Back to Home / #uml / 2008 / 12 / Prev Day | Next Day
#uml IRC Logs for 2008-12-18

---Logopened Thu Dec 18 00:00:14 2008
00:32-!-balbir_ [~balbir@122.167.212.247] has quit [Ping timeout: 480 seconds]
02:12-!-balbir_ [~balbir@59.145.136.1] has joined #uml
04:16-!-Plouj [~Plouj@red.cs.yorku.ca] has quit [Read error: Operation timed out]
04:21-!-Plouj [~Plouj@red.cs.yorku.ca] has joined #uml
09:28-!-balbir_ [~balbir@59.145.136.1] has quit [Ping timeout: 480 seconds]
09:34-!-tchan [~tchan@c-67-173-9-133.hsd1.il.comcast.net] has quit [Quit: WeeChat 0.2.7-dev]
09:34-!-tchan [~tchan@c-67-173-9-133.hsd1.il.comcast.net] has joined #uml
09:56-!-Basic [~Basic@c-75-73-131-27.hsd1.mn.comcast.net] has quit [Quit: Basic]
10:14-!-Nem^ [~Nem@dslb-084-056-208-180.pools.arcor-ip.net] has joined #uml
10:46-!-xinming-host [~hyy@125.109.240.21] has joined #uml
11:00<xinming-host:#uml>http://rafb.net/p/ndTGpC43.html <-- Anyone here can help me this out please?
11:07<xinming-host:#uml>hmm, BTW, I saw that I do a typo about udb, It should be ubd, After I tried with ubd, It's almost the same error
11:14<xinming-host:#uml>my stupid, I didn't compile the UBD driver in.
11:26-!-jdike [~jdike@pool-96-252-45-50.bstnma.fios.verizon.net] has joined #uml
11:26<jdike:#uml>Hi guys
11:31-!-rob0 [~rob0@cardinal.lizella.net] has quit [Ping timeout: 480 seconds]
11:58-!-rob0 [~rob0@cardinal.lizella.net] has joined #uml
12:23-!-balbir_ [~balbir@122.167.204.221] has joined #uml
12:25-!-Basic [~Basic@32.178.10.226] has joined #uml
12:37<xinming-host:#uml>Anyone here can tell me why won't uml create pts for me?
12:38<xinming-host:#uml>./linux con=pts blabla, <-- This won't create any pts device, and uml seems to be hung
12:41-!-Petter [~Petter@75.52.217.234] has joined #uml
12:44-!-Basic [~Basic@32.178.10.226] has quit [Ping timeout: 480 seconds]
12:52<jdike:#uml>add con0=fd:0,fd:1
12:52<jdike:#uml>you'll see more output
12:59<xinming-host:#uml>jdike: thanks. :-) hmm, BTW, Is it possible to get the virtual console list from command line rather than check log by hand please?
13:00<jdike:#uml>mconsole will tell you
13:01*xinming-host:#uml is trying
13:04<jdike:#uml>uml_console config con1
13:04<xinming-host:#uml>jdike: Sorry, I didn't see how mconsole get the list, Any hint?
13:04<xinming-host:#uml>jdike: Oh, Thnkas
13:04<jdike:#uml>will tell you how console1 is hooked up to the host
13:06<xinming-host:#uml>jdike: May I ask what does the first OK mean for when we do config?
13:06<xinming-host:#uml>oops
13:06<xinming-host:#uml>my stupid
13:07<xinming-host:#uml>ERR || OK.
13:07<jdike:#uml>it just means it communicated with the UML
13:07<jdike:#uml>right
13:10<xinming-host:#uml>After I tried various virtualization tech, I found that uml is almost the most right one as paravirtualization thing for desktop. :-)
13:15<xinming-host:#uml>BTW, may I know what the difference between pts and pty for con? When I try pts, It worked, But when I am doing with pty, There will be errors
14:29-!-balbir_ [~balbir@122.167.204.221] has quit [Ping timeout: 480 seconds]
15:37-!-bunk [~bunk@cs181140183.pp.htv.fi] has joined #uml
15:47<bunk:#uml>anyone here who could help me getting an own UML kernel working?
