HornbillTechnology
Blog

A WordPress update will not fix PHP2Shell

On 21 October 2026, PHP disclosed two critical flaws that can run code on a server before WordPress starts. What to check this week, and when a move off PHP is the saner plan.

On 21 October 2026, the PHP project shipped an emergency security release. Two flaws in PHP itself, CVE-2026-109618 and CVE-2026-109619, now called PHP2Shell, let someone on the internet run code on a server or read what that server is holding in memory. No login is required.

WordPress.org said the same day that no WordPress update is planned, because WordPress is not where the bug lives. A theme update will not reach it. Neither will Wordfence, a page builder, or a “security hardening” plugin. Those all run as PHP, and the flaw is in the PHP engine that starts handling the request before your site’s code does.

If you run a company site, a WooCommerce shop, or a booking site on WordPress, this week is about the PHP version. A redesign can wait. An unpatched server should not.

What actually broke

PHP still runs a large share of the web, including WordPress, WooCommerce, Drupal, Joomla, Magento, and Laravel. Researchers at Wiz showed they could take over current WordPress, Drupal, and Joomla installs, and that a plain “hello world” PHP page was enough. The application did not have to be badly written. The engine underneath it was.

Both issues are rated critical (CVSS 9.2). They are memory-safety bugs in how PHP recovers when a request runs out of memory. The practical result, confirmed by the PHP project, CISA, and Datadog:

Attempts to use the bugs were reported in the days before the public patch. CISA’s first alert said it had not yet confirmed successful break-ins. That is not a reason to wait. Flaws this widely deployed get scanned quickly, and a lot of business sites are still on PHP versions that will never receive a normal fix.

What this does to a WordPress site

WordPress is a stack. Only the top of it is yours.

LayerExamplesWho can update itDoes “Update WordPress” fix PHP2Shell?
ContentPages, products, usersYouNo
Plugins and themeWooCommerce, forms, page buildersYouNo
WordPress itselfCore 6.x or 7.xYou, or auto-updatesNo
PHP7.4, 8.1, 8.2, 8.3…Your host, or whoever runs the serverThis is the layer that has to change
ServerApache, NGINX, LiteSpeed, LinuxYour hostNo, unless they also ship a patched PHP

A shop with forty plugins and a brochure site with three plugins are exposed the same way. Plugin count does not decide whether PHP2Shell applies. It decides how painful the repair is.

If someone does get in, the damage is the usual kind, and it is ugly on a small business site: new admin users, spam pages that hijack Google rankings, skimming on a checkout, customer lists copied out, mail sent from your domain, and a host that suspends the account. Memory disclosure matters even when the front page still looks fine. Database passwords and payment keys live in that process.

There is a second, quieter impact we see more often than a dramatic hack. Sites that have been left alone for years are pinned to PHP 7.4 or 8.1 by one plugin or an old custom theme. Those branches do not have a normal security release for this bug. The choice becomes: break the plugin by moving PHP, or leave a critical hole open. That is the moment a “we will update it next quarter” plan stops working.

Check this before you plan a rebuild

In WordPress, open Tools → Site Health → Info → Server and read the PHP version. Many hosting panels show the same number.

What you seeWhat it means
8.5.12 or newer, 8.4.27 or newer, 8.3.36 or newer, or 8.2.35 or newerYou are on a PHP release that includes the fix. Confirm the host actually restarted PHP afterwards. One worker still on the old build is still exposed.
Same branch, older number (8.3.30, 8.2.20, 8.4.10)The fix is missing. Ask the host to move you to the patched release on that same branch. That jump is a security patch, and it is much less likely to break plugins than a major upgrade.
8.1, 8.0, 7.4, or anything olderThe PHP project did not ship a fixed release for these. You need a supported branch, or a host that can show you a real backport. PHP 8.2 itself only receives security fixes through 31 December 2026, so 8.3 or newer is the version worth aiming at.
A version number that looks old, but PHP comes from Ubuntu, Debian, or a host that backportsThe number may not move when the fix lands. Ask which package version you are on, and whether it includes both CVE-2026-109618 and CVE-2026-109619.

While you wait on the host:

If your host has already moved you to a fixed PHP and the site is simple, stop here. Patching is the right outcome. You do not need a new website to survive this month.

The problem that survives the patch

PHP2Shell is a bad week. It is also a picture of how these sites age.

