--- Day changed --- Log opened Wed May 28 00:00:02 2003 00:01 < caker> i've wasted days trying to get rid of a kernel bug since 2.4.19-pre 00:01 < david> caker: what is it? 00:01 < caker> i think i'm going to revert to 2.4.18 00:02 < caker> it causes the kernel to "pause" for seconds to minutes when performing heavy I/O .. when the machine is paused, NOTHING is going on 00:02 < caker> check out #elevator 00:02 < caker> so, now I'm faced with getting bridging/netfilter/skas to work in 2.4.18 00:04 < caker> david: happen to remember which ebtables patch and bridge_nf patches apply correctly to 2.4.18? 00:04 < david> caker: why not just use a -ck kernel then? 00:05 < david> caker: can't you just get rid of the patches between 2.4.18 and 2.4.19pre1 which 'break' this thing? 00:09 < caker> david: me and a bunch of other people have tried to find the offending patch .. but it's too integral to undiff 00:09 < caker> david: what's the deak with -ck? 00:09 < caker> s/deak/deal 00:10 < david> caker: the site says that the latency2 patch cleans it up somewhat 00:10 < david> http://members.optusnet.com.au/ckolivas/kernel/elevator.html 00:11 < caker> david: ok, the ck patch the read_latency2 patch doesn't fix it 00:11 < caker> david: i've tried all of them 00:12 < caker> david: there is no known fix currently .. also, check out the lkml thread: 00:12 < caker> http://www.ussg.iu.edu/hypermail/linux/kernel/0305.3/0509.html 00:14 < david> hrm 00:22 < david> http://linux.bkbits.net:8080/linux-2.4/diffs/drivers/block/ll_rw_blk.c@1.29?nav=index.html|ChangeSet@-2y|cset@1.160|hist/drivers/block/ll_rw_blk.c 00:22 < david> looked at that 00:23 < caker> yeah, now I'm checking out "fix-pausing-2" : 00:23 < caker> http://lists.insecure.org/lists/linux-kernel/2003/Mar/att-3040/fix-pausing-2.patch 00:25 < caker> david: that's a nice link, btw, thanks 00:25 < david> that is 'apparently' the offending patch 00:26 < david> although it wouldn't reverse in 2.4.20 00:26 < caker> too many patches after it, i think 00:26 < david> probably 00:28 < david> caker: you might want to ask Jeff to port host-skas3 to 2.5 00:29 < david> caker: I assume he ported the arch/um components to 2.5 00:29 < caker> david: he'd probably just say he's going to wait until Uber-Skas 00:29 < caker> david: but I'll inquire next time I see him 00:29 < david> ok 00:31 < mistik1> *sigh* 00:31 < david> mistik1: problems? 00:31 < mistik1> not at all 00:31 < mistik1> Just listening to you guys go at it :) 00:31 < david> mistik1: not really 'going at it' 00:31 < mistik1> like a sesspool of knowledge 00:31 < david> heh 00:32 < mistik1> Its all good man :) 00:32 < david> mistik1: doewsn't seem to fix the problem though 00:33 < mistik1> hey is that rc3 patch or source available or just the binary I saw in the ftp 00:33 < mistik1> err, www 00:36 < david> mistik1: for UML? 00:36 < mistik1> yea 00:36 < david> mistik1: I distribute a patch against 2.4.21-rc3 00:36 < mistik1> ok 00:36 < david> working on -rc5 now 00:37 < mistik1> wooo 00:38 < mistik1> I have been running some test since 2 days now and from what i've read, I really need to rebuild the kernel on the host with the skas patch 00:38 < caker> damn .. redhat's latest 2.4.18 kernel contains the ptrace fix :( I'm using mod_skas, so I guess I could patch the one-liner into it! 00:38 < david> mistik1: right 00:38 < caker> I'm probably nuts for using redhats 2.4.18 in the first place 00:38 < caker> but it's got ebtables and netfilter 00:39 < mistik1> I wondow how the skas-3 with go over a mandrake kernel 00:39 < mistik1> s/with/will/ 00:39 < caker> mod_skas rocks 00:39 < mistik1> mod_skas? 00:39 < caker> (for situations like these) 00:40 < caker> skas as a module, so you can build it with any kernel source tree 00:40 < caker> and ... insmod skas 00:40 < caker> :-) 00:40 < mistik1> hmmmm 00:40 < david> caker: do you have the patch for that? 00:40 < david> or, rather, the code? 00:40 < caker> yup 00:40 < caker> i couldn't find a link to it anymore 00:40 < david> caker: can you stick it somewhere so I can put it on kernels.usermodelinux.org? 00:40 < caker> but I had it stashed away 00:40 < caker> lemme find 00:41 < mistik1> that would be far preferable to me right now 00:41 < mistik1> what version does mod_skas support 00:41 < mistik1> skas version that is 00:41 < david> mistik1: it's based on host-skas3 00:41 < caker> skas 3 .. you'll have to modify it for the ptrace fix fix .. I might have that done in a second... 00:42 < caker> http://www.rogerbinns.com/modskas3/ 00:42 < david> caker: you need to fix the ptrace fix anyway - It breaks too much stuff to leave a box running like that 00:43 < mistik1> thanks caker 00:43 < caker> david: i agree 00:43 < caker> mistik1: thank Roger Binns ! :-) 00:43 < mistik1> :) 00:44 < david> mistik1: make sure you read the instructions 00:45 < caker> here's the patch for mod_skas for the ptrace fix fix: 00:45 < caker> *** proc_mm.c 2003-05-28 00:32:25.000000000 -0400 00:45 < caker> --- ../mod-skas3/proc_mm.c 2002-12-18 04:01:35.000000000 -0500 00:45 < caker> *************** static int open_proc_mm(struct inode *in 00:45 < caker> *** 136,138 **** 00:45 < caker> 00:45 < caker> - mm->dumpable = 1; 00:45 < caker> spin_lock(&mmlist_lock); 00:45 < caker> --- 136,137 ---- 00:45 < caker> actually, i did that the wrong way :-) 00:45 < caker> muhahahaaaaaaaaaa 00:45 < david> heh 00:46 < david> caker: wow, going to be complicaed to reverse that one ;-) 00:46 < caker> hehe 00:46 < caker> here is the corrected patch, just for fun: 00:46 < caker> *** ../mod-skas3/proc_mm.c 2002-12-18 04:01:35.000000000 -0500 00:46 < caker> --- proc_mm.c 2003-05-28 00:32:25.000000000 -0400 00:46 < caker> *************** static int open_proc_mm(struct inode *in 00:46 < caker> *** 136,137 **** 00:46 < caker> --- 136,138 ---- 00:46 < caker> 00:46 < caker> + mm->dumpable = 1; 00:46 < caker> spin_lock(&mmlist_lock); 00:47 < david> ok, time for bed 00:47 < caker> david: nite david, thanks 00:48 < caker> mistik1: if you get any "unresolved symbols" errors, add the function to the "syms" file 00:51 < mistik1> ok 01:38 < caker> someone else here was talking about the pauses in 2.4.20 .. 