Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-110754 Overview
CVE-2026-110754 is a memory leak vulnerability in the Text::CSV_XS Perl module affecting versions before 1.65. The flaw resides in the getline function, which fails to free the result array and parsed field data when a record encounters a syntax error. Each malformed record retained in memory accumulates over the lifetime of the process. A long-running Perl service that parses attacker-controlled CSV input can exhaust available memory by receiving repeated malformed records. The weakness is classified as a Missing Release of Memory after Effective Lifetime [CWE-401].
Critical Impact
Attackers who can submit untrusted CSV data to a long-running Text::CSV_XS consumer can trigger memory exhaustion and denial of service without authentication.
Affected Products
- Text::CSV_XS for Perl versions prior to 1.65
- Perl applications and services that call getline on untrusted CSV streams
- Long-running daemons and web services that process user-supplied CSV content
Discovery Timeline
- 2026-10-16 - CVE-2026-110754 published to NVD
- 2026-10-20 - Last updated in NVD database
Technical Details for CVE-2026-110754
Vulnerability Analysis
The Text::CSV_XS XS implementation builds each parsed record by allocating a new Perl array (AV) and populating it with scalar values for each field. When the parse succeeds, getline wraps the array in a reference and returns it to the caller, transferring ownership. When the parse fails, getline returns undef and does not release the array or the field scalars that were already pushed onto it.
The leak compounds with every malformed record. A persistent worker process, API endpoint, or ETL pipeline that accepts CSV uploads will continue to grow its resident set size until the operating system terminates it or the host runs out of memory. Because the condition requires only a syntactically invalid record, triggering the leak costs the attacker almost nothing.
Root Cause
The root cause is a missing SvREFCNT_dec call on the in-progress row array in the failure path of getline inside CSV_XS.xs. Perl's reference counting relies on explicit decrements for memory not transferred to the caller. The pre-patch code only released the array when the parse completed and a reference was returned to the Perl layer.
Attack Vector
The attack vector is network-reachable in deployments where CSV data arrives from untrusted sources, such as file upload endpoints, message queues, or log ingestion pipelines. The attacker submits a stream of CSV records containing deliberate syntax errors (unterminated quotes, invalid escape sequences, or malformed field separators). Each record increases the victim process memory footprint. User interaction is required in the sense that the application must call getline on the attacker-supplied input.
}
av_push (avr, newRV_noinc ((SV *)row));
- if (n >= length && skip >= 0)
+ if (n >= length && skip >= 0) {
+ row = NULL;
break; /* We have enough */
+ }
row = newAV ();
}
+ if (row)
+ SvREFCNT_dec ((SV *)row);
while (n > length) {
SvREFCNT_dec (av_pop (avr));
n--;
Source: GitHub Commit Patch. The patch tracks the in-progress row pointer, nulls it when the row has been transferred to the result array, and calls SvREFCNT_dec on any remaining allocation before returning.
Detection Methods for CVE-2026-110754
Indicators of Compromise
- Steady, unexplained growth in resident set size (RSS) of Perl processes that call Text::CSV_XS::getline.
- Repeated out-of-memory kills or malloc failures in application logs for CSV-processing services.
- Spikes in CSV parse-error counters correlated with memory pressure on the same host.
Detection Strategies
- Inventory Perl environments and enumerate installed Text::CSV_XS versions using cpan -D Text::CSV_XS or package manager queries.
- Instrument CSV ingestion endpoints to log parse failure rates and alert on sustained high error ratios from a single source.
- Correlate process memory metrics with request logs to identify long-running consumers that grow in proportion to malformed input volume.
Monitoring Recommendations
- Add resident memory and heap-size thresholds to monitoring for any service that calls Text::CSV_XS on external data.
- Track getline return values at the application layer and record the source IP or job ID that produced malformed records.
- Enable process restart or supervisor-based memory limits (for example, systemd MemoryMax) to contain exploitation attempts while patches are deployed.
How to Mitigate CVE-2026-110754
Immediate Actions Required
- Upgrade Text::CSV_XS to version 1.65 or later on all Perl hosts that process untrusted CSV data.
- Restart long-running Perl daemons and workers after the upgrade so the patched XS module is loaded.
- Apply input rate limiting and size caps on CSV upload endpoints until patching is confirmed across the fleet.
Patch Information
The fix is included in Text::CSV_XS1.65, released by H.Merijn Brand. The upstream change log entry reads: getline (): free the result array when the parse fails. Install the update from CPAN with cpan Text::CSV_XS or through the distribution package manager. See the MetaCPAN Release Changes and the GitHub Commit Patch for the full change set.
Workarounds
- Reject or discard CSV payloads that exceed a sensible size threshold before passing them to getline.
- Run CSV parsing in short-lived worker processes so leaked memory is reclaimed on process exit.
- Enforce cgroup or systemd memory limits on affected services to force termination before host-wide exhaustion.
# Upgrade Text::CSV_XS to the patched release
cpan install Text::CSV_XS@1.65
# Verify installed version
perl -MText::CSV_XS -e 'print $Text::CSV_XS::VERSION, "\n"'
# Example systemd hardening for a Perl CSV worker
# /etc/systemd/system/csv-worker.service.d/limits.conf
[Service]
MemoryMax=512M
Restart=on-failure
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

