If a Debian 12 server or VM started freezing a few minutes after its last reboot, the kernel update that arrived on 15 November 2026 is the likely cause. The package is linux-image-6.1.0-55, and it can crash machines that use the ext4 filesystem with a “kernel NULL pointer dereference”.
Key takeaways
- Debian 12’s kernel update of 15 November (6.1.0-55, version 6.1.192-1) can freeze machines that use ext4.
- The crash shows as a kernel NULL pointer dereference in ext4_mb_load_buddy_gfp, minutes or hours after boot.
- Booting the previous kernel, 6.1.0-54, and making it the GRUB default keeps affected systems stable.
- Update, November 17, 2026: the fixed kernel, 6.1.0-56 (6.1.192-2), shipped on November 16. Move to it, because 6.1.0-54 lacks the new security fixes.
This guide shows how to check whether you are running the affected kernel and how to boot back into the previous one so it sticks. It also explains what to do when the fixed package ships. As of this writing there is a tested fix but no official package yet.
What happened
On 15 November, Debian’s Long Term Support team published DLA 4906-1, a security update for the linux package in Debian 12 “bookworm”. It moved the kernel to version 6.1.192-1, which installs as linux-image-6.1.0-55-amd64 (or the matching -cloud-amd64, -arm64 or -686-pae package). It fixed a long list of CVEs, so many machines installed it automatically.
Within hours, a critical bug report titled “System freeze after update to linux-image-6.1.0-55-amd64” was opened. By Monday morning, Debian’s kernel team had merged six more reports into it. The story also reached the Hacker News front page.
The reports describe the same pattern:
- The machine boots normally on 6.1.0-55.
- Some time later the console shows
BUG: kernel NULL pointer dereference, usually insideext4_mb_load_buddy_gfp. - Services such as SSH, mail and web stop responding, and often only a hard reset works.
- Booting the previous kernel, 6.1.0-54 (6.1.190-1), makes the system stable again.
Reporters saw it on bare-metal servers and on Xen, KVM and Hyper-V virtual machines, on amd64, arm64 and 32-bit i686. One admin wrote that crashes came anywhere from 25 seconds to about 5 hours after boot, “so hosts that look fine may just not have hit it yet.”
By the numbers
6.1.192-1the affected package version
25 secondsearliest reported crash
5 hourslatest reported crash
Why the kernel crashes
The cause is a small backporting mistake, explained in an upstream patch posted to the ext4 mailing list on 2 November.
Stable kernels like 6.1 receive fixes copied back from newer kernels. Some ext4 changes written for kernel 6.4 and later were copied into the 6.1 line, first in the upstream 6.1.191 release. In newer kernels, a memory function called __filemap_get_folio() signals failure with an error code. In 6.1 it returns NULL (an empty pointer) instead. The copied code checked for the error code, missed the NULL, and then tried to use it.
In practice this hits when the kernel has dropped some of ext4’s cached allocation data, which happens normally under memory pressure. The next time the filesystem allocates or frees blocks, it reads the empty pointer and the kernel oopses. Every process writing to that filesystem then hangs, which is why the machine looks frozen rather than rebooting cleanly.
The patch author also found and fixed a similar mistake in the ntfs3 driver. It needs much rarer conditions to trigger.
Did you know?
The bug came from fixes copied back from newer kernels. In kernel 6.4 and later a memory function reports failure with an error code, but in 6.1 it returns NULL, so the copied check missed it.
Are you affected?
You are in the risk group if all three are true:
- You run Debian 12 (bookworm).
- The running kernel is 6.1.0-55 from package version 6.1.192-1.
- Your root or data filesystem is ext4, which is the Debian default.
Check the first two in a terminal:
uname -r
uname -v
Output like 6.1.0-55-amd64 and #1 SMP PREEMPT_DYNAMIC Debian 6.1.192-1 (2026-11-14) means you are on the affected build.
To see whether it already crashed during an earlier boot, search the previous boot’s kernel log. This only works if your journal is kept on disk:
sudo journalctl -k -b -1 | grep -E "NULL pointer|ext4_mb_load_buddy"
Bug reporters also posted a two-line command that triggers the crash in seconds. Don’t run it on a machine you care about: on an affected kernel it hangs every writer on that filesystem until you power-cycle.
Debian 13 “trixie” uses the 6.12 kernel line, which is not the package in this bug, and the upstream explanation is specific to 6.1.

How to roll back to 6.1.0-54
These steps assume your machine boots with GRUB and that the previous kernel is still installed. That is the usual case: each new kernel ABI installs as a separate package, so the upgrade to 6.1.0-55 leaves 6.1.0-54 in place.
1. Confirm the previous kernel is installed.
dpkg --list 'linux-image-6.1.0-5*' | grep ^ii
You want to see a linux-image-6.1.0-54-... line. If it’s missing, skip ahead to “Common errors and fixes”.
2. Boot it once from the GRUB menu (console access). If you can reach the screen or the hypervisor console, reboot and pick Advanced options for Debian GNU/Linux, then the entry ending in 6.1.0-54-amd64 (or your flavor, such as -cloud-amd64). This is what the first reporter did to get a stable server back.
3. Make it the default so automatic reboots don’t undo it. First list the exact menu titles on your machine:
grep -E "^submenu|^\s*menuentry" /boot/grub/grub.cfg | cut -d"'" -f2
Then edit /etc/default/grub and set GRUB_DEFAULT to the submenu title, a > sign and the entry title, copied exactly from that list. On a standard amd64 install it looks like this:
GRUB_DEFAULT="Advanced options for Debian GNU/Linux>Debian GNU/Linux, with Linux 6.1.0-54-amd64"
The GRUB manual documents this “submenu>entry” form. Apply it and reboot:
sudo update-grub
sudo reboot
4. Check that it worked.
uname -r
It should now print 6.1.0-54-amd64 (or your flavor). If the box stays up past the point where it used to freeze, you are back on a working kernel.

When the fixed kernel ships
Update, November 17, 2026: the fixed kernel is out. Debian closed bug #1154101 on November 16 with version 6.1.192-2, and bookworm-security now carries it as linux-image-6.1.0-56-amd64, which the linux-image-amd64 package depends on (checked against the bookworm-security archive). The fix came with a new 6.1.0-56 package name, so a normal sudo apt update && sudo apt full-upgrade installs it alongside 6.1.0-55. As of this update, no follow-up LTS advisory to DLA 4906-1 had been posted. After installing it, set GRUB_DEFAULT=0 again, run sudo update-grub and reboot, as described below, so GRUB stops booting 6.1.0-54.
Before the fix shipped, it was close but not released. On the morning of 16 November, Debian’s kernel team linked the upstream patch, opened a merge request against the bookworm security branch and posted an unsigned amd64 test build. Several affected admins replied that the test kernel ran without crashes, and one confirmed the quick reproducer no longer hangs it.
These were the two places to watch for the official package:
- The bug report, which Debian closed on November 16, 2026 with version 6.1.192-2.
- The Debian LTS advisories page, where a follow-up linux advisory is expected (none had been posted as of November 17, 2026).
When it lands, update and reboot onto the new kernel, then undo the pin:
sudo apt update && sudo apt full-upgrade
Set GRUB_DEFAULT=0 again in /etc/default/grub, run sudo update-grub and reboot. Then confirm with uname -v that you’re on the new package version, not 6.1.192-1.
Don’t use apt-mark hold as a shortcut. Before the release, the test build kept the 6.1.0-55 name, so a hold could have blocked the fix. If you held a kernel package anyway, release it with sudo apt-mark unhold <package> before you upgrade.
Did you know?
Debian’s test build kept the 6.1.0-55 name, but the released fix arrived as a new package, linux-image-6.1.0-56-amd64, so a normal full-upgrade picks it up.
Common errors and fixes
- The machine freezes before you can edit GRUB. Use the console (or your cloud provider’s serial or VNC console) to pick 6.1.0-54 from the menu first, then make the change from the stable kernel.
- No 6.1.0-54 entry in the menu. The old kernel was removed. Restoring a snapshot or backup from before 15 November is the fastest route back; one reporter had to do exactly that.
GRUB_DEFAULTchange ignored. You probably edited the file but didn’t runsudo update-grub, or the title doesn’t match your menu exactly. Copy it from thegrepoutput, including the>between the submenu and the entry.- No GRUB menu appears on the console. Many installs hide it or count down quickly. Per the GRUB manual, pressing Esc, F4 or holding Shift during the countdown shows the menu when the timeout style is hidden or countdown.
What it means for you
This is a regression, not an attack: nobody exploited anything. But it hit servers that patched exactly as they’re supposed to. A few habits limit the damage next time a kernel update goes wrong:
- Keep one older kernel installed. It is your one-step escape hatch. Don’t purge old kernels right after an update.
- Have console access ready. For VMs and cloud servers, know where your provider’s emergency console is before you need it. Our guide to setting up SSH keys covers the remote side.
- Stage updates. Let kernel updates reach a test machine or a small group of servers a day before the rest, especially if unattended upgrades also reboot.
- Don’t stay on the old kernel forever. 6.1.0-54 lacks the security fixes in DLA 4906-1. Treat the rollback as temporary and move to the fixed package once it ships.
Bad updates aren’t unique to Linux. Windows had its own November update printing and activation problems, and patching web servers quickly still matters, as the Apache HTTP/2 flaw showed. More fixes and set-up guides live in Software & How-To.
Facts in this article are as of 16 November 2026, 16:00 PKT (11:00 UTC). This post will be updated when Debian releases the fixed kernel.
Frequently asked questions
Why does my Debian server freeze after the latest kernel update?
The 6.1.0-55 kernel (6.1.192-1) from 15 November contains a backporting mistake in ext4. When cached allocation data is dropped, the kernel reads a NULL pointer and oopses, and every process writing to that filesystem hangs.
How do I boot an older kernel in Debian?
At the GRUB menu choose Advanced options for Debian GNU/Linux and pick the older entry. To keep it, set GRUB_DEFAULT in /etc/default/grub to the submenu title, a > sign and the entry title, then run sudo update-grub.
Is Debian 13 trixie affected by the 6.1.0-55 bug?
Debian 13 uses the 6.12 kernel line, which is not the package in this bug, and the upstream explanation of the fault is specific to the 6.1 kernel.
Should I stay on kernel 6.1.0-54?
Only until the fixed package ships. 6.1.0-54 lacks the security fixes in DLA 4906-1, so treat the rollback as temporary, then upgrade and set GRUB_DEFAULT back to 0.
Quick quiz
Which filesystem is involved in the 6.1.0-55 crash?
- Btrfs
- ext4
- XFS
The oops happens inside ext4_mb_load_buddy_gfp, part of ext4’s block allocator.
Before the fix shipped, why was apt-mark hold on linux-image-6.1.0-55-amd64 a bad idea?
- The fix could have arrived as a new version of that same package
- Holds delete the old kernel
- apt-mark doesn’t work on kernels
The test build kept the 6.1.0-55 name, so a hold could have blocked the fix. It finally shipped as 6.1.0-56.
What is the safe way to check if you are affected?
- Run the drop_caches and dd reproducer
- Reinstall Debian
- Run uname -r and uname -v
uname shows the kernel build; the reproducer can hang the filesystem until a power cycle.
Sources
- Debian bug #1154101: System freeze after update to linux-image-6.1.0-55-amd64
- DLA 4906-1: linux security update (Debian LTS announcement)
- linux-ext4: Fix NULL folio dereferences from IS_ERR() checks in 6.1.191
- Debian kernel team merge request 2168: ext4: fix NULL folio dereference in mballoc
- Debian LTS security advisories
- Debian Wiki: Long Term Support
- GNU GRUB manual: the default variable
- Debian bookworm-security archive (linux-image-amd64 6.1.192-2)





Write a response