| === tomreyn_ is now known as tomreyn | ||
| Eickmeyer | Hi everyone! I've got my Ubuntu Studio hat on, and I'm not happy that the lowlatency kernel is officially lagging behind the generic kernel in jammy. This is what I feared would happen as a result of splitting the source package. | 15:25 |
|---|---|---|
| Eickmeyer | I'm seeing not only a divergence of version numbers, but the lowlatency kernel is taking far more time in -proposed. | 15:26 |
| Eickmeyer | This makes no sense as there's very little difference in build flags. | 15:27 |
| xnox | it failed automated testing | 15:58 |
| xnox | and you will notice many cloud kernels failed too | 15:58 |
| xnox | generic is now on 25, for which lowlatency will be respun too | 15:59 |
| xnox | jammy has no support, and no guarantees that various kernel flavours would go out together. Unlike SRU cycles which we aim to release all kernels together. | 16:00 |
| xnox | Eickmeyer: what problems arise from different lowlatency build numbers? | 16:01 |
| xnox | does your daily iso fail to build? | 16:01 |
| Eickmeyer | xnox: No, I'm mostly worried about the security implications in addition to the grub menu ordering if both generic and lowlatency are installed, such as cases where people use ubuntustudio-installer on other flavors. Ubuntu Studio isn't just a flavor, it's also a toolset/configuration. | 16:02 |
| Eickmeyer | We have a package that mitigates the ordering (ubuntustudio-lowlatency-settings), but people don't necessarily have to install it. | 16:03 |
| Eickmeyer | As you know, grub always picks the highest version number of any given kernel as the default. | 16:04 |
| Eickmeyer | BUT, I'm mostly worried that it will fall-behind with security patches as more attention will be given to the generic kernel, unlike before when they were in the same source it forced both to have equal attention. | 16:05 |
| xnox | Eickmeyer: all releases have -generic kernels that are below and above lowlatency; meaning all releases can have unpredictable grub menu, unless grub.d conf snippet is specified as to which flavour is prefered | 16:06 |
| xnox | all our kernels have the same support time frame | 16:06 |
| xnox | jammy has no security support; and you notice that all jammy kernels at the momement are behind security fixes compared with impish/focal/bionic/xenial | 16:07 |
| xnox | in stable releases; generic and lowlatency are spun together, using the same source code (and security fixes) doh | 16:07 |
| xnox | Eickmeyer: your concerns and worries are unjust and missplaced. | 16:07 |
| xnox | the current migration delays of lowlatency appears to be mostly driven by introduction of arm64 support; which appears to be regressing in test results within jammy release. | 16:08 |
| Eickmeyer | xnox: So, when jammy is stable, it generic and lowlatency kernels will be spun together? Is that what I'm understanding? | 16:08 |
| xnox | yes. | 16:10 |
| Eickmeyer | I'm just looking for reassurances, I'm sorry if i come across as accusatry. | 16:10 |
| Eickmeyer | Ok, that's all I needed to know. | 16:10 |
| xnox | and GRUB_FLAVOUR_ORDER is required if one always wants "generic" or "lowlatency" as the top boot entry | 16:10 |
| xnox | because for example in focal today 5.4.0-XX-generic < 5.4.0-XX-lowlatency < 5.15.0-XX-generic < 5.15.0-XX-lowlatency in grub's mind | 16:11 |
| xnox | depending on which meta's one has installed. | 16:11 |
| Eickmeyer | Yeah, we've got a mechanism in place, and also adds "threadirqs" to the lowlatency kernel command line, because even though it's built with the right flags, threadirqs isn't activated by default (ever). | 16:11 |
| xnox | (in case somebody installed intentionally or accidentally linux-(generic|lowlatancy|virtual)[[-hwe-20.04]-edge] | 16:12 |
| Eickmeyer | xnox: Either way, thanks for the reassurance. | 16:13 |
| xnox | Eickmeyer: at least there _is_ lowlatency in jammy-release | 16:14 |
| xnox | Eickmeyer: we still don't have some kernels in jammy-release that need to go on the iso =))))) | 16:14 |
| Eickmeyer | Yeah, I saw oem is like that currently. | 16:14 |
| tjaalton | because I don't understand why some tests are failing :) | 16:18 |
| tjaalton | generic -25 failed testing too, but it was manually overridden | 16:22 |
| xnox | some of them are weird indeed | 16:25 |
| tjaalton | like, what's the hwclock test? I don't see it on any other kernel, unstable included | 16:27 |
| tjaalton | not true, it's on oem-5.14 too | 16:37 |
| === fling_ is now known as fling | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!