user-mode-linux-user Mailing List for User-mode Linux kernel port (Page 794)

On Sun, 19 Nov 2000, Jeff Dike wrote:
> phi...@in... said:
> > This is the command I used to start uml:
> >   ./linux ubd0=/dev/hda6
> > and it doesn't resemble the documentation too much... 
> 
> No it doesn't.  I guess there's a bootable filesystem on your /dev/hda6, and 
> I'm somewhat amazed that UML managed to boot it (most distros run hwclock or 
> something similar, which runs some privileged instructions, which UML doesn't 
> handle yet).
> 
> Did you grab a bootable filesystem from my site?  Uncompress it, rename it to 
> root_fs and run "linux", and it will boot.  Don't rename it, and run "linux 
> ubd0=root_fs_whatever", and it will boot.  Or "linux ubd7=root_fs_whatever 
> root=ubd7".
> 
> It's pretty straightforward.

I used the root_fs_debian2.2_small from your site.  I didn't know about
devfs, and even when I did know about devfs, finding useful
documentation on devfs wasn't easy.  I'm currently running it without
devfs, and that may explain why I can only boot in 'single' mode at the
moment.

It seems straightforward now, but it didn't when I had to guess about
it.  I spent a lot of time trying variations of 

	linux root=root_fs_whatever

because the error messages seemed to suggest that's what I should do. 
Also, I was unable to make the connection between ubd/0 and ubd0 on my
own, and there is nothing about this in the howto, or the step-by-step
procedures on the site.

How about building a 'with devfs' and 'without devfs' debian kernel, so
at least that variable goes away.  At least it would be worth
mentioning the need for configuring devfs here:

  http://sourceforge.net/project/shownotes.php?release_id=15092

I note that it does say "if you have trouble, try the Red Hat kernel
instead", but...

Another thing worth mentioning is "if you have trouble booting, then try
to boot to single mode".  You may consider that obvious, but when there
are so many new variables to consider, sometimes it helps to state the
obvious.

Anyway, I have enough now to do the VFS development I had in mind. 
Now - on to figuring out how to gdb it.

-- 
Daniel

xi...@bo... said:
> It looks like it's stuck in __wait4() for some reason. 

That's where it's supposed to be.

You could try getting a backtrace the old way.  Look at 
current_task.thread.extern_pid, and kill -STOP it.  With gdb attached to the 
tracing thread 'call detach(that_pid)'.  If it segfaults, do it again.  Detach 
from the tracing thread and attach to the other pid.  If the attach hangs, 
kill -CONT it.  Then get a backtrace.

You might see that the UML debugger is a huge improvement over what preceded 
it :-)

Tonight, I did get a situation where the kernel debugger hung.  It turned out 
that it was getting an infinite stream of segfaults for some reason.  If I can 
reproduce it, I'll see what was going on and fix it.

				Jeff

Jeff Dike <jd...@ka...> writes:

> No it doesn't.  I guess there's a bootable filesystem on your /dev/hda6, and 
> I'm somewhat amazed that UML managed to boot it (most distros run hwclock or 
> something similar, which runs some privileged instructions, which UML doesn't 
> handle yet).

I'm running on a RedHat 6.2 filesystem, that I just copied from one of my
machines. I did nothing special to get it to boot. So it seems I'm lucky
too. :)

-- 
Best regards
    Christian Laursen

phi...@in... said:
> This is the command I used to start uml:
>   ./linux ubd0=/dev/hda6
> and it doesn't resemble the documentation too much... 

No it doesn't.  I guess there's a bootable filesystem on your /dev/hda6, and 
I'm somewhat amazed that UML managed to boot it (most distros run hwclock or 
something similar, which runs some privileged instructions, which UML doesn't 
handle yet).

Did you grab a bootable filesystem from my site?  Uncompress it, rename it to 
root_fs and run "linux", and it will boot.  Don't rename it, and run "linux 
ubd0=root_fs_whatever", and it will boot.  Or "linux ubd7=root_fs_whatever 
root=ubd7".

It's pretty straightforward.

