Skip to main content
Mallory
HighPublic exploitVulnerability

CVE-2026-105437: Heap Buffer Overflow in Git xdiff Three-Way Merging

Published Nov 10, 2026 · Updated today
Trending this week#33of 427,813 vulnerabilities
Mention activity · 30 days45 mentions · peak 17 in a dayToday
CVSS7.5High · v3.1
EPSS0.40%35th percentile
CISA KEVNoNot listed
In the wildNonePublic exploit, no exploitation reported
Vendor
Git-Scm
IdentifiersCVE-2026-105437CWE-122 · Heap-based Buffer Overflow

CVE-2026-105437 is a heap-based buffer overflow in Git's xdiff textual three-way merge machinery. The xdl_fill_merge_buffer() sizing pass accumulates the required merge-output size in a 32-bit signed integer before allocating the output buffer. An attacker-controlled repository can combine an extremely large positive conflict-marker-size attribute with many separate merge conflicts, generating output exceeding 4 GiB from relatively small inputs. When the size wraps to a small positive value, Git allocates an undersized buffer and writes largely attacker-controlled output outside its bounds, corrupting the heap. Exposed commands include git merge, git rebase, git cherry-pick, and git merge-tree. Git hosting services performing automated mergeability checks are particularly exposed when they obtain attributes from untrusted repository trees.

Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Analyst brief

Impact, mitigation & remediation

What it means and what to do now: patch path, interim mitigations, and the assume-compromise checklist.

Impact

What an attacker gets, and what they’ve been doing with it.

Successful exploitation can crash Git and disrupt local merge operations or server-side mergeability processing. Out-of-bounds writes can occur both beyond the allocation and before its beginning. Because the overflow contents are largely attacker-controlled, more serious memory corruption and potentially code execution in the Git process are possible, including processes spawned by GitLab Gitaly. Reliable code execution is not established.

Mitigation

If you can’t patch tonight, do this now.

If upgrading is temporarily impossible, add '* conflict-marker-size=7' to each repository's $GIT_DIR/info/attributes to override attacker-controlled in-tree attributes. Alternatively, disable server-side mergeability checks for untrusted repositories. Avoid textual three-way merges of untrusted repository content until a fix or effective attribute override is deployed. These workarounds do not repair the underlying arithmetic flaw.

Remediation

Patch, then assume compromise.

Upgrade to Git 2.56.1 or a patched release for the installed maintenance line: 2.50.2, 2.51.3, 2.52.1, 2.53.1, 2.54.1, or 2.55.1. Vendor-maintained packages may contain backported fixes, so verify the distribution's security update rather than relying solely on the upstream version number. Update applications and automation services that bundle Git. For affected GitLab installations, install 19.3.8, 19.4.6, or 19.5.4, as appropriate for the release branch. The upstream fix uses size_t for merge-output calculations, checks sizing additions for overflow, and rejects results exceeding the representable output-size limit.

Public exploits

Exploits

1 exploit validated by Mallory's analysis. Hidden: 1 rejected by Mallory's analysis as not a working exploit.

View more in app →
CVE-2026-105437This standalone repository supplies a working crash-oriented exploit and a reproducible Git-hosting simulation for the claimed CVE-2026-105437. Its 24 files comprise seven Python/C/shell source files, a Makefile, two Containerfiles, documentation and metadata, and ten evidence logs. The main client, poc/poc.py, generates malicious repository history and exposes check, exploit, push, mitigation, and residual-probing commands. The payload combines an oversized conflict-marker-size attribute with hundreds of conflicts so xdiff/xmerge.c's 32-bit output-size accumulator wraps before allocation. Supplied logs show a Git 2.56.0 segmentation fault and an ASAN-confirmed out-of-bounds write, while Git 2.56.1 completes without the same corruption. These are supplied artifacts, not independently reproduced results. lab/ implements a custom Gitaly-like HTTP service and Git daemon rather than a full GitLab deployment. Its container builds vulnerable, fixed, and ASAN variants. Makefile and run-lab.sh orchestrate Podman-based comparisons and evidence collection. poc/fuzz_xmerge.c directly exercises xdl_merge(); poc/hunt_0day.py compiles and drives a sanitizer parameter matrix; poc/probe_marker_size.c reproduces neighboring marker-size arithmetic in an isolated harness. The normal PoC image lacks the compiler and Git source trees needed for its standalone 0day command, whereas the Makefile's dedicated sweep uses the larger lab image. The sweep also labels cases without selected sanitizer reports as OK, including timeouts, so it cannot establish comprehensive safety. The mitigation evidence distinguishes a bare 'conflict-marker-size=7' line, which does not override the malicious attribute, from a path-qualified override, which does. Documentation identifies checked size_t arithmetic as the upstream fix and notes that correctly sized multi-gigabyte output can still exhaust resources. No demonstrated RCE or newly confirmed memory-safety vulnerability in the fixed build is present; the repository's potential-RCE language exceeds its demonstrated crash capability. Network names are lab targets, with no identified exfiltration or command-and-control service. Parts of README.md and poc/poc.py were truncated in the supplied material. Repository ref and archive size were not supplied; size_bytes=0 denotes unknown.googlesky · Disclosed Nov 11, 2026 · python · c · network · file · localOperational
Exposure surface

Affected products & vendors

Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.

VendorProductType
Git-ScmGitapplication
FAQ

Frequently asked questions

Is CVE-2026-105437 exploited in the wild?
As of Nov 17, 2026, Mallory has no reports of CVE-2026-105437 being exploited in the wild. It is not listed in CISA KEV.
Is there a public exploit for CVE-2026-105437?
Yes. Mallory tracks 1 public exploit or proof of concept for CVE-2026-105437, the earliest disclosed Nov 11, 2026.
Is there a patch for CVE-2026-105437?
Upgrade to Git 2.56.1 or a patched release for the installed maintenance line: 2.50.2, 2.51.3, 2.52.1, 2.53.1, 2.54.1, or 2.55.1. Vendor-maintained packages may contain backported fixes, so verify the distribution's security update rather than relying solely on the upstream version number.
What products are affected by CVE-2026-105437?
Git-Scm Git.
How severe is CVE-2026-105437?
CVE-2026-105437 has a CVSS v3.1 score of 7.5 (High) and an EPSS score of 0.40% (35th percentile) as of Nov 17, 2026.
Activity feed

Recent activity

45 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.

View more in app →
+21 more in the timeline. Sign up to see them
Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Mallory platform

The version that knows your environment.

This page is what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which adversaries are exploiting it right now, and which detections to deploy tonight.
Exposure mappingQuery your assets running an affected version, and investigate the blast radius.
Threat actor evidenceEvery report linking this CVE to a named adversary.
Associated malwareMalware families linked to this CVE, with evidence and IOCs.
Detection signaturesYARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Vendor-by-vendor mappingCross-references every affected SKU, including bundled OEM variants.
Social activity19Community discussion across Reddit, Mastodon, and other social sources.