15:47<caker:#uml>!ask
15:47<linbot:#uml>Don't ask to ask; just ASK
15:48<bunk:#uml>64bit hostkernel, 32bit host userspace, 32bit uml
15:49<bunk:#uml>linux-2.6.24-rc7 from the UML website works fine
15:49<bunk:#uml>self-compiled 2.6.27.3 fails at seemingly random (during each bootup different) parts during init
15:50<jdike:#uml>there's a patch you should try
15:50<jdike:#uml>hold on...
15:51<jdike:#uml>fails how, BTW?
15:51<bunk:#uml>sometimes "Kernel panic - not syncing: Attempted to kill init!"
15:51<bunk:#uml>sometimes:
15:51<bunk:#uml>INIT: Entering runlevel: 2
15:51<bunk:#uml>: : No such file or directory
15:52<bunk:#uml>or:
15:52<bunk:#uml>/lib/lsb/init-functions: line 184: 251 Segmentation fault $TPUT hpa 60 > /dev/null 2>&1
15:52<bunk:#uml>(these 3 examples are from 3 different bootups)
15:53<bunk:#uml>"any started program segfaults with a probability of 50%" would very roughly describe what I'm seeing
15:54<jdike:#uml>hmm
15:54<bunk:#uml>my commandline is: ./linux-2.6.27.3 ubda=uml-fs mem=2G eth0=tuntap,,,192.168.0.254
15:55<jdike:#uml>what happens if you tune down the physmem?
15:55<jdike:#uml>does 2.6.24 run with 2G?
15:56<bunk:#uml>yup, options are the same
15:56<bunk:#uml>mem=200M fails the same way
15:57<jdike:#uml>OK
15:59<jdike:#uml>361371201b60ffd686a694c848c1d5ad6061725f
15:59<jdike:#uml>that's a git commit, BTW :-)
15:59<bunk:#uml>already got it in Linus' tree
16:00<jdike:#uml>you're running 2.6.27, right?
16:00<bunk:#uml>2.6.27.3
16:00<jdike:#uml>right, that one's in 2.6.28-rc8
16:01<jdike:#uml>not sure if it matches your symptoms though
16:01<bunk:#uml>(2.6.27.8 with defconfig had the same problems)
16:02<bunk:#uml>doesn't help
16:03<bunk:#uml>64bit hostkernel, 32bit host userspace is supported?
16:03<jdike:#uml>yes
16:03<jdike:#uml>that really has nothing to do with UML
16:04<jdike:#uml>as long as the kernel runs 32 bit userspace OK, 32-bit UML will be fine
16:04<rob0:#uml>I do that on my laptop :)
16:04<rob0:#uml>(so I can have all my 4G RAM)
16:04<bunk:#uml>what puzzles me is how completely random it is where exactly it breaks
16:05<bunk:#uml>BTW: sometimes the host dmesg says "general protection ip:f7eadf84 sp:1405bd24 error:0 in libc-2.7.so[f7e81000+155000]"
16:06<bunk:#uml>both host and guest are roughly Debian lenny
16:12<bunk:#uml>is there anything in the hostkernel that might confuse UML
16:13<bunk:#uml>?
16:23<bunk:#uml>I assume it's not a problem that my host has 8 GB RAM and 4 GB shm?
16:25-!-ram [~ram@pool-71-245-96-245.nycmny.fios.verizon.net] has quit [Quit: Leaving]
16:49-!-ferret_0567 [~ferret_05@cpe-70-120-93-94.satx.res.rr.com] has joined #uml
17:01<jdike:#uml>back
17:01<jdike:#uml>Father showed up and I had to be sociable for a while...
17:03-!-hfb [~hfb@pool-96-247-109-136.lsanca.dsl-w.verizon.net] has joined #uml
17:05<jdike:#uml>that smacks of memory corruption, although I'd expect kernel crashes rather than userspace crashes
17:07<bunk:#uml>am I the only one with this problem?
17:07<jdike:#uml>the only one I've heard of
17:09<bunk:#uml>are there any host kernel or host userspace settings UML does not like?
