ESPAÑOL
Presentación

Programa

Desarrollo

ENGLISH

Presentation

Programa

Desarrollo

riel	for questions and answers, you can go to #qc
riel	remember that during the lecture, you can always ask questions in #qc
* leonardop is back (gone 00:33:36)
riel	if you have any questions, you can ask them at any time during or after the lecture, in the #qc channel
FloodeR	Question to riel: Don't you think that it be a little dangerous to "play" with stable kernel series?
riel	FloodeR: ok, I'll answer that one ;)
FloodeR	Thanks :)
JALH	like changing the vm...
hensema	riel: just for reference, what kernel was current in juli 2001? 2.4.9?
HoraPe	(to be answered in #qc, what's UML?)
riel	could you cut'n'paste on #qc what the last line was you got from me ?
hensema	User Mode Linux
sarnold	uml == user mode linux .. run linux on top of linus as a user process
riel	bah, server problems it seems ;(
HoraPe	riel, "the extra..."
Ap	<riel> the extra stability and new features meant that many kernel developers started switching to Alan's kernel and making patches for Alan's kernel
dabeej	the extra stability and new features meant that many kernel developers started switching to Alan's kernel and making patches for Alan's kernel
riel	HoraPe,Ap: ok, thanks
dabeej	hey riel, continue
HoraPe	ok, i didn't get the acronym :-)
anders	HoraPe: http://user-mode-linux.sourceforge.net/
Bugblue	riel:we miss a much....
onki	Bugblue: don't worry, I'm running a log
* Bugblue is doing the same
* Bugblue just did /ignore *!*@* (joins,modes,parts,quits)
onki	riel, we have seen rumours about two vm's (on the kernel list it was suggested a.f.a.i.k) would that be a good idea in your opinion?
Bugblue	that should look a bit better
Bugblue	*it was a hint*
onki	maybe a little advanced (my question)
riel	onki: I'll try to answer that question ;)
onki	k, thnx
HoraPe	Bugblue, tnx
Bugblue	HoraPe: smodes is also a good idea
bernardo	hol
bernardo	hola
bernardo	alguién sabe el nombre de algún software serio para componer musica
hensema	bernardo: english please
viZard	bernardo, canal de preguntas para #umeet
setepo	riel: what is happening with XFS? XFS will be added in 2.5?
HoraPe	hensema, not needed to write in english...
HoraPe	riel reads (more or less) spanish and portuguese
hensema	Yeah, but not all people on this channel do and questions are copied to #linux by riel...
Bugblue	in Xchat users can click on the conference button (the [c] right beside the inputline where you type text) just choose #linux and click on this button. It will automatically ignore everything that is not normal text
riel	well, this is the first half of this lecture, if you have questions about what happened in 2.4 last year, please ask them in #qc
Ikarus	Bugblue: it will show mode and nick changes
viZard	he's gone, see ?
HoraPe	hensema, when riel answers you can understand the answer... it's something :-)
Bugblue	Ikarus: it will not show joins, parts and quits...
Ikarus	Bugblue: correct
dabeej	i have question!
dabeej	the vm you had in earlier releases of 2.4
t00R__	viZard =)
thobias_	riel: what do you think about most distributions start using Alan's kernel
hensema	riel: lots of people tend to be unhappy about the critical bugs in released kernels, like 2.4.11 and 2.4.15. Do you think anything can be done in order to prevent such bad releases?
thobias_	?
dabeej	which release would you call your best
wol	riel: I've followed the 2.4 development and I'm nervous that critical bugfixes seem never to get integrated.  Do you think this will change now Marcelo is in charge?
Army	Bugblue : when 'll they be online?
peter111	buenas
cox	riel: the bugs you are talking about do affect "normal" servers?
viZard	peter111, shhh
dabeej	and would you continue working on your vm in future
manaha	riel what are the major differences between yours and marcelo VM (i've had no time to look at marcelo VM :(  )
* fluxie is away. idle for 30m (time: 01:35pm)
Army	manaha : its Andrea VM not Marcelo's
manaha	excuses
velco	Army: it's Marcelo's now ;)
manaha	marcelo...
t00R__	riel: has the vm from AA been tested publicly ? (mjesus asks this)
dabeej	??
viZard	jeje
Army	velco : owkee velco scores one point :)
Folken	that was like invoking a spirit...
dabeej	marcelo isnt around riel
riel	seems like it ;(
Folken	"if you're here knock 3 times"
dabeej	=(
dabeej	riel: im gonna go sit in "the channel"
wol	riel: (for instance: the periodic complaints about not being able to tell how Andrea's VM works, and recently it was said that this was because doc patches are discarded.)
dabeej	i got someone logging for me
dabeej	good job so far
