Server installer should not use -server kernel for non-PAE CPU's

Bug #227869 reported by Ben Collins on 2008-05-07
This bug affects 1 person
Affects Status Importance Assigned to Milestone
base-installer (Ubuntu)

Bug Description

Binary package hint: base-installer

The -server CD will install -server kernels on CPU's where it cannot run. Spefically, when PAE is not supported (On VIA C3 CPU's for example).

The checks for whether or not to install the -server kernel should be changed to something more simplistic. The primary difference between -server and -generic, for CPU support is the PAE requirement in the -server kernel. Any CPU that supports PAE should get the -server kernel, and any that do not, should get the -generic kernel.

This probably only applies to 32-bit x86, since x86_64 can be assumed to have PAE support.

I've attached two cpuinfo's, one with, and one without PAE support. All other checks for -server kernel usage, can likely be removed (-server and -generic CPU level support is the same).

Ben Collins (ben-collins) wrote :
Ben Collins (ben-collins) wrote :
Colin Watson (cjwatson) wrote :

The reason we do it with preseeding rather than by run-time detection is because otherwise the server kernel ends up being installed for *all* PAE-capable machines that perform a netboot installation, which I don't think is appropriate. Unfortunately you can't assume that the -server kernel should be used if it's available and the system supports it, otherwise this would be a lot simpler.

We need some way to select the -server kernel if the system is capable of it, if it's available, *and* we're installing from the server CD. Right now, there's no mechanism that would permit this, so one would have to be invented.

Changed in base-installer:
importance: Undecided → Medium
status: New → Triaged
Forest (foresto) wrote :

Adding my voice to this bug. It's rather a pain for a project I'm working on, which uses Hardy server on low powered server boxes with VIA Nehemiah CPUs. Installation completes, but boot fails with a message like this:

Starting up...
This kernel requires the following features not present on the CPU:
0:6 0:8
Unable to boot -please use a kernel appropriate for your CPU.

That's not a very helpful message, nor is it particularly pleasant to have to mount a freshly installed partition in recovery mode and manually replace the kernel and related packages.

It's great that Ubuntu can take advantage of the latest greatest hardware, but let's not forget that low powered machines continue to be valuable and in widespread use, especially where energy consumption matters.

For the record, Hardy server fails similarly when installing as a virtualbox guest, complaining only about feature "0:6".

Tim Bishop (geodog-twjb) wrote :

Adding my voice as well, and asking about workarounds.

I ran into the same problem with installing 8.04.1 server onto a USB disk for use with a VIA CPU for a low power, silent, personal server.

Is there a workaround or easy way to switch kernels? I'm not super technical, but if someone pointed me to directions I could probably follow them.

Thanks in advance.

Me too. It's easy to find out whether the cpu has pae or not, and select the right kernel for it. So why is this not happening? At least tell me I need PAE support before you install or reboot so I can know that I need to install a different kernel before I waste my time failing to reboot.

reread. It's not so simple. But its lack is *awfully* annoying.

David Anderson (dma-pern) wrote :

This is still happening in 8.10 server edition. Surely it is simple to put a check in a version that is designed to be run as server to see that prerequisite conditions are met?

tags: added: iso-testing
Jean-Baptiste Lallement (jibel) wrote :

Confirmed during iso-testing of 10.04.2 Ubuntu Server i386 Testcase: Install (JeOS on KVM)

Changed in base-installer (Ubuntu Lucid):
importance: Undecided → Medium
status: New → Confirmed
status: Confirmed → Triaged
Changed in base-installer (Ubuntu Lucid):
milestone: none → ubuntu-10.04.3
Colin Watson (cjwatson) on 2011-07-19
Changed in base-installer (Ubuntu Lucid):
milestone: ubuntu-10.04.3 → ubuntu-10.04.4
Martin Pitt (pitti) wrote :

Not fixed in precise yet, no assignee, and the 10.04.4 window is closing fast, dropping milestone.

Changed in base-installer (Ubuntu Lucid):
milestone: ubuntu-10.04.4 → lucid-updates
Rolf Leggewie (r0lf) wrote :

lucid has seen the end of its life and is no longer receiving any updates. Marking the lucid task for this ticket as "Won't Fix".

Changed in base-installer (Ubuntu Lucid):
status: Triaged → Won't Fix
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers