user-mode-linux-user Mailing List for User-mode Linux kernel port (Page 795)

This is probably easy to do, but I'm not (yet) a gdb guru.

Anyways, I typically start uml with the debug option so I get a nice
xterm w/gdb in it. I then load a module into the kernel after it all
boots up. But when I try to set a breakpoint in some function in that
module it tells me that the function is not defined.

Is there any way to tell gdb that the module is now part of the kernel?

.adam

-- 
[ Adam Lazur, NOW Monkey                          <al...@pr...> ]
[ Progeny Linux Systems                             http://progeny.com ]

smu...@ho... said:
> I was wondering - has anyone tried the Debian Hurd distribution in
> user mode  linux?  Could it even in theory be done? 

Do you mean running the Hurd inside UML?  That would require a user-mode port 
of the Hurd.

Do you mean booting UML off a Debian Hurd filesystem?  If Debian Hurd is 
Debian - the linux kernel + the Hurd kernel, then that's just booting Debian, 
which works fine.  If the Debian Hurd filesystem can't be booted on a Linux 
kernel, then UML can't boot it.

				Jeff

I was wondering - has anyone tried the Debian Hurd distribution in user mode 
linux?  Could it even in theory be done?
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.

Share information about yourself, create your own public profile at 
http://profiles.msn.com.

al...@pr... said:
> On my Debian box (net-tools 1.57) ifconfig doesn't like the -n option
> when trying to setup the ptp networking stuff. I pulled the -n arg out
> of arch/um/drivers/umn_user.c and all is working now. 
> I've attached a patch just in case somebody else has the same problem...

Patch applied.  Thanks.

				Jeff

I didn't see this in the archives... so

On my Debian box (net-tools 1.57) ifconfig doesn't like the -n option
when trying to setup the ptp networking stuff. I pulled the -n arg out
of arch/um/drivers/umn_user.c and all is working now.

What's the -n option to ifconfig do? I'm assuming I don't need it as
it's not part of my ifconfig and things seem to be working w/out it :)

I've attached a patch just in case somebody else has the same problem...

.adam

-- 
[ Adam Lazur, NOW Monkey                          <al...@pr...> ]
[ Progeny Linux Systems                             http://progeny.com ]

wst...@po... said:
> Jeff; any requests for improvement?  Could these get out to the web
> site or cvs at some point?

I haven't gone through it in any detail yet.

I'll put it out there someplace.  I haven't decided where yet.

				Jeff

After I generated the patch, I noticed some stuff I didn't want to release, so 
I edited the patch by hand, and of course, messed it up.

So, the new patch is available at http://sourceforge.net/project/filelist.php?g
roup_id=429 as usual.

				Jeff

Good evening, Robert, Jeff,

On Thu, 2 Nov 2000, Robert Lipe wrote:

> > >  I did the mkrootfs and fed it the mountpoint of my RH6.2 cdrom.
> > 
> > My site makes no claim about mkrootfs being able to deal with RH6.2.  That's 
> > no accident - it works for 6.0 and 6.1.  I'm surprised that it ran and 
> 
> From your first hint (thanx!) I was trying to spend some time with
> MAKEDEV and teaching it about things that it seemed like it should have
> known.   This tells me I should walk away from using a RH 6.2 media for
> root_fs, right?

	Not at all.  rh62 and 70 work just fine with the attached.  Others
are in testing.

> > Bill Stearns has hacked and wacked on mkrootfs to the point that it can 
> > produce filesystems from practically any RPM-based distro.  It's attached 
> > below.
> 
> Interesting.  It seems to be using max, addline, networkof, broadcastof,
> chage, and a number of other utilities that don't exist on my system.  

	chage is part of the shadow-utils package; you should install this
on your system.  The others are provided by the "functions" file, which is
attached along with a fresh copy of the mkrootfs program (a few small
changes from Jeff's recent post).
	"functions" simply needs to be in the current directory.

> So went back to the provided 'mkrootfs', slammed in a RH6.1 CD, and
> tried that.   That brought satisfaction. 

