Re: [9fans] 9pm/Plan 9 kernel in user space

Re: [9fans] 9pm/Plan 9 kernel in user space

[Richard Miller]

Today I said:

> I am currently running a Plan 9 server
> on a Xen virtual machine provided by a 'virtual hosting' company
> (http://www.blackcatnetworks.co.uk).

Coincidentally I've just heard from another (smaller) company
http://strugglers.net/wiki/Xen_hosting which also is willing to
"give it a go" with Plan 9.

-- Richard

Re: [9fans] 9pm/Plan 9 kernel in user space

[Charles Forsyth]

>>For servers tho, it could be a normal userspace program listening on a socket. Boring but possibly easier (cheaper)

that's roughly what is done (possibly roughly) by the odbc server on windows for Inferno,
supported a libstyx (which implements the same protocol).  there are python and (several) java client
libraries as well.

Re: [9fans] 9pm/Plan 9 kernel in user space

[Richard Bilson]

On 10/19/06, Sascha Retzki <sretzki@gmx.de> wrote:
> I can comment on that. Getting the facts, Microsoft once wrote some API so that anyone, with the right SDK, could write virtual file system drivers - they actually presented that stuff via a ftpfs, which made me optimistic that, if one gets this SDK, 9p4win32 could be written.

It seems like if you already have access to the Windows DDK, then

    http://www.acc.umu.se/~bosse/

has the rest of the pieces needed to do such a thing, and examples.

Of course, I can't confirm that it actually works. I took a look at
the examples and decided that I didn't have the stomach for it.

Re: [9fans] 9pm/Plan 9 kernel in user space

[Sascha Retzki]

> Hello!
> 

Hai,

> P.S.: ahh - and: how about 9P Clients and Servers for MS-Win*?

I can comment on that. Getting the facts, Microsoft once wrote some API so that anyone, with the right SDK, could write virtual file system drivers - they actually presented that stuff via a ftpfs, which made me optimistic that, if one gets this SDK, 9p4win32 could be written.
The big pro about that would be that a 9p-client would be 100% integrated in the Operating System - you would be able to see the 9p-connection/server in 'My Computer' like any other drive (virtual or not). I think that's what Nokia uses for their PC-Suite btw, in case you are a customer of this great company and their great.. eh.. Anyway

Then I saw the price of that SDK. It's arround hundred-and-something United States bucks, just for the SDK. If you are seriously gonna look into this, drop me a private mail and I will search for the presentations I collected on that topic, etc.


For servers tho, it could be a normal userspace program listening on a socket. Boring but possibly easier (cheaper)

Mfg, Sascha

Re: [9fans] 9pm/Plan 9 kernel in user space

[C H Forsyth]

>>The idea of Plan 9 in userspace still attracts me for several reasons:

i'd have found it useful myself to have a User Mode Plan 9 that could run
as a replacement for User Mode Linux, implemented a similar way
(so it could be virtually booted by sites hosting UML), but tidily.
given the way Plan 9 is organised, i thought it would be much
less work than it was originally for them to do User Mode Linux.

Re: [9fans] 9pm/Plan 9 kernel in user space

[Richard Miller]

Another possible alternative is running both Linux and Plan 9 under
Xen.

> - running Plan 9 alongside my Linux Desktop with reasonable speed

This works fine, if you are content with using VNC to access the
Plan 9 desktop (Xen does not yet export a virtual framebuffer which
would be easily usable as /dev/screen).

> - running CPU and Auth servers on Linux Servers with otherwise small
>   workload

This works extremely well.  I am currently running a Plan 9 server
on a Xen virtual machine provided by a 'virtual hosting' company
(http://www.blackcatnetworks.co.uk).  I'll post more details after
it's been tested a while longer.

> - doing same thing under Windows, eventually starting under Cygwin

Xen is no help with this one, however.

-- Richard

Re: [9fans] 9pm/Plan 9 kernel in user space

[Russ Cox]

If you are running Linux, you can use gcc.
cl is the Microsoft C compiler.

Russ

Re: [9fans] 9pm/Plan 9 kernel in user space

[Georg Lehner]

"Russ Cox" <rsc@swtch.com> writes:
...
> You'd be better off just porting fossil (venti is ported)
> to the current Plan 9 from User Space, which does
> not depend on a central user-space kernel.
...

