Description of problem:

When using a read-only nfs share as the root filesystem, the fedora-readonly service stalls forever at "A start job is running for Configure read-only root support".

I was able to resolve the problem to my satisfaction by commenting some lines that were not present in Fedora 18, but I want to let you know about the change in behavior after Fedora 18.

Version-Release number of selected component (if applicable):

initscripts-9.47-1.fc19.x86_64

How reproducible:


Steps to Reproduce:
1. Prepare a Fedora setup to be used as a stateless read-only root filesystem and export it with nfs:

A: Modify the variables in /etc/sysconfig/readonly-root:
READONLY=yes
TEMPORARY_STATE=yes

B: Remove all lines in /etc/fstab

2. Prepare an init ram filesystem for network boot with these commands:

yum -y install dracut-network
dracut -N initrd-fc19.img `uname -r`

3. Make the kernel and the new init ram filesystem available to the diskless client and have it boot the Fedora kernel with theses arguments (we use pxelinux):

initrd=diskless/F19/initrd-fc19.img
ip=dhcp
root=nfs:132.207.12.3:/store/diskless/F19GIGL:nolock

Actual results:
The start sequence stalls at the following message:
"A start job is running for Configure read-only root support"

Expected results:
In Fedora 17 and Fedora 18, this method would provide a Fedora startup that worked exactly as it did on the originally installed computer.


Additional info:

I was able to make it work again in Fedora 19 by commenting these lines in
/usr/lib/systemd/fedora-readonly:

# if mount | grep -q /var/lib/nfs/rpc_pipefs ; then
#        mount -t rpc_pipefs sunrpc /var/lib/nfs/rpc_pipefs && service rpcidmapd restart
# fi


So, I guess the problem might be with the restart of the rpcidmapd service.