TheGrebs.com:IRC Logs:#xen:2006:02:21

---	Log	opened Tue Feb 21 00:00:32 2006
00:45	\|-\|	unicks_vodoo [~saneax@59.93.247.97] has quit [Quit: using sirc version 2.211+KSIRC/1.3.11]
01:31	\|-\|	rusty [~rusty@bh02i525f01.au.ibm.com] has quit [Quit: Client exiting]
01:44	\|-\|	hollisb [~hollisb@adsl-67-115-102-122.dsl.sntc01.pacbell.net] has joined #xen
02:15	\|-\|	sdog [~sdog@62.58.98.210] has joined #xen
02:26	\|-\|	mooli [~muli@nesher3.haifa.il.ibm.com] has quit [Quit: My damn controlling terminal disappeared!]
02:31	\|-\|	sdog [~sdog@62.58.98.210] has quit [Quit: Everything is a Fscking DNS Problem]
02:31	\|-\|	sdog [~sdog@62.58.98.210] has joined #xen
02:45	\|-\|	zwane [~zwane@S0106001217d33fa6.vc.shawcable.net] has joined #xen
02:59	\|-\|	hollisb [~hollisb@adsl-67-115-102-122.dsl.sntc01.pacbell.net] has quit [Quit: leaving]
03:14	\|-\|	pvanhoof [~pvanhoof@mailhost.newtec.be] has joined #xen
03:22	\|-\|	zwane [~zwane@S0106001217d33fa6.vc.shawcable.net] has quit [Remote host closed the connection]
03:29	<Shaun>	hmm
03:29	<Shaun>	xm list becuase very slow when say you send shutdowns to 20 guests
03:39	\|-\|	sdague_ [~sdague@serenity.dague.org] has joined #xen
03:39	\|-\|	sdague [~sdague@serenity.dague.org] has quit [Read error: Connection reset by peer]
03:40	\|-\|	sdague_ changed nick to sdague
03:44	\|-\|	xzu_ [~otto@brucester.a20.net] has joined #xen
03:45	\|-\|	xzu [~otto@brucester.a20.net] has quit [Read error: Connection reset by peer]
03:51	\|-\|	tschwinge [cpts2523@l2.cip.ei.uni-stuttgart.de] has quit [Remote host closed the connection]
04:06	\|-\|	ivan [~ikelly@83-71-24-103.b-ras1.lmk.limerick.eircom.net] has joined #xen
04:24	\|-\|	vanhoofp [~pvanhoof@mailhost.newtec.be] has joined #xen
04:24	\|-\|	pvanhoof [~pvanhoof@mailhost.newtec.be] has quit [Ping timeout: 480 seconds]
04:24	\|-\|	vanhoofp changed nick to pvanhoof
05:00	\|-\|	rusty [~rusty@ppp61-121.lns1.cbr1.internode.on.net] has joined #xen
05:01	\|-\|	rusty [~rusty@ppp61-121.lns1.cbr1.internode.on.net] has quit [Quit: ]
05:07	\|-\|	Truck [~constitua@cs181064149.pp.htv.fi] has quit [Quit: Leaving]
05:16	\|-\|	mdday [~mdday@cpe-024-163-120-222.nc.res.rr.com] has joined #xen
06:55	\|-\|	johnlev [~johnlev@nwkea-socks-2.sun.com] has quit [Quit: Leaving]
07:07	\|-\|	athomas [~athomas@host86-136-81-69.range86-136.btcentralplus.com] has joined #xen
07:37	\|-\|	mancdaz [~mancdaz@blueice2n1.uk.ibm.com] has joined #xen
07:39	<mancdaz>	hi
07:40	<mancdaz>	I'm just starting to look at xen - I thought I understood that you needed 2 linux kernels - one for dom0 and one for domU, but on xen-3 that seems not to be the case
07:40	<mancdaz>	so do I use the same kernel for dom0 and domU
07:41	<mancdaz>	ie vmlinux-x-x-x-x-FC5hypervisor
07:41	<riel>	the kernel-xen-<numbers>-FC5guest RPM is smaller
07:42	<mancdaz>	so it is correct that there is a separate kernel for the guests?
07:42	<riel>	yes
07:42	<riel>	not strictly needed, but quite practical
07:43	<mancdaz>	the kernels are now called hypervisor and guest instead of dom0 and domU
07:43	<riel>	yeah
07:43	<mancdaz>	was it dom0 and domU in xen-2 then - maybe I've been mixing up my documentation
07:43	<riel>	no, it was dom0 and domU in FC4
07:43	<mancdaz>	aha
07:44	<riel>	it got changed to hypervisor and guest while I was on holidays ;)
07:44	<mancdaz>	I've been playing on fc4 and fc5
07:44	<mancdaz>	hence the confusion
07:44	<mancdaz>	good holiday?
07:45	<mancdaz>	you could in theory use the hypervisor kernel to create a guest?
07:46	<riel>	yes, you could do that
07:48	<mancdaz>	what would be the significance in terms of priviledges?
07:48	<riel>	none, privilege is determined by the hypervisor not the kernel
07:49	<mancdaz>	you mean the xen hypervisor, as opposed the the linux hypervisor kernel?
07:50	<mancdaz>	which one is the real hypervisor
07:50	<riel>	xen
07:50	<mancdaz>	I think it's confusing to call the dom0 kernel hypervisor
07:51	<riel>	shrug
07:51	<mancdaz>	well, are they both hypervisors?
07:52	<mancdaz>	sorry - I'm being annoying. I'm just trying to get my head round how it all works
07:55	<riel>	yeah, you are ;)
07:56	<riel>	just the name changed
07:56	<riel>	it's still the dom0 equivalent kernel
07:57	<movement>	it does seem a bit of a weird name.
07:59	\|-\|	mdday [~mdday@cpe-024-163-120-222.nc.res.rr.com] has quit [Read error: Connection reset by peer]
08:02	\|-\|	athomas [~athomas@host86-136-81-69.range86-136.btcentralplus.com] has quit [Quit: Leaving]
08:15	\|-\|	mdday [~mdday@cpe-024-163-120-222.nc.res.rr.com] has joined #xen
08:26	<mancdaz>	when I install a new guest using anaconda (in the fc5t3 snapshot this is possible) do I need to create swap space in the virtual drive or will it use the host systems swap space?
08:26	<tonfa>	you have to create it i think
08:28	<riel>	yeah, you can create it in the virtual drive
08:28	<riel>	just like on install of a physical system
08:28	\|-\|	dme_ [~dme@host-84-9-60-139.bulldogdsl.com] has quit [Ping timeout: 480 seconds]
08:29	<mancdaz>	so there is no way for it to share the dom0 swap?
08:29	<riel>	hahahaha
08:29	<riel>	LOL
08:30	<tonfa>	mancdaz: this is not the same system, it would be like sharing the same part of the memory
08:31	<mancdaz>	ok I'm just getting to grips with all this virtualization stuff
08:31	<mancdaz>	I'm sure VM on s390 is able to share paging space
08:32	<tonfa>	because page out is probably done by the hypervisor
08:32	<tonfa>	riel: correct me if i'm wrong
08:33	<riel>	pageout is done by each guest
08:33	<riel>	it's just not possible for multiple kernels to share the same swap space
08:33	<riel>	regardless of whether or not you're virtualizing
08:33	<mancdaz>	ok
08:34	<riel>	and even if you could, it'd be a very dumb thing to do
08:35	<riel>	clustered storage needs communication between the various kernels involved
08:35	<riel>	communication requires a communication path, which requires memory
08:35	<riel>	and by the time you need swap, you might not have that memory left
08:40	\|-\|	toks [~toks@bi01p1.nc.us.ibm.com] has joined #xen
08:42	\|-\|	sdog [~sdog@62.58.98.210] has left #xen []
08:42	\|-\|	pea [~pea@ppp-69-223-79-28.dsl.dytnoh.ameritech.net] has quit [Read error: Connection reset by peer]
08:43	\|-\|	dme [~dme@host-84-9-60-139.bulldogdsl.com] has joined #xen
08:50	\|-\|	ivan_ [~ikelly@83-71-72-159.b-ras1.lmk.limerick.eircom.net] has joined #xen
08:53	\|-\|	ivan [~ikelly@83-71-24-103.b-ras1.lmk.limerick.eircom.net] has quit [Ping timeout: 480 seconds]
08:53	\|-\|	johnlev [~johnlev@nwkea-socks-1.sun.com] has joined #xen
08:57	\|-\|	pea [~pea@ppp-69-223-82-199.dsl.dytnoh.ameritech.net] has joined #xen
09:01	\|-\|	zwane [~zwane@S0106001217d33fa6.vc.shawcable.net] has joined #xen
09:01	\|-\|	zwane [~zwane@S0106001217d33fa6.vc.shawcable.net] has quit [Remote host closed the connection]
09:01	\|-\|	toks [~toks@bi01p1.nc.us.ibm.com] has quit [Remote host closed the connection]
09:05	\|-\|	dansmith [~dan@129.33.1.37] has joined #xen
09:06	\|-\|	Basic_py [~Basic@warden.real-time.com] has quit [Quit: Leaving]
09:15	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has joined #xen
09:29	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has quit [Quit: Ex-Chat]
09:32	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has joined #xen
09:35	\|-\|	rharper [~rharper@pixpat.austin.ibm.com] has joined #xen
10:02	<mancdaz>	ok I've managed to install a guest fc5t3 system, but now when I try and create a domain I get an error
10:02	<mancdaz>	"using config file /etc/xen/guest1"
10:02	<mancdaz>	Error: [Errno 17] File exists
10:05	\|-\|	stekloff [~stekloff@pool-71-245-110-186.ptldor.fios.verizon.net] has joined #xen
10:09	\|-\|	jerone [~jerone@pixpat.austin.ibm.com] has joined #xen
10:12	<tessier>	Anyone using X under xen yet? I'd like to run my desktop in a domU instead of dom0
10:13	\|-\|	cableman [~cableman@3E6B4910.rev.stofanet.dk] has joined #xen
10:14	<jerone>	tessier: you have to run your desktop in dom0 .. unless you are goin to use VNC or some other remote client to access it via a domU
10:14	<iamlost>	tessier: yes but
10:14	<iamlost>	tessier: i use Xvnc
10:14	<iamlost>	and i use fbvnc as the client
10:14	<iamlost>	fbvnc on dom0 :(
10:14	<tessier>	jerone: I know. I was hoping that had changed.
10:14	<iamlost>	tessier: there are patches
10:15	<iamlost>	but it stil goes thru vnc
10:15	<jerone>	iamlost: those patches are still being worked ... aliguori sits right behind me
10:15	<jerone>	iamlost: he wrote those patches
10:15	<iamlost>	jerone: :)
10:16	<tessier>	If vnc is required even with the patches I don't see the point
10:16	<iamlost>	jerone: any chance of being able to render the paravirtual framebuffer onto the real one?
10:16	<aliguori>	tessier, no, it's not
10:16	<iamlost>	by sharing the fb betwen guests.
