
> - in the routeirq case, the hda_intel IRQ as well as the
>   IRQ in the error message change to 21
> 
> So it might be related to the hda_intel MSI check.

To disable MSI from the beginning, set disable_msi=1 module option for
snd-hda-intel.


Takashi

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Olivier Nicolas]

Takashi,

You are right, setting disable_msi=1 as an option for snd-hda-intel 
module solve my problem.

Thanks

Olivier

Takashi Iwai wrote:
> At Thu, 9 Nov 2006 07:49:56 +0100,
> Adrian Bunk wrote:
>> On Wed, Nov 08, 2006 at 01:44:29PM +0100, Olivier Nicolas wrote:
>>
>>> Hi,
>> Hi Olivier,
>>
>>> 2.6.19-rc5 does not boot properly, I have tried pci=routeirq, irpoll 
>>> without success.
>>>
>>> Full details (.config, dmesg, /proc/interrupts) are in 
>>> http://olivn.trollprod.org/2.6.19-rc5-irq.tar.gz
>> thanks for your report!
>>
>> I might be wrong, but looking at the dmesg:
>> - irq 22 is the hda_intel IRQ
>> - the "irq 22: nobody cared" is immediately before the
>>   "hda_intel: No response from codec, disabling MSI..."
>> - in the routeirq case, the hda_intel IRQ as well as the
>>   IRQ in the error message change to 21
>>
>> So it might be related to the hda_intel MSI check.
> 
> To disable MSI from the beginning, set disable_msi=1 module option for
> snd-hda-intel.
> 
> 
> Takashi

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Stephen Hemminger]

On Thu, 9 Nov 2006 07:49:56 +0100
Adrian Bunk <bunk@stusta.de> wrote:

> On Wed, Nov 08, 2006 at 01:44:29PM +0100, Olivier Nicolas wrote:
> 
> > Hi,
> 
> Hi Olivier,
> 
> > 2.6.19-rc5 does not boot properly, I have tried pci=routeirq, irpoll 
> > without success.
> > 
> > Full details (.config, dmesg, /proc/interrupts) are in 
> > http://olivn.trollprod.org/2.6.19-rc5-irq.tar.gz
> 
> thanks for your report!
> 
> I might be wrong, but looking at the dmesg:
> - irq 22 is the hda_intel IRQ
> - the "irq 22: nobody cared" is immediately before the
>   "hda_intel: No response from codec, disabling MSI..."
> - in the routeirq case, the hda_intel IRQ as well as the
>   IRQ in the error message change to 21
> 
> So it might be related to the hda_intel MSI check.

More likely the MSI management routines don't work for disabling MSI.
I am debugging a problem where MSI doesn't work across suspend/resume,
I suspect the base MSI code needs fixing.


-- 
Stephen Hemminger <shemminger@osdl.org>

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Eric W. Biederman]

Stephen Hemminger <shemminger@osdl.org> writes:

> On Thu, 9 Nov 2006 07:49:56 +0100
> Adrian Bunk <bunk@stusta.de> wrote:
>
>> On Wed, Nov 08, 2006 at 01:44:29PM +0100, Olivier Nicolas wrote:
>> 
>> > Hi,
>> 
>> Hi Olivier,
>> 
>> > 2.6.19-rc5 does not boot properly, I have tried pci=routeirq, irpoll 
>> > without success.
>> > 
>> > Full details (.config, dmesg, /proc/interrupts) are in 
>> > http://olivn.trollprod.org/2.6.19-rc5-irq.tar.gz
>> 
>> thanks for your report!
>> 
>> I might be wrong, but looking at the dmesg:
>> - irq 22 is the hda_intel IRQ
>> - the "irq 22: nobody cared" is immediately before the
>>   "hda_intel: No response from codec, disabling MSI..."
>> - in the routeirq case, the hda_intel IRQ as well as the
>>   IRQ in the error message change to 21
>> 
>> So it might be related to the hda_intel MSI check.
>
> More likely the MSI management routines don't work for disabling MSI.

Well brand new MSI handling could be buggy, the disable hypothesis doesn't
make much sense on boot up.

> I am debugging a problem where MSI doesn't work across suspend/resume,
> I suspect the base MSI code needs fixing.

Rethinking the interfaces maybe I don't think the code is broken I think
most likely it is just not designed to do the right thing.  I remember
thinking how horrible that code is, when I read through it.

Eric

RE: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Lu, Yinghai]

           CPU0       CPU1       
  0:        214      24856   IO-APIC-edge      timer
  1:          0        359   IO-APIC-edge      i8042
  6:          0          5   IO-APIC-edge      floppy
  8:          0          0   IO-APIC-edge      rtc
  9:          0          0   IO-APIC-fasteoi   acpi
 12:          0        103   IO-APIC-edge      i8042
 14:          0        128   IO-APIC-edge      ide0
 16:          0          0   IO-APIC-fasteoi   libata
 17:          1          2   IO-APIC-fasteoi   bttv0
 20:         22       6469   IO-APIC-fasteoi   libata
 21:         11      99989   IO-APIC-fasteoi   ehci_hcd:usb2, HDA Intel
 22:          0          1   IO-APIC-fasteoi   libata, ohci_hcd:usb1
 23:          0          0   IO-APIC-fasteoi   libata
308:          8       2378   PCI-MSI-edge      eth1
309:          0          9   PCI-MSI-edge      eth1
310:          0          9   PCI-MSI-edge      eth1
311:          4       2401   PCI-MSI-edge      eth0
312:          0          0   PCI-MSI-edge      eth0
313:          0          1   PCI-MSI-edge      eth0
NMI:         74         47 
LOC:      25024      24991 
ERR:          0

But according to the irq router in pci conf, 
the ehci and audio share to use irq 20.
sata0 and ohci use irq 23
sata1 use irq 22, (MAC1 share it but kernel will use MSI)
sata2 use irq 21, (MAC0 share it but kernel will use MSI)

the ACPI report wrong GSI for them?

