Fixing it on shared hosting
On shared hosting, PHP is installed and updated by the provider and you have no access to the server. Your job is therefore to verify: you need to know that the fix has actually been applied.
- Check the PHP version. In cPanel, look at “Select PHP Version” or the MultiPHP Manager; in Plesk, the domain's “PHP Settings” page shows it. In WordPress: Tools › Site Health › Info › Server.
- Do not rely on the number alone. If it is 8.5.12, 8.4.27, 8.3.36, 8.2.35 or later, you are fine. If it is lower, it may still be patched (the provider may have backported the fix), but only the provider can confirm that.
- Ask for a written answer. You can send the email below to customer support as it is.
- Switch if you are on PHP 8.1 or older. Take a backup, try it on a staging site if you can, then set the version to 8.3 or 8.4 in the control panel. Old themes and plugins may break on a newer PHP, so afterwards check your important pages, and the checkout if you run an online store.
- If the site is behind Cloudflare, Cloudflare has rolled out its PHP2Shell rules on the free plan too. On Pro or higher plans, make sure the managed rules block rather than only log. This is an additional layer, not a fix: the filter can be bypassed through the server's direct IP address.
Sample email to your hosting providerSubject: PHP security update (CVE-2026-109618, CVE-2026-109619)
Dear Support Team, which PHP version is running on the hosting account for [domain]? Have you installed the PHP security release of 21 October 2026 (CVE-2026-109618, CVE-2026-109619)? If the fix was applied without a change in the version number, please confirm in writing which package build contains it. Thank you.
What does not help on shared hosting: updating WordPress, the theme or the plugins (a good thing to do, but it does not fix this flaw), installing a security plugin, or disabling session.upload_progress.enabled in a .user.ini file or with ini_set(). According to the PHP advisory, these settings take effect too late: by then the attacker's request has already been processed.
Fixing it on a VPS and in containers
On your own server the whole chain is yours: updating the package, replacing the running processes and verifying the result. Skipping the restart is the most common mistake: the patched PHP is already on disk, but the old one is still running in memory.
Ubuntu and Debian
First find out which PHP serves the web. php -v only shows the command-line build; websites often run on a different branch.
# Which PHP-FPM branches are running? (the socket name tells you)
ls /run/php/
php-fpm8.3 -v
# Update (from the ondrej/sury repository and from the Ubuntu or Debian security repository alike)
sudo apt update
apt list --upgradable 2>/dev/null | grep php
sudo apt upgrade
# Restart: every running PHP branch and Apache (if you use mod_php)
sudo systemctl restart php8.3-fpm
sudo systemctl restart apache2
# Verify: the version and the start time (it must be after the update)
php-fpm8.3 -v
systemctl show -p ActiveEnterTimestamp php8.3-fpm
If you use the distribution's own PHP package (Ubuntu 24.04, for example, ships 8.3 by default), the version number does not jump to 8.3.36; the fix arrives as a backport. In that case, look for the CVE number in the package changelog:
zgrep -c "CVE-2026-109618" /usr/share/doc/php8.3-common/changelog.Debian.gz
If the result is 0, the fix is not installed yet: check whether the security repository is enabled, or wait for the distribution's announcement.
AlmaLinux, Rocky Linux, RHEL
sudo dnf upgrade 'php*'
sudo systemctl restart php-fpm httpd
php-fpm -v
With the Remi repository the procedure is the same, using package and service names such as php83-php-fpm.
Docker and containers
PHP running in a container is not updated together with the host. The image has to be rebuilt from a fresh base image and the old containers replaced. If the Dockerfile pins an exact version (for example php:8.3.35-fpm), change that too.
docker compose build --pull
docker compose up -d
docker compose exec app php -v # replace “app” with the name of your own service
If you cannot update today: temporary mitigation
The PHP team recommends a few settings for the period until the update can be scheduled. These reduce the risk but do not eliminate it. They go into php.ini or the PHP-FPM pool file, and every PHP process must be restarted afterwards.
; php.ini (.user.ini and ini_set() have no effect here)
session.upload_progress.enabled = 0
fatal_error_backtraces = 0 ; PHP 8.5 only
post_max_size = 8M ; 32M if the site has file uploads
upload_max_filesize = 6M ; always smaller than post_max_size
; PHP-FPM pool, for example /etc/php/8.3/fpm/pool.d/www.conf
php_admin_value[memory_limit] = -1
memory_limit = -1 removes the trigger of CVE-2026-109618 (hitting the memory limit), but in exchange a single request can consume the server's entire memory. Apply it only if the memory of the PHP processes is limited by something else as well (a container or cgroup limit) and the web server also filters request sizes (client_max_body_size in nginx). CERT-EU recommends temporarily taking anything that cannot be updated off the public internet, while CISA would have systems on PHP 8.1 and older switched off until they are patched.
Have you been hacked? How to check
The update rules out future attacks, not past ones. Since the first attempt took place on 17 October, it is worth reviewing the logs from that day onwards. Each sign may be harmless on its own, but together they are suspicious.
- Memory error at request start-up: an “Allowed memory size … exhausted” error that reports “Unknown on line 0” instead of your code, especially in series, within one process.
- Crashing PHP processes: SIGSEGV (signal 11) in the PHP-FPM log.
- Unusual requests: large, chunked POST requests with many or deeply nested form fields, and a long or binary PHPSESSID cookie.
- A shell running as the web server: if the
www-data user runs sh, bash, curl or wget, that is the strongest sign. - New PHP files in upload and cache folders, where there should be no PHP at all.
# Memory error at request start-up
sudo zgrep -h "Allowed memory size" /var/log/php*-fpm.log* /var/log/nginx/error.log* | grep -c "Unknown on line 0"
# Crashing PHP processes
sudo zgrep -h "SIGSEGV" /var/log/php*-fpm.log*
# Is the web server user running a shell or a downloader?
ps -u www-data -o pid,lstart,cmd | grep -E "sh -c|bash|curl|wget"
# PHP files created in upload folders since 17 October
sudo find /var/www -path "*upload*" -name "*.php" -newermt 2026-10-17
If any of these returns a suspicious hit, do not clean up straight away, because you would destroy the evidence along with the traces. Save the logs, replace every password and key (database, email, payment and API keys), and restore the site from a clean backup taken before 17 October. It is worth calling in a specialist for this.