user-mode-linux-devel

Re: [User-mode-linux-devel] Portability

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

> > Posix compatibility is definitely not enough.  I think the 
> > biggest problem would be the ability to intercept and modify 
> > Linux system calls.  Nothing works without that.
>
> Ah, a light dawns.  I wondered (without wondering enough to learn the
> Linux and uml internals) how you virtualized the machine.  This
> combines with the comments about debugging yourself to explain the
> fundamentals... 

Actually, with an OS that maps reasonably well to Linux, it might not be hard 
to replace the top layer (the bit that intercepts system calls) with its 
equivalent.  It would run its own binaries and map those syscalls onto Linux 
syscalls.

Everything below that layer it Posix, I think.  Except maybe some of the 
mmaping that I do.

				Jeff

Re: [User-mode-linux-devel] Portability

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

la...@no... said:
> I wonder if a uml-light would be possible by just redirecting the
> shared library system call functions? 

vfork has to run on a different stack, although you could just make vfork a 
synonym for fork in the shared library.  You'd also have to find a way to 
locate kernel addresses (or maybe just a single kernel address) from the 
shared library.

That also doesn't help static executables.

> Great, but I'll be damned if I can find it. :) 

There's a way to get my page hit counts up.  Just tell people that there's a 
link to their site somewhere on mine :-)
				Jeff

Re: [User-mode-linux-devel] Portability

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

Glenn Chambers <GCh...@pr...> writes:
> > Posix compatibility is definitely not enough.  I think the 
> > biggest problem would be the ability to intercept and modify 
> > Linux system calls.  Nothing works without that.
> Ah, a light dawns.  I wondered (without wondering enough to
> learn the Linux and uml internals) how you virtualized the
> machine.  This combines with the comments about debugging
> yourself to explain the fundamentals...

I wonder if a uml-light would be possible by just redirecting
the shared library system call functions?

Re: [User-mode-linux-devel] Portability

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

Jeff Dike <jd...@ka...> writes:
> > Jeff, am I allowed to advertise my project here? 
> Sure.  It's related, and you've got a pointer to my stuff on your site :-)

Great, but I'll be damned if I can find it. :)

I'll just point out that I'm doing this user-mode port with a slightly
different angle.  It's might be duplicated effort, but hey, this is for
fun.

My Linux/a386 project is just like uml, but with a separate
CPU/hardware layer called a386.  I think this will make it easier to
port to other Unices, as well as make it easier to port other
operating systems to user-mode.  NetBSD/a386 is my next target when
Linux/a386 stabilizes.

See
        http://a386.nocrew.org/
and
        http://linux.a386.nocrew.org/
(new locations; Jeff, please update)

> Posix compatibility is definitely not enough.  I think the biggest problem 
> would be the ability to intercept and modify Linux system calls. 

I don't think ptrace() is in POSIX at all, so that is a huge problem.

RE: [User-mode-linux-devel] Portability

[Glenn C. <GCh...@pr...>]

> Posix compatibility is definitely not enough.  I think the 
> biggest problem would be the ability to intercept and modify 
> Linux system calls.  Nothing works without that.

Ah, a light dawns.  I wondered (without wondering enough to
learn the Linux and uml internals) how you virtualized the
machine.  This combines with the comments about debugging
yourself to explain the fundamentals...

Cool.

Re: [User-mode-linux-devel] Portability

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

> Jeff, am I allowed to advertise my project here? 

Sure.  It's related, and you've got a pointer to my stuff on your site :-)

> Please note that Cygwin can't implement mmap(MAP_FIXED) for Windows 95/
> 98.  That's probably a Very Bad Thing if you want to run an operating
> system in user mode. 

That would hurt.  You definitely want to be able to map pages and have them go 
where you want.

Posix compatibility is definitely not enough.  I think the biggest problem 
would be the ability to intercept and modify Linux system calls.  Nothing 
works without that.  There is also code that is written with knowledge about 
how the system call path works.  I'm going to get rid of that, but for now, 
it's a definite portability problem.

				Jeff

Re: [User-mode-linux-devel] Portability

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

Glenn Chambers <GCh...@pr...> writes:
> Does user-mode linux require a linux host, or is Posix
> compatability enough?

