| --- | Log | opened Mon Jan 01 00:00:52 2007 |
| 01:30 | |-| | buggs [~noidentd@shadow.splashground.de] has quit [Read error: Connection reset by peer] |
| 01:31 | |-| | buggs [~noidentd@shadow.splashground.de] has joined #xen |
| 05:15 | |-| | buggs [~noidentd@shadow.splashground.de] has quit [Read error: Connection reset by peer] |
| 05:15 | |-| | buggs [~noidentd@shadow.splashground.de] has joined #xen |
| 06:13 | |-| | KriP [kp@yancey.unixworld.dk] has joined #xen |
| 06:22 | |-| | mejlholm [~mejlholm@port79.ds1-abc.adsl.cybercity.dk] has joined #xen |
| 06:25 | |-| | cableman [~cableman@0x503fe15b.hrnxx3.adsl-dhcp.tele.dk] has joined #xen |
| 06:45 | |-| | cableman [~cableman@0x503fe15b.hrnxx3.adsl-dhcp.tele.dk] has quit [Ping timeout: 480 seconds] |
| 06:57 | |-| | cableman [~cableman@0x503fe15b.hrnxx3.adsl-dhcp.tele.dk] has joined #xen |
| 07:34 | |-| | cableman [~cableman@0x503fe15b.hrnxx3.adsl-dhcp.tele.dk] has quit [Quit: Leaving] |
| 07:34 | |-| | cableman [~cableman@0x503fe15b.hrnxx3.adsl-dhcp.tele.dk] has joined #xen |
| 08:31 | |-| | agrif_away changed nick to agriffis |
| 08:59 | |-| | agriffis changed nick to agrif_away |
| 09:12 | |-| | tarawa [~tarawa@h199.8.55.139.ip.alltel.net] has joined #xen |
| 09:16 | |-| | swingle [~swingle@41.Red-81-40-78.staticIP.rima-tde.net] has joined #xen |
| 09:16 | <swingle> | hi |
| 09:18 | |-| | mulix [~mulix@87.69.40.180] has joined #xen |
| 09:18 | |-| | tarawa [~tarawa@h199.8.55.139.ip.alltel.net] has left #xen [Leaving] |
| 09:19 | <swingle> | ive got a little problem |
| 09:19 | <swingle> | im looking for a library that allows me to manage domains without having python installed |
| 09:20 | <swingle> | ive read there is something called vm-tools but the link i got is broken and i cant find it |
| 09:20 | <aliguori> | doesn't exist |
| 09:20 | <aliguori> | vm-tools only works for very old versions of xen |
| 09:20 | <swingle> | theres also something called migration tools |
| 09:21 | <aliguori> | never heard of that |
| 09:21 | <swingle> | i can send it to you if you want to take a look |
| 09:21 | <swingle> | its fairly small |
| 09:22 | <aliguori> | is there a website? |
| 09:22 | <swingle> | i think there is |
| 09:22 | <swingle> | gimme a second |
| 09:22 | <swingle> | http://www.diku.dk/~jacobg/ <- second link in the current projects section |
| 09:23 | <aliguori> | oh |
| 09:23 | <aliguori> | those are jacob's tools |
| 09:23 | <aliguori> | yeah, they can do very basic stuff |
| 09:24 | <swingle> | very basic means start, stop and save? |
| 09:24 | <swingle> | thats all i need |
| 09:24 | <aliguori> | i don't know about save |
| 09:25 | <swingle> | btw is there a reason to not provide a C library? |
| 09:25 | <swingle> | or is it only that nobody bothered writing it |
| 09:26 | <aliguori> | the maintainers don't want one |
| 09:26 | <aliguori> | you can't have two tool chains active at a given time either.. that's part of the problem |
| 09:27 | <swingle> | if by toolchain you mean xend+vm+xenstore ... that takes me to the next question, is a user space daemon really needed? |
| 09:28 | <aliguori> | in xen today, yup |
| 09:28 | <aliguori> | b/c xen is stupid in so many ways :-) |
| 09:28 | <swingle> | i havent read xend but for what i know about it it spends most of the time sleeping doesnt it? |
| 09:29 | <swingle> | i mean its used when you want to do something but its not doing real job while you are working inside your VM |
| 09:30 | <aliguori> | that's perhaps a fair characterization |
| 09:31 | <swingle> | my problem is that i need to use xen in an ambedded system and i dont want to install python on it |