viZard	riel: somebody asks me, what's a VM?
wol	viZard: virtual memory subsystem.
sarnold	virtual memory system .. it handles running more programs at once than can fit in memory
viZard	thx
peter111	riel: yo quiero aprender a cambiarme la Ip ¿tienes algun programa?
peter111	traducelo ahi viZard
t00R__	peter111: ifconfig ;-)
Folken	riel , and what about the -rc proposed in LKML?
peter111	ifconfig ·· comando desconocido
fernand0	peter callese, por favor
Army	riel : how is the janitor-project doing?
viZard	peter111: ahora estamos en medio de una conferencia, dejalo para despues, si?
onki	riel, additional question to hensema's question, is there some sort of list kernal hackers are using to test a new release?
cox	a kernel QA team is being needed :)
Folken	riel , but that give us the other problem discussed, no one will test -rc's
* hensema will
* Army will also when he has time
Rawsock	i've been testing since 2.4.0 and got no prob
dmc	The main thing is to have -preX kernels that are guaranteed not to have new features before they become a release...
Rawsock	but can't test too well
velco	velco is now known as Lorette
wol	Might it make more sense (at least in 2.<odd> series) to give up on -pre and just keep incrementing the SUBLEVEL?
riel	wol: maybe
AdamK	A myriad of different release naming schemes has been discussed nearly to death on LKML
sarnold	there are only 255 sublevels to play with in odd series (at least, as currently implemented..) .. using -pre allows more than 255 kernelseasily :)
AdamK	Anyone whos interested, feel free to check an archive
Army	wol : you meen something like 2.4.17.1 and 2.4.17.2?
JALH	that's what zealos uses :>
HoraPe	it's a port of linux to linux or to posix?
wol	*shrug* I come from a universe where test releases are called "20010407 snapshot" and we spend _months_ testing each point release, so the kernel versioning system has never really worked for me.
velco	HoraPe: to linux
HoraPe	ie, can i run uml in a bsd (or nt) ?
JALH	yes
JALH	I believe it runs on bsd
velco	HoraPe: has linux specific ptrace calls
JALH	not sure about nt
JaL	it's a linux-kernel patch
wol	Army: no, just 2.4.17, 2.4.18, 2.4.19, ... keep bumping the smallest number until it works.
velco	HoraPe: it could be _ported_ though
JALH	HoraPe, talk to jdike on #kernelnewbies on irc.openprojects.net
Jae	Will UML be added to the linux source tree?
JALH	it was in the -ac ones
Bugblue	riel: what are the advantages instead of using 'chroot' or in HP-linux so called: 'compartments' ?
dre	this isn't like tue impletation of compaq on his tru64 that can run several o.s. at the same time ??
Bugblue	riel: and could't we do it already with vmware (and others like that?)
dre	(the uml )
onki	riel, if you would provide a 'kernel instance' with UML would the admin still have control or would the user have to much freedom?
Jae	riel: will UML be added to 2.5 ?
kroks	but you have ulimit for preventing that kind of starving resources
hensema	Jae: like he said: he cannot look into the future
lennert	riel: jeff dike said at LK2001 he would submit uml for 2.5 any day now
* hensema counts on UML being added, though ;-)
JALH	hey lennert!
* Bugblue bets hensema for a bbq-beer
Army	kroks: ulimit is nothing compared with prm
lennert	hey JALH
Army	kroks : prm gives you more control about what every one is allowed todo
cox	riel: what about the security provided by the NSA Linux distrib compared to UML?
FloodeR	riel: The process of generate uml virtual enviroments, can starve the machine, or while you don't use it they are stoped? Imagine a university machine
Army	FloodeR : prm is taking care of that
lennert	running >5 umls on a moderate box will noticably slow it down
kroks	Army: i dont know anything about prm, thank you for telling me about it
lennert	(because of the uml timer tick)
FloodeR	Um
Folken	lennert , moderate box being what?
lennert	Folken: pentium iii 800 or so
FloodeR	I'm thinking in at least 50 umls
Folken	lennert , not so bad...
lennert	we need a tickless kernel.. :)
Army	kroks: with prm the kernel can limit things, hopefully will it make 2.5
Bugblue	or
TJ	hey congratz for conference its very interesting
TJ	:OP
lennert	as it is, idle uml's take up too much resources
Sorvin	riel : meaning, Inter-Mezzo is just .. well.. Offline Files for linux .. ? :)
Bugblue	we need a IBM Z-series mainframe.... what can run about 40.000 concurent linux instances
Bugblue	without any problem except the money
velco	the overhead that uml instroduces is mostly due to the redirection of system calls, besides system calls slowdown all the rest works the same way, with the same speed.
