Linux – How to make a SquashFS file the root file system when doing an iPXE boot

linuxnfspxesquashfsUbuntu

I have a system that works perfectly fine as iPXE boot server. Now I'm trying to make it use a SquashFS file as a root file system, instead of an NFS file system.

The current system uses iPXE configuration as the following (relevant lines shown), and it's working fine:

:retry
kernel http://${next-server}/installdcos/ubuntu_os/vmlinuz root=/dev/nfs nfsroot=${next-server}:/installdcos/ubuntu_os/nfsroot ro ip=dhcp BOOTIF=%(mac)s console=ttyS0,38400n8 console=ttyS1,9600n8 console=tty0  || goto retry
initrd http://${next-server}/installdcos/ubuntu_os/initrd.gz || goto retry
boot || goto retry

Now, instead of the booted system use /installdcos/ubuntu_os/nfsroot as the root file system, I want it to use a single SquashFS as root file system (I mean the system should mount it and then use it as the root file system).

For this, first I've created a file ubuntu_os.squashfs that holds the contents of nfsroot directory using mksquashfs. Then I've placed this ubuntu_os.squashfs file in /installdcos/ubuntu_os/nfsroot directory.

Now, in my iPXE configuration, I want to say: please get the ubuntu_os.squashfs via network, mount it as a SquashFS file system, and use it as your root file system (so that you will not have go over NFS again for invoking commands that reside in the root file system).

But when I try the following change for the kernel line in the configuration:

kernel http://${next-server}/installdcos/ubuntu_os/vmlinuz root=/dev/nfs nfsroot=${next-server}:/installdcos/ubuntu_os/nfsroot/ubuntu_os.squashfs ro rootfstype=squashfs ip=dhcp BOOTIF=%(mac)s console=ttyS0,38400n8 console=ttyS1,9600n8 console=tty0  || goto retry

And do an iPXE boot from a machine, it starts the iPXE boot process, and after printing some lines, it gets stuck at the following error:

  Begin: Retrying nfs mount ... Begin: Running /scripts/nfs-premount ... done.
  mount: Not a directory
  done
  Begin: Retrying nfs mount ... Begin: Running /scripts/nfs-premount ... done.
  mount: Not a directory
  done
  Begin: Retrying nfs mount ... Begin: Running /scripts/nfs-premount ... done.
  mount: Not a directory
  done

The error message makes sense, indeeed, /installdcos/ubuntu_os/nfsroot/ubuntu_os.squashfs is not a directory.

So what kind of parameter and value combination should I write for that kernel configuration line?

I've checked documentation such as http://www.tldp.org/HOWTO/BootPrompt-HOWTO-3.html and https://www.kernel.org/doc/Documentation/filesystems/nfs/nfsroot.txt, but could not figure out what magic incantation I should pass to kernel (and the rest of the SquashFS boot tutorials I found consider the boot operation taking place from disk, and not PXE, or doing it LiveCD, which is not exactly like my case).

Best Answer

For understanding what's going on you need to understand how Linux PXE boots.

The kernel vmlinuz is transferred
The initrd initrd.gz is transferred
The kernel mounts initrd and starts its init script

initrd is an ultra-minimal Linux system containing the minimal functionality for connecting to (NFS) or retrieving (HTTP/CIFS) the "real" OS (in your case contained within ubuntu_os.squashfs), mounting it and finally "chrooting" into it.

A PXE ready init script is in charge of parsing the kernel variables, starting the net services, handling NFS, HTTP, CIFS, etc. In most of the cases a particular initrd.gz is not able to deal with all the protocols mentioned above then you need to "customize" your initrd.gz or to create a "complementary" initrd providing the missing features when needed.

