/srv/irclogs.ubuntu.com/2008/04/03/#ubuntu-kernel.txt

tjaaltonrtg: lrm is now uploaded, and I need some sleep :)00:10
infinitytjaalton: Danke.00:25
=== smb_away is now known as smb_tp
=== _stijn_ is now known as C
=== C is now known as _stijn_
=== asac_ is now known as asac
krautmoin09:10
munckfishHi, I hope this is on-topic enough - I'm updating kboot for ps3 (LP bug 146230). It comes bundled with it's own kernel source (2.6.24) ...11:29
ubotuLaunchpad bug 146230 in ps3-kboot "update ps3-kboot to 1.4.1" [Medium,In progress] https://launchpad.net/bugs/14623011:29
munckfishCould someone offer any advice on what I should do about "syncing" it with the Hardy kernel?11:30
munckfishE.g. should I take a snapshot of the hardy kernel and include that within the ps3-kboot package? Or is there a more elegant solution?11:30
munckfishAlso how important is it that the kboot kernel and ubuntu kernel are the same?11:31
munckfishthx11:31
phoenix_Hi there, I reported a bug #199934, which is causing hardy to panic upon reboot in the raid device driver - and the bug is still 'undecided' and I just wanted to know what I can do, to get some attention to get it fixed for -release11:52
ubotuLaunchpad bug 199934 in linux-meta "Kernel Panic, in gdth (RAID) driver on reboot" [Undecided,New] https://launchpad.net/bugs/19993411:52
munckfishcjwatson: have you got a moment to chat about ps3-kboot?12:02
cjwatsonmunckfish: BenC will really be better for that, once he's around12:07
cjwatson(he's US Eastern time)12:07
munckfishcjwatson: ok thx I'll look out for him later. thx12:11
Whoopiesoren: Hi, I'd have a question about the virtual kernel package. When I use the vmxnet driver in the VM, I have to select "wired network" each time I log in. Is the driver lacking some features which NetworkManager needs?12:54
BenCGood morning everyone13:30
amitk_Morning BenC 13:33
rtgamitk_: I see the wimax firmware disappeared.13:42
BenCbrb, testing my vzw evdo card13:48
BenCsweet, evdo card works14:14
munckfishBenC: Hi have you got a moment to talk about ps3-kboot?14:34
munckfishI'm working on LP bug 14623014:35
ubotuLaunchpad bug 146230 in ps3-kboot "update ps3-kboot to 1.4.1" [Medium,In progress] https://launchpad.net/bugs/14623014:35
BenCmunckfish: Sure14:37
munckfishGreat14:37
munckfishbasically I'm half way to getting it to compile14:37
munckfishas is, from upstream14:37
munckfishNow, it bundles kernel 2.6.2414:37
munckfishWhat should I do about "syncing" it with the Hardy kernel?14:38
munckfishShould I take a snapshot14:38
munckfishof the hardy kernel source14:38
munckfishand put that into the source tarball?14:38
munckfishOr is there a more elegant solution?14:38
BenCNo14:38
BenCjust use what they have, stock14:38
munckfishAha, ok14:38
BenCit's only used for first stage loader, it's not what the system uses to boot14:39
BenCthe end boot results will be out powerpc64-smp kernel14:39
BenC*our14:39
munckfishOk great that makes things simpler :)14:39
munckfishRight then, well I'll cont. working on it.14:39
munckfishI have all evening Friday put aside14:40
munckfishhopefully I'll have something to show next week14:40
munckfishthanks14:40
rtgBenC: so who do you have service with for your evdo card?14:47
BenCrtg: verizon14:49
rtgUSB?14:49
BenCrtg: it had to be activated via windows, but easy to setup after14:49
amitk_when are you guys getting WiMAX rollouts?14:49
BenCrtg: it's USB, but it's one of the dell provided cards that fits into the internal wwan slot14:49
rtgBenC: I think I have one of those.14:49
BenCpretty sure jose sent you one when he sent me one14:50
WhoopieBenC: Hi, do you perhaps know why NetworkManager doesn't activate "wired network" with the vmxnet driver? After each login, I have to enable it.15:10
mjg59It doesn't provide carrier detect?15:13
Whoopiemjg59: aha, ok. I read something about it on http://blogs.gnome.org/dcbw/2008/02/18/hey-vmware-set_netdev_dev-wants-a-word-with-you/15:22
Whoopieit seems that vmxnet supports carrier detection, but NetworkManager is not able to find it out.15:23
thoreauputicmjg59: thanks for getting the ball rolling on bug 201591 - looks like all is well with 2.6.24-14 :)18:36
ubotuLaunchpad bug 201591 in linux "atyfb regression - screen blank except for blinking cursor after fbcon vtswitch " [Medium,Fix released] https://launchpad.net/bugs/20159118:36
mjg59thoreauputic: Excellent18:37
thoreauputicfinally! Now I can pursue my eccentric little no-X project with a hardy version18:38
