Two remote code execution vulnerabilities in KSMBD, the SMB server built into the Linux kernel, were published to public mailing lists this week with no fix yet merged into the mainline kernel. Both carry a CVSS score of 9.0, according to Trend Micro's Zero Day Initiative (ZDI), which filed the reports.
The more dangerous of the pair, tracked as ZDI-CAN-33774, can be exploited without any network authentication. Any host running KSMBD on port 445 is exposed. The defect is a race in the way the server assigns its handler table: a burst of short SMB1 packets followed by an SMB2 packet inside the same TCP write can cause worker threads to process SMB1 data with an SMB2 handler, reading beyond the end of the smaller buffer. A route to writing past the buffer also opens up once an attacker has access to a .
The second flaw, ZDI-CAN-33549, is an integer underflow in the SMB Direct path. A client can send RDMA buffer descriptors in which the first is longer than the declared payload, wrapping the server's size counter and letting a later memory copy run past the end of an allocated buffer. Exploiting it takes three conditions: an RDMA-capable interface, either hardware or the rxe software driver; a writable KSMBD ; and a client that negotiates SMB Direct.
Neither issue has a named fix in mainline or in linux- , the integration tree that feeds it, as of Thursday. Linux news site Phoronix reported that SMB Direct maintainer Stefan Metzmacher has submitted a patch for ZDI-CAN-33549 aimed at the development cycle, though no other outlet has confirmed it. Mainline fixes typically take weeks to reach stable branches and the kernels shipped by distributions. Neither flaw has a CVE identifier, because the kernel project issues those only after a stable-tree fix exists.
Administrators can check whether the module is loaded with systemctl status ksmbd or lsmod | grep ksmbd . Red Hat Enterprise Linux does not ship KSMBD by default, according to Red Hat's support documentation.
The disclosure drew a sharp response from stable-kernel maintainer Greg Kroah-Hartman, who answered three of the five kernel reports within minutes of one another. On the VMware GPU driver report he wrote: "For some reason you sent this to public lists, and did not include a patch, so there's not much we can do with this." Seventeen seconds later, on the NVMe driver report: "Again, no patch, and sent to a public list? Please fix your submission process."
ZDI credited TrendAI, Trend Micro's AI scanner, as the finder. The kernel's security documentation, revised in May 2026 after guidance from Linus Torvalds, states that bugs surfaced by AI tools are by definition already public and should go to subsystem developer lists. The reports followed that logic, copying public lists alongside the private kernel security address but attaching no patches.
Three of the five kernel reports have nothing to do with SMB. ZDI-CAN-33367, found by researcher p1r0x, is a local privilege escalation in the networking stack that requires local access. ZDI-CAN-34068 is a local privilege escalation in the VMware virtual GPU driver, and ZDI-CAN-34076 is an information disclosure in the NVMe target driver. A sixth report, ZDI-CAN-32005, covers a buffer overflow in ConnMan, a network daemon that sits outside the kernel.
Note: Severity figures reflect the ZDI reports as published; only the two KSMBD issues were assigned CVSS 9.0.
The exposure is concentrated among file servers that have enabled KSMBD, which is not part of a default Red Hat installation but is available in mainline kernels and many distribution builds. The window matters most for internet-facing SMB endpoints, where the pre-authentication flaw requires no credentials at all. Until patches land in stable and vendor kernels, the practical mitigations are to disable or unload the module where it is not needed, restrict port 445 to trusted networks, and avoid enabling SMB Direct on hosts that do not require RDMA transport.
The episode also highlights a procedural fight inside kernel development. Public disclosure without an accompanying patch leaves maintainers with little to act on and gives attackers a head start, while the project's own policy pushes AI-discovered bugs toward open lists. The two KSMBD reports now sit in that gap, rated critical and still unfixed upstream.
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