	Excellent!  I'm still including the new versions as I know that
others on this list use these tools and I wanted to get the new versions
out.  Jeff; any requests for improvement?  Could these get out to the web
site or cvs at some point?
	Cheers,
	- Bill

---------------------------------------------------------------------------
	If it happens once, it's a bug.  If it happens twice, it's 
a feature.  If it happens more than twice, it's a design philosophy.
(Courtesy of Slashdot)
--------------------------------------------------------------------------
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
--------------------------------------------------------------------------

> >  I did the mkrootfs and fed it the mountpoint of my RH6.2 cdrom.
> 
> My site makes no claim about mkrootfs being able to deal with RH6.2.  That's 
> no accident - it works for 6.0 and 6.1.  I'm surprised that it ran and 

From your first hint (thanx!) I was trying to spend some time with
MAKEDEV and teaching it about things that it seemed like it should have
known.   This tells me I should walk away from using a RH 6.2 media for
root_fs, right?

> Bill Stearns has hacked and wacked on mkrootfs to the point that it can 
> produce filesystems from practically any RPM-based distro.  It's attached 
> below.

Interesting.  It seems to be using max, addline, networkof, broadcastof,
chage, and a number of other utilities that don't exist on my system.  


So went back to the provided 'mkrootfs', slammed in a RH6.1 CD, and
tried that.   That brought satisfaction. 

Thanx much!
RJL

rob...@us... said:
>  I did the mkrootfs and fed it the mountpoint of my RH6.2 cdrom.

My site makes no claim about mkrootfs being able to deal with RH6.2.  That's 
no accident - it works for 6.0 and 6.1.  I'm surprised that it ran and 
pretended to give you a real filesystem.

Bill Stearns has hacked and wacked on mkrootfs to the point that it can 
produce filesystems from practically any RPM-based distro.  It's attached 
below.

				Jeff

> Warning: unable to open an initial console.

This is from the kernel not being able to open /dev/console in the filesystem 
you gave it.

> process 1 (sh) segfaulted at address 0x0, ip = 0x8066f5c

Why is process 1 a shell?  It should be init.  Looking at init/main.c show 
this:
	execve("/sbin/init",argv_init,envp_init);
	execve("/etc/init",argv_init,envp_init);
	execve("/bin/init",argv_init,envp_init);
	execve("/bin/sh",argv_init,envp_init);

It looks like it fell through the various places it could find init, and fell 
back on the shell.

So, have a good look at that filesystem and figure out why there's no console 
or init.  BTW, mkrootfs assumes devfs, so it doesn't create any device files, 
which might explain the console.  The missing init is tougher to explain.

And maybe the kernel is being too eager to panic here.  It probably shouldn't 
panic on repeated process segfaults.  The segfault should have been delivered 
and it should have killed the shell (and why it's segfaulting is another issue 
:-), which would kill the kernel in this case anyway.

				Jeff

Hello,

I'm sure this is pilot error, but I'm not finding the answer in the doc,
the FAQ, the list archives, or the HOWTO, so I'd appreciate any help
offered.

I just pulled rh-package-2.4.0-test10.tar.bz2 and installed it on my
RH6.2 system.  It's running stock redhat kernel-2.2.16-3 and examination
of the kernel shows the patches to ptrace and friends are present.  I
did the mkrootfs and fed it the mountpoint of my RH6.2 cdrom.  Intuition
tells me that I'm headed for trouble with new kernels and "old"
user-space tools, but I don't think I'm getting that far yet.   Since the
host system doesn't seem to have devfs, I start with the 'devfs=nomount'
flag.   Unfortunately, it crashes out of the gate:

