Bug#110717: more on partitioning drama
The "no device name for disk ... partition .. (skipped)" message seems to be
triggered if the kernel hasn't managed to re-read the partition table off
the disk (and hence /proc/partitions is out of date).
The code in partition_config.c goes to some lengths to unmount and swapoff
appropriately before repartitioning in order to avoid this situation; my guess
is that for some reason this isn't working out. I think the prime suspect
would be swapon_partitions not containing the right data; in a quick look at
the code it seemed that the only place this really gets set is a section
of libfdisk/fdisk.c that's #ifdef'd out.
It would probably be worth having partition_disk() issue a BLKRRPART ioctl
itself after fdisk runs, with failure here having the same effect as
"root_on_disk", namely concluding that something is wrong and rebooting
the machine.
Reply to: