{"id":"CVE-2026-90236","published":"2026-09-17T17:17:19.533","lastModified":"2026-09-17T17:17:19.533","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nNFSD: Release the export reference when reaping open stateids\n\nnfs4_put_stid() releases the svc_export tracked in\nnfs4_stid.sc_export, but free_ol_stateid_reaplist() frees open and\nlock stateids by calling ->sc_free() directly, bypassing that path.\nAn open stateid takes an sc_export reference in nfs4_open() and a\nlock stateid takes its own in init_lock_stateid(); both reach\nfree_ol_stateid_reaplist() through their normal teardown, the open\nstateid via release_open_stateid() and the lock stateid via\nnfsd4_release_lockowner(), each through put_ol_stateid_locked().\nThe reference is therefore never dropped, pinning the export and\nblocking unmount for the lifetime of the stateid.\n\nRelease sc_export in free_ol_stateid_reaplist() the way\nnfs4_put_stid() does. ->sc_free() runs once per stateid, and a\nstateid reaches free_ol_stateid_reaplist() or nfs4_put_stid() but\nnever both, so the reference is dropped exactly once. Revoked\nstateids reach this path with sc_export already cleared by\ndrop_stid_export(), so they are skipped rather than double-freed.\n\nnfs4_put_stid() itself read sc_export before acquiring cl_lock.\ndrop_stid_export() clears that field and releases the reference\nunder cl_lock, so a concurrent revocation could drop the export in\nthe window between the read and the final put, releasing the same\nreference twice. Read sc_export while cl_lock is held so the two\npaths serialize and the reference is released exactly once.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/6480bd703684ed3f760e9e36c4a699033c0806ae","tags":[]},{"url":"https://git.kernel.org/stable/c/7458727fd7cb79d09e5120b6fecc1ad11cdb0946","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}