On November 18, 2026, the nginx project released nginx 1.30.7 (stable) and 1.31.9 (mainline), fixing 10 security issues. Three matter most:
- CVE-2026-123007: memory corruption in HTTP/2 flow control;
- CVE-2026-123633: a use-after-free in HTTP/2;
- CVE-2026-118899: an infinite loop while reading a request body over TLS 1.3.
Researchers have nicknamed the first two Nginx2Shell, because they can let an attacker run code on a server without logging in. The third can freeze nginx so it stops serving visitors. The same release also fixes memory flaws in nginx's HTTP/3, resolver and mail proxy code. nginx is one of the most widely used web servers in the world, so this guide covers who is affected, how to check your server, and how to patch it today.
Key takeaways
- nginx 1.30.7 (stable) and 1.31.9 (mainline), released November 18, 2026, fix 10 security issues.
- CVE-2026-123007 and CVE-2026-123633 are HTTP/2 flaws that can lead to remote code execution without logging in. Researchers have shown it working on a server serving only static files.
- CVE-2026-118899 can freeze nginx with a crafted TLS 1.3 connection, and turning off HTTP/2 does not stop it.
- The same release fixes use-after-free flaws in HTTP/3 (CVE-2026-121810), the resolver (CVE-2026-123634) and the mail proxy (CVE-2026-113425).
- If nginx -v still shows 1.24.0 or 1.22.1 after you upgrade, your server runs your distribution's nginx package. See "Ran the upgrade and nginx is still on 1.24.0 or 1.22.1?" for your two options.
- WebGee hosting platforms and managed servers are patched by our engineers. Unmanaged cloud and dedicated servers need to be upgraded by their owners.
What this means for WebGee customers Hosting (cPanel, WordPress, WordPress Cloud and reseller): already patched by our engineers, no action needed. Managed servers: our engineers upgrade and patch them automatically, so there's nothing you need to do and no need to contact us. Unmanaged cloud and dedicated servers: you manage the software, so if nginx runs on your server, directly, through Plesk or in Docker containers, please follow the steps below today.
The three main flaws in plain terms
CVE-2026-123007: HTTP/2 flow-control memory corruption
HTTP/2 lets the browser tell the server how much data it's ready to receive, using messages called WINDOW_UPDATE. By sending a flood of them, an attacker can push nginx's internal counter past its limit, which makes nginx write data outside the memory it set aside. In the best case the nginx worker crashes. In the worst case, researchers showed it can be turned into remote code execution, even on a server that only serves plain static files, with no application behind it.
It affects nginx 1.9.5 to 1.31.8, a bug dating back to 2015, whenever HTTP/2 is enabled. F5, which maintains nginx, rates it 9.2 out of 10 (Critical).
CVE-2026-123633: HTTP/2 use-after-free
This one affects nginx used as a reverse proxy, the common setup where nginx sits in front of an application such as Node.js, PHP-FPM or a container. When nginx runs short of free connection slots, it can close an HTTP/2 connection that is still in use and then keep using that freed memory. An attacker who can cause that situation can corrupt memory and, potentially, run code. It affects nginx 1.19.7 to 1.31.8 with HTTP/2 enabled.
CVE-2026-118899: TLS 1.3 infinite loop
With HTTPS and TLS 1.3 enabled, which is the default on modern servers, a specially crafted connection can trap an nginx worker in an endless loop while it reads a request body. The worker uses 100% CPU and stops handling visitors. Repeat it for each worker and the site goes down. It doesn't let anyone run code, but it is an easy denial of service. It affects nginx 1.13.0 to 1.31.8, and turning off HTTP/2 does not protect you.
Three more flaws fixed in the same release
These only affect servers using specific nginx features, but they're fixed by the same upgrade, so check whether they apply to you:
- CVE-2026-121810: HTTP/3 use-after-free. If you've turned on HTTP/3 (a
listen ... quicline in your configuration), nginx can use freed memory when it runs short of connection slots. Turning off HTTP/2 does not cover this one. - CVE-2026-123634: resolver use-after-free, in nginx 0.6.18 to 1.31.8. It affects servers where nginx looks up hostnames itself, which is the case when your configuration has a
resolverline, typically withproxy_passto an address held in a variable. Two lookups for the same name at the same moment can make nginx use freed memory. - CVE-2026-113425: mail proxy use-after-free, rated Major, in nginx 0.5.33 to 1.31.8. It only affects nginx used as a mail proxy, with a
mail { }block in the configuration. Most web servers don't use this.