$ ./linux devfs=nomount 
tracing thread pid = 16617
Linux version 2.4.0-test10-1um (jd...@cc...) (gcc version egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)) #1 Tue Oct 31 23:13:06 EST 2000
On node 0 totalpages: 4096
zone(0): 0 pages.
zone(1): 4096 pages.
zone(2): 0 pages.
Kernel command line: devfs=nomount root=/dev/ubd0
Calibrating delay loop... 526.91 BogoMIPS
Memory: 16100k available
Dentry-cache hash table entries: 2048 (order: 2, 16384 bytes)
Buffer-cache hash table entries: 1024 (order: 0, 4096 bytes)
Page-cache hash table entries: 4096 (order: 2, 16384 bytes)
Inode-cache hash table entries: 1024 (order: 1, 8192 bytes)
VFS: Diskquotas version dquot_6.4.0 initialized
POSIX conformance testing by UNIFIX
Linux NET4.0 for Linux 2.4
Based upon Swansea University Computer Society NET3.039
Starting kswapd v1.8
pty: 256 Unix98 ptys configured
RAMDISK driver initialized: 16 RAM disks of 4096K size 1024 blocksize
loop: enabling 8 loop devices
User-mode Linux network interface 0.005 (eth0)
User-mode Linux network interface 0.005 (eth1)
User-mode Linux network interface 0.005 (eth2)
User-mode Linux network interface 0.005 (eth3)
NET4: Linux TCP/IP 1.0 for NET4.0
IP Protocols: ICMP, UDP, TCP
IP: routing cache hash table of 512 buckets, 4Kbytes
TCP: Hash tables configured (established 1024 bind 1024)
NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
Initializing stdio console driver
Initializing software serial port version 0
serial line 0 assigned pty /dev/ptyp1
ssl receive thread is pid 16624
devfs: v0.102 (20000622) Richard Gooch (rg...@at...)
devfs: devfs_debug: 0x0
devfs: boot_options: 0x2
VFS: Mounted root (ext2 filesystem) readonly.
Warning: unable to open an initial console.
process 1 (sh) segfaulted at address 0x0, ip = 0x8066f5c
process 1 (sh) segfaulted at address 0x0, ip = 0x8066f5c
Kernel panic: Double fault on 0x0 - panicing because it wasn't fixed the first time




So there must be something I'm doing wrong.   Can anybody spot it?

This looks like a Very Cool way to do development! (At least once I can
make it work :-)

Thanx,
RJL

The user-mode port of 2.4.0-test10 is available.

The stack overflows seen in test9 are fixed.  The stack is now allocated as 
four pages, the top two used as a kernel stack, the third is inaccessible and 
acts as a guard page, and the lowest page contains the task structure.

Host devices can again be mounted inside the virtual machine.  This was broken 
a few releases ago when I made the block driver check io requests against the 
device size.

It will no longer crash if the main console is not a terminal.

I fixed a race which was causing strange kernel memory faults.

In the sources (the patch and cvs), but not the binaries, there is the 
beginning of a hostfs filesystem.  This gives you access to the host root 
filesystem.  Doing 'mount none /wherever -t hostfs' will mount the host root 
filesystem on /wherever.  Right now, you can mount it and cd into it, but ls 
will crash the kernel.

The project's home page is http://user-mode-linux.sourceforge.net

The project's download page is http://sourceforge.net/project/filelist.php?grou
p_id=429

				Jeff

wst...@po... said:
> On Mon, 30 Oct 2000, Rob Farber wrote:
> 	FYI: I have never been able to use a 2.4 kernel as the host kernel.
>
> Interesting.  Could you give a little more detail on what problems
> you get when you try to use a 2.4.x host kernel?

Yeah.  What he said.

I know of two problems with 2.4.0-testx kernels, both of which are fixed as of 
my current CVS.  One which is not my problem at all is the test8 ptrace "fix". 
 That totally broke UML.  test8 can't run UML without backing out that change. 
 Linux backed it out in test9, so it's no longer a problem.

The other is the increased signal frame size over 2.2.  It's about 500 bytes 
bigger, and it's not hard to get three or four of them on the stack.  When 
that happens, and half of a 4K page is occupied by signal frames, it's a 
little hard to fit everything else onto the remaining half page.

So, if you're seeing something else, you need to provide some details.

