jjohansen | hallyn: can I throw your tested by on the patch | 04:46 |
---|---|---|
hallyn | jjohansen: yup! | 04:48 |
alkisg | Good morning all | 05:56 |
alkisg | So currently I'm at: problem verified between vanilla 4.15rc4 and 4.15rc5, and I've also done two "git bisect good" steps after that. Each "git bisect" step takes me 1 hour, so I might hopefully finish it today... | 05:59 |
f_g | alkisg: welcome to the world of kernel bisecting, where progress can be slow and tiresome :) at least your reproducer is easy and fast - I've had bisects where the test was "do the following non-trivially automatable steps in several VMs, some of which run Windows", with the chance of triggering the problem being about 30%. | 07:00 |
f_g | so each build took ~10 minutes, then rebooting/kexecing, then preparing the VMs, then running the test (that took about 20 minutes), then repeat the preparation and test at least 5 times unless the issue triggered | 07:01 |
f_g | that was a fun week and a half ;) | 07:01 |
alkisg | Meh... well, at least I can do a few things while I'm building the kernel... but I'd really appreciate a monster PC right now, that could build it in under half an hour... | 07:01 |
thresh | :S | 07:01 |
alkisg | One of the builds needed 88 mins because I was doing some other work simultaneously :( | 07:02 |
alkisg | Testing time :) (unfortunately I can only reproduce it on my work pc, not any spare ones!) | 07:14 |
alkisg | 3rd bisect => good. I'm afraid the problem will be the pti patches that came right before rc5. | 07:20 |
f_g | alkisg: three good beside rc4? or including? | 07:28 |
alkisg | f_g: rc4 good, rc5 bad, then git bisect good 3 times | 07:33 |
alkisg | $ git bisect good | 07:35 |
alkisg | Bisecting: 20 revisions left to test after this (roughly 4 steps) | 07:35 |
alkisg | [ed1bbc40a0d10e0c5c74fe7bdc6298295cf40255] x86/cpu_entry_area: Move it to a separate unit | 07:35 |
f_g | alkisg: sounds like you should be done soon, and that it's like one of the PTI changes. | 07:37 |
alkisg | well, in 4 hours... :( | 07:37 |
f_g | out of curiosity, did you test -rc8 as well? if not, you might want to add that to your next reboot cycle.. | 07:39 |
alkisg | Yes, that was one of my very early tests, but only using the mainline .deb, not with compiling | 07:40 |
f_g | nevermind then | 07:41 |
alkisg | (I also tested the daily deb) | 07:41 |
alkisg | git status ==> HEAD detached at ed1bbc4, You are currently bisecting, | 09:36 |
alkisg | ...and yet that gave me a kernel named 4.14? Is that possible? | 09:36 |
alkisg | (while bisecting between 4.15rc4 and 4.15rc5) | 09:36 |
alkisg | In bisection step 3, I got this: linux-image-4.15.0-rc4-custom_4.15.0-rc4-custom-6_i386.deb | 09:38 |
alkisg | Then moving forward (git bisect good), I got this: linux-image-4.14.0-custom_4.14.0-custom-7_i386.deb | 09:38 |
alkisg | ...can that be because some older commit got merged later on, or does it sound like I did something wrong? | 09:39 |
alkisg | Anyway continuing with: $ git bisect good/ Bisecting: 10 revisions left to test after this (roughly 3 steps)/[9c294ec08408ed90c0f2d994a7979366675e3734] Merge tag 'powerpc-4.15-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux | 09:41 |
f_g | alkisg: yes, that is usually an artifact of bisecting into some merged branch which hasn't been rebased prior to getting merged. the git DAG in linux-stable looks rather interesting :P | 10:07 |
alkisg | ty! I got worried for my sanity there for a while :D | 10:13 |
alkisg | $ git bisect good / Bisecting: 5 revisions left to test after this (roughly 3 steps) / [24e3a7fb60a9187e5df90e5fa655ffc94b9c4f77] libnvdimm, btt: Fix an incompatibility in the log layout | 11:27 |
