>>>>> "JH" == Jack Howarth <howarth@xxxxxxxxxxxxxxxxx> writes: JH> Has anyone heard anything specific yet about the origin of JH> this bug introduced with the 2.6.10 kernels? All I know is that it broke USB key mounting on over a hundred machines for me, but that I was able to hack around it. My observations are recorded at the end of https://bugzilla.redhat.com/beta/show_bug.cgi?id=119140, but I think the underlying cause of the problem originally reported there is different than the one which started with 2.6.10. Perhaps you could try this patch to /etc/hotplug/hotplug.functions. It's just a nasty hack, but I've rolled it out across my network and so far the problem has gone away. This really needs to be fixed within the kernel. --- a/etc/hotplug/hotplug.functions.orig 2005-01-31 11:57:59.000000000 -0600 +++ a/etc/hotplug/hotplug.functions 2005-01-31 12:06:45.000000000 -0600 @@ -146,6 +146,7 @@ # parameters ... handle per-device parameters in apps # (ioctls etc) not in setup scripts or modules.conf LOADED=true + JUST_LOADED=true fi else # This module is already loaded @@ -167,7 +168,19 @@ mesg "missing kernel or user mode driver $MODULE " fi if echo "$MODULE" | grep -q "usb-storage" > /dev/null 2>&1 ; then - [ -x /usr/sbin/updfstab ] && /usr/sbin/updfstab + if [ -x /usr/sbin/updfstab ]; then +# INITIAL_DEVICES=$(ls -1 /sys/bus/usb/drivers/usb-storage | wc -l) + INITIAL_SCSI=$(cat /proc/scsi/scsi | wc -l) + COUNT=10 # (big) upper bound of the loop + while [ $COUNT -gt 0 ]; do + sleep 1 +# [ "$(ls -1 /sys/bus/usb/drivers/usb-storage | wc -l)" -gt "$INITIAL_DEVICES" ] && + [ "$(cat /proc/scsi/scsi | wc -l)" -gt "$INITIAL_SCSI" ] && break + COUNT=$(($COUNT - 1)) + done + fi + + /usr/sbin/updfstab fi done } JH> Hopefully it will be addressed in 2.6.11. That's not going to happen unless somebody fixes it. I haven't seen others complaining about it, so I thought it was somehow related to something I was doing. Are others seeing this bug? - J<
