[uml-devel] Re: What to do with uml_switch?

[uml-devel] Re: What to do with uml_switch?

[Steve Schmidtke]

Michael Richardson wrote:
>  uml_netjig is a tool which does the following:
>   1) moves packets using "daemon" mode a la uml_switch.
>  2) has an expect-friendly command line interface  3) can create 
>*multiple* switches and keep them seperate

If they are separate, why not just run multiple uml_switches?

>  4) can record all packets to a .pcap file
>  5) can play packets from a .pcap file at a given rate     (or under 
>command line interface control)
>  6) can optionally answer ARP queries for imaginary nodes.

I like these 3, but could you not just connect another UML guest to the 
switch and run arp -s pub and tcpdump to accomplish the same thing?

>  7) can talk to tuntap device(s) as uml_switch could.
>
>  My patches refactor uml_switch to get uml_netjig + uml_switch'

I am still not clear if uml_netjig is a separate program running on top of 
uml_switch or inside it, but (and I apologize for being dense if this is 
what you are doing) I think the proper way to accomplish what you want is to 
create a "-monitor <socket>" option to uml switch to filter all network 
traffic through.  Then create a monitor program that can attach/detach from 
a uml_switch at will through the monitor socket.  A hook to filter packets 
entering uml_switch and a new function to inject packets (with/without 
updating MAC tables) might be all that's necessary.  All your traffic 
management and expect code could reside in the external monitor program.

Steve schmidtke




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: What to do with uml_switch?

[mcr]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Steve" == Steve Schmidtke <steve_schmidtke@hotmail.com> writes:
    Steve> Michael Richardson wrote:
    >> uml_netjig is a tool which does the following: 1) moves packets
    >> using "daemon" mode a la uml_switch.  2) has an expect-friendly
    >> command line interface 3) can create *multiple* switches and keep
    >> them seperate

    Steve> If they are separate, why not just run multiple uml_switches?

  a) because one sometimes needs to inject packets onto multiple networks
     in a specific order. uml_netjig permits this.

  b) once one has written uml_netjig to listen to the command channel
     and all of the sockets each switch has, there is little reason not
     to combine things. It doesn't add much complexity, and it makes
     managing the switch from expect much easier.

  {oh, I forgot, if you have the libnetdissect branch of tcpdump, you
can tcpdump all networks as well, which you can not always do from a
UML. Particularly if one is debugging an issue where a UML crashes as a
result of a packet-of-death}

    >> 4) can record all packets to a .pcap file 5) can play packets
    >> from a .pcap file at a given rate (or under command line
    >> interface control) 6) can optionally answer ARP queries for
    >> imaginary nodes.

    Steve> I like these 3, but could you not just connect another UML
    Steve> guest to the switch and run arp -s pub and tcpdump to
    Steve> accomplish the same thing?

  Sure, no problem :-)
  Please send:
	 a) MORE RAM
	    16-way system

	 b) Paulo to come and debug why UMLs to "rogue", and clog up
	    memory and system state. 
	    (Remember we start and stop some 1000 of them every night)
	    The more that we run, the worse the situation.

  Starting a UML to answer ARP and ICMP echo requests is silly.
  Particularly since our test system now takes 2-3 hours to run on a
multigigahertz box with lots of ram.

    Steve> I am still not clear if uml_netjig is a separate program
    Steve> running on top of uml_switch or inside it, but (and I
    Steve> apologize for being dense if this is what you are doing) I
    Steve> think the proper way to accomplish what you want is to create
    Steve> a "-monitor <socket>" option to uml switch to filter all

  Please go read the paper.

  uml_netjig is a superset of uml_switch.
  The refactoring of it means that uml_switch has just a main() that
is much simpler than that of uml_netjig.

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbO0h4qHRg3pndX9AQFxFQQA38+UCjmR4YnOGatlUFDYtbW/e+qWHAU/
4oDV9euDJiF6Tw7D4qDQ7eLMpjFvcAmDCTPGbfhXowbWDrXpGFhXZ7SxqQ0AzUIt
yyGFQeSrRp9GMvyDrQrmenI6q4rY6TtZItG1FcNz0E2g7VtnNY2YXN21+jT/Uujx
0ldP3wvZX2E=
=2gqt
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: What to do with uml_switch?

[Steve Schmidtke]

Michael Richardson wrote:
>     Steve> I like these 3, but could you not just connect another UML
>     Steve> guest to the switch and run arp -s pub and tcpdump to
>     Steve> accomplish the same thing?
>
>   Sure, no problem :-)
>   Please send:
>	 a) MORE RAM
>	    16-way system
>
>	 b) Paulo to come and debug why UMLs to "rogue", and clog up
>	    memory and system state.
>	    (Remember we start and stop some 1000 of them every night)
>	    The more that we run, the worse the situation.
>
>   Starting a UML to answer ARP and ICMP echo requests is silly.
>   Particularly since our test system now takes 2-3 hours to run on a
>multigigahertz box with lots of ram

Agreed, it would be an unwieldy solution.  And it's obvious you need fine 
grained control to manage packets in your test environment, sorry if it 
sounded like I was telling you what you should do.  What I was trying to 
point out (and thinking out loud) is that an ARP generator can be connected 
*to* uml_switch, it doesn't have to be built *into* uml_switch.  Strip down 
the UML kernel's daemon_user.c and what you have is a framework to talk 
socket datagrams to a uml_switch.  Add code to accept and generate arbitrary 
ethernet frames, add you now have an independant pluggable diagnostic tool 
anyone can use on the fly.  This is a solution that would probably satisfy 
95% of us mere mortals.

Steve Schmidtke




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Steve" == Steve Schmidtke <steve_schmidtke@hotmail.com> writes:
    Steve> Agreed, it would be an unwieldy solution.  And it's obvious
    Steve> you need fine grained control to manage packets in your test
    Steve> environment, sorry if it sounded like I was telling you what
    Steve> you should do.  What I was trying to point out (and thinking
    Steve> out loud) is that an ARP generator can be connected *to*
    Steve> uml_switch, it doesn't have to be built *into* uml_switch.