| 09:31 | <swingle> | i wonder if i could write/find some utilities and not use a daemon at all |
| 09:47 | <riel> | see http://libvirt.org/ |
| 09:47 | <riel> | you may be able to get away without the Xen tools |
| 09:48 | <swingle> | looks like that talks to xend for what ive read |
| 09:48 | <riel> | I think it makes direct hypervisor calls for many things |
| 09:48 | <riel> | but maybe not all |
| 09:48 | <swingle> | ill check it out |
| 09:49 | <swingle> | is it really that painfull to use xc and xs libraries directly? |
| 09:49 | <riel> | dunno |
| 09:50 | [~] | riel tries to stay away from userland stuff ;) |
| 09:50 | <swingle> | :) |
| 09:50 | <swingle> | ok ill rebuild that question |
| 09:50 | <swingle> | would it be that painfull to use directly hypervisor calls for everything? |
| 09:51 | <riel> | probably |
| 09:51 | <riel> | you'd have to construct the domain's initial pagetables yourself |
| 09:51 | <riel> | and do all kinds of other subtle stuff |
| 09:51 | <swingle> | ic |
| 09:51 | <riel> | using the xc and xs libraries is probably going to be easier |
| 09:58 | <swingle> | :q |
| 09:58 | <swingle> | ooops |
| 09:58 | <swingle> | wrong window |
| 10:26 | <swingle> | riel, looks like libvirt can do everything i need thx for the pointer |
| 10:27 | |-| | swingle [~swingle@41.Red-81-40-78.staticIP.rima-tde.net] has quit [Quit: Leaving] |
| 10:43 | |-| | Shoragan [~shoragan@datenfreihafen.org] has joined #xen |
| 11:19 | |-| | segher [~segher@84-105-60-229.cable.quicknet.nl] has joined #xen |
| 14:05 | |-| | wogri [~wogri@131.130.7.148] has joined #xen |
| 14:07 | |-| | pvanhoof [~pvanhoof@d54C0EE14.access.telenet.be] has joined #xen |
| 14:12 | <wogri> | hi xen developers, I'd like to ask a driver specific question, I hope that is OK. I'm having issues with xen in combination with an Intel PIIX board. I'm using the libata (ata_piix) and AHCI driver, I don't get any "bad" messages during the boot, but disk performance is OUTSTANDING slow - e.g. recovery of an ext3 2G-partition takes about 60 seconds. I am also having huge performance issues in dom0, so it's not domU related, but I don't have those issues if I boot |
| 14:13 | <wogri> | is this a known bug, or do you think it's an kernel configuration issue? |
| 14:14 | <wogri> | xen version: 3.0.3 - linux 2.6.29 |
| 14:14 | <matti> | .29? |
| 14:14 | <matti> | ;] |
| 14:15 | <wogri> | damn :) |
| 14:15 | <wogri> | 2.6.16.29, sorry :) |
| 14:18 | |-| | aw [~awilliam@c-24-9-84-32.hsd1.co.comcast.net] has joined #xen |
| 14:25 | <wogri> | hm... no ideas - anybody? in that case I might go back to vserver on this host again (I have several xen installations without problems, but this piece of hardware won't work with XEN I guess)... |
| 14:26 | <riel> | you may have to wait for Xen to be ported to a newer kernel |
| 14:26 | <riel> | and/or use the Fedora Xen kernel |
| 14:26 | <wogri> | why? |
| 14:26 | <riel> | because that one is based on 2.6.18 (or 2.6.19 for the latest) |
| 14:27 | <wogri> | and that one fixes my issues you think? |
| 14:27 | <wogri> | is it something known to the developers? |
| 14:27 | <riel> | it sounds like it could be an irq routing problem |
| 14:27 | <riel> | the kind of thing where "old kernel" does not always go well with "new board" :) |
| 14:27 | <riel> | but newer kernels work |
| 14:29 | <wogri> | hm... hopefully they do. I don't have a problem waiting, as long as it will work. |
| 14:29 | <wogri> | i'm compiling 3.0.4 right now, but I guess I can stop doing that, as it's still 2.6.16.33. |
| 14:30 | |-| | MarkW [~MarkW@maw48.kings.cam.ac.uk] has joined #xen |
