[SOLVED] Sun Ultra 40 possible ACPI problem

DrKrieger · 05-03-2014, 11:20 AM

I've got an old Sun Ultra 40 setup for audio and video processing (and some other stuff) for my church, I've got Slackware64 14.1 installed and running with minimal issues, except one big one; if I do any kind of reboot, power off, etc. short pulling the power cord, the system won't boot past RAM check, just sits there at the boot screen with CPU info.
If I pull the power plug and plug it in, it turns on like a jet and boots up perfectly fine.

If I disable ACPI in the BIOS settings I can boot and reboot all day long, but things don't work the way I need them too. (nouveau for one, haven't tried nvidia driver yet)
I compiled 3.12.18-huge from testing but no difference.
Before I installed Slackware to the workstation I had Linux Mint installed and it worked fine (as far as booting and rebooting is concerned anyways, otherwise it sucked)

When I first got the workstation I put 14.0 32bit on it and the machine was quite happy, so I might try using 3.2.29 from 14.0 to see what happens.
I can't do a lot today because we have church service, but I can definitely take note of any suggestions from anyone, cause I'm kind of at a lost for right now.
I'll go ahead and grab 3.2.29 kernel packages, install and setup extlinux to show the kernel, I'll be able to test it later this afternoon.

metaschima · 05-03-2014, 12:21 PM

What is the output right before it hangs during boot ? Can you post a full boot 'dmesg' to pastebin or similar.

I would run memtest86 first.

DrKrieger · 05-03-2014, 01:07 PM

It never gets to the point of booting the kernel after a reboot.
Here's a paste of dmesg -T after booting the system this week.
http://vpaste.net/wfTp8

I'll run memtest after services are over

metaschima · 05-03-2014, 03:34 PM

So it doesn't even boot the kernel ? Do you use lilo ? If so, make sure 'compat' is NOT in the lilo.conf. This causes some older machines to hang, that's why it is not there by default.

The things I notice in the dmesg are:

Code:

[Sat May  3 09:51:57 2014] Nvidia board detected. Ignoring ACPI timer override.
[Sat May  3 09:51:57 2014] If you got timer trouble try acpi_use_timer_override

[Sat May  3 09:51:57 2014] Your BIOS doesn't leave a aperture memory hole
[Sat May  3 09:51:57 2014] Please enable the IOMMU option in the BIOS setup
[Sat May  3 09:51:57 2014] This costs you 64 MB of RAM

[Sat May  3 09:51:57 2014] spurious 8259A interrupt: IRQ7.

[Sat May  3 09:51:57 2014] ACPI Exception: AE_NOT_FOUND, While evaluating Sleep State [\_S2_] (20130725/hwxface-571)
[Sat May  3 09:51:57 2014] ACPI: (supports S0 S1 S3 S4 S5)

[Sat May  3 09:51:57 2014] PCI: Ignoring host bridge windows from ACPI; if necessary, use "pci=use_crs" and report a bug

[Sat May  3 09:51:57 2014] acpi PNP0A03:00: ACPI _OSC support notification failed, disabling PCIe ASPM
[Sat May  3 09:51:57 2014] acpi PNP0A03:00: Unable to request _OSC control (_OSC support mask: 0x08)

[Sat May  3 09:51:57 2014] HPET not enabled in BIOS. You might try hpet=force boot option

[Sat May  3 12:21:43 2014] nouveau E[   PFIFO][0000:03:00.0] DMA_PUSHER - ch 1 [X[941]] get 0xbeef0200 put 0x00027af0 state 0xc002018c (err: MEM_FAULT) push 0x00000000
[Sat May  3 12:57:37 2014] nouveau E[   PFIFO][0000:03:00.0] DMA_PUSHER - ch 1 [X[941]] get 0xbeef0200 put 0x00020f20 state 0xc002018c (err: MEM_FAULT) push 0x00000000

Some recommendations:
Add 'acpi_use_timer_override' to the boot options, because timer issues can cause hangs.
Use the nvidia driver if your card is supported.
You should enable IOMMU and HPET in the BIOS if they are listed, but this probably won't fix this exact issue.
You may want to post the output of 'cat /proc/interrputs' to see what is causing the spurious interrupt, but this probably isn't causing the hang.

That's all I can think of for now.

number22 · 05-03-2014, 04:00 PM

you need to try some kernel parameters in appends of lilo.conf file, such as "nopat"; "acpi=x86" etc... google kernel parameters for acpi, lot of options.

DrKrieger · 05-03-2014, 04:47 PM

@metaschima: No, bootloader doesn't even load, BIOS starts, shows CPU info and stops before starting RAM check, unless I pull the power cable and boot.
I left memtest to run, I'll hopefully be able to check it tomorrow or one day this next week. (otherwise it will be Saturday before I can try anything)
I downloaded everything I need to setup the nvidia driver, and I know IOMMU option is in the BIOS setup, don't remember seeing HPET but I'll hunt for it; I'll make a note of catting /proc/interrupts.

I use extlinux for bootloader.

@number22: I'll definitely be trying kernel parameters, hopefully one day this week.

metaschima · 05-03-2014, 05:46 PM

Before the bootloader ? Well, then it's probably a hardware issue.

DrKrieger · 05-04-2014, 11:45 AM

Then why don't any issues show up while the system is running or with Linux Mint?

DrKrieger · 05-05-2014, 10:49 AM

Well this is strange; I'm here on the Sun Ultra 40 and after 43 1/2 hours of running memtest and 13 passes, no errors, I press esc and reboot, instead of taking the slackware dvd out I just use it to boot the system, I went ahead and installed nvidia driver and set kernel options in extlinux.conf, I reboot and it boots normally. I boot into the 3.10.17 kernel and check the nvidia driver and make sure everything works the way I need it to and it does, I reboot and boot stops before the BIOS RAM check again.

So it appears I can boot using the CD, but not the hard drive? wtf?

Also, I checked IOMMU and it was already enabled and set to 64mb and I couldn't find HPET anywhere. :/

DrKrieger · 05-05-2014, 12:18 PM

OK, I must have been remembering what I did wrong.
I've got nvidia driver 304.88 running instead of nouveau and I've rebooted 3 times now with no issue.
So I guess nouveau was the culprit.