  Yes, that could be done.
  If you needed more sophistication than we have, then you'd want to do
exactly that.

    Steve> Strip down the UML kernel's daemon_user.c and what you have
    Steve> is a framework to talk socket datagrams to a uml_switch.  Add

  That's probably not the best idea. We put the interfaces in two places
right now... Best would be if a refactored uml_switch exported an
interface that replaced daemon_user.c :-)

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbc1w4qHRg3pndX9AQGyIgQAgnkAvJ1JWlsugt+VwwJz5Qg/qvv81lXa
nxrZ7gHWo4qn4oKE0bQS7FWgdNLnFbFm2Ii7VwYUQs9XVrYkIv0LN+DfjpRhBA3g
DB/KvaBkAAm6WNpRaoQ5hhKiBAtRE/ej3Wp/IU7EVw0Zz7e1OqplZUFlJ8dYvth4
kXOxWroli5Q=
=Izak
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Steve Schmidtke]

Michael Richardson wrote:
>     Steve> Strip down the UML kernel's daemon_user.c and what you have
>     Steve> is a framework to talk socket datagrams to a uml_switch.  Add
>
>   That's probably not the best idea. We put the interfaces in two places
>right now... Best would be if a refactored uml_switch exported an
>interface that replaced daemon_user.c :-)

It would be useful if daemon_user.c could be run in a synchronous mode such 
that packets are guaranteed delivered to the switch and guaranteed delivered 
to a host.  The switch could then mangle packets in a deterministic fashion 
for testing.

It would also be interesting if packets could be passed in shared memory.  
I'm wondering if it would be faster and more reliable at the same time.

Steve Schmidtke




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Steve" == Steve Schmidtke <steve_schmidtke@hotmail.com> writes:
    Steve> It would be useful if daemon_user.c could be run in a
    Steve> synchronous mode such that packets are guaranteed delivered
    Steve> to the switch and guaranteed delivered to a host.  The switch
    Steve> could then mangle packets in a deterministic fashion for
    Steve> testing.

  Yes, that would be useful.
  From a practical point of view:
       s/SOCK_DGRAM/SOCK_STREAM/

  not a very useful way to operate if there is any network load, but it
may be okay for regression testing. Also realize that Linux Unix domain
datagram sockets are mostly reliable. (BSD does not have this property)

    Steve> It would also be interesting if packets could be passed in
    Steve> shared memory.  I'm wondering if it would be faster and more
    Steve> reliable at the same time.

  On a uniprocessor, perhaps.
  Evidence is that on multiprocessor systems, shared memory is not
faster than copying because of cache concurency costs.

  My opinion is that this is premature optimization.

  I having been thinking of booting all my UMLs up to single-user
prompt, swsusp them, and restore swsusp each time instead of booting
them. That may be a larger win for me.

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [




-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbd1I4qHRg3pndX9AQG/HgP7BTfeATngfiD8qffUreEmSVYgmxFM/ZYz
cvS/qJjLVVghjL9Lf0KG56rOMqmVkQ47A7wE52coVLlHgXEIz0KX2j8lvuw2M3yg
4neNAulXa+yiIGf/U/GoxCe/K74cW0//tCV/Jlcl9g68AKzQ952AzEOtqy7D4hXc
sdxEKTXk//4=
=mf5Q
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Werner Almesberger]

Steve Schmidtke wrote:
> It would be useful if daemon_user.c could be run in a synchronous mode such 
> that packets are guaranteed delivered to the switch and guaranteed delivered 
> to a host.  The switch could then mangle packets in a deterministic fashion 
> for testing.

BTW, for testing purposes, one may also want to consider umlsim,
which gives you even more control over what happens to the packets
(including delay, etc.).

Right now, the networking support still isn't great, but it's
getting there ...

- Werner

-- 
  _________________________________________________________________________
 / Werner Almesberger, Buenos Aires, Argentina         wa@almesberger.net /
/_http://www.almesberger.net/____________________________________________/


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Blaisorblade]

On Tuesday 07 December 2004 17:06, Steve Schmidtke wrote:
> Michael Richardson wrote:
> >     Steve> I like these 3, but could you not just connect another UML
> >     Steve> guest to the switch and run arp -s pub and tcpdump to
> >     Steve> accomplish the same thing?

> >   Sure, no problem :-)
> >   Please send:
> >  a) MORE RAM
> >     16-way system

> >  b) Paulo to come and debug why UMLs to "rogue", and clog up
> >     memory and system state.
> >     (Remember we start and stop some 1000 of them every night)
> >     The more that we run, the worse the situation.
??? Could you give more detail? If you're getting host memory leaks, maybe 
updating SKAS could help... please send a separate message on that... there 
was with -V1 a big memory leak in some unusual situations.
> >   Starting a UML to answer ARP and ICMP echo requests is silly.
> >   Particularly since our test system now takes 2-3 hours to run on a
> >multigigahertz box with lots of ram
>
> Agreed, it would be an unwieldy solution.  And it's obvious you need fine
> grained control to manage packets in your test environment, sorry if it
> sounded like I was telling you what you should do.  What I was trying to
> point out (and thinking out loud) is that an ARP generator can be connected
> *to* uml_switch, it doesn't have to be built *into* uml_switch.

> Strip down 
> the UML kernel's daemon_user.c and what you have is a framework to talk
> socket datagrams to a uml_switch.  Add code to accept and generate
> arbitrary ethernet frames, add you now have an independant pluggable
> diagnostic tool anyone can use on the fly.  This is a solution that would
> probably satisfy 95% of us mere mortals.

It seems reasonable to me too, and a good application of the Unix KISS 
principle...

This way, it is also easier to control the frame injection, since the injector 
is a separate program.

However I've seen the timing argument and the discussion went much further 
while I was away, which is good... as I said, I just wanted to make you start 
working on that, not to work on it myself, since I have 0 experience for 
this.

