--- Day changed

--- Log opened Thu Aug 07 00:00:01 2003
01:54 -!- silug [~steve@64.240.156.225] has joined #uml
02:16 -!- mdz_ [~mdz@h00034795fb2d.ne.client2.attbi.com] has quit [Read error: Connection timed out]
02:21 -!- mdz_ [~mdz@h00034795fb2d.ne.client2.attbi.com] has joined #uml
02:39 -!- pirlouit [~peter@duvel.drunkcoders.com] has joined #uml
03:16 -!- srinath [~srinath@61.95.225.98] has joined #uml
03:16 -!- srinath [~srinath@61.95.225.98] has left #uml []
03:30 -!- G2_ [~G2@193.195.148.66] has joined #uml
03:31 < magnet> hello.
03:31 < magnet> I'm looking for some help on how to setup my network.
03:47 < G2_> Morning
03:48 < magnet> hello.
03:48 < G2_> Go ahead
03:49 < magnet> good.
03:49 < magnet> I tried they way in the howto,
03:49 < magnet> but I always get :
03:49 < magnet> root@(none):~# /sbin/ifconfig eth0 192.168.0.5 up
03:49 < magnet> SIOCSIFFLAGS: Invalid argument
03:51 < G2_> that's fine, if you ifconfig again it should have the ip address in there
03:52 < magnet> umm.hang on,I'm trying.
03:54 < magnet> I think I'm not getting the kernel's boot parameter thing.
04:03 < magnet> should I assign the same ip to /dev/tap0 on my host system ?
04:03 < magnet> rather than on the guest ?
04:16 < G2_> did you do ubd0=root_fs tuntap=eth0,,,192.168.0.5 
04:17 < magnet> yes.
04:17 < G2_> don't know then
04:17 < magnet> I use :
04:17 < magnet>  linux devfs=mount eth0=ethertap,tap0,fe:fd:0:0:0:1,192.168.0.25
04:17 < caker> magnet: on the host, set the tap as the same IP as your host's eth0
04:17 < magnet> caker I try.
04:17 < caker> magnet: feed that same IP into your ./linux command line
04:18 < caker> then boot it up, and inside, use a DIFFERENT ip :-)
04:18 < caker> you'll have to add routes and a bunch of other stuff .. it's a pain.  I recommend using a bridge on the host
04:19 < magnet> wait a min , brb.
04:37 -!- G2_ [~G2@193.195.148.66] has quit [Quit: Client exiting]
05:04 -!- G2 [~G2@193.195.148.66] has joined #uml
05:08 < magnet> caker: are you still here ?
05:33 -!- ticallion [~ticallion@213.175.160.36] has joined #uml
06:19 < wojci> magnet: You can setup networking on the host before starting your UML. It is also described in the howto. 
06:20 < wojci> magnet: I can spoonfeed you with the commands I use, if you like ;)
06:25 < ticallion> what turned me on was umm tun/tap, didn't like the bridging thing in david's tut
06:49 < magnet> wojci: that would be really cool.
06:49 < magnet> I'm going to lunch now, I'll be back a bit later,you can mail me at magnet@sitadelle.com
06:49 < magnet> or pvt me :)
06:53 < wojci> magnet: Give me a minute ..
06:55 < wojci> I wrote some perl to setup UMLs on my system - I can get the script to output commands to configure the host/network inside the UML :)
06:55 < wojci> For the host:
06:56 < wojci> root@umlbox:~# uml_cntrl.pl -pretend -start 0
06:56 < wojci>  /pack/UML-tools/bin/tunctl -b -u uml0 -t tap0
06:56 < wojci>  /sbin/ifconfig tap0 arp mtu 1500 10.0.0.50 up
06:56 < wojci>  /sbin/route add -host 10.1.0.50 dev tap0
06:56 < wojci>  /usr/sbin/arp -Ds 10.1.0.50 tap0 pub
06:56 < wojci>  /usr/bin/screen -S box0:BOXX0 -d -m /bin/su - uml0 -c "/pack/UML-root/kernel/linux-2.4.20/linux-3 tty_log_fd=3 ubd0=/pack/UML-root/fs/0/Debian-3.0r0.ext2 ubd1=/pack/UML-root/fs/0/swap.img root=/dev/ubd/0 eth0=tuntap,tap0,fe:fd:0:0:0:50,10.0.0.50 mem=64M umid=0 uml_dir=/pack/UML-root/var/run 3>>/pack/UML-root/var/logs/0
06:56 < wojci>  .ttylog.log"
06:56 < wojci> For networking inside the UML box:
06:57 < wojci> root@umlbox:~# uml_cntrl.pl -vmconfig 0
06:57 < wojci> # Interface:
06:57 < wojci> ifconfig eth0 10.1.0.50 netmask 255.255.255.0 up
06:57 < wojci> # Routes:
06:57 < wojci> route add -net 10.0.0.0 netmask 255.255.255.0 dev eth0
06:57 < wojci> route add default gw 10.0.0.50
06:57 < wojci> That is all ;)
06:59 < wojci> I also use iptables to forward packets between tap0 and the host.
07:00 < wojci> I hope that this helps you in some way - I only used information contained in the howto to write my script/setup the UML boxen.
07:07 -!- ticallion [~ticallion@213.175.160.36] has quit [Ping timeout: 490 seconds]
07:17 -!- G2 [~G2@193.195.148.66] has quit [Quit: Client exiting]
07:37 -!- G2_ [~G2@193.195.148.66] has joined #uml
08:37 -!- ghimlet^ [~hy@80.116.23.134] has joined #uml
08:53 < magnet> thanks a lot wojci , I'm gonna try that.
09:15 < magnet> wojci: GREAT this seems to works.
09:28 < magnet> now I need to understand what I did lol.
09:31 -!- jdike [~jdike@129.170.249.155] has joined #uml
09:31 < jdike> hi guys
09:35 < magnet> hi.
09:35 < green> Hi Jeff!
09:35 < green> jdike: I see that memleak fixes I sent you some time ago were magically lost somehow
09:36 < jdike> green: hmmm
09:36 < jdike> green: did I apply them?
09:36 < jdike> green: I vaguely recall that I did
09:36 < green> Well, I sent those to you and and you said they are ok so you will apply.
09:37 < green> btw, I have some microoptimisations patch to pass to you ;) Replaces "current->task_info" with "current_task" ;)
09:37 < jdike> green: I've still got that patch in my todo folder
09:38 < jdike> green: (the memleak one)
09:38 < david-work> hello
09:38 < jdike> hi david
09:38 < green> hi david
09:38 < jdike> green: yeah, those need doing
09:39 < paq_> hey guys
09:39  * green was trying to find task_info leak for most of the day without much success :(
09:39 < green> task_struct leak, even
09:39 < jdike> green: I looked for that, without success
09:40 < jdike> green: the process exits with an extra reference on its task_struct
09:40 < jdike> green: and all the references and dereferences are in generic code iirc
09:41 < green> jdike: funny thing is that ther is not always this extra reference
09:41 < jdike> green: no?
09:41 < jdike> green: and those task_structs get freed properly?
09:42 < green> e.g. I tried to do forceful decrement of usage in release_thread() and got slab corruption warning, the location where "usage" is in task_struct was decremented by one
09:44 < green> If I check the value to be > 1, before I decrement, I hit the assertion about "task == current" shound not be true
09:45 < green> Also I checked on real kernel, we seems to always get into release_thread with p->usage of 1. (at least all three times I did this, this was true)
09:45 < green> so osmetimes they are freed and sometimes they are not
09:47 < jdike> s/jobs/lots/
09:47 < green> hm?
