--- Day changed --- Log opened Fri Jun 27 00:00:02 2003 01:52 -!- mistral [mistral@212.159.71.212] has joined #uml 02:05 < Pahan> Hrm. I ran an infinite loop in a Python interpreter under a UML under a shell with ulimit -t 5, and I got a kernel panic. 02:10 < Pahan> Oh, right, ulimit -t is deadly. 03:25 -!- gump [gump@81.5.136.90] has quit [Read error: Connection reset by peer] 03:29 -!- gump [gump@81.5.136.90] has joined #uml 03:31 -!- gump [gump@81.5.136.90] has quit [Client Quit] 04:06 -!- ido [~Ido@bzq-218-140-160.cablep.bezeqint.net] has quit [Ping timeout: 496 seconds] 05:40 -!- Lathiat [lathiat@seven.sixlabs.org] has quit [Ping timeout: 490 seconds] 06:00 -!- aw [~andrew@inhibition.dilex.net] has quit [Ping timeout: 496 seconds] 06:00 -!- Lathiat [~Lathiat@thump.bur.st] has joined #uml 06:03 -!- ido [~Ido@bzq-218-140-160.cablep.bezeqint.net] has joined #uml 06:05 -!- aw [~andrew@inhibition.dilex.net] has joined #uml 06:18 < ido> i've booted the initrd slackware 9.0 06:18 < ido> on a UML 06:18 < ido> on cat /proc/partitions, i can see: 06:18 < ido> major minor #blocks name 06:18 < ido> 98 0 10000 ubd/disc0/disc 06:18 < ido> 98 1 8176 ubd/disc0/part1 06:18 < ido> but there isn't such a device /dev/ubd* 06:18 < ido> so where does ubd exist ? 06:19 < Lathiat> devfs magic 06:21 < ido> could you please elaborate ? 06:22 -!- Lathiat_ [lathiat@130.95.13.25] has joined #uml 06:24 < ido> Lathiat: could you please elaborate on 'devfs magic' ? 06:24 < ido> not sure if it has to do with it, but i ran the UML with devfs=nomount 06:25 < green> so create /dev/ubd*yourself? 06:26 < Lathiat> mount -t devfs dev /dev ?? 06:26 < Lathiat> or 06:26 < Lathiat> mknod /dev/yubd* 06:26 -!- Lathiat [~Lathiat@thump.bur.st] has quit [Quit: brb] 06:27 < ido> oh, i see. 06:27 < ido> so /proc/partitions is a kernel thingy 06:27 < ido> and devfs takes the info from /proc/partitions and makes devices for them, right ? 06:28 < ido> hmm, sweet. 06:28 -!- Lathiat_ [lathiat@130.95.13.25] has quit [Quit: Lost terminal] 06:28 < ido> oh wait. no it doesn't. 06:28 -!- Lathiat [lathiat@130.95.13.25] has joined #uml 06:29 < ido> does it ? 06:29 * green wonders if "processor detected hard error" MCE error means something dangerous, like should I trash the processor? ;) 06:30 < ido> eh. 06:32 < ido> where does devfs take its input ? 06:32 < Lathiat> err 06:32 < Lathiat> from the kernel 06:32 < Lathiat> mount -t devfs dev /dev 06:32 < ido> i mean, when i ran the kernel with fake_ide and fakehd, it devfs stil made /dev/ubd's 06:32 < ido> the question to be more accurate, is why did it make both of them ? 06:33 < ido> when i cat'd /proc/partitions i saw only hd/* devices, not ubd/*'s 06:33 < green> if you want kernel to really think of ubd devices as hda ones, you need to add "ubd=3" to your uml commandline 06:34 < green> then access to /dev/hda* will result in accessing ubd 06:59 < ido> where is the ubd=3 option documented ? 07:00 < Lathiat> it isnt? *grin* 07:00 < ido> haven't seen it on ./uml --help 07:00 -!- DnsInfector [~DnsInfect@ARennes-204-1-11-50.w81-53.abo.wanadoo.fr] has joined #uml 07:00 < ido> or anywhere else 07:02 < ido> btw, is there a way to set where 10* tempfiles made by the UML will be placed ? 07:03 < Lathiat> export TEMP or TMP? 07:04 < ido> and if i want to place them in another directory (other then the one set on TMP).. 07:05 < Lathiat> TMP=/blah ./linux 07:05 < Lathiat> will set it for that instance 07:07 < ido> ah. ok. 09:08 -!- litost [~sombitch@phynp6.phy-astr.gsu.edu] has joined #uml 10:53 -!- DnsInfector [~DnsInfect@ARennes-204-1-11-50.w81-53.abo.wanadoo.fr] has quit [Read error: Connection reset by peer] 11:35 -!- Getty [~Getty@metaluna4.de] has quit [Read error: Connection reset by peer] 12:30 -!- DnsInfector [~DnsInfect@ARennes-204-1-11-50.w81-53.abo.wanadoo.fr] has joined #uml 12:30 < silug> isn't that TMPDIR? 12:34 -!- DnsInfector [~DnsInfect@ARennes-204-1-11-50.w81-53.abo.wanadoo.fr] has left #uml [] 12:35 -!- DnsInfector [~DnsInfect@ARennes-204-1-11-50.w81-53.abo.wanadoo.fr] has joined #uml 12:50 -!- vertigo_ [~vertigo@frontier.xs4all.nl] has joined #uml 13:30 -!- lah [~z@AVATAR.internet.ufg.ac.at] has joined #uml 13:52 < Pahan> Hrm, how concurrent are SMP UMLs on a single-CPU host? 13:54 < david> Pahan: how do you mean 13:55 < david> Pahan: each 'cpu' on the UML gets scheduled on the host like any other process 13:55 < Pahan> Well, is it more likely to expose any wonderful race conditions than a single-CPU UML> 13:55 < Pahan> ? 13:55 < Pahan> Hrm. 13:55 < Pahan> Guess not? 13:56 < david> Pahan: yes, because the UML scheduler will schedule two processes at the same time 13:56 < Pahan> Mmmh. 13:56 < david> Pahan: the fact that they don't occur at the same time on the host doesn't matter 14:03 -!- aw [~andrew@inhibition.dilex.net] has quit [Quit: ChatZilla 0.8.23 [Mozilla rv:1.3.1/20030522]] 14:27 -!- yeti [~yeti@p3EE2297C.dip.t-dialin.net] has joined #uml 15:25 -!- revenger1 [revenger@p508B2554.dip.t-dialin.net] has joined #uml 15:32 -!- revenger_ [revenger@p508B2ACC.dip.t-dialin.net] has quit [Ping timeout: 490 seconds] 15:49 -!- shak [~shak@pc1-hudd4-5-cust78.hudd.cable.ntl.com] has joined #uml 15:49 < shak> Hi 15:50 < yeti> hi! 15:50 < shak> do you know what kernel options I need for the /dev/loop support that david mentions in the howto? 15:51 < yeti> i dont know the howto? 15:51 < yeti> that loop device for mounting files? 15:51 < shak> yes 15:51 < yeti> section block devices 15:51 < shak> its at uml.openconsultancy.org 15:51 < shak> ah, I see 15:52 < shak> "loopback support" 15:52 < yeti> yes 15:52 < shak> thanks 15:52 * shak can't believe he didnt see that 15:53 < yeti> my prototype of carrier system with 2 umls where one is firewall is up 15:53 < shak> nice 15:53 < shak> Im just making a LFS solution for the root_fs for mine 15:53 < yeti> i plugged my old firewall back to dsl but the uml-ish one had a successfull test run 15:54 < shak> :) 15:54 < yeti> i need to het ised to iptables... 15:54 < yeti> het->get 15:54 < yeti> my old fw is ipchains based 15:54 < shak> yes, iptables isnt something that I use that regularly 15:54 < yeti> and i need some redirections to be able to phone thru my fw 15:55 < shak> you have VoIP? 15:55 < yeti> and i need my phone or the international calls to my gf are too expensive 15:55 < shak> I fancy a VoIP phone 15:55 < yeti> i use gnomemeeting and she uses openphone (after experiments with netmeeting) 15:55 < shak> ah :) 15:56 < yeti> i think openphone is the choice... 15:56 < yeti> more control, less bloat... 15:56 < shak> :D 15:57 < Pahan> Why don't you use Vonage and NOT screw with broken open sores crap? 15:57 < yeti> now i need some ideas to get my umls up and down in startup scripts... i dont like the screen acrobatics i saw... 15:58 < yeti> maybe i put each uml on a real tty of the host ;-) 15:58 < Pahan> Last time I went through all this, none of available applications that use SIP were compatible with Cisco stuff. 15:58 < shak> Pahan: open source is good, you need to be able to connect to it 15:58 < shak> the cisco SIP stuff is second rate. 15:58 < yeti> vonge? whazzat? 15:59 < Pahan> yeti: www.vonage.com 15:59 < Pahan> shak: openphone is third-rate. It barely works. 16:00 < yeti> it works. 16:02 < yeti> and i exactly need a connection to my girlfroend... nothing else... i might even use kinda netpipes between the soundcards... i dont need to talk to others except her... i am a hermit... 16:02 < yeti> kinda. 16:03 < yeti> shak: why lfs 16:03 < yeti> ? 16:03 < shak> because its small and fast and custom :) 16:04 < yeti> that's what i think about debian 16:05 < yeti> i dont care about modifying deb-sources to fit my needs and having an own additional repository 16:05 < yeti> and gentoo cured me from optimizing... 16:05 < yeti> gentoo robbed all the time i saved with debian... 16:09 -!- shak [~shak@pc1-hudd4-5-cust78.hudd.cable.ntl.com] has quit [Quit: new kernel!] 16:15 -!- shak [~shak@pc1-hudd4-5-cust78.hudd.cable.ntl.com] has joined #uml 16:15 < shak> right 16:15 < shak> Im going to attempt to install onto a mounted device now, this should be fun 16:16 < yeti> why? 16:16 < shak> well, I'm going to do this 16:16 < shak> dd if=/dev/zero of=rootfs count=1 bs=1k seek=$[1*1024*1024] 16:16 < shak> which is my 1GB filesystem 16:16 < shak> mkreiserfs that filesystem 16:16 < shak> # mount -o loop rootfs mnt/new.rootfs then do that 16:16 < shak> and install LFS onto it 16:17 < yeti> everyday business for me such stuff :-) 16:17 < shak> yus :) 16:17 < shak> was that a why it was going to be fun? 16:18 < ido> :( 16:18 < ido> :) 16:18 < ido> eh 16:18 < yeti> is fun the firt few times... 16:18 < yeti> ...first... 16:18 < shak> its always fun playing with linux 16:18 < ido> does sound like fun 16:18 < Pahan> LFS sucks, though. 16:18 < Pahan> Well, the third time you do it, it does. 16:18 < shak> whyso? 16:18 < shak> I've done it ~15 times 16:18 < shak> and I like it 16:18 < Pahan> Heh. 16:18 < ido> which distro do you prefer ? 16:19 < Pahan> I use Debian. Usually, it works without having to spend hours troubleshooting it. 16:19 < yeti> :-) 16:20 < ido> eh. *BSD would run under bochs, right ? 16:22 < yeti> i use debian to get results and gentoo to waste time :-) 16:23 < ido> lol. 16:24 < yeti> i cant even get kde or gnome compiled on gentoo atm... 16:24 < yeti> they have upgraded something to death again... 16:30 < yeti> btw: stage1 of gentoo 1.4_rc1 was built on debian :-) 16:31 < yeti> but that doesnt really mater... it would have been possible on any other linux... it is a chroot action. 16:38 -!- yeti [~yeti@p3EE2297C.dip.t-dialin.net] has left #uml [...breeding some more umls...] 17:40 < shak> 108MB so far for my LFS system 17:41 < ido> eh 17:41 < shak> its size.. 17:41 < shak> is 108MB so far 17:41 < ido> why not just install slackware ? 17:41 < shak> I want a custom one :) 17:41 < shak> and this is faster 17:41 < shak> and lighter 17:50 -!- litost [~sombitch@phynp6.phy-astr.gsu.edu] has quit [Quit: ERC v2.91 $Revision: 1.239 $ (IRC client for Emacs)] 18:24 -!- vertigo_ [~vertigo@frontier.xs4all.nl] has left #uml [no more! no more!] 18:29 < lah> green? 18:41 < mistik1> dev_ip_addr - device not assigned an IP address 18:41 < mistik1> I got the error after ifconfig runs in my UML 18:41 < mistik1> anyone seen that one before 18:41 < mistik1> btw: the networking works, even with the error 19:36 -!- mistral [mistral@212.159.71.212] has quit [Ping timeout: 488 seconds] 19:43 -!- Lathiat [lathiat@130.95.13.25] has quit [Quit: brb] 19:43 -!- Lathiat [lathiat@seven.sixlabs.org] has joined #uml 19:45 -!- Lathiat [lathiat@seven.sixlabs.org] has quit [Client Quit] 19:46 -!- Lathiat [lathiat@seven.sixlabs.org] has joined #uml 20:26 < ido> hmm 20:26 < ido> one UML, and it REALLY slows my system 20:27 < caker> then you've done something wrong :-) 20:27 < ido> Kernel panic: switch_mm_skas - PTRACE_SWITCH_MM failed, errno = 9 20:27 < ido> Kernel panic: switch_mm_skas - PTRACE_SWITCH_MM failed, errno = 9 20:27 < ido> Kernel panic: Segfault with no mm 20:27 < ido> Kernel panic: Segfault with no mm 20:27 < ido> Kernel panic: Segfault with no mm 20:27 < DnsInfector> or your system has weak CPUs ;) 20:27 < caker> you need the ptrace patch for your skas patch 20:27 < ido> athlon XP 1700 20:28 < ido> what ptrace patch ? 20:28 < caker> ido: http://kernels.usermodelinux.org/host/host-skas3.ptrace.patch.bz2 20:28 < DnsInfector> skas patch nicely improve performances 20:29 < ido> is that a skas with a ptrace patch ? 20:29 < ido> or do i patch in ontop of skas ? 20:29 < caker> You're running a kernel which has the ptrace security fix in it .. skas needs it too, or else you get that message 20:29 < ido> oh 20:29 < caker> That patch I linked to is skas and ptrace fix (it's only one line) all in one 20:30 < ido> oh 20:31 < ido> could it be that the UML is slow since i dont have much RAM ? 20:31 < caker> yup 20:31 < DnsInfector> by default it takes 16M iirc 20:31 < caker> I think it's 32 20:32 * DnsInfector usually give 128Mo for rh72 root 20:32 < ido> wow 20:34 < ido> if the kernel is already patched with skas 20:35 < ido> should i delete it and reopen the kernel sources from scratch (and then patch with the patch) 20:35 < caker> Yes 20:35 < caker> Or 20:35 < ido> or could i patch the patched kernel 20:35 < caker> if you have the source to your kernel still, just add the one liner 20:35 < caker> and recompile 20:35 < ido> which line exactly ? 20:37 < caker> http://marc.theaimsgroup.com/?l=user-mode-linux-user&m=105060615322517&w=2 20:38 < ido> thanks :) 20:38 < caker> np 20:38 < ido> damnt, now i gotta reboot ;p 20:39 < caker> [root@host1 root]# uptime 20:39 < caker> 20:36:35 up 13 days, 14:41, 3 users, load average: 0.42, 0.45, 0.41 20:40 < caker> with 34 UML's running on it 20:40 < shak> what spec machine? 20:40 < caker> That one is just a p4 2.4Mhz with 2.5GB ram 20:40 < shak> :o 20:40 < shak> nice :D 20:40 < caker> others are xaxons 20:40 < shak> you finding better performance with xeons? 20:40 < caker> j/k btw 20:40 < shak> :p 20:40 < ido> wow 20:40 * shak didnt get that 20:40 * shak slap 20:40 < caker> Yes, especially ram/bus speed - 533 vs 400 20:41 < shak> how many UMLs on a dual xeon box (2.4Ghz, 2GB RAM) would you say? 20:41 < caker> I don't over commit ram, so that is the limiting factor 20:42 < ido> i've got a simple athlonXP1700 with currently 128mb ram (256 were donated to another box) will buy more when i'll get more cash. 20:42 < ido> ej 20:42 * DnsInfector feels ridiculous with his dual MP2400+ + 1GbRam 20:42 < ido> where do you run that box at ? 20:42 < caker> off my aol account 20:42 * shak has a dual MP1800+ 0.5GB RAM 20:43 < ido> running that at home ? 20:43 < caker> just kidding 20:43 < caker> no it's at a datacenter 20:43 < ido> eh. thought so. 20:43 -!- DnsInfector_ [~DnsInfect@ARennes-204-1-1-200.w193-251.abo.wanadoo.fr] has joined #uml 20:43 < caker> my home computer is a lowely 1Ghz Athlon 512MB ram 20:44 < ido> ;p 20:44 < ido> lemme try giving the UML more RAM 20:44 < ido> it should run at least as fast as the host, right ? 20:44 < caker> Try adding ram to your machine first :-) 20:44 < ido> i mean, when there aren't any other processes on running on the host 20:44 < caker> UML runs 20:44 < caker> that was weird 20:45 < caker> sorry - uml runs as fast to 3x as slow as the host, i'd estimate - depends on task 20:45 < ido> how come ? 20:45 < caker> Because there's a kernel running on a kernel 20:46 < ido> hmm. doesn't a x86 emulater such as bochs run faster then being x3 slower then the host ? 20:46 < ido> s/emulator 20:46 < caker> i wouldn't think so 20:46 < caker> because then its: 20:46 < caker> a kernel running on emulated hardware running on a kernel 20:47 < ido> my point it, isn't the UML supposed to be much faster then a x86 emulator, which emulates hardware 20:47 < caker> Yes, correct 20:47 < ido> but x3 slower then the hosts still sounds quite slow ;p 20:47 < caker> 3x slower **for some things** 20:48 < caker> like 20:48 < caker> block io 20:48 < caker> kernel calls 20:48 < ido> then people running UML's on p2 400's actually feel as if they're working on *pentium mmx's* ? 20:48 < ido> eh. 20:49 < caker> What are your intentions with UML? 20:49 < ido> any way to speed up things ? any sort of cache mechanism ? 20:49 < caker> skas and using tmpfs are the first two 20:49 < caker> Others are just little tweaks here and there 20:50 < ido> running a few hosts on my HOME network. one acting as a web/mail server, one as a public server, and another one as a playground 20:50 < ido> which tweaks ? 20:50 < caker> I dunno - all of the normal performance tweaks apply 20:50 < caker> hdparam 20:50 < caker> etc 20:50 < caker> er, hdparm 20:50 < ido> oh:) those 20:51 < ido> thought you had UML specific tweaks 20:51 -!- DnsInfector [~DnsInfect@ARennes-204-1-11-50.w81-53.abo.wanadoo.fr] has quit [Ping timeout: 492 seconds] 20:51 < caker> Well, just don't build enormous UML kernels 20:51 < caker> don't use sync mode on your uml devices 20:52 < ido> hmm, how do you define enormous ? 