				Jeff

Good day, Rob,

On Mon, 30 Oct 2000, Rob Farber wrote:

> 	This is consistent across three machines running a redhat 6.2 base
> installation. The 2.2.17 kernel was built on each machine, a 500 Mhz Celeron
> Dell Optiplex with 128 MB RAM. The user-mode kernel, linux-2.4.0-test9, was
> built on each machine
> 
> 	FYI: I have never been able to use a 2.4 kernel as the host kernel.

	Interesting.  Could you give a little more detail on what problems
you get when you try to use a 2.4.x host kernel?  What kernels, what error
messages you see, if the uml enviroment hangs, where, what would normally
show up next in the uml boot process, can you boot it with
"init=/bin/bash" on the uml command line, which root filesystem or
distribution you use on the host and in uml...
	Thanks!
	Cheers,
	- Bill

---------------------------------------------------------------------------
	FLASH!  Intelligence of mankind decreasing!
	Details at...uh, when the little hand is on the....
(Courtesy of Joe <jg...@t-...>)
--------------------------------------------------------------------------
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 Mon, 30 Oct 2000, Rob Farber wrote:

> 	I have a simple test which consistently causes a "Kernel panic: Stack
> overflowed onto current_task page" on my machines. It breaks a working
> 2.4.0-test9 user-mode kernel. All I do is allocate 1024 integers on the
> stack in a kernel routine i.e. "int killer[1024];". 

Which is exactly your problem. Linux, on all of the various architectures
that it runs on, has a very small stack for memory efficiency reasons.
IIRC, on Linux for x86, the stack size is approximately six kilobytes
(eight kilobytes - sizeof(struct task_struct)). Here is a recent message
from Jeff Dike on the subject of user-mode Linux kernel stacks:

--- begin quote ---

Date: Mon, 16 Oct 2000 23:13:33 -0500
From: Jeff Dike <jd...@ka...>
To: use...@li...
Subject: CVS checkins

This batch does the following:

Allocates larger kernel stacks.  There is also a guard page between the two
pages of stack and the task structure.  Any overly long stacks will now
segfault at the place that they grew to big.

Redid a couple of procedures to reduce their stack frame size.

The kernel debugger no longer sees SIGSEGV.

                                Jeff

--- end quote ---

It nicely explains what you are seeing, huh?

davez

-- 
Dave Zarzycki
http://thor.sbay.org/~dave/

Hi,

	I have a simple test which consistently causes a "Kernel panic: Stack
overflowed onto current_task page" on my machines. It breaks a working
2.4.0-test9 user-mode kernel. All I do is allocate 1024 integers on the
stack in a kernel routine i.e. "int killer[1024];". The host kernel is
2.2.17.

	The test is to build ramfs into the user-domain kernel in the 2.4.0-test9
kernel.

Test it works:

	-boot the user-mode kernel which has ramfs
	- mount ramfs -t ramfs /mnt
	- mkdir /mnt/a
	the mkdir succeeds.
	- halt

Test it fails:
	- add the line: "int killer[1024];" to the routine ramfs_mkdir in
fs/ramfs/inode.c.
	- recompile "make linux. cp linux .."
	- boot the user-mode kernel
	- mount ramfs -t ramfs /mnt
	- mkdir /mnt/a
	Kernel panic: Stack overflowed onto current_task page
	The user-mode kernel is now dead.


	This is consistent across three machines running a redhat 6.2 base
installation. The 2.2.17 kernel was built on each machine, a 500 Mhz Celeron
Dell Optiplex with 128 MB RAM. The user-mode kernel, linux-2.4.0-test9, was
built on each machine

	FYI: I have never been able to use a 2.4 kernel as the host kernel.

	By the way, A very useful tool you have developed Jeff. :-)

	Many thanks!
	Rob

nd...@fr... said:
> ld -r --start-group  --end-group -o um_drivers.o 
> ld: no input files 

You unconfigured all the drivers?

If so, I think it's somewhat reasonable for something to blow up somewhere 
along the line because there's no way that a kernel with no drivers can be 
useful.

