Re: [uml-user] Detail with one config warning

[Balaco B. <bal...@im...>]

On Mon, Nov 23, 2015, at 13:35, Richard Weinberger wrote:
> Am 22.11.2015 um 16:39 schrieb Balaco Baco:
> >>>>>>> Since I have no root access in the host machine, I must use 'slirp'.
> >>>>>>> Right? Now I'll install it (https://en.wikipedia.org/wiki/Slirp) from
> >>>>>>> source and pass the full path to UML in the command line, e.g.:
> >>>>>>
> >>>>>> Yes.
> >>>>>>
> >>>>>>> eth0=slirp,,/home/me/bin/
> >>>>>>
> >>>>>> Documentation says: ethn=slirp,ethernet address,slirp path
> >>>>>>
> >>>>>> That said, the slirp tool is old and AFIAK unmaintained.
> >>>>>> qemu forked it a long time ago and has it integrated into its
> >>>>>> source.
> >>>>>>
> >>>>>
> >>>>> Documentation of UML also said that ethernet address is optional. I did
> >>>>> both ways, none worked (and it took me a few minutes and tries to find
> >>>>> out that not all MAC addresses are possible to use for a virtual machine
> >>>>> like UML; is this at least mentioned in some UML doc? It shouls, IMHO).
> >>>>>
> >>>>> So, is it possible that I run an UML with full Internet access, and
> >>>>> without having root access in the host machine? Or am I trying something
> >>>>> that from the start was not possible?
> >>>>>
> >>>>> If it is not possible, is it easy to set up some form of file sharing
> >>>>> between the host and UML machine? (again, this must *not* depend on host
> >>>>> root access!)
> >>>>
> >>>> As I said, if you don't have root on the host, you have to use slirp.
> >>>>
> >>>
> >>> And I am using slirp. But it do not work yet. I'm not sure what is
> >>> wrong. May you help me?
> >>
> >> Just gave it a try. Works here.
> >>
> >> ./linux ubda=../Downloads/Debian-Squeeze-AMD64-root_fs mem=1G
> >> eth0=slirp,FE:FD:01:02:03:04,/home/rw/work/slirp/slirp-1.0.17/src/slirp
> >>
> >> root@uml:~# ifconfig eth0 10.0.2.15 up
> >> root@uml:~# route add default gw 10.0.2.15
> > 
> > Do not work for me. And to point directly to the slirp program instead
> > of its folder (which is understood by many as the *path* for anything,
> > not the things themselves, with full path) is news for me, and something
> > that should be pointed in the docs. I think many others will understand
> > it as I did before you said this.
> 
> "Does not work" is not a valid error description.
> Please find out what exactly does not work.

There is no clear error. In this paragraph I just pointed to something
in the instructions that mislead me to give the path to the folder
*where* slirp was, instead of the slip program itself, its file. The
most I could see that did not work I said right after, below in the
previous message. Please assume and try to see that I'm trying my best
to follow your directions, but I still could not do it although you just
said you did.

You may also ask me to directly do things or steps that you do (maybe
it's something I don't know or do that make it don't work).


> > Now there is a few lines about slirt and ETH0:
> > 
> > =======================
> > [autodetect SLIP/CSLIP, MTU 1500, MRU 1500, 115200 baud]
> > 
> > SLiRP Ready ...
> > Unsupported device type 256 for "eth0"
> > Failed to bring up eth0.
> > done.
> > 
> > ========================
> 
> DHCP cannot work on SLIRP.
> That's why you have to configure the interface manually.
> 
> > What can I do?
> > 

If configure the interface manually is to type the commands (inside UML)
 below, that I did before:

root@uml:~# ifconfig eth0 10.0.2.15 up
root@uml:~# route add default gw 10.0.2.15

It did not work.

What information is useful for you? I don't know. So, since you seem to
have thought I did not try enough before, now I'm sending the whole boot
log, and each step I did before and repeated right now, after I read
your message.


> >> Then you can use networking. Please not that slirp does not have a
> >> performance.
> >> Depending on your slirp package you can tweak it.
> > 
> > I have compiled it from source. Do you know exactly where and how I can
> > tweak it? If so, please say how, so I don't have to search and find what
> > is already known.
> 
> AFAIK Debian folks have some patches in their package.

I choose the best distro I know for the intended uses (RedHat and sister
is usually not that good, but I don't discard using them to start). I
did not use Fedora or RedHat that even the blog shows because I don't
want and don't need graphical interface. Just CLI is perfect, just fine.

If you can point which patch or which feature Debian patched I can work
around it. Right now I know that it may have some problem, but I can
even look for help my self or anywhere else because I don't know what it
is - not much clues I can tell.

> Thanks,
> //richard

Ironically, doing the same steps I did before it seemed to work today.
:-/ I hate when this happens... hopefully it won't stop tomorrow, with
another ironic happening.

Now, with a close to work situaton, I guessed that it just needed to
find the DNS the host uses (easy, fine), and then I repeat the route add
command. But this is not true. Before I update UML's /etc/resolv.conf
the only IP I can ping from UML is 127.0.0.1. What else do I need to do
to have network in UML?

Thank you,

-- 
  Balaco


-- 
http://www.fastmail.com - Choose from over 50 domains or use your own