| 14:30 | |-| | mode/#xen [+o MarkW] by ChanServ |
| 14:30 | <@MarkW> | Yo all. |
| 14:30 | <@MarkW> | aliguori: ping |
| 14:30 | <@MarkW> | An Happy New Year to everybody! |
| 14:35 | |-| | asti [~ahowells@omg.plz.dont.haxx.us] has quit [Quit: splat] |
| 14:38 | <aliguori> | MarkW: howdy |
| 14:38 | <aliguori> | MarkW: happy new year to you too :-) |
| 14:39 | <@MarkW> | aliguori: have a good new years eve? |
| 14:39 | <wogri> | riel, I just got a hint, that debian has a 2.6.18 in their repository, too. I'm installing it right now, let's see how it goes. thanks for your advice. |
| 14:39 | <aliguori> | actually, I just crashed. been traveling for the past week and just got in |
| 14:39 | <aliguori> | MarkW: and you? |
| 14:39 | <@MarkW> | aliguori: I went to a new years party at a friends' house |
| 14:40 | <@MarkW> | But was semi-asleep most of the evening. |
| 14:40 | <aliguori> | sweet |
| 14:40 | <aliguori> | hehe :-) |
| 14:40 | <@MarkW> | Just got back from parents house yesterday. |
| 14:40 | <aliguori> | yeah, I was also at my parents over the holidays |
| 14:40 | <@MarkW> | Which implies I'd a) gotten up too early in the morning and b) taken some anti-travel-sickness pills |
| 14:41 | <aliguori> | MarkW: :-) I had to hop between my parents and my gf's parents so I had a couple of flights and then a few hour drive from houston. I think that's what got me :-) |
| 14:41 | <@MarkW> | The combination of those two generally makes me pretty drowsy... when I travelled down to my parents I was falling asleep randomly and not knowing what bits of the night's drive I was dreaming |
| 14:41 | <@MarkW> | I thought I saw mexicans with orange jackets by the road |
| 14:42 | <@MarkW> | But since nobody else saw them I think I probably dreamt that :-) |
| 14:43 | <aliguori> | hehhe |
| 14:43 | <riel> | wogri: yeah, debian uses the fedora patches |
| 14:43 | <aliguori> | I'd reckon that'd be a dream :-) |
| 14:43 | <aliguori> | riel: +1 on your post to xen-devel btw |
| 14:44 | <wogri> | woohooo. I'm pretty excited. hopefully it will work. |
| 14:44 | <@MarkW> | aliguori: although the next day I went into a supermarket and half the staff were dressed as mexicans. No idea why. |
| 14:44 | <@MarkW> | aliguori: ponchos, sombreros, the lot |
| 14:45 | <@MarkW> | aliguori: Anyhow, I've been trying out the paravirt framebuffer stuff. Pretty cool |
| 14:45 | <aliguori> | MarkW: hehe.. in the uk? that's interesting |
| 14:45 | <aliguori> | MarkW: yeah, markus did a great job updating it to xenstore. still needs a fair bit of work though so it'll support multiple resolutions and such |
| 14:45 | <@MarkW> | aliguori: Yeah. After dreaming them during the car journey, I was quite weirded out to see more of them in the shops! |
| 14:46 | <riel> | aliguori: basically, if it's not upstream by the end of the year, they shouldn't bother porting it from 2.6.16 :) |
| 14:46 | <aliguori> | riel: you mean, end of 07 right :-) |
| 14:46 | <riel> | yeah |
| 14:46 | <aliguori> | as the end of the year was yesterday :-) |
| 14:47 | [~] | aliguori wonders if vmware will ever publish the xen vmi rom |
| 14:47 | <aliguori> | i was playing with VMI this past week. it's pretty neat |
| 14:47 | <@MarkW> | aliguori: I gotta say, VMI is a pretty neat idea. |
| 14:47 | <@MarkW> | I've not looked at the implementation, but the concept does make some sense |
| 14:48 | <riel> | it's a bit too low level though, for higher level optimizations |
| 14:48 | <aliguori> | MarkW: I really like that the ROM exposes almost all of the paravirt_ops.. it gives you a ton of flexibility without modifying the kernel. it still has some warts but i was able to get a minimal rom going pretty easily |
