user-mode-linux-user

Re: [User-mode-linux-user] security ??

[Jeff D. <jd...@ka...>]

> If you decide _not_ to finish the migration to jail, I'd sincerely
> love to hear about where those holes in the razor wire even _might_
> exist; it'll make the article more useful.

It's definitely on my list.  I just haven't done anything about it yet.

Big hole #1:
	The tracing thread's stack is in kernel physical memory.  All a hostile  
process needs to do is zero it out or something, and the tracing thread 
crashes.  Since it was ptracing the hostile process and intercepting its 
system calls, once it's gone, the hostile process is no longer under control.  
It now has access to the host system (as whatever user ran the kernel, not as 
root).

Big hole #2:
	When a process is running process code, its system calls are intercepted.  
When it is running a system call or trap handler in the kernel, they are not.  
There is a field in the thread structure which keeps track of this.  A hostile 
process could locate its kernel stack (at the bottom of which is its task 
structure, which contains the thread structure), and flip that field to fake 
the tracing thread into believing that its system calls are not to be 
intercepted.  It now has access to the host system.

All of the other attacks that I can think of which involve corrupting kernel 
data also involve crashing the kernel without letting the attacking process 
out of jail.  That's not good, but it's not terrible, especially if the jail 
is just for that one process.

				Jeff

Re: [User-mode-linux-user] security ??

[William S. <wst...@po...>]

On Tue, 14 Mar 2000, Jeff Dike wrote:

> It is possible right now for a hostile process to break out of the user-mode 
> virtual machine and gain access to the hosting machine.
> 
> However, this is fairly easy to fix.  It would not be hard to turn the 
> virtual machine into a secure jail.
> 
> > It's essentially a software sandbox; a safe place to play with
> > untrusted apps.  I plan to write an article for LinuxMonth expanding
> > on that idea and showing how to do it.

	Interesting; I hadn't realized that.
	While there might be situations where it would be fun to migrate
an app back and forth between uml and host (I can't think of any, but that
doesn't mean there aren't any), I would personally have a preference that
uml be a self contained locked down environment.  I see some real use as a
sandbox environment.
	If you decide _not_ to finish the migration to jail, I'd sincerely
love to hear about where those holes in the razor wire even _might_
exist; it'll make the article more useful.

> It is good for that, since a hostile program would have to specifically attack 
> the kernel.  I don't think many such attacks have been implemented since uml 
> is still fairly new...

	I suppose the obvious points of attack would include any place
where ./linux trusts information provided inside uml and acts on that
information in the host OS.  Unfortunately, it means that ./linux probably
_shouldn't_ allow a user in uml to assign the IP address to be used in the
host OS; one might have a certain amount of fun by assigning 127.0.0.1 to
umn. :-)

previous post, Bill:
> The ethertap could be brought up once during host system boot and left
> until uml-linux is started.  Just for example:  Boot scripts bring up
> tap0 w/ 192.168.0.254, ptp -> 192.168.0.253, and add a route to .253.
>  /dev/tap0 is given mode 660, owned by root/uml-group.

previous post, Jeff:
	If you're doing serious network stuff, you'd want a bunch of them,
and I'd be somewhat concerned about potentially locking out other ethertap
users. Also, the fact that both ends are configured a boot time before a
uml runs means that the uml net setup scripts need to change to just
accept whatever IP address the uml's /dev/tap has.  It also means that the
IP address needs to be read out of the network driver somehow and fed back
into ifconfig so that the network layer knows its own IP address.

	The administrator of the system would be responsible for
allocating as many as would be needed ahead of time and providing routing
table entries to the uml's on the other ends of the taps.  In fact, each
uml can be told which tap to connect to, which would be the tap that has
the right routing table for that uml.  I still think that passing:
./linux tap0=192.168.0.253,/dev/tap12
	is elegant enough that the uml can know what to open in the host
OS and what IP address to assign to it's own tap.
	If this turns out to be a useful approach, I'd be glad to help
with the documentation for the host OS side of it too.