The release also fixes four lower-risk issues in the stream, try_files, PROXY protocol and ssl_preread code. Upgrading covers all ten at once.
How urgent is this? No attack code had been published at the time of writing, and no attacks in the wild had been reported. But the researchers demonstrated working remote code execution, and flaws like this are usually attacked within days of the details spreading. Treat it as a patch-today issue.
Step 1: check your server
Connect to your server over SSH and run these checks. The first shows your nginx version; anything below 1.30.7 (or below 1.31.9 on the mainline branch) is affected. The others show which of the risky features are turned on.
# 1. Which nginx version is installed?
nginx -v
# 2. Is HTTP/2 enabled anywhere in your configuration?
sudo nginx -T 2>&1 | sed "s/#.*//" | grep -nE "listen[[:space:]].*http2|^[[:space:]]*http2[[:space:]]+on[[:space:]]*;"
# 3. What do clients actually negotiate? (replace the IP with your server's)
openssl s_client -connect 127.0.0.1:443 -alpn h2 </dev/null 2>&1 | grep -E "ALPN protocol:|No ALPN"
# 4. HTTP/3, resolver or mail proxy in use?
sudo nginx -T 2>&1 | sed "s/#.*//" | grep -nE "listen[^;]*quic|^[[:space:]]*resolver[[:space:]]|^[[:space:]]*mail[[:space:]]*\{"If the second command prints any lines, or the third prints ALPN protocol: h2, HTTP/2 is on. The fourth shows whether HTTP/3, the resolver or the mail proxy is in use. Don't forget nginx running somewhere less obvious: Plesk uses nginx in front of Apache by default, and many Docker images and Kubernetes ingress controllers are built on nginx too.
Which nginx package do you have?
This decides how you upgrade in step 2. If nginx -v shows something like nginx/1.24.0 (Ubuntu), or an older version such as 1.22.x or 1.24.x, you're almost certainly on your Linux distribution's own package, not the official nginx.org one. To be sure, check where it came from:
# Debian / Ubuntu: which repository is the installed version from?
apt policy nginx
# AlmaLinux / Rocky / RHEL: who built the installed package?
rpm -qi nginxIn apt policy, look at the lines under the installed version (marked ***): an nginx.org address means the official package, while an Ubuntu or Debian address means the distribution's. In rpm -qi, nginx.org builds have .ngx in the Release line; anything else is the distribution's package.
Step 2: upgrade nginx
The real fix is upgrading to nginx 1.30.7 or 1.31.9. Take a snapshot of your server first, so you can roll back in seconds if anything goes wrong.
If you use the official nginx.org packages
The fixed versions are already available. Update and restart:
# Debian / Ubuntu with the nginx.org repository
sudo apt update && sudo apt install --only-upgrade nginx
# AlmaLinux / Rocky / RHEL with the nginx.org repository
sudo dnf upgrade nginx
# Check the configuration, then restart so every worker runs the new code
sudo nginx -t && sudo systemctl restart nginx
nginx -vIf you use your Linux distribution's own nginx package
Ubuntu, Debian, AlmaLinux, Rocky and other distributions ship their own, older nginx versions and add security fixes to them, keeping the old version number. On the day of the release, those fixed packages were still on their way. Install updates as soon as they appear, and check whether your installed package includes the fixes:
# Debian / Ubuntu: does the package changelog mention the fixes?
sudo apt update && apt changelog nginx 2>/dev/null | grep -E "CVE-2026-(123007|123633|118899|121810|123634|113425)"
# AlmaLinux / Rocky / RHEL
rpm -q --changelog nginx | grep -E "CVE-2026-(123007|123633|118899|121810|123634|113425)"No result means your package isn't fixed yet. Apply the stopgap in step 3 now, and either wait for your distribution's update or switch to the official nginx.org repository, which already has the fixed versions.
Plesk and Docker
- Plesk: run Tools & Settings › Updates and install the latest updates; Plesk ships its own nginx build. Until the update arrives, use the stopgap below through Plesk's nginx settings.
- Docker and Kubernetes: pull the patched nginx image as soon as it's published, rebuild any image based on
nginx, and redeploy. Restarting an old container doesn't update it.
Ran the upgrade and nginx is still on 1.24.0 or 1.22.1?
You ran the upgrade commands, saw no errors, and nginx -v still shows the old version: nginx/1.24.0 (Ubuntu) on Ubuntu 24.04, nginx/1.22.1 on Debian 12, or the distribution's own version on AlmaLinux, Rocky and other RHEL-family servers.
