--- Day changed --- Log opened Thu Jun 05 00:00:01 2003 00:00 < Lathiat> that annoys me :P 02:02 < mistik1> anyone awake 02:02 < caker> mistik1: barely 02:02 < mistik1> hey caker 02:02 < mistik1> wget http://music.geeksinthehood.net:81/~mistik1/ifup-bridge-0.2.sh 02:03 < mistik1> I'll write up a readme and tar it up when I get home 02:03 < mistik1> or maybe just as comments 03:00 -!- snide [~snide@AMontsouris-108-1-14-194.w80-13.abo.wanadoo.fr] has joined #uml 05:08 -!- avoozzzl is now known as avoozl 05:24 -!- sssssssh____ [~chatzilla@68.17.231.210] has quit [Ping timeout: 488 seconds] 05:31 < Getty> mmhhh.. 05:31 < Getty> what ways you use for having some kind of "flexibel" filesystem for the uml? 05:31 < Getty> hostfs suxx, it has to less options for translating userrights 05:32 < avoozl> you could nfs i suppose 05:32 < Getty> bad... 05:32 < Getty> other plan please ;) 05:32 < avoozl> patch hostfs :) 05:32 < Getty> i don't wanna involve other host for this 05:32 < Getty> baeh... ;) 05:32 < Getty> what about LVM, files and so? 05:32 < avoozl> haven't tried --- Log closed Thu Jun 05 05:38:58 2003 --- Log opened Thu Jun 05 05:39:10 2003 --- Log closed Thu Jun 05 06:39:01 2003 --- Log opened Thu Jun 05 06:39:14 2003 07:43 < snide> i bt an hanging uml (when networking), and it seems that it hangs (eats 100% CPU) in munmap().... 07:54 -!- ced [~delaunoi@descartes.info.ucl.ac.be] has joined #uml 08:38 -!- Phazeman|AFK is now known as Phazeman 08:41 -!- arch [arch@68.51.49.60] has joined #uml 08:55 -!- litost [~mondain@131.96.158.138] has joined #uml 09:03 -!- ced [~delaunoi@descartes.info.ucl.ac.be] has quit [Quit: Client exiting] 09:11 -!- Cliff [none@63.93.88.67] has joined #uml 09:11 < Cliff> can anyone help me with networking and umld? 09:50 -!- Cliff [none@63.93.88.67] has quit [Quit: -=SysReset 2.53=-] 09:50 -!- jdike [~jdike@dhcp05.ists.dartmouth.edu] has joined #uml 09:50 < jdike> hi guys 09:51 < snide> jdike: hi 09:59 < ichilton> jdike: hi jeff 09:59 < ichilton> hi snide 10:17 -!- Phazeman [~Phazeman@217.132.3.195] has left #uml [Client exiting] 10:18 < green> Hi Jeff! 10:20 < snide> ichilton: hi. 10:20 < jdike> Hey Oleg 10:20 < snide> ichilton: i'm finishing my uml_mv and uml_bmv ;-) 10:20 < jdike> green: do you use xconfig? 10:21 < snide> jdike: btw, are there some COWv3 header specs ? 10:21 < jdike> snide: nothing formal 10:22 < jdike> snide: there's an implementation by James McMechan 10:22 < ichilton> snide: cool....what are they? :-) 10:24 < snide> ichilton: <-- u don't even know what it is ;-p 10:25 < snide> ichilton: actually, uml_mv is when u move a cow backend ( i posted it some time ago ) 10:27 < snide> ichilton: uml_bmv is when u want to switch a cow to a completely different backend [ for example, u have C1 & C2 COW files w/ the backend as B1. then u uml_moo C1 to have B2 as its backend. With uml_bmv u can also move C2 to use B2 as ur backend. ] 10:27 < jdike> snide: how does it differ from cp -p? 10:27 < green> jdike: No, I use menuconfig 10:29 < jdike> snide: that makes no sense 10:29 < snide> ichilton: it's maybe not so useful at first sight. but imagine that u have 20 uml that are almost the same, and u do a apt-get upgrade in each. Many blocks are dupes 10:29 < ichilton> snide: ahh, I remmber now! 10:29 < jdike> snide: uml_moo causes a COW file to get mashed into the backing file, so it is invalid afterwards (with -d anyway) 10:30 < jdike> snide: this is sort of a partial uml_moo? 10:31 < snide> jdike: depends of what u call "sort of", but it can be seens as it, yes. 10:32 < jdike> snide: i.e. take a set of COW files, and one backing file, and merge as much as possible from the COW files into the backing file? 10:32 < snide> jdike: it makes a brand new COW file (recalculates all the bitmaps sectors) 10:32 < snide> jdike: that's the goal 10:33 < jdike> snide: OK, cute 10:33 < snide> jdike: but for now, it only takes a moo'ed backfile. ;-) 10:33 < jdike> snide: with the apt-get example, you're sort of dependend on block allocation being similar