20:52 < caker> Just build stuff as modules (which is insecure) or don't include stuff you don't need - that helps to build a smaller application 20:52 < caker> and, turn off all the kernel debugging opts 20:52 < ido> does it really matter ? i mean, excpet being a bit bloated in the host's memory ? 20:53 < caker> well, a UML kernel with debugging compiled in is like 32Megs, that's like 1/8 your hosts's ram :) 20:53 < caker> er 1/4 even 20:53 < ido> are modules considered insecure since they could be changed and then could lead to other exploits ? 20:54 < caker> basically 20:54 < ido> welp, 32mb's before being stripped, right? eh 20:54 < caker> they could hi-jack the uml process on the host 20:54 < ido> oh. hmm. haven't thought of that. 20:54 < ido> thats bad ;p 20:55 < ido> how do i check if i'm using sync mode ? 20:55 < caker> there's a uml kernel .config option .. grep for UBD or something 20:55 < shak> night 20:55 < caker> always_use_sync or something 20:55 < caker> shak: nite 20:56 * caker is afk for a bit 20:56 -!- gump [gump@81.5.136.90] has joined #uml 20:57 < ido> CONFIG_BLK_DEV_UBD_SYNC 20:57 < ido> isn't set ;p 20:57 * ido is rebooting 20:57 -!- DnsInfector_ [~DnsInfect@ARennes-204-1-1-200.w193-251.abo.wanadoo.fr] has quit [Quit: Client exiting] 21:03 -!- ido [~Ido@bzq-218-140-160.cablep.bezeqint.net] has quit [] 21:03 -!- ido [~Ido@bzq-218-140-160.cablep.bezeqint.net] has joined #uml 21:08 -!- revenger1 [revenger@p508B2554.dip.t-dialin.net] has quit [Ping timeout: 496 seconds] 21:13 < ido> could i dynamicly assign more RAM to a UML with mconsole ? 21:14 < david> no 21:23 < Pahan> You can! 21:24 < Pahan> Create a swap partition, put it on a tmpfs and assign it to UML. 21:24 < ido> oh. nice trick. 21:26 < ido> i need to unmount the current swap partition, right ? 21:26 < Pahan> No, you can add one more. 21:27 < ido> but wont it first fill up the first, and only then use the second swapfs (which sits on a tmpfs) 21:27 < Pahan> If you aren't running out of memory, why do you need more? 21:28 < Pahan> You can juggle swap priorities, actually. 21:29 < ido> the first swap fs sits on the HD, which is rather slow 21:29 < Pahan> Like I said, you can change the priorities dynamically. 21:29 < Pahan> I think. 21:29 < ido> i think so to, not sure how though. 21:30 < Pahan> swapon 21:30 < ido> ouu. doens't quite matter. that was only for theoretaical knowhow. 21:31 < ido> 04:29:20 up 17 min, 3 users, load average: 3.19, 1.59, 0.72 21:32 < ido> is there a way to benchmark my UML ? 21:33 < Pahan> Not that I know of. 21:33 < ido> in terms of CPU and disk read||write speed ? 21:33 < Pahan> hdparm can measure disk speed. 21:33 < Pahan> -t and -T 21:34 < ido> foobar root # hdparm /dev/hda2 21:34 < ido> /dev/hda2 not supported by hdparm 21:34 < ido> eh 21:34 < Pahan> Heh heh. 21:35 < ido> http://lbs.sourceforge.net/ 21:37 < ido> http://www.iozone.org/ 21:37 < ido> try this 21:37 < ido> would be interesting to make a comparision 21:37 < ido> both on hosts and UML's. 21:37 < ido> on different hardware 21:38 < ido> mind trying that ? - any of you guys --- Log closed Sat Jun 28 00:00:00 2003