That's because your server runs your distribution's nginx package, not the one from nginx.org. Distributions freeze the version number for the life of a release and only add (backport) fixes to it, so apt upgrade or dnf upgrade can never take you to 1.30.7. The nginx.org commands above only do something once the nginx.org repository is set up, which is why they ran without errors and changed nothing.
You have two options.
Option A: stay on your distribution's package
Keep the package you have and install your distribution's update as soon as it's published. Use the changelog check above to see whether your installed package lists the CVE IDs, and keep the stopgap in step 3 in place until it does.
Option B: switch to the nginx.org package
This gets you nginx 1.30.7 now. The easiest route is our upgrade script. Start with check: it changes nothing, and tells you your distribution, which nginx package you have, whether your current configuration passes nginx -t, which modules it uses and what will need attention after the switch. It ends with "Ready to upgrade" or a list of things to fix first. Then run upgrade, which asks you to confirm before it changes anything:
# See what would happen (changes nothing)
wget -qO- https://security.webgee.com/upgrade-nginx.sh | sudo bash -s -- check
# Switch to nginx.org and upgrade
wget -qO- https://security.webgee.com/upgrade-nginx.sh | sudo bash -s -- upgrade
# Changed your mind later? Put the previous nginx and configuration back
wget -qO- https://security.webgee.com/upgrade-nginx.sh | sudo bash -s -- rollbackThe script:
- checks before it changes anything: it refuses to start if your current configuration already fails
nginx -t, or if switching would remove packages other than nginx's own; - adds the official nginx.org repository for your distribution and installs the latest stable nginx from it (add
--mainlinefor the 1.31 branch). On Debian, Ubuntu, the RHEL family and SLES it verifies nginx.org's signing key first; - backs up
/etc/nginxto/etc/nginx.bak-<date>-<time>and keeps your existingnginx.conf(the packaged one is saved next to it asnginx.conf.ngx-dist); - brings your modules across: if your configuration uses GeoIP, image filter, njs, Perl, XSLT or OpenTelemetry, it installs the nginx.org version of the module and loads it for you. If you use a module nginx.org doesn't provide (headers-more or Brotli, for example), it stops without changing anything;
- keeps your sites online: on Debian, Ubuntu and RHEL-family servers, nginx keeps serving while the packages are swapped and is restarted once at the end;
- rolls back by itself: it tests the new nginx with
nginx -tand restarts only if the test passes. If it fails, the script puts your previous nginx package and configuration back automatically; - keeps a log of everything it does in
/var/log/webgee-nginx-upgrade.log; - supports Debian, Ubuntu, the RHEL family (AlmaLinux, Rocky, Oracle Linux, CentOS Stream), Amazon Linux 2 and 2023, SLES and openSUSE Leap, and Alpine. Fedora isn't supported, because nginx.org doesn't publish packages for it;
- refuses to run on control-panel servers (cPanel, Plesk, DirectAdmin, CyberPanel, HestiaCP, CloudPanel and similar) unless you force it with
--force.
What changes when you switch packages:
- Distro modules are replaced. Your distribution's module packages (
libnginx-mod-*on Debian and Ubuntu) are removed. The ones nginx.org also ships are installed from nginx.org and loaded automatically, and the stream and mail modules are built into nginx.org's nginx. - Different defaults. The nginx.org package runs as the
nginxuser rather thanwww-data, and usesconf.dinstead ofsites-enabled. Because the script keeps your existingnginx.conf, your current user andsites-enabledsetup carry on working; these only matter if you switch to the packagednginx.conf. - A brief restart at the end, typically about a second.
Jumping from 1.22 or 1.24 to 1.30 You may see a warning that "listen ... http2" is deprecated: replace it with "http2 on;" inside the server block. nginx 1.30 also changes the proxy default to HTTP/1.1 with keep-alive for backend connections, so test your proxied applications after the upgrade. The script's check lists the lines this affects.
On Plesk or another control panel? Don't use the script. Update nginx through the panel's updater, as described under "Plesk and Docker" in step 2.
Step 3: the stopgap if you can't upgrade today
Turning HTTP/2 off closes the two code-execution flaws. If you use HTTP/3, turn that off as well by removing the quic listen lines. Visitors still connect normally over HTTP/1.1; sites may load a little slower. This does not protect against the TLS 1.3 loop, the resolver flaw or the mail proxy flaw, so it is only a bridge until you upgrade.