01:38 < caker> fix-pausing-2 patch seems to fix it 01:41 -!- mistik1 [rasta@68.192.32.70] has quit [Read error: No route to host] 01:51 -!- mistik1 [rasta@ool-44c02046.dyn.optonline.net] has joined #uml 01:52 < mistik1> ok host all patched up and rebooted 01:52 < mistik1> :) 01:53 < caker> mistik1: what did you do? 01:53 < mistik1> I decided to go with the patch rather than the module 01:53 < caker> mistik1: always a better choice :-) 01:54 < mistik1> I run redhat but I dont run the redhat kernels anyway :0 01:55 < green> hehe 02:00 < mistik1> hey caker, what is the best method to use if you have a single machine(host) that is using eth0 to connect to a DSL (so routable IP) and would like to have networking in your UML 02:01 < caker> mistik1: probably the net helper thing; although I've never used it 02:01 < mistik1> *nod* 02:03 < green> probably the NAT on host and tuntap networking should be the best ;) And eay. Or if you have separate routable IP for uml then tuntap networking + proxyarp on host 02:06 -!- Aike [~aikedejon@217.122.0.97] has joined #uml 03:27 * BB whistles as 2 servers he rebooted last night nevercame back 03:27 * caker finds remote serial console invaluable 03:28 < BB> its at a customers office ;) 03:31 * green remotely upgraded rh 7.3 to rh9 during last night. Haev not tried reboot yet, though ;) 03:32 < BB> heh 03:32 < BB> i'd managed to screw slip somehow 03:33 < green> remote console it truely invaluable in all such cases 03:33 < BB> heh i really cant see them paying tho... so its tough 03:34 < caker> I absolutely love the Baytech equipment 03:35 < caker> their remote power control/remote serial combo unit RPC-2 goes for around $100 usd on ebay 03:35 < caker> 4 power ports, 4 serial ports, 1 ethernet port. 03:36 < BB> heh they wouldnt buy anything second hand incase it broke.. been there b4 03:37 < caker> well they can pay $500 for new ones then :-p 03:37 < caker> I'd rather have 5 used ones, blah blah blah .. 03:37 < caker> people are weird like that, though - I don't get it 03:37 < BB> but then u hit on that being too expensive , we got them to buy an apcpower thing and a moxa serial console for data center luckily tho 03:37 < caker> nice 03:38 < BB> had to use both so far heh 03:38 < caker> My machines are about 1200 miles away - I recompiled/rebooted about 15 kernels today 03:39 < caker> all experimental 03:39 < caker> There's no other way to go :-) 03:40 < BB> heh 03:40 < green> kernels are easy. you have "lilo -R" 03:40 < BB> yeah 03:40 < BB> altho i have nfi how you do that in grub 03:41 < green> do not use grub. period 03:41 * green seen more than one time that grub declined to recognise valid fs as fs with kernel image after unclean shutdown 03:42 < caker> strange, considering your /boot partition would rarely be dirty .. 03:42 < BB> i dont just have a m8 with errors i think might be kernel related, so i said just boot other one temp .. turns out he has grub so i left him to it 03:42 < caker> grub over serial is pretty simple 03:43 < BB> heh i think he has remote power no serial tho 03:43 * BB likes serial bios :) 03:43 < green> caker: when you are trying testing kernels it well may be dirty ;) 03:43 < caker> green: that's the only thing I could think of :-) 03:44 < green> caker: also grub over serial will help you very little if grub itself says that it does not recognise the fs ;) 03:44 < caker> well, that's why I keep a duplicate /boot partition on the mirror drive synced, just in case :-) 03:44 < caker> only takes a few MBs 03:45 < caker> which reminds me .. i should go sync it 03:45 < green> well, if it is mirror and synced, it will be in the same state then ;) 03:45 < caker> synced means me copying it by hand .. :-) 03:45 < green> ah ;) 03:45 < caker> but it's got at least a kernel or two that would get me going 03:46 * green sticks to lilo that does not claims it is super smart and just does the job 03:47 < caker> yeah .. just going along the whole remote-rescue theme... 03:47 < caker> I also thought about burning a remote-rescue CD-rom, and placing it in each server. Of course, telling the bios to boot off cd last, until there is a problem 03:49 < green> won't work 03:49 < caker> I can access BIOS over serial port, OR use the boot loader 03:49 < green> of the boot loader is present it wil ltry to boot first, fail to load kernel/hang and no control gets to CD 03:49 < green> accessing bios over serial p[ort might wrk indeed 03:49 < caker> yup 03:50 < caker> it rocks 03:50 < green> the only thing that I seem to still like in sun hardware is that all of it seems to be able to talk on serial port by default if no keyboard is attached 03:51 < caker> that's cool... I never knew how to make the kernel do it until recently.. fairly easy! 03:51 < BB> heh i have this monster sun box with a tvi9110 ontop ;) 03:52 < caker> I also thought.. if you have a bunch of machines, you could daisy chain PC1:COM1 to PC2:COM2, PC2:COM1 to PC3:COM2, etc.. 03:52 < caker> you wouldn't need an access server, just a neighboring machine that's running 03:52 < caker> and a few serial null cables 03:52 * green was restoring corrupted disklabel on some sun box remotely, and the box did not even boot because of corrupted disklabel. And if that was a PC, that would become complete nightmare 03:53 < green> and those smart sun guys even store on disk a copy of disklabel. Nobody seems to store copy of mbr on pc 03:53 < BB> heh 03:53 < BB> with a raq3 you even get a kernel in flash ;) 03:54 < BB> but sun just stole that from cobalt :P 03:54 < green> BB: I had installation CD in CD drive, so just booted off that in single mode 03:55 < BB> green heh, i was planning on doing that and leaving it for the redbus guys to put in if i needed it 03:55 * green actually misses the point why people still spend lots of money on those used E250/E450 instead of buying cheap PC (p4 based for example) that will greatly outperform those sun boxes 03:56 * BB has an ultra1 and and antique 670MP 03:56 < caker> a few years ago I did some work back at my local high school - They were sold some SGI Origin 200s for about $30,000... they asked me a few weeks ago if I wanted them for free now! 03:56 * green have seen only one sun box in person. that's my old sparc classic ;) 03:56 < green> caker: if they don't - I want it. No, I do not want to pay postage ;) 03:56 < BB> heh 03:57 < caker> no kidding 03:57 < caker> I looked up the cost for them on ebay - they're going for like $200 bucks! 03:57 < BB> green we used to have IPX at uni 03:57 < green> caker: if they pay for the shipping, I can get it ;) 03:57 < caker> the only thing they did on them was fileserving for about 4000 students, and run my perl script that I wrote 03:58 < green> BB: hehe. I run linux on my classic. Also this is the only linux on sparcs I have access to these days, other people still choose solaris for unknown reason 03:58 < caker> I believe I am retrieving them in a few weeks, sorry mates 03:59 < BB> green i had linux on the ultra1 until it refused to boot one day. the 670mp has debian *shudder* on it atm 04:00 < green> hehe. 04:01 < green> so you reverted to solaris on ultra ? ;() 04:06 < BB> green nah, its sat here until i have more time to hook up something to the back of it 04:06 < BB> it will only work on a monster old 17" i have, which is burried somewhere in the room 04:07 < green> рурую 04:07 < green> hm. 04:07 * green have this nice sun monitor->vga monitor convertor on the desk 04:08 < BB> heh i have a converter but it doesnt help the resolution 04:08 < green> sure. but at least you can use some modern monitor ;) 04:08 < BB> it is semi modern it just weighs a ton 04:08 < green> hehe 04:08 < green> so just hookup serial console instead ;) 04:11 < BB> heh 04:11 < BB> it took pride place in the back of the raq3 it needed it daily :P 04:11 * green misses those days when I was just a sysadmin 04:15 < BB> blah i'm going for a shower this can wait until i get to work 04:34 -!- ElectricElf [david@elf.noc.oftc.net] has quit [jupiter.oftc.net ozone.oftc.net] 04:35 -!- ElectricElf [david@elf.noc.oftc.net] has joined #uml 04:49 -!- ElectricElf [david@elf.noc.oftc.net] has quit [Remote host closed the connection] 04:50 -!- ElectricElf [david@elf.noc.oftc.net] has joined #uml 04:55 * BB2 wodners how 2 routes can dissapear overnight from a server that hadnt been rebooted 05:00 < green> interface down/up will achieve the same result 05:39 < Getty> he green you there? 05:40 < Getty> "ah its not easy being green" 05:40 < Getty> i must sample this part of the shrek interview for him.... 05:41 * green is here. sort of 05:42 < Getty> hey green, how are you? 05:42 < green> I am fine, thank you. 05:43 < green> (hm, this dialogue promptly reminds me my english classes at school ;) ) 05:43 -!- Electric1lf [david@CPE0040f6b429d8-CM400049226055.cpe.net.cable.rogers.com] has joined #uml 05:43 < Getty> green: lol 05:43 < Getty> green: but i'm really interested anyway :) 05:43 -!- ElectricElf [david@elf.noc.oftc.net] has quit [] 05:44 -!- Electric1lf is now known as ElectricElf 05:44 < Getty> green: wanna explain me something? :-) 05:44 < green> well, is it something interesting? ;) 05:44 < Getty> green: or other question first, ever worked with tmpfs/ramfs on linux? 05:44 < Getty> green: mmhh.. memory in UML 05:45 < green> tmpfs/ramfs works just fine ;) we do some benchmarks with it ;) 05:45 < Getty> can i use it without reboot the machine for adding kernel parameters? 05:45 < Getty> (the host) 05:46 < green> it == what? 05:46 < Getty> tmpfs/ramfs 05:46 < green> if you have this stuff compiled in, then you can use it rught now. RAMFS is compiled into 2.5 by default and there is no way to tur that off, I think 05:46 < Getty> we have 2.4.20 05:47 < Getty> let me check dmesg, should be a message there, right? ;) 05:47 < green> then check your .config if you have it enabled 05:47 < green> no. 05:47 < Getty> debian kernel 05:47 < green> better check /proc/filesystems 05:47 < Getty> nodev ramfs ? 05:47 < Getty> nodev tmpfs? 05:47 < green> yup, so you have both 05:51 < Getty> mh 05:51 < Getty> and how add i space then? all i found was docu with kernel parameters 05:53 < green> add space? 05:53 -!- ElectricElf [david@elf.noc.oftc.net] has quit [Ping timeout: 245 seconds] 05:53 < Getty> i mean, getting some ramfs/tmpfs for use it with uml ;) 05:53 < green> well, you just "mount none /mntpnt -t ramfs (or tmpfs)" ;) 05:53 < Getty> ah and thats "all" ram? 05:54 < caker> mount tmpfs /mntpnt -t tmpfs -o size=64m 05:54 < green> ramfs is all ram, mpfs can also go to swap and it is slower because of that 05:54 < Getty> cool :) 05:54 < caker> it defaults to half your ram, or something 05:54 < Getty> tmpfs much better then... 05:54 < Getty> cool 05:54 < green> tmpfs defaults to half of ram, not sure of ramfs 05:54 < caker> Getty: check this url out 05:54 < caker> http://www-106.ibm.com/developerworks/library/l-fs3.html 05:55 < Getty> cool 05:55 < Getty> coooool :) 05:56 < Getty> tmpfs 65536 0 65536 0% /data/uml/umshadow/ram 05:56 < Getty> do you can guess my next question? ;) 05:56 < green> no ;) 05:56 < Getty> come on, you're an intelligent guy, you will know :) 05:56 < caker> how to I make uml use the tmpfs mount? 05:56 < green> you know, the batteries in my telepathic receiver are long dead and it is pretty hard to find this kind of batteries in local stores ;) 05:57 < Getty> yeah, but anyway you see the mountpoint ;) 05:57 < Getty> caker: good guy :) 05:57 < caker> Getty: use a command like this: 05:58 < caker> TMP=/tmp/dir/mount/point ./linux arg1=foo arg2=bar ... 05:58 < Getty> ah cool 05:58 < Getty> this all theme is much undocumented, i should write some short info about that ;) 05:58 < green> just make sure there is enough space for all the ram in mntpoint, or uml will die with sigbus 05:59 < Getty> caker: did i understand right that he is not much using the TMP? 05:59 < Getty> mmhh 05:59 < Getty> mem=64M and tmpfs size=64M 05:59 -!- ElectricElf [david@CPE0040f6b429d8-CM400049226055.cpe.net.cable.rogers.com] has joined #uml 05:59 < Getty> should work, or? ;) 05:59 < caker> yes 06:00 < Getty> can i check that he is using the dir? 06:00 < caker> lsof | grep umshadow 06:00 < green> df will tell you there is some space used ;) 06:00 < caker> look for a deleted file in your tmp mount 06:01 < Getty> mmhhh... my host seems some kind of "slow"..... 06:03 < Getty> system up 06:04 < Getty> using the RAM :) 06:04 < Getty> cool 06:04 < Getty> but anyway... the load of the "host" is really "high" 06:04 < caker> Getty: what happened to skas mode? 06:05 < Getty> i don't use skas 06:06 < Getty> i can't reboot host at the moment 06:08 < Getty> yeah.. with tmpfs the host is much more loaded then before 06:08 < Getty> regulary the UML doesn't make "load" 06:08 < Getty> 0.01 or something 06:08 < Getty> now (after the system is up and "calm")... i still have 0.10 06:15 < Getty> ok now... 0.02 06:25 < Getty> mmhhh.. 06:25 < Getty> interessting question 06:25 < Getty> UML: Mem: 63795200 52871168 10924032 0 659456 45281280 06:25 < Getty> Host: tmpfs 67108864 56270848 10838016 84% /data/uml/umshadow/ram 06:25 < Getty> (total / used / free in both case) 06:35 < Getty> i mean... 06:35 < Getty> shouldn't it be the same? 06:36 < green> no if not all of the ram is used yet inside of uml ;) 06:51 < Getty> mh? 06:51 < Getty> look at the used 06:51 < Getty> in the UML there are 52.... used 06:51 < Getty> and in the Host the tmpfs there is 56... used 06:52 < Getty> if i count right, if the UML ram ist full its bigger then the tmpfs 06:54 -!- caker [~null@68.52.196.167] has quit [Ping timeout: 488 seconds] 07:41 -!- arch [arch@68.51.49.60] has joined #uml 08:43 -!- Aike [~aikedejon@217.122.0.97] has quit [Ping timeout: 488 seconds] 08:48 -!- caker [~null@68.52.196.167] has joined #uml 09:31 -!- Aike [~aikedejon@217.122.0.97] has joined #uml 09:44 -!- Valen_ [~valen@sprocket.hosting365.ie] has joined #uml 09:44 < Valen_> Hi...anyone know if the RH9 kernels have the skas3 patch in them ? 09:45 < david> they don't 09:45 * Valen_ is just trying UML for the first time, and going for the path of least resistance 09:45 < green> no, they don't have it 09:45 < green> hi david 09:45 < david> green: morning 09:45 < Valen_> Smeg. The last skas3 patch doesn't go against the redhat kernels nicely at all. 09:46 < david> Valen_: so use the module 09:47 < Valen_> david: ooh. that exists ? 09:47 < david> yeah 09:47 < david> kernels.usermodelinux.org/host/ 09:48 < david> http://kernels.usermodelinux.org/host/mod-skas3-20030107.tgz 09:48 < david> specificly that 09:49 < Valen_> sound, thanks. 09:49 < green> I think it is only suitable for RH8, not RH9 09:50 < Valen_> Ah. Well, I'll give it a go. If it's compilable, I'll do it. If not, I can do some hacking. Before giving up & crying. 09:51 < Valen_> How important is skas anyway ? More for piece of mind, or real performance ? 09:51 < green> the RH9 changes are fairly not trivial 09:51 < green> skas is mostly for performance 09:57 < Valen_> You are right....it's not going into the RH9 kernel. Will that be remedied over time ? 09:57 < Valen_> It looks like it assumes "task" is global 10:37 < arch> mornin 11:12 -!- smcavoy [~smcavoy@207.61.160.163] has joined #uml 11:36 < mistik1> RawR!! 11:37 < mistik1> Kudos to Novell! 12:17 < Valen_> ? 12:18 < mistik1> http://www.novell.com/news/press/archive/2003/05/pr03033.html 12:23 < Valen_> Anyone good with the briding software to get an IP up ? 12:24 < Valen_> Wow, it'd be turn for the books if Novell want to enforce copyrights etc. 12:33 -!- raceme is now known as register 12:45 < Valen_> Can anyone guess why a newbie's UML would die at startup ? Last line is "Kernel panic: outer trampoline didn't exit with SIGKILL 12:45 < Valen_> " 12:56 < david> *sigh* 12:56 < david> Valen_: this is a known problem with UML on RH9.0 kernels 12:57 < david> Valen_: either grab a newer kernel or patch the one you have 13:01 < Valen_> david: newer UML, or newer redhat kernel ? 13:01 < david> UML 13:01 < david> Valen_: did you build your own kernel or get a binary? 13:03 < Valen_> I just got the 2.4.19 RPM from the sf.net page 13:04 < david> ah, ok 13:04 < Valen_> Could you recommend a modern version ? There seems to be a bewildering set at http://www.stearns.org/uml/ 13:04 < david> you might want to grab one of my kernels from kernels.usermodelinux.org 13:04 < Valen_> Basically, i'm looking for something that isn't too advanced, but is stable. 13:04 < david> Valen_: it needs to be 2.4.20-um4 or later I believe 13:05 < Valen_> david: they are all source, yeah ? 13:05 < david> no 13:05 < david> I have binaries 13:06 < Valen_> ah, of course. the "nomods/linux" file is what I'm looking for. 13:06 < david> nod 13:06 < Valen_> This means I need some of the other binaries that the RPM provided though, yes ? 13:06 < david> Valen_: nah, just install the RPM and overwrite the /usr/bin/linux with the one from my site 13:07 < Valen_> OK, sweet. 13:08 < david> I should probably build some vanilla 2.4.20 kernel binaries for people 13:08 < david> since this trampoline bug is a pita 13:08 < Valen_> Sweet. I'm getting a little further. 13:08 < Valen_> VFS: Cannot open root device "ubd0" or 62:00 13:08 < Valen_> I'm running : "/usr/bin/linux ubd0=/home/test1/myrootfs,/usr/local/uml/uml-rh-9.img ..." 13:08 < Valen_> Is that OK ? 13:09 < Valen_> myrootfs is blank, the second one is the read-only pristine one. 13:11 < Valen_> Aha. "Failed to read COW header from COW file "/home/test1/myrootfs", errno = -22" 13:12 < Valen_> How do you add a COW header into a file that you have just run mk2efs on ? 13:14 -!- pasky [pasky@pasky.noc.xs26.net] has joined #uml 13:18 < david> uh 13:18 < david> you don't 13:18 < david> delete /home/test1/myrootfs 13:19 < david> UML will create it for you 13:19 < Valen_> Ah, it gets made ? 13:19 < david> yep 13:19 < Valen_> Wow. Computers can do that ? 13:19 < david> Valen_: sometimes 13:20 < Valen_> Heh. Kick ass. It came up, but is moaning about the /etc/fstab being wrong. Considering the box this filesystem came from was RAID1 setup, that's not bad. 13:20 < Valen_> what are the device files i should be mounting ? /dev/hda1 ? /dev/udba ? 13:21 < david> /dev/ubd0 13:22 < Valen_> Hmm. No mention of the major/minor numbers in linux/Documentation/devices.txt 13:24 -!- pasky [pasky@pasky.noc.xs26.net] has quit [Ping timeout: 492 seconds] 13:27 -!- pasky [pasky@pasky.noc.xs26.net] has joined #uml 13:31 < Valen_> Ah, it lives. Nearly. 13:31 < Valen_> UML booted RH9 as far as "Starting console mouse services: (no mouse is configured)", then stopped 13:42 < david> Valen_: kill it and run it with 'con=xterm' as a command line option 13:42 < david> then the gettys will start up in xterms 13:44 < Valen_> Ah, nice. 13:44 < Valen_> Making sure I have only one or two defined in /etc/iniittab, yeah ? 13:45 < Valen_> I'm still having curious problems with devices though. I can't swap to ubd1, even though I made it with mknod 98 1 13:45 < david> ubd1 has a minor of 16, not 1 13:45 < Valen_> aha. 13:45 < david> ubd's can have partitions, which are 1-15 13:46 < caker> I've been using "ubd=3" lately, which fakes your ubd devices as major 3 (/dev/hdX) inside UML .. i like 13:50 < caker> Yes! the guys on the lkml have provided a fix for "process stuck in __lock_page" ! 13:50 < david> caker: Andrew Morton's fix? 13:50 < david> http://marc.theaimsgroup.com/?l=linux-kernel&m=105410846604499&w=2 13:51 < caker> later in the thread, someone posted three things to try .. as it turns out, you need all three 13:52 < david> nod 13:52 < caker> http://marc.theaimsgroup.com/?l=linux-kernel&m=105411752313126&w=2 13:52 < Valen_> david: after about ten minutes wondering why it wasn't working, i realised that the machine I'm playing with is headless, and doesn't have X installed, never mind xterm. 13:52 < caker> It's been in the kernel for over a year 13:52 < Valen_> What's the next neatest way of getting a terminal ? 13:52 < david> Valen_: pts 13:52 < david> Valen_: con=pts, which is the default 13:53 < david> then use 'uml_mconsole' to find out the /dev/pts/X for each terminal 13:53 < david> uml_mconsole config con1 13:53 < david> then 'screen /dev/pts/X' 13:53 < Valen_> I was adding "con=pty con0=fd:0,fd:1" to the line, and it seemed that the term I started "linux" from was the console, but it hung after a while. 13:54 < david> it didn't 'hang' 13:54 < david> you just didn't have a console on tty0 13:54 < david> Valen_: what are the getty lines from your inittab? 13:55 < Valen_> 1:2345:respawn:/sbin/mingetty tty1 13:55 < Valen_> So I just add one for tty0, and all is happy ? 13:57 < david> nod 13:57 < david> assuming it can open /dev/tty0 13:58 < Valen_> Just checking. If you have twenty people running off a COW filesystem, and a new glibc comes out...there is no nice way of applying that to the old "pristine" image, is there. You ahve to install it in all twenty UMLs ? 13:59 < david> Valen_: yep 13:59 < david> caker: seems like a pretty basic fix :-/ 14:00 < Valen_> Bah. I suppose I can understand that. 14:00 < caker> Yeah, nothing I would have found - these guys probably looked at it for only a few minutes and had something 14:07 < Valen_> OK, I've a UML up. Next step networking. 14:07 < Valen_> # ifconfig eth0 192.168.0.236 up 14:07 < Valen_> Failed to open /dev/net/tun, errno = 13 14:07 < Valen_> SIOCSIFFLAGS: Permission denied 14:07 < caker> Valen_: chmod 777 /dev/net/tun 14:07 < Valen_> I'd /dev/net/tun on the host chgrp'd to uml, and the user this is running as is in tun... 14:07 < caker> Valen_: on the host :-) 14:07 < Valen_> caker: Well, why not.. 14:08 < Valen_> Now I'm scared. Can't ping out, and I have no idea why. The bridging stuff scares me. 14:10 < BB> heh ;) 14:11 < Valen_> OK, when I do; ifconfig uml-int0 192.168.0.236 netmask 255.255.255.0 up 14:12 < Valen_> on the host, that's the IP that the UML can use, when it brings up eth0, yeah ? 14:12 < caker> no that should be the IP address of your host 14:12 < Valen_> Oh. when I did that before, it made an almighty mess of things. 14:12 < caker> you could use a different IP in the same subnet 14:13 < caker> but make sure the UML get's it's own 14:13 < Valen_> from the uml.openconsultancy.org page, there is a line 14:14 < Valen_> "ifconfig eth1 0.0.0.0 promisc up" 14:14 < Valen_> This scares me. eth1 is my main IP. Does that mean that suddenly I drop off the network ? 14:14 < david> Valen_: yes 14:14 < david> Valen_: write a script to setup the bridge, so it doesn't disappear for good :-) 14:14 < BB> david you mean write a script pray for no typo's ;) 14:14 < david> BB: right 14:15 < Valen_> david: well, I wrote a script, based off that page I mentioned, and it dropped off the network. 14:15 < Valen_> Some messing with IPs later, and it didn't anymore. But of course, now it doesn't work. 14:15 < Valen_> Deep magic. 14:15 * BB kept typing brctl in as btctl which didnt help too much 14:15 < Valen_> :) 14:16 < Valen_> Ok, so stop me if this is stupid; ifconfig uml-int0 netmask 255.255.255.0 up 14:16 < Valen_> tunctl -u test1 -t uml-test1 14:16 < Valen_> ifconfig uml-test1 0.0.0.0 promisc up 14:16 < Valen_> brctl addif uml-int0 uml-test1 14:16 < Valen_> brctl addbr ext0 14:16 < Valen_> ifconfig eth1 0.0.0.0 promisc up 14:16 < Valen_> brctl addif ext0 eth1 14:17 < Valen_> tunctl -u test1 -t ext-test1 14:17 < Valen_> ifconfig ext-test1 0.0.0.0 promisc up 14:17 < Valen_> brctl addif ext0 ext-test1 14:17 < caker> that looks wacked 14:17 < Valen_> And that should give me a ext-test1 device ? 14:17 * Valen_ looks at caker like a spaniel that's crapped on the carpet and knows he did it 14:17 < Valen_> what did I do wrong ? 14:19 < caker> Here is my startup script 14:19 < caker> /usr/sbin/brctl addbr br0 14:19 < caker> /usr/sbin/brctl stp br0 off 14:19 < caker> grrr 14:19 < caker> - /sbin/ifconfig eth0 down 14:19 < caker> - /sbin/ifconfig eth1 down 14:19 -!- Maniac [~daManiac@209.5.247.105] has joined #uml 14:19 < caker> - /usr/sbin/brctl addif br0 eth1 14:19 < caker> - /sbin/ifconfig eth1 0.0.0.0 up 14:19 < caker> - /sbin/ifconfig br0 64.5.53.6 netmask 255.255.255.0 up 14:19 < caker> - /sbin/route add default gw 64.5.53.1 14:20 < caker> that gets the host bridge running and the host back on the net 14:20 < caker> then, create your tap device, with IP as the hosts's, and add it to the bridge 14:20 < Valen_> what is "64.5.53.6" 14:20 < caker> that's my real ip 14:20 < caker> yours will be 192.168.x.x 14:21 < Valen_> Ah. I think the reason my setup was different to yours is that I intend having twenty UMLs on this soon :) 14:21 < caker> no different 14:22 < caker> just run those first, and then for each uml: create the tap, ifconfig it to hosts's ip, add it to bridge, turn on proxy_arp 14:22 < Valen_> Hmm. No one mentioned proxy arp. 14:23 < Valen_> Any chance you could mail me your startup script ? 14:23 * BB doesnt have proxy_arp turned on and all seems well? 14:23 < caker> you mean how I launch a UML? 14:23 < Valen_> caker: well, more importantly all the brctl stuff. 14:23 < caker> my host startup script is those 8 lines above 14:23 < Valen_> Oh. 14:24 < Valen_> so this should do me; 14:24 < mistik1> this may sound silly in light of things, however, Is it posable for a bridge device to work with DHCP? 14:24 < Valen_> ifconfig eth0 down 14:24 < Valen_> ifconfig eth1 down 14:24 < Valen_> brctl addbr uml-int0 14:24 < Valen_> brctl setfd uml-int0 1 14:24 < Valen_> brctl sethello uml-int0 1 14:24 < Valen_> brctl stp uml-int0 off 14:24 < Valen_> ifconfig uml-int0 192.168.0.235 netmask 255.255.255.0 up 14:24 < Valen_> route add default gw 192.168.0.1 14:24 < Valen_> tunctl -u test1 -t uml-test1 14:24 < Valen_> ifconfig uml-test1 0.0.0.0 promisc up 14:24 < Valen_> brctl addif uml-int0 uml-test1 14:24 < david> Valen_: no 14:24 < david> Valen_: oh, wait, yes 14:24 < caker> mistik1: yes, i've got that working - but it only will be picked up on the host the UML is running, so far for me 14:24 < david> mistik1: a bridge forwards ethenret frames, so if you have a bridge, your UML can DHCP on the LAN 14:25 < mistik1> Let me clarify.. 14:25 < Valen_> suddenly the old man from monty python & the holy grail throws david off the bridge 14:25 < caker> it only works on the same machine for me because of some iptable rules i have 14:25 < mistik1> Lets say eth0(host) that I would like to use for the bridge is DHCP, an this work? 14:25 < caker> that keep people from ifconfig'ing IPs that aren't theirs 14:26 < caker> oh 14:26 < mistik1> s/an/can/ 14:26 < caker> you could write a script that looks at the current IP before ifconfig'ing eth0 down to set up the bridge 14:26 < caker> ? 14:26 < mistik1> this is what I was thinking actually 14:27 < mistik1> and have a cronjob that checks for changes and resets the bridge accordingly? 14:27 < caker> i'd edit the file that runs when dhcp changes 14:27 < caker> or set one up 14:28 < mistik1> true 14:28 < caker> You **might** be able to make an ifup-bridge0 and use RedHats/your distros config stuff 14:28 < david> the device to do dhcp on is br0, not eth0 14:29 < mistik1> I know 14:29 < david> eth0 doesn't function as a layer 3 device anymore 14:29 < mistik1> It is the MAC issue why I was asking this 14:29 < caker> it still goes out your ethernet device, so it will us ethat MAC addy? 14:29 < caker> ? 14:29 < david> no 14:30 < david> the bridge has it's own MAC 14:30 < caker> hmm 14:30 < david> a 00:FF:x:x:x:x 14:30 < mistik1> indeed 14:30 < david> so just hwaddr it 14:30 < david> change the eth0 to something else 14:30 < mistik1> awesome 14:30 < david> change the bridge to that of eth0 14:30 < david> and you're done 14:30 < caker> that rules 14:30 < david> doesn't matter what eth0 is, as long as it's unique 14:30 < mistik1> rockin 14:31 < caker> I have a little perl function that will take an IP and give you a unique MAC addy: 14:31 < caker> sub mac_from_ip { 14:31 < caker> my $ip = shift; 14:31 < caker> my $tmp_mac="fe:fd"; 14:31 < caker> my @blocks = split(/\./, $ip); 14:31 < caker> foreach $block (@blocks) { 14:31 < caker> $hex = sprintf("%0x",$block); 14:31 < caker> $hex = "0".$hex if length($hex) == 1; 14:31 < caker> $tmp_mac.=":" . $hex 14:31 < caker> } 14:31 < caker> return $tmp_mac; 14:31 < caker> } 14:31 < caker> probably from umld :-p 14:33 < mistik1> thanks david 14:35 * Valen_ mutters that his startup script killed the machines network, so he's doing *something* wrong. 14:35 * BB has a rebooting uml lol 14:38 < david> np 14:42 -!- Maniac is now known as Maniac-werk 14:42 < Valen_> Any idea why I'm taking down the network, when I try bring up the bridge ? 14:42 -!- Maniac-werk is now known as Maniac 14:44 < david> Valen_: because you have to ifconfig the eth device without an IP 14:44 < david> which will drop it off the face of the planet 14:45 * Maniac notes uml rocks 14:45 -!- Maniac [~daManiac@209.5.247.105] has left #uml [Client exiting] 14:46 < BB> doh 14:46 < BB> its a night for toast! :( 14:47 < Valen_> david: is the bridge not supposed to give it an IP in its place ? 14:47 < Valen_> as in; 14:47 < Valen_> brctl addbr uml-int0 14:47 < Valen_> ifconfig uml-int0 192.168.0.235 netmask 255.255.255.0 up 14:47 < Valen_> route add default gw 192.168.0.1 14:47 < Valen_> ifconfig eth0 0.0.0.0 promisc up 14:48 < david> the bridge gets an IP 14:48 < david> the eth0 doesn't 14:48 < Valen_> does eth0 need one ? Or is the ip the bridge gets for the UML box ? 14:48 < david> no, eth0 doesn't have one 14:48 < Valen_> so, I should be able to ssh to the bridge IP ? 14:49 -!- Maniac [~daManiac@209.5.247.105] has joined #uml 14:49 < david> the IP the bridge gets should be the IP from eth 14:49 < david> Valen_: yes 14:49 < david> Valen_: without a UML running, with a bridge it should behave as normal 14:49 < Valen_> Hmm. Well, as is, when I run those four lines, the machine drops off the network. 14:50 * BB rings clara.net and has a beer 14:50 < david> you need to add eth0 to the birdge 14:50 < david> brctl addif uml-int0 eth0 14:50 < Valen_> Hmm. Ok. 14:52 * BB will remember no screwing with the wrong server next time 14:54 < Valen_> WOOOHOOOO 14:54 < Valen_> All works 14:56 < Valen_> david, caker, thanks a lot!! 14:56 < Valen_> I'll be back, annoying you again... 14:56 < david> np 14:58 -!- Cliff [none@209.179.238.204] has joined #uml 14:59 < Cliff> is it possible to expand a filesystem once it's been created? 14:59 < Cliff> i got one of the filesystem's from the sourceforge UML site, and it's too small to do anything 14:59 < Valen_> Cliff: well, you can just make a new one. 14:59 < david> dd if=/dev/zero of= bs=1k count=1 seek=$[512*1024] 14:59 < david> that makes it 512Mb 14:59 < Cliff> oh thanks 15:00 < david> then run resize2fs on it 15:00 < Cliff> i thought that dd might be able too 15:00 < BB> Cliff http://www.linux-sxs.org/pipermail/linux-users/2002-November/010762.html 15:00 < Valen_> Cliff: you can mount the filesystem file outside UML with: mount -o loop filesystem.img /mnt/fs 15:00 < Cliff> how does dd know how far to seek? 15:00 < BB> just dont blame me if that breaks 15:00 < Cliff> haha 15:00 < Valen_> Cliff: so, mount the old one, make a new one, copy the old one into the newer FS. 15:00 < Cliff> ok 15:01 < Valen_> I've just made a nice RH9 + APT one, and it works quite nicely. It gzips to 129MB. 15:01 < Valen_> It expands as a 2GB one, with 540MB used. 15:04 < Valen_> david: Can I send the linux processes a sigsusp, and take a copy of the COW filesystem files, to make backups ? Or would I really have to take the boxes down ? 15:04 -!- Cliff [none@209.179.238.204] has quit [Read error: Connection reset by peer] 15:05 < david> Valen_: uml_mconsole stop 15:05 < david> Valen_: uml_mconsole sysrq s 15:05 < david> copy the stuff 15:05 < david> Valen_: uml_mconsole go 15:23 < Valen_> david: wow, cool. Is there a nice way of telling them to "init" 0 ? I'm doing some startup/shutdown scripts... 15:23 < david> no 15:23 < david> you can send 'halt' and 'cad' 15:23 < david> cad is ctrl-alt-del 15:24 < david> the 'ca' section of inittab 15:24 < david> I usually make it shutdown the UML 15:24 < Valen_> david: ah, sweet. Ok, I'll do that. Much nicer. that "cad" can be sent with uml_mconsole too, yeah ? 15:24 < david> nod 15:24 < david> uml_mconsole cad 15:28 < Valen_> sweet. 15:28 < Valen_> Damn, they have thought of everything 15:31 < Valen_> Don't suppose you've an idea how I could get a status of all UMLs running ? 15:32 < david> status? 15:32 < david> use 'ps' 15:32 < david> :-) 15:32 < Valen_> Yeah, just wondering could I get a nice "these are your running VMs" type screen. No worry. 15:33 < david> Nope 16:20 -!- shark [~shark@dsl-209-162-200-223.dsl.easystreet.com] has joined #uml 16:33 -!- cameron [~cameron@crosstalk.otago.ac.nz] has quit [Ping timeout: 492 seconds] 16:34 * BB gives redbus a pat on the back for fixing :) 17:12 -!- smcavoy [~smcavoy@207.61.160.163] has quit [Remote host closed the connection] 17:17 < caker> david: did have problems with the route bug? 17:18 < david> caker: no 17:18 < david> caker: I just don't *want* problems 17:19 < caker> I've been suprised how many "extra" patches are available for vanilla 2.4 series. Some of them seem fairly serious 17:19 < caker> and none of them seem to make their way in to main 17:20 < david> marcello is a dip 17:21 < caker> so i've heard :( 17:21 < david> none of the good patches to make 2.4 useful end up in the vanilla kernel tree 17:21 < caker> david: do you do anything to limit num/processes or file descriptors for your host/umls? 17:21 < david> no 17:22 < caker> I've thought about using ulimit or modifying fs.h's constants 17:23 < david> you could ulimit, but I don't see the point right now 17:23 < caker> what am i missing then? 17:24 < david> caker: my ulimit sets max open files to 1024 17:24 < david> my kernel allows ~157k open files 17:24 < caker> is that ulimit set in the shell that launches your umls? 17:25 < david> yeo 17:25 < david> p 17:29 -!- livio [~livio@200.162.231.47] has joined #uml 17:29 -!- smcavoy [~smcavoy@grimlock.drive-megawheels.net] has joined #uml 17:31 < livio> david: Which fix for ll_rw_blk.c did you include in your UML kernel? The 3 small patches from Andrew Morton? 17:32 < caker> livio: where are you getting that from? 17:33 < livio> caker: HIs e-mail to uml-devel from some time ago.. Subject: UML Kernel -2.4.21-rc5-djc1-5um 17:35 < caker> well, it's definately not the three small patches from lkml this morning 17:36 < caker> actually, is it? 17:36 < livio> I don't know... that's why I asked :-) 17:36 < caker> you could download his diff and look 17:36 < livio> Yeah... I'll take a look. 17:37 < caker> livio: coolio 17:40 -!- Maniac [~daManiac@209.5.247.105] has quit [Quit: Client exiting] 17:40 < livio> caker: Hummm.. looks like he included patches 1 and 3 of the ones sent by akpm... 17:41 < caker> but not the add_wait_queue mod? 17:41 -!- Aike [~aikedejon@217.122.0.97] has quit [] 17:41 < caker> because I think actually 1+2 fix the problem, not sure if 3 made much diff 17:41 < livio> caker: Nope, my bad! He included all 3. 17:41 < caker> damn, that was fast of him :-) 17:41 < caker> btw, I hated that bug, soooo glad it's fixed 17:42 * caker stops himself from ranting 17:42 < livio> caker: Yeah... but by his e-mail, I got the impression that a "oficial" fix had been released... and I thought I hadn't received some LKML mail... 17:42 < caker> livio: I didn't see that one 17:43 < livio> caker: Yeah, me neither. I think there is still no "oficial" fix for it. 17:43 < caker> livio: it's only a three line fix> 17:43 < caker> ? 17:43 < caker> i don't see much more official than that :-) 17:45 < livio> caker: Hehehhe... we'll Jens and Andrew were kind of sceptical about that fixing anything... so that may just be "covering" up some other problem. Anyway... I was just curious. 17:45 < caker> livio: thanks for the insight 17:45 < caker> livio: hopefully, whatever it was/is they'll find it 17:46 < livio> caker: I sure hope so.. I mean, it's been around for a pretty long time now. Also, hope Marcelo lets it in before 2.4.21... 17:51 -!- smcavoy [~smcavoy@grimlock.drive-megawheels.net] has quit [Quit: ACK!] 17:54 -!- pasky [pasky@pasky.noc.xs26.net] has quit [Ping timeout: 492 seconds] 18:00 -!- shark [~shark@dsl-209-162-200-223.dsl.easystreet.com] has quit [Quit: Client Exiting] 18:33 -!- cameron [~cameron@crosstalk.otago.ac.nz] has joined #uml 18:54 < BB> lol so how many would you ? lol 18:55 < caker> BB: depends on how much I'm paid 18:55 < BB> caker erm i've done it again havnt i lol 18:55 < caker> heh 18:56 < BB> but trust me with some of these you wouldnt need paying lol 18:56 * BB now goes and discuses in the correct channel 18:56 < caker> BB: i have no idea what you're talking about :| 18:58 < mistik1> david: you around? 19:01 < mistik1> ok, I get the bridge setup except one thing, with the earlier situation of having only a single routable IP what IP would I give eth0(uml) if I made my CABLE use br0 19:01 < mistik1> that is the only thing that escapes me 19:04 < caker> set up an internet subnet and masq it through your cable modem's ip 19:04 < caker> s/internet/network/ 19:14 < mistik1> based on where would I run this MASQ 19:14 < mistik1> on the host? 19:14 < caker> yes. You are saying that you only have one machine on your cable modem? 19:14 < mistik1> right 19:15 < caker> yes, .. set up a 192.168.x.x network on your host machine, then use iptables to set up masq 19:15 < caker> http://www.tldp.org/HOWTO/IP-Masquerade-HOWTO/firewall-examples.html#RC.FIREWALL-2.4.X 19:15 < mistik1> hmm 19:15 < mistik1> like eth0:1 192.168.foo 19:15 < caker> you only have one IP from your ISP, right? 19:15 < caker> yeah 19:15 < mistik1> yes 19:16 < mistik1> I know iptables pretty well :) 19:16 < caker> use 192.168.1.1 for eth0:1, and that becomes the "gateway" for the UMLs 19:16 < mistik1> just have not done a lot of bridging 19:16 < mistik1> rocking 19:16 < caker> you could probably use another tap device instead of eth0:1 .. 19:16 * caker shrugs 19:17 < mistik1> I would only need to change my $ext_if=br0 from eth0 19:17 < caker> think so 19:17 < mistik1> $int_if=tap1 for example 19:17 < caker> you can't ifup the bridge with two IPs, so you'll need another one, i think 19:17 < caker> (another device) 20:11 < mistik1> hmm 20:27 < david> david: hello 20:27 < david> mistik1: er, hello 20:27 < mistik1> hey 20:27 < david> if your eth0 is for your cable Internet 20:27 < david> don't bridge onto eth0 20:28 < david> since your ISp will probably only give you one IP 20:28 < david> setup a bridge with a 192.168.0.x IP 20:28 < david> masquerade that 20:29 < mistik1> bugger all 20:29 < mistik1> I think I get it 20:29 < mistik1> let me run it down ok 20:29 < mistik1> eth0 = routable 20:30 < mistik1> br0 = non-routable 20:30 < mistik1> add tap0 to br0 20:30 < mistik1> run uml against tap0 20:31 < mistik1> I could use uml_switch on it also 20:31 < mistik1> damn but this is fun 21:12 < david> I'd not use uml_switch 21:29 -!- ASY [~someone@216.221.51.54] has joined #uml 21:29 -!- ASY is now known as ASY- 21:59 -!- ASY- is now known as ASY 22:00 < mistik1> so just go ahead and use a different tap for each uml 22:02 < david> that's what I do 22:04 < mistik1> ok 22:05 < mistik1> you know the more I look at this the more it seems I dont even need a bridge 22:05 < mistik1> or *may not* 22:05 < mistik1> I'll let you know 22:05 < david> if you don't use a bridge 22:05 < david> and have lots of taps 22:05 < david> then you need lots of /32 routes 22:05 < mistik1> I'm ahh 22:06 < mistik1> so I'll test with one for now 22:06 < mistik1> I'll let you know how it goes 22:41 -!- livio [~livio@200.162.231.47] has quit [Quit: Download Gaim [http://gaim.sourceforge.net/]] --- Log closed Thu May 29 00:00:00 2003