| 14:49 | <aliguori> | riel: i think it depends on what high level optimizations your talking about. it's got a pretty reasonable interface for mmu batching and stuff |
| 14:49 | <aliguori> | in fact, i'm hoping to try out the pte update batching with kvm this evening |
| 14:49 | <riel> | aliguori: another issue would be time management, both using the hypervisor as clock source (instead of wanting timer interrupts) and keeping track of steal time |
| 14:50 | <aliguori> | riel: right, vmi defines a virtual timer to deal with that. timer's can be probed easily enough |
| 14:51 | <aliguori> | it's nice though that you can ignore those things to start out with though |
| 14:51 | <riel> | true |
| 14:51 | <riel> | coopvirt++ |
| 14:53 | |-| | cableman [~cableman@0x503fe15b.hrnxx3.adsl-dhcp.tele.dk] has quit [Quit: Leaving] |
| 14:57 | <@MarkW> | aliguori: using the mouse in X over VNC is kinda weird though |
| 14:57 | <@MarkW> | aliguori: does it not use absolute positioning or something similar? The guest and my vnc client seem to have rather different ideas about where the pointer should be... |
| 14:58 | <@MarkW> | Or have I missed a config option somewhere? |
| 14:58 | <riel> | MarkW: your guest OS expects relative mouse movements and not absolute events? |
| 15:00 | <@MarkW> | Well, I expected the paravirt mouse driver would just do the appropriate thing... The HID subsystem in Linux can cope with absolute positioning events, right? |
| 15:01 | <aliguori> | MarkW: it can, but the stupid X evdev driver can't |
| 15:01 | <aliguori> | MarkW: I've authored a VNC extension that allows VNC to use relative coordinates |
| 15:01 | <@MarkW> | Hmmm. Maybe X11 isn't coping... ISTR X11 support for USB devices with abolute positions is relatively recent. |
| 15:01 | <@MarkW> | aliguori: Ah, as I feared. Is that likely to get fixed in X11 at some stage? |
| 15:01 | <aliguori> | I've already got a patch for QEMU |
| 15:01 | <aliguori> | MarkW: yeah, I think some of the RH guys are looking at it |
| 15:02 | <@MarkW> | aliguori: Good :-) |
| 15:02 | <@MarkW> | aliguori: You're producing a lot of cool stuff for Qemu these days :-D |
| 15:02 | <@MarkW> | I rather like the way all the virtualisation projects are stimulating Qemu growth |
| 15:02 | <@MarkW> | I hope somebody hires Fabrice at some point |
| 15:05 | <aliguori> | MarkW: qemu's a lot of fun b/c I can use it on my laptop. I don't want to try and figure out how to get a Xen ubuntu kernel working on my laptop |
| 15:05 | <@MarkW> | Mmmm, it's extremely convenient. |
| 15:05 | <@MarkW> | I've been using it to make CentOS disk images |
| 15:06 | <@MarkW> | But I generally only run Xen on my test machines. Just bought a new box so that I can have test machine at home |
| 15:07 | <@MarkW> | Used to use Intel's facilities in Cambridge, but since they've shut down... |
| 15:07 | <aliguori> | MarkW: I've got a ton of machines running Xen at work, so at home I tend to play with other things :-) |
| 15:07 | <aliguori> | MarkW: yeah, I heard about that, that sucked :-/ |
| 15:08 | <@MarkW> | aliguori: too right! No more free coffee and soda |
| 15:08 | <@MarkW> | I'm not sure how I'll survive now |
| 15:12 | <@MarkW> | Caffiene buffer underflow... |
| 15:13 | <aliguori> | hehe |
| 15:18 | <riel> | btw, the xendomains initscript is pretty cool |
| 15:18 | <riel> | 2 of the 3 guests on my system have a higher uptime than dom0 |
| 15:18 | <@MarkW> | riel: lol :-) |
| 15:19 | <@MarkW> | riel: That is pretty cool. Who was it that added the suspend/resume support to it? |
| 15:19 | <riel> | dunno |
| 15:19 | <riel> | save/restore from the initscripts is pretty nice |
| 15:19 | <@MarkW> | Yeah, they've got pretty clever. |
| 15:20 | <@MarkW> | When I knocked up the first version they were fairly dumb and just shutdown the guests |
| 15:20 | <@MarkW> | But suspending them instead is really really cool ;-) |
| 15:20 | <riel> | if only the dom0 kernel ran in ring 0, so it could suspend laptops and do cpu frequency control, etc... :) |
| 15:21 | <riel> | then we could remove half the code from the hypervisor |
| 15:21 | <riel> | and still run the same guests |
| 15:22 | <aliguori> | riel: there's nothing stopping someone from implementing a linux kernel module that supports the Xen guest ABI |
| 15:22 | <riel> | indeed |
| 15:23 | <riel> | I'm wondering what would be more work |
| 15:23 | <riel> | implementing such a module |
| 15:23 | <riel> | or keep modifying Xen, to keep up with the kernel |
| 15:23 | <riel> | and new hardware |
| 15:27 | <@MarkW> | Somebody could probably port Xen to run as a Linux kernel module, with a bit of hackery... |
| 15:30 | <hensema> | and then call it KVM? |
| 15:30 | <riel> | that's a very different thing |
| 15:31 | <riel> | KVM only does full virtualization |
| 15:31 | <hensema> | my point exactly ;-) |
| 15:32 | [~] | riel points hensema at http://virt.kernelnewbies.org/TechComparison :) |
| 15:33 | <hensema> | yeah, but implementing the xen guest abi in a loadable module would give something quite similar to kvm, wouldn't it? |
| 15:33 | <hensema> | except for the full virt / paravirt |
| 15:34 | <riel> | yes, it would |
| 15:55 | |-| | wogri [~wogri@131.130.7.148] has quit [Quit: leaving] |
| 15:56 | <aliguori> | riel: the xen abi is definitely complex.. |
| 15:58 | |-| | mejlholm [~mejlholm@port79.ds1-abc.adsl.cybercity.dk] has quit [Quit: Leaving] |
| 16:06 | |-| | aliguori [~anthony@cpe-70-112-17-156.austin.res.rr.com] has quit [Ping timeout: 480 seconds] |
| 16:15 | |-| | infernix [~nix@spirit.infernix.net] has quit [Ping timeout: 480 seconds] |
| 16:16 | |-| | aliguori [~anthony@cpe-70-112-17-156.austin.res.rr.com] has joined #xen |
| 16:16 | |-| | KriP [kp@yancey.unixworld.dk] has quit [Quit: Leaving] |
| 16:41 | |-| | infernix [~nix@spirit.infernix.net] has joined #xen |
| 16:48 | <tab> | MarkW: hey ! |
| 17:11 | <@MarkW> | aliguori, riel: actually, istr some company had developed a xen-compliant lkm-based hypervisor for embedded products |
| 17:11 | <@MarkW> | but i've not heard anything since the initial announcement |
| 17:11 | <@MarkW> | tab: yo dude, 'sup? |
| 17:13 | <@MarkW> | Sorry for disappearing - was watching the season finale of Torchwood :-) |
| 17:13 | <riel> | MarkW: Virtual Iron? don't they just use Xen itself? |
| 17:13 | <@MarkW> | riel: not VI, they do indeed just use Xen itself (with some patches) |
| 17:13 | <@MarkW> | I can't remember the company name... |
| 17:15 | <@MarkW> | googling... |
| 17:15 | <riel> | me too :) |
| 17:15 | <@MarkW> | http://linuxdevices.com/news/NS3509840836.html |
| 17:15 | <@MarkW> | " Sombrio Systems is alpha-testing a loadable kernel module and hard real-time application aimed at transforming a running Linux kernel into a Xen-compatible hypervisor. The Xen Loadable Module (XLM) can be used to build multi-function embedded devices in which each function runs in a protected domain, the company says. " |
| 17:15 | <riel> | ohhhh |
| 17:16 | <@MarkW> | Seemed to work very much as if Xen had been ported to run as an LKM, but I remember seeing some suggestion that it was original code. |