alkisg | Yey for grub-reboot, it allows to continue this remotely :) | 15:52 |
alkisg | $ git bisect good/ Bisecting: 2 revisions left to test after this (roughly 2 steps) /[f6c4fd506cb626e4346aa81688f255e593a7c5a0] x86/cpu_entry_area: Prevent wraparound in setup_cpu_entry_area_ptes() on 32bit | 15:54 |
alkisg | So, I'm guessing I'll find that the problem was https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?id=caf9a82657b313106aae8f4a35936c116a152299, the pti preparation patches. | 16:01 |
alkisg | What next? File a bug report in Ubuntu? Or directly upstream? Or both? | 16:01 |
alkisg | Or "blame" that commit on github? | 16:02 |
alkisg | Also, would I be able to test the latest version without that huge patch? I'm not sure it can be cleanly un-applied in the latest version... | 16:04 |
mattsm | do any scripts / automation exist to copy a ubuntu kernel to a remote machine and boot into it?/ | 16:23 |
alkisg | mattsm: I don't know, but it shouldn't be hard to do this with ssh/scp/grub-reboot... | 16:37 |
mattsm | alkisg yea i was just curious... for grub-reboot I'd need to add an entry for it is all i suppose | 16:39 |
alkisg | mattsm: normally, installing a kernel add the grub entr | 16:39 |
alkisg | grub-reboot then just selects that for the next reboot | 16:40 |
mattsm | well, i need to copy kernel + modules and add grub entry... and try it out | 16:40 |
mattsm | was just hoping for a script that copied stuff to a target machine existed | 16:41 |
alkisg | mattsm: ah, you're not using a .deb package that does all that automatically? | 16:41 |
mattsm | alkisg im attempting to rebuild a kernel on a remote machine and copy it over | 16:43 |
TJ- | mattsm: you can also reboot into another kernel using kexec - but best to test it on the actual target hardware in person first before relying on it for remote reboot :) | 16:58 |
mattsm | TJ-: i've got remote power reset, so that would be OK | 16:58 |
TJ- | mattsm: in which case after rsync-ing the /boot/vmlinuz-$VERSION, /boot/initrd.img-$VERSION and /lib/modules/$VERSION/* you can do "sudo kexec -l /boot/vmlinuz-$VERSION --initrd /boot/initrd.img-$VERSION --reuse-cmdline" (that loads them into memory) you can do either "sudo systemctl reboot" or for a more abrupt switch "sudo kexec -e" | 17:01 |
mattsm | TJ-: that sounds like it will work pretty good, also it seems like the kernel build instructions do in fact build a deb as well | 17:01 |
TJ- | Although recently I've found kexec seems to be broken on every hardware and kernel version combination I've tried | 17:02 |
mattsm | a major downside | 17:03 |
TJ- | mattsm: if you're doing "fakeroot debian/rules binary" yes that generates the packages in the parent directory | 17:04 |
mattsm | does anyone know where the zesty kernel configs are? | 17:04 |
mattsm | TJ-: yea im playing with that | 17:05 |
mattsm | i suppose that will put the config somewhere? | 17:05 |
TJ- | mattsm: they're combined at build-time from fragments (-common + -flavour specific) | 17:05 |
TJ- | mattsm: the generated config will be in the build directory as .config | 17:05 |
mattsm | TJ- nice, i think i have a zesty kernel building | 17:06 |
OliverO | Hi everyone! | 17:12 |
OliverO | Any idea why v4.14.14 (released 2 days ago) build does not show up in http://kernel.ubuntu.com/~kernel-ppa/mainline? | 17:12 |
apw | alk | 17:43 |
chiluk | OliverO it could be getting held up by requisite changes to gcc in order to enable retpoline. If I was the kteam that's what I would expect. | 17:43 |