Jeff, am I allowed to advertise my project here?

> I've been wondering if one could compile this baby on
> FreeBSD, or Darwin, or (flinch) Cygwin...

Please note that Cygwin can't implement mmap(MAP_FIXED) for
Windows 95/98.  That's probably a Very Bad Thing if you want
to run an operating system in user mode.

[User-mode-linux-devel] Portability

[Glenn C. <GCh...@pr...>]

Does user-mode linux require a linux host, or is Posix
compatability enough?

I've been wondering if one could compile this baby on
FreeBSD, or Darwin, or (flinch) Cygwin...

[User-mode-linux-devel] cvs updated to 2.3.36

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

No functional changes on this one.  Just dropped in the .36 patch and it 
worked.

The tag is v2_3_36.

				Jeff

[User-mode-linux-devel] more cvs stuff

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

I updated cvs to 2.3.35.

I also added v2_3_32, v2_3_33, v2_3_34, and v2_3_35 tags to the appropriate 
revisions.  In theory (i.e. I haven't tested this), this means that, if at any 
point in the future, you need a 2.3.32 kernel and nothing else will do, then 
you can "co -r v2_3_32 linux" and get the appropriate sources.

				Jeff

[User-mode-linux-devel] Latest checkins

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

Here are the highlights:

Everything that needed it was updated to 2.3.32.

It now handles low-memory situations better.  Processes are now killed if 
something unrecoverable happens like a memory allocation failure during exec 
and a fault not being handled due to a lack of memory.

Closed up some memory leaks.

Made the default physical memory size 16 meg.

				Jeff

[User-mode-linux-devel] Current status

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

I've traced the various booting problems with the defaults to running short of 
memory.  By default, the kernel configures itself with 8 meg.  Something 
happened around 2.3.30 which caused it to run short of memory sooner or behave 
differently when it ran short.  I didn't notice because I was running with 64 
meg getting the kernel to behave when running apache.

So, as a workaround with the existing kernel, you can specify a larger memory 
with "mem=" on the command line.  I booted it with "mem=16M".  The memory size 
can go to slightly less than 768 meg.  I've only tested it up to 256 meg, 
though.

				Jeff

[User-mode-linux-devel] checkins

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

arch/um/Makefile - buglet fix
arch/um/main.c - prints a newline before exiting
arch/um/kernel/syscall_kern.c - Hooked up sys_ipc
arch/um/include/kern_util.h
arch/um/include/user_util.h
arch/um/kernel/exec_kern.c 
arch/um/kernel/exec_user.c 
arch/um/kernel/process.c 
arch/um/kernel/process_kern.c 
arch/um/kernel/signal_user.c 
arch/um/kernel/syscall_user.c 
arch/um/kernel/time.c 
arch/um/kernel/time_kern.c 
arch/um/kernel/trap_kern.c 
arch/um/kernel/trap_user.c 
arch/um/kernel/um_arch.c 
include/asm-um/processor.h 
include/asm-um/system.h - These files were all involved in handling preemption

				Jeff

[User-mode-linux-devel] The public launch

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

I sent announcements to slashdot, lwn, and linuxtoday, as well as putting one 
on freshmeat.  Only slashdot has yet to run it.

Looking at the "reads" numbers on the stories on linuxtoday, it was sort of in 
the middle of the pack wrt its neighbors.

The freshmeat announcment (http://freshmeat.net/news/1999/12/20/945707666.html)
 generated 522 click-throughs to the home page and 65 to the download site.  
There's also an interesting comment there.  I think it's serious, but I'm not 
sure.

I don't get logs on the current site (mv.com), but I can judge the level of 
activity with 'ls -ult' (which shows last access times) on the html files, and 
it's been 'right now' all day.

So, I'd say it's been fairly successful.  Too bad about slashdot, though.

				Jeff

[User-mode-linux-devel] debugging proxy

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

I have hacked up a debugging proxy called ptproxy.  This program spawns
two processes.  Both are debugged by the proxy, and the debugging calls
made by one of the processes are redirected through the proxy to the other
process.