	Have I missed something in the uml environment?  You seem to be
referring to telling umn/tap12 in uml what IP address it has.  Does it
really need to know at that low a level?  Couldn't umn/tap12 simply act
promiscuous (or the pointopoint equivalent) and pass everything up to the
network layer so that the network layer could decide what needs to be
handled locally?
	I understand that ifconfig in uml needs to know what IP address to
assign to the interface, but that could be handed in on the kernel command
line.

previous post, Bill:
> The right to access host networking is now down to whether ./linux is
> able to read and write to a preconfigured /dev/tap0.  Is that a more
> elegant approach?

previous port, Jeff:
	It's better, but I'm not all that happy with it. What security
problems arise from letting normal users set the ptp address (assuming
that address isn't already used) of an otherwise configured /dev/tap and
letting them send and receive packets over it? That's what I'd really like
to be able to do.

	I don't think the unprivileged user would need to set anything on
the host tap0.  All of the taps could have the same address (much like
having the same remote address on a terminal server for 253 dialup users),
and individual routes to the uml IP's could be set, one through each host
tap.  Since the addresses are being allocated ahead of time to allow the
routes to be put in place, the ptp addresses could be set at that time as
well.
	I don't particularly like the idea of preallocating network
devices that may not be used either.  If we don't, however, either ./linux
has to be given special privileges or the user needs special rights to
make those changes.  I'd rather see the devices preconfigured and left to
sit until used.

> > The ultimate way to accomplish this would be to have all a shared,
> > read-only filesystem with all the files that no-one needs to write to
> > and a much smaller writeable root with symlinks for directories and
> > writeable /var, /tmp, /etc, etc. directories.  Each uml mounts its own
> > writeable root and this shared space.
> 
> That's a decent idea.  I was thinking about some kind of an overlay filesystem 
> to do the same thing.

	Jeff, if you can code a true union filesystem that could someday
be part of the mainstream kernel, I would be forever in your debt.  I
wrote a long list of uses that could have, not the least of which back
then and here too would include the ability to mount a writeable small
ext2 partition on top of a readonly cdrom or readonly root_fs.
	I'm quite serious - that would be fantastic and would be truly
useful even outside of uml.

> > Jeff - if the file on the host OS is read-only for the user running
> > UML, can I mount it read-only in UML?  That would seem like a
> > reasonably secure way of making sure none of the umls can corrupt or
> > modify the shared filespace.
> 
> Yes.  What one kernel must not do is mount a device that another kernel has 
> mounted read-write.

	Agreed - they'd all have to be RO.

> He's talking about sticking bind in a uml, which could work well.  It's not 
> like a million-hit-per-day web server.

	True!
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"``Threads are like salt.  You like salt, I like salt, but we eat a
lot more pasta than salt.''  The thread guys are trying to tell you that
diet of salt is a good idea.  They are wrong, don't listen, eat more 
pasta and be happy."
	-- Larry McVoy <lm...@bi...>
--------------------------------------------------------------------------
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
--------------------------------------------------------------------------

Re: [User-mode-linux-user] security ??

[Jeff D. <jd...@ka...>]

la...@no... said:
> In Linux/a386, I just replaced the asm("hlt") with
> a386_interrupt_wait(), which sleeps until woken up by a signal or
> until the next timer interrupt (aka SIGTVALRM) would have occurred. 

That's more or less what I'm planning.

What complicates things is I'd have to have the input thread start delivering 
signals to sleeping idle threads when there's some input to handle, and that's 
racy.

This will be especially handy with MP emulation.  Currently, you'll have 2 or 
4 (or whatever) idle threads all spinning when they have nothing to do.

				Jeff

Re: [User-mode-linux-user] Any Debian experts out there?

[Jeff D. <jd...@ka...>]

> cant you just install it somewhere and then use dd to make an image of
> that disk ?? 

I could.  But I'd also like to ship that procedure as a script (i.e. my 
mkrootfs script is essentially what I'm asking for with RH 6.0) for people who 
have the media but don't want to download a whole filesystem.

A real install and dd is also a pain in the a**.

				Jeff

Re: [User-mode-linux-user] Any Debian experts out there?

