Back to Home / #uml / 2007 / 10 / Prev Day | Next Day
#uml IRC Logs for 2007-10-17

---Logopened Wed Oct 17 00:00:06 2007
00:45|-|aroscha [] has joined #uml
00:47|-|aroscha [] has quit []
01:05|-|alb [] has joined #uml
01:07|-|albertito [] has quit [Ping timeout: 480 seconds]
01:57|-|balbir [~balbir@] has joined #uml
02:29|-|balbir [~balbir@] has quit [Read error: Operation timed out]
02:35|-|alb [] has quit [Ping timeout: 480 seconds]
02:46|-|Ancalagon [] has quit [Ping timeout: 480 seconds]
03:04|-|tyler29 [] has joined #uml
03:33|-|aroscha [] has joined #uml
03:55|-|ftumch [~James@] has quit [Quit: Goodbye.]
03:56|-|ftumch [] has joined #uml
04:13|-|tyler29 [] has quit [Ping timeout: 480 seconds]
04:21|-|duns_s [] has joined #uml
04:23<duns_s>is there any way to circumvent the 2 gb hostfs limit?
04:26<duns_s>and i cannot mount humfs, is there something i need to add in my uml box?
04:31|-|tyler29 [] has joined #uml
04:33<duns_s>Magotari: yes, but currently it is possible to prevent the machine to access the local network via iptables on the host
04:33<duns_s>and hostfs is really fast
04:34<duns_s>is humfs better?
04:34<duns_s>mount says it does not know humfs inside my uml machine.
04:40<Magotari>duns_s: Sorry for just typing 4 characters, I am on the phone here.
04:40<Magotari>I have no humfs experience, so I cannot say anything about it. I think it would be slower than hostfs.
04:41<Magotari>An idea could be to use a loopback mount, but it kinda begs for file corruption.
04:41<Magotari>A file containing a filesystem, connected to UML as ubdb, also loopback mounted on the host.
04:41<Magotari>Might be a disaster, but that is the only idea I have at the moment.
04:42<duns_s>Magotari: i've tried that with ext3 ro mount on the host
04:42<Magotari>Back to phone for now.
04:47<duns_s>it works
04:47<duns_s>i just don't get why hostfs needs the 2 gb limit, it is restricted by the filesystem on the host anyway
04:47<duns_s>isn't it?
04:48<duns_s>if i try to create a file which i am not allowed to, the host's filesystem blocks hostfs from doing it and hostfs simply throws that error
04:52<duns_s>the problem with loopback mounting ro is, that ext3 (and most likely any fs) does not check for filesystem changes not made by itself
04:53<duns_s>so if i create or delete files inside the vm, i get errors on the host mounted loopback ext3
05:42|-|duns_s [] has quit [Ping timeout: 480 seconds]
05:43|-|tyler29 [] has quit [Ping timeout: 480 seconds]
06:00|-|tyler29 [] has joined #uml
06:57|-|aroscha [] has quit [Quit: aroscha]
07:16|-|aroscha [] has joined #uml
08:28|-|dang [] has joined #uml
08:29|-|tyler29 [] has quit [Ping timeout: 480 seconds]
08:43|-|tyler29 [] has joined #uml
09:14|-|hfb [] has joined #uml
09:26|-|aroscha [] has quit [Quit: aroscha]
09:41|-|dang [] has quit [Quit: Leaving.]
09:46|-|jdike [] has joined #uml
09:46<jdike>Hi guys
09:54[~]dgraves hands jdike the newspaper and leave.
09:54<dgraves>your turn!
09:55<jdike>there's a dog here whose nose needs hitting?
10:00<dgraves>its a fly.
10:00<dgraves>i kept trying to swat it.
10:08<jdike>well, systemtap works again
10:15<jdike>now I can start poking around the host kernel to figure out where these bogus SIGTRAPs are coming from
10:16[~]dgraves suspects aliens.
10:17[~]jdike suspects dgraves is a few tentacles short of a full UFO
10:19<dgraves>jdike: you're not the first one, but so far, the official investigation and govt coverup shows nothing.
10:23<jdike>well I guess we need some devoted wackos to uncover THE REAL TRUTH!!!
10:40[~]dgraves wouldn't like to be uncoverd.
10:41<jdike>we wouldn't like that either
10:42<dgraves>jdike: man, what did you do to this channel yesterday?
10:42<dgraves>there's no one here!
10:43<jdike>you don't want to knwo
10:45<dgraves>jdike: can you think of an actual usage case for double mounting either the same device or a different device on the same mnt point?
10:47|-|ram [] has joined #uml
10:47<jdike>not offhand
10:49<dgraves>guess we'll deny that.
10:53<jdike>I can't reproduce the crashme problem any more
10:53<jdike>once I have the weapons ready, working, and aimed, the bug goes into hiding
10:56<dgraves>of course.
