Assar Westerlund wrote:
>
> Wes Peters <[EMAIL PROTECTED]> writes:
> > "Daniel C. Sobral" wrote:
> > >
> > > Just to expand a little bit more, some distributed filesystems *do
> > > not* have a unique identifier like the inode.
> >
> > So then the FreeBSD client software should create one? Do they just assign
> > a random number as the st_ino when stat'ing the file?
>
> If there's none, you of course have to create one. As long as you
> keep giving the same `va_fileid' to the same file (by remembering what
> files you have seen), I guess that's ok. But then I don't know of any
> distributed filesystem that acts this way (what's `same' in the text
> above?). What filesystems are like that?
>
> Looking at some existing file systems:
>
> NFS - the server returns a 32-bit file-ID
The only one I had available, which didn't look like a problem.
> AFS/Arla - files are identified by (cell, volume, vnode,
> uniquifier) which is hashed down to a 32 bit fileno
> Coda - same as AFS/Arla
Are hash collisions handled reasonably? I.e. does the test for st_dev/
st_ino uniquely identifying the file during an entire single mount session
hold true? If so, no problem. If not, I can name at least one popular
security manager program that is going to have a conniption fit. ;^)
--
"Where am I, and what am I doing in this handbasket?"
Wes Peters Softweyr LLC
[EMAIL PROTECTED] http://softweyr.com/
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
