another init.xz issue

Tom Elliott

@george1421 has_usb_nic=1

bmaster001

Whene I add that, it asks me to unplug and replug my device into the usb port and press enter. Can’t do that, so I just pressed enter

Anyway: the result is the same: another random mac address…

[Tue Jun 21 root@fogclient ~]# ip addr show
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue qlen 1
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast qlen 1000
    link/ether a6:3a:48:bb:24:65 brd ff:ff:ff:ff:ff:ff
    inet 10.1.14.75/16 brd 10.1.255.255 scope global eth0
       valid_lft forever preferred_lft forever

Have you seen the link to some driver-source I posted above? http://lxr.free-electrons.com/source/drivers/net/usb/smsc95xx.c#L788

george1421

@bmaster001 I saw the kernel driver, but not sure what you wanted to tell me. I assume the driver is built into the current FOS kernel otherwise you would not get an ip address.

I think we need to get the kernel developers to look at this issue. They may need to check to see if the current kernel supports this specific nic or one that is close. The device (hex) ID you provided below will help. I think I’m at the end of my ability to debug this @Developers ??

bmaster001

@george1421 said in another init.xz issue:

@bmaster001 I saw the kernel driver, but not sure what you wanted to tell me. I assume the driver is built into the current FOS kernel otherwise you would not get an ip address.

In that source you can see that the driver tries to get a MAC address from the EEPROM but if that fails, it generates a random one. I thought that it might help to know that it’s the driver who does this, and not some weird config option somewhere…

Another thing that I’ve tried is to boot in debug mode, and manually add a host to the fog server with the random mac. Then schedule a capture task, and run ‘fog’ on the host. It fails with the “unknown request type :: Null” error. Maybe you know if there’s a way around that?

Thanks a lot for you help anyway, I’ve learned a lot! Let’s hope someone else with a deeper knowledge of this stuff is willing to look at this!

FYI: I’ll be on vacation from coming friday until the 3rd of july…

george1421

@bmaster001 Gotcha on the drive. I did not read it line for line. (that is a silly feature)

As for your hack around this. If you look in grub for the debug entry, there is a kernel parameter isdebug=yes If you copy and paste that onto the end of the capture / deploy line. Your capture deploy entry will be in debug mode. Do that boot, and then try your trick of updating the mac address in fog, schedule a capture/deploy. Then on the FOS console key in fog. That might just work.

bmaster001

That seems to work! I’m now stuck on the ntfs inconsistency again that’s not fog’s fault… I’ll have to find a way to fix that too.

Before doing what you described, I took a loog at the contents of /bin/fog . It starts like this:

### If USB Boot device we need a way to get the kernel args properly
if [[ $boottype == usb && ! -z $web ]]; then
    mac=$(getMACAddresses)
    wget -q -O /tmp/hinfo.txt "http://${web}service/hostinfo.php?mac=$mac"
    [[ -f /tmp/hinfo.txt ]] && . /tmp/hinfo.txt
fi $

The $web part seems to be a problem, because that variable doesn’t exist. So it skips the getMACAddresses-part…

george1421

@bmaster001 When you enter in debug mode none of the fog scripts have run. So the $web won’t be set until . /usr/share/fog/lib/funcs.sh has been called. I know it is confusing, but I ran into the same issue when developing the code snippet you posted.

$web should point to your fog server when its set. Actually <fog_server_ip>/fog (this comes from the kernel parameter in the GRUB config file). The funcs.sh script extracts the kernel parameters and then sets the bash variables to match the kernel parameters. Then my code calls the fog server to pick up the other dynamic kernel parameters than can’t be supplied by the static grub.cfg file. That is where the hostinfo.php comes in. The type variable is set by the hostinfo.php script, fwiw.

bmaster001

It’s a difficult mechanism to understand. A lot of unfamiliar technologies, scripts and boot-arguments working together. It’ll take a while for me to better understand it. But it’s a very interesting subject!

The very first line in the fog script is the call to funcs.sh. when I run that on the command line, and do set, I see no web variable:

_=/usr/share/fog/lib/funcs.sh
acpi=off
boottype=usb
consoleblank=0
has_usb_nic=1
init=/sbin/init
isdebug=yes
ismajordebug=0
loglevel=7
name=has_usb_nic
oIFS='
'
ramdisk_size=127000
root=/dev/ram0
rootfstype=ext4
value=1
var=has_usb_nic=1

george1421

@bmaster001 hey that’s cool (not).