My only advice is to start packaging at least the initial work... Michael, you 
had offered to actually maintain uml_switch, so why don't you start merging 
at least the earliest Steve's cleanups?

Maintainance does not only mean development, also caring about such details 
and merging good patches from anybody else...
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Blaisorblade" == Blaisorblade  <blaisorblade_spam@yahoo.it> writes:
    Blaisorblade> My only advice is to start packaging at least the
    Blaisorblade> initial work... Michael, you had offered to actually
    Blaisorblade> maintain uml_switch, so why don't you start merging at
    Blaisorblade> least the earliest Steve's cleanups?

  Steve, please point me at your patches/cleanups.
  Please understand that I may just start from my merged uml_netjig/uml_switch.
 
  I'm happy to go through your patches and adapt them to my code.
  Where do I put the result?
  
- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbuMloqHRg3pndX9AQEiJwP/blgTjaghFovqUtw/NlO0kokpvYctud3M
R4eP1uiCs1YpLhlbjq+ObLfjWhxj/+XBYWuYk4qxM8GmyUxOPmj9V/yRRHJopHQ2
fSnxutMjDowmq9PkGe885yekkgEN+UHkbM0RGVPOODC7RtgFar3V52dMsvqLGWFD
DEyXNooJ+mY=
=Wr1F
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Steve Schmidtke]

Michael Richardson wrote:
> >>>>> "Blaisorblade" == Blaisorblade  <blaisorblade_spam@yahoo.it> writes:
>     Blaisorblade> My only advice is to start packaging at least the
>     Blaisorblade> initial work... Michael, you had offered to actually
>     Blaisorblade> maintain uml_switch, so why don't you start merging at
>     Blaisorblade> least the earliest Steve's cleanups?

I was hoping to have a working -monitor <port> option before I spoke up, but 
that's going to take a while to sort out.

So, in the mean time, I'd like to announce that I would like to take over as 
maintainer of uml_switch, if that's ok.

My intent is to have uml_switch as a functional, virtual, network switch, 
solidly performing the tasks normally expected of a real switch.  I would 
also like to see it as a platform that can support extensions for networking 
experiments and diagnostic tools.  Oh. and instead of "uml_router" or 
"uml_switch", I would like to propose a new name, "User Mode Switch", or 
"ums" for short. :)

>   Steve, please point me at your patches/cleanups.
>   Please understand that I may just start from my merged 
>uml_netjig/uml_switch.

At this point, it might be easier for me to incorporate the uml_netjig 
changes than the other way around.  I found 5 patches (1-5) for the 
uml_netjig foundation in the mailing list archive, and I'm OK with the basic 
premises I have seen:  that uml_switch should support multiple switches, and 
that  uml_switch should have a script friendly command interface for 
controlling those switches.  If we can agree that the other features like 
tcpdumps and fake ARPs, etc. are not core to the nature of a switch (but can 
be patched in when needed), then I would be happy to maintain the above 
premises as well.

Steve Schmidtke




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Steve" == Steve Schmidtke <steve_schmidtke@hotmail.com> writes:
    Steve> uml_netjig changes than the other way around.  I found 5
    Steve> patches (1-5) for the uml_netjig foundation in the mailing
    Steve> list archive, and I'm OK with the basic premises I have seen:

  Great.
  Let me know when I should diff against what I have, and re-submit more
patches.

    Steve> for controlling those switches.  If we can agree that the
    Steve> other features like tcpdumps and fake ARPs, etc. are not core
    Steve> to the nature of a switch (but can be patched in when
    Steve> needed), then I would be happy to maintain the above premises
    Steve> as well.

  I'd rather see the multiple switch support in a seperate binary.
  You'll see that in my patches uml_netjig and uml_switch each have
their own poll(2)-loop. That's because the uml_switch is significantly
simpler.

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbyvKYqHRg3pndX9AQGwsAQA1U/+gyxKFjOfBeBSo1T4/9QUIYcK9QWS
h0NI8NWNMTWT3E/nHSms01ZjYfiusWf7eWba7FMs8d2jj09d7fojoSDYFHxslOw2
6voXfQwQdwAHhfvpPBgjA6+FZtQizliz7zH+pGgpnRIEOnvx7ZCpiZ5z/fDL08eL
z8TkqkSpvoU=
=rmIm
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Blaisorblade]

On Sunday 12 December 2004 01:11, Michael Richardson wrote:
> >>>>> "Blaisorblade" == Blaisorblade  <blaisorblade_spam@yahoo.it> writes:
>
>     Blaisorblade> My only advice is to start packaging at least the
>     Blaisorblade> initial work... Michael, you had offered to actually
>     Blaisorblade> maintain uml_switch, so why don't you start merging at
>     Blaisorblade> least the earliest Steve's cleanups?
>
>   Steve, please point me at your patches/cleanups.
>   Please understand that I may just start from my merged
> uml_netjig/uml_switch.
>
>   I'm happy to go through your patches and adapt them to my code.
>   Where do I put the result?
I can add a link from my homepage (hopefully even on the main UML site) to get 
it widely used (if you want to do it through SourceForge it's also possible, 
but maybe sucks a bit more). Just remember me to do it when you've found a 
site.

I think Jeff could also give you an account on www.user-mode-linux.org (the 
host is actually jdike.stearns.org, so CC William Stearns on the request) 
like for my page...

Just don't think I will put it directly on my page - it's just to avoid more 
pain for me (updating the site sucks horribly).
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Blaisorblade]

On Sunday 05 December 2004 21:20, Steve Schmidtke wrote:
> Michael Richardson wrote:
> >  uml_netjig is a tool which does the following:
> >   1) moves packets using "daemon" mode a la uml_switch.
> >  2) has an expect-friendly command line interface  3) can create
> >*multiple* switches and keep them seperate

> If they are separate, why not just run multiple uml_switches?

Probably easier to manage... simplicity is good, but not as important, or 
important in the same way, for userspace problems. Don't apply on uml_switch 
the same inclusion criteria you use for the kernel.