10:16	<aliguori>	iamlost, yeah, that's what i do
10:16	<aliguori>	iamlost, no, that's silly :-)
10:16	<jerone>	aliguori: haha
10:17	<iamlost>	aliguori: is it?
10:17	<aliguori>	iamlost, domU gets it's own framebuffer memory
10:17	\|-\|	Basic_py [~Basic@gatekeeper.real-time.com] has joined #xen
10:17	\|-\|	Solist [~Solist@mvil2.bb.netvision.net.il] has joined #xen
10:17	<iamlost>	aliguori: what i'd like to be able to do is alt-f1 is dom0, alt-f2 is domU1, altf3- is domU2, etc
10:17	<aliguori>	iamlost, sends updates to dom0, dom0 blts the updated area of memory into an SDL buffer
10:17	\|-\|	Solist [~Solist@mvil2.bb.netvision.net.il] has quit [Quit: ]
10:17	<tessier>	aliguori: ah, ok. That makes sense then. It just seems like nothing should be done in dom0 except administer the hardware and guest OS's.
10:17	<tessier>	Ideally dom0 would never even talk to the Internet.
10:17	<aliguori>	iamlost, it's SDL, so you could setup SDL to use fbdev and do just that
10:17	<DV>	aliguori: hi
10:18	<aliguori>	DV: howdy
10:18	<iamlost>	aliguori: hmmm...ok. i guess that works.
10:18	<tessier>	Ability to run X on a domU and ability to overcommit memory are the only two things on my Xen wishlist.
10:18	\|-\|	Solist [~Solist@mvil2.bb.netvision.net.il] has joined #xen
10:18	<DV>	aliguori: would you mind spending 5mn on the xml-rpc status ?
10:18	<aliguori>	tessier, overcommit is hard in Xen :-(
10:18	\|-\|	mejlholm [~mejlholm@port79.ds1-abc.adsl.cybercity.dk] has joined #xen
10:18	<aliguori>	DV: sure
10:19	<tessier>	aliguori: So it seems.
10:19	<aliguori>	DV: which portion, the Xend side or the libvirt side?
10:19	<DV>	aliguori: both, I assume it's not yet in the xend side
10:19	<aliguori>	DV: on the Xend side, I have a patch that 100% passing regressions.. going to port up to tip today and hopefully submit
10:19	<aliguori>	hopefully it will be soon
10:20	<DV>	okay, then libvirt, I now depend on libxml2 already so
10:20	<aliguori>	DV: as i mentioned before, I have some libxml2 code. since the basic ops are the same, it should be easy to plumb libvirt to use it
10:20	<aliguori>	DV: the only thing I'm not sure about (and I have to look into), is how to make it use both
10:20	<aliguori>	the s-expression and the xml-rpc stuff
10:20	<iamlost>	aliguori: i would have prefered like a frontend and backend fb device
10:20	<iamlost>	e.g. actual low level virtualization of the fb
10:20	<DV>	you mean on the xend side ?
10:21	<DV>	I would test first a basic xml-rpc and then fallback to sexpr
10:21	<aliguori>	iamlost, why?
10:21	<DV>	for the libvirt side
10:21	<aliguori>	iamlost, you still have to display the backend somehow
10:21	<aliguori>	iamlost, no, i mean on the libvirt side
10:21	<iamlost>	aliguori: so i coudl get rid of SDL on my dom0 ;)
10:21	<aliguori>	er, s/iamlost/DV
10:22	<aliguori>	iamlost, just write a new client.. the code that displays to SDL is like < 100 lines
10:22	<DV>	okay overload :-)
10:22	<iamlost>	aliguori: ah, ok :)
10:22	<aliguori>	DV: heh
10:22	[~]	DV waits for a bit ...
10:22	[~]	aliguori is good at doing a bunch of things at once :-)
10:22	<jerone>	ls
10:22	<aliguori>	er
10:22	<aliguori>	is not
10:23	<aliguori>	DV: so, there's already a connect function for libvirt, i guess what we need to figure out is how to let the user say they want to connect to a S-Expression/HTTP version of Xend or an XML-RPC version
10:23	<aliguori>	assuming we want to support both
10:23	<aliguori>	we could just probe....
10:23	<aliguori>	to see what protocol it's using
10:24	<DV>	aliguori: I would not expose that at the API level
10:24	<DV>	aliguori: the library should autodetect
10:24	<aliguori>	DV: yeah, I agree. I really hadn't thought of that before to be honest with you
10:24	<DV>	aliguori: try an XML-RPC, if failing fallback to sxpr
10:24	<aliguori>	it's actually pretty easy.. just make an XML-RPC, if you get a 404, go to sxpr
10:25	<aliguori>	yeah
10:25	<aliguori>	okay :-)
10:25	<DV>	don't worry should be simple
10:25	\|-\|	Solist [~Solist@mvil2.bb.netvision.net.il] has quit [Quit: Leaving]
10:25	<DV>	in other news Create seems to work fine with sexpr
10:25	<DV>	but I wonder how you remap this to xmlrpc
10:26	<aliguori>	DV: right now, it's just the same RPC just over a different transport
10:26	<aliguori>	DV: so it should be minimal change (you still pass S-Expression configs for instance)
10:26	<DV>	you mean you send the SExpr ?
10:26	<DV>	ah ... less impressive than I though
10:26	\|-\|	maks [~max@baikonur.stro.at] has joined #xen
10:26	<johnlev>	don't suppose anybody's written up how xen interacts with udev system? especially the backend agent?
10:27	<DV>	that would have forced to define a Real API
10:27	<DV>	but it's probably still too early :-)
10:27	<aliguori>	johnlev, not that I know of..
