Progress installing on IBM LS21 "Blade" machine
tom at samplonius.org
Sun Mar 18 04:27:14 UTC 2007
----- "Ivan Voras" <ivoras at fer.hr> wrote:
> First, things that don't work:
> * 64-bit kernels. With or without ACPI, on 6-stable or on 7-current,
> kernels in AMD64 mode don't finish booting, or in the best case boot
> can't run SMP (they don't find additional CPUs, but mptable
> is correct - see previous posts). There's even a sort-of regression
> 7-current: while 6-stable without ACPI boots but finds only one CPU,
> 7-current kernel hangs with or without ACPI, either during USB bus'
> detection or just after "parallel port" detection (which is obviously
> not present on blades but still detected...).
The one thing about IBM blades, is that the various controllers for all kinds of IO devices may exist, but are simply not wired. FreeBSD is quite picky (or stupid) about device probing. If it finds a device, it basically demands that it works. It would be great if drivers could simply bail, if the devices were there, but didn't work, rather than just hanging.
You can disable the parallel port device from the boot prompt. The same goes for any other phantom hardware that FreeBSD can't grok.
> * One irritating umass device. It seems that there's an embedded
> (USB mass storage) device in the blade or blade center which is
> in device tree but doesn't respond to any probes, thus hanging the
> process for upto 15 minutes until all timeouts expire. First time
> happened I almost gave up and pronounced it a lost cause, but it
> to be a harmless (if irritating) timeout issue. I've built a kernel
> without umass support, but that means I also lost the built-in CD/DVD
> drive in the chasis.
The umass device might exist, sometimes. This might be part of the remote CD and floppy support. So when you use the remote console to map your local floppy or CD or ISO to the blade, the remote console software makes the device appear as a umass device to the blade. I bet if you used the remote console, and had it share your local floppy to the device, the FreeBSD install would pass.
But a 15 minute timeout for a USB device to respond is insane. 10 seconds would be more realistic.
More information about the freebsd-stable