Running fossil+venti on a Linux Kernel surely is an interresting
alternative: more efficient then my first idea, plus the same benefit:
lots of hardware drivers not supported in Plan 9.

The idea of Plan 9 in userspace still attracts me for several reasons:
- running Plan 9 alongside my Linux Desktop with reasonable speed
- running CPU and Auth servers on Linux Servers with otherwise small
  workload=20
- doing same thing under Windows, eventually starting under Cygwin

 I'll look into drawterm.


Charles Forsyth <forsyth@terzarima.net> writes:

>>"cl"? (Under Windows, you might expect your compiler to be called "cl.exe")
>
> plain cl is fine: it's just  not in his search path
> or perhaps not even on his machine

cl is not on my machine, where may I be able to get it from?

Thanks and

Regards,

        Jorge-León

Re: [9fans] 9pm/Plan 9 kernel in user space

[Charles Forsyth]

>"cl"? (Under Windows, you might expect your compiler to be called "cl.exe")

plain cl is fine: it's just  not in his search path
or perhaps not even on his machine

Re: [9fans] 9pm/Plan 9 kernel in user space

[Russ Cox]

You could get that version compiling with some
work, but I don't think it's a very good idea.
You'll still need to write a lot of code to deal
with things like process management, which a
user level kernel can't actually help with.
Having device files for the disks only gets you so far.

You'd be better off just porting fossil (venti is ported)
to the current Plan 9 from User Space, which does
not depend on a central user-space kernel.
http://swtch.com/plan9port/

If you still want a user-space kernel, the drawterm code
is essentially that, though it doesn't contain any
provisions for running "system calls" on behalf of
other processes.
http://swtch.com/drawterm/

Russ

Re: [9fans] 9pm/Plan 9 kernel in user space

[Andrew Simmons]

>     cl -c nologo -W3 ....   
>     cl: '/bin/cl'file does not exists
>     ... etc.
> 
Isn't just telling you that your system can't find a C compiler called 
"cl"? (Under Windows, you might expect your compiler to be called "cl.exe")

> Is there a chance to get this "Windows hosted" Plan 9 implementation
> compiled and running these days?  Can it reasonable be used to create
> a "Linux/Unix" hosted Plan 9 Kernel?
> 
Is this what you are looking for?
http://swtch.com/plan9port/

[9fans] 9pm/Plan 9 kernel in user space

[jorge-plan9]

Hello!

I'm very new to Plan 9, just installed three or so Inferno's, a PC
with Plan 9 and a QEMU instance under Linux with Plan 9.

As a systems/network administrator in a small software company I'm
interested in simple mainenance and setup of file shares in a
heterogeneous environment: Linux, Windows, some BSD's,...

9P2000, fossil and venti look great to me in this respect, however
there are lots of missing pieces to get a pure 9P driven environment
working, especially on the MS-Windows side.

While playing around with Plan 9 and Inferno at home, and with OpenVZ
at work, the idea struck me, that with the Plan 9 Kernel it should be
"easy" to write IO drivers which make use of a hosting OS-Environment,
e.g. Linux. and load the Kernel into use space: simulate hard disks
with Linux block devices, the console with character devices, the
screen with the VirtualConsole, a VNC-framebuffer or an X11-Windows,
etc.

Such a Plan 9 would be much faster then QEMU, probably very portable
to any Unix-derivate, and I could run as much of them as I liked, just
with OpenVZ.

I found out in http://swtch.com/usr/local/plan9/9pm/README that such a
thing had once be done under Windows, and tried to compile the
provided sources with my QEMU Plan 9, however failed to get over the
very first hoops:

    cl -c nologo -W3 ....   
    cl: '/bin/cl'file does not exists
    ... etc.

Is there a chance to get this "Windows hosted" Plan 9 implementation
compiled and running these days?  Can it reasonable be used to create
a "Linux/Unix" hosted Plan 9 Kernel?

Thanks for all, to bring in a fresh wind into the OS-Scene,

Regards,

        Jorge-Len

P.S.: ahh - and: how about 9P Clients and Servers for MS-Win*?