09:49 < jdike> green: the last reference is dropped right after the call to release_thread
09:49 < jdike> green: so I'd expect the refcount to be 1 there
09:49 < green> yup, be usually we get there with refcount of 2. (in uml)
09:50 < green> actually I found the comment at the end of dup_task_struct to be strange. They sed refcount to 2, and say that 1 is for release_task() and another one is for them, but I have not found where they decrement it, unfortunatelly
09:59 < green> oh, great. sourceforge still does not accepts mails from me.
09:59 < jdike> green: you never figured out what was happening with that?
10:00 < green> yup
10:11 < green> have you received the patch?
10:12 < jdike> green: you sent something just now?
10:13 < green> no, I sent it like just before I complained about sourceforge
10:13 < jdike> green: OK, haven't checked mail since then, hold on a bit
10:13 < jdike> green: I'm preparing mail to Linus so he can ignore me again
10:14 < jdike> Linus is torvalds@odsl.org now?
10:15 < green> yup
10:25 -!- Jax [~computer_@148.100.212.187] has joined #uml
10:30 < wojci> Anyone noticed this: If I login as $user using the console(trough screen in this case) and then login to the same UML using openssh, everthing I write to ssh is outputted on the console. I tried 2.6.0-test2 where this does not happend - I can only make this "work" in 2.4.20.
10:32 < wojci> At first I thought that the problem was my debian stable image, but the problem disappears when I use a 2.6.
10:33 < attriel> lol.  thats actually kinda amusing :o
10:33 < wojci> It is?
10:33 < attriel> if, perhaps, something to make security people rip their hair out :)
10:33 < wojci> Oh .. great ;)
10:33  * wojci searches for a nice screenshot 
10:34 < wojci> http://wojci.dk/download/UML-strange.jpg 
10:34 < wojci> There you go :)
10:34 < attriel> heh.  i'd guess that screen/etc isn't 'locking' the tty (pty?) and that ssh is ending up with the same one somehow
10:35 < attriel> of course, i have no idea why that'd happen or only happen sometimes :o
10:35 < attriel> does a second ssh also echo to console or just the first?
10:35 < wojci> It happends *every time* in my setup.
10:36 < attriel> weird
10:36 < attriel> i can see that being annoying yah
10:36 < wojci> It would be nice if somebody else could try my setup, and see if the problem appears for them too.
10:37 < jdike> wojci: strange
10:37 < jdike> wojci: what's the console on the left?
10:37 < wojci> jdike: "screen -r -d", uml running inside it.
10:38 < jdike> wojci: so that's the main console?
10:39 < wojci> jdike: Yes.
10:42 < jdike> wojci: do you have tty logging happening?
10:42 < wojci> jdike: Yes. I belive it also "works" without logging enabled.
10:43 < jdike> wojci: can you check?
10:43 < wojci> jdike: Yes. Doing a compile right now.
10:43 < jdike> wojci: coz loggging is the only reason I can see for output to be copied somewhere else like this
10:44 < wojci> jdike: Are you by any chance a UML developer?
10:44 < jdike> wojci: if you log in somewhere else, does that output appear there as well?
10:44 < wojci> jdike: Second thing to check, noted :)
10:45 < jdike> wojci: I am the UML developer
10:45 < wojci> Oh ;)
10:45 < wojci> My kernels are: Linux umlbox 2.4.21 #2 Tue Jun 17 00:52:27 CEST 2003 i686 unknown for the box that runs UMLs.
10:46 < wojci> The UML itself: Linux stable.wojci.dk 2.4.20-5um #12 Sat Aug 2 15:59:02 CEST 2003 i686 unknown
10:50 < jdike> wojci: what's in that window at the top which is also seeing your echo commands?
10:50 < jdike> wojci: looks like syslog
10:50 < wojci> Just an app I wrote.
10:50 < wojci> Reads UML logfiles.
10:52 < jdike> wojci: yeah, just noticed that
10:52 < jdike> wojci: you planning on releasing that sometime? :-)
10:52 < wojci> The perl program you like to is b0rked. It gets confused by the exec records :/
10:52 < wojci> Bah .. ignore that sentence 
10:53 < jdike> wojci: you need a later version
10:53 < wojci> I was talking about the perl program I found linked to on the main UML site.
