[php-src] Issue #24657: Log Security-Relevant Request Rejections and Mitigated Exploit Attempts

Issue: https://github.com/php/php-src/issues/24657 Author: hackerfactor ### Description ### Problem PHP can encounter malicious or malformed requests that trigger known vulnerabilities or exercise security-sensitive code paths. Fixes may prevent memory corruption, stale-state reuse, or other unintended behavior, but the administrator often receives no reliable indication that an attack was attempted. The PHP script cannot distinguish a valid request from a mitigated attack. CVE-2026-109618 and CVE-2026-109619 demonstrate why this matters. A patched PHP runtime may prevent the vulnerable behavior. However, it provides no clear indication that the request triggered a security-relevant condition. There are a few examples where mitigated attacks are logged. For example: - "PHP message: PHP Warning: Missing boundary in multipart/form-data POST data in Unknown on line 0" - "PHP message: PHP Warning: PHP Request Startup: Input variables exceeded 1000. To increase the limit change max_input_vars in php.ini. in Unknown on line 0" The text in Unknown on line 0 is a great indicator of a mitigated attack. Unfortunately, the running PHP script cannot readily determine if the error message applies to the current request. When no diagnostic is emitted, administrators are left to infer what happened from web-server access logs, worker failures, and other indirect evidence. An application script cannot reliably report conditions that occur before it starts executing. ### Proposed enhancement Provide a consistent mechanism for PHP to log security-relevant events when it detects and rejects, mitigates, or safely terminates processing of a potentially malicious request. Allow this information to be retrieved by the running application. I've attached a working proof of concept, including source code that compiles into a PHP .so extension. It checks for errors in the current worker thread and exposes a queryable function that returns an array of potential concerns to the running PHP script. This demonstrates one way PHP could make otherwise inaccessible diagnostic information available to applications. The mechanism should be extended to cover other anomalies detected by the PHP engine. There should be a general mechanism rather than a collection of CVE-specific messages. Potential event categories include: - Malformed or suspicious multipart/form-data requests. - Requests rejected because they violate parser or protocol constraints. - Security-sensitive error and bailout paths that are safely handled by the runtime. - Known exploit conditions that are detected and prevented by a security fix. - Other runtime-detected conditions that could indicate an exploitation attempt. The goal is not to label every malformed request as an attack. Log messages should describe the condition detected and the action PHP took, without asserting malicious intent when that cannot be established. The running PHP script can use this information to determine whether the request indicates a possible attack. ### Logging requirements Ideally, security-relevant events would: 1. Be available to the system administrator even when the application script never starts or terminates abnormally. 2. Be written to a configurable error log or dedicated security log. 3. Be accessible to the running PHP script. 4. Identify the event category, the mitigation or rejection action, and the relevant request or worker context. 5. Avoid logging sensitive request bodies, credentials, or other unnecessary user data. 6. Be configurable to control verbosity and prevent log flooding from repeated attacks. 7. Use consistent, documented event identifiers so administrators can monitor and correlate events across PHP versions. When PHP cannot safely emit a diagnostic, the documentation should explicitly identify that limitation and describe any available alternative. ### Motivation Security patches should prevent exploitation, but administrators also need visibility into attacks against their systems. Without logging, a successfully mitigated attack can be indistinguishable from an ordinary request. This information is useful for incident response, identifying targeted systems, measuring attack activity, verifying that mitigations are being exercised, and correlating events with web-server logs. The proposed feature would improve PHP's security observability across multiple vulnerabilities, rather than addressing only CVE-2026-109618 and CVE-2026-109619.