In your case your init script still think it has to mount an NFS directory and not a file; then if you want to use NFS you should mount the directory where your ubuntu_os.squashfs is located and next mounting the ubuntu_os.squashfs file. THis of course means patching init (and/or its associated components)

If you do not mind offering ubuntu_os.squashfs on a CIFS share you can see what Serva does for PXE booting Ubuntu live distros; you will see there all the parameters for CIFS booting (I'm related to Serva development)

i.e. Ubuntu LTS 14.04 Desktop Live

[PXESERVA_MENU_ENTRY]
asset    = Ubuntu LTS 14.04 Desktop Live
platform = amd64
kernel   = /NWA_PXE/$HEAD_DIR$/casper/vmlinuz
append   = showmounts toram root=/dev/cifs initrd=/NWA_PXE/$HEAD_DIR$/casper/initrd.lz,/NWA_PXE/$HEAD_DIR$/casper/INITRD_N11.GZ boot=casper netboot=cifs nfsroot=//$IP_BSRV$/NWA_PXE_SHARE/$HEAD_DIR$ NFSOPTS=-ouser=serva,pass=avres,ro ip=dhcp ro

Related Solutions

Linux – Doubts About the Linux Root File System

The mount command takes information about current mounts from /etc/mtab.

In the past, mtab was a normal file re-created after every boot and updated by the mount command – so it wouldn't have a rootfs / entry simply because the rootfs is never explicitly mounted; it just always exists. (Such a mtab sometimes also has duplicate entries, or entries for filesystems that aren't mounted anymore...)

Many current distros now symlink mtab to /proc/self/mounts, which (like everything else in /proc) is directly generated by the kernel. Because of this, it always shows everything that is currently mounted, including both rootfs / and /dev/blah /.

You can use cat /proc/self/mounts on all distros to compare the mounts. (There also is /proc/self/mountinfo, which uses an incompatible syntax but adds more detail.)

Linux – How to get linux to stop complaining when the nfs server is down

Edit: Since the problem only occurs when the NFS server goes down AFTER the share has already been successfully mounted, the previous answer does not help.

My only suggestion now is to use autofs to mount the NFS shares on-demand, instead of mounting them on boot. This way, the system will only attempt to mount the NFS shares when you try to access them, and will automatically unmount them after a period of inactivity.

There is a guide on how to use autofs available here, but below I will list the steps you should need to follow to get something running.

Make sure autofs is installed. The package is usually just called 'autofs'

On Ubuntu, you can install it by running apt-get install autofs as root.
Add the following entry (as root) to /etc/auto.master
```
/home/Shares /etc/auto.shares
```

Create the file /etc/auto.shares (as root), with the following content:

TV_SHOWS    -rsize=8192,wsize=8292,timeo=14,intr,rw    192.168.1.200:/mnt/Raid/TV_SHOWS
Movies      -rsize=8192,wsize=8292,timeo=14,intr,rw    192.168.1.200:/mnt/Raid/MOVIES
DocuTV      -rsize=8192,wsize=8292,timeo=14,intr,rw    192.168.1.200:/mnt/Raid/DocuTV

Remove the static mount points for these directories from /etc/fstab
Reload the autofs service. On Ubuntu, you can run service reload autofs as root to do this.
Test to make sure it works. Run ls /home/Shares/TV_SHOWS and see if it lists your files.

Note: Due to additional information added to the question, the following answer no longer applies. I will leave it here, however, in case it helps someone in the future.

You could try adding the nofail option to the entires in /etc/fstab.

From the mount manpage:

nofail

    Do not report errors for this device if it does not exist.

This should cause the system to attempt to mount the device on boot, but give up if it is unable to do so. You can then manually attempt to mount the device later.

There is another option that may be useful: retry (nfs manpage)

retry=n

    The number of minutes that the mount(8) command retries an NFS mount
    operation in the foreground or background before giving up. If this
    option is not specified, the default value for foreground mounts is 2
    minutes, and the default value for background mounts is 10000 minutes
    (80 minutes shy of one week). If a value of zero is specified, the
    mount(8) command exits immediately after the first failure.

By default, NFS mounts are foreground mounts, and thus should fail after 2 minutes of attempts. You could always manually set this to a certain amount of time to cause it to fail more quickly if the NFS server is unavailable, which combined with the nofail option, could get it to work as you want.

Best Answer

Related Solutions

Linux – Doubts About the Linux Root File System

Linux – How to get linux to stop complaining when the nfs server is down

Related Question