| rrq | fsmithred: would you have a link for that libre kernel ISO you have problem with? I though to check it out (unless you are past the problem already) | 12:13 |
|---|---|---|
| fsmithred | rrq, I haven't uploaded it. Want to do magic-wormhole? | 13:49 |
| fsmithred | I pinned mount, fdisk, libfdisk1 and uuid-runtime to the older versions, and that works for now. | 13:56 |
| fsmithred | drivers/nvme/target/nvme-loop.ko and drivers/spi/spi-loopback-test.ko are the only things with "loop" in their name in the initramfs. | 14:14 |
| fsmithred | This is the same in one that works and one that doen't work. | 14:14 |
| fsmithred | broken iso: https://get.refracta.org/files/experimental/refracta_13_minx_ndb_libre_i386-20250525_1523-BROKEN.iso | 14:34 |
| fsmithred | https://get.refracta.org/files/experimental/refracta_13_minx_ndb_libre_i386-20250525_1523-BROKEN.iso.sha256 | 14:34 |
| fsmithred | login user:user or root:root if you can get there. | 14:35 |
| rrq | there should be a "loop.ko" (unless it's builtin .. check e.g modules.builtin) | 14:39 |
| fsmithred | yeah, it's there in modules.builtin | 14:42 |
| rrq | hmm but the initramfs doesn't have /dev/loop[0-9] devnodes? | 14:45 |
| LeePen | Is this a variant of https://bugs.debian.org/1055776? | 14:46 |
| LeePen | Or maybe not (misread versions) | 14:48 |
| rrq | don't think it's the same | 14:48 |
| fsmithred | 2.38 works ok, 2.41 does not. I don't know about 2.39. | 14:48 |
| rrq | if devnodes are missing it might need a script added to /etc/initramfs-tools/scripts/init-top/ to mknod them | 14:49 |
| fsmithred | fsconfig() is where the error occurred | 14:49 |
| rrq | what is fsconfig()? | 14:49 |
| fsmithred | https://get.refracta.org/files/misc/mount_squashfs_error_2025-05-25_18-00-38.png | 14:52 |
| rrq | well first error is by losetup | 14:53 |
| rrq | does that /script/live-realpremount script have an losetup command? | 14:55 |
| rrq | (I have it donwloaded now....) | 14:56 |
| fsmithred | I'm not finding that script | 14:59 |
| rrq | nevermind. there's an issue to losetup from that mount /run/live/medium, which also cannot be unounted | 15:15 |
| rrq | not sure why | 15:18 |
| rrq | comparing a working ISO with this broken; is the kernel version the only difference? | 15:29 |
| fsmithred | no, the main difference is the version of those four packages I pinned. | 15:34 |
| fsmithred | some of the other kernels that I thought were ok had the older mount in the initramfs. | 15:35 |
| fsmithred | I can't find fsconfig in any files in util-linux or squashfs-tools | 15:36 |
| rrq | I think inprinciple the problem is that /dev/sda1 (the ISO) has got mounted onto both /run/live/medium and / where the latter was via "run-init" (which I guess is a pivot_root variant), while in "shared" mode; somehow making it lock itself from unmounting | 15:40 |
| rrq | I think fsconfig is a syscall (or nearabouts) .. doesn't explain why there is a problem | 15:52 |
| fsmithred | The problem is not present in amd64 build with the latest util-linux et. al. | 16:07 |
| fsmithred | and debian kernel, not libre | 16:07 |
| fsmithred | I spoke to libre folks and they haven't changed anything in their de-blobbing | 16:08 |
| rrq | right. yes it appears to be an issue with mounts that now has gone away | 16:10 |
| rrq | by "cat /proc/1/mountinfo" you can see that the moutns are not "shared" (it would say "shared" otherwise), and probably my ramblings about that is of no value | 16:11 |
| fsmithred | what do you mean "now has gone away"? | 16:13 |
| fsmithred | you got it to boot? | 16:13 |
| rrq | well I meant that it looks like tied to the particular kernel version | 16:13 |
| rrq | though it's getting late so I'll drop off here while still awake :) | 16:14 |
| fsmithred | ok | 16:15 |
| fsmithred | I need to go out anyway. | 16:15 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