10:53 < wojci> At the time I tried to use it, there were no later version, so I wrote my own.
10:54 < jdike> wojci: hmmm
10:55 < wojci> I can release it, but it needs some more work. The gui sucks at some points.
10:57 < jdike> wojci: you referring to playlog.pl from the uml_utilities tarball?
10:57 < jdike> green: with a quick look, that patch looks fine
10:58 < wojci> In my last comment I was referring to my own application. In my previous comments I was referring to playlog.pl.
10:58 < jdike> wojci: about being confused by exec records?
10:58 < wojci> Yes.
10:59 < green> jdike: also it applies cleanly to my tree and the resulting kernel boots fine ;)
10:59 < jdike> wojci: playlog has known about exec records for at least two releases afaics
10:59 < wojci> jdike: I could have downloaded some outdated version.
11:03  * jdike send Linus updates to ignore
11:03 < green> hehe.
11:04 < wojci> jdike: The problem disappers when tty_logging is turned off.
11:04 < green> add "[CRITICAL SECURITY UPDATE]" in subject ;)
11:04 < jdike> green: hehe
11:04 < wojci> jdike: I can give you a good config and a bad config ;)
11:04 < jdike> green: that would likely make world headlines
11:05 -!- thy_ [koubt9am@u-pl4.ms.mff.cuni.cz] has joined #uml
11:05 < jdike> green: which isn't the desired result
11:05 < jdike> wojci: what's your UML command line?
11:06 < wojci>  /usr/bin/screen -S box0:BOXX0 -d -m /bin/su - uml0 -c "/pack/UML-root/kernel/linux-2.4.20/linux-without-tty-logging tty_log_fd=3 ubd0=/pack/UML-root/fs/0/Debian-3.0r0.ext2 ubd1=/pack/UML-root/fs/0/swap.img root=/dev/ubd/0 eth0=tuntap,tap0,fe:fd:0:0:0:50,10.0.0.50 mem=64M umid=0 uml_dir=/pack/UML-root/var/run 3>>/pack/UML-root/var/logs/0.ttylog.log"
11:07 < jdike> wojci: that looks OK, you are getting records in your log file when this happens?
11:08 < wojci> jdike: Yes. This command line was for my last try without tty logging enabled.
11:09 < jdike> wojci: and the stuff on the console is the text, not log records...
11:09 < wojci> jdike: Yes, only text.
11:13 < wojci> jdike: Here is the config: http://wojci.dk/download/linux_2.4.20_uml_config_bad
11:14 < wojci> Got that by using "--showconfig".
11:42 < jdike> wojci: doesn't happen for me
11:49 -!- G2_ [~G2@193.195.148.66] has quit [Quit: Client exiting]
11:50 < wojci> jdike: Now .. what do you suggest that I can do to clear up this situation?
12:02 < jdike> wojci: gdb UML to see where those writes are coming from
12:14 < wojci> jdike: OK - I will do that later tonight. 
12:14 < wojci> I prefer to use gvd -;)
12:19  * wojci detaches .. bbl
12:33 -!- ghimlet^ [~hy@80.116.23.134] has quit [Quit: Uscita dal client]
12:45 < miah> \disconnect
12:45 -!- miah [miah@sifl.digivill.net] has quit [Quit: Disconnecting]
13:56 -!- the_p0pe [~rewt@156.34.60.73] has joined #uml
14:23  * jdike makes a new 2.4 UML patch
14:28 < Jax> for 2.4.22-rc1?
14:31 < jdike> Jax: 2.4.20-7
14:38 < Jax> oh
14:40 < Jax> is the um architecture in 2.6? i see that you have patches for 2.6.0-test*, but i also have an article on features saying that "virtualizing linux" is now a feature
14:41 < jdike> Jax: it's in 2.6, just a very old version
14:54 < attriel> jdike: is hostfs still hosed for 2.6, or does it work again?