thoreauputicanyway, I thought I'd drop in and say thanks, because you took the trouble to start the process with your patch etc.18:39
mjg59Yeah, it was my fault to begin with18:40
thoreauputicreally?18:40
thoreauputicah well - all's well that ends well and other conventional sentiments...18:41
thoreauputic:)18:41
mjg59Hm18:44
mjg59acx111 just blew away my kernel18:44
=== doko_ is now known as doko
dhavalany clue where I will find Tim Gardener?20:36
rtgdhaval: yes?20:37
dhavalrtg, regarding BUG 18822620:37
ubotuLaunchpad bug 188226 in linux "Kernel should use CONFIG_FAIR_CGROUP_SCHED" [High,Fix released] https://launchpad.net/bugs/18822620:37
dhavalCONFIG_FAIR_CGROUP_SCHED is arch independent20:38
rtgdhaval: as far as I could tell, it is not supported on non x86/x86_64 arches (which I found a bit surprising)20:38
dhavalrtg, it is arch independent. i have resolved fair_cgroup_sched bugs on ia64 and ppc20:39
rtgI went through the menuconfig options manually, but no CGROUPS selection presented itself.20:39
dhavalrtg, you need to have CONFIG_CGROUP on for that20:39
rtgis the a prerequisite ?20:40
dhavalit depends on that20:40
dhavalwell, how else would you do cgroup scheduling :)20:40
rtghmm, lemme recheck.20:40
rtgI was in  a hurry at the time.20:40
dhavalhmm. you had us a bit worried20:40
dhavalrtg, i still suggest, keep cgroup_sched as default20:40
rtgshit. you know thats gonna cause another ABI bump.20:40
dhavalitwill save you a lot of bugs like that boinc one20:40
dhavalrtg, hmm?20:41
rtgits not a big deal, but does cause some process work for the archive admins.20:41
dhavalok. well i have been hollering for cgroup_sched for months now :)20:42
rtgI have to point out that you are a small (but vocal) minority. I have not been able to extract an informed opinion from the server group.20:43
dhavaltrust me, it would make life a lot easier. you would have old behavior and if you wanted group scheduling, you just have to mount it20:43
dhavalrtg, heh, i wrote part of that code :)20:43
rtgas I pointed out in the LP report, its the addition of user space configuration that is causing me problems. Its more of a policy restriction given the development stage we are at. Hardy is about to release.20:44
dhavalrtg, i have been hollering for months. i don't have any issues.20:45
dhavalbut you are going to get a number of "bugs" because people are not going to be aware of the changed behavior20:45
rtgI'll get CGROUPS in the server image, but I'm not willing to mess with the desktop until we start Intrepid next month.20:45
rtgThose who really care can run the server image on their desktop.20:45
dhavalrtg, people are *NOT* going to be aware of the new scheduler behavior20:46
dhavalthey are going to find that they get 50% cpu and daemons get 50%20:46
dhavaland nice can't help them there.20:46
dhavalyou will get bugs raised on that, and you will have to come up with workarounds.20:46
dhavalthe suggested solution is to shift to CGROUP_SCHED, which as I mention i have been suggesting for months now :)20:47
rtgin all honesty, 99% of desktop users don't care. I'll point out the alternatives when I have to.20:47
dhavalrtg, ok, i will let you be the judge on that. its your decision in the end..20:48
rtgCGROUPS is definitely in the cards for Intrepid.20:48
dhavalrtg, btw, could you mention that fair_cgroup_sched is arch independent on that thread, esp since that might come up on google searches and we don't want ubuntu bz to give wrong results20:49
rtgyes - as part of updating the configs.20:49
dhavalrtg, might I also suggest to try playing with the memory controller, which allows you to control memory.. :)20:49
rtgi'll get it into the LP report.20:49
dhavalrtg, thanks!20:49
rtgdhaval: what about  CONFIG_CGROUP_NS and CONFIG_CPUSETS ?20:53
dhavalrtg, cpusets will allow you create groups of CPUs/memory banks20:53
dhavalyou can then put tasks and set them to run only on those cpusets20:53
rtgdhaval: so you recommend enabling those options as well?20:54
dhavalns is still a black hole for me.20:54
dhavalrtg, ns, i suggest you get in touch with their developers.20:54
dhavali believe cpusets is quite stable.20:54
rtgdhaval: you would be just as happy with just cpusets?20:54
dhavalrtg, did not get the question20:55
rtgdhaval: sorry, l meant would you be ok with only enabling CONFIG_CPUSETS ?20:55
rtgdo you actually use either option in practice?20:55