Lets confirm that your grub.cfg has this line.

menuentry "1. FOG Image Deploy/Capture" {
 echo loading the kernel
 linux  $myimage loglevel=$myloglevel initrd=init.xz root=/dev/ram0 rw ramdisk_size=127000 keymap= web=$myfogip/fog/ boottype=usb consoleblank=0 rootfstype=ext4 isdebug=yes
 echo loading the virtual hard drive
 initrd $myinits
 echo booting kernel...
}

Note the kerenel parameter web=$myfogip/fog/

[edit] note web will only be there when you are in capture/deploy menu mode not in the debug mode from the menu entry.

george1421

@bmaster001 said in another init.xz issue:

It’s a difficult mechanism to understand. A lot of unfamiliar technologies, scripts and boot-arguments working together. It’ll take a while for me to better understand it. But it’s a very interesting subject!

I understand this (linux) is a alien world to most windows folks. But you see there IS a lot of magic that goes into make FOG functional and universal.

You have the unfortunate situation of having very strange hardware (maybe one of a kind) that make imaging difficult to manage. FOG has a lot of instruments in the band, its takes a bit of cohering to get them to all play the same tune.

Sebastian Roth

@bmaster001 said:

In that source you can see that the driver tries to get a MAC address from the EEPROM but if that fails, it generates a random one. I thought that it might help to know that it’s the driver who does this, and not some weird config option somewhere…

Great find! This is really unusual. Have never seen this before! Why would if not be able to read the MAC I wonder?!? Random MAC is kind of breaking the whole FOG logic of identifying hosts via MAC address.

bmaster001

@george1421 said in another init.xz issue:

[edit] note web will only be there when you are in capture/deploy menu mode not in the debug mode from the menu entry.

Ok, that explains why the $web variable doesn’t exist in debug-mode, and why the fog script doesn’t work then.

I understand this (linux) is a alien world to most windows folks. But you see there IS a lot of magic that goes into make FOG functional and universal.

You have the unfortunate situation of having very strange hardware (maybe one of a kind) that make imaging difficult to manage. FOG has a lot of instruments in the band, its takes a bit of cohering to get them to all play the same tune.

I do have some linux knowledge, and understanding some scripts would not cause too many problems. But it’s the (network)boot-process that I don’t know anything about. Like you say, trying to understand that, together with the fog scripts, and some weird hardware, makes the step too big. If I really want to learn how fog works, I’d better start with some normal hardware

bmaster001

@Sebastian-Roth said in another init.xz issue:

@bmaster001 said:

In that source you can see that the driver tries to get a MAC address from the EEPROM but if that fails, it generates a random one. I thought that it might help to know that it’s the driver who does this, and not some weird config option somewhere…

Great find! This is really unusual. Have never seen this before! Why would if not be able to read the MAC I wonder?!? Random MAC is kind of breaking the whole FOG logic of identifying hosts via MAC address.

Exactly. And I’m pretty sure there IS a mac address in the nic: when I enable uefi and let it network boot, I see always the same mac address. Would it be possible that it uses a different nic with uefi disabled?

Sebastian Roth

@bmaster001 said:

Would it be possible that it uses a different nic with uefi disabled?

Can’t imagine that being the case. Well on the other hand I’ve never seen a driver generating a random MAC for a real hardware NIC… So what do I know!

lol, just found this. Seems like the driver used to generate a new random MAC even on ifup/ifdown. Along that I found something interesting in the kernel code. See here (line 771):

/* maybe the boot loader passed the MAC address in devicetree */

Here you find a description of what that of_get_mac_address function does. I wasn’t able to find out if you can actually set this via iPXE or as a kernel parameter. Anyone an idea?

bmaster001

We’ve decided for now to just boot to Windows, and set the thing up so that we can test it “on the floor”. If it behaves like it should, we’re gonna buy some other ones to replace older devices currently in use. When we do that, I’ll have more time to play with capturing/deploying it so make our life easier So you guys have plenty of time to think about this weird piece of hardware, and I’ll get back to this thread when we have more devices to play with!

Thanks all for the help so far!!

Sebastian Roth

@bmaster001 said:

So you guys have plenty of time to think about this weird piece of hardware, and I’ll get back to this thread when we have more devices to play with!

I don’t think we are able to push this anywhere without having the hardware around or someone like you who are willing to test things out.

another init.xz issue

158

12.0k

17.3k

155.2k