getfsstat visibility patch

Matthew Dillon dillon at
Wed Feb 2 12:45:06 PST 2005

:As extension of the original patch, this removes (with one exception)
:all code from filesystems which deals with f_mntonname. Most debug
:statements use f_mntfromname instead, which should be enough to
:identify the filesystem. I didn't use cache_fullpath in the
:shutdown path to avoid possible blocks, since cache_fullpath does
:a local allocation. I changed cache_fullpath to work for pure kernel
:Both getfsstat and fhstatfs check the visibility, the other stat calls
:just have the (chroot) relative path handling.
:I'm still thinking about how to best handle the remaining use in
:cd9660fs, but I think using cache_fullpath doesn't hurt.
:Matt, what do you think?

    I like it.  For a moment I considered, well, what if someone renamed
    a parent directory above a mount point, but then I realized that the new
    methodology would report the correct path in that case while the old one
    would no longer be correct (and what sane sysop would ever do that

    Go for it.  It's precisely cleanups like this that I had hoped the
    namecache code would make possible.


More information about the Submit mailing list