> >  4) can record all packets to a .pcap file
> >  5) can play packets from a .pcap file at a given rate     (or under
> >command line interface control)
> >  6) can optionally answer ARP queries for imaginary nodes.

> I like these 3, but could you not just connect another UML guest to the
> switch and run arp -s pub and tcpdump to accomplish the same thing?

Needing another UML guest to do something which can be accomplished in a 
simpler way (supposing it's not too badly coded / heavy / intrusive in the 
patch) could even be reported as a bug, on the reverse!

Another problem: it's doable even without support for "pcap transport" in UML 
(which is currently broken in 2.6)? I know what to do to fix that (build 
problem only), only it's a bit ugly to fix - or requires pcap maintainers to 
fix up some screwage:

a symbol which *should* be static is global, in the static libraries it gets 
obviously exported, it's called vmap, conflicts with the kernel functions 
with the same name, and since it's also a common symbol (i.e. it can share 
the space of the function, see LD manual) it crashes UML silently, by trying 
to overwrite the vmap() code when you use it - funny breakage, isn't it? 
Would you imagine somebody to be able to screw up things so much? And is 
there a polite way to get it fixed? Yes, but needs time to settle down

I've recently learned to fix that with objcopy --keep-global, just haven't 
done it yet).

> >  7) can talk to tuntap device(s) as uml_switch could.
> >
> >  My patches refactor uml_switch to get uml_netjig + uml_switch'

> I am still not clear if uml_netjig is a separate program running on top of
> uml_switch or inside it, but (and I apologize for being dense if this is
> what you are doing) I think the proper way to accomplish what you want is
> to create a "-monitor <socket>" option to uml switch to filter all network
> traffic through.  Then create a monitor program that can attach/detach from
> a uml_switch at will through the monitor socket.  A hook to filter packets
> entering uml_switch and a new function to inject packets (with/without
> updating MAC tables) might be all that's necessary.  All your traffic
> management and expect code could reside in the external monitor program.
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Blaisorblade" == Blaisorblade  <blaisorblade_spam@yahoo.it> writes:
    Blaisorblade> Needing another UML guest to do something which can be
    Blaisorblade> accomplished in a simpler way (supposing it's not too
    Blaisorblade> badly coded / heavy / intrusive in the patch) could
    Blaisorblade> even be reported as a bug, on the reverse!

  Particularly, if you are debugging the ARP code *in* the kernel :-)

    Blaisorblade> Another problem: it's doable even without support for
    Blaisorblade> "pcap transport" in UML (which is currently broken in
    Blaisorblade> 2.6)? I know what to do to fix that (build problem
    Blaisorblade> only), only it's a bit ugly to fix - or requires pcap
    Blaisorblade> maintainers to fix up some screwage:

  I don't use the pcap transport.

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbcu6YqHRg3pndX9AQGLrgP8CsnnEaRDDkZPI7yzSv1qBsRjlgoFe275
beWCJRq4soxm+u1KhWxxthgV6teI/nLz8sR7u56YmDENcxlUjfDoBJIq0qkLyuQs
v6C2bA6LazlNFulFBvL4dCNeMnsyzvZTzudh/k3NvRbSItKtoS32GvhtPLsCcq5H
qfZbVH5wnm0=
=5jVX
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] What to do with uml_switch?

[Blaisorblade]

It is apparent that something is wrong with uml_switch (though I don't know 
the code and I'm not going to express technical opinion, only "management" 
one).

I've just seen about a dozen patches to Steve Schmidtke, two different 
rewrites from Gerd and Felix, and a request from Michael to merge some code 
(uml_netjig) from OpenSWAN/FreeSWAN (IIRC) into it; and it could be something 
very interesting for SuSE customers, I think - I only know that Open/FreeSWAN 
it's about VPN, nothing more - so maybe Gerd could be interested in this.

Michael, could you start posting it here, instead that to Jeff (maybe you did 
and I didn't notice, if it was time ago, but now it's the moment to repost 
it).

Felix, even if your newer code is not yet "releasable", it can help anyway... 
even just describing the changes you did and why... design ideas are more 
important than code itself sometimes.

Steve, merging at least the earlier of your patches (they are splitup The 
Right Way(tm)) might be done soon, without lots of discussions.

If you feel so, you might split uml_switch from the rest (but remember the 
user point of view - two packages doubles the setup time).

Anyway rename the uml_router folder to uml_switch, please... time to fix it!

Clearly, Jeff is current working on completely different things, and it's a 
right thing for him to do - he is not the "I'll do everything about UML" boy, 
and time is short.

It is not even the "UML dictator". So I asked him if someone else could take 
over the maintainance of uml_switch, and he basically agreed.

The only doubt is:

1) he wants to appoint somebody he really trusts as the uml_switch maintainer.

2) I think that, within the UML community, there are enough experienced 
developers that can work together and get it right themselves, even deciding 
for a CVS development model, even because userspace code has different 
stability requirements from kernelspace one (not that uml_switch can be 
faulty, just it's easier to make it correct than with UML).

This, in the most abstract form, is a well-known debate - think to "The 
Cathedral and The Bazaar" by Eric S. Raymond.

I was simply astonished when I read "Think to users as co-developers". I've 
never been against that, but reading it so clearly explained made me think.

Now, I've not started distributing SKAS patches because someone told me to do 
that. Ingo Molnar was working on it on incompatible ways, so I took his work 
and made it backwards compatible, did some other cleanups and published it 
all over. Then it become more and more serious... and one week ago Jeff said 
"I've just been happy with you taking over the maintainership over it".

If you are good guys (and you are), I think you can sort out by discussing it 
here (I'm just against the creation of a separate ML for that development - 
things don't work that way).

So, the ball is over to you. In this moment, Gerd Knorr seems the one who has 
the time and the position (working in SuSE) to manage it well (he can push 
updates to its users and so on, for instance) but it's just my little 2 
cents.

Or you might use our wiki (http://uml.harlowhill.com/) as distribution 
center... how much of you didn't ever know it exists? Well, it being so 
unknown is the biggest bug we experience now.
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Blaisorblade" == Blaisorblade  <blaisorblade_spam@yahoo.it> writes:
    Blaisorblade> Michael, could you start posting it here, instead that
    Blaisorblade> to Jeff (maybe you did and I didn't notice, if it was
    Blaisorblade> time ago, but now it's the moment to repost it).

  I can repost it if you like. 
  Give me a day to find my patches again.

  uml_netjig is a tool which does the following:
  
  1) moves packets using "daemon" mode a la uml_switch.
  2) has an expect-friendly command line interface 
  3) can create *multiple* switches and keep them seperate
  4) can record all packets to a .pcap file
  5) can play packets from a .pcap file at a given rate 
     (or under command line interface control)
  6) can optionally answer ARP queries for imaginary nodes.
  7) can talk to tuntap device(s) as uml_switch could.

  My patches refactor uml_switch to get uml_netjig + uml_switch'

    Blaisorblade> Felix, even if your newer code is not yet
    Blaisorblade> "releasable", it can help anyway... even just
    Blaisorblade> describing the changes you did and why... design ideas
    Blaisorblade> are more important than code itself sometimes.

  Please tell me what your code does...

    Blaisorblade> Steve, merging at least the earlier of your patches
    Blaisorblade> (they are splitup The Right Way(tm)) might be done
    Blaisorblade> soon, without lots of discussions.

  I have a lot of refactoring patches.
  I am willing to spend the time getting consensus among people who
have done other things, and integrating their patches with mine.

  See http://www.sandelman.ottawa.on.ca/SSW/freeswan/fsumltesting/
for an OLS paper on how uml_netjig was (and continues to be) used.
  We invoke about 170 tests every night on the Openswan code base.
some use as many as 7 UMLs connected via uml_netjig to get the job
done. 

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbKDsoqHRg3pndX9AQHAlQP8COxACR4RJFG6lvmhyGTwU0AVi0Xnm53k
1fVm6EfEhXid/vde5RbzM1aoKqA9fCOWJ5GyCVZltNxr1S8P9nuHb9UK7BkNQUKd
ezbjPZ55irSn+yonyOwT/ljHbzh/lFd5kxO5BmregjoRZdOUtqXfHAqQupX6lBix
XJiG4UVo5pk=
=wYSb
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: What to do with uml_switch?

[Gerd Knorr]

>   uml_netjig is a tool which does the following:
>   
>   1) moves packets using "daemon" mode a la uml_switch.
>   2) has an expect-friendly command line interface 

Sounds ok.

>   3) can create *multiple* switches and keep them seperate

Hmm, not sure I like that idea.  Is this with vlan tagging support?
How can this be configured?  Is the command mode the only way, or can
you also use some config file or command line switches for that?

>   4) can record all packets to a .pcap file
>   5) can play packets from a .pcap file at a given rate 
>      (or under command line interface control)
>   6) can optionally answer ARP queries for imaginary nodes.

Hmm, I don't think it is a good idea to put that kind of testing code
into the switch daemon.  I'd make the switch daemon behave as close as
possible to a real switch, then use the normal tools you use for real
networks as well.  I think uml_switch should just support a monitor port
which will see all traffic (even the switched one), that is the usual
way how it's done on physical hardware.

(4) may make sense neverless, to avoid packages being lost on load
peaks.

(5) + (6) I'd keep separated for sure.  Can be done by either some
separate utility which connects to the switch daemon the same way the
uml machines do that, and then injects the packets / answers arp and so
on (probably the better option for test runs).  Or just connect the
switch daemon via tuntap device and setup proxy-arp on the host machine
to simulate the imaginary nodes, and also inject packets via tuntap
network interface.

  Gerd

-- 
#define printk(args...) fprintf(stderr, ## args)



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: What to do with uml_switch?

[Steve Schmidtke]

Gerd Knorr wrote:
> >   3) can create *multiple* switches and keep them seperate
>
>Hmm, not sure I like that idea.  Is this with vlan tagging support?
>How can this be configured?  Is the command mode the only way, or can
>you also use some config file or command line switches for that?

I thought about adding vlan support, but then realized if I wasn't going to 
code up some sort of multi-membership system (union ports?), it would be 
simpler just to run multiple uml_switches.  It should be obvious I don't 
know what I'm talking about, so I left it alone. :)

> >   4) can record all packets to a .pcap file
> >   5) can play packets from a .pcap file at a given rate
> >      (or under command line interface control)
> >   6) can optionally answer ARP queries for imaginary nodes.
>
>Hmm, I don't think it is a good idea to put that kind of testing code
>into the switch daemon.  I'd make the switch daemon behave as close as
>possible to a real switch, then use the normal tools you use for real
>networks as well.  I think uml_switch should just support a monitor port
>which will see all traffic (even the switched one), that is the usual
>way how it's done on physical hardware.

I'm eating my own dog food (since I too suggested a monitor port), and I've 
coded up a framework for doing just that.  I need to test it more before 
submitting a patch for review and comment.

Steve Schmidtke




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Gerd" == Gerd Knorr <kraxel@bytesex.org> writes:
    >> 3) can create *multiple* switches and keep them seperate

    Gerd> Hmm, not sure I like that idea.  Is this with vlan tagging
    Gerd> support?  How can this be configured?  Is the command mode the
    Gerd> only way, or can you also use some config file or command line
    Gerd> switches for that?

  No VLANs. 
  It creates multiple sockets, and you tell the UML(s) about the multiple
interfaces. 

  Please read the mentione paper.

    Gerd> Hmm, I don't think it is a good idea to put that kind of
    Gerd> testing code into the switch daemon.  I'd make the switch
    Gerd> daemon behave as close as possible to a real switch, then use
    Gerd> the normal tools you use for real networks as well.  I think
    Gerd> uml_switch should just support a monitor port which will see

  Not practical, not reliable.
  Often, not even possible if you wish to play packets that Linux simply
can not easily produce. 

