The Linux kernel's NFSv3 server did not properly validate the nanoseconds field (tv_nsec) in client-supplied atime and mtime timestamps for SETATTR, CREATE, MKDIR, SYMLINK, and MKNOD operations. Although the nseconds field was well-formed as a uint32, it could be out of the valid range (tv_nsec must be less than NSEC_PER_SEC). This invalid value was stored verbatim in the inode timestamps, leading to corruption of on-disk metadata timestamps in ext4 and XFS filesystems due to overflow and clobbering of seconds fields. The fix involves checking these timestamps in the NFS procedure handlers and returning NFS3ERR_INVAL for invalid values before any object creation or attribute changes occur, preventing metadata corruption.

Clients could cause corruption of file timestamps on disk by sending out-of-range nanosecond values in NFSv3 SETATTR and create operations. This affects the integrity of filesystem metadata, potentially causing incorrect file times to be stored and reported. There is no indication of remote code execution or denial of service from the provided data.

A fix is available that validates client-supplied atime and mtime nanosecond values in the NFSv3 server handlers and rejects out-of-range values with NFS3ERR_INVAL before any changes are made. Systems should apply the official Linux kernel update that includes this validation to prevent metadata corruption. No additional mitigation is required as the patch prevents the issue at the protocol handling level.