14:55 < jdike> attriel: haven't tried it recently, but it works afaik
14:57 < attriel> ok; was looking at user-mode-linux.sourceforge and it said it was hosed in 2.5 b/c of a filesystem rewrite somewhere, so figured I'd see if that'd been fixed :o
14:57 < attriel> (I just got it to start booting, yay me :)  so now i'm looking for ways to b reak it :o)
15:02 -!- AquaJo [~torbofh@80.130.240.223] has joined #uml
15:24 < jdike> 2.4.20-7 is out there
15:30 < AquaJo> hi jeff
15:37 -!- jdike [~jdike@129.170.249.155] has quit [Ping timeout: 501 seconds]
16:01 < the_p0pe> i wish i had enough hd space on my laptop to run all teh uml boxen i want to.
16:01 < the_p0pe> i should just buy a usb hdd.
16:01 < AquaJo> *g*
16:02 < the_p0pe> i have like 20uml's runnign in my basement. i emulated my work network so i can telecommute better.
16:02 < the_p0pe> somethign like that anyway. i just wished it worked on ARM :/
16:02 -!- G2 [~gdive23@212.159.59.85] has joined #uml
16:02 < AquaJo> *g*
16:03 < AquaJo> <- is kepping a few more umls running
16:03 < the_p0pe> anyone know about someone pokign at a ARM port ?
16:03 < the_p0pe> i'velooked at it. i can't do that sort of shite.
16:03 < the_p0pe> not yet, anyway
16:03 < AquaJo> neither I can
16:04 < AquaJo> I'd like, but I havn't enouh time to lear it
16:04 < AquaJo> :-(
16:04 < the_p0pe> learn what ? porting ?
16:04 < AquaJo> first of all c
16:04 < the_p0pe> oh.
16:04 < AquaJo> then kernel hacking
16:04 < AquaJo> and then porting uml
16:05 < AquaJo> <- uses pascal ans the bash for nealy all his problems
16:05 < AquaJo> I'd like to learn so much more
16:05 < AquaJo> but I can't buy time
16:08 -!- jdike [~jdike@129.170.249.155] has joined #uml
16:08 < AquaJo> rehi jeff
16:09 < jdike> hi
16:10 < jdike> slight network difficulties here...
16:12 -!- thy_ [koubt9am@u-pl4.ms.mff.cuni.cz] has quit [Quit: Client exiting]
16:31 -!- Jax [~computer_@148.100.212.187] has quit [Quit: going home]
16:51 < caker> hello
16:53 < jdike> caker: hi
16:53 < caker> heya jeff
16:54 < caker> jdike: i have a question about uml_exitcode -- is it *only* ever set through the proc interface?
16:55 < jdike> caker: yes, otherwise UML exits 0
16:56 < jdike> caker: (unless there's some disaster during startup)
16:56 < caker> jdike: ok -- interested in a patch that adds a command line option like "exit with a code on reboot" instead of rebooting?
16:56 -!- Nimaj [~Nimaj@148.100.211.243] has joined #uml
16:56 < caker> that lets whatever is managing the UMLs do stuff between reboots
16:57 < jdike> caker: a "don't reboot on reboot" switch seems kinda perverse
16:58 < caker> jdike: it makes some sense, since I'd like to "trap" reboots and do stuff in the hosts's environment before it execs again...
16:58 < jdike> caker: I've considered a mechanism to stall a UML as it reboots, though
16:58 < caker> jdike: as it is right now, I have reboot patched to poweroff
16:59 < caker> jdike: stall as in just a delay, and perhaps some notification?
16:59 < jdike> caker: would that work?
16:59 < caker> well, the host doesn't know it was a reboot and not a poweroff, hence the exitcode :-)
16:59 < jdike> caker: not just a delay
16:59 < jdike> caker: a notification, plus it won't go anywhere until mconsole says so
17:00 -!- Nimaj is now known as Jax
17:00 < caker> jdike: in my case, some of the UML command line options might/would change between "reboots"
17:01 < jdike> caker: hmmm
17:01 < caker> i guess I could just set uml_exitcode AND call poweroff on reboot ..