=== ben_r_ is now known as ben_r | ||
mattsm | What gcc version does Ubuntu intent to use for building zesty kernel? | 17:52 |
mattsm | and is there a way to tell fakeroot which version to use? | 17:55 |
tomreyn | !zesty | mattsm | 18:25 |
ubot5 | mattsm: Ubuntu 17.04 (Zesty Zapus) was the 26th release of Ubuntu. Support ended on January 13th, 2018. See !eol and !eolupgrade | 18:25 |
tomreyn | just in case you're not aware of the EOL status | 18:25 |
tomreyn | other than that, to actually answer your question, my guess would be: the gcc version available in zesty. | 18:26 |
mattsm | hmm which tree is in the next version? | 18:27 |
tomreyn | !info gcc artful | 18:27 |
ubot5 | gcc (source: gcc-defaults (1.173ubuntu1)): GNU C compiler. In component main, is optional. Version 4:7.2.0-1ubuntu1 (artful), package size 5 kB, installed size 64 kB | 18:27 |
mattsm | !artful | mattsm | 18:28 |
ubot5 | mattsm, please see my private message | 18:28 |
tomreyn | note my guess can be very wrong, i'm not sure how / where default kernel image packages are built. | 18:29 |
mattsm | you made it clear to me i was confused as well | 18:30 |
mattsm | i wanted the next kernel version, so bionic beaver...? | 18:30 |
tomreyn | but i would indeed expect that building them with the gcc version of the same ubuntu release will work without an issue | 18:30 |
tomreyn | bionic is going to be 18.04 lts | 18:31 |
mattsm | right | 18:31 |
tomreyn | zesty was 17.04, artful is 17.10 | 18:31 |
tomreyn | so 'next' can mean many things ;) | 18:31 |
mattsm | well, there is an ubuntu-bionic tree in the same place | 18:38 |
alkisg | $ git bisect bad / Bisecting: 0 revisions left to test after this (roughly 1 step) / [613e396bc0d4c7604fba23256644e78454c68cf6] init: Invoke init_espfix_bsp() from mm_init() | 18:38 |
alkisg | At least it's NOT the pti preparation patches :) | 18:39 |
mattsm | Does anyone know if there is a wiki for getting changes into an Ubuntu kernel release.... | 18:56 |
TJ- | mattsm: bug report, plus email to the kernel-team mailing list | 19:11 |
mattsm | TJ- perfect, thanks | 19:13 |
alkisg | $ git bisect bad / Bisecting: 0 revisions left to test after this (roughly 0 steps) / [92a0f81d89571e3e8759366e050ee05cc545ef99] x86/cpu_entry_area: Move it out of the fixmap | 19:56 |
alkisg | ...so I guess I'll file this to bugzilla.kernel.org... | 19:57 |
alkisg | https://bugzilla.kernel.org/show_bug.cgi?id=198529 | 20:20 |
ubot5 | bugzilla.kernel.org bug 198529 in x86-64 "Reboot on kernel load due to 92a0f81d" [High,New] | 20:20 |
TJ- | alkisg: best to send an email to the maintainer and sub-system mailing list (using "scripts/get_maintainer.pl path/to/source/file.c" ) | 20:21 |
alkisg | TJ-: after reporting the bug, bugzilla said it notified his mail | 20:22 |
alkisg | Is that enough, or should I also send it to the sub system mailing list? | 20:22 |
TJ- | alkisg: oh, cool, i've not noticed it do that for my reports! | 20:22 |
alkisg | Email sent to: bugsfx@gmail.com, tglx@linutronix.de | 20:22 |
alkisg | That second mail there is the committer | 20:22 |
TJ- | alkisg: yeah, that's Gleixner :) | 20:23 |
alkisg | I saw a follow up commit where fixes a breakage in 32bit, and he blames "the moron that wrote the code", and I think that's again himself... sounds like he has a sense of humor :D | 20:24 |
TJ- | alkisg: he signed off under protest on a commit for 4.14 that fixed a regression for nvidia that I reported with " Despised-by: Thomas Gleixner <tglx@linutronix.de>" (commit 87df26175e67) | 20:28 |
alkisg | Haha | 20:29 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!