Boot hangs after init.xz on Dell OptiPlex 9030 All-In-One
-
We have several teaching labs full of computers that we’ve been using Fog with for many years (thanks!), but we recently got a batch of Dell Optiplex 9030 computers, and they’re refusing to pxe boot. They hang right after:
bzImage… ok
init.xz… okWe’ve been using undionly.kpxe for the machines, but I’ve tried switching to ipxe.kpxe and intel.kpxe with the same results. I’ve also tried downloading the latest versions of these, but it still didn’t work.
I notice that the screen flashes momentarily after init.xz, as though it’s trying to do something else, but nothing ever happens. The only log entries I see on the server look like this:
Jun 28 13:28:38 mrwizard-apn dhcpd: DHCPREQUEST for 192.168.104.56 (192.168.104. 2) from 74:e6:e2:e1:ab:1f via eth1 Jun 28 13:28:38 mrwizard-apn dhcpd: DHCPACK on 192.168.104.56 to 74:e6:e2:e1:ab: 1f via eth1 Jun 28 13:28:38 mrwizard-apn in.tftpd[2015]: Error code 0: TFTP Aborted Jun 28 13:28:38 mrwizard-apn in.tftpd[2016]: Client 192.168.104.56 finished intel.kpxe
Notice the “TFTP Aborted” message. I’ve tried fiddling with the tftp block size, but that doesn’t seem to have any affect either.
I’d appreciate any advice anyone has to offer.
UPDATE:
I tried booting 64-bit tiny core linux from a USB stick on this computer, and it also hangs. In this case, it stops after “Can’t find IRQ for PCI INT A; probably buggy MP table”. -
@Bryan-Wright Increase the Kernel loglevel in FOG configuration in the web UI and see if you get more output this way.
-
@sebastian-roth
OK, with the kernel log level cranked up to 7, I see the same error message that tiny core shows me “Can’t find IRQ for PCI INT A…”. Here’s photo of the screen after it hangs:
-
@Bryan-Wright Ok, so FOS boot as far as the other tiny Linux OS you tried:
can't find IRQ for PCI INT A; probably buggy MP table
Not sure what this means. Searching the web I find many posts with this message but mostly with way older Kernel versions than what we use with FOS.
By the way, which version of FOG do you use and which Linux kernel version?
Try updating to the latest kernel (5.10.34 as of now) and see if that works on the Dell OptiPlex 9030 All-In-One.
-
@sebastian-roth
This is fog 1.5.5. I’ve just now tried downloading and booting the 5.10.34 kernel, but it behaves the same (“probably buggy mp”).I suspect the solution will involve some combination of acpi and apic flags on the kernel. I’ve tried a few, but haven’t found the right mix yet.
-
@bryan-wright Is your firmware (bios) updated on this computer? This is typically a firmware issue.
-
@george1421
Yes, that was one of the first things we tried. The firmware is the latest available. -
@george1421 @Sebastian-Roth
Aha! The magic seems to be “acpi=noirq”. Now I just need to figure out how to get fog to apply that to only this subset of machines… -
@bryan-wright
I’ll open a new topic on the forum about this, but the problem I’m running into now is that we have some kernel arguments set under “general settings” that we use for all clients. These are “pci=nomsi acpi=off”. But these settings won’t work with the recently problematic machines. For those I’d like to just have “acpi=noirq”. It looks like the per-host kernel arguments get concatenated onto the global ones, though.In any case, temporarily setting the global kernel args to “acpi=noirq” allows me to image the currently problematic machines.
-
@bryan-wright said in Boot hangs after init.xz on Dell OptiPlex 9030 All-In-One:
It looks like the per-host kernel arguments get concatenated onto the global ones, though.
Yes, correct, all kernel args are pulled together. Tough I can’t tell you why it was coded this way back in the days. I fear changing this behaviour in FOG will cause trouble for many people. But you could modify
/var/www/html/fog/lib/fog/bootmenu.class.php
on your FOG server to skip the global kernel args when it’s set for a specific host.