Also, there are descriptions of all the command line options (the UML-specific 
ones) on my site.

				Jeff

Hi, I just managed to get uml up - after quite a lot of random trying. 
Is there some documentation on ubdx ?

This is the command I used to start uml:

  ./linux ubd0=/dev/hda6

and it doesn't resemble the documentation too much...

-- 
Daniel

Jeff Dike <jd...@ka...> writes:

> xi...@bo... said:
> > The debugger fires up allright, and I can break it with ^C.
> > However, after I hang the kernel, I can no longer break it. 
> 
> Try attaching to the tracing thread to see if it's in some kind of infinite 
> loop.
> 
> That's the only thing I can think of offhand which would cause gdb to hang.

It looks like it's stuck in __wait4() for some reason.

This is the backtrace:

Program received signal SIGINT, Interrupt.
0x100cb8c9 in __wait4 ()
(gdb) bt
#0  0x100cb8c9 in __wait4 ()
#1  0xbffff8fc in ?? ()
#2  0x100be3ad in signals (init_proc=0x100be858 <start_kernel_proc>, sp=0x10147ffc) at trap_user.c:100
#3  0x100bed48 in linux_main (argc=7, argv=0xbffffa04) at um_arch.c:191
#4  0x10001277 in main (argc=7, argv=0xbffffa04, envp=0xbffffa24) at /usr/src/linux-2.4.0-test10uml/arch/um/main.c:66
(gdb) 

-- 
Best regards
    Christian Laursen

xi...@bo... said:
> The debugger fires up allright, and I can break it with ^C.
> However, after I hang the kernel, I can no longer break it. 

Try attaching to the tracing thread to see if it's in some kind of infinite 
loop.

That's the only thing I can think of offhand which would cause gdb to hang.

				Jeff

Christian Laursen <xi...@bo...> writes:

> Jeff Dike <jd...@ka...> writes:
> 
> > Run it with 'debug' on the command line.  This will fire up the kernel 
> > debugger.  When it hangs, ^C the debugger and get a backtrace.  See 
> > http://user-mode-linux.sourceforge.net/debugging.html for details on using the 
> > debugger if necessary.
> 
> Thank you, I'll try that.

Well, things seem to be a little complicated here.

The debugger fires up allright, and I can break it with ^C.

However, after I hang the kernel, I can no longer break it.

It dies on a SIGTERM, but that's about it.

Any suggestions?

-- 
Best regards
    Christian Laursen

Jeff Dike <jd...@ka...> writes:

> xi...@bo... said:
> > I succesfully applied the reiserfs patch to uml and it just works. The
> > only problem is, that it hangs on umount.
> 
> That's cool.  That's something I've been meaning to do, but haven't got around 
> to.  Did you need to install any usermode stuff, like a new mount or anything?

No, not at all.

You need the reiserfs tools of course to build the filesystem, but I had already
done that on the host.

> Run it with 'debug' on the command line.  This will fire up the kernel 
> debugger.  When it hangs, ^C the debugger and get a backtrace.  See 
> http://user-mode-linux.sourceforge.net/debugging.html for details on using the 
> debugger if necessary.

Thank you, I'll try that.

-- 
Best regards
    Christian Laursen

yd...@al... said:
> > serial line 0 assigned pty /dev/ptyp0
> He he, that surely is the problem, I don't have such a line on a
> test11 kernel, although I do have it on test10 :} 

You mean that the host is test11 and it doesn't have /dev/ptyp*, but your 
test10 host does?

> Ah, my first real bug on UML ?   

And what's the alleged bug?

The option for turning on the virtual serial line is CONFIG_SSL, which is in 
the 'General Setup' menu.

				Jeff

xi...@bo... said:
> I succesfully applied the reiserfs patch to uml and it just works. The
> only problem is, that it hangs on umount.

That's cool.  That's something I've been meaning to do, but haven't got around 
to.  Did you need to install any usermode stuff, like a new mount or anything?

Run it with 'debug' on the command line.  This will fire up the kernel 
debugger.  When it hangs, ^C the debugger and get a backtrace.  See 
http://user-mode-linux.sourceforge.net/debugging.html for details on using the 
debugger if necessary.