				Jeff

Hi, 

Has anyone seen this before?  

make[2]: Entering directory `/usr/src/linux-2.4.0-test9/lib'
make[2]: Nothing to be done for `all_targets'.
make[2]: Leaving directory `/usr/src/linux-2.4.0-test9/lib'
make[1]: Leaving directory `/usr/src/linux-2.4.0-test9/lib'
make CFLAGS="-D__KERNEL__ -I/usr/src/linux/include -Wall
-Wstrict-prototypes -O2 -fomit-frame-pointer   -U__i386__ -Ui386
-D__arch_um__ -DSUBARCH=\"i386\" -DNESTING=0 -fno-strict-aliasing " -C
/usr/src/linux/arch/um/drivers
make[1]: Entering directory `/usr/src/linux-2.4.0-test9/arch/um/drivers'
rm -f um_drivers.o
ld -r --start-group  --end-group -o um_drivers.o
ld: no input files
make[1]: *** [um_drivers.o] Error 1
make[1]: Leaving directory `/usr/src/linux-2.4.0-test9/arch/um/drivers'
make: *** [_dir_/usr/src/linux/arch/um/drivers] Error 2

Regards,

Nick Dronen

ma...@ex... said:
> Remounting root device with read-write enabled.
> The following two processes then eat all my resources: 

Put 'debug' on the command line, 'c' the gdb that comes up, ^C it when it 
hangs, and send in the stack trace.

And if you're feeling adventurous, you can step it until you see what the 
infinite loop is.

				Jeff

Hi,

When I try to compile my own 2.4.0-test9 kernel and run, I fail.  I use
root_fs_slackware_7.1_small as root filesystem and my host is Redhat 7
(kernel 2.2.16).  If I use the precompiled kernel from the
linux-2.4.0-test9.tar.gz archive everything works perfectly, but when I
compile a kernel my self it fails.  It prints the following and then
hangs.

zsh% ./run_linux 
tracing thread pid = 1346
Linux version 2.4.0-test9-1um (bla@bla) (gcc version 2.96
20000731 (Red Hat Linux 7.0)) #2 Fri Oct 27 16:22:02 CEST 2000
On node 0 totalpages: 4096
zone(0): 0 pages.
zone(1): 4096 pages.
zone(2): 0 pages.
Kernel command line: root=/dev/ubd0
Calibrating delay loop... 327.02 BogoMIPS
Memory: 16100k available
Dentry-cache hash table entries: 2048 (order: 2, 16384 bytes)
Buffer-cache hash table entries: 1024 (order: 0, 4096 bytes)
Page-cache hash table entries: 4096 (order: 2, 16384 bytes)
Inode-cache hash table entries: 1024 (order: 1, 8192 bytes)
POSIX conformance testing by UNIFIX
Linux NET4.0 for Linux 2.4
Based upon Swansea University Computer Society NET3.039
NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
NET4: Linux TCP/IP 1.0 for NET4.0
IP Protocols: ICMP, UDP, TCP, IGMP
IP: routing cache hash table of 512 buckets, 4Kbytes
TCP: Hash tables configured (established 1024 bind 1024)
Initializing RT netlink socket
Starting kswapd v1.8
pty: 256 Unix98 ptys configured
RAMDISK driver initialized: 16 RAM disks of 4096K size 1024 blocksize
loop: enabling 8 loop devices
early initialization of device teql0 is deferred
User-mode Linux network interface 0.005 (eth0)
User-mode Linux network interface 0.005 (eth1)
User-mode Linux network interface 0.005 (eth2)
User-mode Linux network interface 0.005 (eth3)
Initializing stdio console driver
Initializing software serial port version 0
serial line 0 assigned pty /dev/ptyp0
ssl receive thread is pid 1353
devfs: v0.102 (20000622) Richard Gooch (rg...@at...)
devfs: devfs_debug: 0x0
devfs: boot_options: 0x0
VFS: Mounted root (ext2 filesystem) readonly.
Mounted devfs on /dev
INIT: version 2.78 booting
/etc/rc.d/rc.S: Testing filesystem status: Read-only file system
Parallelizing fsck version 1.18 (11-Nov-1999)
/dev/ubd/0: clean, 7539/25688 files, 48494/102400 blocks
Remounting root device with read-write enabled.


