On Fri, 2003-02-14 at 00:04, H. Peter Anvin wrote:
> Ogden, Aaron A. wrote:
> > Sun Solaris has been doing this for a while now.
> 
> The Solaris VFS is very, very different; they seem more willing to slow 
> their VFS down for these kinds of features.

Very much so.

> 
> > Maybe autofs5 should be constructed to allow this feature?
> > Ian's ghosting patches for autofs4 seem very stable but they
>  > don't work on all kernel versions, the autofs4 kernel source is 

As far as I know only 2.4.20 has broken the patch.

> changing too much.
> 
> Until they've been through a heavy grinder, I don't trust them.  During 
> autofs v3 development after the dcache was added I would find bugs that 
> would show only after several days of running on a hundred heavily 
> loaded SMP machines with a 1-second timeout.  Writing race-free code is 
> extremely hard.

You know I agree with you but I am trying and I have been running may
patch on a number of machines for ages. It plays well with the original
v4 and your v3.

And yes there are unusual problems that I cannot nail. However, it is
not at all clear whether they are caused by my kernel patch, the autofs
daemon or nfs. Until the nfs client is rock solid I will probably not
nail any of the problems.

That said we need this functionality in the automounter. Linux will not
cut it in the enterprise without it.

At least I have something to work with!

-- 
Ian Kent <[EMAIL PROTECTED]>

_______________________________________________
autofs mailing list
[EMAIL PROTECTED]
http://linux.kernel.org/mailman/listinfo/autofs

Reply via email to