This is useful for the user-mode kernel, because it's debugging itself
to catch system calls.  This makes it impossible to directly attach a
debugger, because a process can only be debugged by one debugger at a
time.  With (an adaption of) ptproxy, the kernel can spawn a debugger
to debug itself.

Availability:
cvs -d :pserver:ano...@cv...:/usr/local/cvsroot login
(empty password)
cvs -d :pserver:ano...@cv...:/usr/local/cvsroot co ptproxy

[User-mode-linux-devel] User-mode port 0.2-2.3.31

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

What's new with the kernel:
	it's up to 2.3.31
	swapping works
	memory size and block device configuration has been added
	a few more system calls are hooked up
	virtual consoles are supported and come up in xterms
	I added some missed headers whose lack caused build problems for some people
	the patch is located at ftp://user-mode-linux.sourceforge.net/pub/user-mode-li
nux/patch-2.3.31.gz
	
What else is new:
	This is now an official project, with mailing lists, cvs, and everything else 
a self-respecting project should have, hosted on sourceforge.net.
	Rusty wrote a nice HOWTO (http://www.mv.com/ipusers/karaya/uml/UserModeLinux-H
OWTO.html), and this official project stuff is mostly his doing.
	The mailing lists are use...@li... and 
use...@li...; archives and subscription info are 
at http://www.sourceforge.net/mail/?group_id=429
	The web site has been redone, but because of domain name confusion, it is 
still located at http://www.mv.com/ipusers/karaya/uml/index.html.

				Jeff

[User-mode-linux-devel] Updates to arch/um/main.c, arch/um/tlb.c and arch/um/syscall_kern.c

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

Here are the log entries:

RCS file: /cvsroot/user-mode-linux/linux/linux/arch/um/main.c,v
----------------------------
revision 1.2
date: 1999/12/16 17:16:57;  author: jdike;  state: Exp;  lines: +2 -2
Shrank the size of the space argument.  At 256, it somehow needed two execs
to pad out the command line enough, and that confused the argument parser.

RCS file: /cvsroot/user-mode-linux/linux/linux/arch/um/kernel/syscall_kern.c,v
----------------------------
revision 1.2
date: 1999/12/16 19:09:11;  author: jdike;  state: Exp;  lines: +4 -0
Hooked up fchown.
----------------------------

RCS file: /cvsroot/user-mode-linux/linux/linux/arch/um/kernel/tlb.c,v
----------------------------
revision 1.2
date: 1999/12/16 19:08:30;  author: jdike;  state: Exp;  lines: +19 -6
Fixed a bug which interpreted kernel pte flags according to the user pte
definitions, resulting in write-only kernel page tables.

[User-mode-linux-devel] the code is in CVS

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

I messed it up, but not fatally.

To put it in a pool, get a 2.3.31 pool with its root at ./linux/linux, and:

cvs -du...@cv...:/cvsroot/user-mode-linux co 
linux

I couldn't get anonymous CVS to work, but YMMV.

The nested linux directories are what I messed up.  I intended for there to be 
only one.

This will drop the user-mode sources into the appropriate places in the kernel 
pool.

Run 'make linux' and you will be off and running.

				Jeff

[User-mode-linux-devel] HOWTO change

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

I checked in the following patchlet to the HOWTO:

Index: doc/UserModeLinux-HOWTO.sgml
diff -c doc/UserModeLinux-HOWTO.sgml:1.3 doc/UserModeLinux-HOWTO.sgml:1.4
*** doc/UserModeLinux-HOWTO.sgml:1.3    Mon Dec  6 23:06:55 1999
--- doc/UserModeLinux-HOWTO.sgml        Tue Dec  7 16:27:26 1999
***************
*** 233,239 ****
        # mkdir /mnt/loop
        # mount -o loop ./root_fs /mnt/loop
        #
! <tscreen><verb>
  
  <item> Copy your filesystem across:
  
--- 233,239 ----
        # mkdir /mnt/loop
        # mount -o loop ./root_fs /mnt/loop
        #
! </verb></tscreen>
  
  <item> Copy your filesystem across:
  
				Jeff