Re: Outstanding /dev/.udev bugs for /run migration
On 13.01.2012 03:18, Roger Leigh wrote:
> All are currently broken. Whether they cause severe breakage depends
> upon the individual case. Some are working apparently OK, e.g.
> mdadm. Others are doing broken things to try and create device nodes
For mdadm an upstream patch is needed. I already removed all refs to
/dev/.udev from surrounding scripts and plan to patch remaining bit
in the source. It will still use /dev/.udev to be able to work on
older systems but will check both it and /run/udev. FWIW. I'd love
to assume udev is always present but for mdadm it is not a good idea
since it may be called in initramfs without udev or when udev is
somehow broken, and mdadm needs to be functioning regardless.
Besides, when mdadm detects lack of udev it merely creates /dev/md*
nodes itself, which does not do any (visible) harm - udev just
This missing bit is why I didn't close the bug in question when
did initial changes (#627774 and #644319).