  Secondly, I do not put these things into uml_switch.
  Rather, I *refactor* things so that uml_switch and uml_netjig share
back end code. uml_switch has none of the above things, only the same
port.c code.

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbcxHYqHRg3pndX9AQG3/gQAx6XsUJv3AcpqG/c7ZWPFJzZZhXILAmS4
VT1mpY3GS2BZNFEiPhaxoA9tEb1vxWiJtWhe5yGxa/oh0iy4ESp22gPlQDN/p4Iv
bwaW+CxJHE6tVi61ybdaI/9qVrWxBlzZ0CVVuTZBNuonYRQwIn15hn1JlOcahxGB
ov0jytcoSQE=
=Fj0A
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Gerd Knorr]

>     Gerd> Hmm, not sure I like that idea.  Is this with vlan tagging
>     Gerd> support?  How can this be configured?  Is the command mode the
>     Gerd> only way, or can you also use some config file or command line
>     Gerd> switches for that?
> 
>   No VLANs. 

Would be cool I think, so you could debug vlan code in the kernel using
uml for example.

>   It creates multiple sockets, and you tell the UML(s) about the multiple
> interfaces. 

That is clear.  How to you tell uml_switch which sockets it should
create?

>   Please read the mentione paper.

I've looked at it, it doesn't answer the configuration question through.
Your openswan test setup seems to use command mode.  It's not clear
whenever that is the only way to do it.  It shouldn't, for running tests
that likely is perfect, but for other applications I just want a static
configuration and uml_switch shouldn't wait for commands on stdin/out.

>   Secondly, I do not put these things into uml_switch.
>   Rather, I *refactor* things so that uml_switch and uml_netjig share
> back end code. uml_switch has none of the above things, only the same
> port.c code.

Oh, so for uml_netjig you just link in some more some object files which
do the arp answering, packet injection and so on?  That wasn't clear to
me.  This makes much more sense ...

  Gerd

-- 
#define printk(args...) fprintf(stderr, ## args)


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Gerd" == Gerd Knorr <kraxel@bytesex.org> writes:
    Gerd> Hmm, not sure I like that idea.  Is this with vlan tagging
    Gerd> support?  How can this be configured?  Is the command mode the
    Gerd> only way, or can you also use some config file or command line
    Gerd> switches for that?
    >> No VLANs.

    Gerd> Would be cool I think, so you could debug vlan code in the
    Gerd> kernel using uml for example.

  Having uml_switch decode vlan tags, and split it up into multiple
logical switches, would certainly be possible.

  The log-to-pcap file might also be interesting if it could log
specific tags to specific files. That would certainly permit very easy
regression testing.
  (And you don't want to use the VLAN code in the kernel to test the
vlan code in the kernel...)

    >> Secondly, I do not put these things into uml_switch.  Rather, I
    >> *refactor* things so that uml_switch and uml_netjig share back
    >> end code. uml_switch has none of the above things, only the same
    >> port.c code.

    Gerd> Oh, so for uml_netjig you just link in some more some object
    Gerd> files which do the arp answering, packet injection and so on?
    Gerd> That wasn't clear to me.  This makes much more sense ...

  Yes.
  The idea is that uml_switch doesn't change. 
  Adding a bazillion command line switches seemed to be a loss to me.

  No doubt some have gotten bit by being unable to use mcast between
UMLs while, on say, a train. lo isn't multicast capable (at least on 2.4
host, I don't know about 2.6), and sticking in a network card just to
get multicast seems silly.
  We have a script called "localswitches", which invokes uml_netjig
under expect, and then invokes the appropriate UMLs, with the right
environment variables. The environment variables are clearly a bit
locally relevant, since they get plugged into the UML command line(s).

marajade-[nightly/testing/pluto/basic-pluto-06] mcr 1027 %localswitches east west

...

spawn /mara4/openswan-2/nightly/testing/utils/uml_netjig/uml_netjig --cmdproto -t 
OK netjig>NEWSWITCH north
OK 3 LINES
ARPREPLY=0
UML_north_CTL=/tmp/umlyntoYX.d/north/ctl
UML_north_DATA=/tmp/umlyntoYX.d/north/data
OK netjig>NEWSWITCH south
OK 3 LINES
ARPREPLY=0
UML_south_CTL=/tmp/umlyntoYX.d/south/ctl
UML_south_DATA=/tmp/umlyntoYX.d/south/data
OK netjig>NEWSWITCH northpublic
OK 3 LINES
ARPREPLY=0
UML_northpublic_CTL=/tmp/umlyntoYX.d/northpublic/ctl
UML_northpublic_DATA=/tmp/umlyntoYX.d/northpublic/data
OK netjig>NEWSWITCH southpublic
OK 3 LINES
ARPREPLY=0
UML_southpublic_CTL=/tmp/umlyntoYX.d/southpublic/ctl
UML_southpublic_DATA=/tmp/umlyntoYX.d/southpublic/data
OK netjig>NEWSWITCH east
OK 3 LINES
ARPREPLY=0
UML_east_CTL=/tmp/umlyntoYX.d/east/ctl
UML_east_DATA=/tmp/umlyntoYX.d/east/data
OK netjig>NEWSWITCH west
OK 3 LINES
ARPREPLY=0
UML_west_CTL=/tmp/umlyntoYX.d/west/ctl
UML_west_DATA=/tmp/umlyntoYX.d/west/data
OK netjig>NEWSWITCH public
OK 3 LINES
ARPREPLY=0
UML_public_CTL=/tmp/umlyntoYX.d/public/ctl
UML_public_DATA=/tmp/umlyntoYX.d/public/data
OK netjig>NEWSWITCH admin
OK 3 LINES
ARPREPLY=0
UML_admin_CTL=/tmp/umlyntoYX.d/admin/ctl
UML_admin_DATA=/tmp/umlyntoYX.d/admin/data
OK netjig>


- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbfFEIqHRg3pndX9AQFTKQP+I29St4Vr/2jOOLtLjRKCsULLF13q+IZR
Nws422J1sXbdmTsSxBEbjS0zhbRiXYqZmP8Q11G4oHxEC+zz8JPuYgU2XsfEnKhk
oDsuk89evlOgRkx3zo3S+iIbCA1gikdiiB7z3XVpGn31qmon6IJNuXlamoqjzjxd
9jAM9lu8r9A=
=c2Fv
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Steve Schmidtke]

Michael Richardson wrote:
>     Gerd> Hmm, I don't think it is a good idea to put that kind of
>     Gerd> testing code into the switch daemon.  I'd make the switch
>     Gerd> daemon behave as close as possible to a real switch, then use
>     Gerd> the normal tools you use for real networks as well.  I think
>     Gerd> uml_switch should just support a monitor port which will see
>
>   Not practical, not reliable.

