HIGH
smb OplockBreak Leak
CVE-2026-97557
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
KernelScan AI7.5HIGH
01Description
In the Linux kernel, the following vulnerability has been resolved: smb: client: avoid leaking refcount in cifs_queue_oplock_break() cifs_queue_oplock_break() unconditionally takes a reference on the target file before queueing cifs_oplock_break(). Only that work item decreases the reference counter again. If another oplock break arrives while that work is still queued, queue_work() will return false and not queue this second work item. As a result, we will never reach the point to drop the file reference again and are leaking this reference. This can be triggered when interacting with a slow-responding server. As a result, later unmount operations for this file system will fail with BUG: Dentry ... still in use (1) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) kernel BUG at fs/super.c:777! Fix this by only incrementing the reference count if the work has been queued successfully. Taking it after queue_work() is safe because all three callers hold tcon->open_file_lock across the call and _cifsFileInfo_put() decrements under that same lock, so a worker that starts the handler in the window cannot drop the reference before it has been taken.
02KernelScan AI Analysis
Risk summary
A reference-count leak in the CIFS/SMB client's oplock break handler allows a malicious or slow SMB server to prevent filesystem unmount by sending overlapping oplock break notifications. Each leaked reference pins an inode indefinitely, eventually triggering a kernel BUG when the admin attempts to unmount. Any system that mounts SMB/CIFS shares and is network-connected to an untrusted or compromised server is at risk.
Vulnerability analysis
The CIFS client's oplock break queuing function unconditionally increments a reference count on the target file handle before attempting to queue the break handler. When a second oplock break arrives while the handler is still queued, the queueing call is not performed (the work item is already pending) but the extra reference was already taken and is never released — only the handler itself drops it. This leaked reference keeps the file inode alive, causing subsequent unmount attempts to fail with a kernel BUG assertion. The fix moves the reference increment to after the successful queueing call, so the count is only raised when the handler will actually run and eventually drop it. The vulnerability is reachable from the network: the trigger is a remote SMB server (or a man-in-the-middle on the SMB connection) sending oplock break notifications faster than the client processes them. No local privileges are required on the client; only an existing CIFS mount and network connectivity to the server.
Lifecycle
03Fix Versions
| Branch | Introduced | Fixed in | Patch commit |
|---|---|---|---|
| 3.16 | 3.16.72 | 3.17 | 735a3a610136 |
| 4.9 | 4.9.171 | 4.10 | de2a6bcff265 |
| 4.14 | 4.14.114 | 4.15 | af0193bbf1ac |
| 4.19 | 4.19.37 | 4.20 | 2ed2c29dd959 |
| 5.0 | 5.0.10 | 5.1 | dfe7b750c7e0 |
| 6.1 | — | 6.1.189 | 9f2e63f1b2d5 |
| 6.6 | — | 6.6.158 | — |
| 6.12 | — | 6.12.111 | — |
| 6.18 | — | 6.18.53 | — |
| 7.2 | — | 7.2.7 | — |
| mainline | — | 7.3-rc3 | — |
| cip/6.12 | — | 6.12.111-cip32 | af0193bbf1ac |