> Furthermore I've configured uml to be present on my LAN. I run the
> daemon like this 'um_eth_net_util eth0 100', but I can only reach my
> gateway from inside uml. The host machine and the other machines on
> the LAN are not reachable. 

/me summons wstearns

				Jeff

Hi guys!

I've been playing around with uml the last couple of days,
and I have to say, I'm very impressed.

It is indeed easy to get it up and running, so I decided
to do a little experimentation.

I succesfully applied the reiserfs patch to uml and it just
works. The only problem is, that it hangs on umount.

Have anybody else tried this combination and found a solution.
If not I guess I'll have to play a bit with gdb. (I'm an
experienced C programmer, so I guess I can figure that out.)


Furthermore I've configured uml to be present on my LAN.
I run the daemon like this 'um_eth_net_util eth0 100', but
I can only reach my gateway from inside uml. The host machine
and the other machines on the LAN are not reachable.

Do I have to do anything special to get this working?

I really like the idea about having the uml's running
in the same ip-range as the other machines, so that I don't
have to set up special routing.

Thanks in advance.

-- 
Best regards
    Christian Laursen

Good morning, Yann,

On Thu, 16 Nov 2000, Yann Dirson wrote:

> As for routing, I must say I had problems using the tap device from
> 2.2.18pre21.  Here's what I get:
>
> # net-tools/um_eth_net_util tap0 100
> /dev/tap0: No such device
> PHY: tap0
> phy close
>
> The module isn't autoloaded, and an explicit load attempt gives:
>
> # modprobe ethertap
> /lib/modules/2.2.18pre21vm7/net/ethertap.o: init_module: Device or resource busy
> Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters
> /lib/modules/2.2.18pre21vm7/net/ethertap.o: insmod /lib/modules/2.2.18pre21vm7/net/ethertap.o failed
> /lib/modules/2.2.18pre21vm7/net/ethertap.o: insmod ethertap failed
>
> ... which I kinda find strange - which damn resource does it try to get ?!

	Looking at the code in
/usr/src/linux-2.2.18/drivers/net/ethertap.c, there's only one thing that
can cause an -EBUSY to be returned on insertion:

int init_module(void)
{
        dev_ethertap.base_addr=unit+NETLINK_TAPBASE;
        sprintf(devicename,"tap%d",unit);
        if (dev_get(devicename))
        {
                printk(KERN_INFO "%s already loaded.\n", devicename);
                return -EBUSY;
        }
        if (register_netdev(&dev_ethertap) != 0)
                return -EIO;
        return 0;
}

	It seems the module is already loaded.  Is it?  Is it possibly
compiled straight into the kernel?

> > the case and you want to have the uml host act as a bridge.  You have two
> > choices.
> > 	The first is to do real bridging in the host kernel.  From
> > Configure.help:
>
> Hm... I thought net_util would do that itself - though I had doubts
> after my attempts.

	net_util has the sole job of getting packets back and forth
between the uml kernel and the host kernel.  It has no responsibilities
for routing, packet forwarding, packet filtering, or proxarp that may also
be needed depending on your requirements.
	Think of um_eth_net_util as the virtual ethernet card on the host
kernel side of the internal virtual ethernet network.

> Anyway, these should be documented on the networking page, where only
> the net_util command-line is given.

	If you look back in he archives of this list, I posted a recipe
for Ethernet networking a few weeks back.

> > 	The second choice is to perform proxyarp.  The host will need to
> > stand in for all machines on both sides, acting as a standin for all arp
> > requests.  You'll have to do the arp commands by hand.
> > 	I've attached a writeup I did about proxyarp.  I hope it'll
> > explain the process enough that you can give it a try yourself.
>
> Thanks much for this, it is indeed an interesting text.  However I
> don't think it can be applied to my case.
>
> I'll have a look at the kernel bridging stuff, then.

