ICH9 boot problem

[Dhaks]

Hello Experts,

I am trying to boot Ubuntu with ICH9 chipset using VirtualBox 5.0.10, have booted same image with PIIX3 but having problems with ICH9.
I copied over vdi file from previous builds.

Ata1 master: VBOX CD-ROM ATAPI-6 CD-ROM/DVD-ROM
Ata2 master: unknown device
Ata2 slave: unknown device
Ata3 master: unknown device
Ata3 slave: unknown device

Using same setup as on PIIX3, SATA -> AHCI -> Port 0.

Thanks.

[mpack]

Why are you asking here? Surely Ubuntu's support for a given motherboard chipset (and whether you can flip back and fore willy nilly) is something you should raise on the Ubuntu forums?

By "here" I meant the VirtualBox forums, but I could also ask what this has to do with the "OSE" forum on this site! Moving this to "Linux Guests".

[Dhaks]

I felt it was more like environmental thing that I may have missed. Sure, let me try putting in Ubuntu forum.
I doubt Ubuntu support is not extended for ICH9 though, but I could be wrong. Thanks.

[Dhaks]

Thanks for moving it to Linux.

[Dhaks]

BTW, I ended up seeing the same error message with Windows 10 guest OS as well.

ata2 master: Unknown device
ata2 slave: Unknown device
ata3 master: Unknown device
ata3 slave: Unknown device

Any suggestions on why this would show up ? Thanks.

[michaln]

Any suggestions on why this would show up ? Thanks.

No VBox.log, no answers, sorry...

Statistically speaking there's a high chance you're doing something wrong. But without the log, I won't even try to guess what.

[Dhaks]

Please find attached VBox log tar file. Thanks.

[michaln]

I have no idea what you did. The VM configuration looks roughly correct but the BIOS never found any disks. The "ataX master/slave: Unknown device" messages are normal because you have no IDE devices.

Either the BIOS in the guest is hung or it didn't find anything, I can't quite tell. It definitely didn't get as far as trying to boot anything. You should compare with VBox.log files from a working setup.

Since you're building everything yourself and making significant modifications, no one who doesn't sit in your office can give you much direct help. If you can reproduce the problem on an actual 5.0.X release then we can look into it. My only general suggestion is "don't make too many changes at once".

[Dhaks]

I have absolutely no idea why setup behaves the way it is behaving. Hang does happen at the very early stage of the boot.

I first added some debug messages to ich9pciBiosInitDevice which showed that there wasn't enough space left for devices E1000, VMMDev, AHCI. This wasn't the case when compared with previous logs, only other end of the passthrough device. But, strictly speaking my pluggable device doesn't really ask much of resource, so not sure why the memory allocation problem.

I ran multiple experiments
1. Increased memory hole to 2G, I could see that getting reflected with my device registration. Start address moved to 0x80000000, but, nothing changed wrt E1000, VMMDev, AHCI devices.
2. Moved my device position down to the last but one (0x1E) since AHCI(0x1F) occupies device position 31. Expectation was BIOS would start allocating from 0 and so somehow would accommodate E1000, VMMDev but Nothing happened.

Even though other end of my pluggable device is new, but is very much aligned with PCI protocol.

Not sure what else to try.

[michaln]

I have absolutely no idea why setup behaves the way it is behaving. Hang does happen at the very early stage of the boot.

This is the first time you mentioned it's a hang. Is the BIOS in the guest stuck? What is it doing? You can debug it with the VM debugger. Try removing devices from the VM until something else happens.

Not sure what else to try.

Have you tried the obvious, an official 5.0.x build from Oracle?

[Dhaks]

That's because I felt problem was due to disk not being accessible for some reason and error messages had some internal meaning. Now that you mentioned the reason, since there were no pending requests between my device to the platform and it's probably "hung" inside the BIOS.

I will try remove devices but last time when I experimented, it triggered assertions and was inclined to think that these devices as "required".

Can you please confirm whether these are the right options, Thanks.

VirtualBox-5.0.10/out/linux.amd64/release/bin/VBoxManage modifyvm vm_name --ahci off
VirtualBox-5.0.10/out/linux.amd64/release/bin/VBoxManage modifyvm vm_name --e1000 off
VirtualBox-5.0.10/out/linux.amd64/release/bin/VBoxManage modifyvm vm_name --vmmdev off

[Dhaks]

Using SCSI controller booted to the desktop. Thanks.

[michaln]

I can only tell you what the options do. You have to know what you want/need. In general, I would highly recommend using the GUI for managing VMs because it has a lot of logic built into it preventing misconfiguration. With the command line tools, you have to know exactly what you're doing.

So you don't know why your VM doesn't work with AHCI but it works with SCSI. That's very strange and not a good sign. But since I see no such problem, there's not much I can do for you I'm afraid.