Public bug reported:

1) Release of Ubuntu
Ubuntu 26.04.1 LTS (Resolute Raccoon), Server install.

2) Package version
linux-image-generic:
  Installed: 7.0.0-43.43
  Candidate: 7.0.0-43.43

Running kernel: 7.0.0-43-generic #43-Ubuntu SMP PREEMPT_DYNAMIC Fri Oct
23 13:25:25 UTC 2026

3) What I expected to happen
The system should remain stable serving large mmap'd model files from ext4 via 
the NVIDIA 595 open driver, without kernel panics during mm teardown, futex, or 
ext4 writeback paths.

4) What happened instead
The host has suffered repeated wild-pointer kernel panics on 7.0.0-38, 
7.0.0-42, and 7.0.0-43-generic:

- BUG: unable to handle page fault in zap_pte_range / folio_remove_rmap_ptes 
during process exit
- BUG: unable to handle page fault with a wild RIP in the futex path
- BUG: unable to handle page fault in ext4_bio_write_folio during writeback
- BUG: Bad page state in kswapd0

vmcores and dmesg are saved in /var/crash/. A near-identical public
report exists at LP #2173885 (7.0.0-43-generic, NVIDIA 595 OOT, large
mmap'd models, NULL-pointer crash in lruvec_stat_mod_folio ->
folio_remove_rmap_ptes -> zap_pte_range).

Root cause / fix requested
The upstream fix is commit 9b25f381de6b ("ext4: unmap invalidated folios from 
page tables in mpage_release_unused_pages()"). It is in mainline 7.1+ and 
7.3.x, but the upstream Linux 7.0 stable branch is EOL (last release 7.0.14, 
2026-06-27), so it has not backported into Ubuntu 7.0 automatically. I verified 
it is absent from 7.0.0-43.43 and proposed 7.0.0-44.44.

I am requesting a cherry-pick of 9b25f381de6b into the Ubuntu Resolute
7.0 kernel series. A backport patch against Ubuntu-7.0.0-44.44 is
attached.

Regression risk
Very low: +15/-1 lines in fs/ext4/inode.c, only in the invalidate=true branch 
of mpage_release_unused_pages(), using the standard unmap_mapping_pages() 
helper.

Attachments
- ext4-stale-pte-fix-ubuntu-7.0.patch — proposed backport
- release.log — lsb_release, uname, /proc/version_signature, apt-cache policy 
linux-image-generic
- lspci-vnvn.log — lspci -vnvn

Testing offered
I can test a proposed 7.0.0-4x kernel build on the affected host and report 
whether the wild-pointer panics stop.

** Affects: linux (Ubuntu)
     Importance: Undecided
         Status: New

** Attachment added: "patch (cheery picked, LLM generated) & logs"
   
https://bugs.launchpad.net/bugs/2174770/+attachment/6018939/+files/logs_and_patch.zip

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2174770

Title:
  SRU request: cherry-pick ext4 stale-PTE fix (commit 9b25f381de6b) into
  Ubuntu 26.04 7.0 kernel

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2174770/+subscriptions


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

Reply via email to