	Best of luck - I hope you find what you need.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	Q: Will the tcp/ethernet SMP scaling changes be back-ported to 2.2.x?
	Mingo: yes, all SMP changes in 2.3 will be backported to 2.2 in the
next few months, but to not confuse it with 2.2 it will be named '2.4' ;)
	-- Ingo Molnar <mi...@ch...>
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts,
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

On Thu, Nov 16, 2000 at 12:07:30AM +0100, Yann Dirson wrote:
> The module isn't autoloaded, and an explicit load attempt gives:

Hm, I was wrong.  Module "tap0" is autoloaded.  Thanks to diald.deb
which added to my modules.conf:

alias tap0      ethertap
options tap0    -o tap0 unit=0
alias tap1      ethertap
options tap1    -o tap1 unit=1

Looks like these lines are correct...


> # modprobe ethertap
> /lib/modules/2.2.18pre21vm7/net/ethertap.o: init_module: Device or resource busy
> Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters
> /lib/modules/2.2.18pre21vm7/net/ethertap.o: insmod /lib/modules/2.2.18pre21vm7/net/ethertap.o failed
> /lib/modules/2.2.18pre21vm7/net/ethertap.o: insmod ethertap failed

Removing the tap0 module allows ethertap to be loaded.  However they do not have the same size:

tap1                    2288   0  (unused)
ethertap                2304   0  (unused)

Looks like I still have to learn about tap devices... and that once
again the networking docs for UML lack some key details :(

Regards,
-- 
Yann Dirson    <yd...@al...> |    Why make M$-Bill richer & richer ?
debian-email:   <di...@de...> |   Support Debian GNU/Linux:
                                    | Cheaper, more Powerful, more Stable !
http://ydirson.free.fr/             | Check <http://www.debian.org/>

On Mon, Nov 13, 2000 at 09:27:15PM -0500, William Stearns wrote:
> 	The serial line doesn't actually provide a usable serial port.

Ah yes, I started to lack some sleep when I did those tests :(


> As the uml boots, look for a line like the following:
> 
> serial line 0 assigned pty /dev/ptyp0

He he, that surely is the problem, I don't have such a line on a
test11 kernel, although I do have it on test10 :}

Ah, my first real bug on UML ?  
Jeff, do you confirm ?  Foreseeing virtual beer to be openned :)

And I don't see an option that would turn this on.

Best regards,
-- 
Yann Dirson    <yd...@al...> |    Why make M$-Bill richer & richer ?
debian-email:   <di...@de...> |   Support Debian GNU/Linux:
                                    | Cheaper, more Powerful, more Stable !
http://ydirson.free.fr/             | Check <http://www.debian.org/>

Hi Bill,

On Mon, Nov 13, 2000 at 09:18:55PM -0500, William Stearns wrote:
> 	I'd be interested to know why you want to bridge packets between
> the real and the virtual ethernet lans - is there a reason why setting up
> a second IP network would be less preferable?

Yes.  I want in the end a UML virtual host to act as a firewall
between the local LAN (part which could be done by routing) and the
internet comming on another ether, which obviously I do not want to
connect to anything else.

For now the config I described is just used for testing.


As for routing, I must say I had problems using the tap device from
2.2.18pre21.  Here's what I get:

# net-tools/um_eth_net_util tap0 100
/dev/tap0: No such device
PHY: tap0
phy close


The module isn't autoloaded, and an explicit load attempt gives:

# modprobe ethertap
/lib/modules/2.2.18pre21vm7/net/ethertap.o: init_module: Device or resource busy
Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters
/lib/modules/2.2.18pre21vm7/net/ethertap.o: insmod /lib/modules/2.2.18pre21vm7/net/ethertap.o failed
/lib/modules/2.2.18pre21vm7/net/ethertap.o: insmod ethertap failed

... which I kinda find strange - which damn resource does it try to get ?!


> the case and you want to have the uml host act as a bridge.  You have two
> choices.
> 	The first is to do real bridging in the host kernel.  From
> Configure.help:

Hm... I thought net_util would do that itself - though I had doubts
after my attempts.

Anyway, these should be documented on the networking page, where only
the net_util command-line is given.


