[13:10] why the ddebs for linux-image are 15Kb only ? anyone knows ? [13:11] I see ddebs.ubuntu.com/dists was moved to dists.old [13:30] rafaeldtinoco: are you looking at the real ABI linux-image package; or just the meta which indeed is empty usually? [13:30] I was using info from Packages to generate BTF files for HWE kernels [13:30] https://github.com/aquasecurity/btfhub/tree/main/ubuntu/20.04/x86_64 [13:31] because Ubuntu Bionic and Focal don't enable BTF by default [13:31] script stopped working recently that is all I see #) [13:32] BTW updating dwarves for Bionic and Focal will allow you to enable embedded BTF into HWE kernels for both. That would save eBPF vendors a lot of trouble ^. [13:33] wget http://ddebs.ubuntu.com/dists/bionic-updates/main/binary-amd64/Packages -O bionic-updates -> now package points to a 15Kb ddeb file [13:34] https://github.com/aquasecurity/btfhub#ubuntu (about eBPF support topic) [14:25] rafaeldtinoco: i have backpowered dwarves to enable BTF by default [14:25] rafaeldtinoco: but SRU team is not accepting it =( [14:26] rafaeldtinoco: i'm not sure you are doing things right. [14:26] rafaeldtinoco: linux-image..... blah blah contains no debug symbols. but depends on unsigned ddebs package. you need unsigned linux image ddeb for ddebs [14:27] Package: linux-image-unsigned-5.0.0-61-generic-dbgsym [14:27] ah ok.. then it makes more sense.. I recently changed that [14:27] Installed-Size: 5749733 [14:27] because of a regex change. thanks xnox [14:27] sorry for the noise then [14:27] and rbasak ^ if you could help reconsidering dwarves backport [14:28] to enable BTF in HWE kernels for Bionic and Focal [14:28] that will open a new world to Ubuntu (regarding k8s and eBPF) [14:28] for already installed base at least =) [14:28] xnox: thx a lot [14:42] hi! fyi, I think that linux-firmware needs an update for i915 for hwe-5.11: https://bugs.launchpad.net/ubuntu/+source/linux-firmware/+bug/1889137/comments/12 [14:42] Launchpad bug 1889137 in linux-firmware (Ubuntu) "HWE kernel is missing firmwares" [Undecided, Confirmed] [14:44] tjaalton: hey, fyi since you did the last linux-firmware update for focal ^ [14:58] jdstrand: right, those are harmless though [14:59] although the noise isn't nice [15:19] tjaalton: oh? it seemed scary since I use i915 [15:19] but ok [15:40] jdstrand: guc isn't used at all before 5.15, and even there it's only enabled for gen11 and up (Ice Lake ->) [15:41] huc is used for h.264 encoding [15:41] required, I mean. and that's not a common use case [15:43] hmm no, HEVC/H.265 [16:24] rafaeldtinoco: the dwarves *SRU* is blocked on https://bugs.launchpad.net/ubuntu/+source/dwarves-dfsg/+bug/1912811/comments/7. Nothing else to reconsider I don't think? [16:24] Launchpad bug 1912811 in dwarves-dfsg (Ubuntu Hirsute) "Update dwarves-dfsg in focal to version 1.21 from hirsute" [Undecided, Confirmed] [16:25] rafaeldtinoco: separately I'd like to discuss the backports pocket as a better destination, and it would be good to discuss that before accepting the SRU. [16:25] rbasak: I'll think on that commit through. Safest way would be to have 'pahole' binary statically compiled added to old dwarves perhaps. [16:25] without changing dwarves entirely for the rest of the OS. [16:26] rafaeldtinoco: maybe it would be easier to push into backports though? Any reason that is not acceptable? [16:26] rbasak: compiling kernel with -backports ? [16:26] that was my -1 on that idea (before) =( [16:26] and I did see the -backports thread indeed [16:26] Oh, the kernel has to build-depend on this? [16:26] I still have to get back to it though [16:27] rbasak: yep, the pahole binary is used to play with DWARVES/BTF on kernel generation (kernel build-deps on dwarves and it is old for HWE builds) [16:27] Ah [16:27] OK, so SRU then, but that means even more care needs taking! [16:27] tjaalton: ack, thanks! [16:27] And, does it mean we need to SRU verify against a rebuilt kernel? [16:27] rbasak: are you super -1 on having a newer static pahole (being compiled together with old dwarves source) added to the binary pkg ? [16:28] the only change to final package would be a static binary added (that kernel could use if present) [16:28] and the delta would be to add the pahole binary generation logic [16:28] rafaeldtinoco: definitely not -1 right now. I don't fully understand the situation so cannot have an opinion yet. [16:28] ok [16:28] Ill work on it and describe it fully in the bug. [16:28] Thanks! [16:28] If backports cannot work then I'm open to an SRU still. [16:28] and about -backports, will get back to the thread soonish. Sorry on the dealys [16:29] rbasak: that will enable BTF for Bionic and Focal [16:29] and that is a key competitive feature against centos/fedora/alma [16:29] fyio [16:29] I just thought backports would be easier, but if that can't work, then it's certainly worth considering the SRU. [16:29] nice! tku! === mhcerri0 is now known as mhcerri [17:08] rbasak: backports no good, as it needs to be kernel's build-dep. [17:08] rbasak: nothing but the kernel uses those packages, and kernels have stop using them, because they are too old. [17:24] xnox: lots of good explanations to go in the bug please, thanks :)