[Jon B. <ben...@di...>]

cant you just install it somewhere and then use dd to make an image of
that disk ??


ion++


On Tue, 14 Mar 2000, Jeff Dike wrote:

> I've a whole bunch of CDs with various distros on them and I'm planning on 
> putting together a bunch of filesystems from them so that people can play with 
> a new distro by downloading the appropriate filesystem.
> 
> Right now, I'm playing with Debian.  What I need is some procedure for 
> starting with a CD and an empty filesystem and ending with a filesystem that 
> will boot a reasonable system.
> 
> I've got the Debian install tools in another filesystem (the Debian filesystem 
> that I distribute - my outer system has RH on it).  In order to get dpkg to 
> even think about working in the empty filesystem, I had to do a bunch of 
> mkdirs and touch a bunch of files that it expected to already be there.
> 
> After that, it kind of went through some motions and gave me more errors which 
> I didn't understand.
> 
> So, what's the official way to do this?  There's probably some kind of 
> bootstrap environment that the install uses to get dpkg up and running.
> 
> And if anyone has the equivalent answers for other distros, I've got the 
> following:
> 	Debian 2.1
> 	Slackware 7.0
> 	Stampede Linux Edition 3
> 	Linux-Mandrake 7.0
> 	Caldera OpenLinux 2.3
> 	Red Hat 6.1
> 	Suse 6.3
> 
> 				Jeff
> 
> 
> 
> _______________________________________________
> User-mode-linux-user mailing list
> Use...@li...
> http://lists.sourceforge.net/mailman/listinfo/user-mode-linux-user
> 

[User-mode-linux-user] Any Debian experts out there?

[Jeff D. <jd...@ka...>]

I've a whole bunch of CDs with various distros on them and I'm planning on 
putting together a bunch of filesystems from them so that people can play with 
a new distro by downloading the appropriate filesystem.

Right now, I'm playing with Debian.  What I need is some procedure for 
starting with a CD and an empty filesystem and ending with a filesystem that 
will boot a reasonable system.

I've got the Debian install tools in another filesystem (the Debian filesystem 
that I distribute - my outer system has RH on it).  In order to get dpkg to 
even think about working in the empty filesystem, I had to do a bunch of 
mkdirs and touch a bunch of files that it expected to already be there.

After that, it kind of went through some motions and gave me more errors which 
I didn't understand.

So, what's the official way to do this?  There's probably some kind of 
bootstrap environment that the install uses to get dpkg up and running.

And if anyone has the equivalent answers for other distros, I've got the 
following:
	Debian 2.1
	Slackware 7.0
	Stampede Linux Edition 3
	Linux-Mandrake 7.0
	Caldera OpenLinux 2.3
	Red Hat 6.1
	Suse 6.3

				Jeff

Re: [User-mode-linux-user] security ??

[lars b. <la...@no...>]

Jeff Dike <jd...@ka...> writes:
> > and the idle thread tends to soak up a lot of CPU time
> It busy-waits, just like any normal idle thread.  I'm planning on using the 
> kernel's APM support to turn the busy-waiting into sleeping.  This will make 
> uml put a lower load on the system.

In Linux/a386, I just replaced the asm("hlt") with a386_interrupt_wait(),
which sleeps until woken up by a signal or until the next timer interrupt
(aka SIGTVALRM) would have occurred.

Re: [User-mode-linux-user] security ??

[Jeff D. <jd...@ka...>]

I replied privately (by accident) with this:

It is possible right now for a hostile process to break out of the user-mode 
virtual machine and gain access to the hosting machine.

However, this is fairly easy to fix.  It would not be hard to turn the 
virtual machine into a secure jail.

> It's essentially a software sandbox; a safe place to play with
> untrusted apps.  I plan to write an article for LinuxMonth expanding
> on that idea and showing how to do it.

It is good for that, since a hostile program would have to specifically attack 
the kernel.  I don't think many such attacks have been implemented since uml 
is still fairly new...

> one could still not use it to directly gain root in the host OS.

Right.  You're root inside, but outside you're just a normal user.  So 
something that manages to break out goes from being root inside to being 
unpriviledged outside.

> I believe I remember reading that when a disk sector is being
> requested, the rest of uml is held up

The block driver is completely synchronous at this point.  I don't know how 
this affects performance, but it probably doesn't help.

> and the idle thread tends to soak up a lot of CPU time

It busy-waits, just like any normal idle thread.  I'm planning on using the 
kernel's APM support to turn the busy-waiting into sleeping.  This will make 
uml put a lower load on the system.

> The ultimate way to accomplish this would be to have all a shared,
> read-only filesystem with all the files that no-one needs to write to
> and a much smaller writeable root with symlinks for directories and
> writeable /var, /tmp, /etc, etc. directories.  Each uml mounts its own
> writeable root and this shared space.