The following two processes then eat all my resources:

zsh% ps 1362 1346
  PID TTY      STAT   TIME COMMAND
 1346 pts/0    S      0:30 ./linux [(tracing thread)]
 1362 pts/0    R      0:36 ./linux [/sbin/mount]                    


Thanks,
/Martin Jacobsson


---
.signature: No such file or directory

Good day, Jeff,

On Tue, 24 Oct 2000, Jeff Dike wrote:

> wst...@po... said:
> > 	I don't know of any reason why any of the UML networking approaches
> > would have trouble with masq.  The upper level networking code doesn't
> > care how the packets actually get in and out of the system. 
> 
> I just tried reaching a second box (my laptop) from a UML with the ethertap 
> interface, and failed.  The packets reached the host just fine, but went no 
> further as far as I could tell.
> 
> Doing the same with the slip interface worked just fine.
> 
> Both of those are local, so no masquerading is involved.
> 
> Do you have a recipe for setting this sort of thing up for the ethertap 
> interface?

	How about this; in one window, as root:

[root@sparrow /root]# ifconfig tap0 192.168.210.254 netmask 255.255.255.0 broadcast 192.168.210.255 arp up
[root@sparrow /root]# um_eth_net_util tap0 100
TAP: tap0

	Set up masquerade (2.4 host kernel):

[root@sparrow /root]# iptables -A POSTROUTING -t nat -s 192.168.210.0/24 -o wvlan0 -j MASQUERADE
[root@sparrow /root]# iptables -L POSTROUTING -t nat
Chain POSTROUTING (policy ACCEPT)
target     prot opt source               destination         
MASQUERADE  all  --  192.168.210.0/24     anywhere           
[root@sparrow /root]# iptables -L POSTROUTING -t nat -n -x -v 
Chain POSTROUTING (policy ACCEPT 5172 packets, 316377 bytes)
    pkts      bytes target     prot opt in     out     source          destination
       0        0 MASQUERADE  all  --  *      wvlan0  192.168.210.0/24 0.0.0.0/0
[root@sparrow /root]# 

	If you're using a 2.2.x host kernel, try something like:

/sbin/ipchains -A forward -s 192.168.210.0/24 -j MASQ

	Or for 2.0.x ipfwadm host kernel (hmm, just how would you be
running uml on this ;-):

/sbin/ipfwadm -F -a accept -m -S 192.168.210.0/24

	In UML:
bash# ifconfig eth0
eth0      Link encap:Ethernet  HWaddr FE:FD:C0:A8:D2:8D  
          inet addr:192.168.210.141  Bcast:192.168.210.255 Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:3 errors:0 dropped:0 overruns:0 frame:0
          TX packets:3 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:100 
          Interrupt:4 

bash# route -n
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
192.168.210.254 0.0.0.0         255.255.255.255 UH    0      0        0 eth0
192.168.210.0   0.0.0.0         255.255.255.0   U     0      0        0 eth0
127.0.0.0       0.0.0.0         255.0.0.0       U     0      0        0 lo
0.0.0.0         192.168.210.254 0.0.0.0         UG    0      0        0 eth0
0.0.0.0         192.168.210.254 0.0.0.0         UG    1      0        0 eth0

	The above were created by entering the following lines:
bash# less /etc/sysconfig/network-scripts/ifcfg-eth0 
DEVICE=eth0
ONBOOT=YES
IPADDR=192.168.210.141
GATEWAY=192.168.210.254
NETMASK=255.255.255.0
NETWORK=192.168.210.0
BROADCAST=192.168.210.255


