CVE-2026-53399

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> nfsd: release layout stid on setlease failure<br /> <br /> nfs4_alloc_stid() publishes the new stid into cl-&gt;cl_stateids via<br /> idr_alloc_cyclic() under cl_lock before returning to<br /> nfsd4_alloc_layout_stateid(). When nfsd4_layout_setlease() then<br /> fails, the error path frees the layout stateid directly with<br /> kmem_cache_free() without ever calling idr_remove(), leaving the<br /> IDR slot pointing at freed slab memory. Any subsequent IDR walker<br /> (states_show, client teardown) dereferences the dangling pointer.<br /> <br /> The correct teardown for an IDR-published stid is nfs4_put_stid(),<br /> which removes the IDR slot under cl_lock, dispatches sc_free<br /> (nfsd4_free_layout_stateid) to release ls-&gt;ls_file via<br /> nfsd4_close_layout(), and drops the nfs4_file reference in its<br /> tail.<br /> <br /> A second issue blocks that switch: nfsd4_free_layout_stateid()<br /> unconditionally inspects ls-&gt;ls_fence_work via<br /> delayed_work_pending() under ls_lock, but<br /> INIT_DELAYED_WORK(&amp;ls-&gt;ls_fence_work, ...) currently runs only<br /> after the setlease call. On the setlease-failure path the<br /> destructor would touch an uninitialized delayed_work.<br /> <br /> nfsd4_alloc_layout_stateid()<br /> nfs4_alloc_stid() /* idr_alloc_cyclic under cl_lock */<br /> nfsd4_layout_setlease() /* fails */<br /> nfs4_put_stid()<br /> nfsd4_free_layout_stateid()<br /> delayed_work_pending(&amp;ls-&gt;ls_fence_work) /* needs INIT */<br /> nfsd4_close_layout() /* nfsd_file_put(ls-&gt;ls_file) */<br /> put_nfs4_file()<br /> <br /> Fix by hoisting the ls_fenced / ls_fence_delay / INIT_DELAYED_WORK<br /> initialization above the nfsd4_layout_setlease() call, and replace<br /> the manual nfsd_file_put + put_nfs4_file + kmem_cache_free cleanup<br /> with a single nfs4_put_stid(stp).