dhavalrtg, i am not sure if desktop users would be interested in cpusets.20:56
dhavalor ns proxy.20:56
dhavalas a user, i don't use cpusets or ns20:56
rtgdhaval: ok, since their use is optional I'll enable them as well.20:57
dhavali do use cgroup scheduling though when i want to control how my system is running as well as the memory controller to trap firefox/openioffice20:57
dhavalrtg, this is for intrepid right?20:59
blueyedDoesn't cpusets depend on cgroup?20:59
dhavalblueyed, yep, all thesedepend on cgroup20:59
blueyeddhaval: thanks for bringing this issue up again here.20:59
blueyedrtg: I've been also pondering about this, since 2.6.24 hit Hardy..20:59
rtgdhaval: I was going to enable cpusets and ns for Hardy server.20:59
dhavalblueyed, i did not know about the existence of this channel, otherwise i would have been hollering here earlier21:00
blueyedit's really a regression for users/desktops..21:00
dhavalblueyed, well, we can agree to disagree :)21:00
dhavalblueyed, i call it expected behavior21:00
dhavalrtg, i just talked to upstream folks, they like the idea, it will get it wider coverage21:00
blueyeddhaval: I'm for cgroups.. because it's the old behaviour, I mean.21:00
dhavalrtg, thanks a lot for listening!21:00
dhavalblueyed, yeah, i aree21:01
rtgdhaval: the new kernel package should show up by morning (UTC-8)21:01
dhavalrtg, thanks, i should set up hardy-rawhide or something like that on a virtual machine21:01
blueyedrtg: with cgroups only for server, still?21:02
dhavalblueyed, rtg i would still suggest desktop as well, but its your decision in the end, not mine21:02
rtgdhaval, blueyed: its actually not _all_ my decision. the kernel team discussed the ramifications, and being a generally conservative bunch, decided to only enable CGROUPS for servers.21:03
dhavalrtg, i wish they would have involved us in the discussion as well21:04
dhavalwe could have provided you with more information21:04
rtgwell, better late then never :)21:04
dhavalbeing conservative, i would have suggested cgroup and not user21:04
blueyedrtg: I've scammed the irclogs about this, and it seemed to just have been rushed over.21:04
dhavaluser is more dangerous, because of the "bugs" that will surely crop up as blueyed will confirm21:04
rtgblueyed: it may well have been, being somewhat ignorant of the issue myself.21:04
blueyedrtg: not fair.. I've been here before, there have been several bug reports it, etc..21:05
dhavalrtg, fwiw, why don't you guys get upstream folks involved as well?21:05
blueyedIt has just been ignored and then "too late now..."..21:05
dhavalblueyed, i agree..21:05
rtgits a matter of volume. there are few of us, and literally thousands of bug reports.21:05
rtgthe stuff that gets my attention are hard failures.21:06
blueyedrtg: you could have saved some time, by listening to this issue earlier.. ;) "upstream" (dhaval) came to us even..21:06
dhavalrtg, hmm. that is where getting upstream more closely involved would help21:06
dhavalrtg, fwiw, i don't use ubuntu and this experience made me feel like using ubuntu even lesser21:07
rtgcome on you guys, quit beating me up. I can on;y do so much.21:07
dhavalrtg, sorry! not meant to beat you up.21:08
dhavalbut lesson to be learnt. do involve upstream21:08
dhavalwe all want linux to get better21:08
blueyedrtg: sorry, we're not beating.. but finally somebody seems to listen, so..21:08
rtgI'm working on it. like I said, I'll have a CGROUPS server release in the morning.21:09
dhavalrtg, if you do hit scheduler bugs, do report it up, most of the folks are very responsive21:09
dhavalrtg, fwiw, it holds for all subsystems21:09
rtgwill do21:09
dhavalremember, we all want to make Linux better :)21:10
blueyedrtg: thanks. Do you understand the issue for desktop users, too? Having processes in the background (cron, daemons, especially owned by root [because it gets 2000 by default in user_sched]) makes the foreground/user's tasks slow. IMHO it's even more important for the desktop then for the server.21:12
blueyedrtg: can you please bring this up in the kernel-team again?21:12
dhavalblueyed, thanks for putting it up so clearly, thta's what i have been trying to say21:13
dhavalrtg, i have to agree with blueyed 21:13
rtgblueyed: I'll raise the issue. I'll even post a PPA kernel for testing.21:13
dhavalwhat is PPA?21:13
blueyedrtg: great. I would have put one in my Personal Package Archive (PPA) anyway.21:13