Can you disable the acpi and check the interrupt and lspci?

YH

RE: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Lu, Yinghai]

Eric,

Can you confirm:
>From 2.6.19-rc1, the irq for devices that using io-apic will not change
between pci=routeirq and not.

Olivier,
Can you send lspci -vvxxx for pci=routeirq too?

YH

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Olivier Nicolas]

Lu, Yinghai wrote:
> Eric,
> 
> Can you confirm:
> From 2.6.19-rc1, the irq for devices that using io-apic will not change
> between pci=routeirq and not.
> 
> Olivier,
> Can you send lspci -vvxxx for pci=routeirq too?

All files are available here http://olivn.trollprod.org/19-rc5/

RE: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Lu, Yinghai]

Andi,

The two patches solve the problems that irq nobody care.

They are already in your tree. But first one I wonder if you put correct
one in your tree.


YH

[PATCH] x86_64 irq: reuse vector for __assign_irq_vector 
http://lkml.org/lkml/2006/10/26/38
[PATCH] x86_64 irq: reset more to default when clear irq_vector for
destroy_irq
http://lkml.org/lkml/2006/10/28/16

Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Eric W. Biederman]

"Lu, Yinghai" <yinghai.lu@amd.com> writes:

> Andi,
>
> The two patches solve the problems that irq nobody care.
>
> They are already in your tree. But first one I wonder if you put correct
> one in your tree.

YH.  This is a completely different problem.  The irq is properly setup
and received but none of the drivers wanted it.

Eric

RE: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Lu, Yinghai]

That didn't fix the bug.



-----Original Message-----
From: Olivier Nicolas [mailto:olivn@trollprod.org] 
Sent: Friday, November 10, 2006 12:03 PM
To: Lu, Yinghai
Subject: Re: 2.6.19-rc5 x86_64 irq 22: nobody cared

Bad day today,

Kernel compiled with the first parch irq_mcp55.diff

and with the disable_msi option removed.



#options snd-hda-intel disable_msi=1


ACPI: PCI Interrupt Link [ASA2] enabled at IRQ 21
__assign_irq_vector: irq=15, vector=79, domain=000000ff, mask=00000003,
cpu_online_map=00000003
ACPI: PCI Interrupt 0000:00:0d.2[C] -> Link [ASA2] -> GSI 21 (level,
low) -> IRQ 21
PCI: Setting latency timer of device 0000:00:0d.2 to 64
ata5: SATA max UDMA/133 cmd 0xC400 ctl 0xC002 bmdma 0xB400 irq 21
ata6: SATA max UDMA/133 cmd 0xBC00 ctl 0xB802 bmdma 0xB408 irq 21

ACPI: PCI Interrupt Link [AAZA] enabled at IRQ 21
ACPI: PCI Interrupt 0000:00:0e.1[B] -> Link [AAZA] -> GSI 21 (level,
low) -> IRQ 21
__assign_irq_vector: irq=139, vector=91, domain=000000ff, mask=00000003,
cpu_online_map=00000003
__assign_irq_vector: irq=139, old_vector=91, domain=000000ff,
mask=00000003, cpu_online_map=00000003
PCI: Setting latency timer of device 0000:00:0e.1 to 64
irq 21: nobody cared (try booting with the "irqpoll" option)

Call Trace:
 <IRQ>  [<ffffffff8025a055>] __report_bad_irq+0x35/0x90
 [<ffffffff8025a2d3>] note_interrupt+0x223/0x280
 [<ffffffff8025ad41>] handle_fasteoi_irq+0xb1/0xf0
 [<ffffffff8020b17c>] call_softirq+0x1c/0x30
 [<ffffffff8020d1ba>] do_IRQ+0x8a/0xe0
 [<ffffffff802092f0>] default_idle+0x0/0x50
 [<ffffffff8020a571>] ret_from_intr+0x0/0xa
 <EOI>  [<ffffffff80209319>] default_idle+0x29/0x50
 [<ffffffff8020939b>] cpu_idle+0x5b/0x80
 [<ffffffff8050039c>] start_secondary+0x50c/0x520

handlers:
[<ffffffff8807f150>] (nv_generic_interrupt+0x0/0xc0 [sata_nv])
Disabling IRQ #21
ALSA sound/pci/hda/hda_intel.c:543: hda_intel: No response from codec,
disabling MSI...

So kernl assign irq 21 to sata2, and later share irq 21 with audio.

But audio get MSI, and at that time, it may do sth bad to irq21 that it
is still shared with SATA2.

YH

RE: 2.6.19-rc5 x86_64 irq 22: nobody cared

[Lu, Yinghai]

-----Original Message-----
From: ebiederm@xmission.com [mailto:ebiederm@xmission.com] 

>YH.  This is a completely different problem.  The irq is properly setup
>and received but none of the drivers wanted it.

The irq  with ioapic are shared with sata2 and audio, and later the
audio driver get the irq, and then try to get MSI.

The MAC first shared irq (ioapic) with SATA and it later transfer to use
MSI without problem.

YH