Linux Kernel CVE-2026-64156

The vulnerability resides in the Linux kernel's netfs layer (network filesystem abstraction) and its AFS (Andrew File System) client implementation. The netfs_write_single() function handles writing a single folio to a network filesystem backend, while afs_single_writepages() coordinates writeback for AFS directory and symlink inodes. Under WB_SYNC_NONE (asynchronous, non-blocking writeback), if a lock cannot be acquired immediately the write may be intentionally skipped - but the bug is that the inode's dirty flag is not re-asserted afterward. The VFS writeback infrastructure, which may have already cleared that flag, will then never retry the write. The developers note this is 'really only theoretical for directories' because afs_single_writepages() holds ->validate_lock around the call, serializing concurrent access. No CWE is formally assigned, but the root cause resembles CWE-664 (Improper Control of a Resource Through its Lifetime) - specifically a missed state re-assertion after a conditional skip. Affected CPE: cpe:2.3:a:linux:linux. Only kernels containing commit 6dd80936618c4ff852d4db73aca400351d9bd9f0 and lacking the fix commits are vulnerable.