10:57<dgraves>wouldn't you if i had a nuke pointed at you?
10:58<jdike>but it is inconvenient
10:58<dgraves>try sweets!
10:58[~]dgraves puts out bug juice for the bug.
12:22<Magotari>When I found the bug I wanted to be more useful than just a report, so I started stracing UML. Gigabytes of logs later, it did not even flinch...
12:23<Magotari>No crashes, no error messages, no nothing.
12:30<dgraves>Mmm.... timing issues. :)
12:35|-|the_hydra [~the_hydra@] has joined #uml
12:37<jdike>except everything relevant is the same
12:37<jdike>same host kernel, same UML
12:37<jdike>nevermind that the thing just did a FC5 -> FC6 upgrade
12:37[~]the_hydra smells problem...
12:40<jdike>actually, the mysterious lack of a problem
12:42<Magotari>I wish you could do COW backing on real filesystems, without UML. I would love to be able to back out changes to my real systems after they happen, but I don't always feel like running a virtual machine.
12:42<jdike>port the COW driver to the host
12:42<Magotari>Savestates, especially if hotpluggable, would be damn handy.
12:42<jdike>there's also dm_snap, I think
12:42<the_hydra>jdike: almost said the same thing...
12:42<the_hydra>Magotari: read about device mapper...
12:43<Magotari>Yeah, much easier said than done, porting. I tried porting once, and I ended up at little more than: int main(){}
12:43<the_hydra>Magotari: or use filesystem like ZFS which supports snasphots
12:43<Magotari>Not knowing C does not help. Grrrr.
12:43<the_hydra>jdike: how are you today?
12:46<jdike>some joker decided to try humfs and discovered that it doesn't work
12:46<jdike>so I'm sitting here waiting for humfsify to finish making me a filesystem
12:48<jdike>it ain't healthy, but I can get a shell
12:48<Magotari>By the way... Does the sound relay thing work, or was it only on my computer that I get horrible noise?
12:49<jdike>lrwxrwxrwx 1 root root 15 Jun 2 21:54 rc.sysinit -> rc.d/rc.sysinit
12:49<jdike>?--------- ? ? ? ? ? rc0.d
12:49<jdike>?--------- ? ? ? ? ? rc1.d
12:49<jdike>I haven't played with it in years, but it has worked
12:50<the_hydra>?? what were you doing jeff? :)
12:52<jdike>seeing how well humfs works
12:53|-|mgross [] has joined #uml
13:00|-|balbir [~balbir@] has joined #uml
13:04<jdike>and I trashed the ubd filesystem I used to populate the humfs mount
13:05<jdike>I forgot you need to make a separate copy of the data before running humfsify
13:07|-|Baltam8 [] has quit [Remote host closed the connection]
13:07[~]jdike sure hopes humfsify -r works
13:07<the_hydra>jdike: keep the good work!
13:08<jdike>keep me away from anything you care about
13:09<jdike>looks OK
13:11<Magotari>I just had a thought. Would it work if I put a rootfs file in nfs and used UML from a different computer? I think that could be the way my school gets Linux classes which are not limited to being a user.
13:20|-|aroscha [~aroscha@] has joined #uml
13:21|-|Baltam8 [] has joined #uml
13:21|-|aroscha [~aroscha@] has quit []
13:27<jdike>try this again
13:27[~]jdike recovered the ubd image
13:40|-|aroscha [~aroscha@] has joined #uml
13:47|-|mgross [] has quit [Ping timeout: 480 seconds]
13:49<jdike>solution #1 - don't use an old buggy humfsify
14:03|-|aroscha [~aroscha@] has quit [Ping timeout: 480 seconds]
14:04|-|tyler29 [] has quit [Ping timeout: 480 seconds]
14:15|-|remus [~remus@] has quit [Remote host closed the connection]
14:16|-|the_hydra [~the_hydra@] has left #uml []
14:17|-|aroscha [] has joined #uml
14:18|-|tyler29 [] has joined #uml
14:33|-|remus [~remus@] has joined #uml
14:38|-|balbir [~balbir@] has quit [Ping timeout: 480 seconds]
14:46|-|aroscha [] has quit [Quit: aroscha]
15:02|-|tyler29 [] has quit [Ping timeout: 480 seconds]
15:23|-|tyler29 [] has joined #uml
15:44|-|remus [~remus@] has quit [Remote host closed the connection]
15:46|-|tyler29 [] has quit [Ping timeout: 480 seconds]
15:47|-|remus [~remus@] has joined #uml
16:03|-|tyler29 [] has joined #uml
16:34|-|tyler29 [] has quit [Ping timeout: 480 seconds]
16:36|-|newbie [] has joined #uml
16:41<newbie>Is it possible to set a static ip AND default network route in one step while launching uml. My goal is to launch kickstart over the network so I can easily build a nice custom uml machine which has everything exactly the way we set machines up here
16:42<newbie>i know the static ip part should be easy - just don't know the exact eth0=blah syntax that will be needed
16:43<newbie>I would be happy with a pointer to that sort of info and info on if I can setup the default route as well - the wiki uml network page does not mention details on these things
16:57<jdike>there's nothing UML-specific there
16:58<jdike>an ifconfig-eth0 file will do the right thing
17:02|-|hfb [] has quit [Quit: Leaving]
17:03<Magotari>As lame as it can be, but I usually just put an "ifconfig eth0 up; route add default gw" into my local rc file. It works. UML inside is just a normal system, you can use it as such.
