CVE-2025-66516 PDF Form Extraction
How the CVE-2025-66516 patch changes PDF form extraction in NES for Apache Solr 8.11.x, and how to opt back in where form text is needed.
CVE-2025-66516 PDF Form Extraction
CVE-2025-66516 is an XML external entity (XXE) flaw in the way Apache Tika parses the XFA form inside a PDF. Solr reaches it through Solr Cell (ExtractingRequestHandler) and the DataImportHandler's TikaEntityProcessor.
NES for Apache Solr 8.11.x 8.11.4-solr-8.11.7 adopts the upstream fix, SOLR-17888: Tika no longer parses AcroForm or XFA form content in PDFs unless extractAcroFormContent is set explicitly. Upgrading needs no code or configuration change, but it does change what gets indexed.
What Changes
Solr starts, uploads and imports succeed, and page text and metadata are indexed as before. What changes is the extracted text:
- Filled-in form field values are no longer indexed. A search for a value someone typed into a PDF form stops matching documents indexed after the upgrade.
- Existing documents keep their form text until they are reindexed, so the index is mixed until then.
- XFA-only extraction is turned off too.
ifXFAExtractOnlyXFA=trueis ignored unlessextractAcroFormContent=trueis also set, so XFA PDFs on those handlers now index their page text instead of their form data. - Nothing is logged when form text is dropped.
| Configuration | After the upgrade |
|---|---|
No tika.config or parseContext.config (the Reference Guide setup) | Form text is no longer indexed |
tika.config with PDFParser params that don't set extractAcroFormContent | Form text is no longer indexed; the other PDFParser params still apply |
parseContext.config with a PDFParserConfig entry that doesn't set extractAcroFormContent | Form text is no longer indexed; the other properties still apply |
stream.type=application/pdf (Solr Cell), or a DIH entity with a parser attribute | Form text is no longer indexed |
extractAcroFormContent=true set explicitly | Unchanged: form text is indexed |
Body text, metadata, non-PDF formats, schemas and APIs are not affected.
Opting Back In
Set extractAcroFormContent explicitly on each handler or entity that needs form text. Do this only where the PDFs come from sources you trust: the opt-in turns XFA parsing back on, and with it the exposure this CVE describes. Leave public uploads and shared folders on the default.
Solr logs a warning when a handler opts in:
solr-1 | 2026-11-18 22:19:24.366 WARN (xxxxxxxxxx-25) [ x:documents] o.a.s.h.e.ParseContextConfig extractAcroFormContent=true is explicitly set, which may be vulnerable to CVE-2025-54988 XXE attacks. Ensure you trust all PDF sources or disable XFA parsing.
In a handler's tika.config, or a DIH entity's tikaConfig:
<parser class="org.apache.tika.parser.pdf.PDFParser">
<params>
<param name="extractAcroFormContent" type="bool">true</param>
</params>
</parser>
In a parseContext.config:
<entry class="org.apache.tika.parser.pdf.PDFParserConfig" impl="org.apache.tika.parser.pdf.PDFParserConfig">
<property name="extractAcroFormContent" value="true"/>
</entry>
The setting has no effect on releases without the patch, so it can be deployed before the upgrade.
Upgrade Checklist
- List the handlers and DIH entities that index PDFs, and decide which need form text.
- On those that do, and that receive only trusted PDFs, set
extractAcroFormContent=trueas shown above. - Upgrade to
8.11.4-solr-8.11.7. - Reindex PDFs on the routes that no longer extract form text, if stored documents should match the new behavior.
Additional Information
You can read more about the CVE and its implications in the CVE-2025-66516 vulnerability entry.
- Apache Solr security news: CVE-2025-66516
- Solr 8.11 Reference Guide: Uploading Data with Solr Cell using Apache Tika