Loading vulnerability…
CVE-2021-26708 is a race-condition vulnerability in the Linux kernel VMware VSOCK implementation, in the AF_VSOCK code path. Incorrect locking introduced with VSOCK multi-transport support permits concurrent operations on shared socket state to race. The flaw affects Linux kernels before 5.10.13 and can be exploited by a local low-privileged attacker for privilege escalation. Public research has demonstrated that exploitation can bypass SMEP and SMAP protections.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos (1 hidden).
Small 4-file repository centered on a single C proof-of-concept (`poc.c`) for CVE-2021-26708, a Linux kernel AF_VSOCK use-after-free caused by a race condition. Supporting files are documentation (`README.md`), a captured KASAN crash log (`kasan_report.txt`), and the MIT license. The PoC is a local kernel bug trigger rather than a full privilege-escalation exploit. `poc.c` implements the exploit logic. It creates AF_VSOCK sockets, optionally starts a local VSOCK listener on `VMADDR_CID_LOCAL:1234` (`-load` mode), then in normal mode orchestrates a race using three threads: a `writer` thread calling `setsockopt(... SO_VM_SOCKETS_BUFFER_SIZE ...)`, a `connecter` thread issuing a concurrent `connect()` to `VMADDR_CID_HOST+3:5678`, and a `gater` thread calling `setsockopt` with a pointer into a userfaultfd-managed page. The userfaultfd mechanism intentionally blocks kernel access to the setsockopt argument until the operator presses `g`, widening the race window so the concurrent connect and buffer-size update paths overlap. This structure is consistent with a deterministic crash/reproduction PoC for a kernel race. The included `kasan_report.txt` corroborates successful triggering of the vulnerability, showing a KASAN use-after-free in `virtio_transport_notify_buffer_size`, with allocation and free traces through `vsock_stream_connect` and related AF_VSOCK code paths. No shellcode, command execution, or post-exploitation payload is present. The repository’s purpose is vulnerability reproduction and analysis for Linux kernel 5.10.12 in a QEMU/KVM vsock-enabled environment, not weaponized exploitation.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
9 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
A Linux kernel race-condition vulnerability due to improper locking in net/vmw_vsock/af_vsock.c.
A Linux kernel race-condition vulnerability caused by incorrect locking in net/vmw_vsock/af_vsock.c.
A previously disclosed Linux kernel vulnerability referenced as an exploitation-related example.
A Linux kernel vulnerability referenced as prior work demonstrating arbitrary read via corruption of msg_msg metadata fields.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.