> 	The second choice is to perform proxyarp.  The host will need to
> stand in for all machines on both sides, acting as a standin for all arp
> requests.  You'll have to do the arp commands by hand.
> 	I've attached a writeup I did about proxyarp.  I hope it'll
> explain the process enough that you can give it a try yourself.

Thanks much for this, it is indeed an interesting text.  However I
don't think it can be applied to my case.

I'll have a look at the kernel bridging stuff, then.

Best regards, and thanks for the info
-- 
Yann Dirson    <yd...@al...> |    Why make M$-Bill richer & richer ?
debian-email:   <di...@de...> |   Support Debian GNU/Linux:
                                    | Cheaper, more Powerful, more Stable !
http://ydirson.free.fr/             | Check <http://www.debian.org/>

Matt Clay wrote:

> However, when you run fdisk you do need to set the cylinders from the
> advanced menu (seems that the cylinders should be half the size of the
> disk image in megabytes: i.e. 500 mb partition should have 250
> cylinders).  Then you can create your primary partition and run setup.

Ok, apparently this isn't the case.  As long as I don't run mke2fs on
the filesystem before running UML, I can simply create my partition and
run setup, doing the format from setup.

-- 

 - Matt Clay
 - Cowboyz.com
 - (503) 241-1990

Jeff Dike wrote:

> ma...@co... said:
> > "mke2fs root_fs_slackware"
> > then later...
> > "Run fdisk on /dev/ubd1 to put a partition table on the new
> > filesystem"
> 
> > It seems strange to me that you run fdisk on an image that has been
> > formatted with mke2fs.
> 
> Jeez, people sure don't like my writing today.  You're both right.  These
> will be fixed.

I tried a few more installations and found that for installing Slackware
7.1 using the Slackware kit, you do not need to format the filesystem
after you create it with "dd" (you do it during setup instead).

However, when you run fdisk you do need to set the cylinders from the
advanced menu (seems that the cylinders should be half the size of the
disk image in megabytes: i.e. 500 mb partition should have 250
cylinders).  Then you can create your primary partition and run setup.

On a side note, setting the time zone doesn't seem to work during the
install.  You need run "timeconfig" after you've booted up UML for the
first time after the install.

-- 

 - Matt Clay
 - Cowboyz.com
 - (503) 241-1990

Good evening, Yann,

On Tue, 14 Nov 2000, Yann Dirson wrote:

> The provided debian root_fs has an inittab entry for serial getty.
> However, can't find a "serial" config option in the config tree - the
> whole "character devices" menu doesn't show up for me.
>
> Did I overlooked something once again ?

	I think so, but don't worry about it.  I do it all the time with
UML, to the point that Jeff's procmail scripts have an entire killfile
section devoted to me. ;-)
	The serial line doesn't actually provide a usable serial port.
Instead, it connects to a pty on the host system, allowing you to log into
your uml right from the host without using the console or xterms.  As the
uml boots, look for a line like the following:

serial line 0 assigned pty /dev/ptyp0

	Once a getty has been started in the uml, you can use minicom on
the host to talk to /dev/ptyp0.  I think something like

minicom -p /dev/ptyp0

	should let you log into the uml from the host.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	Failure is not an option. It comes bundled with your Microsoft
product.
	-- Ferenc Mantfeld
(Courtesy of Jim Thompson <ji...@wa...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts,
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

Good evening, Yann,
	I'd be interested to know why you want to bridge packets between
the real and the virtual ethernet lans - is there a reason why setting up
a second IP network would be less preferable?  For example, could you
leave your real ethernet at 10.x.x.x and put the umls on 192.168.0.x/24?
Taking that approach would allow you to simply route between the two
networks.
	I could see that perhaps you don't wish to make any routing
changes to the machines on the real ethernet network - lets assume that's
the case and you want to have the uml host act as a bridge.  You have two
choices.
	The first is to do real bridging in the host kernel.  From
Configure.help:

802.1d Ethernet Bridging
CONFIG_BRIDGE
  If you say Y here, then your Linux box will be able to act as an
  Ethernet bridge, which means that the different Ethernet segments it
  is connected to will appear as one Ethernet to the participants.
  Several such bridges can work together to create even larger
  networks of Ethernets using the IEEE 802.1 spanning tree algorithm.
  As this is a standard, Linux bridges will cooperate properly with
  other third party bridge products.
  In order to use the ethernet bridge, you'll need the bridge
  configuration tools; see Documentation/networking/bridge.txt for
  location. Please read the Bridge mini-HOWTO for more information.

