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-2026-110752 Overview
CVE-2026-110752 affects Text::CSV_XS versions before 1.65 for Perl. The flaw allows memory exhaustion through the fragment method, which allocates one array entry per index of an RFC 7111 selector before reading any data. An attacker who supplies a crafted row=, col=, or cell= selector such as row=1-400000000 triggers allocation of hundreds of millions of elements. The process terminates with an uncatchable Out of memory error that eval cannot trap. Applications passing request-supplied fragments to the fragment or csv interfaces are exposed to remote denial of service [CWE-770: Allocation of Resources Without Limits].
Critical Impact
Unauthenticated remote attackers can crash Perl applications processing CSV data by submitting a fragment selector with large integer bounds, with no authentication or user interaction required.
Affected Products
- Text::CSV_XS for Perl, all versions prior to 1.65
- Perl applications that pass untrusted fragment selectors to the fragment() method
- Perl applications that invoke the csv interface with user-controlled fragment arguments
Discovery Timeline
- 2026-10-16 - CVE-2026-110752 published to NVD
- 2026-10-20 - Last updated in NVD database
Technical Details for CVE-2026-110752
Vulnerability Analysis
The fragment method in Text::CSV_XS implements RFC 7111, which defines URI fragment selectors for CSV documents. The method parses row=, col=, and cell= selectors and validates only their syntax and range order. It does not validate that selector bounds correspond to actual data size.
For each selector range, the implementation expands the integer bounds into an in-memory array with one element per selected index. A selector like row=1-400000000 applied to a two-line file allocates 400 million slots before any CSV row is parsed. The allocation occurs eagerly, driven solely by the selector integers rather than by data content.
Additionally, an explicit end column in a cell= selector is not clamped against the actual row width. This produces the same unbounded allocation pattern for column selectors that reference positions beyond the row.
Root Cause
The root cause is missing resource bounding in the fragment expansion logic. The code sized its internal structures using the raw selector integers rather than clamping them against the data being processed. Because the allocation failure manifests as a Perl Out of memory fatal error, calling code cannot recover using eval blocks.
Attack Vector
Exploitation requires only the ability to supply the fragment parameter to an application that forwards it to Text::CSV_XS. Web services that expose CSV filtering with user-supplied RFC 7111 selectors are directly reachable over the network without authentication. A single HTTP request containing a selector with a large upper bound is sufficient to terminate the worker process.
# Fragment of the official patch - CSV_XS.pm
# Source: https://github.com/cpan-authors/Text-CSV_XS/commit/d9d3d66f9ff6a8a1bc255a6ef05100f9b102b92b.patch
my ($tlr, $tlc, $brr, $brc) = @{$s};
$r < $tlr || ($brr ne "*" && $r > $brr) and next;
!defined $lc || $tlc < $lc and $lc = $tlc;
- my $rr = $brc eq "*" ? $#{$row} : $brc;
+ # Do not pad past the data: an explicit column beyond the
+ # row would otherwise allocate one slot per selector integer
+ my $lr = $#{$row} > $tlc ? $#{$row} : $tlc;
+ my $rr = $brc eq "*" ? $#{$row} : $brc > $lr ? $lr : $brc;
$row{$_} = $row->[$_] for $tlc .. $rr;
The patch clamps $brc (bottom-right column) against the actual row width $lr, preventing allocation of slots for indices that do not exist in the data. See the GitHub Commit Patch for the full diff.
Detection Methods for CVE-2026-110752
Indicators of Compromise
- Perl worker processes terminating with Out of memory! messages in application or system logs
- HTTP request logs containing fragment=row=, fragment=col=, or fragment=cell= parameters with large numeric upper bounds
- Sudden spikes in resident set size (RSS) of Perl processes immediately before process death
- Repeated restarts of Perl-based web workers (FastCGI, PSGI, mod_perl) under PID supervisors
Detection Strategies
- Inspect query strings and POST bodies for RFC 7111 fragment selectors with ranges exceeding reasonable row or column counts
- Correlate OOM-killer events in dmesg or /var/log/messages with the Perl process cgroup or user
- Audit application code for calls to Text::CSV_XS::fragment() and csv(..., fragment => ...) that pass unsanitized input
Monitoring Recommendations
- Alert on process memory growth exceeding configured limits for CSV-handling services
- Monitor HTTP 5xx error rates and worker restart counts on endpoints that accept CSV fragment parameters
- Centralize Perl warn and die output to a log aggregation layer to surface recurring Out of memory failures
How to Mitigate CVE-2026-110752
Immediate Actions Required
- Upgrade Text::CSV_XS to version 1.65 or later from CPAN
- Audit every call site that forwards HTTP-supplied data to the fragment argument of Text::CSV_XS
- Reject or clamp fragment selectors in application code before they reach the module
Patch Information
The fix is included in Text::CSV_XS 1.65, released 2026 by H.Merijn Brand. The ChangeLog entry reads fragment (): do not size memory by the selector integers. See the MetaCPAN Release Changes and the GitHub Commit Patch.
Workarounds
- Strip or reject the fragment parameter at the web tier until the upgrade is deployed
- Validate selector integers against the known maximum row and column counts of expected input
- Enforce per-process memory limits using ulimit -v, systemd MemoryMax=, or container cgroup limits to bound damage
# Upgrade Text::CSV_XS to the patched release
cpanm Text::CSV_XS@1.65
# Verify the installed version
perl -MText::CSV_XS -E 'say Text::CSV_XS->VERSION'
# Enforce a per-process memory ceiling for the Perl worker (example: 512 MB virtual)
ulimit -v 524288
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

