Hi Mika,
I've returned from vacation. Good news on bug isolation.
Booting grml 1.0 final CD-ROM with "noudev" made USB behave normally. So: the bug may be udev rules rather than a kernel regression. (We can discuss privately.)
About the fscking test: it would not have helped us here, while other tests had real help potential (2.6.20-stable releases, kernel flags, boot options). The next beta will have 2.6.22 anyway, and being a full distro, can deploy to my users.
2.6.21-plus-grml-1.0 is not a distro. We need pre-built packages and modules, not some kernel++ that requires me to build them. I can't support users that way. Small -stable increments, maybe; that was why I asked.
.12 does not include anything else than a bugfix for GEODE-AES: http://kernel.org/pub/linux/kernel/v2.6/ChangeLog-2.6.20.12
The particular delta number did not matter. My thrust was that recent and significant USB activity in 2.6.20-x (stable patches) meant USB bugs were *NOT* specific to our hardware.
Wernfried - impossible perfection is beside the point. A beta program needs to avoid sweeping assumptions about weird hardware and kernel regressions. Otherwise there's no utility in it. Many bugs are just dumb typos in scripts and config files...like udev rules.
Thanks again, M