rtgdhaval: Personal Package Archive.21:13
dhavalblueyed, thanks!21:13
dhavalrtg, thanks!21:13
blueyed(or installed -server), but all that won't help normal users.21:14
dhavalblueyed, i would suggest custom sched-devel kernel :)21:14
blueyedThank you, dhaval. Also with bugging me about properly working around in boinc.. :)21:14
blueyedsched-devel?21:15
dhavalblueyed, that is the tree for the next kernel release. (currently aimed for 2.6.26)21:15
blueyeddhaval: don't ask for too much.. ;)21:15
dhavalblueyed, ah, np. (that part of the code i had written (and believe me, sysfs is a big pain to work with!), so all the hard work had to be seen in action :P )21:16
dhavalrtg, blueyed are there any resources on how i can setup the ubuntu-devel branch?21:17
dhavallet me just set it up in a virtual machine21:17
rtgdhaval: there is some wiki info. lemme find it.21:17
rtghttps://wiki.ubuntu.com/KernelMaintenance21:18
dhavalrtg, oh, i meant something similar to fedora's rawhide which is a daily build of how fedora looks?21:18
rtgdhaval: not knowing how the rawhide stuff works, I'm not sure what to tell you. I have not enabled the daily kernel build stuff yet. Its on my list for Intrepid.21:20
rtgthere is agit repo.21:20
rtgs/agit/a git/21:20
dhavalrtg, oh ok, so what fedora does is that it builds the whole distro every night and you do a yum update to get updated fully. is there something similar for ubuntu?21:21
dhavalif i can set that up, I can get you more feedback than I can do now21:21
rtgdhaval: no, the distro is only built on demand, e.g., when a package update is uploaded.21:21
dhavalhmm. ok.21:22
dhavalright, that's similar to rawhide.21:22
dhavalwhat i meant by rebuild was that you have a daily build. the updates are pushed out once aday21:23
rtgthe updates are pushed to the archive as soon as an archive admin pushes them, about once an hour.21:23
rtgthat is, after the build completes.21:24
dhavalok, so is there some documentation about that? i can't really find my way through the ubuntu wiki. Contribute brought me to a completely unexpected page :(21:24
rtgdhaval: its all based on the debian way of doing things. Its such a deep topic I'm not sure where to tell you to start. I've only been doing it for just over a year.21:26
dhavalrtg, oh. ok, proably another day then.21:27
dhavalrtg, i won't be lurking about here, but feel free to let me know by email if I can help21:27
cradekyay, my framebuffer console (radeon) works again with -1421:27
rtgdhaval: ok. thanks for the pointers.21:27
dhavalrtg, np. hope to see the cfs group scheduler in action and making people more happy with its responsiveness. btw, i've given a pointer to a daemon in the LP mail if you want to simulate user scheduling with cgroup_sched21:29
dhavalrtg, i did have one more question21:31
rtgdhaval: yes?21:32
dhavalrtg, how does one get packages in ubuntu? i'm working on a userspace library for control groups known as libcg, (http://libcg.sf.net)21:32
dhavaldo you need sponsors and such stuff?21:32
rtgdhaval: right, they are called MOTU sponsors.21:33
dhavalif this is totally off topic, just point me to the right channel21:33
rtgdhaval: is there an #ubuntu channel? Its likely a good place to start.21:33
dhavalthat seems like a user channel and /join #ubuntu-devel gave me * ubuntu-devel :That channel doesn't exist21:34
dhavalhmm. i seemed to have made a stupid mistake21:34
dhavalmised the #21:34
rtgdhaval: if that doesn't work out, send an email to Daniel Holbach <dholbach@ubuntu.com>21:35
dhavalrtg, thanks! i found #ubuntu-motu21:36
rtgok, I'm ducking out for a bit.21:36
dhavalhmm. that channel seems awfully quiet21:37
dhavalrtg, ok, thanks!21:37
blueyeddhaval: just get the beta release (or daily live cd), install it and then you can upgrade from there, as the package get built/uploaded.21:45
dhavalblueyed, link to the daily live cd?21:45
blueyeddhaval: http://cdimage.ubuntu.com/daily-live/current/21:45
blueyed(from google :p)21:45
dhavalblueyed, thanks!21:45
dhavalblueyed, too lazy, its too late here. i am supposed to be asleep about 2 hours back21:46
=== dhaval is now known as dhaval_away
tjaaltonis lbm meant to be a staging area for proposed updates to lum?21:57
liquidhackzirssi quit22:03
tjaaltonblueyed: you did the acpi-support update? I'm not sure if that's to blame, but the suspend hotkey doesn't work on my thinkpad anymore since some recent update22:06
