On 14-Apr-2000 H. Peter Anvin wrote:
> Actually, that's something that autofs v3 *does* keep in the dentry; by
> keeping a negative dentry which eventually gets invalidated (by using a
> timer.) That works great in the case of requests that fail quickly. I
> don't worry too much about stuff that fail during mount, since the
> mount tends to be a rate limiter anyway.
That works fine if the daemon can tell its going to fail before anything
is mounted, but once it creates the mountpoint it has to rmdir it on
failure. The rmdir blows it away completely, so there's nothing to
remember something went wrong. I'm thinking of a simple cache which
answers questions like "asking for foobar won't work, because it failed
in the last 30 seconds".
The mount is a rate limiter, but it's a pretty expensive rate limiter.
With autofs v3 there was only one mount per mountpoint, so it wasn't that
bad. With v4 there could be hundreds (then mount 201 fails). Hundreds
of mounts, hundreds of umounts, hundreds of hosts... Lots of activity.
I guess it's somewhat fortunate that there's a limit of 256 deviceless
filesystems (only 256 minors on major 0) - it stops things getting too
pathalogical.
J
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Apr 15 2000 - 21:00:24 EST