Joselito1	who is -rc
MCArkan	riel: won't intermezzo takes too much network resources if it duplicates files ?
Army	Bugblue : money? just order it :)
* JALH hands riel cvs :-)
Army	Bugblue : just say you want to see the machine running before you buy :)
cdub	Joselito1, -rc means release candidate
Rawsock	um JALH
sarnold	joselito1: -rc == release candidate
HoraPe	what happens when both the laptop and the server have modified the same file?
Bugblue	Army: I did and they wanted to put a test machine here (for a month)
Rawsock	Intermezzo is a transparent filesystem layer i suppose
JALH	cvsfs
JALH	:)
Rawsock	lika a HURD translator i suppose
Army	Bugblue : when can I come for a visit? :)
lennert	HoraPe: intermezzo doesn't deal with that yet
Rawsock	( doh, said that word again )
HoraPe	what does it does when that happens? croaks?
lennert	Rawsock: yes, but does need some per-filesystem support code
lennert	intermezzo uses either an own algorithm, or the rsync protocol (librsync)
lennert	(for replication)
lennert	replcation changes are tracked by keeping a KML, kernel modification log
JALH	Leenooks S&M!
lennert	this is a sort-of journal of modifications, but it's not circularly rewritten.
lennert	it's more akin to db-style archive logging
JALH	it uses a luser-space lib?
Rawsock	um sweet
lennert	JALH: it consists of two components, presto and lento
lennert	JALH: presto is the kernel part, lento the userspace part
JALH	ah
lennert	JALH: lento takes care of forwarding changes
lennert	JALH: lento is, incidentally, written in perl
Rawsock	:) andante ma non troppo
JALH	aiee
* JALH takes a look :)
Ikarus	In PERL, sick
Sorvin	riel : will it replace PAM ?
onki	riel: could you explain the difference between iptables and LSM? is LSM arranging security on the user part?
mulix	sorvin, it's a userspace component
mulix	pam is userspace, lsm is kernel space
peter	me voy...
Sorvin	i know, but will it make PAM unneeded ?
peter	chao a todos por aqui!!!!!
mulix	not likely
mulix	because some things are better checked in userspace
mulix	and some are better checked in kernel space
Sorvin	i see.
Sorvin	thanks :)
mulix	sure, dont mention it :)
cdub	riel, sarnold: NSA (SELinux), standard POSIX.1e capabilities ;-)
Army	riel: do you know if anything from example openwall or lids are going to be included in 2.5?
Folken	SELinux bd included...
cdub	iptables, is just network, LSM, uses iptables for some of the network hooks
cdub	sorry, netfilter
onki	cdub: thnx
cox	did the NSA create the LSM?
onki	netfilter indeed
sarnold	<cox> did the NSA create the LSM?
Sorvin	mulix : does that mean that all the security checks that are being done when open() is called, will be simply a list of LSM modules to be invoked ?
cdub	NSA presented to 2.5 kernel summit.
rapid	riel : code on linux kernel is a mess... its planed to fix
cdub	but LSM is a combined effort of NSA and many others
Rawsock	hum .. Loading magiclantern.o ?
rapid	things around? move funcions etc..?
mulix	Sorvin, i'm not sure about the lsm architecture
mulix	if it's a series of modules, or a series of hooks or what
Sorvin	k
mulix	perhaphs some of the developers could tell you, or just check out the source :)
kroks	does LSM  means that when you make a program you dont have to care about security?
* hensema laughs
Folken	'use the source Luke!' (TM)
mulix	that's *always* the right answer, folken :)
onki	kroks, that would be cool :)
MCArkan	how reliable can the NSA members be ?
mulix	sarnold, than how is it implemented?
Folken	yeh :)
cdub	kroks, no, in fact, you may want to make your app _aware_ of the new security features in the kernel module
cdub	but, you can also use an LSM to confine code that you are unsure of.
sarnold	<mulix> sarnold, than how is it implemented?
mulix	sarnold, while we are at it, will lsm allow hijacking system calls?
Sorvin	sarnold : of course not. but im asking this if developers could "plug-in" security checks of their own easily when open() is called
cdub	or check http://lsm.bkbits.net ;-)
kroks	cdub: ok, but ive read that if you want to open a file in a secure way, you have to do more thinks that only call open(). will LSM implement that kind of security matters?
mulix	ok, sarnold, thanks very much.
sarnold	<cdub> or check http://lsm.bkbits.net ;-)
kiwnix	can LSM do kernel stack protections?
kiwnix	as solaris do
hensema	sarnold: can LSM be used to RAISE permissions, instead of denying access to certain resources?
cdub	heheh, hesema, yes, but not at the same granularity as denying.
sarnold	since we don't need the lecture to last forever, and it is supposed to be *riel's* lecture, I'll hang out in #lsm-dev on irc.openprojects.net to answer more questions :)