in the UMLs 10:33 < Getty> hey green, how are you? :) everything cold out there? ;) 10:34 < snide> jdike: yup. i know. but i tested some time ago, and acutally it's actually not so random 10:34 < snide> jdike: hence the idea to do this ;-) 10:34 < green> Getty: Well, the snow have dissolved already, and the temperature is slowly getting higher which is good. Kind of dissapointing to have snow in June 10:34 < Getty> green: lol, thats russia ;) 10:35 < snide> green: too much secret climatic experiments there ;-))) 10:35 < green> Getty: still this is not Siberia ;) 10:35 < Getty> green: yeah you're right ;) 10:36 * green is preparing to go through the joys of replacing bad ram with something not bad 10:36 < snide> green: u know the badram kernel patch i presume 10:36 < green> seems there is no good ram on the market at all, whenever I buy soem RAM, I get bad one for first attempt 10:37 < green> snide: sure. But I paid money for good ram ;) 10:37 < green> snide: and I am not sure that badram patch works with highmem 10:37 < snide> green: u know... quality isn't anymore what is was ;-( 10:37 < snide> s/is/it/ 10:37 < green> yeah, I noticed that already 10:38 < snide> jdike: an even better way would be to detect the least changed block, and to merge that ;-) 10:39 < jdike> snide: least changed == most similar to what it was before? 10:40 < snide> ichilton: i do know why it didn't work my ndb-server patching attemps to make it recognise COW files... --> *u cannot use it on localhost* 10:40 < green> jdike: I se that Linus have not took your patches? Perhaps throw them at Andrew Morton? 10:40 < snide> jdike: no. to choose to most recurrent sector among the COW farm ;-) 10:41 < jdike> green: yeah, maybe 10:41 < jdike> snide: OK, that makes sense 10:42 < snide> jdike: btw, i do have a strange behavior... my uml hangs on heavily using the network. [ and most surprisingly, it dies on munmap ] 10:43 < snide> jdike: even more interesting is that i can use any uml version... ;) 10:43 < jdike> snide: it hangs or dies? 10:43 < snide> jdike: hangs. 10:43 < snide> jdike: and eating 100% host cpu 10:44 < snide> jdike: and when i kill it, it gives me something about being unable to close /proc/mm ;-( 10:44 < jdike> snide: have you looked at the stack? 10:44 * green seems to have permanent problem to get invitation letter from some canadian company to get canadian visa. Oh the crazy world 10:45 < snide> green: ;) 10:46 < snide> jdike: just a sec... i'll make crash another ;) 10:51 < snide> jdike: so, what should i do ? 10:51 < snide> jdike: gdb ... attach ... bt ? 10:52 < jdike> snide: tt or skas mode? 10:53 < snide> jdike: skas 10:53 < jdike> snide: running under gdb? 10:54 < snide> jdike: i attached gdb to the 100% pid 10:54 < snide> it gives me : 10:54 < snide> #0 0xa000207d in mmap () 10:54 < snide> #1 0xa0032fce in os_map_memory () 10:54 < snide> #2 0xa0016432 in map_memory () 10:54 < snide> #3 0xa0024508 in flush_tlb_kernel_range_skas () 10:56 < jdike> snide: OK 10:56 < jdike> snide: A bunch of people have hit this lately 10:56 < jdike> snide: David (and other people) have a fix 10:57 < snide> jdike: did i miss something on the ml ? 10:58 < jdike> snide: yeah, it's there too 10:58 < jdike> snide: a COW file rounding error 11:00 < snide> jdike: oh.... i do have something like that also : /dev/ubd/disc0:end != nsectors 11:01 < snide> snide: a COW file rounding error <-- that one tells me something... 11:12 < snide> jdike: i found the COWv3 implementation, but no filerounding ;-( 11:15 < jdike> snide: hmmm 11:15 < jdike> snide: see if you can wake up David 11:17 < snide> jdike: is it the 'There's a patch/bodge which seems to alleviate a buffer overrun in the COW...' ? 11:18 < jdike> snide: URL? 11:18 < snide> jdike: *bitmap_len_out = (size + 8*sectorsize - 1) / (8 * sectorsize); 11:18 < snide> jdike: in ubd_user:78 11:18 < jdike> snide: Maybe 11:19 < snide> jdike: yeah, i lookup in the thread, and the bt looks almost as ugly as mine ;-p 11:19 < snide> jdike: so, i'm quite confident... *checking: patching & recompiling* 11:30 < snide> jdike: when do u think the modules will be supported again ? [ i mean, is it in the upper or lower part of ur task list ;-pp ] 11:35 < jdike> snide: probably lower part right now 11:37 < snide> jdike: ok 11:45 < caker> jdike: hello 11:45 < jdike> caker: how's it going? 11:46 < caker> jdike: pretty well - have some ppl hammering my system - but networking is hosed - cant figure it out 11:51 < jdike> caker: hosed how? 11:51 < caker> jdike: I'm using bridge/tap devices - all works for a while (minutes to a day), then the IP is inaccessible - I believe after arp times out 11:52 < caker> if I log into console of the UML and start pinging stuff, the switch/host/something picks up the ARP/IP again and it works for a while longer... 11:58 < jdike> caker: arp is enabled on the bridge devices? 11:59 < caker> how do you enable arp? 11:59 < jdike> caker: if ifconfig doesn't say NOARP then it's on 12:02 < caker> jdike: its on :( 12:02 < caker> devices are in promisc (except bridge), proxy_arp is off for everything 12:03 < caker> tried with ip_forwarding on/off (no difference) 12:15 < Getty> strange strange 12:16 < jdike> caker: are arp who-has messages floating around when you try to reach an unreachable UML? 12:16 < caker> no 12:16 < caker> but I see other arp requests 12:16 < Getty> must uml_net run as root or as the uml user? 12:16 < caker> both outside and inside uml 12:17 < jdike> caker: they should be 12:18 < caker> jdike: I get arp responses if I'm inside my own network, but not if I'm pinging from outside (when the uml is unreachable) 12:18 < Getty> mh? 12:18 < Getty> i didn't get networking to run... strange.. its all the same like on my first uml but it doesn't work 12:19 < jdike> caker: when you ping from outside, does the gateway make who-has requests? 12:19 < caker> doesn't seem like it 12:20 < jdike> caker: well, that would be your first problem 12:20 < caker> jdike: ok, it did it that time 12:22 < Getty> hmpf... 12:22 < jdike> caker: then, if you see arp requests at the bridge device, and no responses go out, then the bridge isn't generating them 12:22 < jdike> Getty: uml_net has to be setuid root 12:22 < Getty> ok then this is right, too... 12:22 < Getty> where is the mistake then... 12:23 < caker> jdike: I see plenty of who-has requests. but only a handful of arp replies 12:23 < Getty> ip_forward is set to 1 12:23 < Getty> device is setup right (same config) changed ips to fit and host-ip as gateway 12:23 -!- smcavoy [~smcavoy@207.61.160.163] has joined #uml 12:24 < jdike> caker: are any of the arp replies the right ones? 12:24 < caker> jdike: yes, looks correct when it replies 12:25 < jdike> caker: and the reply is seen at the gateway? 12:25 < jdike> caker: and there's a new entry in its arp table? 12:25 < caker> I didnt't just check, but that's what I saw last night 12:25 < caker> My switch had to have it in the arp table before anything would work 12:30 < caker> ok .. I made the switch try and ping an UML's ip. I'm running tcpdump arp on the host, and inside the uml 12:31 < caker> I see the UML respond to this: arp who-has li-34.members.linode.com tell switch.linode.com 12:31 < caker> but on the host, I never see the response 12:32 -!- snide [~snide@AMontsouris-108-1-14-194.w80-13.abo.wanadoo.fr] has quit [Quit: going back home.....] 12:32 < jdike> caker: the switch is uml_switch? or a physical switch? 12:32 < caker> cisco 12:34 < jdike> caker: so the request made it through the bridge to the UML, the UML responded and the reply didn't make it back out through the bridge 12:34 < jdike> caker: ? 12:34 < david> jdike heya 12:34 < caker> jdike: that looks like it's correct .. 12:34 < jdike> caker: well, looks like the bridge is fscked 12:35 < caker> jdike: great :-) 12:37 < david> hello 12:37 < caker> hey 12:42 < Getty> i can't find the f*&%$ing mistake 12:43 < caker> Getty: chmod the tun device on host, create the tap device as the user that runs UML 12:43 < Getty> caker: tun device has correct rights (660 root.uml-net) 12:43 < caker> make sure you only have one uml_net installed - uml searches paths, i believe 12:43 < Getty> caker: tap device is generated by uml_net (which works correct like i see on host networks config) 12:43 < Getty> caker: i can ping the device in the UML 12:43 < caker> Getty: what err do you get? 12:44 < Getty> caker: and i can ping the UML ip on the host 12:44 < Getty> caker: but the UML can't get outside, he even can't ping the host 12:44 < Getty> ip_forward is 1 12:44 < jdike> Getty: does it have a route to the host? 12:45 < Getty> of course network setup is correct 12:45 < Getty> 194.77.163.64 0.0.0.0 255.255.255.192 U 0 0 0 eth0 12:45 < Getty> 0.0.0.0 194.77.163.125 0.0.0.0 UG 0 0 0 eth0 12:45 < Getty> .125 is the host 12:45 < Getty> .67 is the UML 12:45 < Getty> this is route table of UML 12:45 < david> Getty: does the host know how to route to the UML? 12:45 < jdike> Getty: what do you see when you tcpdump the tap device? 12:46 < Getty> david: yeah route on host is ok: 12:46 < Getty> 194.77.163.67 0.0.0.0 255.255.255.255 UH 0 0 0 tap1 12:46 < Getty> 194.77.163.64 0.0.0.0 255.255.255.192 U 0 0 0 eth0 12:46 < Getty> 0.0.0.0 194.77.163.65 0.0.0.0 UG 0 0 0 eth0 12:46 < Getty> jdike: let me check this 12:46 < david> what IP does tap1 have? 12:46 < Getty> tap1 Link encap:Ethernet HWaddr 00:FF:AB:2F:00:DF 12:46 < Getty> inet addr:194.77.163.67 Bcast:194.77.163.255 Mask:255.255.255.255 12:47 < Getty> (its generated by uml_net) 12:47 < david> that's wrong 12:47 < Getty> yeah i see that 12:47 < david> tap1 should have the IP of the host 12:47 < david> not the IP of the UML 12:47 < caker> yup 12:47 < Getty> mh? 12:47 < Getty> AH!!! 12:47 < Getty> yes! 12:47 < Getty> know i see my mistake! 12:47 < Getty> thx 12:47 < david> :-) 12:47 < Getty> nohup su - $UMLUSER -s /bin/bash -c "TMP=$UMLDIR/ram exec $UMLKERN \ 12:47 < Getty> mem=32M \ 12:47 < Getty> umid=$UMLNAME \ 12:47 < Getty> con1=pty con=null \ 12:47 < Getty> eth0=tuntap,,,194.77.163.67" \ 12:47 < Getty> >$UMLDIR/bootup.log & 12:47 < Getty> that was the error 12:48 < david> right 12:48 < Getty> ah ok, now next time it will go faultless i think 12:48 < Getty> now i have done all faults ;) 12:48 < jdike> Getty: if you looked at dmesg, you'd see it complaining about that 12:49 < Getty> mh 12:50 < Getty> i didn't looked into the kern.log/dmesg really 12:50 < Getty> but anyway i see there is no error into that 12:50 < jdike> Getty: should be 12:50 < Getty> Jun 5 17:27:55 metazone kernel: Using /lib/modules/2.4.18-686/kernel/drivers/net/tun.o 12:50 < Getty> Jun 5 17:27:55 metazone kernel: * ifconfig tap1 194.77.163.67 netmask 255.255.255.255 up 12:50 < Getty> [...] 12:50 < Getty> Jun 5 17:27:55 metazone kernel: * arp -Ds 194.77.163.67 eth0 pub 12:50 < Getty> Jun 5 17:55:14 metazone kernel: Virtual console 1 assigned device '/dev/ptyp0' 12:51 < Getty> no error in the [...] 12:51 < caker> jdike: i was wrong, the arps are getting from the uml, to my switch .. when I do "clear arp-cache" on the switch, it asks for everything again, and get's responses 12:51 < caker> jdike: perhaps it's time to call my provider? 12:51 < jdike> Getty: it would have been later, at UML ifconfig time 12:51 < Getty> jdike: there is nothing in the kern.log..... really :) 12:52 < jdike> printk("The tap IP address and the UML eth IP address" 12:52 < jdike> " must be different\n"); 12:52 < Getty> hehe 12:52 < Getty> yeah show me the source ;) 12:53 < ASY> david: could i ask you for a snippet of the netdef table... 12:53 < Getty> jdike: could be that he doesn't write that into kern.log 12:54 < jdike> Getty: everything else made it 12:54 < Getty> strange strange 12:54 < Getty> anyway, going on to next point: sparse files 12:55 < Getty> but i will go on this point at home, with some nice smoke ;) 12:56 < david> ASY: the struct or the content? 12:56 < ASY> content 12:56 < david> INSERT INTO netdef VALUES ('maeve','public',0,'a040700',24,1,1,168036097); 12:56 < ASY> thanks 13:05 < ASY> still need some help :) bridge... is this a bridge id? ipblk? what does 24 mean in this context. lets say i have 192.168.0.x network .1 is gw and .2-.128 is my free range... 13:11 < david> ASY: bridge is the name of the bridge interface 13:11 < david> ipblk is the size of your netblock in CIDR notation (24 = /24 = 255.255.255.0) 13:12 < david> ipcnt appears to be unused at the moment 13:13 < ASY> k 13:13 < david> ipskip is the number of IPs at the start of the netblock to skip 13:13 < david> if it's one, then the first IP used by UMLd is .2 13:13 < david> (e.g. if .1 is the gateway 13:13 < ASY> k. understood. thanks. 13:14 < ASY> i was wondering if you have any plans on putting this on CVS and if there is a way to report bugs if any found etc. 13:15 < david> I might put it on sourceforge eventually 13:15 < david> I've not worked on it much recently due to having to spend my time doing other things 13:15 < ASY> i know you have a house to take care of :) we are buying one... moving in 1st of july :) 13:16 < david> wow, cool 13:16 < david> I've been working on running cables in the house these last few weeks 13:16 < david> quite a time drain 13:17 < ASY> 802.11a is gonna help me for a while... 13:17 < ASY> i am basicaly intending to use umld full blast here. so you will be hearing from me quite a bit :) 13:18 < ASY> it is quite a nice system... i just wish we did not have to completely reverse engineer it :) 13:19 < ASY> once i am setup i will throw my notes in your direction. 13:40 < david> ASY: documentation is on my list, somewhere :-) 13:40 -!- sjj [~sjj@64.142.27.18] has joined #uml 13:41 * jdike pulls the COW stuff into a separate file 13:41 < jdike> preparation for COW3 14:02 < david> jdike: any change of a command-line COW creation tool? 14:03 -!- radical [~kernet@202.88.184.39] has joined #uml 14:03 < jdike> david: not from me any time soon 14:03 < jdike> david: it'd be easy enough, though 15:06 < caker> Is the host the "default gateway" for the UMLs when using a bridge? 15:08 < jdike> caker: don't think so 15:08 < jdike> caker: it should have the same gateway as the host 15:08 < caker> On the phone with my datacenter .. they had suggested that 15:18 -!- david [~david@207.166.203.138] has quit [Ping timeout: 488 seconds] 15:37 -!- david [~d@207.166.203.138] has joined #uml 15:39 < jdike> new utilities release on the way 15:45 < david> jdike: can you post it on uml.org? 15:45 -!- david changed the topic of #uml to: User Mode Linux | 2.4.20-5um, 2.5.69-1um, 2.4.21-rc7-djc1-5um | http://user-mode-linux.sf.net | http://usermodelinux.org | http://usermodelinux.co.uk | http://uml.openconsultancy.com | http://www.stearns.org/slartibartfast | #uml Archives: http://www.theshore.net/uml/irc/ 15:47 < jdike> david: yup, there's an associated UML release - I'll post them both 15:47 < david> jdike: ok, cool 15:48 < david> jdike: anything other than cow changed in it? 15:48 < jdike> david: tt-mode lcall syscall disablement 15:49 < david> jdike: ah, ok 15:50 < david> jdike: I'll try to get a binary out tonight 15:59 < smcavoy> I'm having some troubles booting the Debian 3.0 image (downloaded from sourceforge). I keep getting 'fsck.ext2: No such file or directory while trying to open /dev/ubd/0' 16:00 < smcavoy> but the system seems to have mounted the drive '/dev/ubd/0 on / type ext2 (rw)' 16:00 < smcavoy> any ideas? 16:00 < BB1> do you have a device /dev/ubd/0 ? do you need devfs mounted? 16:01 < smcavoy> well mount does report it as mounted, I executed with ubd0=[path to fs] 16:01 < smcavoy> BB1: hmmm devfs seems to make sense 16:02 < smcavoy> BB1: what's the proper fstab entry for it? 16:03 < BB1> smcavoy devfs /dev devfs defaults,noauto 0 0 ? 16:03 < BB1> u might not want the noauto i guess 16:03 < smcavoy> cool. 16:04 < BB1> brb just making sure the black smoke from the kitchen is just food and not the furniture 16:04 < jdike> smcavoy: that's probably too late 16:04 < jdike> smcavoy: devfs=mount on the command line might help 16:05 < smcavoy> jdike: that did it. 16:05 < smcavoy> jdike: so the kernel doesn't automatically try and mount it? 16:07 < jdike> smcavoy: it's a config option 16:14 < smcavoy> it doesn't seem to be recognizing the tun device, I've setup a bridge, added 1 of the host ifs to it, created a tun device, added it and configured it with 0.0.0.0 promisc up 16:14 < smcavoy> any idea why it would cause a problem? 16:15 < david> is /dev/net/tun 666? 16:16 < david> and define 'recognising' 16:16 < smcavoy> yes 16:17 < smcavoy> when I boot into the uml, I cannot configure eth0. I get 'eth0: ERROR while getting interface flags: No such device' on boot up 16:18 < smcavoy> ... I think I missed something in my start up command :) 16:19 < david> what is your command line? 16:19 < david> you want something like 16:19 < david> eth0=tuntap,yourtapdevice 16:19 < david> did you make a device with tunctl? 16:19 < david> jdike: btw, tunctl is misleading -Maybe it should be renamed tapctl, since it creates a tap device, not a tun device 16:19 < smcavoy> I completly missed eth0= 16:19 < david> that'd do it 16:23 < smcavoy> hmm... now that I added it, the packets don't seem to be passing-thru the bridge... should the host bridge device be assigned an IP on the LAN? 16:24 < smcavoy> tcpdump shows the device is sending arp requests for the IP I tried to ping from the uml.. 16:26 < david> smcavoy: if you have eth0, for example 16:26 < david> you give the IP from eth0 to the bridge 16:26 < david> remove the IP from eth0 16:26 < david> add eth0 to the bridge 16:26 < david> and the host will work as normal 16:29 < jdike> david: maybe but no one else has complained :-) 16:30 * BB1 is having a bad day.. burnt food.. went to shop to buy replacement and came back with a bottle of vodka and no food .. 16:31 < david> jdike: btw, when you mirror stuff to my box, make sure you chmod 644 it because umask is set to 077 16:31 < jdike> BB1: vodka 16:31 < jdike> BB1: sounds like food to me 16:31 < jdike> BB1: or at least you'll forget about being hungry 16:31 -!- sjj [~sjj@64.142.27.18] has quit [Quit: leaving] 16:32 < jdike> david: rsync -p will override umask, right 16:32 < jdike> ? 16:32 < BB1> jdike heh good plan, untill the flatm8 comes back and smells the burnt smell all over 16:33 < david> jdike: should do 16:33 * jdike fixes the Makefile 16:38 < david> jdike: the UML build process is -j safe, right? 16:38 < jdike> david: should be 16:39 < david> jdike: nope, isn't :-( 16:39 * BB1 found that out heh 16:39 < jdike> david: send the log 16:39 < david> jdike: let me finish compiling my kernel 16:39 < david> jdike: have you sent out the -um6 announcement? 16:39 < jdike> david: not yet 16:40 < david> jdike: http://mirror.usermodelinux.org/uml/uml-patch-2.4.20-6.bz2 16:41 < david> didn't work, now it does 16:41 < jdike> david: what about that? 16:41 < jdike> david: that's the patch I just released, right? 16:41 < david> right 16:41 < david> but I couldn't download it from there because the permissions were screwy 16:53 < jdike> Sent out the announcements 17:10 < david> wooh 17:10 < david> got bins built 17:13 -!- avoozl [~jrk@vengeance.et.tudelft.nl] has left #uml [bbl] 17:18 -!- smcavoy [~smcavoy@207.61.160.163] has quit [Quit: ACK!] 17:24 < jdike> david: posted on uml.org 17:26 -!- jdike [~jdike@dhcp05.ists.dartmouth.edu] has quit [Quit: Leaving] 17:28 < Pahan> How come I can't have two UML machines using same tap device? 17:29 < Pahan> I want the host and all UML machines to have a private segment. 17:29 < Pahan> When Debian attempts to bring eth0 up on the second machine, I get: 17:29 < Pahan> Configuring network interfaces... TUNSETIFF failed, errno = 16SIOCSIFFLAGS: Device or resource busy 17:29 < Pahan> TUNSETIFF failed, errno = 16SIOCSIFFLAGS: Device or resource busy 17:30 < caker> Pahan: give each UML its own TAP, and put all the TAPs onto a bridge 17:31 < Pahan> The host lacks bridging support and I am reluctant to install a kernel remotely. 17:31 < Pahan> Why can't a tap be shared? It's really weird. 17:32 < BB1> they are point to point 17:33 < Pahan> I thought they emulated an Ethernet segment? :"( 17:34 < Pahan> The kernel certainly considers tap devices broadcast links. 17:36 < Pahan> Ooooh, looks like the switch daemon connects to a tap device. 17:42 < caker> This arp stuff is driving me bonkers 17:43 < Pahan> What is wrong with it? 17:44 < caker> long story.. using bridge, the IPs of the UMLs work for a while, but then don't - I must ping them in my internal network before they start working again 17:44 < caker> they all have public IPs 17:44 < caker> same subnet 17:45 < caker> I'm now wondering if my provider's switch has proxy arp disabled, or if that even matters 17:45 < Pahan> Proxy ARP should be off. There is no reason for it to be on. 17:45 < BB1> why do you need to do anything with arp if your using bridge? 17:45 < caker> Pahan: thought so .. 17:46 < caker> BB1: the ips aren't accessible after some time, and it's only until some arp traffic goes around that they start working again 17:46 < Pahan> caker: Where are you trying to access them from? 17:46 < BB1> caker i dont do with arp and mine seem to be working .. strange 17:47 < caker> I've got four places: a machine next to the one with the problem (actually both have the problem), the switch the two machines are connected from, another machine in NJ, and my machine here in nashville 17:48 < Pahan> Check to see whether your gateway is sending ARP requests. 17:49 < caker> If I attempt to ping one of the inaccessible IPs from *outside*, I get no arp traffic. 17:49 -!- litost [~mondain@131.96.158.138] has quit [Quit: Run Away!! Run Awaayyyy...] 17:49 < Pahan> Broken router. 17:49 < caker> if I ping from inside (switch, or one of the machines) the arp traffic goes around and then everything works 17:50 < caker> I get no arp traffic from the gateway to either of the two machines - I don't know how to tcpdump on a cisco switch to see if it's going to the switch 17:50 < caker> (my switch) 17:50 * BB1 says it works... then checks and his UML machine has crashed lol 17:50 < Pahan> Check your IP addressing. If the gateway is not sending ARP requests, it doesn't think the destination machine is on the interface. 17:51 < caker> The datacenter people said they have a direct IP route to my switch, so she said it doesn't involve ARP until it hits my switch/machines 17:51 < caker> ?? 17:53 < Pahan> What's the router interface IP address and subnet masks and what addresses/masks are you using on UML machines? 17:54 < caker> the gateway from my datacenter is a .1 address, 255.255.255.0, my switch is .4/255.255.255.0, and the UMLs are configured to use a .20-90 IP with 255.255.255.0 netmask 17:54 < caker> all my machines use the .1 as the gw 17:54 < caker> all public IPs 17:55 < Pahan> Oh, freaking weird. 17:56 < Pahan> Do you have access to the router? 17:57 < caker> Not theirs, no 17:57 < caker> I'm thinking the IP doesn't work until they have an entry in the arp-cache on their router 17:57 < caker> because I can clear out the arp (and mac) cache on my switch, and stuff keeps pinging (once I get it to work) 17:57 < caker> their arp timeout is 2min 17:58 < Pahan> Hrm. So the router sends ARP requests for all physical machines but not for UMLs? 17:58 < caker> I can only see what gets to my machines, so I'm not sure what's getting to the switch, but yes. 17:59 < Pahan> ARP requests are broadcast. You should see all of them. 17:59 < caker> I agree 17:59 < Pahan> Well, there's a stupid and inappropriate solution, of course. 18:00 < caker> write a cron job to ping everything? 18:00 < Pahan> arpspoof from dsniff ran on each UML machine to advertise its MAC address :P 18:02 < caker> What if it was going to their gateway, but it was my switch? 18:03 < Pahan> switches don't care about ARP 18:03 < Pahan> Hrm. 18:04 < Pahan> Switches shouldn't care about multiple MAC addresses appearing on a single interfaced, unless there's too much free time on admin's hands. 18:04 < caker> well, that's not me, so.. 18:05 < caker> my switches config is about as bare-bones as you can get 18:12 < caker> just called them -- their gateway DID have an entry, and the correct one, at that 18:12 < caker> so looks like the problem is on my end 18:12 < Pahan> Hrm. 18:12 < caker> I think the HOST is is unaware 18:13 < caker> once the host knows the arp, it works 18:16 < BB1> shoudlnt the uml do its own arp reply? 18:17 < caker> spamming: 18:17 < caker> 18:13:24.622032 arp who-has li-60.members.linode.com tell switch.linode.com 18:17 < caker> 18:13:24.622321 arp reply li-60.members.linode.com is-at fe:fd:40:5:35:3c 18:17 < caker> 18:13:25.594843 arp who-has gateway.linode.com tell li-60.members.linode.com 18:17 < caker> 18:13:25.595126 arp reply gateway.linode.com is-at 0:0:c:7:ac:1 18:17 < caker> 18:13:29.464056 arp who-has 64.5.53.100 tell up1.linode.com 18:17 < caker> 18:13:30.809362 arp who-has switch.linode.com tell li-60.members.linode.com 18:17 < caker> 18:13:30.857291 arp reply switch.linode.com is-at 0:3:e3:47:57:40 18:18 < caker> that's the output from "tcpdump arp" when I ping'ed the IP from the switch - then it started working again 18:19 < caker> hmmm 18:19 < caker> br0 Link encap:Ethernet HWaddr 00:30:48:51:B9:89 18:19 < caker> inet addr:64.5.53.6 Bcast:64.255.255.255 Mask:255.255.255.0 18:19 < caker> bcast is wrong, no? 18:19 < BB1> looks wrong to me 18:21 < mistik1> definately 18:21 < mistik1> still having this problem I see caker 18:22 < caker> well, that took networking down 18:24 < mistik1> changing the bcast 18:24 < BB1> :/ 18:24 < caker> someone who uses a bridge show me what their broadcast is? 18:24 < mistik1> ack! I dont have one setup 18:24 < mistik1> lemme fire one up somewhere 18:24 < BB1> inet addr:217.158.138.163 Bcast:217.158.138.191 Mask:255.255.255.224 18:25 < caker> my machine is not very happy right now 18:25 < BB1> i would expect yours to be 64.5.53.255 if your on a /24 18:25 < caker> yeah 18:26 < mistik1> inet addr:192.168.0.1 Bcast:192.168.0.255 Mask:255.255.255.0 18:27 < mistik1> that's what I get from one a just brought up 18:27 < caker> thanks BB and mistik1 18:27 < mistik1> np 18:37 < BB1> cool i had no complaints of uml machine being down for like 9hrs heh 18:38 -!- radical [~kernet@202.88.184.39] has quit [Remote host closed the connection] 18:38 < caker> I don't need any routes (besides default gw) or crazyness inside the UMLs when you use a bridge, right? 18:39 < BB1> nop 18:51 < caker> ARP 18:52 < BB1> bARP 19:07 -!- pwagland [~paul@igloo.xs4all.nl] has joined #uml 19:08 -!- pwagland [~paul@igloo.xs4all.nl] has quit [Remote host closed the connection] 19:18 -!- david [~d@207.166.203.138] has quit [Ping timeout: 488 seconds] 19:21 < caker> um .. what if I compiled in VLAN support onto the bridge 19:21 < caker> er, I mean, what if I take it off :-) 19:22 < BB1> shouldnt make any diff 19:22 < caker> well, arp 19:22 < caker> arp is my new curse word 19:22 < BB1> heh 19:26 -!- sssssssh_____ [~chatzilla@68.154.235.182] has joined #uml 19:26 < mistik1> arp you too man! 19:26 < mistik1> could work ;) 19:26 < caker> stick it up your arp 19:27 < BB1> arphole 19:27 < mistik1> lol 19:28 < caker> that's where all my arps are going .. into the arphole 19:28 < BB1> ;) 21:07 -!- BB1 [~chris@217.79.122.210] has quit [Ping timeout: 492 seconds] 21:37 -!- Z-Wing [zedders@zanzar.net] has quit [Ping timeout: 490 seconds] 22:46 -!- david-roam [~d@207.166.203.138] has joined #uml 22:46 -!- david-roam is now known as david 22:49 < david> hello 22:52 < caker> hello 23:00 -!- BB1 [~chris@217.79.122.210] has joined #uml 23:26 < caker> since I've been trying to fix this all day, I narrowed it down to the MAC address table (not ARP) on the switch times out, and then the IP is inaccessible 23:53 < david> strange 23:53 < david> is something blocking ARPs? 23:53 < caker> nope 23:54 < caker> if I clear the MAC cache on the switch, I can make the IPs die 23:55 < caker> I tried a newer kernel, now about to try an old one (perhaps even redhat 9's)... I'm about to resort to writing a cron job that pings all the machines every two minutes --- Log closed Fri Jun 06 00:00:00 2003