This is more of an old kernel bug inherited from BSD/SunOS: After trading an NFS partition on a Solaris server, I get: # strace -e mount mount -o remount /loc mount("loc:/loc", "/loc", "nfs", MS_REMOUNT|0xc0ed0000, 0x8057620) = 0 Yet nothing is remounted, we are still looking at the old partition: # df -h /loc Filesystem Size Used Avail Use% Mounted on loc:/loc 29G 29G 949M 97% /loc # touch /loc/share/tmp/test touch: /loc/share/tmp/test: Read-only file system Workaround: # umount /loc # mount /loc # df -h /loc | tee /loc/share/tmp/test Filesystem Size Used Avail Use% Mounted on loc:/loc 36G 29G 7.7G 79% /loc
Assigning to kernel, since it happens with just the syscall, as well.
Works ok in newer kernels
Ah wait, if you change the server side stuff it can fail. Remount remounts an existing handle be it a disk or whatever. The same is true with remounting a renamed file system