Seven more kdump reboots since my 2026-09-06 comment (#8–#9), all on
7.0.0-31-generic with MGLRU still disabled (lru_gen/enabled → 0x0000).
P1 sysctl + THP=never mitigation applied 2026-09-06 did not stop recurrence.

Additionally, two earlier captures from 2026-08-08 were recovered from
/var/crash retention when archiving logs on 2026-09-15 (#10–#11).

**Burst:** four crashes on 2026-09-13 within ~11 hours (including one with
dozens of WARN stack traces from python3 before final Oops at lru_activate).

**Tonight (2026-09-15 22:15 JST):**
- Oops: page fault at __page_cache_release.part.0+0xac/0x240
- Comm: grok-bot (Cursor/Electron internal thread)
- Path: do_madvise → folios_put_refs → __page_cache_release
- Boot uptime: ~31 min (1849 s)

Recurrence table update (#10–#18):

| Date (JST)        | Uptime   | RIP                         | Comm             
 | Path hint        |
|-------------------|----------|-----------------------------|-------------------|------------------|
| 2026-08-08 00:30  | ~11.5 h  | __page_cache_release+0xac   | DedicatedWorker  
 | madvise          |
| 2026-08-08 13:55  | ~13.4 h  | isolate_lru_folios+0x121    | kswapd0          
 | LRU reclaim      |
| 2026-09-13 11:51  | ~2.3 h   | __page_cache_release+0xac   | krita            
 | page cache       |
| 2026-09-13 16:19  | ~4.4 h   | isolate_lru_folios+0x121    | kswapd0          
 | LRU reclaim      |
| 2026-09-13 22:34  | ~20 min  | lru_activate+0x91           | python3          
 | after WARN storm |
| 2026-09-13 22:59  | ~25 min  | __page_cache_release+0xac   | tokio-rt-worker  
 | exit_mmap        |
| 2026-09-14 19:11  | ~6 min   | __page_cache_release+0xac   | cursor           
 | do_madvise       |
| 2026-09-14 23:12  | ~4 h     | __page_cache_release+0xac   | MainThread       
 | exit_mmap        |
| 2026-09-15 22:15  | ~31 min  | __page_cache_release+0xac   | grok-bot         
 | do_madvise       |

Note: six of these nine share the **exact same RIP offset** (+0xac) at
__page_cache_release.part.0 — strong evidence of one underlying folio/LRU bug.

Hardware unchanged. Not OOM. Full kdump dmesg archived locally (host
backup script).

Still waiting for 7.0.0-32+ with upstream MGLRU v2 backport.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2162722

Title:
  general protection fault in lru_gen_del_folio during exit_mmap
  (ThreadPoolForeg)

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2162722/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to