<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Sysadmin on kalfaoglu.net</title><link>https://blog.kalfaoglu.net/tags/sysadmin/</link><description>Recent content in Sysadmin on kalfaoglu.net</description><generator>Hugo</generator><language>en</language><lastBuildDate>Sat, 09 May 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://blog.kalfaoglu.net/tags/sysadmin/index.xml" rel="self" type="application/rss+xml"/><item><title>When AI Reads Your Database Code: MariaDB's CVE-2026-32710 and What to Do About It</title><link>https://blog.kalfaoglu.net/posts/2026-05-09-mariadb-cve-2026-32710-json-schema-en/</link><pubDate>Sat, 09 May 2026 00:00:00 +0000</pubDate><guid>https://blog.kalfaoglu.net/posts/2026-05-09-mariadb-cve-2026-32710-json-schema-en/</guid><description>&lt;p&gt;A heap buffer overflow in MariaDB&amp;rsquo;s &lt;code&gt;JSON_SCHEMA_VALID()&lt;/code&gt; function went unnoticed for years — until AI-assisted code analysis flagged it earlier this year. The flaw, now tracked as &lt;a href="https://nvd.nist.gov/vuln/detail/CVE-2026-32710"&gt;CVE-2026-32710&lt;/a&gt;, was &lt;a href="https://securityonline.info/mariadb-json-schema-validation-buffer-overflow-vulnerability-cve-2026-32710/"&gt;disclosed on March 19, 2026&lt;/a&gt;, and patches landed the same day. If your server runs MariaDB 11.4.x or 11.8.x and you haven&amp;rsquo;t applied the update yet, this is the one to prioritise this week.&lt;/p&gt;
&lt;h2 id="what-the-bug-actually-does"&gt;What the Bug Actually Does&lt;/h2&gt;
&lt;p&gt;The problem lives in &lt;code&gt;json_get_normalized_string()&lt;/code&gt; inside &lt;code&gt;sql/json_schema_helper.cc&lt;/code&gt;. The function allocates a fixed 128-byte heap buffer and then copies a JSON string value into it using &lt;code&gt;strncpy&lt;/code&gt; — without first checking whether the value fits. If an attacker crafts a JSON schema with a string field longer than that buffer, the heap overflows.&lt;/p&gt;</description></item><item><title>Postfix 3.11.2 Patches a 20-Year-Old Buffer Over-Read — and an AI Found Most of the Rest</title><link>https://blog.kalfaoglu.net/posts/2026-05-07-postfix-3112-cve-2026-43964-en/</link><pubDate>Thu, 07 May 2026 00:00:00 +0000</pubDate><guid>https://blog.kalfaoglu.net/posts/2026-05-07-postfix-3112-cve-2026-43964-en/</guid><description>&lt;p&gt;On May 4, 2026, Wietse Venema released Postfix 3.11.2, 3.10.9, 3.9.10, and 3.8.16. If you run a mail server, this is the update you actually want to read — not because the CVSS score is alarming (it isn&amp;rsquo;t), but because one of the bugs patched in this release has been sitting in the codebase since 2005.&lt;/p&gt;
&lt;h2 id="the-cve-worth-knowing-about"&gt;The CVE Worth Knowing About&lt;/h2&gt;
&lt;p&gt;&lt;a href="https://cve.threatint.eu/CVE/CVE-2026-43964"&gt;CVE-2026-43964&lt;/a&gt; is an off-by-one error in how Postfix handles enhanced status codes. If an SMTP access table, policy server, DNSBL response, or milter returns a bare status code — something like &lt;code&gt;5.7.2&lt;/code&gt; without any text following it — the daemon reads past the end of the allocated buffer. The result is a process crash.&lt;/p&gt;</description></item><item><title>Apache HTTP/2 Double-Free (CVE-2026-23918): What You Need to Do Before Friday</title><link>https://blog.kalfaoglu.net/posts/2026-05-06-apache-http2-cve-2026-23918-en/</link><pubDate>Wed, 06 May 2026 00:00:00 +0000</pubDate><guid>https://blog.kalfaoglu.net/posts/2026-05-06-apache-http2-cve-2026-23918-en/</guid><description>&lt;p&gt;A memory-corruption bug in Apache HTTP Server&amp;rsquo;s HTTP/2 implementation was publicly disclosed this week, and the details are ugly enough that you should stop reading this sentence and go check your Apache version right now. Done? Good. Let&amp;rsquo;s talk about what&amp;rsquo;s actually going on.&lt;/p&gt;
&lt;h2 id="the-bug"&gt;The Bug&lt;/h2&gt;
&lt;p&gt;&lt;a href="https://app.opencve.io/cve/CVE-2026-23918"&gt;CVE-2026-23918&lt;/a&gt; is a double-free vulnerability in &lt;code&gt;mod_http2&lt;/code&gt;, specifically in the stream cleanup path of Apache httpd 2.4.66. A double-free happens when code tries to release the same chunk of memory twice &amp;mdash; a classic mistake that corrupts internal allocator state and typically leads to crashes, and sometimes worse.&lt;/p&gt;</description></item></channel></rss>