User Feedback :: Still USB boot problems with the final 2.1 :(



As far as I could see there are a number of hardware configurations in which USB booting is not working :(

However in MOST of them the USB itself is working or partially working after the full boot - so I took a look at the minirt24.gz from DSL - is it normal that the driver modules from minirt24.gz (which are the only one used at boot time) are DIFFERENT from precisely the same module in the full image ? I feel quite strange especially about the uhci part which seems to be strongly related to the problem :(

I managed to make a minirt24.gz that works on some of machines on which the default one was not working but the problem does NOT seem to be totally solved by updating ALL the modules in minirt24.gz - actually my best results are with my very first attempt in which only about half of them were updated !!!
You can compare the files by creating an md5sum checksum for one and then comparing it to the other file.

This will tell you if it is the EXACT same file even if the timestamp is different.

But the modules should be exactly the same unless you are using an old version of the usb boot floppy.  Or if you manually UPGRADE your usb drive KNOPPIX file and do not upgrade the rest (linux24 minirt24.gz etc).
There are many issues with USB booting. Here is some background as we have tried to address some of them.

One that became apparent with 2.0 is that we tried acpi=off as a default boot parameter. We found some systems wouldn't boot USB especially the ebox3 machines. This in fact is the only time John or I have seen an issue with USB booting. That is not to say there are not other issues. With 2.1 we dropped this parameter. The ebox3 machines boot usb ok. Now, I have already seen a post that 2.0 worked them and not 2.1 So some machines may need it others may not. We can't have it both ways.

Another common issue is the timing of the USB ports.
Front ports are slower than back ports to register the device.
We made the bootfloppy-usb.img, which is specifically for booting USB devices have a delay to improve time to register.

The standard cdrom images, of course have a much wider array of boot devices to support and an order of searching devices and we don't want to unnecessarily want to delay the bootup for everyone.

Then we used to require net access for installation to pendrives, this allowed the script to download the special bootfloppy-usb image to be used for the pendrive. But users wanted not to require net access, so now the script use the standard boot image files.

Then there is the issue, as is always the case, John and I do not have every conciveable device to test with. That is why we now have an open public release candidate testing. Still, when we cannot reproduce or see the issue it can become difficult to solve.

But we are both committed to trying to constantly improve DSL with each release. But as with any software product perfection is elusive.

On all my dual boxen, using the parameter
"acpi=off" keeps the dreaded kernel oops away at boottime.
( using noacpi does NOT work successfully for me.. )

But yes, then not only is USB booting affected,
but even detecting or reading a pendrive fails.

I now use the combination 'acpi=off' with 'noapic' in my
grub boot menu.lst, which solves both problems for me.

Boots successfully, AND sees/allows use of USB devices.

Hope it helps..

ke4nt
Let me also if I may quote the master himself, Klaus Knopper:

"...some boards have serious errors in their implementation of the ACPI standard, and Linux tries to provide more or less automatic work-arounds to turn off features like incorrect interrupt routing or CPU frequency mis-calculations."

Some times it is help to try all of these boot options together:
acpi=off noapic pnpbios=off pci=bios

and sometimes you may have to use
acpi=force

or

irqpoll
Next Page...
original here.