	In UML:
bash# ping home.netscape.com
PING www-mv.netscape.com (205.188.247.65) from 192.168.210.141 : 56(84) bytes of data.
64 bytes from 205.188.247.65: icmp_seq=0 ttl=48 time=266.1 ms
64 bytes from 205.188.247.65: icmp_seq=1 ttl=48 time=8266.0 ms
64 bytes from 205.188.247.65: icmp_seq=2 ttl=48 time=7265.6 ms
64 bytes from 205.188.247.65: icmp_seq=3 ttl=48 time=6273.3 ms
64 bytes from 205.188.247.65: icmp_seq=4 ttl=48 time=5276.0 ms
64 bytes from 205.188.247.65: icmp_seq=5 ttl=48 time=4278.6 ms
64 bytes from 205.188.247.65: icmp_seq=6 ttl=48 time=3296.0 ms
64 bytes from 205.188.247.65: icmp_seq=7 ttl=48 time=2298.8 ms
64 bytes from 205.188.247.65: icmp_seq=8 ttl=48 time=1309.6 ms
64 bytes from 205.188.247.65: icmp_seq=9 ttl=48 time=312.2 ms

--- www-mv.netscape.com ping statistics ---
11 packets transmitted, 10 packets received, 9% packet loss
round-trip min/avg/max = 266.1/3884.2/8266.0 ms

	Other client apps seem to work just fine as well.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"Microsoft has done more for the fault tolerance industry than any
other company.  They have made end-users very tolerant of faults". 
(Courtesy of "Deliduka, Bennet" <ben...@st...>)
--------------------------------------------------------------------------
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
--------------------------------------------------------------------------

wst...@po... said:
> 	I don't know of any reason why any of the UML networking approaches
> would have trouble with masq.  The upper level networking code doesn't
> care how the packets actually get in and out of the system. 

I just tried reaching a second box (my laptop) from a UML with the ethertap 
interface, and failed.  The packets reached the host just fine, but went no 
further as far as I could tell.

Doing the same with the slip interface worked just fine.

Both of those are local, so no masquerading is involved.

Do you have a recipe for setting this sort of thing up for the ethertap 
interface?

				Jeff

ri...@co... said:
> How will ethertap function in combination with IP masquerading?

Haven't tried it.

> I suspect a lot of UML users will have only ONE ipv4 address for their
> machine and will be using private addresses for the UMLs... 

One of them being me - I've never tried masquerading a virtual machine through 
the ethertap interface, though.

> > > It seams to me that this would be more desirable. 
> > Why?
> Better IP address setup stuff? 

I have no idea.  There was a debate on this topic a while ago, and the people 
who seemed to know what they were talking about thought that ethertap was the 
way to go.

Unfortunately, the official list archiver seems not have remembered that 
thread.

Gerald Britton was pretty emphatic that ethertap was superior to slip, and 
there was a comment by Alan Cox on lkml a while ago to the same effect.

				Jeff

Hi Rik, Jeff,

On Tue, 24 Oct 2000, Rik van Riel wrote:

> On Thu, 19 Oct 2000, Jeff Dike wrote:
> > hea...@em... said:
> > > Has anyone tried doing networking with ppp over a pseudo-tty instead
> > > of using slip?  
> > 
> > Not that I know of.  I chose slip over ppp when I first did the
> > umn driver and I can't remember why.
> > 
> > Anyway, ethertap-type interface are preferable to slip/ppp.
> 
> How will ethertap function in combination with IP masquerading?

	I don't know of any reason why any of the UML networking
approaches would have trouble with masq.  The upper level networking code
doesn't care how the packets actually get in and out of the system.

> I suspect a lot of UML users will have only ONE ipv4 address
> for their machine and will be using private addresses for the
> UMLs...
> 
> > > It seams to me that this would be more desirable. 
> > 
> > Why?
> 
> Better IP address setup stuff?

	Perhaps, and the ability to tunnel other protocols than just IP.  
In all honesty, though, I don't see any reason to continue discussing the
ppp/slip/point-to-point path; the ethernet networking code is working
marvelously well.  Can you see an advantage to the slip approach that I've
missed?  The second cup of coffee hasn't kicked in yet and perhaps I'm
being particularly dense today. ;-)
	Cheers,
	- Bill

---------------------------------------------------------------------------
	An imperfect plan executed violently is far superior to a perfect
plan.
	-- George Patton
(Courtesy of Thomas Roell <ro...@xi...>)
--------------------------------------------------------------------------
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
--------------------------------------------------------------------------