On Sat, Jan 18, 2003 at 12:19:09AM -0500, der Mouse wrote:

> > Mounting a nullfs filesystem with an non-absolute mountpoint results
> > in a parmanent mount entry which umount cannot handle anymore.
> 
> See umount's -R option.

Ah interesting, thanks. So it was part of nullfs design decisions to make
no path sanity checking at all? Perhaps we could make it at least check
for extra '/'s, and non-absolute paths or such?

Possibly indirectly related to this I noticed that reverse name cache for
inode->filename can only optionally be enabled with NAMECACHE_ENTER_REVERSE
at kernel compilation time, and that it only seems to apply to linux procfs
simulation for memory maps display. I thought that it might be useful if
fstat(1) could optionally benefit from that as well, with a new flag...

Or are only library filenames reversed in the name cache? I couldn't check
the source about this yet...

Thanks,
Matt