17:03 < jdike> caker: at the reboot notification, you could force a halt, I suppose
17:03 < jdike> caker: why are you rebooting with new args?
17:04 < jdike> caker: just curious
17:04 < caker> jdike: for example: someone adds a hard drive to their UML through my web interface .. then they need to "reboot" (i'm not doing any hotplugging yet :-)...
17:05 < jdike> caker: Oh
17:05 -!- G2 [~gdive23@212.159.59.85] has quit [Quit: Client exiting]
17:05 < jdike> caker: you're not going to be rebooting forever, right?
17:05 < caker> jdike: or IP addresses, or change RAM, or ...
17:05 < caker> jdike: sorry?
17:06 < jdike> caker: you're going to have hotplug implemented at some point
17:06 < jdike> caker: RAM is the only tricky case you mentioned
17:06 < caker> jdike: hotplugging via mconsole is very low on my priorities list
17:07 < jdike> caker: too bad, because that would be pretty cool
17:07 < caker> jdike: i agree -- bad ass in fact
17:07 < jdike> caker: how about an mconsole way of changing the arguments?
17:07 < jdike> caker: then it would reboot on the new args
17:08 < caker> Honestly, I only use mconsole for my shutdown (then sync sync halt) procedure
17:08 < caker> I've had trouble with mconsole hanging up .. I believe it's a race condition .. 
17:08 -!- Jax is now known as stbe
17:09 < caker> jdike: i've only hit it twice, but I think if someone "poweroff" their UML, AND initiates my shutdown procedure, something goes wrong
17:09 < jdike> caker: well, tell me how to make it happen, and I'll fix it
17:09 -!- stbe is now known as Jax
17:09 < jdike> caker: the machine hangs, or the mconsole client?
17:10 < jdike> caker: (machine == UML)
17:10 < caker> the mconsole client hangs, and at least one of the UML threads sticks around
17:10 < jdike> caker: send mail to uml-user and I'll add that to my list of things to look at
17:11 < caker> jdike: ok 
17:43 < AquaJo> good night
17:43 -!- jdike [~jdike@129.170.249.155] has quit [Quit: Leaving]
17:46 -!- AquaJo [~torbofh@80.130.240.223] has quit [Quit: Die Stärke eines Menschen mißt man daran, wie er mit seinen Schwächen umgeht.]
18:14 -!- ticallion [~ticallion@213.175.160.8] has joined #uml
19:00 -!- ticallion [~ticallion@213.175.160.8] has quit [Quit: later]
19:20 -!- ichilton [~ian@80.6.255.203] has quit [Ping timeout: 490 seconds]
19:25 -!- ichilton [~ian@pc3-stoc3-4-cust203.midd.cable.ntl.com] has joined #uml
19:35 -!- shpetros [~shpetros@pc-80-194-96-72-hy.blueyonder.co.uk] has joined #uml
19:35 < shpetros> hello all
19:37 < Jax> hi
19:45 -!- shpetros [~shpetros@pc-80-194-96-72-hy.blueyonder.co.uk] has left #uml [Client exiting]
21:32 -!- com4 [~com4@12-209-152-183.client.attbi.com] has joined #uml
21:33 < com4> is there a quick guide of how to setup UML to have the host be the firewall for the uml instances
21:33 < com4> so the uml instances have local ips
21:34 < com4> like, i don't really understand how you'd link the internal with the external, for a normal router, you have 2 nics
21:34 < com4> but i don't understand what the 2nd nic should be
21:34 < Jax> someone said that david's documentation is good for networking, but i haven't gotten that far
21:34 < Jax> http://uml.openconsultancy.com or something like that
21:34 < com4> i looked at his, but i odn't understand what's going on
21:34 < com4> yea
23:24 -!- the_p0pe [~rewt@156.34.60.73] has left #uml []
--- Log closed Fri Aug 08 00:00:00 2003