Bug#883829: Fw: Bug#883829: systemd: Fully qualified host names in fstab results in mounts failing

[Duncan Hare]

 Michael


u-boot is a bootloader. The u-boot script load  the kernel from a server,, constructs the kernel commandline, 
and the u-boot script is complete when the kernel starts, well before these failures. 

In the kernel command line there is a dns server address, 192.168.1.12 passed to the linux kernel on the kernel command line.  

How the DNS server address is passed to resolveconf by the kernal I do not know. That's an interface between the kernel and systemd.
As I wrote, I can define the Ethernet interface as dhcp if that's what is required to populate the files.
Regards 
Duncan Hare

714 931 7952

      From: Michael Biebl <biebl at debian.org>
 To: Duncan Hare <dh at synoia.com>; 883829 at bugs.debian.org 
 Sent: Friday, December 8, 2017 10:33 AM
 Subject: Re: Bug#883829: systemd: Fully qualified host names in fstab results in mounts failing
  
Am 08.12.2017 um 01:44 schrieb Duncan Hare:
> Michael
> 
> Full DNS is provided by an MS Windows server 2012R2.
> 
> Because it is net boot, the network in enabled before the kernel is
> loaded by u-boot,
> and the kernel starts and mounts the file system over NFS.
> 
> Networking parameters, including DNS servers are pass through the kernel
> parameters.
> 
> Kernel parms are (these include a lot of raspberry pi stuff), all one
> line no carriage returns
> 
> dwc_otg.lpm_enable=0 console=serial0,115200 console=tty1 root=/dev/nfs
> rootfstype=nfs nfsrootdebug elevator=deadline rootwait
> nfsroot=192.168.1.10:/nfsroot/r.32.test,tcp,vers=3
> ip=192.168.1.132:192.168.1.10:192.168.1.1:255.255.255.0:abcdef::off:192.168.1.12:192.168.1.22
> 
> 192,168,1,132 is the raspberry pi machine
> 192.168.1.10 is the linux file server
> 192.168.1.1 is the default gateway
> 255.255.255.0 is the netmask
> abcdef is the hostname in this example
> 192.168.1.12 is one dns server
> 192.168.1.22 is the second dns server
> 
> We don't use /dev/nfs in fstab, due to a bug in debian stretch mount
> command (a previous bug report) - which has undergone much change
> because of the
> introduction of fiilesystem ids replacing device names.
> 
> the parameters are completed as above by u-boot script.
> 
> U-boot does a dhcp command to get the parameters.

So if I understand you correctly, there is a u-boot script which sets up
/etc/resolv.conf. How exactly does this u-boot script look like and how
is it started. How do you ensure that this script runs before systemd
attempts the NFS network mounts?

This looks like another case of a misconfiguration.
Systemd can not magically know that it has to wait for a u-boot script
to finish before it attempts any network mounts.
This is for you, the admin, to setup correctly or the u-boot script
(assuming it ships a systemd service) to hook into the correct targets.



> The Linux networking is left as configured in Debian. It can be
> configured as manual or static
> in /etc/networking/interfaces with little or no change in behavior. If
> configure with dhcp Linux
> just get another lease on the same Ethernet address, because the
> mac address is the same for u-boot and linux.
> 
> In this configuration networking is the prereq to file systems because
> we are booting a ZFS machine, with as much of the file system ro as possible
> to have a very secure system. At the application layer we plan to run
> Citrix which provides excellent application control.
>  

That's not relevant to this issue.


-- 
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?


   

   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/pkg-systemd-maintainers/attachments/20171208/e21c7e6f/attachment-0002.html>