17:09<bunk:#uml>or what changed in uml between 2.6.24-rc7 and 2.6.27?
17:15<bunk:#uml>sometimes the kernel crashes
17:15<bunk:#uml>http://pastebin.com/m5a33d113
17:16<jdike:#uml>that patch went into 2.6.27.9, not .8
17:16<bunk:#uml>what patch?
17:16<jdike:#uml>361371201b60ffd686a694c848c1d5ad6061725f
17:16<bunk:#uml>I applied it and it didn't help
17:17<jdike:#uml>that's init dying somehow, not the kernel itself
17:17<jdike:#uml>OK
17:17*bunk:#uml becomes kinda blind after midnight...
17:18<jdike:#uml>there's nothing on the host which should confuse UML like this
17:19<jdike:#uml>Just so I can play with this here - 64-bit host, any recent 32-bit UML, defconfig?
17:19<bunk:#uml>yes
17:19<jdike:#uml>OK
17:19<bunk:#uml>UML fs is a Debisn lenny, 350GB
17:19<jdike:#uml>OK
17:19<jdike:#uml>I doubt that has much to do with it
17:20<jdike:#uml>unless lenny is completely borked, of course :-)
17:20<bunk:#uml>but then I wouldn't be the first to report it...
17:21<jdike:#uml>yeah
17:22<bunk:#uml>make ARCH=um SUBARCH=i386
17:23<jdike:#uml>That hasn't worked for me
17:23<jdike:#uml>I build on 32-bit and copy it over
17:23<bunk:#uml>my host userspace is 32bit
17:23<jdike:#uml>Oh yeah
17:24<jdike:#uml>that'll help
17:24<jdike:#uml>do you even need SUBARCH?
17:24<jdike:#uml>your uname should say i686 or something 32-bit-ish
17:24<bunk:#uml>no, uname says x86_64
17:25<jdike:#uml>hmm
17:25<bunk:#uml>that's expected in a 64bit kernel 32bit userspace setup
17:25<jdike:#uml>OK
17:25<jdike:#uml>unexpected by me, but what do I know
17:31<bunk:#uml>anything interesting in http://pastebin.com/m7624e28c ?
17:33<jdike:#uml>Detected host type: x86_64 (GDT indexes 12 to 15)
17:33<jdike:#uml>questionable, but I think that may be OK
17:34<jdike:#uml>that would explain lots of userspace segfaults though
17:34<jdike:#uml>but that should be consistent
17:34<jdike:#uml>any chance you could bisect it?
17:35<bunk:#uml>I'll try
18:07-!-SNy [9426e8112a@bmx-chemnitz.de] has quit [Ping timeout: 480 seconds]
18:36<bunk:#uml>536788fe2d28e11db6aeda74207d95d750fb761f is first bad commit
18:36<bunk:#uml>commit 536788fe2d28e11db6aeda74207d95d750fb761f
18:36<bunk:#uml>Author: Jeff Dike <jdike@addtoit.com>
18:36<bunk:#uml>Date: Fri Feb 8 04:22:07 2008 -0800
18:36<bunk:#uml> uml: runtime host VMSPLIT detection
18:46<bunk:#uml>jdike: workaround that fixes it for 2.6.27.10: cp arch/um/os-Linux/sys-x86_64/task_size.c arch/um/os-Linux/sys-i386/task_size.c
19:28<bunk:#uml>on 32bit: (long) 0x7fc0000000 = 0xC0000000
19:29<bunk:#uml>and the latter was the HOST_VMSPLIT_3G value
19:30<rob0:#uml>oh nifty, I know that when I had a nonstandard vmsplit some years ago, UML didn't work.
19:30<bunk:#uml>I don't understand what this code is all about, and I might be wrong, but I have the dumb feeling that it worked before was more accidentally...
19:33<bunk:#uml>rob0: UML now automatically tries to adapt for that; but whatever it does for 32bit guest on 64bit hostkernel seems strange
19:39<rob0:#uml>x86_64 doesn't have that vmsplit issue, does it?
