[04:59] <dilinger> fabbione: when do you think you'll get around to doing a new kernel for this box?
[04:59] <dilinger> i've got a machine hooked up to its serial console right now that i want to bring back downstairs
[04:59] <dilinger> which i'll do as soon as there's little chance of the sunfire needing any sort of kernel upgrade
[05:50] <fabbione> dilinger: i already copies the sources there.. i will start the build in a few minutes
[05:51] <fabbione> oh hold on
[05:51] <fabbione> i just got a mail from davem
[05:51] <dilinger> yea?
[05:51] <fabbione> It doesn't fix a hang, it fixes userspace FPU register corruption.
[05:51] <fabbione> The hang on SMP Ultra-III chips will not be fixed by this patch.
[05:51] <fabbione> That problem would require at least 2 or 3 weeks with physical
[05:51] <fabbione> access to an SMP Ultra-III system.
[05:51] <fabbione> I'm just assuming your Ultra-III box was SMP since I do know about
[05:51] <fabbione> a hang on those systems.  It's a bug related to how the chip processes
[05:51] <fabbione> cross-cpu calls and L2 cache transactions.  It wedges and the only
[05:51] <fabbione> way to unwedge a wedged cpu is to try to gain ownership of some L2
[05:52] <fabbione> cache lines that cpu has.  And for this you have to guess, ie. try to
[05:52] <fabbione> figure out what data structures and pieces of the address space the
[05:52] <fabbione> hung cpu might have in it's L2 caches.
[05:52] <fabbione> I hope you can see why it's so difficult to code up a fix without
[05:52] <fabbione> physical access to a machine on which this can be triggered  :-) 
[05:52] <fabbione> If your Ultra-III is not SMP I'd be curious to learn about it if you
[05:52] <fabbione> can still trigger the hang.
[05:52] <dilinger> he's in california, right?
[05:52] <fabbione> SF afaik
[05:53] <dilinger> ask him if he plans to come to ny anytime soon :)
[05:54] <fabbione> dilinger: is that a sunfire280r?
[05:54] <dilinger> yep
[05:54] <dilinger> (or montreal)
[05:54] <dilinger> (for ubz)
[05:54] <fabbione> ok hold on.. i am talking with him on another chat
[05:55] <fabbione> console isn't enough.. he needs access for real..
[05:59] <dilinger> right
[05:59] <dilinger> that's why i'm asking if he's going to be anywhere near me anytime soon
[05:59] <dilinger> i'll give him the machine
[06:01] <dilinger> is he on irc?
[06:02] <fabbione> yeah.. but not on feenode ;)
[06:02] <fabbione> (pasting here and there)
[06:02] <fabbione> davem he can't fedex it to SF? :)
[06:02] <fabbione> davem I'd pay for shipping, I could do that
[06:04] <dilinger> ugh
[06:04] <dilinger> i suppose i could, but i have no idea where the nearest place i could fedex from is
[06:05] <dilinger> he realizes it's 80lbs, right? :)
[06:06] <fabbione> i think he knows.. i did send an email to both of you
[06:06] <dilinger> ok
[06:07] <fabbione> so you can discuss it directly
[06:07] <fabbione> (on his request)
[06:07] <fabbione> pointless for me to bounce info around
[06:07] <dilinger> appropriate .sig
[06:10] <dilinger> fabbione: i would still like to see it used as a buildd or something, i dunno what davem plans to do w/ it in the long run.  but, the original owner of the machine would probably be pleased w/ davem owning it
[06:10] <fabbione> dilinger:  i think you can still agree with davem for him to have it while he fixes the bug
[06:10] <fabbione> and later send it to a hosting facility
[06:10] <fabbione> perhaps he knows even where to host it
[06:13] <dilinger> lame, fedex.com requires an account to get rates
[06:23] <fabbione> dilinger: about ubuntu buildd.. we might get official buildd's at the datacenter
[06:23] <fabbione> if we don't, we will find hosting for it
[06:25] <dilinger> fabbione: *shrug*.  well, ubuntu buildd or whatever; mainly i want to see it used for something that actually benefits the community
[06:26] <fabbione> yup.. i understand that :)
[06:26] <dilinger> if davem actively maintains it, i suppose that's a huge benefit.  but if it's just sitting in a room unused alongside 30 other sparcs that he owns.. :)
[06:27] <fabbione> nah he is already dedicating hw to ubuntu :) see his last email
[06:57] <fabbione> ehehe
[08:09] <fabbione> dilinger: did the machine change ip again+
[08:09] <fabbione> ?
[08:13] <dilinger> ugh
[08:14] <dilinger> i dunno what's up w/ this
[08:14] <dilinger> maybe give it a little time and it'll come back
[08:15] <fabbione> ok
[08:24] <dilinger> fabbione: if it's not back by morning, i'll switch its ip again
[08:25] <dilinger> 'night
[08:25] <fabbione> dilinger: can you please do it on the fly?
[08:25] <fabbione> because i did restart the buildd on it :)
[08:25] <dilinger> huh
[08:25] <dilinger> interesting
[08:25] <dilinger> try it now
[08:26] <fabbione> yup
[08:26] <fabbione> what happened?
[08:26] <dilinger> i'm.. not sure
[08:26] <fabbione> was it in OBP mode?
[08:26] <dilinger> no, it's up and running
[08:26] <dilinger> i couldn't reach it from the outside world
[08:26] <dilinger> but i could reach it from inside the network
[08:27] <fabbione> ah
[08:27] <fabbione> so what did you do?
