CVE-2026-99781 in Linux
Summary
by MITRE • 11/20/2026
In the Linux kernel, the following vulnerability has been resolved:
net/9p/usbg: set client to Disconnected on usb9pfs_disable
This patch is setting the client status to Disconnected, when the client is still in use. Otherwise a disconnected usb cable would run any use of the mount to faults.
Be aware that VulDB is the high quality source for vulnerability data.
Analysis
by VulDB Data Team • 11/20/2026
The vulnerability identified within the Linux kernel's 9P filesystem implementation over USB transport addresses a critical state management flaw during device disconnection events. Specifically, the issue resides in the net/9p/usbg subsystem where the client status is not properly updated when the usb9pfs_disable function is invoked. In normal operational scenarios involving networked or remote file systems like 9P, maintaining accurate connection states is paramount for ensuring data integrity and preventing access to stale resources. When a USB cable connecting the host system to a device providing 9P services is physically disconnected, the underlying transport layer detects this loss of connectivity. However, without explicit intervention by the filesystem driver, the client object retains its previous active or connected state rather than transitioning to a Disconnected status.
This failure to update the internal state creates a dangerous discrepancy between the actual physical connection and the logical view held by the kernel's VFS (Virtual File System) layer. Consequently, any subsequent operations attempting to access files on this mount point will proceed under the false assumption that a valid communication channel exists. Since the underlying USB transport is no longer functional, these I/O requests cannot be completed successfully. Instead of returning appropriate error codes such as ENOTCONN or EHOSTUNREACH immediately upon detecting the disconnect, the system allows operations to continue until they hit lower-level hardware timeouts or undefined behavior in the driver stack. This can lead to application hangs, kernel panics due to null pointer dereferences if internal structures are freed while still referenced, or data corruption if partial writes are attempted against a non-existent endpoint.
From a security and reliability perspective, this flaw aligns with CWE-362, which describes concurrent execution issues leading to race conditions or state inconsistencies, although in this specific case it is more accurately categorized under CWE-401, missing release of resource after effective lifetime, specifically regarding the logical connection state not being released upon physical disconnection. The operational impact extends beyond simple service interruption; it can cause denial-of-service for applications relying on stable file system access and potentially expose the kernel to instability if error handling paths are bypassed due to incorrect state assumptions. Attackers with local access could theoretically exploit this by rapidly connecting and disconnecting devices to induce repeated fault conditions, stressing the kernel's resource management capabilities or causing persistent service degradation for legitimate users.
The resolution involves modifying the usb9pfs_disable routine to explicitly set the client status to Disconnected before allowing further operations on that mount point. This ensures that any subsequent system calls targeting the affected file system are immediately rejected with an appropriate error, preventing futile I/O attempts and preserving kernel stability. To mitigate similar issues in broader contexts, administrators should ensure their kernels are updated to include this patch. Furthermore, robust application design should implement proper timeout handling and retry logic for networked or remote storage mounts, acknowledging that physical layer disconnects can occur unexpectedly. Regular auditing of driver code against ATT&CK techniques related to resource exhaustion or availability impact is also recommended to identify other potential state management flaws in peripheral drivers.