10:28	<aliguori>	DV: yeah, a real api is the next step
10:28	<aliguori>	DV: remember, this was an incremental plan.. i don't want to go off and do an api in a vacuum either (like this portion has been)
10:28	<DV>	but Ian can Co. are they ready for this ? I feel some resistance :-)
10:29	<DV>	yeah, sure :-)
10:29	<aliguori>	as far as I know, their opinions haven't changed
10:29	<aliguori>	the only people complaining are the lanl guys
10:29	<DV>	i.e. "nothing is frozen" ?
10:30	<aliguori>	hehe
10:30	<aliguori>	xen doesn't have freezes :-)
10:30	<aliguori>	i dunno, we'll see. i don't know when the next release is going to be
10:30	<aliguori>	i have to head into the office, i'll be back in 5 minutes
10:30	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has quit [Quit: Ex-Chat]
10:30	<DV>	okay, no worry. I was wondering how I should plan to waste my time
10:42	\|-\|	yorkedork [~thom@adsl-221-0-120.rmo.bellsouth.net] has joined #xen
10:44	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has joined #xen
10:47	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has quit [Quit: ]
10:48	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has joined #xen
10:58	<DV>	aliguori: re
10:59	\|-\|	Bicster [~Bicster@ip195.planetfall.COM] has joined #xen
10:59	\|-\|	noidd [~red@66.193.15.70] has joined #xen
10:59	<Bicster>	hi noidd
11:00	<noidd>	hey Bicster
11:00	<aliguori>	DV: yes
11:00	<Bicster>	aliguori!
11:00	<aliguori>	howdy Bicster
11:00	<noidd>	I have a random question about Xen functionality - IE, is this functionality in Xen (if it is I havn't found it in the documentation).
11:01	<noidd>	I run all my VM's with read-only root filesystems. I do this by having the Xen hypervisor enforce the vbd to be read-only.
11:02	<noidd>	When I need to do upgrades or changes I want to be able to set that device to read-write, then I can do a mount -o rw,remount
11:02	<noidd>	make my changes
11:02	<noidd>	mount -o ro, remount /
11:02	<noidd>	and then set permissions back to read-only
11:03	<noidd>	I see vbd create and vbd destroy, is there a way to vbd-modify?
11:03	<aliguori>	noidd, nope. that might be tricky to expose in linux anyway. i don't know of devices that alter between read-only/read-write
11:04	<Bicster>	noidd, is it possible to use lvm, and use lvchange to make the volume read-only instead of using Xen to do it?
11:04	<Bicster>	I don't know if lvm can make that change on the fly, but it's worth trying
11:05	<noidd>	I don't know, but i'll test it!
11:05	<Bicster>	seems not to work here
11:06	<Bicster>	it let me change it to read only and then I proceeded to write to it
11:06	<noidd>	lemme run a test, 10 secs
11:09	<aliguori>	Bicster, if that did work, i fear all sorts of bad things would happen as the guest would do writes() that would succeed but that would never be committed to disk
11:09	<Bicster>	aliguori, yeah, it could be ugly :)
11:09	<aliguori>	and if anything did a write and the a read assuming the write worked...
11:09	<aliguori>	i think things would start panic()'ing
11:10	<noidd>	Well, I wouldn't switch the device permissions until after I remounted it read-only
11:10	<Bicster>	but if the fs's are mounted readonly it wouldn't matter, since presumably only a cracker would be doing writes at that point, so a panic might be desired
11:10	<noidd>	but other users may be less, strict :-)
11:10	\|-\|	gerrit [~gerrit@129.33.1.37] has joined #xen
11:10	<noidd>	oh, I see what you're sayinmg
11:11	<Bicster>	aliguori, are we back to ARCH=xen with hg xen-unstable tip?
11:11	<Bicster>	or did I fsck up my patching
11:12	<Bicster>	linux-2.6.16-rc4-xen-765b0657264d
11:12	<Bicster>	I get no XEN options in kernel config
11:12	<Bicster>	unless I do ARCH=xen
11:13	<Bicster>	arrgh
11:13	<Bicster>	this blows
11:13	<Bicster>	there was much rejoicing when ARCH=xen went away
11:15	<noidd>	You're right, xen keeps writing to that device. Interesting.
11:17	\|-\|	pvanhoof [~pvanhoof@mailhost.newtec.be] has quit [Remote host closed the connection]
11:20	[~]	Bicster pokes aliguori
11:33	\|-\|	yorkedork [~thom@adsl-221-0-120.rmo.bellsouth.net] has quit [Quit: Leaving]
11:34	\|-\|	pvanhoof [~pvanhoof@mailhost.newtec.be] has joined #xen
11:38	\|-\|	pvanhoof [~pvanhoof@mailhost.newtec.be] has quit [Remote host closed the connection]
11:41	<Bicster>	can anyone hint on the future direction of Xen kernels? ARCH=xen or not?
11:54	<ivan_>	anyone know much about events/interrupts in xen?
11:54	<ivan_>	w
12:02	\|-\|	Bicster [~Bicster@ip195.planetfall.COM] has quit [Remote host closed the connection]
12:07	\|-\|	bernarde [~bernarde@32.104.18.240] has joined #xen
12:52	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has quit [Quit: Ex-Chat]
12:59	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has joined #xen
13:09	\|-\|	noidd [~red@66.193.15.70] has quit [Quit: Leaving]
13:14	\|-\|	pvanhoof [~pvanhoof@d54C0FBBD.access.telenet.be] has joined #xen
13:15	\|-\|	ivan [~ikelly@86.41.193.236] has joined #xen
13:18	\|-\|	ivan_ [~ikelly@83-71-72-159.b-ras1.lmk.limerick.eircom.net] has quit [Ping timeout: 480 seconds]
13:38	\|-\|	cableman [~cableman@3E6B4910.rev.stofanet.dk] has quit [Remote host closed the connection]
13:52	<DV>	mdday: I could not find it from reading the patch, can you confirm that the sysfs version files are readable by anybody on Dom 0 ?