numhello, i've done an hardy upgrade, which downloaded a new kernel 2.6.24-14, when i now boot my notebook it hangs on every new boot on a different driver. once at the iwl3945 driver loading, once at the ricoh-mmc driver. may there be something i miss or isn't done automaticaly?22:24
Nafallonum: I would try a memtest :-)22:38
numa memtest?22:39
numNafallo: but a memtest that urly during boot?22:40
Nafallonum: esc during the grub menu and choose memtest22:40
numNafallo: you suppose a damaged memory?22:41
Nafallothat's one of the first things I check. sounds like some kind of hardware issue at least.22:42
numi can't believe this, since i'm on this machine with the kernel version below, xx-1222:43
numand no problems22:43
* Nafallo shrugs22:43
NafalloI just say what it sounds like to me. and never said that it was the actual issue.22:44
numNafallo: i thank you22:44
numi appreciate any ideas, help22:44
Nafallo:-)22:45
numwhat i also saw was, that my wireless is a iwl3945, and kernel -12 shows the device as 3945ABG and the new kernel -14 shows it as 3945AB22:46
blueyedtjaalton: I don't think it was acpi-support. There have been some generic hotkeys been reported lately.. do you have a bug report? What laptop do you have?22:59
Ngis there intended to be another kernel build before release? or is 14 likely to be it barring any disasters?23:23
rtgNg: one more ABI bump tonight.23:24
rtgI did n't get CGROUPS set correctly on all arches.23:25
Ngerk23:25
Ngoh well, I was probably going to have to maintain my own lum, I can do the same with the kernel. damn new hardware ;)23:25
rtgthe joys of open source?23:26
Ngit cuts both ways, I've had all of the major hardware glitches worked out within a fairly short time of contacting upstream projects, but it's been harder to get the fine details locked down enough to push for inclusion23:26
Ngand I can make easily enough make a PPA for people with the same hardware23:28
rtgNg: what hw do you have that we are not supporting?23:28
smb_tpNg: You solved your camera problem23:29
smb_tp?23:29
Ngrtg: audio chipset and usb camera23:29
Ngsmb_tp: I think so, yeah23:29
Ngwell I didn't, upstream did :)23:29
smb_tpNg: I guess somewhere in this wretched find_control. Maybe you can give me the pointer. Or did you just get the latest driver?23:30
Ngsmb_tp: https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/200990/comments/1023:30
ubotuLaunchpad bug 200990 in linux "[hardy] oops when running cheese on Thinkpad X300" [High,Triaged] 23:30
smb_tpNg: Just missed that by seconds, hmpf. ;-)23:32
Nghehe23:33
NgI've not tried applying it to hardy source yet, working with upstream tarballs is a lot quicker for building one tiny driver ;)23:33
smb_tpNg: yeah, but looks tiny enough to be ok and if that fixes your oops its one less on the bad list. rtg: would that be alright to push into lum?23:35
rtgsmb_tp: looks fine. It demonstrably fixes the issue?23:36
smb_tprtg: So says Ng but i can put that in and make a test module23:37
rtgsmoh, I missed that port. I trust Ng. (must be getting late)23:37
rtgs/smoh/oh/23:37
smb_tprtg: It really is. :)23:38
Ngif you can make a test module with our version of uvc I'd be very happy to test it since I've only tested it with the latest upstream svn. I could try building it myself, but it might take some time ;)23:39
rtgsmb_tp: make it so. I'll get it before I upload later this evening.23:39
smb_tprtg: I'll try my very best...23:39
Ngthanks :)23:40
infinityNg: What did you do to earn this bizarre level of trust?23:44
infinityNg: Has rtg *seen* terminator?23:44
Nginfinity: I'll have you know that more than one respected member of the distro team uses it! ;)23:45
infinityNg: Can you name them, or have they begged for anonymity? :)23:46
rtginfinity: I trust him because he could delete all of my accounts. wait - is that a bad thing? I could go on vacation again.23:46
infinityrtg: I can prevent you from ever building anything again; do my patches get similar treatment? :)23:47
rtginfinity: only if you attach them to the LP report :) its so easy that way...23:47
infinityrtg: That's so much more effort than just directly committing them.23:48
Nginfinity: I might choose to protect them from association with the unfair treatment terminator gets, and the irony is that the haters all use gnome-terminal, which is hardly a paragon of quality ;)23:48
infinityNg: xterm's where it's at!23:48

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!