Re: Possible PNP issue on Dell PowerEdge 4200 unless noapic used

Re: Possible PNP issue on Dell PowerEdge 4200 unless noapic used

[Jonathan Nieder]

(just cc-ing linux-kernel so any work on this gets archived)
trouble daemon wrote:

> Greetings,
>
> Hello Adam and Bjorn. I am sending this email to you as suggested by
> Jonathan regarding a problem that I was having with Linux on a pair of
> old Dell PowerEdge 4200 machines that I own (yes, they still run!).
>
> I have left the most recent reply between Jonathan and I quoted below,
> but you can find the original bug report at the Debian bug tracker
> under bug #586494 [1]. As a brief recap, I have been required to use
> the noapic kernel option in these systems in order to allow them to
> boot, and appears to have originated after a commit to the kernel
> relating to EISA (although it has come to my attention now that PNP
> may be related as well).
>
> As suggested, I have attached a copy of the "quick & dirty" config I
> used for a vanilla 3.1.0-rc7 kernel I compiled, along with the boot
> log of the machine, and a diff against the apic log of a working boot
> of the same kernel when using noapic. One thing that I should mention
> is that on these dells, I also usually use the pnpbios=off option to
> hide the pesky warnings on boot, in addition to noapic.
>
> I hope these logs help you in some way. If there is anything else you
> might be interested in knowing or testing, don't hesitate to ask!

Original report: http://thread.gmane.org/gmane.linux.kernel/1000807
Logs: http://bugs.debian.org/cgi-bin/bugreport.cgi?msg=86;bug=586494

> Dan McGrath
>
> PS - Not a problem Jonathan, thanks!
>
> [1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=586494

Re: Possible PNP issue on Dell PowerEdge 4200 unless noapic used

[Bjorn Helgaas]

[-- Attachment #1: Type: text/plain, Size: 2104 bytes --]

[Resending since gmail made it HTML, which LKML thinks is spam.  sigh]

On Sat, Sep 24, 2011 at 6:28 PM, Jonathan Nieder <jrnieder@gmail.com> wrote:
>
> (just cc-ing linux-kernel so any work on this gets archived)
> trouble daemon wrote:
>
> > Greetings,
> >
> > Hello Adam and Bjorn. I am sending this email to you as suggested by
> > Jonathan regarding a problem that I was having with Linux on a pair of
> > old Dell PowerEdge 4200 machines that I own (yes, they still run!).
> >
> > I have left the most recent reply between Jonathan and I quoted below,
> > but you can find the original bug report at the Debian bug tracker
> > under bug #586494 [1]. As a brief recap, I have been required to use
> > the noapic kernel option in these systems in order to allow them to
> > boot, and appears to have originated after a commit to the kernel
> > relating to EISA (although it has come to my attention now that PNP
> > may be related as well).
> >
> > As suggested, I have attached a copy of the "quick & dirty" config I
> > used for a vanilla 3.1.0-rc7 kernel I compiled, along with the boot
> > log of the machine, and a diff against the apic log of a working boot
> > of the same kernel when using noapic. One thing that I should mention
> > is that on these dells, I also usually use the pnpbios=off option to
> > hide the pesky warnings on boot, in addition to noapic.
> >
> > I hope these logs help you in some way. If there is anything else you
> > might be interested in knowing or testing, don't hesitate to ask!
>
> Original report: http://thread.gmane.org/gmane.linux.kernel/1000807
> Logs: http://bugs.debian.org/cgi-bin/bugreport.cgi?msg=86;bug=586494
>
> > Dan McGrath
> >
> > PS - Not a problem Jonathan, thanks!
> >
> > [1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=586494

Dan, can you try the attached patch?  I'm not an MPS expert, but from
inspecting f8924e770, it seems suspect that we used to always set the
mp_bus_not_pci bit, then clear it if we found a PCI bus.  After
f8924e770, we set mp_bus_not_pci for ISA buses, clear it for PCI
buses, and leave it alone otherwise.

