=== JanC_ is now known as JanC | ||
=== zyga is now known as zyga-ubuntu | ||
ogra | uh, oh ... | 12:26 |
---|---|---|
ogra | the current linux package has all the config scripts (splitconfig.pl etc) set to non-executable | 12:27 |
ogra | debian/scripts/misc/../config-check: Permission denied | 12:27 |
ogra | (thats what i get after running fdr editconfigs on a xenial source package ... (before there was the same error about splitconfig.pl whihc i manually made executable) | 12:28 |
ogra | something seems to have dropped all exec bits in the scripts dir | 12:28 |
ogra | apw, ^^^ want a bug ? | 12:32 |
ogra | (thats with 4.4.0-80 on xenial) | 12:33 |
apw | ogra, nope that is expected ... you should be working in the git tree not a source pacakge | 12:34 |
ogra | pffft | 12:34 |
apw | ogra, or fi you are working in the source package you need to know debian packages lose execute bits on things, because diff | 12:34 |
apw | i suppose we might be able to fix that with the reconstruct stuff, maybe | 12:35 |
* apw will card that | 12:35 | |
ogra | well, it used to work in the past ... thats definitely new behaviour | 12:35 |
ogra | (and it breaks all the cross compilation howtos that use the source deb ) | 12:35 |
apw | ogra, that is definatly not new behaviour, it always does that for any package with a .orig, and has since i join 10 years ago | 12:36 |
apw | it _only_ breaks changing the config, not building | 12:36 |
ogra | well, i did re-build the deb since years when trying out stuff on arm | 12:36 |
apw | as we do not assume or need executable bits on those | 12:36 |
ogra | which always includes editconfigs when i do that | 12:37 |
ogra | so i'm sure this used to work (though admittedly i dont think i cross built a kernel since trusty) | 12:37 |
ogra | but well, i'll hack in the exec bits ... | 12:38 |
apw | yeah hackedy hack | 12:39 |
ogra | hrm ... | 12:40 |
ogra | config-check freaks out ... amout arm64 and powerpc ... which i didnt even touch | 12:41 |
ogra | *about | 12:41 |
ogra | wow, even when i start from zero and run editconfigs without doing any change it explodes | 12:44 |
ogra | apw, oh, ignore me ... | 12:58 |
* ogra just noticed https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel#Modifying_the_configuration | 12:58 | |
ogra | :P | 12:58 |
apw | ogra, i am always willing to ignore you :) | 13:07 |
cjwatson | apw: exec bits> only true for source format 1.0; 3.0 preserves exec | 13:12 |
apw | cjwatson, true enough, we are very backward in that sense | 13:12 |
apw | cjwatson, not that 3.0 preserves everything either, wihch is annoying | 13:13 |
cjwatson | apw: ogra probably perceives it as new behaviour because at various times in the past the kernel has been a native source package | 13:13 |
apw | cjwatson, most of the time in devel too | 13:13 |
ogra | oh, it isnt native anymore ? i didnt even notice that | 13:15 |
ogra | but yeah, that would explain it | 13:15 |
apw | ogra, once it has an .orig it isn't native ... which occurs at release of the base kernel version | 13:16 |
ogra | apw, right, but it used to | 13:16 |
apw | each new devel kernel is native, then gains an orig before release | 13:16 |
=== koza is now known as koza|away | ||
tomreyn | thanks for the help some ~ 20 hours ago, CJ | 18:38 |
shewless | Hello. running the hwe kernel for 16.04.2. Looking at this problem: http://people.canonical.com/~inaddy/lp1328088/. Basically linux namejspace creation is slow.. which is causing me problems with my openstack deployment | 20:08 |
shewless | Trying to play around with kernel options related to RCU_NOCB_CPU and rcu_nocbs= but not having much luck so far | 20:09 |
shewless | wondering if someone can help make sure the options I'm setting in grub are taking effect | 20:09 |
shewless | We've seen that creating namespaces (multiple) is less efficient on systems with more cores.. so a 1 core system does way faster than a 40 core system | 20:11 |
shewless | I tried to do this to force these RCUs to happen only on CPU 3 but it didn't seem to do anything for me: GRUB_CMDLINE_LINUX_DEFAULT="CONFIG_RCU_NOCB_CPU_ALL=n isolcpus=3 rcu_nocbs=3" | 20:12 |
shewless | do I need to build the kernel myself with some options enabled in order for this to work? I'm open to that but not sure what options I would need to enable | 20:13 |
shewless | hmm I suppose since https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=728dba3a39c66b3d8ac889ddbe38b5b1c264aec3 | 20:21 |
shewless | that RCU stuff won't be valid.. since it's not using rcu_lock anymore | 20:21 |
shewless | hmm. rcu_lock is still used in some of that code so it might have an impact? | 20:26 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!