That's a decent idea.  I was thinking about some kind of an overlay filesystem 
to do the same thing.

> Jeff - if the file on the host OS is read-only for the user running
> UML, can I mount it read-only in UML?  That would seem like a
> reasonably secure way of making sure none of the umls can corrupt or
> modify the shared filespace.

Yes.  What one kernel must not do is mount a device that another kernel has 
mounted read-write.

He's talking about sticking bind in a uml, which could work well.  It's not 
like a million-hit-per-day web server.

				Jeff

Re: [User-mode-linux-user] security ??

[William S. <wst...@po...>]

On Tue, 14 Mar 2000, Jon Bendtsen wrote:

> how secure is usermode linux ??

	In the sense that the uml environment cleanly partitions
applications, it adds a lovely level of security for untrusted apps.  
It's essentially a software sandbox; a safe place to play with untrusted
apps.  I plan to write an article for LinuxMonth expanding on that idea
and showing how to do it.
	The fact that (with a few current minor quirks) uml can be run as
a non-root user in the host OS means that even if the entire uml could be
somehow compromised (and even that's a real stretch for me), one could
still not use it to directly gain root in the host OS.

> i was thinking at using it to implement a virtual linux for each and every
> service my computer runs, and then have the actualy linux at the computer
> to run backups and checking the log files to see what happens.

	A nice idea, with a few considerations.  I hope others will chime
in with their ideas too:

	UML appears to have somewhat of a performance overhead that I'm
sure Jeff is working on.  Networking can have noticeable delays, I believe
I remember reading that when a disk sector is being requested, the rest of
uml is held up (please clarify, Jeff; it's a hazy memory), and the idle
thread tends to soak up a lot of CPU time (although it's not clear whether
that truly has any effect on either uml or the host).  Granted, my only
testing has been on a machine with a slow CPU and very little memory;
others may find that it flies!
	For that reason, I'd suggest that a good half-way point might be
to run all daemons that need to be run as root in a uml.

	The ultimate way to accomplish this would be to have all a shared,
read-only filesystem with all the files that no-one needs to write to and
a much smaller writeable root with symlinks for directories and writeable
/var, /tmp, /etc, etc. directories.  Each uml mounts its own writeable
root and this shared space.  The local root filesystem could include just
the locally run daemons.
	It would take quite a bit of work to do correctly, but it would be
a fun project.
	Jeff - if the file on the host OS is read-only for the user
running UML, can I mount it read-only in UML?  That would seem like a
reasonably secure way of making sure none of the umls can corrupt or
modify the shared filespace.

	Cheers,
	- Bill

---------------------------------------------------------------------------
	Having Microsoft give us advice on open standards is like W.C. Fields 
giving moral advice to the Mormon Tabernacle Choir
	-- Scott McNealy, Sun Microsystems Inc.
(Courtesy of Michael Remski <mr...@ix...>)
--------------------------------------------------------------------------
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
--------------------------------------------------------------------------

[User-mode-linux-user] security ??

[Jon B. <ben...@di...>]

how secure is usermode linux ??

i was thinking at using it to implement a virtual linux for each and every
service my computer runs, and then have the actualy linux at the computer
to run backups and checking the log files to see what happens.



ion++

Re: [User-mode-linux-user] lmbench numbers?

[Jeff D. <jd...@ka...>]

la...@no... said:
> Has anyone tried lmbench in uml? 

Not as far as I know.  What does it measure?

			Jeff

[User-mode-linux-user] lmbench numbers?

[lars b. <la...@no...>]

Has anyone tried lmbench in uml?

[User-mode-linux-user] 2.3.51 updates

[Jeff D. <jd...@ka...>]

I've added a couple of bug fixes which allow the network device to be shut 
down by hand, and I've hooked up the chroot and creat system calls.

I've updated the packages, the patch, and kernel.

				Jeff

[User-mode-linux-user] 2.3.51 available

[Jeff D. <jd...@ka...>]

Nothing new in this one.  Just the update to 2.3.51.

The patch, kernel, and RH package are updated.  The Debian package will be 
updated as soon as I copy the filesystem to sourceforge.

				Jeff

[User-mode-linux-user] (no subject)

[Jeff D. <jd...@ka...>]

I've updated the patch and kernel at http://sourceforge.net/project/filelist.ph
p?group_id=429

The user-visible change is:
	The virtual console xterms now disappear when you log out and reappear when 
the new getty starts up.  They will also disappear for good if you comment 
them out of inittab and 'kill -1 1' which is what I'm fixing with this change. 
 If anyone really prefers the old behavior, let me know, and maybe we can work 
something out.  I think the new behavior is more correct.

				Jeff

[User-mode-linux-user] the user-mode kernel and expect

[Jeff D. <jd...@ka...>]

Unsurprisingly, they work well together.  

I'm chasing the fsck crash.  Since it only happens when fsck needs to fix up a 
dirty filesystem, and, even then, most of the time, it works, I needed to 
repeatedly boot the system, log in, cause some file activity, and crash it.

After a fair amount of experimentation and learning tcl, I ended up with this 
expect script:

set send_slow {1 .2}
for {} 1 {} {
    set timeout -1
    spawn linux fhd0=../../root_fs_deb fhd1=swap
    expect "usermode login:" 
    send -s "root\r"
    expect "Password:"
    send -s "root\r"
    expect "usermode:~#"
    send -s "dd if=/dev/disk/0 of=/mnt/data &\r"
    expect "usermode:~#"
    sleep 1
    system "kill -9 `ps uax | grep linux | awk '{print \$2}'`"
}

Pretty clean and simple once you know the syntax.

				Jeff

[User-mode-linux-user] Stupid (user-mode) kernel tricks

[Jeff D. <jd...@ka...>]

Boot the kernel, and log in on two or three consoles.
As simultaneously as possible, start a "ls -alR /" in each one.
Watch the system switch from one to the other every few seconds...

This looks like a good way to graphically demonstrate time-sharing.  Most 
kernels switch too fast for time-sharing to be visible.  Not mine :-)

				Jeff

[User-mode-linux-user] 0.10-2.3.49 is available

[Jeff D. <jd...@ka...>]

I think this one fixes the SIGIO problem.
I hooked up a handful of system calls, and put back the changing of external 
process names to their internal names.

The patch and the kernel are available from the project download page:
	http://sourceforge.net/project/filelist.php?group_id=429

				Jeff

[User-mode-linux-user] 2.3.48 is out

[Jeff D. <jd...@ka...>]

The 2.3.48 user-mode kernel is available, as I'm sure lots of you have noticed 
from freshmeat.

The binaries have devfs support in them (pure devfs - there is no /dev, except 
for the mount point).

Get while the getting's good at http://sourceforge.net/project/filelist.php?gro
up_id=429.

				Jeff

Re: [User-mode-linux-user] 2.3.48_devfs

[Jeff D. <jd...@ka...>]

> > > it's perfect for my hardware.
> > 
> > Blazing fast processors and tons of everything :-) ?
> 	*smile*  P90, 40M maxed out laptop, and tons 'o nothing! 

I developed this thing on a P66, and it kind of dragged, especially when I was 
doing SMP (that machine died a few weeks ago, and I was forced to upgrade :-).

> 	Awwww, and I _liked_ the xterms.  That's a really cool feature.
> Could they at least be an option? 

I do, too.  I'm not turning them off in the kernel.  If you really like them, 
go into /etc/inittab in the package fs, and uncomment the '1' and '2' lines.  
The xterms will be back, and the kernel will start crashing on you.

> ... run "cd /", then "ls -alR" on an xterm... crash with Unimplemented
> syscall : 163
>                            Untested (18468) [0x10154c68]:
> syscall_kern.c line 662 on the console... press Ctrl-C on gdb:

> Program received signal SIGINT, Interrupt. 0x10085529 in __wait4 ()
> (gdb) print current_task.comm $1 = "ls\000h\000\000r\000\000\000\000\00
> 0\000\000\000" 

That's fine.  Just knowing that it was ls -alR would have been good.  I 
thought you were maybe running some obscure little program that I hadn't tried 
yet.

> 	I'll mail you my devfsd rpm I just finished preparing under separate
> cover; it's available on request for anyone else. 

I'm doing that differently now.  I'll see if I can slide this rpm thing in 
somehow.

				Jeff

Re: [User-mode-linux-user] 2.3.48_devfs

[William S. <wst...@po...>]

On Wed, 1 Mar 2000, Jeff Dike wrote:

> > it's perfect for my hardware.
> 
> Blazing fast processors and tons of everything :-) ?

	*smile*  P90, 40M maxed out laptop, and tons 'o nothing!

> > 	- 2.3.48_devfs-uml is more stable than 2.3.46-uml; the latter crashed
> > within a few minutes and sometimes before it finished booting.
> > 2.3.48_devfs stays up longer.
> 
> I don't remember fixing any bugs like that between .46 and .48.  Maybe bugs 
> were fixed in the rest of the kernel.  One thing that will help stability when 

	Very likely.

> I put the packages out is that virtual consoles are turned off.  My forking 
> and execing of xterm is causing memory corruption.  I have a fix in mind, and 
> that will be added pretty quickly.

	Awwww, and I _liked_ the xterms.  That's a really cool feature.  
Could they at least be an option?

> > 	- For those without a devfs capable root_fs, the quick and easy
> > workaround is to add devfs=nomount to the command line, ala ./linux
> > devfs=nomount 
> 
> So you're one of the ones that grabbed the devfs kernel.  My release notes 
> said not to do that :-)  I put it out for completeness, and for the masochists 
> out there.

	*hand raised*  Me, Teach, me!  I'm a software masochist!
	I'm playing for the moment, and the only precompiled one you had
was w/ devfs.  Luckily, the "devfs=nomount" basically returns the kernel
to it's normal behavior for backwards compatibility.
	I'll mail you my devfsd rpm I just finished preparing under
separate cover; it's available on request for anyone else.

> > 	- I'm sure it doesn't surprise you that I'm getting occasional errors
> > on the uml console.  The latest (on 2.3.48_devfs w/ debian root_fs):
> > Unimplemented syscall : 163
> >                            Untested (16019) [0x10154c68]:
> > syscall_kern.c line 662 
> 
> Something did a mremap.  Not all of the system calls are hooked up yet.  If 
> you see that again, attach gdb to the tracing thread, and look at 
> current_task.comm, and tell me what the command line was for it.  I like to 

	And here's where I get to publicly admit that I'm a complete
novice at debugging.  I don't know how to use gdb.  Oh well, dive in and
learn:

[root@sparrow uml]# gdb linux-2.3.48_devfs 18245
[snip]
Attaching to program: /home/wstearns/uml/linux-2.3.48_devfs, Pid 18245
0x10085529 in __wait4 ()
(gdb) c
Continuing.

... run "cd /", then "ls -alR" on an xterm... crash with
Unimplemented syscall : 163
                           Untested (18468) [0x10154c68]: syscall_kern.c line 662
on the console... press Ctrl-C on gdb:

Program received signal SIGINT, Interrupt.
0x10085529 in __wait4 ()
(gdb) print current_task.comm
$1 = "ls\000h\000\000r\000\000\000\000\000\000\000\000"

	Any closer?
	Just for reference, this is 2.3.48_devfs with the debian root_fs
from SourceForge.

> > I'll try putting together my own root_fs at some point, but need a
> > working loop device in my host kernel first; 2.3.48-i386 doesn't quite
> > seem to have it yet.
> 
> Where do you get your kernels from?  My RH kernels have always had the loop 
> device.

	My apologies.  The kernel itself _has_ the loop device, but I'm
spending time in the 2.3.4x kernels (see "software masochist",
above).  The loop mount is _there_ but seems unstable at this time.
	Thanks again for all your work.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"As a computer I find your faith in technology amusing."
(Courtesy of Gerhard Mack <gm...@im...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at: http://www.pobox.com/~wstearns/
--------------------------------------------------------------------------

Re: [User-mode-linux-user] 2.3.48_devfs

[Jeff D. <jd...@ka...>]

> it's perfect for my hardware.

Blazing fast processors and tons of everything :-) ?

> 	- 2.3.48_devfs-uml is more stable than 2.3.46-uml; the latter crashed
> within a few minutes and sometimes before it finished booting.
> 2.3.48_devfs stays up longer.

I don't remember fixing any bugs like that between .46 and .48.  Maybe bugs 
were fixed in the rest of the kernel.  One thing that will help stability when 
I put the packages out is that virtual consoles are turned off.  My forking 
and execing of xterm is causing memory corruption.  I have a fix in mind, and 
that will be added pretty quickly.

