IBM ZPRO ACPI Hang and No Shutdown

[mkgreene]

Running 3.14 using /APIC /SMP system sometimes hangs after WPS starts and even if it does not hang network card does not work. Using /PIC /SMP system boots and runs.

Using /APIC or /PIC system will not shutdown (power off). This did work with 3.09 I believe.

Logs attached.

MikeG

[mkgreene]

Forgot, system info:

​https://www-304.ibm.com/systems/support/supportsite.wss/docdisplay?lndocid=MIGR-45139&brandind=5000004

Not using SCSI and have ATI video card.

[pasha]

Log when using /APIC ?

[mkgreene]

This log is APIC without SMP. When I put /APIC back in config.sys with /SMP, I get a system trap during boot.

[pasha]

For use /APIC with /SMP you need to remap IRQ48

[eco]

e-co:

RUS: [Pasha]: ну я пока в этом тикете ничего не вижу

No more advices..

1) acpi.log -> 3:1.0 14E4:16A7 'PCI1' IRQs: PCI:11 PIC:'LNKC':11 APIC:48 MSI

the device is going use reserved IRQ 48 in SMP APIC. It necessary remap

2) e-co: How do you start acpidaemon? try without it, try start it from startup.cmd

[eco]

e-co:

MikeG continues:

"I can disable all USB and it boots. Do you mean with /ACPI option and no USB? "

e-co:

I recommend disable USB Legacy in BIOS Setup, try boot /SMP /APIC.

[mkgreene]

I will try what you suggested. I am not concerned with the /APIC switch operation, but I would like the poweroff option to work again on this system.

[mkgreene]

I turned off USB and added /APIC switch. System traps before WPS starts.

[pasha]

You have 92 IRQ in your MB. IBM kernel support 32 IRQ only. So you need remap:

3:1.0 14E4:16A7 'PCI1' IRQs: PCI:11 PIC:'LNKC':11 APIC:48 MSI - broadcom network card

But you have problem with remap, because have:

0:31.1 8086:24CB 'ICHX' IRQs: PCI:0 PIC:'LNKC':11 APIC:18 - EIDE controller

So, you can't use IRQ11 for remap. try IRQ13 . Also, you can insert your netcard to another slot. It be change IRQ for broadcom and do possible easy for remap.

[mkgreene]

I will try to remap to IRQ13. The network card is built in to the motherboard.

[mkgreene]

I tried remap REMAP 48 to 13 which still resulted in a trap. I have reverted back to ACPI 3.09 which will only run PIC, but will power off system. Also, the system is more responsive (faster) than 3.14.

I recommend closing this ticket because I intend to stay with ACPI 3.09.

[pasha]

3.14 don't boot with psd=acpi.psd /SMP /PIC? Where is stopping/trapped?

[mkgreene]

No! System boots with 3.14 using /SMP /PIC but it will not poweroff. Using 3.09 the system powers off and runs a bit faster.

[pasha]

You probe ACPI29012009.ZIP with PowerOff.exe and it don't work?

[mkgreene]

Pasha,

I did not know about ACPI29012009.ZIP but I did just try it. No good, as soon as I type poweroff the system hangs. Version 3.09 seems to work with /PIC and powers the system off. I just installed eCS RC6 on a IBM XSeries 335 and every thing works fine - except ACPI. Again, if I use v3.09 that system works with /PIC. The XSeries 335 is a few years old but in is a main stream IBM server - WTF?

MikeG

[Steven Levine]

Mike,

Please review

http://svn.netlabs.org/acpi/wiki/TicketProcessingGuide

and

​http://cz.ecomstation.ru/projects/acpitools/?action=logs

It is unreasonable to expect issues to be resolved unless you supply Pasha with the data he needs. This means new logs when you test a new builds and proper handling of the ticket fields. If you leave the ticket closed, Pasha is supposed to ignore it. He's plenty busy working on the open tickets for the 3.15 milestone. Since you have a hang, you might need to work with a debug build or a system dump to capture the data Pasha needs. I not sure which will provide better test data.

Also, it should be obvious, but regressions are never an intended work product. However, Intel continues to make significant changes to the ACPICA sources that provide a lot of the acpi.psd functionality. Most times, these changes resolve open issues, but regressions can and will happen.

[mkgreene]

I tried acpipre2.zip from betazone. As a note I have the patched B57.OS2 driver installed.

When I boot with the /APIC and /SMP options the system traps early in the boot process. If I boot with /APIC and no SMP (single CPU) the system boots (logs IBM6221_APIC_NOSMP.ziq). Booting with only /SMP switch (PIC mode) works and power off from xworkplace also works (logs: IBM6221_PIC_SMP.ziq).

[mkgreene]

with SMP and PIC mode

[mkgreene]

APIC mode single CPU

[David Azarewicz]

This ticket is for an unsupported version. Have you tried the new 3.20.01 release?

[David Azarewicz]

Ticket closed due to no response from reporter.