# nginx 1.25.1 and newer: in each server block
http2 off;
# Older versions: remove "http2" from every listen line
# before: listen 443 ssl http2;
listen 443 ssl;
# HTTP/3: remove or comment out QUIC listeners
# listen 443 quic reuseport;
# Then test and reload:
# sudo nginx -t && sudo systemctl reload nginxWatch out for this gotcha If a listen line still says "http2", the newer "http2 off;" setting in the same server block is ignored. Remove http2 from the listen lines too, then re-run the ALPN check from step 1: it should now say "No ALPN negotiated". If you serve gRPC, disabling HTTP/2 breaks it, so upgrade instead.
Step 4: confirm, and look for trouble
- Run
nginx -vagain and confirm 1.30.7, 1.31.9 or a distribution package whose changelog lists the fixes. - Check the error log for worker process exited on signal 11 (a segmentation fault). A burst of these before you patched can be a sign someone tried the memory flaws.
- Look for workers stuck at 100% CPU in
top, the sign of the TLS 1.3 loop. - If you suspect a break-in, check for unknown users, SSH keys, cron jobs and processes, and restore from a snapshot taken before the suspicious activity.
# Crashed nginx workers in recent logs
sudo grep -h "exited on signal 11" /var/log/nginx/error.log* 2>/dev/null | tail -20
# Any nginx worker pinned at high CPU right now?
ps -eo pid,pcpu,etime,cmd --sort=-pcpu | grep "[n]ginx: worker" | headChecklist for unmanaged servers
- Find every server and container running nginx, including Plesk and Docker
- Check the version with nginx -v
- Confirm which package is installed, the distribution's or nginx.org's (apt policy nginx or rpm -qi nginx)
- Take a snapshot before changing anything
- Upgrade to nginx 1.30.7 or 1.31.9, or a distribution package that lists the CVEs
- Restart nginx so every worker runs the new code
- If you can't upgrade today, turn HTTP/2 and HTTP/3 off and confirm with the ALPN check
- Check the error log for signal 11 crashes and look for signs of a break-in
Who does what
- WebGee hosting (cPanel, WordPress, WordPress Cloud and reseller): already patched by our engineers. Nothing to do.
- Managed servers: our engineers upgrade and patch nginx automatically. Nothing to do, and no need to contact us.
- Unmanaged cloud and dedicated servers: you manage the server's software, so the upgrade is yours to do, using the steps above.
Would you rather not handle patches like this yourself? Moving to a managed WebGee server or managed hosting puts security updates in our engineers' hands, and we migrate your sites for free.
Frequently asked questions
Which nginx versions are affected?
CVE-2026-123007 affects nginx 1.9.5 to 1.31.8, CVE-2026-123633 affects 1.19.7 to 1.31.8, CVE-2026-118899 affects 1.13.0 to 1.31.8, CVE-2026-123634 affects 0.6.18 to 1.31.8 and CVE-2026-113425 affects 0.5.33 to 1.31.8. Fixed versions are 1.30.7 (stable) and 1.31.9 (mainline).
Is my WebGee hosting affected?
No action is needed. Our engineers have already patched WebGee's hosting platforms, including cPanel, WordPress, WordPress Cloud and reseller hosting.
I have a managed WebGee server. Do I need to do anything?
No. Our engineers upgrade and patch managed servers automatically, so you don't need to contact us or change anything.
I have an unmanaged cloud or dedicated server. What should I do?
On unmanaged servers you are responsible for software updates. If nginx runs on your server, directly, through Plesk or inside Docker containers, check the version with nginx -v and upgrade to 1.30.7 or 1.31.9 by following this guide. If you can't upgrade today, turn HTTP/2 and HTTP/3 off until you can.
I upgraded but nginx -v still shows an old version. Why?
Your server is running your Linux distribution's own nginx package, such as 1.24.0 on Ubuntu 24.04 or 1.22.1 on Debian 12. Distributions keep the version number and only backport fixes, so apt or dnf upgrade will never reach 1.30.7. Either check your package's changelog for the CVE IDs, or switch to the nginx.org package. See "Ran the upgrade and nginx is still on 1.24.0 or 1.22.1?" for both options.
Is turning off HTTP/2 enough?
No. It blocks the two HTTP/2 flaws, but not the HTTP/3 flaw (turn off QUIC listeners too), the TLS 1.3 loop, the resolver flaw or the mail proxy flaw. It is a stopgap until you upgrade, not a fix.
Are these flaws being exploited?
At the time of writing, the researchers had shown working attacks in their labs but had not published exploit code, and no attacks in the wild had been reported. Flaws like this are usually attacked quickly once details spread, so patch now.