[08:27] <dilinger> once i netcat'd out to a machine on the internet, it was reachable again
[08:27] <fabbione> ahhh
[08:27] <fabbione> arpcache problems
[08:27] <dilinger> so it sounds like some arp or routing issue
[08:27] <fabbione> no big deal
[08:27] <dilinger> yea
[08:27] <fabbione> i can add a ping every N minutes
[08:27] <fabbione> if you don't mind
[08:27] <dilinger> yea, i just started something like that
[08:27] <fabbione> perfect :)
[08:27] <dilinger> dilinger@sunfire:~$ while (sleep 180); do echo foo | nc mouth.voxel.net 22; done
[08:28] <fabbione> thanks a lot
[08:28] <dilinger> and now, i shall sleep
[08:28] <dilinger> np
[08:28] <fabbione> good night
[08:28] <dilinger> 'night
[12:04] <chmj> do we provide 2.6.13 ?
[12:06] <fabbione> no
[12:07] <chmj> thought so 
[12:27] <johnm> fabbione: do you guys provide a hardened version of the ubuntu sources? (grsec/pax, etc)
[12:28] <fabbione> nope
[12:28] <johnm> fabbione: any plans to?
[12:28] <fabbione> nope :)
[12:28] <johnm> how come?
[12:28] <fabbione> but you can ask for discussion
[12:28] <fabbione> because the patches are a pain to maintain in time
[12:28] <johnm> Wiki = best place?
[12:28] <fabbione> johnm: kernel-team@lists.ubuntu.com 
[12:29] <fabbione> for a fast discussion
[12:29] <johnm> :) they can be... but in honesty they're no worse than some of the stuff which is teasing mainsteam inclusion.
[12:29] <fabbione> wiki to formalize
[12:29] <johnm> that an open list?
[12:29] <johnm> tbh, it ideally means toolchain changes anyways, like the inclusion of pie and ssp./
[12:31] <fabbione> oh we did discuss that to death
[12:31] <fabbione> no you need to subscribe
[12:31] <johnm> pie/ssp? or the hardened stuff?
[12:31] <fabbione> we will not apply these patches in main
[12:31] <fabbione> yes
[12:31] <johnm> what came out of it?
[12:31] <fabbione> toolchain related..
[12:32] <fabbione> that hardened can be a derivate Ubuntu product
[12:32] <fabbione> but it won't hit mainline if it doesn't hit upstream first
[12:32] <johnm> ah I see.
[12:32] <fabbione> there are tons of reasons for that
[12:33] <fabbione> search the mailing list archive will give you a good bunch of stuff
[12:33] <fabbione> iirc it was discussed on ubuntu-devel mailing list
[12:33] <johnm> Yeah, we had similar discussions but pie/ssp is part of the base toolchain. (although we do give options to ignore using it).
[12:33] <johnm> cool, I'll take a look
[12:33] <fabbione> food time
[12:33] <fabbione> later
[12:33] <johnm> tbh, it isn't a requirement for the hardened-kernel stuff anyways. it's merely an extra attack vector which is being nailed down
[01:10] <makx> fabbione: klibc-sparc64-signals.diff causes build failures
[01:11] <makx> they are due too duplicate defines.
[01:11] <makx> gcc-3.3 doesn't cure that for me.
[01:11] <fabbione> makx: strange.. it was required here when building in 64 bit
[01:12] <makx> cpu             : TI UltraSparc IIi (Sabre)
[01:12] <fabbione> makx: i think it might be due to different l-k.h
[01:12] <fabbione> makx: cpu has nothing to do with it.. it's the build env that makes the difference
[01:13] <makx> fabbione: thought so did you apply an kernel patch too.
[01:13] <fabbione> makx: i don't do l-k-h, that's a question for jbailey 
[01:13] <makx> #if defined(__KERNEL__)
[01:14] <makx> #define __old_sigaction         old_sigaction
[01:14] <makx> #define __old_sigaction32       old_sigaction32
[01:14] <makx> #else
[01:14] <makx> #define __old_sigaction         sigaction
[01:14] <makx> #define __old_sigaction32       sigaction32
[01:14] <makx> #endif
[01:14] <fabbione> makx: i don't have the code right under my eyes
[01:14] <makx> fabbione: sure :)
[01:14] <fabbione> building in 64bit did require that patch for me
[01:14] <fabbione> 32bit no
[01:14] <fabbione> now
[01:15] <fabbione> the gcc thing is another story
[01:15] <makx> the gcc thing is to make it working, linking?
[01:15] <fabbione> exactly
[01:15] <makx> davem doesn't like gcc-4.0 irc
[01:15] <fabbione> it FTBFS with breezy gcc-4.0 due to some linking error to libgcc
[01:15] <fabbione> so we did switch to gcc-3.3 or 3.4
[01:16] <fabbione> in 32 bit
[01:16] <makx> does ubuntu has build logs?
[01:16] <fabbione> so that works fine as a combination
[01:16] <fabbione> makx: yes
[01:16] <fabbione>  http://bld-3.mmjgroup.com/buildLogs/k/klibc/ <-
[01:16] <makx> nice :)
[01:16] <fabbione> but not all of them are there
[01:16] <fabbione> some tests have been done offline
[02:20] <dilinger> fabbione: everything still ok?
[02:20] <fabbione> dilinger: yes! that machine is fast
[02:21] <fabbione> if we can't find a place where to host it, i will take it here..
[02:21] <fabbione> i have bw and ips
[02:21] <dilinger> haha
[02:21] <dilinger> alright, i'm off to work.  later.
[02:22] <fabbione> i am off to nap and school
[02:22] <fabbione> later