Re: "VFS: no free inodes - contact Linus"

Bill Hawes (whawes@star.net)
Thu, 16 Oct 1997 13:00:50 -0400

David Luyer wrote:
> Since file-max is 4096 and inode-max is 32768 (inode-nr is normally
> 16-20k), I guess that means it had to go thru the code path
> nr_inodes<max_inodes && grow_inodes() != 0, ie, got an ENOMEM trying to
> grow_inodes(). Is this the problem? Is there a solution? (just using the
> "best" from before when grow_inodes() fails would look like an obvious
> one...)

There may be some conditions under which the current search algorithm
will fail. If this problem is repeatable for you (and you're willing to
experiment), I can send you a patch to play with.

Regards,
Bill