WordPress is a reasonable way to publish. The trouble starts when the public site, the editor, the shop, the forms, and twenty other people’s plugins all share one PHP process. Every plugin is code that runs on each request. Every abandoned plugin is a reason the host cannot move you to a current PHP. The admin login, xmlrpc.php, and the user API sit on the same host as the homepage. When the engine underneath has a flaw, all of that is in scope at once, and your own code never gets a vote.

The next five years on that stack look like this: another PHP branch goes end-of-life, a page builder stops shipping updates, a payment plugin lags a year behind, and the security work becomes “please ask the host if they backported the fix.” Some teams can live with that. A lot of business sites cannot, because nobody owns the stack. A freelancer built it, a nephew picked the theme, and the hosting panel is the only documentation.

That is the migration we actually do. Not a slogan about “modern frameworks.” A site whose editor still works on Monday, and whose public pages are no longer a PHP application.

What Next.js, Astro, and Payload change

Hornbill builds custom sites on Next.js, Astro, and Payload CMS, hosted on AWS. None of those run on PHP. PHP2Shell does not apply to them, because the vulnerable interpreter is not in the request path.

That sentence has limits, and it is worth being plain about them. JavaScript has its own bugs. A stolen admin password still opens the dashboard. A careless cloud setup still leaks a database. Moving stacks is not a force field. It removes a whole class of problem: memory-corruption bugs in a C engine that parses the request before your application exists, plus the habit of bolting on features as unvetted plugins.

Here is how the three pieces split the work.

Astro is what we use when the site is mostly pages. Company sites, destination pages, articles, menus of services. Astro sends HTML and very little JavaScript, so the browser is not downloading a page builder on every visit. Content can still be edited. The public site does not have to boot a CMS on every hit. There is no wp-login on the same origin as the marketing pages unless we deliberately put an admin there.

Next.js is what we use when the site has to do something: accounts, booking, search that depends on live data, a checkout, a dashboard. Pages are rendered on the server, which is what you want for Google and for first load, and the interactive parts are React where a page actually needs them. The trekking sites we ship are this shape: trip pages, itineraries, enquiries, and payments, with the front end in Next.js.

Payload CMS is the piece people worry about losing. Editors do not want Markdown and a deploy. Payload is a TypeScript admin on Node, with drafts, live preview, media, roles, and fields shaped around the business. Trek itineraries, product specs, branch pages, staff profiles. Your team writes and publishes. They do not install plugins to get a contact form, and they do not share a login screen with every other WordPress site on a shared host.

Under that, we use PostgreSQL when the data is relational, S3 for images, and AWS for the hosting itself. Patching, backups, and monitoring sit with us, billed with clear monthly pricing, with support worldwide. The same promise we already make for hosting: you should not have to learn the panel to stay safe.

A few differences that matter on a Tuesday, not in a keynote:

What a move off WordPress looks like

We do not start by deleting the current site. The first conversation is whether you should move at all.

Stay on WordPress if the PHP version is supported, the plugin list is short, someone is actually updating it, and the site does what the business needs. We host PHP sites too. This bug is an argument for a current runtime and a person who owns it, not an argument that every WordPress install is junk.

Move if any of these are true:

A typical Hornbill migration then runs in this order.

  1. Inventory. Pages that rank, forms that receive real enquiries, payments, logins, and the plugins that are only still installed because someone was scared to remove them.
  2. Content model. Posts and pages go into Payload. Products, trips, branches, or staff become real fields, not a row of shortcodes.
  3. Front end. Astro when it is a content site. Next.js when there is booking, an account, or a workflow. The design is rebuilt to match the brand, not dropped in as a theme skin.
  4. Addresses. Old URLs redirect. Titles, descriptions, and the sitemap come with the new site, so search does not have to rediscover you from scratch.
  5. Launch on AWS. Previews before publish, media in S3, backups, and a dashboard your team can use without us on the call.

WooCommerce is the honest exception. A catalog, order history, and a payment provider are a project, not a weekend export. We will say so at the start. A five-page company site is a different job from a shop, and the quote should show that.

If this is your site

Two useful outcomes, and you can pick either.

If you only need the PHP version checked and the host pushed onto 8.2.35, 8.3.36, 8.4.27, or 8.5.12, do that today. Write to your host with the version from Site Health and those four release numbers. If they cannot move you, and the site is on PHP 8.1 or older, take it offline until they can.

If the patch is blocked by the site itself, or you already know WordPress is costing more to keep alive than it returns, talk to Hornbill. We will tell you whether to patch and stay, or to rebuild on Astro or Next.js with Payload. Either answer is a real plan. “We will look at it after Dashain” is how a site ends up on PHP 7.4 in the middle of the next advisory.