13:53	<aliguori>	DV: you can set the perms on sysfs files
13:53	<mdday>	depends upon the mode that you give the files when yo create them.
13:53	<aliguori>	usually it's 644
13:54	<DV>	okay, I was wondering if it was subsystem specific :-)
13:54	<mdday>	btw, was patch 2/3 linewrapped? I swear it wasnt but gregkh says it is
13:54	<DV>	aliguori: so it should usually work for me :-)
13:54	<aliguori>	mdday, they had hard tabs btw.. you should probably use soft tabs for the next patch
13:54	[~]	DV currently does an hypercall to extract the version but it's limited to root :-\
13:55	<aliguori>	DV: yeah, i was thinking about a non-root domain socket xml-rpc interface to solve this problem (like xenstore has)
13:55	<aliguori>	but i haven't stopped to think about what info should be readable by non-root useres
13:55	<mdday>	I use __ATTR_RO which sets it to 0444, meaning you should be able to read
13:55	<aliguori>	perhaps just the domain info stuff
13:55	<mdday>	anyone should be able to read
13:56	<DV>	aliguori: well I want to be able to provide the version as user as much as possible
13:56	<aliguori>	mdday, yours was the first code i've ever seen to use the __label__ extension :-)
13:56	<mdday>	aliguori: but 2/3, it isn't linewrapped, right?
13:56	<DV>	laptop:~/libvirt/src -> ./virsh version
13:56	<DV>	Compiled against library: libvir 0.0.4
13:56	<DV>	Using library: libvir 0.0.4
13:56	<DV>	Using API: Xen 3.0.0
13:56	<DV>	cannot extract running Xen hypervisor version
13:57	<mdday>	aliguori: cool, huh :) I like that extension, makes a lot of sense. but I'm going to take it out cause gregkh doesn't know what it is.
13:57	<aliguori>	mdday, 2/3? not for me
13:57	<mdday>	yeah, wtf?
13:58	<DV>	mdday: I saw __ATTR_RO but I wasn't sure
13:58	<aliguori>	mdday, 3/3 is whitespace munged thoug
13:58	<ivan>	can you use the "int" assembly instruction in a xen kernel?
13:58	<mdday>	aliguori: thanks, mutt for me from now on
13:59	<DV>	mutt rules :-)
13:59	<aliguori>	i take that back, looks like thunderbird has some bugs
13:59	<mdday>	DV: right, when I first submitted I initialized the structure manually and 0444 was there in plain text. But we are supposed to use the macro and it hides the mode, which is 0444.
14:00	\|-\|	bernarde [~bernarde@32.104.18.240] has quit [Read error: Connection reset by peer]
14:00	<mdday>	aliguori: the patch should be good, right. thunderbird modifies it when displaying but it you look at the source text it is good.
14:00	<aliguori>	mdday, yeah, i saved it and opened it in emacs...
14:01	<mdday>	exactly, it's good. wtf???
14:03	<mdday>	DV: you got any ideas on what other stuff you want out of sysfs for xen?
14:13	<DV>	well, so far I keep libvirt interfaces minimals
14:13	<DV>	and try to not expose too much details
14:14	<DV>	so ATM I don't have a good idea
14:14	<DV>	I use direct hypercall for domain runtime information when run as root
14:15	<DV>	(to avoid the round-trip though http/python/http :-) if those where exposed
14:15	<DV>	in dom0 at the fs level that could lower the cost
14:16	<DV>	but I'm not sure it's really useful ATM
14:17	<DV>	aliguori: speaking of "what info should be readable by non-root user", I'm more wondering about
14:17	<DV>	aliguori: what RPC to xend should be allowed by non-root users
14:17	<DV>	s/by/to/
14:17	<aliguori>	DV, yes, this is whta I was asking before...
14:18	<aliguori>	DV: i think, at least, the domain info querying functions should be available to non-root users
14:18	<DV>	oh, maybe I misunderstood
14:18	<DV>	well currently creating domains though httpd works very well for users :-)
14:18	<aliguori>	sorry :-) it's trivial to add a second XML-RPC service on a different unix socket that has global r/w permissions that only exposes a subset of functionality
14:19	<DV>	yes trivial but not done ATM :) the problem is what to filter out
14:19	<aliguori>	DV: well, don't you always want domain creation to be done by root?
14:20	<DV>	well, when I run my regression tests I prefer to not run as root
14:20	<DV>	but it's anecdotical, not a serious case
14:20	<iamlost>	(15:16:02) neo10: Problem was with switch and arp cache :-/ .. fuck planet?s switch ... very stupid mistake
14:20	<iamlost>	(15:17:45) neo10: It?s funny now :-)
14:21	<DV>	aliguori: if the notion of uid could be associated to a running domain that might be interesting from a permission POV
14:22	<DV>	"everything is a file" isn't a bad model
14:22	<aliguori>	DV, yeah, was thinking of suggesting PAM authentication for the XML-RPC service but I'm not sure how to approach that without adding tons of mandatory access control crap on top of it
14:22	\|-\|	zwane [~zwane@S0106001217d33fa6.vc.shawcable.net] has joined #xen
14:22	<aliguori>	mm, xen fuse management interface :-)
14:22	<DV>	yeah, and I'm pretty sure SELinux is gonna show up too at some point
14:22	<aliguori>	yeah...