[-- Attachment #2: mps-fix --]
[-- Type: application/octet-stream, Size: 618 bytes --]

commit 7ed1c1f44bc58d9e6a01d7e970a8ad25ece3df06
Author: Bjorn Helgaas <bhelgaas@google.com>
Date:   Sun Sep 25 06:09:31 2011 -0600

diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c
index 9103b89..0741b062 100644
--- a/arch/x86/kernel/mpparse.c
+++ b/arch/x86/kernel/mpparse.c
@@ -95,8 +95,8 @@ static void __init MP_bus_info(struct mpc_bus *m)
 	}
 #endif
 
+	set_bit(m->busid, mp_bus_not_pci);
 	if (strncmp(str, BUSTYPE_ISA, sizeof(BUSTYPE_ISA) - 1) == 0) {
-		set_bit(m->busid, mp_bus_not_pci);
 #if defined(CONFIG_EISA) || defined(CONFIG_MCA)
 		mp_bus_id_to_type[m->busid] = MP_BUS_ISA;
 #endif

Re: Possible PNP issue on Dell PowerEdge 4200 unless noapic used

[trouble daemon]

[-- Attachment #1: Type: text/plain, Size: 2429 bytes --]

Heya,

On Sun, Sep 25, 2011 at 8:24 AM, Bjorn Helgaas <bhelgaas@google.com> wrote:
> [Resending since gmail made it HTML, which LKML thinks is spam.  sigh]
>
> On Sat, Sep 24, 2011 at 6:28 PM, Jonathan Nieder <jrnieder@gmail.com> wrote:
>>
>> (just cc-ing linux-kernel so any work on this gets archived)
>> trouble daemon wrote:
>>
>> > Greetings,
>> >
>> > Hello Adam and Bjorn. I am sending this email to you as suggested by
>> > Jonathan regarding a problem that I was having with Linux on a pair of
>> > old Dell PowerEdge 4200 machines that I own (yes, they still run!).
>> >
>> > I have left the most recent reply between Jonathan and I quoted below,
>> > but you can find the original bug report at the Debian bug tracker
>> > under bug #586494 [1]. As a brief recap, I have been required to use
>> > the noapic kernel option in these systems in order to allow them to
>> > boot, and appears to have originated after a commit to the kernel
>> > relating to EISA (although it has come to my attention now that PNP
>> > may be related as well).
>> >
>> > As suggested, I have attached a copy of the "quick & dirty" config I
>> > used for a vanilla 3.1.0-rc7 kernel I compiled, along with the boot
>> > log of the machine, and a diff against the apic log of a working boot
>> > of the same kernel when using noapic. One thing that I should mention
>> > is that on these dells, I also usually use the pnpbios=off option to
>> > hide the pesky warnings on boot, in addition to noapic.
>> >
>> > I hope these logs help you in some way. If there is anything else you
>> > might be interested in knowing or testing, don't hesitate to ask!
>>
>> Original report: http://thread.gmane.org/gmane.linux.kernel/1000807
>> Logs: http://bugs.debian.org/cgi-bin/bugreport.cgi?msg=86;bug=586494
>>
>> > Dan McGrath
>> >
>> > PS - Not a problem Jonathan, thanks!
>> >
>> > [1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=586494


> Dan, can you try the attached patch?  I'm not an MPS expert, but from
> inspecting f8924e770, it seems suspect that we used to always set the
> mp_bus_not_pci bit, then clear it if we found a PCI bus.  After
> f8924e770, we set mp_bus_not_pci for ISA buses, clear it for PCI
> buses, and leave it alone otherwise.

Success! :) Booted up like a charm in 3.1.0-rc7 with the patch applied
and no kernel boot params.

Thank you for figuring it out *bow*

[-- Attachment #2: mps-fix --]
[-- Type: application/octet-stream, Size: 618 bytes --]

commit 7ed1c1f44bc58d9e6a01d7e970a8ad25ece3df06
Author: Bjorn Helgaas <bhelgaas@google.com>
Date:   Sun Sep 25 06:09:31 2011 -0600

diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c
index 9103b89..0741b062 100644
--- a/arch/x86/kernel/mpparse.c
+++ b/arch/x86/kernel/mpparse.c
@@ -95,8 +95,8 @@ static void __init MP_bus_info(struct mpc_bus *m)
 	}
 #endif
 
+	set_bit(m->busid, mp_bus_not_pci);
 	if (strncmp(str, BUSTYPE_ISA, sizeof(BUSTYPE_ISA) - 1) == 0) {
-		set_bit(m->busid, mp_bus_not_pci);
 #if defined(CONFIG_EISA) || defined(CONFIG_MCA)
 		mp_bus_id_to_type[m->busid] = MP_BUS_ISA;
 #endif