17:04<Magotari>I use tuntap networking, and that seems to do the trick. The helper program will take care of the host side, or use a preconfigured device.
17:05<Magotari>You can edit your distro's network setup files to get it working right too. UML, other than being an imaginary computer, is a normal system.
17:16<caker>can also assign a specific mac address to each UML, and use a dhcp server which will assign ips based on the mac of the request
17:23<newbie>caker: yes, that's what I'm looking for since I want to boot from a kickstart image ....I can do it using a static IP on the command line but need the routing to work or I get no where.. hmmm not sure kickstart is in the standard linux kernel boot options...not even sure how to find out
17:31<jdike>there needs to be something in the filesystem which matches an IP and default route to a MAC
17:31<jdike>and you put the MAC on the UML command line
17:31|-|jdike [] has quit [Quit: Leaving]
17:32<caker>set your UML image to set up networking via DHCP, and set up a dhcp server like I mentioned
17:33<caker>boot it, sends out dhcp request, assigns IP *and* default route. done.
17:33<newbie>i have dhcp-fwd but it wants to know interface names to listen to - it's obvious I can give it the real eth0 but I don't know what tuntap will assign until uml comes up
17:33<caker>bridging on the host is the way to go...
17:34<newbie>caker: no I can't setup anything on the root filesystem - it's all being kickstarted from a small boot image which then goes out over the network to do the real install
17:34<newbie>ahh bridging hmm...I will check that out
17:36<newbie>Great! Thanks again - I will jump on that some time tomorrow - looks like a nice article
17:36<newbie>(I'm trying to set this stuff up on the side since I think it will be helpful for various things)
17:36<newbie>that are work related... anyway.. thanks
17:37<Magotari>I need to try bridging one day. Ah, one day...
17:37<Magotari>What are the benefits above tuntap? Just curious
17:41<caker>you mean using over uml_switch or whatever it's called?
17:41<Magotari>No, over tuntap.
17:41<caker>it uses tuntap
17:42<Magotari>Ok, over pure tuntap then, without bridging. Sorry, my networking never was strong, so I have trouble expressing myself.
17:43<caker>AFAIK, you need to do some route magic if you just create a tap device on the host and assign it to a UML
17:43<caker>another option is to wire it into uml_switch, which requires the userspace stuff to run (slow)
17:43<caker>bridging rocks because it's literally bridging the UML onto your network.
17:43<Magotari>There is a program which does the routing magic for you, but it needs setuid. uml_net maybe?
17:44<caker>yeah, one of those
17:44<Magotari>Changes some arp, forwarding stuff on the host.
17:44<Magotari>Userspace, but I think it runs just once.
17:45<Magotari>I will have to try a connection one day, to see if my approach really gets UML out onto the network, or even The Network.
17:47<Magotari>I see bridging lets through dhcp. This is good.
18:40|-|albertito [] has joined #uml
18:57|-|albertito [] has quit [Quit: q]
19:03|-|albertito [] has joined #uml
19:52|-|albertito [] has quit [Quit: q]
19:56|-|albertito [] has joined #uml
21:08|-|Nem^ [] has quit [Ping timeout: 480 seconds]
21:17|-|Nem^ [] has joined #uml
21:31|-|mgross [] has joined #uml
22:55|-|balbir [~balbir@] has joined #uml
22:58|-|VS_ChanLog [] has left #uml [Rotating Logs]
22:58|-|VS_ChanLog [] has joined #uml
23:03|-|balbir [~balbir@] has quit [Ping timeout: 480 seconds]
23:06|-|mgross [] has quit [Read error: Operation timed out]
23:14|-|balbir [~balbir@] has joined #uml
23:21|-|silug [~steve@] has quit [Read error: Connection reset by peer]
23:26|-|balbir [~balbir@] has quit [Read error: Operation timed out]
23:38|-|balbir [~balbir@] has joined #uml
23:53|-|balbir [~balbir@] has quit [Ping timeout: 480 seconds]
23:55|-|albertito [] has quit [Quit: q]
23:58|-|albertito [] has joined #uml
---Logclosed Thu Oct 18 00:00:36 2007