| 26 Jun 2025 |
 Theuni | I feel you. i'm currently bisecting the 6.12 series for another v9fs bug that is biting us in our nixos tests ... i started out with getting rate limited by the kernel.org git server because I cloned the repo over and over ... 🙈 ... took a bit to see what's going on, but strace showed me that it was throttled at 1 object per second ... which is ... a tarpit I guess ... | 19:43:46 |
 hexa | that sounds bad, yeah 😄 | 19:44:15 |
| Theuni | well. a colleague pointed out that i can leverage a local repo's working dir with the builtin fetchgit, so that' fast now. still a 25 min cycle and about 13 bisect steps left (with the added complication that there is another v9fs bug in the suspected range that I'll have to skip to see where the bug that is biting us was introduced ... ) | 19:45:38 |
| Theuni | so i can't even automate that properly because i need to manually inspect each test failure that always deviates a bit. i guess i need an AI agent /s | 19:46:20 |
 raitobezarius | Given what you are saying, I believe the Lix project already bisected the kernel bug you are looking for, unfortunately, all the commit IDs and answers are trapped behind the kernel lore mailing list browser: https://lore.kernel.org/all/w5ap2zcsatkx4dmakrkjmaexwh3mnmgc5vhavb2miaj6grrzat@7kzr5vlsrmh5/ | 19:51:38 |
| hexa | Christian Theune: ^ | 19:51:57 |
| Theuni | oh? | 19:52:43 |
| * Theuni checks | 19:52:46 |
| Theuni | for me it's behind a guru error 😉 | 19:53:10 |
| raitobezarius | I think you will get a 503 unfortunately | 19:53:10 |
| Theuni | yeah | 19:53:14 |
| Theuni | did this break recently / due to a current event? i've been working with it last week I think ... | 19:53:45 |
| raitobezarius | https://lkml.indiana.edu/hypermail/linux/kernel/2506.1/07850.html | 19:54:24 |
| raitobezarius | thankfully I have found another server | 19:54:31 |
| raitobezarius | To me, this is not a regression, this was always broken, it just takes a lot of I/O in the VM to materialize the regression | 19:54:50 |
| raitobezarius | You can try to revert the "real fix", apply the workaround and verify if this fixes your issues | 19:55:07 |
| raitobezarius | If it does, this is the folio management of 9pfs which is still broken atm | 19:55:15 |
| Theuni | ugh | 19:56:04 |
| Theuni | so yes | 19:56:07 |
| Theuni | in your mail, the last point: "massive io in the guest" is what is triggering in our cases, too | 19:56:22 |
| Theuni | and you mentioning folios makes me weep | 19:56:33 |
| Theuni | i already had a 3 year journey wrapped up of getting an xfs folio issue fixed last year -_- | 19:56:51 |
| raitobezarius | Well, you will be happy to know that the folio code is going to happen even more in more FS subsystems | 19:57:12 |
 Alyssa Ross | Didn't we switch to virtio-fs for VM tests? | 19:57:24 |
| Theuni | linus was not amused when he saw what was going on ... | 19:57:31 |
| Theuni | Alyssa Ross: 24.11 and 25.05 don't seem to be, might it have been more recent? | 19:58:15 |
| Theuni | raitobezarius: do you happen to have a patch set for 6.12.34 around that I can try out? | 19:58:51 |
| Alyssa Ross | Not more recent than 25.05 | 19:59:13 |
| Alyssa Ross | Ah it was buildInLinuxVM, not NixOS tests | 19:59:38 |
| Theuni | hmm. now ... how can i respond to that regression list thread so I get involved there ... o_O | 20:01:56 |