I do want to hear about panics and other strange behavior.  Especially if you 
can reproduce it.

> 	- For those without a devfs capable root_fs, the quick and easy
> workaround is to add devfs=nomount to the command line, ala ./linux
> devfs=nomount 

So you're one of the ones that grabbed the devfs kernel.  My release notes 
said not to do that :-)  I put it out for completeness, and for the masochists 
out there.

> 	- I'm sure it doesn't surprise you that I'm getting occasional errors
> on the uml console.  The latest (on 2.3.48_devfs w/ debian root_fs):
> Unimplemented syscall : 163
>                            Untested (16019) [0x10154c68]:
> syscall_kern.c line 662 

Something did a mremap.  Not all of the system calls are hooked up yet.  If 
you see that again, attach gdb to the tracing thread, and look at 
current_task.comm, and tell me what the command line was for it.  I like to 
watch the newly-hooked-up system calls run once just to make sure that I 
didn't mess something up.

> 	Am I right in guessing that you already know exactly from
> execute_syscall which ones need work, or is it useful to hear about
> these, either on the mailing list or privately? 

Yup.  Either way is fine.

> I'll try putting together my own root_fs at some point, but need a
> working loop device in my host kernel first; 2.3.48-i386 doesn't quite
> seem to have it yet.

Where do you get your kernels from?  My RH kernels have always had the loop 
device.

				Jeff

[User-mode-linux-user] 2.3.48_devfs

[William S. <wst...@po...>]

Good day, all,
	Many thanks to Jeff, Rusty, and all the UML developers and
contributors; I've just tried out the uml kernel and it's amazing!  I see
a lot of interesting possibilities for this, and it's perfect for my
hardware.
	A few notes:
	- 2.3.48_devfs-uml is more stable than 2.3.46-uml; the latter
crashed within a few minutes and sometimes before it finished booting.  
2.3.48_devfs stays up longer.
	- For those without a devfs capable root_fs, the quick and easy
workaround is to add devfs=nomount to the command line, ala
./linux devfs=nomount
	- I'm sure it doesn't surprise you that I'm getting occasional
errors on the uml console.  The latest (on 2.3.48_devfs w/ debian root_fs):
Unimplemented syscall : 163
                           Untested (16019) [0x10154c68]: syscall_kern.c line 662
This showed up while top and ls -alR / were running.
	Am I right in guessing that you already know exactly from
execute_syscall which ones need work, or is it useful to hear about these,
either on the mailing list or privately?

	I'll try putting together my own root_fs at some point, but need a
working loop device in my host kernel first; 2.3.48-i386 doesn't quite
seem to have it yet.
	Thanks again for all your work!
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"Eagles may soar, high and proud, but weasels don't get sucked
into jet engines."
(Courtesy of Mike Andrews <man...@te...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at: http://www.pobox.com/~wstearns/
--------------------------------------------------------------------------

[User-mode-linux-user] 2.3.46 available

[Jeff D. <jd...@ka...>]

All of the packages have been updated.  Get them from the project download 
page: http://sourceforge.net/project/filelist.php?group_id=429

				Jeff

Re: [User-mode-linux-user] modules

[Jeff D. <jd...@ka...>]

> > What do you need them for? 
> 
> I'm in a group working on the global filesystem (www.globalfilesystem.org). 
> I thought it might be nicer for catching panics. It has been
> built statically into the kernel, but by far the most common way is as
> a module.  

Do you want modules badly enough to put
	define_bool CONFIG_UNIX y
in arch/um/config.in and rebuilding?  You probably will also get some 
undefined symbols and will have to go grepping around in my code putting 
EXPORT_SYMBOL() in the appropriate places.

If you do that and it works, I will gladly put the changes in...

> Also, everytime I run ./linux the fsck fails at the beginning. I saw
> something of the sort in the debugging session in the howto but it
> looked like it had been resolved.


Failing as in the kernel crapping out every time it runs fsck, or fsck 
crapping out, but not doing anything strange with the kernel?

If the kernel is crapping out, I need to know how, what, where, etc.

In any case, you can probably work around that by fscking the filesystem from 
outside.
				Jeff