| 17:16 | <@MarkW> | That announcement came out in March, tho and I hadn't heard anything since. |
| 17:16 | <@MarkW> | In fact, that news article is the top google hit for "sombrio systems"... |
| 17:17 | <@MarkW> | So it would seem like they don't have much web presence! |
| 17:18 | <riel> | probably not http://www.sombrio.com/ ... |
| 17:18 | <riel> | Registrant: |
| 17:18 | <riel> | Sombrio Systems Inc. |
| 17:18 | <riel> | ATTN: SOMBRIO.COM |
| 17:18 | <riel> | ... crap |
| 17:18 | <riel> | it is them |
| 17:19 | <@MarkW> | Right, not looking so good then. |
| 17:20 | <@MarkW> | Well, whatever became of sombrio, it still seems to me that some sort of loadable Xen port should be doable, a la colinux |
| 17:20 | <riel> | *nod* |
| 17:21 | <riel> | that would give us all kinds of things for free, that are currently a pain with Xen |
| 17:21 | <riel> | full ACPI, kexec, cpu frequency scaling, suspend/resume, ... |
| 17:21 | <@MarkW> | If we used *actual* colinux, we could even run as a module under Windows (in principle) |
| 17:21 | <@MarkW> | Which would rock. |
| 17:25 | <riel> | ahhh, found an email address |
| 17:25 | <riel> | kind of |
| 17:25 | <riel> | https://mail.gna.org/public/xenomai-core/2006-03/msg00059.html |
| 17:26 | <riel> | Due to ADEOS, Xenomai has the |
| 17:26 | <riel> | infrastructure to support doing things like running eCos on one core and |
| 17:26 | <riel> | Linux on the other, |
| 17:28 | <riel> | too bad there was a Novell guy called "Chris Stone" who gets all the high google ratings :) |
| 17:29 | <@MarkW> | How inconsiderate of him :-) |
| 17:34 | <aliguori> | riel: setting up a basic environment where xen guests could run wouldn't be so bad... it's a lot like lhype (need a special IDT to trap hypercalls, etc). |
| 17:34 | <aliguori> | the thing that will be painful is you still need the blkback/netback drivers |
| 17:34 | <aliguori> | host grant table support, and host event channel support |
| 17:34 | <aliguori> | and those drivers are pretty fugly |
| 17:35 | <riel> | yeah :/ |
| 17:35 | <aliguori> | so from the perspective of being a better LKML citizen, i don't think it helps much |
| 17:35 | <aliguori> | but yeah, you get some things for free.. |
| 17:35 | <riel> | and less code duplication in the hypervisor |
| 17:35 | <aliguori> | I'd rather take something like KVM and start paravirtualizing guests with something like VMI |
| 17:35 | <aliguori> | using hardware emulation at first |
| 17:35 | <riel> | coopvirt |
| 17:36 | <aliguori> | yeah |
| 17:36 | <aliguori> | I can't convince rusty to get over doing emulation in lhype :-) |
| 17:36 | <aliguori> | as that would make it useful much more quickly than paravirtualizing everything |
| 17:37 | <riel> | yeah, starting with full virt and moving towards paravirt makes things easier |
| 17:37 | <riel> | the big issue is memory resizing though |
| 17:37 | <riel> | bbl |
| 17:37 | <aliguori> | riel: not so much actually, but yeah, i have to go too, ttyl |
| 18:23 | <@MarkW> | tab: ping |
| 20:01 | |-| | segher [~segher@84-105-60-229.cable.quicknet.nl] has quit [Quit: Leaving] |
| 20:21 | |-| | infernix [~nix@spirit.infernix.net] has quit [Remote host closed the connection] |
| 21:02 | |-| | pvanhoof [~pvanhoof@d54C0EE14.access.telenet.be] has quit [Remote host closed the connection] |
| 21:12 | |-| | infernix [~nix@spirit.infernix.net] has joined #xen |
| 22:59 | |-| | VS_ChanLog [~stats@ns.theshore.net] has left #xen [Rotating Logs] |
| 22:59 | |-| | VS_ChanLog [~stats@ns.theshore.net] has joined #xen |
| --- | Log | closed Tue Jan 02 00:00:14 2007 |