14:23	<aliguori>	the question is what layer does access control
14:23	<aliguori>	libvirt or something higher
14:23	<DV>	we are doing the same mistake about virtualization than for network
14:23	<aliguori>	?
14:23	<DV>	not thinking of the security/users aspects before integrating it in the OS
14:23	<aliguori>	oh
14:23	<aliguori>	:-)
14:24	<DV>	the Linux hackers are no better than the 90' Unices hacks, that's comforting :-)
14:25	<aliguori>	hehe
14:25	<DV>	if a guy can use qemu and nearly bring down the server wasting CPU and IO
14:25	<DV>	why not allow him to run his OS in Xen, costing cheaper overall
14:26	<aliguori>	well...
14:26	<DV>	what we need is good control, but I wouldn't block the possibility by default
14:27	<aliguori>	yeah, I think this needs some thought
14:28	<DV>	yup ... and the new XML-RPC is probably a good place to plug the policy in
14:29	<DV>	though adding one more piece of the infrastructure in xend is not a good idea
14:29	<DV>	yeah more thinking needed
14:30	<aliguori>	DV, more things in Xend just seems bad to me
14:31	<DV>	policy should certainly be stored in xenstore
14:31	<aliguori>	nooo
14:31	<DV>	:-)
14:31	<DV>	lol
14:31	<aliguori>	I'd like to see xenstore simplified too :-)
14:32	<DV>	more control yes, simplification ...
14:33	<DV>	you can't just fetch a source file at runtime for every rpc call
14:33	<DV>	so policy must be backed up in memory
14:34	<DV>	you can PAM, or xenstore or xend
14:36	\|-\|	tab_ [~tab@darwin.snarc.org] has joined #xen
14:36	\|-\|	tab [~tab@darwin.snarc.org] has quit [Read error: Connection reset by peer]
15:17	<Shaun>	Whats the diffrence between xenU and xen0 really, i noticed that the docs talk about not even using those names anymre and just using -xen or whatever.. i'm just wondering if i do a say make linux-2.6-custom1-build if it will know if i'm building a xenU or xen0
15:19	<iamlost>	xenU is a kernelcopmile donly for domU
15:19	<iamlost>	xen0 is a kernel compiled only for dom0
15:19	<Shaun>	right...
15:19	<iamlost>	xen is a mutlidom kernel - one that works in eithe rmode
15:19	<Shaun>	just wondering what the diffrences are
15:20	<iamlost>	huh?
15:20	<Shaun>	Just wondering why dom0 couldnt boot off of a XenU kernel...
15:21	<iamlost>	because xenU would lack the required drivers
15:21	<iamlost>	usually xenU lacks all hardware drivers, so that doesnt work
15:21	<iamlost>	also it wouldnt have the backends that allow u to run other guests
15:21	<Shaun>	ok, so it's just drivers really, the patches are the same except the configs?
15:21	<iamlost>	and if u add support for that in - basically u've gone from xenU to multidom
15:21	<iamlost>	yes
15:22	<iamlost>	its all in the .config
15:22	<Shaun>	ok, thanks, thats what i was looking for.
15:22	<Shaun>	also in the Makefile, i see this... # You may use wildcards in the above e.g. KERNELS=2.6
15:23	<Shaun>	what would be the point of using KERNELS=2.6
15:23	<iamlost>	when u have multiple kernel sources
15:23	<Shaun>	oh ok
15:23	<iamlost>	e.g. one might be for a stable xen, one might have bleeding edge patches
15:23	<iamlost>	might have more than one domU kernel
15:23	<iamlost>	etc
15:24	<Shaun>	ok, is that only for multiple versions like 2.6.12.6 and 2.6.14.xx, etc
15:24	<Shaun>	or could i do say... 2.6.12.6-custom1 2.6.12.6-custom2
15:25	<Shaun>	i build a few diffrent guest kernels usually, ones a stable-basic, latest-basic, stable-advanced, latest-advanced... (just my naming)...
15:34	\|-\|	pvanhoof [~pvanhoof@d54C0FBBD.access.telenet.be] has quit [Read error: Connection reset by peer]
15:34	\|-\|	pvanhoof [~pvanhoof@d54C0FBBD.access.telenet.be] has joined #xen
15:35	<johnny43>	guys, i have a question again.. how do i boot a guest from CD image and then continue with a network install?
15:36	<johnny43>	the docs say this would work if i'm using image files for the virtual disks, however i am using physical devices directly
15:37	<johnny43>	i did cdrom='isomnage' and boot='d' , but it complains about not being able to open an initial console
15:38	<Shaun>	are you running vmx?
15:39	<johnny43>	no
15:39	<Shaun>	i could be wrong but i beleive thats only supported in vmx...
15:39	\|-\|	zwane [~zwane@S0106001217d33fa6.vc.shawcable.net] has quit [Remote host closed the connection]
15:40	<johnny43>	hmm.. ok
15:40	<johnny43>	then what options are there for installing a guest from scratch?
15:40	<Shaun>	cdrom params i show only in the vmx section of the docs
15:40	<johnny43>	(other the rebooting the whole physical machine)
15:41	<Shaun>	it would be nice of regular guests could be told to use a cdrom image though..
15:41	<iamlost>	johnny43: its possible to do it from cdrom
15:41	<iamlost>	but u need to roll ur own domU kernel for the installer first
15:42	<iamlost>	http://lists.xensource.com/archives/html/xen-users/2005-10/msg00406.html
15:42	<Shaun>	iamlost: what needs to be done?
