| === n0u_ is now known as n0u | ||
| CarlFK | amitk: are you the kernelopps expert? | 02:48 |
|---|---|---|
| CarlFK | http://dpaste.com/93532/ dmesg shows 3 stackdumps | 02:50 |
| amitk | CarlFK: IIRC, you are running kerneloops on a server with no GUI? | 08:42 |
| apw | smb_tp, so we have a bug about extending p4-clockmod to include a newer celeron | 11:56 |
| apw | although it doesn't generally seem to be expected to be that useful, there are good stats reported | 11:56 |
| apw | in the bug for power savings and the like, how do we feel about slurping up something like that | 11:56 |
| apw | i get the feeling p4-clockmod is not loaded automatically even if you have such a cpu | 11:57 |
| smb_tp | apw, IFAIK no, you have to add it manually. About the savings I know there are different opinions. Normally throttling is rather seen as temperature control as the overall energy savings are not positive | 11:59 |
| smb_tp | apw, On the other hand, i the cpu does nothing it can do it half speed without taking longer, I guess | 12:00 |
| apw | yeah i saw those arguments, but some of the users seem to report very good savings on overall power from an idle system, whcih could make some sense | 12:00 |
| apw | yeah it depends some on how the idling is performed | 12:00 |
| apw | if the module is an opt-in thing anyhow it doesn't seem like a bad thing to include support for additional cpus to my mind | 12:01 |
| smb_tp | apw, I think if that change is small, we could add it as a SAUCE patch as hw enablement | 12:01 |
| apw | its a single line, to add a new processor id to p4-clockmod, | 12:01 |
| apw | i am wondering why its not upstream as yet, not managed to find any discussion on that side | 12:02 |
| smb_tp | apw, I have the feeling sometimes, that this is somewhat abandoned due to the savings arguments | 12:03 |
| apw | yeah, and yet we have the module at all ... so it clearly does something for someone | 12:03 |
| apw | even if it is only for heat, then we should have it for heat (we == linux kernel community more than ubuntu) | 12:04 |
| smb_tp | apw, It makes some sense I think. Maybe mjg59 knows a bit more about whether there is some upstream maintenance to that. | 12:05 |
| apw | yeah ... have just found the kerneltrap thread on it, so will go read that and see what the outcome was | 12:06 |
| apw | this thread is pretty compelling as to why you really don't want it, how it makes things worse only | 12:16 |
| smb_tp | apw, What is the reasoning there? | 12:17 |
| apw | that the normal state of things is that when the system is idle it enters C2 state, which consumes the same amount of power as being in a skipped cycle due to throttling | 12:19 |
| apw | so on average the the overall consumption would be the same | 12:19 |
| apw | assuming you don't need more than half the cycles | 12:20 |
| apw | which is normal | 12:20 |
| apw | if you do need all the cycles, then you are actually worse off, as you run as half+idle cost for 2x the time and that can lead you to consume more power not less | 12:20 |
| apw | i think the right thing to do is write a clear and concise summary of this thread and post it in the bug and see what happens | 12:22 |
| smb_tp | apw, The strange thing then is only that it seems to have a measurable effect. Do you know whether C states are implemented on centrinos? | 12:26 |
| apw | that is an interesting question, this thread inplies so, but perhaps there is a bug elsewhere that prevents their use or something | 12:30 |
| gregd | hi guys, just was having some problems with my intel wireless card (was dropping constantly connection). Many people seemed to have similar problems. Found that the newest kernel 2.6.28-rc6 has a fix for it (look for "iwlagn: fix RX skb alignment | 12:38 |
| gregd | " in http://kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.28-rc6). I'm really looking forward for the fix to be incorporated into ubuntu-kernel. But in the meantime, should I compile the 2.6.28-rc6 kernel by myself? | 12:38 |
| amitk | gregd: a 2.6.28-rc6 kernel should be available for jaunty this week. If you can wait, then you could simply install that kernel. | 12:39 |
| gregd | this week seems to more than enough, that would be perfect :) | 12:40 |
| amitk | gregd: then just check back here around Thursday. We are shooting for an upload today. | 12:41 |
| gregd | great! ok will be back later this week! Thanks guys! | 12:41 |
| tjaalton | does hardy have a package to clean up old kernel versions, keeping the current and latest kernel alone? | 13:00 |
| amitk | tjaalton: moi! Not that I know of. There was an attempt to get it in Intrepid. But it was removed at the last minute due to some bugs. | 13:03 |
| tjaalton | amitk: ah ok, thanks | 13:05 |
| tjaalton | I'm using pkgsync anyway, so adding old versions to the maynothave-list works quite well :) | 13:06 |
| === asac_ is now known as asac | ||
| CarlFK | amitk: it's u-desktop, but kerneloops wasn't installed. just installed it. | 14:03 |
| amitk | CarlFK: is the kerneloops-applet running? | 14:04 |
