=== cpaelzer_ is now known as cpaelzer | ||
pekkari | Hi, any easy way to get the kernel config file used for linux-image packages of different series? | 07:38 |
---|---|---|
apw | pekkari, they are installed with the kernels in /boot, and are collected here | 07:39 |
pekkari | apw: fine thanks! | 07:46 |
apw | ppisati, http://kernel.ubuntu.com/~kernel-ppa/config/ | 07:47 |
apw | sorry got distracted getting the url | 07:47 |
apw | pekkari, http://kernel.ubuntu.com/~kernel-ppa/config/ | 07:48 |
apw | even | 07:48 |
apw | xnox, hey ... there is a kernel rebuild in ppa:canonical-kernel-team/ubuntu/unstable which looks to have the right bits in s390x .udebs are you able to confirm | 07:58 |
xnox | apw, that looks good. However, I am wonder if performance will be degraded on x86_64 because: | 08:23 |
xnox | x86 optimised .ko modules do not appear to be in the udeb. | 08:23 |
xnox | ./lib/modules/4.8.0-13-generic/kernel/arch/x86/crypto/crc32-pclmul.ko | 08:23 |
xnox | ./lib/modules/4.8.0-13-generic/kernel/arch/x86/crypto/crc32c-intel.ko | 08:23 |
xnox | hopefully things fallback to generic crc32[c] modules instead. | 08:24 |
xnox | looks good. obviously i'll need d-i to verify, but hopefully things are fine. | 08:24 |
alkisg | Is it possible to install a 4.x kernel in ubuntu 12.04? I tried e.g. with the xenial kernel, but the "kmod" dependency isn't satisfied... | 08:48 |
apw | xnox, yep, that we can fix | 08:48 |
apw | (after beta) | 08:48 |
apw | alkisg, the newest kernel we offer there is lts-trusty 3.13 | 08:49 |
alkisg | apw, thanks, a user reports that he has sound in 16.04 but not in 12.04, even with the -trusty kernel there | 08:50 |
alkisg | (12.04 is his normal installation, 16.04 a live cd) | 08:50 |
alkisg | So I was trying to see if an even newer kernel in 12.04 would solve the issue | 08:50 |
apw | i would guess that the modutils/kmod change is more about the transitional package for the old name going away | 08:52 |
apw | so you might find forcing that on works just fine | 08:52 |
alkisg | Thanks, trying even with broken dependencies... | 08:53 |
alkisg | Nah the user said that the 4.x kernel didn't even boot | 10:00 |
=== zyga_ is now known as zyga-ontheroad | ||
smoser | hey... i just marked https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1626158 as kernel related. | 16:48 |
ubot5` | Ubuntu bug 1626158 in linux (Ubuntu) "image won't boot after upgrading to yakkety's 4.8 kernel because efi" [Critical,Confirmed] | 16:48 |
chalbersma | Is this Ask Ubuntu Accurate (http://askubuntu.com/questions/804111/is-no-reboot-kernel-patching-enabled-in-16-04)? Is there now live kernel patching in Ubuntu 16.04? I asked on #ubuntu and they said they didn't think so but I should check with you guys. | 19:59 |
dsmythies | Ever since mainline kernel 4.7-rc5, I have been observing sometimes extreme load averages and over 2000 kworker processes (8 CPUs and up to 258 threads per cpu, maybe more). | 20:38 |
dsmythies | Particularly immediately after boot. | 20:38 |
dsmythies | Eventually (5 minutes +- a few seconds) the kworker threads go away, but might return depending on what root activity is going on (it doesn't happen for user activities). | 20:38 |
dsmythies | The issue is somehow related to the massive kernel configuration changes between 4.7-rc4 and 4.7-rc5, and if I compile say 4.8-rc7 using the old kernel configuration, the issue does not occur. | 20:38 |
dsmythies | If I do "scripts/diffconfig .config-4.7.0-040700rc4-generic .config-4.7.0-040700rc5-generic" there are 2249 lines of changes. | 20:38 |
dsmythies | I have compiled several kernels with my best guess at some configuration reversions, but so far have not been able to isolate the exact change that introduced the issue. | 20:38 |
dsmythies | This work has all been on my main test computer, which is 16.04 server based. However today the same thing was observed on a 16.10 based Laptop test computer, after it upgraded to 4.8.0-11-generic. | 20:39 |
dsmythies | It has two CPUs and I observe up to 516 koworker threads, when the issue occurs (used to be about 11 max). | 20:39 |
dsmythies | Does anyone know which kernel configuration change is responsible? And if this is behaviour is actually O.K.? | 20:39 |
dsmythies | To be clear the kworker processes are mostly idle, and the load average seems to be high only because the queue happened to be huge on a sample boundary. This is NOT a hogging the CPU issue. | 20:39 |
rtg | dsmythies, try reducing CONFIG_NR_CPUS=8192 to 256 or something low like that. | 20:45 |
dsmythies | rtg: O.K. that was a change between 4.7-rc4 and 4.7-rc5: "NR_CPUS 8192 -> 512", and it is one that I have not yet tried reverting. | 20:49 |
rtg | dsmythies, it just seems like that is a lot of worker threads. What file system do you use ? | 20:50 |
dsmythies | rtg: my main disk is ext4, on both computers. | 20:58 |
rtg | huh, I've not observed that many threads, though I haven't looked recently | 20:59 |
rtg | one of my 4-way servers does have 589 kworker threads | 21:00 |
rtg | dsmythies, if that does turn out to have an impact, please start a bug and assign me to it (timg-tpi). I'm EOD for now | 21:01 |
dsmythies | rtg: on my server, it occurs every half hour, when CRON runs the php clean up. i.e. /usr/lib/php/sessionclean. I also have now created a way to do it simply myself. | 21:02 |
dsmythies | rtg: O.K. | 21:02 |
dsmythies | does anyone know what rtg meant by "(timg-tpi)"? (3 lines above). | 21:31 |
dsmythies | reverting CONFIG_NR_CPUS made no difference. | 21:32 |
apw | dsmythies, that is his launchpad id | 21:33 |
dsmythies | apw, aghh thanks. | 21:43 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!