15:42	<iamlost>	basically that tell su wat to do
15:42	<Shaun>	oh
15:43	<Shaun>	thats not using a cdrom... thats just using a initrd image from a network installer..
15:44	<johnny43>	yes, that's even better than cdrom/image
15:44	<johnny43>	it looks like what's needed is to replace the kernel in the installer with the one used for domU
15:44	<Shaun>	cdrom=/path/to/cdimage is simpler in my opinion than having to grab a initrd image, modify it and then load it..
15:44	<Shaun>	but it'll do the trick..
15:45	<johnny43>	iamlost: thanks, i'll give it a try
15:56	<ivan>	if i do a software interrupt will that be caught by my trap table, or my hypervisor_callback
15:56	<ivan>	?
16:14	\|-\|	anthony_ [~anthony@pixpat.austin.ibm.com] has joined #xen
16:14	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has quit [Read error: Connection reset by peer]
16:14	\|-\|	anthony_ changed nick to aliguori
16:15	\|-\|	athomas [~athomas@hardpress.demon.co.uk] has joined #xen
16:20	\|-\|	anthony_ [~anthony@pixpat.austin.ibm.com] has joined #xen
16:20	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has quit [Read error: Connection reset by peer]
16:20	\|-\|	anthony_ changed nick to aliguori
16:23	\|-\|	mdday [~mdday@cpe-024-163-120-222.nc.res.rr.com] has quit [Quit: mdday]
16:25	\|-\|	mdday [~mdday@cpe-024-163-120-222.nc.res.rr.com] has joined #xen
16:35	\|-\|	mejlholm [~mejlholm@port79.ds1-abc.adsl.cybercity.dk] has quit [Quit: Client exiting]
16:35	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has quit [Ping timeout: 480 seconds]
16:38	\|-\|	athomas [~athomas@hardpress.demon.co.uk] has quit [Quit: Leaving]
16:46	<iamlost>	aliguori: which of these email addresses are current?
16:46	<iamlost>	oh
16:46	<iamlost>	he left
16:54	<Shaun>	How can i tell the host to only use cpu0?
16:54	<Shaun>	right now it's using all 4...
16:55	\|-\|	ivan [~ikelly@86.41.193.236] has quit [Quit: leaving]
16:55	<iamlost>	Shaun: hmm
16:57	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has joined #xen
16:57	<iamlost>	aliguori: which of these email addresses are current?
16:57	<aliguori>	?
17:00	<aliguori>	why does xm-test automatically update a bunch of .py files that are already checked into the tree?
17:00	<aliguori>	it mucks with hg ci :-/
17:00	<iamlost>	:)
17:01	\|-\|	dhendrix [~dhendrix@c-69-247-65-201.hsd1.nm.comcast.net] has quit [Quit: Leaving]
17:11	\|-\|	stekloff [~stekloff@pool-71-245-110-186.ptldor.fios.verizon.net] has quit [Quit: Leaving]
17:12	\|-\|	bestorga [~bestorga@129.33.1.37] has joined #xen
17:13	\|-\|	dansmith [~dan@129.33.1.37] has quit [Quit: Leaving]
17:22	\|-\|	mdday [~mdday@cpe-024-163-120-222.nc.res.rr.com] has quit [Quit: mdday]
17:29	\|-\|	aliguori [~anthony@pixpat.austin.ibm.com] has quit [Quit: Ex-Chat]
17:30	<Shaun>	damit..
17:30	<Shaun>	i keep missing aligouri :)
17:33	<iamlost>	well
17:33	<iamlost>	im still here ;)
17:34	<Shaun>	haha, ya but your awnser was hmm :)
17:35	<johnny43>	iamlost, maybe you can help me again..
17:35	<johnny43>	;)
17:36	<iamlost>	johnny43: oh?
17:37	<iamlost>	Shaun: oh oops
17:37	<johnny43>	i booted a domU with a xenU kernel and an uncompressed initrd from centos4, like this
17:37	<johnny43>	disk = [ 'file:/boot/initrd-centos42pxe,hda1,w', 'phy:/dev/xen/dxbaseline,sda1,w' ]
17:37	<johnny43>	root = /dev/hda1
17:37	<johnny43>	the installer starts, but doesn't see any network or disks :(
17:38	<johnny43>	(/dev/xen/dxbaseline is a LV)
17:38	<iamlost>	where'd u get the kernel from?
17:40	<johnny43>	the xen3 domU kernel, recompiled with a few extra options (for iptables)
17:40	<johnny43>	it works fine on the same machine and against the dom0 under it..
17:40	\|-\|	mull [~arg@user-12l2gns.cable.mindspring.com] has quit [Quit: Download Gaim: http://gaim.sourceforge.net/]
17:41	<iamlost>	johnny43: where is the real cdrom?
17:41	<iamlost>	i see the initrd and the disk u wanna install to
17:41	<iamlost>	but not the cdrom itself
17:41	<johnny43>	i dumped the cdrom idea
17:41	[~]	iamlost is guessing that the installer cant get the tool it needs from the cdrom iso to see the disks or the network
17:41	<johnny43>	i want to do it over ftp
17:42	<johnny43>	aha, i see..
17:42	<johnny43>	on the other hand, it's a PXE image i've used - that one shouldn't normaly require a cdrom..
