CVE-2026-89667

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> nfsd: close shrinker/GC/fsnotify vs per-net shutdown race in filecache<br /> <br /> The shrinker, GC worker, and fsnotify/lease callbacks can unhash an<br /> nfsd_file from the rhashtable and then call<br /> nfsd_file_dispose_list_delayed() to move it to the per-net dispose list.<br /> If nfsd_file_cache_shutdown_net() runs concurrently, its rhashtable walk<br /> misses the already-unhashed file, and its drain of the per-net dispose<br /> list can run before the file has been queued. The file then sits on<br /> the per-net list with no thread to drain it, leaking both the file and<br /> its associated state.<br /> <br /> The GC worker and shrinker already hold nfsd_gc_lock while walking the<br /> LRU, but in the original code they release it before calling<br /> nfsd_file_dispose_list_delayed(). The fsnotify/lease path<br /> (nfsd_file_close_inode) has no synchronization at all.<br /> <br /> Fix this by:<br /> <br /> 1. Widening nfsd_gc_lock in both nfsd_file_gc() and nfsd_file_lru_scan()<br /> to cover the nfsd_file_dispose_list_delayed() call.<br /> <br /> 2. Wrapping nfsd_file_close_inode() in nfsd_gc_lock so that all three<br /> callers of nfsd_file_dispose_list_delayed() hold the lock.<br /> <br /> 3. Adding a spin_lock/unlock(nfsd_gc_lock) barrier in<br /> nfsd_file_cache_shutdown_net() after the purge, so that any<br /> in-progress disposal has fully completed before the per-net list<br /> is drained.<br /> <br /> All operations inside the lock are non-sleeping (rhashtable lookups,<br /> atomic bit/refcount ops, list moves, svc_wake_up), so the spinlock is<br /> appropriate.