Why not reliable?  A monitor port is a diagnostic tool, and there is no 
reason you couldn't run it in a sychronous mode with the switch, i.e. the 
switch is paused as the monitor inspects an IO event (an input packet) and 
either allows it to proceed or be modified in some way (replaced, dropped, 
etc).

I will agree that it wouldn't be fit for all purposes.  There is obviously a 
delay as a packet must be written to the monitor app for inspection, a 
context switch to the monitor, an ACK written from the monitor, a context 
switch back, and the ACK processed, before a packet can be disposed of.  
Does that make a monitor port impractical?  I don't think so.  The real 
world delay I've been seeing so far on my test monitor doing just that for 
ping round trip times is less than a millisecond.  I havn't tried larger 
packets yet.

On the other hand, once the monitor has ACK'd a packet, the switch and 
monitor app are free to continue processing independently, possibly reducing 
overall delay times.

>   Often, not even possible if you wish to play packets that Linux simply
>can not easily produce.

There is no reason you would need to use Linux as your packet source:

  cat tcpdump.pcap | switch_mon_tool --socket /tmp/uml.mon --delay 100ms

is the type of functionality I'm shooting for (above example not to be taken 
literally).

Steve Schmidtke




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Steve" == Steve Schmidtke <steve_schmidtke@hotmail.com> writes:
    Gerd> Hmm, I don't think it is a good idea to put that kind of
    Gerd> testing code into the switch daemon.  I'd make the switch
    Gerd> daemon behave as close as possible to a real switch, then use
    Gerd> the normal tools you use for real networks as well.  I think
    Gerd> uml_switch should just support a monitor port which will see

    >> Not practical, not reliable.

    Steve> Why not reliable?  A monitor port is a diagnostic tool, and

  What would I connect to the monitor port? Another UML? 

  This is the part that isn't practical.
  My test cases sometimes have as many as 7 networks. Remember that we
have 170 test cases, and growing. Remember that we may have packets akin
to "ping-of-death", that may cause the Linux network stack to "oops".
  (this isn't a good, but that's the point of regression tests... is to
make sure that old bugs do not get re-introduced)

  So, we have to start the UMLs for each test, as they may have died.
To make it simple, we start/stop all of the UMLs each time.

  The reason for having the ability to capture and play-back pcap files
in the uml_netjig is to eliminate the need to run as many UMLs.

    Steve> There is no reason you would need to use Linux as your packet
    Steve> source:

    Steve>   cat tcpdump.pcap | switch_mon_tool --socket /tmp/uml.mon
    Steve> --delay 100ms

    Steve> is the type of functionality I'm shooting for (above example
    Steve> not to be taken literally).

  Great. That works for 1 network.

  Now, how do you replay packets where you need an actual time sequence
diagram to make it work? I.e.
	   - send packet A on network 1
	   - send packet B on network 2
	   - send packet C on network 1
	   - send packet D on network 2

  In most cases, we have found round-robin amount input sources works
well, but in that we are disappointed with. 

  Further, remember that you may want to single step the packet sources,
because, you may spend *HOURS* in the debugger (with the UML in
question). That's hard with the method you proposed.

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [






-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbd0NIqHRg3pndX9AQEZFgQA5vDKTWQviO9NSLr9Y6Pkvq23KQbWPop2
GyR04q/ALPlMSbdPqsJfHFsqCXLZF/gmftZH6Cp52XpXX4AO7+7x47aUDG7SW2ug
WxYjBJVqHBgBSJQNHkpcXwswqJPV4fidMyiN0X60PtmW+o6hMa/NXr9LPFcElwVw
FiEFd3ONQ7U=
=F82m
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Steve Schmidtke]

Michael Richardson wrote:
>     >> Not practical, not reliable.
>
>     Steve> Why not reliable?  A monitor port is a diagnostic tool, and
>
>   What would I connect to the monitor port? Another UML?

A program that can inspect and/or create packets --  uml_netjig but with a 
modular API.

>     Steve>   cat tcpdump.pcap | switch_mon_tool --socket /tmp/uml.mon
>     Steve> --delay 100ms
>
>     Steve> is the type of functionality I'm shooting for (above example
>     Steve> not to be taken literally).
>
>   Great. That works for 1 network.
>
>   Now, how do you replay packets where you need an actual time sequence
>diagram to make it work? I.e.
>	   - send packet A on network 1
>	   - send packet B on network 2
>	   - send packet C on network 1
>	   - send packet D on network 2

Thinking out loud, I envision somone being able to write a monitor app that 
would allow a session like this:

  host:$> vmon_switch -i
  Welcome to vapour_monitor version -0.01
  Interactive mode ready.
  > verbose
  verbose mode on.
  >  connect swA /tmp/uml.mon1
  swA connected ok.
  > connect swB /tmp/uml.mon2
  swB connected ok.
  > connect swC /tmp/uml.mo3
  swC connect failed: /tmp/uml.mo3
  > connect swC /tmp/uml.mon3
  swC connected ok.
  > pcap source infile tcpdump.pcap
  infile opened for reading ok.  infile record: 1
  > pcap dest outfile traffic.pcap
  outfile opened for writing ok.
  > pcap dump swA outfile
  dumping switch swA, sending to outfile.
  > pcap single input swA
  1 packets to switch swA ok.  infile record 2.
  > pcap single input swB
  1 packets to switch swB ok.  infile record 3.
  > pcap close outfile
  outfile closed, 10 packets.
  > pcap rewind infile
  infile record 1.
  > pcap single infile swC
  1 packets to switch swC ok.  infile record 2.
  ** EOF from swA **
  > pcap delay 100 ms
  send interval set to 100 milliseconds
  > pcap send infile swC
  ..... 5 packets to switch swC ok.  input position 7.
  > quit
  vapour_monitor Exiting...
  host:$> tcpdump -n -r traffic.pcap