| CarlFK | i just didi sudo apt-get install kerneloops - I don't see anything on the bar - it would be over by the clock, right? | 14:08 |
| amitk | CarlFK: no. It doesn't show up in the tray. | 14:09 |
| amitk | CarlFK: "ps aux | grep kerneloops" | 14:09 |
| CarlFK | yup: /usr/sbin/kerneloops | 14:13 |
| amitk | CarlFK: run kerneloops-applet from cmdline | 14:14 |
| CarlFK | in case it isn't clear: I just now installed it, after the crash (thought it was installed - you asking made me check...) | 14:14 |
| CarlFK | running | 14:14 |
| CarlFK | got dialog | 14:15 |
| CarlFK | 'always' | 14:15 |
| CarlFK | is that it? | 14:15 |
| amitk | yup | 14:15 |
| CarlFK | neat | 14:15 |
| CarlFK | brb | 14:15 |
| amitk | CarlFK: i think when you logout/reboot your computer, the applet should be automatically started. | 14:15 |
| apw | rtg if we were considering pulling a newish driver back into intrepid, which was not already there would it go into the main tree or lbm ? | 14:43 |
| rtg | apw: it could go main tree 'cause there is little chance of regression. what driver? | 14:44 |
| apw | i felt it might, this is a panasonic laptop keys support driver | 14:44 |
| rtg | just be sure to evaluate the regression potential. | 14:45 |
| apw | no idea yet if its is a trivial backport or not, and yep that needs looking into | 14:45 |
| amitk | BenC: rtg: why can't skipabi=true really skip an abi check? The third check in scripts/abi-check is really not necessary in case of a new arch or a quick compile test. | 15:00 |
| rtg | amitk: I never use it, so it's never really annoyed me. | 15:00 |
| rtg | amitk: besides, for a subsystem quick compile test I use a completely different method. | 15:01 |
| amitk | rtg: you never have to do test builds when ABI is not yet updated? | 15:01 |
| rtg | rarely. | 15:02 |
| BenC | amitk: touch and empty file | 15:02 |
| rtg | Try something like this: 'make -C debian/build/build-generic M=`pwd`/drivers/net/wireless' | 15:02 |
| amitk | rtg: it isn't a subsystem compile check, I want to compile the whole shebang but not bother about abi. | 15:03 |
| BenC | amitk: basically you will need to just create empty ABI files for each flavour...the check fails because we never want to upload without the previous abi files | 15:03 |
| amitk | BenC: dpkg-buildpackage doesn't like empty files | 15:03 |
| BenC | amitk: then create one fake symbol entry :) | 15:03 |
| BenC | amitk: in your case an out-of-tree build usually works well for a compile test | 15:04 |
| amitk | BenC: out-of-tree would be ok for compile test. But if I wanted to check for packaging issues, I would need to use a binary-<flavour> target | 15:05 |
| BenC | amitk: abi files with one fake symbol is what you will need | 15:05 |
| BenC | amitk: either that or just grab the abi files from the initial build(s) as stub's | 15:06 |
| BenC | I just nailed down the final changes for the headers packages (the move from include/asm-* to arch/*/include/asm needed to be handled) | 15:06 |
| BenC | fglrx dkms build fine with the new header packages, so I think I'm done and ready for an upload | 15:07 |
| BenC | amitk: let me know when you have everything in place for armel | 15:07 |
| amitk | BenC: armel is all set since last night. | 15:07 |
| BenC | amitk: including the abi files? | 15:07 |
| amitk | BenC: yes. I already used the stub from a local build as the inital abi | 15:08 |
| BenC | I see they exist, so cool | 15:08 |
| amitk | BenC: who needs to be informed about EXTRAVERSION change? Have you checked the change? | 15:08 |
| BenC | amitk: let me commit the stuff I have and then I'll pull | 15:09 |
| BenC | amitk: best bet is ubuntu-devel@ | 15:10 |
| amitk | BenC: build will fail. http://paste.ubuntu.com/76801/ Need to change some hardcoded paths for new EXTRAVERSION | 15:25 |
| BenC | amitk: instead of changing just extraversion, how about we add -ub- to the entire thing, even package name? | 15:38 |
| BenC | amitk: linux-meta will take care of upgrades (where we'll keep it flavour package names with no -ub-) | 15:39 |
| amitk | BenC: but would that reflect in /proc/version? | 15:41 |
| BenC | amitk: well the flavour name gets added to extraversion | 15:43 |
| BenC | right now it's $(abi)-$(flavour)...we just need to change it to $(abi)-ub-$(flavour) | 15:44 |
| BenC | amitk: I have to prep the audio patches for smagoun...if you could get this -ub- thing squared away, it will help make sure we still get an upload today | 15:45 |
| amitk | BenC: ohh right. I think that is a better proposal. | 15:46 |
| === shay is now known as shayx | ||
| evand | Looking over bug 92014, I started to wonder if it's feasible to add some hint in /sys/block to discern between SCSI and SATA devices. Does anyone know if this is possible and whether upstream would accept it? | 18:51 |
| amitk | BenC: -ub added to EXTRAVERSION and package names. Compile tested on amd64. | 20:11 |
| amitk | BenC: can you take over from here? | 20:11 |
| * amitk calls it a day... | 20:19 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!