	The second choice is to perform proxyarp.  The host will need to
stand in for all machines on both sides, acting as a standin for all arp
requests.  You'll have to do the arp commands by hand.
	I've attached a writeup I did about proxyarp.  I hope it'll
explain the process enough that you can give it a try yourself.
	I'm pretty sure that routing between two different IP networks,
real ethernet bridging, or proxyarping will solve your problem;  it's just
a matter of what your priorities are.  Best of luck.
	Cheers,
	- Bill

On Tue, 14 Nov 2000, Yann Dirson wrote:

> What I'm trying to set up is bridging the eth0 virtual network
> interface into the local eth1 network.  Probably I'm missing
> something.
>
> The host which will run the UML machines:
>
> eth1      Link encap:Ethernet  HWaddr 00:50:FC:21:79:99
>           inet addr:10.0.0.1  Bcast:10.0.0.255  Mask:255.255.255.0
>           UP BROADCAST RUNNING PROMISC MULTICAST  MTU:1500  Metric:1
>           RX packets:52449 errors:0 dropped:0 overruns:0 frame:0
>           TX packets:54743 errors:0 dropped:0 overruns:0 carrier:0
>           collisions:520 txqueuelen:100
>           Interrupt:9 Base address:0x6c00
>
> The UML machines have same mask and bcast addresses.
>
> The other real host is there:
>
> root@bylbo:~# ping 10.0.0.2
> PING 10.0.0.2 (10.0.0.2): 56 data bytes
> 64 bytes from 10.0.0.2: icmp_seq=0 ttl=255 time=0.6 ms
> 64 bytes from 10.0.0.2: icmp_seq=1 ttl=255 time=0.4 ms
> 64 bytes from 10.0.0.2: icmp_seq=2 ttl=255 time=0.4 ms
>
> - when I fire up 2 UMLs, configure them as described in the
> networking.html tutorial (everything running as an unpriviledged
> user), they can ping each other.
>
> - when I run the net_util as root, the UML hosts do manage to make
> their eth0 come up (without it running they correctly fail to), but
> then pings don't get through.  If then I "ifconfig eth0 down" one of
> them, then up, then down again, the daemon exits with SIGPIPE (hm,
> indeed it sometimes stays up):
>
> select(6, [3 4 5], NULL, NULL, NULL)    = 3 (in [3 4 5])
> write(2, "connect\n", 8connect
> )                = 8
> accept(3, {sin_family=AF_INET, sin_port=htons(1581), sin_addr=inet_addr("127.0.0.1")}}, [16]) = 6
> read(4, "\0`\214\352\205k\0P\374!y\231\10\0E\0\5\334\375\241 \271"..., 1544) = 1514
> write(2, "04 100 ->", 904 100 ->)                = 9
> write(2, " 00", 3 00)                      = 3
> write(2, " 60", 3 60)                      = 3
> write(2, " 8c", 3 8c)                      = 3
> write(2, " ea", 3 ea)                      = 3
> write(2, " 85", 3 85)                      = 3
> write(2, " 6b", 3 6b)                      = 3
> write(2, " 00", 3 00)                      = 3
> write(2, " 50", 3 50)                      = 3
> write(2, " fc", 3 fc)                      = 3
> write(2, " 21", 3 21)                      = 3
> write(2, " 79", 3 79)                      = 3
> write(2, " 99", 3 99)                      = 3
> write(2, " 08", 3 08)                      = 3
> write(2, " 00", 3 00)                      = 3
> write(2, " 45", 3 45)                      = 3
> write(2, " 00", 3 00)                      = 3
> write(2, " ->", 3 ->)                      = 3
> write(2, " 05", 3 05)                      = 3
> write(5, "\0\0\0\1\0\0\5\352\0`\214\352\205k\0P\374!y\231\10\0E\0"..., 1522) = -1 EPIPE (Broken pipe)
> --- SIGPIPE (Broken pipe) ---
> +++ killed by SIGPIPE +++
>
> ...and the involved UML process enters a loop saying "failed to
> forward 14 to process <pid>, errno = 9", which I can only stop with
> Ctrl-C, which inteerupts the UML kernel itself - hence fsck on reboot.
>
> If I try pings from one real host to a virtual one or vice versa,
> "ping" does not get answers.  However, tcpdump on the 10.0.0.1 host
> reports some ICMP "echo request" coming from the UML, but mostly
> arp messages like:
>
> 02:23:22.372252 arp who-has 10.0.0.1 tell 10.0.0.10
>
> When I run "ping 10.0.0.10" from the host:
>
> 02:47:25.345946 10.0.0.10 > 10.0.0.1: icmp: echo request (DF)
>
> And from the remote real host to the UML:
>
> 02:24:32.314019 arp who-has 10.0.0.10 tell 10.0.0.2

	(The above really look like the standard types of arp problems one
would see before implementing proxyarp or bridging.)

---------------------------------------------------------------------------
	I'm not tense, just terribly, terribly alert.
(Courtesy of "Michael J. Dark" <dar...@ze...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts,
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

What I'm trying to set up is bridging the eth0 virtual network
interface into the local eth1 network.  Probably I'm missing
something.

The host which will run the UML machines:

eth1      Link encap:Ethernet  HWaddr 00:50:FC:21:79:99
          inet addr:10.0.0.1  Bcast:10.0.0.255  Mask:255.255.255.0
          UP BROADCAST RUNNING PROMISC MULTICAST  MTU:1500  Metric:1
          RX packets:52449 errors:0 dropped:0 overruns:0 frame:0
          TX packets:54743 errors:0 dropped:0 overruns:0 carrier:0
          collisions:520 txqueuelen:100
          Interrupt:9 Base address:0x6c00

The UML machines have same mask and bcast addresses.

The other real host is there:

root@bylbo:~# ping 10.0.0.2
PING 10.0.0.2 (10.0.0.2): 56 data bytes
64 bytes from 10.0.0.2: icmp_seq=0 ttl=255 time=0.6 ms
64 bytes from 10.0.0.2: icmp_seq=1 ttl=255 time=0.4 ms
64 bytes from 10.0.0.2: icmp_seq=2 ttl=255 time=0.4 ms


- when I fire up 2 UMLs, configure them as described in the
networking.html tutorial (everything running as an unpriviledged
user), they can ping each other.

- when I run the net_util as root, the UML hosts do manage to make
their eth0 come up (without it running they correctly fail to), but
then pings don't get through.  If then I "ifconfig eth0 down" one of
them, then up, then down again, the daemon exits with SIGPIPE (hm,
indeed it sometimes stays up):

select(6, [3 4 5], NULL, NULL, NULL)    = 3 (in [3 4 5])
write(2, "connect\n", 8connect
)                = 8
accept(3, {sin_family=AF_INET, sin_port=htons(1581), sin_addr=inet_addr("127.0.0.1")}}, [16]) = 6
read(4, "\0`\214\352\205k\0P\374!y\231\10\0E\0\5\334\375\241 \271"..., 1544) = 1514
write(2, "04 100 ->", 904 100 ->)                = 9
write(2, " 00", 3 00)                      = 3
write(2, " 60", 3 60)                      = 3
write(2, " 8c", 3 8c)                      = 3
write(2, " ea", 3 ea)                      = 3
write(2, " 85", 3 85)                      = 3
write(2, " 6b", 3 6b)                      = 3
write(2, " 00", 3 00)                      = 3
write(2, " 50", 3 50)                      = 3
write(2, " fc", 3 fc)                      = 3
write(2, " 21", 3 21)                      = 3
write(2, " 79", 3 79)                      = 3
write(2, " 99", 3 99)                      = 3
write(2, " 08", 3 08)                      = 3
write(2, " 00", 3 00)                      = 3
write(2, " 45", 3 45)                      = 3
write(2, " 00", 3 00)                      = 3
write(2, " ->", 3 ->)                      = 3
write(2, " 05", 3 05)                      = 3
write(5, "\0\0\0\1\0\0\5\352\0`\214\352\205k\0P\374!y\231\10\0E\0"..., 1522) = -1 EPIPE (Broken pipe)
--- SIGPIPE (Broken pipe) ---
+++ killed by SIGPIPE +++

...and the involved UML process enters a loop saying "failed to
forward 14 to process <pid>, errno = 9", which I can only stop with
Ctrl-C, which inteerupts the UML kernel itself - hence fsck on reboot.



If I try pings from one real host to a virtual one or vice versa,
"ping" does not get answers.  However, tcpdump on the 10.0.0.1 host
reports some ICMP "echo request" coming from the UML, but mostly
arp messages like:

02:23:22.372252 arp who-has 10.0.0.1 tell 10.0.0.10


When I run "ping 10.0.0.10" from the host:

02:47:25.345946 10.0.0.10 > 10.0.0.1: icmp: echo request (DF)


And from the remote real host to the UML:

02:24:32.314019 arp who-has 10.0.0.10 tell 10.0.0.2


-- 
Yann Dirson    <yd...@al...> |    Why make M$-Bill richer & richer ?
debian-email:   <di...@de...> |   Support Debian GNU/Linux:
                                    | Cheaper, more Powerful, more Stable !
http://ydirson.free.fr/             | Check <http://www.debian.org/>

Hi,

The provided debian root_fs has an inittab entry for serial getty.
However, can't find a "serial" config option in the config tree - the
whole "character devices" menu doesn't show up for me.

Did I overlooked something once again ?

-- 
Yann Dirson    <yd...@al...> |    Why make M$-Bill richer & richer ?
debian-email:   <di...@de...> |   Support Debian GNU/Linux:
                                    | Cheaper, more Powerful, more Stable !
http://ydirson.free.fr/             | Check <http://www.debian.org/>

da...@th... said:
> Shouldn't it say this(?):
> "This option enables a debugging interface which allows gdb to debug
> the kernel without needing to actually attach to kernel threads." 
>            ^^^^^^^

ma...@co... said:
> "mke2fs root_fs_slackware"
> then later...
> "Run fdisk on /dev/ubd1 to put a partition table on the new
> filesystem"

> It seems strange to me that you run fdisk on an image that has been
> formatted with mke2fs.

Jeez, people sure don't like my writing today.  You're both right.  These will 
be fixed.

Maybe everyone is complaining about my writing because they can't find 
anything wrong with my code :-)

				Jeff

Looking over the install instructions for installing Slackware in UML, I
found this:

"mke2fs root_fs_slackware"

then later...

"Run fdisk on /dev/ubd1 to put a partition table on the new filesystem"

It seems strange to me that you run fdisk on an image that has been
formatted with mke2fs.  Is this step correct, or even necessary?  I
didn't seem to hurt anything.

Any insight on this is appreciated.

-- 

 - Matt Clay
 - Cowboyz.com
 - (503) 241-1990

al...@pr... said:
> Is there any way to tell gdb that the module is now part of the
> kernel? 

I recently figured this out and haven't put it on the web site yet.  What you 
do is insmod (or modprobe) the module.  Then, in gdb look at module_list (and 
go down ->next) until you find the entry for your new module.  It will 
probably be the first entry.

Find the module start by adding the address of that entry, 
sizeof(*module_list), and 8.  Those last two add up to 0x50 (IIRC), and the 
module entry will be on a page boundary, so the actual module start address 
will be 0xzzzzz060 in that case.

In gdb, 'add-symbol-file /path/to/module/on/host <module start address>'

Then everything will work normally.

When you unload it and load in another version, you have to
	symbol-file
which will unload everything, including the main binary, then
	symbol-file /path/to/kernel
	add-symbol-file as above

				Jeff