--- Day changed

--- Log opened Sat Jun 28 00:00:01 2003
02:29 -!- mistral [mistral@jstevenson.plus.com] has joined #uml
04:25 -!- shak [~shak@pc1-hudd4-5-cust78.hudd.cable.ntl.com] has quit [Quit: Client exiting]
04:30 -!- mistral [mistral@jstevenson.plus.com] has quit [Ping timeout: 492 seconds]
04:46 -!- mistral [mistral@jstevenson.plus.com] has joined #uml
05:16  * green slowly starts to hate alpha hardware
05:36 -!- revenger_ [revenger@p508B3973.dip.t-dialin.net] has joined #uml
05:42 -!- Getty [~Getty@spast.de] has joined #uml
06:37 -!- DnsInfector [~DnsInfect@ARennes-204-1-1-200.w193-251.abo.wanadoo.fr] has joined #uml
06:39 < ido> how could i check which socket my UML is using ?
06:40 < ido> i mean, so i could connect to it with mconsole ?
06:44 < green> dmesg from inside it will tell you
06:44 < green> or just assign a permanent name to each uml instance
06:45 < ido> hmm, no, the problem was a space problem
06:45 < ido> i did poitnt it to the real socket
06:46 < green> I did not understand what you have just said ;)
06:48 < ido> when trying to connect to the socket i got an error, so i thought it was a dead socket. apparently, it was the right one, but i didnt have enough free space on the HARDDRIVE,so it gave me the error.
06:48 < ido> (because it couldn't create tmp files)
06:53 < green> ah
08:32 < Lathiat> ido: guess? :)
08:37 < Lathiat> uml_mconsoel halt
08:37 < Lathiat> woops not that one
08:37 < Lathiat> uml_mconsole halt
08:37 < Lathiat> damn i meant th eother one! :)
09:58 -!- pasky [~pasky@pasky.ji.cz] has joined #uml
09:59 -!- pasky [~pasky@pasky.ji.cz] has left #uml []
11:02 < david> caker: around?
11:04 < ido> :) i'm here
12:02 < caker> Hello .. arp poisining is bad ..
14:16 < lah> hum, what's the kernel cmdline option for sync disk IO on/off ?
14:17 < green> I think it is a compile-time option
14:17 < caker> linux --help
14:18 < Pahan> caker: It shouldn't have anything to do with UML.
14:18 < caker> Pahan: it does
14:18 < Pahan> Oh, wait, it is.
14:18 < Pahan> :(
14:23 < lah> okay, appending an s..
14:38 < ido> CONFIG_BLK_DEV_UBD_SYNC
14:38 < ido> ?
14:38 < caker> Keep that off, that way, you can do it on the command line by adding s
14:43 < lah> yeah, that's what I did now
14:43 < lah> I thought I could also disable it from the cmdline
14:44 < caker> nope - it's a quirk
14:44 < caker> or a feature, depending on how you want to look at it :-)
14:54 < lah> i think enabling SYNC as default was a baaad idea
14:54 < lah> *waiting*
14:57  * Lathiat grins at lah
14:57  * Lathiat waits with lah
15:09 < mistik1> caker: you around
15:10 < mistik1> I got my little project to work 
15:10 < mistik1> I have two diskless workstations booted off my UML server
15:19 < mistik1> guys how do I prevent this xterm from running when I start my UM?
15:19 < mistik1> I have no extra VTs configured so I dont even need it
15:29 < ljlane> linux  devfs=mount eth0=tuntap,tap0 con0=pts,fd:1 con=pts .. the two con bits. I believe I copied that directly out of the HOWTO.
15:31 < ljlane> Maybe that's the wrong set.
15:33 < ljlane> There it is, section 5.3:  First, this is how you get rid of those bleeding console xterms by attaching them to host ptys:
15:33 < ljlane> con=pty con0=fd:0,fd:1
15:34 < caker> or send them to null, and just set your main console... con=null con0=fd:0,fd:1
15:50 < ljlane> Interesting. I suddenly understand the con stuff after all this time.
15:53 < mistik1> heh
15:53 < mistik1> me too
15:53 < mistik1> thanks
16:59 -!- Dave\\ [dave@67.81.154.64] has joined #uml
17:01 < Dave\\> hey
17:05 < lah> i think 2 hours vs 8 minutes is quite a difference (rh9 package installation, kernel with sync IO vs kernel without sync plus some optimizations)
17:06 < caker> Geesh
17:07 -!- pirlouit [~peter@64.162.195.202] has quit [Quit: Time to frag!]
17:10 < Lathiat> hah
17:12 < Dave\\> does anybody know if there are rh8 rootfs images available to download?
17:15 < lah> Dave\\: make your own! :)
17:16 < lah> Dave\\: been playing around with it for quite some time now .. and it even nearly worked (with rh9)
17:18 < lah> it's stuck at the post install (rpm execution), but I'll try waiting ;)
17:23 < caker> lah: are you following green's document?
17:24 < lah> caker: you didn't want to say "probe" did you
17:47 < Dave\\> lah, I made one but it's semi-screwy
17:47 < Dave\\> :(
17:47 < caker> what's wrong with it?
17:55 < Dave\\> I have no idea, when I reboot it, sometimes new files written to it in that session will be lost
17:55 < Dave\\> am I doing something wrong?
17:55 < caker> Are you using COW files, and how are you turning off the UML?
17:56 < Dave\\> no, not using COW
17:57 < Dave\\> and I'm turning it off by issuing shutdown from inside the UML
17:59 < caker> and then you're shutting down your host machine properly, too, right? .. that's about the only thing I can think of ...
18:00 < Dave\\> well, the host machine hasn't been shut down during the existence of this :(
18:01 < caker> heh .. i dunno then
18:01 < Dave\\> I've been confused about COW.. would users still be able to modify ro files on the shared fs but those changes would be saved on their private COW file?
18:02 < caker> yes
18:02 < caker> the read-only, and their cow file as a whole become their new disk image
18:02 < caker> the COW is kind of like a journal
18:07 < Dave\\> is the writing delayed/buffered to the rootfs (running rh8)? it's very strange.. files are just disappearing :(
18:07 < caker> Yes, unless you use sync, which is slow
18:08 < Dave\\> is there a way for me to ensure that everything is written before it shuts down?
18:08 < caker> You could call "sync" on the host afterwords, or run your udb devices in always-sync mode
18:09 < Dave\\> okay, thanks
18:54 < mistik1> who is the screen guru here
18:54 < mistik1> what is wrong with this command?
18:54 < mistik1> screen -d -m -S DEVEL -X screen -t UMNODE command
19:24 -!- trappist [~trappist@crtntx1-ar8-4-47-221-175.crtntx1.dsl-verizon.net] has joined #uml
19:25 < trappist> I got rid of the xterms by using "con=pty con0=fd:0,fd:1" but I'd like to have a console in the window I ran uml in - I've done it before but this is the first time I've played with uml in a long time.  What am I missing?
19:26 < trappist> oh heya mistik1
19:38 < ljlane> trappist, We covered that topic a bit earlier today. This should work, works for me: con=null con0=fd:0,fd:1
19:40 < trappist> ljlane: I just now tried that too, with the same results except I get a lot of 'permission denied' trying to open pty's
19:40 < mistik1> hey trappist 
19:40 < trappist> err tty's
19:40 < trappist> /etc/rc.d/rc.sysinit: line 807: /dev/tty2: Operation not permitted
19:41 < ljlane> strange
19:42 < caker> because with using con=null con0=fd:0,fd:1, you've only got tty0 inside uml
19:42 < caker> or you're running devfs (try devfs=nomount on command line)
19:42 < trappist> I'm running devfs... I'll try that
19:43 < caker> or, run devfs, and fix your init scripts :-)
19:43 < trappist> :)
19:44 < trappist> well it used to stop after 'running devfsd actions' now it stops after 'starting kheader'
19:47 < trappist> something in the back of my mind is telling my I fixed this last time with something inside the uml... maybe in the inittab?
19:48 < caker> Yeah, you've got some non-devfs stuff left in there - also check fstab, etc
19:48 < trappist> # Run gettys in standard runlevels
19:48 < trappist> 1:2345:respawn:/sbin/mingetty ttys/1
19:49 < trappist> nothing there for tty0 could that be it?
19:49 < caker> I usually add a 0:2345/respawn:/sbin/mingetty ttys/0 for con0
19:49 < caker> (or tty0 for non-devfs)
19:50 < Lathiat> try vc/0
19:50 < Lathiat> that works for me
19:51 < trappist> Lathiat: in inittab?
19:51 < trappist> not sure what you mean
19:51 < trappist> woohoo!
19:51 < trappist> adding ttys/0 to inittab did the trick :)
19:53 < trappist> ok this is cool
19:55 < Lathiat> hehe
19:55 < Lathiat> www.sixlabs.org/talks/uml/
19:55 < Lathiat> its pretty debian orientated tho
19:56 < trappist> sweet
19:56 < ljlane> Blue on dark black?
19:56 < ljlane> err.. black
20:05 < Lathiat> yeh sorry about that
20:05 < Lathiat> i didnt export :)
20:24 -!- chris_ [~chris@ethlife-a.ethz.ch] has joined #uml
20:25 -!- chris_ [~chris@ethlife-a.ethz.ch] has quit [Quit: [x]chat]
20:25 -!- pflanze [~chris@ethlife-a.ethz.ch] has joined #uml
20:26 < pflanze> Hello
20:26 < pflanze> I'm trying to add skas3 on top of grsecurity.
20:27 < pflanze> (Which I've already done recently, but I've noticed that I did a few spots wrong.)
20:28 < pflanze> The problem is that grsecurity changes the do_mmap2() function to use do_mmap() instead of do_mmap_pgoff().
20:29 < pflanze> And skas3 only changes do_mmap_pgoff() to use an additional *mm parameter, and keeps do_mmap() with original parameters.
20:30 < pflanze> I'm trying to change that but am unsure how.
20:31 < pflanze> I've copied do_mmap and changed that to include the *mm parameter.
20:32 < pflanze> But I'm unsure about the grsecurity checks done there on current->... stuff.
20:32 < pflanze> Anyone knows about the design of the /proc/mm stuff? 
20:33 < pflanze> AFAIK, it is about creating address spaces independent of the calling process, right?
20:35 < Lathiat> i think so
20:35 < Lathiat> user-mode-linux.sf.net/skas.html
20:36 < pflanze> I guess grsecurity should *not* check those 'pseudo' address spaces?
20:37 < Lathiat> xhrug
20:37 < Lathiat> shrug
20:44 -!- DnsInfector_ [~DnsInfect@81.53.251.29] has joined #uml
20:51 -!- DnsInfector [~DnsInfect@ARennes-204-1-1-200.w193-251.abo.wanadoo.fr] has quit [Ping timeout: 490 seconds]
20:59 -!- sssssssh______ [~chatzilla@adsl-19-239-102.jax.bellsouth.net] has quit [Remote host closed the connection]
21:12 -!- pirlouit [~peter@64.162.195.202] has joined #uml
21:14 < mistik1> woohoo!!
21:14 -!- sssssssh______ [~chatzilla@68.19.239.102] has joined #uml
21:15 < mistik1> this gets better daily
21:15 < mistik1> now to figure out how to deal with the originating screen session
21:31 < pflanze> mistik1: you mean, what should you do so that you can leave it alone and log out of the host?
21:32 < mistik1> no
21:32 < mistik1> I'm way past that
21:32 < pflanze> ah sorry :)
21:32 < mistik1> However what I so now is similar to Cameron Kerr's scripts
21:33 < mistik1> screen -dmS $SESIONNAME 
21:33 < mistik1> sleep 2s
21:34 < mistik1> screen -r -X -t "$SESIONNAME console" $umkernel_withargs
21:34 < mistik1> this works fine
21:35 < mistik1> however when you stop uml it still leaves the first screen session in place
21:35 < mistik1> since this script has a start/stop I think i'll do something like this as a kludge
21:36  * pflanze does not understand
21:36 < mistik1> kill `screen -ls |grep -w $SESIONNAME |awk -F. '{print $1}'`
21:36 < pflanze> Anyway, I've never seen it occupy screen after exiting.
21:37 < mistik1> this happens because I create the screen first them send a command to it to be run
21:37 < mistik1> this cause anything run in it to get a new screen window also
21:37 < mistik1> such as VTs
21:40 < mistik1> should running 'mconsole UMID halt' be distructive to mounted filesystems?
21:43 < mistik1> screen needs a -wipe -all command or something like that
21:44 < pflanze> Ah, I see.
21:44 < mistik1> did you test it?
21:44 < pflanze> no.
21:45 < pflanze> So far I just start my uml's with a script that starts screen with it. Nothing for shutdown yet.
21:47 < caker> yes
21:47 < caker> you can to a sysrq s before it to sync the drives
21:47 < caker> i think
21:47 < mistik1> hmm
21:48 -!- trappist [~trappist@crtntx1-ar8-4-47-221-175.crtntx1.dsl-verizon.net] has quit [Read error: Connection reset by peer]
21:48 -!- pirlouit [~peter@64.162.195.202] has quit [Quit: Time to frag!]
21:51 < pflanze> Why not write a daemon in perl that listens for udp broadcasts from the host when the latter wants (them) to shut down?
21:52 < pflanze> I have some such scripts running for other purposes (somthing heartbeat alike).
21:53 < mistik1> for one thing i'm no perl monger
21:54 < pflanze> I could do it.
21:55 < pflanze> I'm still thinking about how to make sure replay attacks won't work, though.
21:55 < pflanze> So no third party can shut down your uml.
21:56 < pflanze> -> means that the uml instances must regularly send pakets by themselves as well, with a changing challenge.
21:59 < david> mother fucker
21:59 < david> one of the UMLs just *won't* fsck it's drive
21:59 < david> er, ubd0
22:03 < Lathiat> hrm
22:09 < mistik1> pflanze: this works so far
22:09 < pflanze> mistik1: what works?
22:10 < mistik1> the sysrq s syncs the disks but still forces fsck on restart as they are not un mounted
22:10 < mistik1> this does
22:10 < mistik1>  stop)
22:10 < mistik1>         shift
22:10 < mistik1>         uml_mconsole "$mc" sysrq s "$@"
22:10 < mistik1>         uml_mconsole "$mc" halt "$@"
22:10 < mistik1> if [ "$screen" = "true" ]; then
22:10 < mistik1>         kill `screen -ls |grep -w $name |awk -F. '{print $1}'`
22:10 < mistik1> fi
22:10 < mistik1>         exit
22:10 < mistik1>         ;;
22:10 < caker> sleep a few after the sync
22:10 < mistik1> there has got to be a better way to send halt
22:10 < mistik1> caker: I will
22:12 < mistik1> If we can send cad to init why cant we send init 0
22:12 < caker> david: boot into single on the UML and do it there?
22:13 < mistik1> mconsole umid halt is brutal
22:14 < pflanze> What do you think about the networking (udp) way? And/or what about serial consoles?
22:14 < mistik1> what about ssh 
22:15 < mistik1> with keys
22:15 < mistik1> ssh -l $user $umlhost halt
22:15 < mistik1> simple enough
22:16 < mistik1> since the user running he UML will have root there anyway
22:16 < caker> i think you'll want poweroff, halt goes to single, no?
22:16 < mistik1> depends on distro I think
22:16 < mistik1> with redhat halt goes to init 0
22:16 -!- pirlouit [~peter@64.162.195.202] has joined #uml
22:17 < caker> what is zero, turn off?
22:17 < mistik1> yep
22:17 < caker> cool ..
22:17 < pflanze> With UDP you would have some sign that the uml is up at the same time.
22:18 < pflanze> But maybe you have better ways for that, dunno.
22:19 < Lathiat> a machien is still pigable in int 0
22:19 < Lathiat> if ur distro doesnt down the interfaces
22:19 < pflanze> Well, the transport layer is one thing (udp, ssh, even email..), deamon/heartbeat stuff the other.
22:19 < mistik1> pflanze: indeed
22:20 < pflanze> But udp is the simplest form of communication, shouldn't break too soon.
22:20 < mistik1> I'm not disagreeing with anything you're saying i'm just not ready for that step yet myself
22:20 < mistik1> what you say makes sense
22:20 < pflanze> (Well serial device would probably even make more sense because of this, but I've not used those until now.)
22:21 < mistik1> have a little heartbeat running clientside
22:21 < mistik1> I even have a little C code here that could do it
22:22 < mistik1> It listens on a certain port and will run what commands I configure it for
22:23 < mistik1> It was a little daemon I was going to use for a webinterface to talk to a local daemon
22:24 < pflanze> Perl is easier to extend :)
22:24 < mistik1> heh
22:24 < pflanze> For me at least.
22:25 < DnsInfector_> easy to extend but not to maintain
22:26 < pflanze> I don't believe C is easier to maintain for such a thing.
22:26 < pflanze> But I'm no C junkie.
22:27 < Pahan> Heh, C and Perl.
22:28 < Lathiat> nah, python kix
22:29 < mistik1> actually python looks cool
22:29 < DnsInfector_> time to sleep (4.30AM here)
22:29 -!- DnsInfector_ [~DnsInfect@81.53.251.29] has quit [Quit: Client exiting]
22:29 < mistik1> I think i'll have to learn that one
22:30 < pflanze> AFAIK, perl is the only non-shell scripting language which is guaranteed to be installed on a debian system.
22:31 < Pahan> Doesn't mean it's any good.
22:31 < Pahan> Besides, this is Debian.
22:31 < Pahan> Python is just an "apt-get install python2.2" away.
22:33 < mistik1> heh
22:33 < pflanze> Which doesn't mean it's any better.
22:33 < Pahan> Except that it is.
22:33 < mistik1> I have nothing against perl
22:33 < Pahan> I have!
22:33 < mistik1> but coding it gives me a headache
22:33 < mistik1> so I dont unless I MUST
22:33 -!- pirlouit [~peter@64.162.195.202] has quit [Quit: Time to frag!]
22:34 < pflanze> So what do you have against perl, Pahan?
22:35 < Pahan> pflanze: It's not readable and it's not simple and I hope it stays away from me.
22:35 < mistik1> haha
22:36 < mistik1> it certain is like a house built with duct tape
22:36 < Pahan> Oh, we have these all over California.
22:36 < Pahan> Houses made from cardboard and snot.
22:41 -!- adama [adama@pie.methinks.co.uk] has joined #uml
22:42  * pflanze would still like to learn lisp (the only real language?)
22:43 < Lathiat> eww :p
22:50 < david> caker: I am in single user mode:-)
22:51 < caker> david: it won't force ?? 
22:52 < david> init-2.05b# fsck -fyvvv /dev/ubd0
22:52 < david> fsck 1.33 (21-Apr-2003)
22:52 < david> e2fsck 1.33 (21-Apr-2003)
22:53 < david> and it gets stuck there
22:53 < caker> using cows?
22:59 -!- pirlouit [~peter@64.162.195.202] has joined #uml
23:06 < david> caker: yeah
23:27 -!- pflanze [~chris@ethlife-a.ethz.ch] has quit [Quit: [x]chat]
23:42 -!- umnode1 [geeks@ool-44c02046.dyn.optonline.net] has joined #uml
23:54 -!- kayl [~jada@6532176hfc248.tampabay.rr.com] has joined #uml
--- Log closed Sun Jun 29 00:00:00 2003