I put way too much time and effort into that example, but it was useful to 
think through.  The point is, a port allows the complexity of diagnostic 
code and traffic analysis to reside externally in the monitor app, and not 
be coded in the switch (the flip side is that the monitor apps don't have to 
know how the switch works).  If "switch_mon_tool" is unsuitable, it can be 
replaced by "vmon_switch" in a heartbeat -- without modifying or even 
restarting uml_switch.

Steve Schmidtke




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: What to do with uml_switch?

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Steve" == Steve Schmidtke <steve_schmidtke@hotmail.com> writes:
    Steve> Thinking out loud, I envision somone being able to write a
    Steve> monitor app that would allow a session like this:

  Okay, this is done already :-)

- -- 
]       ON HUMILITY: to err is human. To moo, bovine.           |  firewalls  [
]   Michael Richardson,    Xelerance Corporation, Ottawa, ON    |net architect[
] mcr@xelerance.com      http://www.sandelman.ottawa.on.ca/mcr/ |device driver[
] panic("Just another Debian GNU/Linux using, kernel hacking, security guy"); [


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys

iQCVAwUBQbeh+oqHRg3pndX9AQGdqgQArmXb6Ha863CRyUfrgJPdK3LLBRfNfRmS
1dnfbu5GJbv9RoLtgazQ9PYA9JKqO9jjrMdrkze+zsxwztprMlQSDBWaLtotW5vW
KUFRRCc09xUXszVBWG4F+DdryaclRqhiYVLPj3IFPMm+7boWuYqCZVdhHngD/VMm
ItoUPSb1QGw=
=O5Yg
-----END PGP SIGNATURE-----


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] RE: What to do with uml_switch?

[Steve Schmidtke]

Blaisorblade wrote:
>It is apparent that something is wrong with uml_switch (though I don't know
>the code and I'm not going to express technical opinion, only "management"
>one).

Yes, the fact that a core file has sat in the directory for almost 2 years 
says something. :)

>Steve, merging at least the earlier of your patches (they are splitup The
>Right Way(tm)) might be done soon, without lots of discussions.

Thanks.  I havn't noticed any problems with the changes I've made so far.

>If you feel so, you might split uml_switch from the rest (but remember the
>user point of view - two packages doubles the setup time).

Who is this directed to?  Who is the maintainer at the moment anyways?

>If you are good guys (and you are), I think you can sort out by discussing 
>it
>here (I'm just against the creation of a separate ML for that development -
>things don't work that way).
>
>So, the ball is over to you. In this moment, Gerd Knorr seems the one who 
>has
>the time and the position (working in SuSE) to manage it well (he can push
>updates to its users and so on, for instance) but it's just my little 2
>cents.

I'd be happy with Gerd as maintainer.  I'd nominate myself, but I have no 
system that I can dedicate to the project.  I also have a hard time saying 
no to interesting ideas.

Steve Schmidtke




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: What to do with uml_switch?

[Gerd Knorr]

> >So, the ball is over to you. In this moment, Gerd Knorr seems the one
> >who has the time and the position (working in SuSE) to manage it well
> >(he can push updates to its users and so on, for instance) but it's
> >just my little 2 cents.

I wish I had more time for uml.  uml isn't the only project I'm working
on, and in the end it's just shifting priorities.  Time I spend on UML I
have to take away somewhere else ...

> I'd nominate myself,

Just go ahead ;)

  Gerd

-- 
#define printk(args...) fprintf(stderr, ## args)


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] RE: What to do with uml_switch?

[Blaisorblade]

On Sunday 05 December 2004 21:00, Steve Schmidtke wrote:
> Blaisorblade wrote:
> >It is apparent that something is wrong with uml_switch (though I don't
> > know the code and I'm not going to express technical opinion, only
> > "management" one).

> Yes, the fact that a core file has sat in the directory for almost 2 years
> says something. :)

> >Steve, merging at least the earlier of your patches (they are splitup The
> >Right Way(tm)) might be done soon, without lots of discussions.

> Thanks.  I havn't noticed any problems with the changes I've made so far.

> >If you feel so, you might split uml_switch from the rest (but remember the
> >user point of view - two packages doubles the setup time).

> Who is this directed to?  Who is the maintainer at the moment anyways?

It's (un)maintained by Jeff Dike, i.e. he's the last one updating the version 
on sf.net (you might call that "official version" - but I'm against the 
"official version" concept). Actually, you can guess he has no time to do 
anything else than bugfixes on it. We haven't yet managed to build one single 
2.4 up-to-date UML tree (i.e. with both *all* security fixes and the work 
against 2.6.9 host). Currently probably the stablest version seems to be one 
of the 2.6.9 incarnations (either it or -bb4).

> >If you are good guys (and you are), I think you can sort out by discussing
> >it
> >here (I'm just against the creation of a separate ML for that development
> > - things don't work that way).

> >So, the ball is over to you. In this moment, Gerd Knorr seems the one who
> >has
> >the time and the position (working in SuSE) to manage it well (he can push
> >updates to its users and so on, for instance) but it's just my little 2
> >cents.

> I'd be happy with Gerd as maintainer.  I'd nominate myself, but I have no
> system that I can dedicate to the project.  I also have a hard time saying
> no to interesting ideas.

Which is bad about stability... however you share that with Jeff Dike, it 
seems :-)

> Steve Schmidtke

-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel