--- Day changed

--- Log opened Tue Jun 17 00:00:02 2003
00:15 -!- shaver [~shaver@abla.tive.org] has joined #uml
00:17 < shaver> does anyone happen to know where I might find a RH9 UML image?
00:19 -!- shaver is now known as zzZZhaver
00:19 < Lathiat> root filesysten>?
00:19 < zzZZhaver> yeah
00:20 < Lathiat> uhh
00:20 < Lathiat> make one? :P
00:21 < zzZZhaver> yeah, I feared it would come to that
00:21 < zzZZhaver> thought I'd check first
00:21 < zzZZhaver> thanks
00:21 -!- zzZZhaver [~shaver@abla.tive.org] has quit [Quit: Client Exiting]
01:27  * alt is away: bed
03:05 -!- BB [~chris@ns1.8086.net] has quit [Ping timeout: 496 seconds]
03:07 -!- BB [~chris@ns1.8086.net] has joined #uml
04:43 -!- bshyu [~bshyu@113.c210-58-173.ethome.net.tw] has joined #uml
06:03 -!- DnsInfector [~DnsInfect@ARennes-204-1-14-4.w81-248.abo.wanadoo.fr] has joined #uml
07:16 -!- yeti [~yeti@62.226.42.25] has joined #uml
07:18 < yeti> bridge debug or logging screws my consoles... does someone know which syslog channe i need to catch or whatever i need to get rid of that?
08:19 < Lathiat> umm
08:19 < Lathiat> not sure sorry
08:30 < yeti> dmesg -n 2
08:30 < yeti> ;-)
08:31 < yeti> increases needed level for messages to go to the console (or everybody)
08:31 < yeti> this is not a bridge specific solution but ok for me and now...
08:32 < yeti> how about addin this to usermode.org's bridge text?
09:11 -!- yeti [~yeti@62.226.42.25] has quit [Quit: Client exiting]
10:02  * green notices a lawsuit initiated by parents that were refused to officially name their son as "BOH fVF 260602"
10:04 < DnsInfector> :)
10:31 -!- thy_ [koubt9am@u-pl7.ms.mff.cuni.cz] has joined #uml
10:40 -!- smcavoy [~smcavoy@grimlock.drive-megawheels.net] has joined #uml
10:46 -!- mistral [mistral@212.159.71.212] has joined #uml
10:48  * alt is back (gone 09:21:24)
11:07 -!- yeti [~yeti@p3EE22A19.dip.t-dialin.net] has joined #uml
11:53 -!- yeti [~yeti@p3EE22A19.dip.t-dialin.net] has quit [Quit: Client exiting]
11:56 -!- sjc [~sjc@80.46.14.25] has joined #uml
12:00 -!- DnsInfector_ [~DnsInfect@ARennes-204-1-10-115.w81-53.abo.wanadoo.fr] has joined #uml
12:03 < caker> david: can I get a blurb on uml.org for my site?
12:03 < david> caker: sure
12:03 < david> caker: mail me with it
12:04 < caker> david: ok thanks
12:08 -!- DnsInfector [~DnsInfect@ARennes-204-1-14-4.w81-248.abo.wanadoo.fr] has quit [Ping timeout: 496 seconds]
12:23 < shak> david: in the document on openconsultantcy, do you have different interfaces for internal and external UML connections?
12:24 < shak> so the first bridge that is setup is for internal, and the second for external?
12:29 < david> shak: yes
12:29 < david> right
12:29 < shak> ok, did you have two physical eths in your host?
12:29 < david> no
12:29 < david> I have six
12:29 < david> :-)
12:30 < shak> heh, well, when you were building the bridge, the "external" bridge needs to correspond to a physical eth1?
12:30 < david> bbl
12:30 < david> shak: right, the external bridge has my eth1 as a port, so stuff goes right out onto my upstream
12:30 < smcavoy> I can't seem to resize any of my filesystems. I use dd if=/dev/zero of=root_fs bs=1k count=1 seek=$[3*1024*1024] which expands it to 3GB. resize2fs says there's not enough space (I've try resizing it to less then 1gb, and still complains).
12:31 < shak> david: thanks a lot
12:31  * shak goes to reboot and add another NIC
12:43 < BB> hmm some guy from Zend just called me up about my complaint .. first time thats ever happend with an online company ;)
12:46 < Lathiat> lol
12:46 < Lathiat> nice
12:46 < Lathiat> what complaint?
12:46 < Lathiat> YOUR ENGINE IS TEH SUXX0R!
12:46 < smcavoy> What's the best command to resize filesystems with?
12:46 < BB> they gave my email address to a reseller, and i wasnt amused
12:47 < BB> well that is after I'd already got a few $K/s worth of software online
12:50 < Lathiat> niiiice
12:50 < Lathiat> whats they say?
12:51 -!- pustota1 [~pustota1@217.224.22.79] has joined #uml
12:51 -!- bshyu [~bshyu@113.c210-58-173.ethome.net.tw] has quit []
13:08 < BB> Lathiat saying they thought it would help me to decide what i wanted to buy...
13:16 < Lathiat> heh
13:17 < caker> smcavoy: what is your fstype?
13:21 < ichilton> BB: Zend?
13:21 < Lathiat> sleep.
13:21 < Lathiat> www.zend.com (php engine?...)
13:22 < ichilton> ah
13:22 < ichilton> Was hoping it wasn't a typo and he ment Zen because i've been looking at them for adsl
13:27 < BB> we're not tried zen yet ;)
13:27 < BB> had griffin, onyx, easynet, nildram do far
13:28 < BB> man i cant type do=so we're=we've
13:31 < david> BB: what did Zen say?
13:32 < BB> david lol zend called not zen :P
13:33 < david> uh
13:33 < david> got confused
13:34  * david goes back to sleep
13:35 < BB> awe
13:37 < caker> david: back in uk or what?
13:38 < smcavoy> caker: I seemed to only have the problem with ext3. both ext2 and reiser seem to resize fine
13:38 < caker> ext3 is ext2
13:38 < caker> more or less
13:38 < ichilton> BB: how do you rate those?
13:39 < ichilton> BB: Nildram is suposed to be excellent
13:39 < ichilton> BB: and Zen offer 8 static ip's for 23.99/mo + vat, but they wont let you change the reverse dns :((((
13:39 < BB> ichilton griffin+easynet=crap tech support, onyx ok but prices in the dark ages - nildran only started last month no problems as yet
13:40 < ichilton> my exchange was just adsl enabled at the end of last month so i'm still undecided whether to just keep cable which is fast and reliable or go into the unknown with adsl and get static ip's
13:40 < david> caker: nope, I live in the US
13:40 < ichilton> BB: let me know how you get on with nildram. I heard they are really good, just that bit more expensive
13:40 < BB> i went balistic once at easynet when it all went tits up, and their phone system couldnt cope.. all that did was get me a new account manager
13:41 < ichilton> david: sell the UK house yet?
13:41 < david> ichilton: yep
13:44 < ichilton> david: cool
13:44 < ichilton> david: without comiong back?
13:52 < shak> I'm going to hear alan cox speaking next week
13:52 < shak> I can't wait
13:52 -!- ansoft [~anelson@207.179.107.18] has quit [Ping timeout: 490 seconds]
14:00 < david> ichilton: er, yeah
14:00 < ichilton> david: nifty
14:19 < smcavoy> caker: less in this case... I think some ext2 fs tools get confused with the journal, and complain about a bad super block
14:19 < caker> weird, i do it all the time
14:19 < caker> smcavoy: let me see which i use
14:20 < caker> dd if=/dev/zero of=$path bs=1M count=1 conv=notrunc seek=$size
14:20 < caker> e2fsck -f -y $path
14:20 < caker> resize2fs $path
14:28 -!- porourke [~porourke@63.160.164.4] has joined #uml
14:44 -!- porourke [~porourke@63.160.164.4] has quit [Quit: Client Exiting]
15:11 -!- yeti [~yeti@p3EE22A19.dip.t-dialin.net] has joined #uml
15:21 < sjc> Hi, I'm having all sorts of hassle trying to get uml working on a Debian box and a Red Hat 9 box. So I've decided to try building it myself. So I get myself a 2.4.20 kernel and the UML patches - of which there are 6. (uml-patch-2.4.20-X). Do I need to apply them all or just one (if so, which?)
15:21 < alt> I just came from getting it to work on Debian
15:21 < alt> works good now :)
15:21 < alt> let me look at my system
15:21 < alt> brb
15:21 < sjc> Which farm animal did you have to offer as a sacrifice? :-)
15:22 < alt> a cow
15:22 < sjc> I have it booting OK on Debian, but can't for the life of me get the networking to work
15:22 < alt> but seriously
15:22 < alt> hee hee
15:22 < alt> yeah
15:22 < alt> symlink /usr/lib/uml/uml_net to /usr/bin
15:22 < alt> that was my problem
15:22 < alt> and make sure the user running UML is in group uml-net
15:23 < alt> uml_net is a SUID binary and debian set perms to 750 so you need to be in the uml-net group
15:23 < sjc> OK, I'll try that
15:23 < alt> I spend about 5 hours beating my head on that :)
15:23 < alt> let me know how it turns out :)
15:23 < alt> http://www.insecure.org/nmap/nmap_changelog.html  <- New version of nmap... look at the 8th item :)
15:24 < BB> i like it
15:25 < alt> fitting justice
15:25 < alt> every software writer/vendor should issue new releases that say just that
15:25 < alt> :)
15:28 < sjc> alt: what networking method are you using - daemon? tun/tap? ethertap?
15:28 < alt> tun/tap
15:29 < sjc> ok, thanks, I'll try that
15:29 < alt> it's the best way
15:29 < alt> make sure you "modprobe tun"
15:29 < alt> I think the uml_net does it for you, but I do it anyhow
15:30 -!- pflanze [~chris@dclient80-218-22-118.hispeed.ch] has joined #uml
15:30 < pflanze> Hello!
15:31 < alt> hi
15:31 < pflanze> I'm here for the first time. Cool, a whole bunch of people here!
15:31 < sjc> So, I do "config eth0=tuntap,,,10.10.10.19" on my console, and on my UML I get "Netdevice 0 : TUN/TAP backend - IP = 10.10.10.19". Which I assume is fine.
15:32 < alt> are you assigning 10.10.10.19 to the host tap and the UML ethernet?
15:32 < alt> oh, I see
15:32 < sjc> So then I try "ifconfig eth0 10.10.10.20 up" and get errors
15:32 < alt> yeah, that looks right
15:32 < alt> what error?
15:32 < sjc> On a login to the UML
15:32 < alt> cut n paste if you can
15:32 < sjc> and that gets me errors -
15:32 < sjc> bacchus:~# ifconfig eth0 10.10.10.20 up
15:32 < sjc> SIOCSIFFLAGS: Invalid argument
15:32 < sjc> SIOCSIFFLAGS: Invalid argument
15:32 < alt> yeah
15:33 < alt> check your groups
15:33 < sjc> And on the UML boot consilt...
15:33 < alt> you probably need to log out and back in to get the new groups
15:33 < sjc> tuntap_open_tramp failed - errno = 22
15:33 < sjc> tuntap_open_tramp failed - errno = 22
15:33 < sjc> OK, will try logging in and out
15:33 < alt> did you add your user to the uml-net group? 
15:34 < alt> that was the biggest problem I had :\
15:34 < alt> that and the path problem
15:34 < sjc> Yes, I should be in the group
15:34 < alt> okay
15:34 < alt> do a "groups" on the host command line
15:34 < alt> and double check it
15:34 < sjc> simon@zog:~$ groups
15:34 < sjc> simon users kev uml-net
15:34 < sjc> simon@zog:~$ 
15:34 < alt> yep :
15:34 < alt> :)
15:34 < alt> is uml_net in your path?
15:35 < pflanze> sjc: are you using a linux binary from the uml website?
15:35 < sjc> no, I'm using the Debian package
15:35 < alt> debian was a pain
15:35  * pflanze thinks
15:35 < BB> was there an issue with debian and uml_net being in a diff place?
15:35 < alt> BB: yes
15:36 < pflanze> In any case: I've had the same problems, sjc. Also using Debian packages.
15:36 < alt> uml_net is in /usr/lib/uml
15:36 < alt> if needs to be symlinked into /usr/bin
15:36 < pflanze> Exactly that's what I guessed. I've solved it manually without the help of uml_net (could provide you with scripts),
15:37 < pflanze> but a symlink probably works around the source of the problem.
15:37 < alt> it does :)
15:38 < sjc> Getting there.... can now pink the host, both the host's main ip address and the other side of the tunnel
15:38 < alt> :)
15:39 < sjc> ok, I can ssh into the UML instance from the host
15:39 < sjc> thanks for your help alt.... I was beginning to despair :-)
15:39 < alt> :)
15:39 < alt> glad I could help
15:39 < alt> I spent 2 days on that 
15:39 < alt> I'm still not quite done
15:43 < sjc> What's not working for you?
15:43 < alt> it's all working great
15:43 < alt> now I need to create my own user install
15:44 < alt> I should say, install linux instead of using a pre-installed root_fs
15:44 < sjc> Ah. I'm trying to resurect a machine from a backup tape because it has an app I care about, but is very hard to install. 
15:44 < sjc> Aren't there loads of make root fs scripts about?
15:45 < alt> probably
15:45 < alt> I haven't got that far
15:45 < alt> I've only patched my own host (skas), built a user kernel to remove hostfs and debugged debian
15:45 < alt> but the debian install I have doesn't use shadow password files :\
15:45 < pflanze> alt: type shadowconfig on
15:46 < alt> oh! :)
15:46 < alt> I didn't know that
15:46 < alt> brb
15:46  * pflanze smiles
15:46 < alt> I'm not much of an admin :\
15:46 < pflanze> Been there :)
15:46 < pflanze> Just 2 days ago :)
15:47 < alt> I know how to get stuff working and fix it but I don't know all the commands :\
15:47 < alt> :)
15:47 < pflanze> I'm going to try to compile my own uml kernel.
15:48 < alt> happy happy :)
15:48 < pflanze> It needs to run with 512MB 'virtual' memory, which the one from debian doesn't.,
15:48 < alt> virtual as in swap or "ram"?
15:48 < alt> I actually have a swap "partition" as well on my umls
15:48 < pflanze> (It says 'CONFIG_HIGHMEM not enabled - physical memory shrunk to ... bytes', 
15:49 < pflanze> which is strange since iirc CONFIG_HIGHMEM usually only matters for >1GB,
15:49 < alt> yeah, I didn't play with that
15:49 < pflanze> but I guess that's the consequence of how uml works, even under skas.)
15:49 < caker> in tt mode, highmem limit is somewhere around 500M, in skas, it's closer to 1GB
15:50 < pflanze> caker: but I'm using skas, and mem=512M makes it shrink to something like 380M. 
15:51 < pflanze> BTW, (maybe related to the amount of mem I want), very frequently the uml kernel hangs in the boot process just after mounting devfs.
15:51 < pflanze> In another ~10 times it will then work once.
15:52 < pflanze> It seems to always work when I give init=/bin/bash, and then type 'exec init'.
15:52 < pflanze> (But the last point not really confirmed yet)
16:00 < pflanze> BTW in my setup, forking programs run *10* times slower in uml/skas than on the host.
16:01 < pflanze> Compilation of a bigger package is 2 times slower.
16:01 < alt> well, you have to expect some overhead
16:01 < pflanze> Sure, it's just that a 10 times hit sounds very hard.
16:01 < alt> I don't know if what are experiencing is typical though
16:01 < alt> yeah
16:01 < alt> I personally found my system to be quite responsive
16:02 < pflanze> Responsive, yes,
16:02 < caker> pflanze: using a tmpfs on the host?
16:02 < pflanze> but not performant in some cases.
16:02 < alt> ah
16:02 < pflanze> caker: yes I do. And I have 2GB swap space.
16:03 < caker> uml likes to generate context switches .. i've been amazed by vmstat's output on occasion
16:03 < pflanze> I remember someone argueing that uml is more efficient than vmware.
16:03 < caker> pflanze: no hardware emulation
16:04 < pflanze> Yes, but I'm using MOL, Mac-on-linux, which is an emulation of a macintosh on ppc, on my laptop in my day-to-day work,
16:04 < caker> MOL accesses the PPC chip, no?
16:04 < pflanze> and applications under it are almost nothing slower than running under native macos.
16:04 < caker> right
16:04 < pflanze> caker: yes, but same thing for vmware, right?
16:05 < caker> MOL will never run on anything but a PPC chip
16:05 < caker> vmware runs on a bunch of host os's
16:05 < pflanze> Sure, but the technology is the same.
16:05 < caker> is there a vmware port to other processors besides i386?
16:06 < pflanze> Dunno.
16:06 < pflanze> What I want to say: emulating hardware can be fast. Dunno if vmware is as good as mol,
16:06 < pflanze> but if it is, 
16:06 < caker> and if you can afford it
16:06 < pflanze> then I'm wondering where this claim comes that uml is faster.
16:06 < caker> vmware is a beast, in my experience
16:08 < pflanze> BTW even under skas, uml seems to use signals.
16:09 < pflanze> I've seen heavy logging of sig11 under grsecurity, until I've hacked the grsecurity logger.
16:09 < caker> I think that UML uses signals for it's threads (user/kernel/io/is there another?) to talk to one-another
16:10  * pflanze is wondering why sig11
16:12 < pflanze> I'm also wondering about multiple buffering issues.
16:13 < pflanze> I guess it's contra productive to cache pages twice, on the host and in uml.
16:13 < caker> disk IO is slower when you use SYNC for the ubd devices
16:13 < caker> not really
16:14 < pflanze> UML reads a page from ubd, holds it in memory, the host has it in memory too, then uml copies it to it's swap, which means, the host gets another copy, which it hold in memory...
16:15 < caker> the uml's kernel uses caching just like a host would, and stuff it "commits" to disk, the host caches it or writes it to disk .. I would think it's just the risk of crashing and losing the cache, not a performance issue
16:15 < caker> that's why we rely on good VM implementation on the host
16:16 < pflanze> You use memory for nothing but making copies in it.
16:16 < caker> cache, not memory :-)
16:16 < caker> same diff - but that's what a host is for, no?
16:16 < pflanze> No, I think you should give the memory of the host to the uml as ram, and let uml read stuff from disk directly.
16:17 < pflanze> I've actually tried to use raw devices instead of files to attach ubd to, and it works for filesystem partitions,
16:17 < pflanze> but not for swap.
16:18 < pflanze> My first tests show no difference, but it might well be that ubd does not do tricks like prefetching data,
16:18 < caker> right
16:18 < pflanze> which means it's way slower than what the normal kernel does on ide.
16:19 < pflanze> So if the uml kernel would really work on raw devices like it would on ide devices,
16:19 < pflanze> performance should be better.
16:19 -!- thy_ [koubt9am@u-pl7.ms.mff.cuni.cz] has quit [Quit: Client exiting]
16:20 < pflanze> Think "zero copy uml" :)
16:20 < caker> right
16:22 < pflanze> How do you switch off SYNC on ubd devices?
16:22 < pflanze> ah,
16:22 < pflanze> seen it in the manpage.
16:22 < pflanze> I've not used it.
16:23 < caker> I forget the .config option name, but it's something like always-sync-ubd-io .. if that's compiled into the kernel, it's always on, and you can't turn it off
16:23 < pflanze> ah
16:23 < caker> if it's not, then, yeah, you can use the command args
16:23 < pflanze> ok I'll have to fetch the debian sources.
16:24  * pflanze thinks about how odd it is: using double buffering, but at the same time sync.. waste memory really for nothing.
16:26 < pflanze> CONFIG_BLK_DEV_UBD=y
16:26 < pflanze> # CONFIG_BLK_DEV_UBD_SYNC is not set
16:26 < pflanze> so it seems not enabled.
16:29 < pflanze> So if you don't use sync, you should probably use ext3, if you don't want your uml filesystems corrupt upon host crash.
16:31 < caker> right - or uml crash...
16:32 < pflanze> uml crash shouldn't be an issue, a written page will be written regardless whether the uml process crashes.
16:32 < caker> i'm not thinking staight
16:32 < pflanze> This is the same thing as with a normal computer.
16:33 < pflanze> uml writes blocks to ubd -> are being sent to host.
16:33 < caker> yeah, i get it - i was wrong :-)
16:33 < pflanze> normal machine writes blocks to ide dev -> are being sent to disk.
16:34 < caker> ext3 useful as the fs type inside uml, if the uml crashes, doesn't matter to host
16:34 < pflanze> Yes, but I guess ext3 is particularly important on the host.
16:34 < pflanze> Since it journals data.
16:34 < pflanze> not only metadata.
16:35 < pflanze> So if uml writes a block out, and another, and inbetween the host crashes,
16:35 < pflanze> the root_fs file is restored in correct order.
16:35 < caker> but will it still be sane to the UML since it thinks all the blocks have been flushed to "virtual disk"
16:35 < pflanze> In this case, the uml linux can then recover the root filesystem using a journal.
16:36 < caker> using sync, it would be sane
16:36 < pflanze> If the host crashes but root_fs is on, say, ext2,
16:36 < pflanze> the order of the writes are not guaranteed,
16:36 < pflanze> and uml tries to recover it's ext3/reiser/whatever filesystem from journals,
16:36 < pflanze> which will fail since the blocks are not written even if uml has been told by the host that they were.
16:37 < pflanze> Yes, using sync would be safe,
16:37 < pflanze> as would be using raw devices.
16:37 < caker> hmm
16:37 < pflanze> But you said that using sync slows it down even more.
16:37 < caker> I'm assuming only because it can't cache ...
16:37 < pflanze> So maybe it's better to do without sync, but to use ext3 on the host.
16:38 < caker> that's my setup
16:38 < pflanze> ok.
16:42 < pflanze> Heh: looks like you can boot uml from hostfs?
16:42 < pflanze> Wondering if this is faster.
16:44 < caker> pflanze: i think there's a hostfs jail now, too
16:45 < pflanze> I'm currently starting my uml's inside a chroot. (Which is probably mandatory anyway to be secure.)
16:45 < pflanze> grsecurity tries to offer secure root chroot's. This could make it possible to start uml as root with hostfs,
16:46 < pflanze> and do all filehandling on the host's filesystem.
16:46 < pflanze> (Which would save double buffering)
16:46 < pflanze> (at a complexity cost of course)
16:46 < caker> running uml as root sounds scary, fairly simple to break out of a chroot, esp as root
16:47 < pflanze> without grsecurity, definitely.