19:39<bunk:#uml>no, 64bit is (currently) enough for everything
20:06<bunk:#uml>not sure whether all i said above makes sense - it's 3 in the morning and _really_ time to go to bed...
20:51-!-jdike [~jdike@pool-96-252-45-50.bstnma.fios.verizon.net] has quit [Quit: Leaving]
22:00-!-hfb [~hfb@pool-96-247-109-136.lsanca.dsl-w.verizon.net] has quit [Quit: Leaving]
22:07-!-Nem^ [~Nem@dslb-084-056-208-180.pools.arcor-ip.net] has quit [Ping timeout: 480 seconds]
22:14-!-balbir_ [~balbir@122.167.204.221] has joined #uml
22:17-!-Nem^ [~Nem@dslb-084-056-242-088.pools.arcor-ip.net] has joined #uml
23:04<xinming-host:#uml>Anyone here would give me a rough performance comparison bwteen KVM and UML only for linux system please?
23:05<catsup:#uml>xinming-host: i don't know anything about it anyway, but it would certainly depend on the type of usage, which you should specify in case someone who knows something comes around to answer your question.
23:07<xinming-host:#uml>catsup: No disk io, As a private server. I am considering converting all my KVM linux guest into UML, As IMO, uml is the one most fit for linux guest.
23:07<xinming-host:#uml>catsup: Only CPU part and ram access etc.
23:13<rob0:#uml>Well, I use kvm for (yuck) Windows when necessary. I never saw the need to try a Linux in it.
23:14<rob0:#uml>kvm can be "headless" like UML, using VNC for the console.
23:19<xinming-host:#uml>rob0: Yes, After I know uml, I don't think Linux should be in KVM either. Just curious of the overhead of UML vs KVM in CPU and ram, As that's what I normally care about
23:21<rob0:#uml>Don't know, and I wouldn't trust benchmarks you might find online, because they might be running a full GUI in the guest OS, which to me seems silly, but seems to be popular.
23:22<rob0:#uml>I'm mainly interested in these things for server & networking functionality.
23:32<catsup:#uml>myself i just use uml to test boot scripts
23:32<catsup:#uml>so i know nothing about performance issues :)
23:47<xinming-host:#uml>Ok, Is this a bug? When we start uml with con=pts con0=fd:0,fd:1, After the system starts, I see Virtual console 1 assigned device '/dev/pts/9' in the log, After that, I use "screen /dev/pts/9", Then, I kill the screen by doing "C-a :quit"
23:48<xinming-host:#uml>After we do this, The tty will disappear
23:49<xinming-host:#uml>The device in /dev/pts wil disappear
23:54<catsup:#uml>screen does that
23:54<catsup:#uml>i didn't know you could attach to a pts like that in screen...
23:56<xinming-host:#uml>catsup: We could. :-)
23:56<xinming-host:#uml>catsup: learnt from setting up socket for KVM.
23:57<catsup:#uml>well anyway screen can attach to an existing pts, but then it destroys the pts when it detaches, then i would say that is a screen bug
23:57<catsup:#uml>well anyway IF screen can ...
23:57<catsup:#uml>it should only destroy the pts if it created it
23:58<xinming-host:#uml>catsup: thanks for the explanation, I ever thought that uml has bug. >_<
23:58<catsup:#uml>haha i think uml might have a bug or two
23:58<catsup:#uml>actually let me ask you
23:59-!-VS_ChanLog [~stats@ns.theshore.net] has left #uml [Rotating Logs]
23:59-!-VS_ChanLog [~stats@ns.theshore.net] has joined #uml
23:59<xinming-host:#uml>catsup: ?
23:59<catsup:#uml>when your uml kernel panics does uml segfault?
23:59<catsup:#uml>for example, if you kill init, does it segfault? it does for me
23:59<xinming-host:#uml>catsup: If there is error, uml will segfault *ALWAYS*. :-)
23:59<xinming-host:#uml>catsup: even in a debug, Just run linux.uml
23:59<xinming-host:#uml>oops
---Logclosed Fri Dec 19 00:00:03 2008