Compiling error under Slackware

Kevin Darcy kcd at daimlerchrysler.com
Fri Sep 15 19:56:38 UTC 2000


Joseph S D Yao wrote:

> On Thu, Sep 14, 2000 at 06:34:29PM -0400, Kevin Darcy wrote:
> > Joseph S D Yao wrote:
> ...
> > > "No space left on device" usually means what it says.
> >
> > Usually. But inode exhaustion can produce that error also, even if you have plenty
> > of disk-block "space" in the filesystem. I've often seen our Level 1 (and
> > sometimes even Level 2) Unix admins forget that little piece of Unix trivia...
>
> Well, true.  But if you do the right kind of 'df' [varies among systems
> but one of -i or -t combined with -k usually does it], you get both.
>
> > (And don't even get me started on "why does 'df' show a different amount of free
> > space than 'du' does?"...)
>
> (1) Sometimes one reports in Kb and the other in "blocks" of 512 bytes.
> (2) On some types of file systems, some portion of the disk ["minfree"]
>     is reserved, so that if the super-user needed to fix something, she
>     or he would have space even when the file system was "full".

Actually, that's a misconception. minfree was created for performance reasons, not just
to provide a reserve in case you run out of regular space. And for performance reasons,
it always needs to be a certain *percentage* of the available blocks, regardless of how
big the disk or disk block size is.

> (3) But 'du' doesn't show free space!  ;-}
> (4) There may be differing calculations of amount of indirect block
>     space used.  It depends on your FS and OS.

That wasn't really meant to be a Unix FS trivia question. But if it was, there are a
couple of answers you missed:

5) Deleted-but-still-open files (counted by "df" but not "du).

6) Sparse ("holey") files ("du" counts their apparent block usage, "df" their real
usage).


- Kevin





More information about the bind-users mailing list