Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

PHP Nov 19 Security Update (18 CVEs)

neighhouseneighhouse Member
edited 4:50PM in Security

EDIT / UPDATE: PHP's full November 19, 2026 security release just dropped and it is way bigger than just phpAfterShell. Today's patch fixes 18 CVEs in total (including 7 critical RCEs) across all supported branches. While phpAfterShell (CVE-2026-126714) still has friction, CVE-2026-129289 is a single-request unauthenticated RCE if you have persistent FastCGI connections enabled.

Full advisories:

  • https://www.aikido.dev/blog/php-security-release-19-november-2026-cve-2026-115560
  • https://depthfirst.com/research/phpaftershell

  • CVE-2026-115560 (CVSS 9.2 - Session Serializer Mismatch RCE): Advisory expanded scope beyond ini_set() handler switching. Any separate apps, FPM pools, vhosts, or CLI cron jobs sharing the same session.save_path or session backend with different session.serialize_handler settings are exposed too.

  • CVE-2026-126714 (CVSS 9.2 - phpAfterShell): Non-atomic pwrite()/ftruncate() session file writes during RFC1867 upload progress allow an interrupted worker to splice old and new session data for deserialization RCE.
  • CVE-2026-129481, CVE-2026-129474, CVE-2026-129182 (CVSS 9.2): Use-after-free RCEs in interrupted session shutdown and error/magic-method backtraces.
  • CVE-2026-129496 (CVSS 9.2): Timezone cache use-after-free RCE triggered when startup diagnostics are written to an error log file.
  • CVE-2026-129291 (CVSS 8.8): PHP-FPM epoll use-after-free local privilege escalation (FPM worker to master/root).
  • CVE-2026-128090 (CVSS 7.7): PHAR OpenSSL signature verification bypass via algorithm downgrade.

CVE-2026-129289 (CVSS 9.2 - PHP-FPM Request Smuggling RCE)

If your front-end web server reuses persistent backend connections to PHP-FPM (or LSAPI) and an endpoint opens php://input without reading the full request body to EOF (or aborts/times out early), PHP fails to drain the leftover bytes before reusing the socket. An attacker can stuff a forged second FastCGI request into those unread body bytes, bypass reverse-proxy routing, and inject PHP_VALUE / auto_prepend_file for instant RCE as the FPM worker. No race condition or gadget chain required.

Who is safe by default:

  • Standard NGINX and Apache configs are not affected (fastcgi_keep_conn and enablereuse both default to off).

Who is exposed:

Anyone using performance-tuned templates, control panels, or load balancers with persistent FPM/LSAPI pooling turned on:

  • NGINX: fastcgi_keep_conn on;
  • Apache (mod_proxy_fcgi): enablereuse=on
  • HAProxy / LiteSpeed (LSAPI): Persistent backend connection reuse enabled

Quick check on your boxes:

grep -rnE "fastcgi_keep_conn|enablereuse" /etc/nginx/ /etc/apache2/ /etc/httpd/ 2>/dev/null

Comments

  • rpqurpqu Member
    edited 4:38PM

    fuck
    finally

    Thanked by 1neighhouse
  • nemnem Member, Host Rep

    There's no good use case for keeping session.upload_progress.enabled on in CGI environments.

  • @rpqu said:
    fuck
    finally

    I apologize, just updated the post, there are more than one CVEs

  • @nem said:
    There's no good use case for keeping session.upload_progress.enabled on in CGI environments.

    Just updated the post :cry:

  • nemnem Member, Host Rep

    @neighhouse said:

    @nem said:
    There's no good use case for keeping session.upload_progress.enabled on in CGI environments.

    Just updated the post :cry:

    Apache (mod_proxy_fcgi): enablereuse=on

    For that matter, I found no improvement in SO_REUSEADDR in organic tests. I bitched about folks implementing this because (a) you're trying to outsmart the neckbeards and (b) they smell worse than you, so yield accordingly.

Sign In or Register to comment.