riel	sarnold: if you want to, maybe we could even schedule an LSM lecture somewhere in the next 2 weeks? ;0
dreim0n	sarnold:  we can make a second lecture on this net if you want
dreim0n	:)
zuez	speaking of filesystems, are you folks planing something like growfs for ext2 partitions?
setepo	riel: XFS will be added in 2.5?
HoraPe	riel, bsd people use something called softupdates, supposed to be a more rational way of ordering writes that has lot of the journaled fs without being so complex, will linux get some fs like that?
wol	related, there's the buffer cache and the page cache, what is the difference? it seems like the buffer cache is slowly going away?
zuez	yeah, i´d also like to point out that small difference :)
lennert	correction, andreas dilger wrote the ext2 online resize patch (for ext2resize, ext2resize.sf.net)
lennert	ted t'so is currently figuring out another way of doing online resizing without needing an offline prepare stage first
basilon	buenas
lennert	he wrote something about this in the LK2001 proceedings
velco	HoraPe: with softupdates you may lose blocks, so have to run fsck sometimes, OTOH jornalled filesystems suffer from seeking to the journal (if it's on the same plate)
zuez	or, you loose files that are cached.
lennert	there are still recovery situations that softupdates doesn't deal with, that's why newer solaris has a logufs
lennert	phase trees are awfully identical to 'shadow paging' as described f.e. in 'transaction processing' by Gray and Reuter (1970-something)
ninjalj	what happened with the patent claim against Phillips?
lennert	Gray basically concluded that shadow paging in system R was a failure
HoraPe	why that conclusion?
lennert	HoraPe: fragmentation
lennert	ninjalj: patent claim?  from netapp?
vituco	riel: Could you tell us about the Networking code in 2.5.x later? Will it change again?
lennert	Transaction Processing : Concepts and Techniques (Morgan Kaufmann Series in Data Management Systems), by Jim Gray, Andreas Reuter
dreim0n	hummm one aobut this phase if changes are done with a veryu little tiem in a file like the oracle.db  that is a large file this can mean that you are working with an out-of-phase db ?
lennert	http://www.amazon.com/exec/obidos/ASIN/1558601902/ref=pd_bxgy_text_1/107-3063278-5990918
lennert	<riel> the filesystem can simply let 10 fsync()s complete on the same phase change
lennert	you can do the same with journalling, it's called batch commit or group commit
lennert	bad for latency, but really good for performance
lennert	from what i've read about WAFL, it's just another implementation of shadow paging
viZard	riel: is or will be LIDS merged with 2.4 or 2.5 kernel ?
sarnold	viZard -- no; one of the reasons why Linus proposed LSM is to avoid questions like this. :) The LIDS people have said they plan on porting LIDS to use LSM in the future, when they have omre free time
cdub	viZard likely LIDS will be ported to LSM
MCArkan	riel: what's the difference with intermezzo ?
HoraPe	how is the hardware thing done?
cox	how you can share the same disks between multiple boxes?
dmc	fiberchannel.
dmc	or any SAN.
cox	aps
cox	thx
zuez	storage area network?
dmc	yep.
zuez	hm
dreim0n	we use fier channel in a little network of 30 compiuters it seem enough
dreim0n	the sotarge an hsg80 of sun with 19teras
FloodeR	riel: What's about stability of opengfs? For production machines?
riel	FloodeR: not quite there yet, but getting better fast
FloodeR	Ok, thanks
dreim0n	in what machines is ready to run ?? only  linux boxes ??
ninjalj	Flooder, both intermezzo and opengfs are marked experimental right now
zuez	how do you prevent us the Systems Administrators with Linux from tuning cahce sizes and stuff?
cox	lol
zuez	thing you don´t really need to worry about iwth fbsd.
peter11	buenas
zuez	yeah
zuez	indeed, riel, m.m is good with fbsd.
zuez	however
zuez	you can bot with 4Gb of RAM
zuez	it should cause a panic tho
zuez	19990604-CURRENT has a fix for it, afaik.
wol	you say this will happen outside the main kernel, but will it be intended to go back at some point, or will it remain a separate patch forever?
riel	I guess this lecture has gone on for too long already ... if you have questions you can answer them in #qc
onki	riel, this raises a lot of questions for me, be prepared to answer them within the next week :) thanx for the lecture, it was good
hensema	riel: thank you for the informative lecture.
dmc	riel: thanks.  Great talk.
MCArkan	riel: thanks
mulix	thanks, riel, it was very interesting

Mas información: