# Debian 12 kernel 6.1.0-55 is freezing machines on ext4. How to roll back and hold it

> The Debian 12 (bookworm) kernel security update shipped on November 15 as DLA-4906-1 is being blamed for hard freezes on ext4 systems, from bare metal to cloud VMs. There's no fixed build yet. If you haven't rebooted into 6.1.0-55, hold it. If you have, boot back into 6.1.0-54.

Published Nov 16, 2026 by TechMeasured Staff · Security · Canonical: https://techmeasured.com/security/debian-12-kernel-6-1-0-55-ext4-freeze-dla-4906-1-roll-back-hold-what-to-do

Debian 12 "bookworm" machines are hanging after rebooting into the kernel from the latest security update. Debian's [LTS security page](https://www.debian.org/lts/security/) lists **DLA-4906-1** for the `linux` package on **November 15**. The [security tracker](https://security-tracker.debian.org/tracker/source-package/linux) shows the matching bookworm-security build as **6.1.192-1**, which installs as the `6.1.0-55` kernel ABI. Users began reporting freezes on Debian bug **#1154101**, and by Monday the report was on the [Hacker News front page](https://news.ycombinator.com/item?id=50514180).

This is a problem for anyone still on Debian 12, which is now in long-term support. A kernel security update is exactly what admins and unattended-upgrades are supposed to install quickly. In this case, the next reboot can leave a server hung and needing a hard reset. As of Monday no fixed build had shipped.

## What's breaking

The most detailed public write-up comes from [ComputingForGeeks](https://computingforgeeks.com/debian-kernel-6-1-0-55-freeze-fix/), which reproduced the crash in its own KVM lab. Its findings:

| | |
|---|---|
| Bad kernel | `linux-image-6.1.0-55-amd64`, 6.1.192-1 |
| Last good kernel | `linux-image-6.1.0-54-amd64`, 6.1.190-1 |
| Symptom | Kernel NULL pointer dereference in `ext4_mb_load_buddy_gfp`, followed by a full hang |
| Time to crash | Reported anywhere from about 25 seconds to about 5 hours after boot |
| Reported on | Bare metal, Xen, Hyper-V, and KVM guests at Hetzner and OVH. One report also covers the `-cloud-amd64` flavour |
| Fixed build | None confirmed as of November 16 |

According to the write-up, a change backported into the 6.1 stable branch altered how the ext4 block allocator handles page-cache folios. A helper now returns NULL when a page isn't cached, but the backported code only checks for an error value. Once memory pressure evicts the relevant bitmap page, the next write to that block group dereferences NULL. That explains why busy, write-heavy machines tend to fail sooner.

The oops often never reaches the journal. ComputingForGeeks found it on the serial console and in EFI pstore (`/var/lib/systemd/pstore/`). If a box simply went unresponsive, look there before assuming hardware.

## What to do

**1. Check what you're running and what's installed.**

```
uname -r
dpkg -l 'linux-image-6.1*' | grep ^ii
```

**2. If 6.1.0-55 is installed but you haven't rebooted, hold the kernel metapackage** so nothing pulls it in. ComputingForGeeks recommends holding the metapackage rather than the versioned image, because the fix may ship under a new ABI number:

```
sudo apt-mark hold linux-image-amd64
apt-mark showhold
```

Cloud and arm64 kernel flavours have their own metapackages, which need their own holds. The article didn't test those.

**3. If you're already on -55, make 6.1.0-54 the default boot entry:**

```
sudo sed -i 's/^GRUB_DEFAULT=.*/GRUB_DEFAULT=saved/' /etc/default/grub
sudo update-grub
sudo grub-set-default 'Advanced options for Debian GNU/Linux>Debian GNU/Linux, with Linux 6.1.0-54-amd64'
sudo grub-editenv list
```

The menu-entry text has to match exactly. A typo silently falls back to the first entry, which is -55. Check `uname -r` after rebooting.

**4. After any freeze, check your filesystems and backups.** One reporter cited in the write-up had to restore a system disk from backup.

**5. Unhold once a fixed build lands.** Confirm that the bug report lists a "Fixed in" version and that `apt-cache policy linux-image-amd64` shows a candidate newer than 6.1.192-1. Then run `sudo apt-mark unhold linux-image-amd64` and upgrade.

## The trade-off

Staying on 6.1.0-54 means going without whatever DLA-4906-1 patched, so treat the hold as temporary. ComputingForGeeks says three reporters found an unsigned amd64 test build (6.1.192-1a~test1) stops the crash. Secure Boot machines will reject it, so most production fleets should wait for the signed package. Hacker News commenters suggested moving to the 6.12 kernel from bookworm-backports, or upgrading to Debian 13 "trixie". Both are bigger changes, and nobody has confirmed either one fixes this particular bug.

Debian has been here before. In December 2023, bookworm's 6.1.64 kernel could [corrupt ext4 filesystems](https://groups.google.com/g/fltkgeneral/c/xjQNQDaykMs), and users were told to hold it until a fixed kernel shipped in point release 12.4. The lesson is the same: on fleets that patch kernels automatically, stage reboots on a canary first.


## Sources

1. [Debian — LTS Security Information (DLA-4906-1, linux, 15 Nov 2026)](https://www.debian.org/lts/security/)
2. [Debian Security Tracker — source package linux (bookworm-security 6.1.192-1)](https://security-tracker.debian.org/tracker/source-package/linux)
3. [ComputingForGeeks — Debian 6.1.0-55 Freeze: Roll Back and Hold](https://computingforgeeks.com/debian-kernel-6-1-0-55-freeze-fix/)
4. [Hacker News — Debian Becomes Unresponsive after 6.1.0-55 Kernel Update (discussion of bug #1154101)](https://news.ycombinator.com/item?id=50514180)
5. [fltk.general — NOTICE: Debian Stable/Bookworm: kernel upgrade currently broken (December 2023 precedent)](https://groups.google.com/g/fltkgeneral/c/xjQNQDaykMs)