17:43	<Shaun>	i use the pxe image all the time, not for domU though, for pxe :)
17:43	\|-\|	Basic_py [~Basic@gatekeeper.real-time.com] has quit [Quit: Leaving]
17:46	\|-\|	dansmith [~dan@c-24-21-32-166.hsd1.or.comcast.net] has joined #xen
17:47	\|-\|	zwane [~zwane@S0106001217d33fa6.vc.shawcable.net] has joined #xen
17:52	\|-\|	jerone [~jerone@pixpat.austin.ibm.com] has quit [Quit: Leaving]
17:56	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has joined #xen
17:57	<iamlost>	Shaun: well aliguori is here now
17:57	<aliguori>	uh oh
17:57	[~]	aliguori hides under his desk
17:57	<iamlost>	lol
17:57	<aliguori>	:-)
17:59	<iamlost>	(17:55:04) Shaun: how can i have dom0 use only cpu0 rather than all 4 cpu's?
17:59	\|-\|	rusty [~rusty@bh02i525f01.au.ibm.com] has joined #xen
18:00	<aliguori>	xm cpu-pin
18:00	<aliguori>	or something similiar to that
18:00	<aliguori>	but performance will suffer
18:01	<iamlost>	why??
18:04	\|-\|	dhendrix [~dhendrix@c-69-247-65-201.hsd1.nm.comcast.net] has joined #xen
18:04	\|-\|	dhendrix [~dhendrix@c-69-247-65-201.hsd1.nm.comcast.net] has quit [Read error: Connection reset by peer]
18:04	\|-\|	pvanhoof [~pvanhoof@d54C0FBBD.access.telenet.be] has quit [Read error: Connection reset by peer]
18:04	\|-\|	sc [~dhendrix@c-69-247-65-201.hsd1.nm.comcast.net] has joined #xen
18:04	\|-\|	pvanhoof [~pvanhoof@d54C0FBBD.access.telenet.be] has joined #xen
18:05	\|-\|	rharper [~rharper@pixpat.austin.ibm.com] has quit [Quit: Leaving]
18:05	\|-\|	muli [~muli@87.69.40.180.cable.012.net.il] has quit [Ping timeout: 480 seconds]
18:10	<aliguori>	iamlost, dom0 handles IO, having dom0 on all processors makes it more likely that you won't get deadlocked (all processors are locked waiting for dom0 to complete io on a single processor)
18:10	<aliguori>	testing an x driver, bbiab
18:10	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has quit [Quit: Ex-Chat]
18:25	\|-\|	probonic [~probonic@host-84-9-223-181.bulldogdsl.com] has joined #xen
18:26	<probonic>	Hi. Does anyone know if its possible to determine how an "xm console" session was ended (i.e. killed with ^[ or by a system halt/reboot) ?
18:26	<probonic>	both circumstances yield a "0" return code
18:26	<tonfa>	maybe xm log
18:26	<tonfa>	or xm dmesg
18:50	\|-\|	dansmith [~dan@c-24-21-32-166.hsd1.or.comcast.net] has quit [Quit: Leaving]
18:51	\|-\|	Basic_py [~Basic@warden.real-time.com] has joined #xen
19:05	\|-\|	mdday [~mdday@cpe-024-163-120-222.nc.res.rr.com] has joined #xen
19:09	\|-\|	pvanhoof [~pvanhoof@d54C0FBBD.access.telenet.be] has quit [Remote host closed the connection]
19:12	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has joined #xen
19:32	\|-\|	muli [~muli@87.69.40.180.cable.012.net.il] has joined #xen
20:08	\|-\|	bestorga [~bestorga@129.33.1.37] has quit [Quit: Leaving]
20:09	<teferi>	hi, guys, I had the intractable problem yesterday
20:09	<teferi>	well, I switched from x86-64 back to just i386, and everything is peachy now
20:10	\|-\|	mdday [~mdday@cpe-024-163-120-222.nc.res.rr.com] has quit [Quit: mdday]
20:10	\|-\|	gerrit [~gerrit@129.33.1.37] has quit [Ping timeout: 480 seconds]
20:10	<Shaun>	aliguori: i though it was a performance improvment to have dom0 on 1 cpu and the guest use the rest of them?
20:13	<Shaun>	awsome.....
20:13	<Shaun>	[root@localhost user1][root@localhost user1]# ps auxf\|grep xend\|wc -l
20:13	<Shaun>	169
20:17	\|-\|	Shaun [~ndci@ip68-5-63-223.oc.oc.cox.net] has quit [Quit: ]
20:17	\|-\|	Shaun [~ndci@ip68-5-63-223.oc.oc.cox.net] has joined #xen
20:26	\|-\|	lilo [~lilo@lilo.usercloak.oftc.net] has quit [Quit: leaving]
20:26	\|-\|	lilo [~lilo@lilo.usercloak.oftc.net] has joined #xen
20:52	\|-\|	Basic_py [~Basic@warden.real-time.com] has quit [Quit: Leaving]
22:07	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has quit [Quit: Ex-Chat]
22:16	\|-\|	johnlev [~johnlev@nwkea-socks-1.sun.com] has quit [Ping timeout: 480 seconds]
22:54	[~]	riel wonders why Xen adds "struct list_head ballooned;" to the page struct
22:54	<riel>	surely the balloon driver could just use the lru list_head ?
22:59	\|-\|	VS_ChanLog [~stats@ns.theshore.net] has left #xen [Rotating Logs]
22:59	\|-\|	VS_ChanLog [~stats@ns.theshore.net] has joined #xen
23:03	\|-\|	aliguori [~anthony@cpe-70-116-13-229.austin.res.rr.com] has joined #xen
23:52	\|-\|	Basic_py [~Basic@warden.real-time.com] has joined #xen
---	Log	closed Wed Feb 22 00:00:22 2006