16:47 < pflanze> But you would *have* to run uml as root, or permissions on hostfs would not work.
16:48 < caker> right
16:50 < smcavoy> Why would you have to run as root for permissions to work?
16:50 < caker> if using hostfs
16:50 < pflanze> smcavoy: if you want to use hostfs as root partition, and i.e. write to /etc, and etc is owned by root,
16:50 < pflanze> the uml process has to run as root on the host.
16:52 < pflanze> (Try it out: run uml as some non-root user, log into uml as root, mount hostfs, list root owned files, try to write to them, -> permission denied.
16:52 < pflanze> Funny :) )
16:55 < smcavoy> I see, does hostfs allow you to "export" particular directories to umls, or does it have to be the whole hostfs
16:56 < pflanze> smcavoy: mount -t hostfs /some/dir/on/the/host /mnt/point/
16:58 < smcavoy> so you could then have a directory on the host containing the root for the uml.
16:58 < pflanze> yes, I think so
16:59 < pflanze> I haven't tried booting from hostfs yet.
17:00 < pflanze> Something like: root@host# chroot /home/uml1 /linux root=/ rootfstype=hostfs
17:12 < pflanze> btw (smcavoy): here's a posting about what we are talking about: http://www.geocrawler.com/archives/3/597/2002/3/0/8258154/
17:13 < pflanze> Looks like currently file owners in hostfs are not correct upon creating files even if running uml as root.
17:16 < Pahan> Do your UML and host userids correspond?
17:18 < pflanze> I would just copy my rootfs contents to the host, cp -a. Then there will be no problem.
17:20 < caker> pflanze: have you run multiple vmware instances with success?
17:20 < pflanze> (Pahan: like root@host# mount /dev/mdX /home/uml1 && mount -o loop root_fs /mnt/tmp && cp -a /mnt/tmp /home/uml1 && chroot /home/uml1 /boot/linux root=/ rootfstype=hostfs)
17:20 < pflanze> caker: dunno. I've never used vmware, only mol.
17:21 < pflanze> afaik, you can have multiple mol instances.
17:21 < caker> weird
17:21 < pflanze> Why?
17:21 < caker> cool weird
17:22 < caker> term of appreciation of mine, sorry
17:22 < pflanze> hehe
17:23 < pflanze> yep, you can, see http://www.maconlinux.org/sshots.html
17:24 < caker> damn
17:24 < caker> looks like fun .. where's my tibook
17:27 < pflanze> There are some issues with the latest versions and macosx that are being discussed.
17:27 < caker> Do you think vmware and MOL are similar in their VM approach?
17:27 < pflanze> I'm still running it with macos9 only. (Would like to try linux in it, soon, but need time for setting another box up for that)
17:27 < pflanze> I think so.
17:28 < caker> I suppose virtual-pc would have been a better choice for my points earlier
17:28 < pflanze> MOL needs a kernel module, then runs a process in userspace.¨
17:29 < pflanze> You list in the mol config, which disk devices are being seen as real disks inside mol.
17:29 < caker> did you hear that vmware was faster overall (fork times, etc), or just disk io? .. i'm curious where the 1:10 host:uml ratio is coming from
17:30 < pflanze> No, I did read that uml was faster than vmware (without numbers or details).
17:30 < pflanze> The fork time difference (10x) I've measured myself, as well as the compilation difference (2x).
17:31 < pflanze> I just don't know if my setup is broken, or if it's state of the uml art.
17:31 < caker> here were my stock kernel compile times:
17:32 < caker> host - 3m11s
17:32 < caker> uml - 4m5s
17:32 < caker> nothing else running
17:32 < pflanze> 4:05 or 4:50?
17:32 < caker> 05s
17:32 < pflanze> That sounds much better.
17:32 < caker> so about a minute worse
17:32 < pflanze> I've compiled the debian mysql4 package.
17:33 < pflanze> gcc-2.95 on both systems, debian woody on both.
17:34 < caker> mysql has many files :-)
17:34 < pflanze> yep, maybe.
17:34 < caker> not sure which takes longer, mysql or stock kernel
17:34 < pflanze> It took 21 mins on the host, 43 mins in uml.
17:35 < pflanze> Regardless whether root_fs was on a reiserfs partition, vfat partition, or a raw device.
17:35 < pflanze> Using reiserfs inside uml and for compilation on the host.
17:36 < pflanze> mem=128M, ~200M swap on the root partition of the uml instance.
17:36 < caker> did you mount a tmpfs someplace and set TMP=?
17:36 < pflanze> yes, mount -t tmpfs -o size=2G /home/uml1/tmp
17:36 < caker> TPM=/home/uml1/tmp ./linux ... 
17:37 < pflanze> No, chroot /home/uml1 linux
17:37 < pflanze> well,
17:37 < caker> er TMP .. ok .. well, skas and the tmpfs trick are the two biggies
17:37 < pflanze> I'm using a perl script for the chroot and uid change.
17:37 < shak> hrm. got a little problem here
17:38 < pflanze> skas is being printed by uml as being used.
17:39 < caker> when you df -h on the host, is some of the tmpfs mount being used? .. probably
17:39 < pflanze> tmpfs                 954M  5.5M  948M   1% /tmp; tmpfs                 1.2G   69M  1.1G   6% /home/uml2/tmp; tmpfs                 2.8G     0  2.8G   0% /home/uml1/tmp
17:39 < pflanze> (sorry about the formatting)
17:40 < pflanze> (currently, uml1 is not running thus no space used)
17:40 < caker> ok
17:40 < shak> yeti: you around?
17:41 < yeti> yes
17:42 < shak> do you have the bridge running for your UML networking devices?
17:42 < yeti> yep
17:42 < yeti> on 3 systems up to now
17:42 < shak> I've got mine set up like david has in http://uml.openconsultancy.com/ 
17:42 < shak> but when I try and ifconfig from within the UML
17:42 < yeti> i did it the debian way
17:43 < shak> I get a permission denied
17:43 < shak> did you get any such problem/
17:43 < yeti> forgot to tunctl ?
17:43 < shak> tunctl -u shak -t uml-shak0
17:43 < shak> shouldn't be a probem?
17:44 < yeti> if it doesnt error and ifconfig shows that dev... i think its ok
17:44 < yeti> i use such simple names like tap4
17:45 < yeti> brctl addif br0 thattap
17:45 < yeti> i am in a hurry... the dog wants her night-walk
17:56 < shak> ok, sorry
17:56 < shak> I had to go afk.
18:11 -!- smcavoy [~smcavoy@grimlock.drive-megawheels.net] has quit [Quit: ACK!]
18:27 -!- yeti [~yeti@p3EE22A19.dip.t-dialin.net] has quit [Quit: Client exiting]
18:48 -!- mistral [mistral@212.159.71.212] has quit [Ping timeout: 488 seconds]
18:51 -!- caker [~null@pcp507591pcs.nash01.tn.comcast.net] has quit [Ping timeout: 492 seconds]
20:16  * alt is away: cleaning how
20:16  * alt is away: cleaning how
20:16  * alt is away: cleaning house
20:17 < DnsInfector_> alt: you could also have a look at mine ;)
20:17 < alt> hee hee
20:17 < alt> mine's enough trouble
20:24 -!- DnsInfector_ [~DnsInfect@ARennes-204-1-10-115.w81-53.abo.wanadoo.fr] has quit [Quit: Client exiting]
21:39 -!- pustota1 [~pustota1@217.224.22.79] has quit [Ping timeout: 488 seconds]
22:41 -!- caker [~null@68.52.196.167] has joined #uml
23:49 < david> hello
23:51 < alt> hello
--- Log closed Wed Jun 18 00:00:00 2003