Re: [PATCH] VFS autmounter support

[H. Peter Anvin]

David Howells wrote:
>>>That _is_ actually true. Doing "ls -l" in that directory would otherwise
>>>cause a mount storm.
>>>
>>
>>It's not.  ls -l and all the GUI tools do lstat(), not stat().
> 
> 
> Sorry... you're correct. That should have been "ls -F" or "ls --color", both
> of which are, I believe, commonly used - _they_ definitely use stat() as well
> as lstat().
> 

Only if S_ISLNK.

> 
>>>follow_link resolving to itself? Surely that'll cause ELOOP very quickly?
>>>And where does this "dummy directory inode" live?
>>
>>Nope.  You can follow_link() nonrecursively.  You need a dummy directory 
>>inode to mount upon anyway.
> 
> You're right about follow_link() not recursing... it would have to recurse
> itself, and so can avoid that. However, if it only ever follows to itself, how
> does that help? That never actually gets you anywhere... It needs to trigger a
> mount at some point.
> 
> Or do you mean it should follow to an arbitrary (disconnected, otherwise it
> changes the topology from what the AFS admin required) dentry with a dummy
> directory inode attached to it?
> 

You could do that if you wanted to, but it doesn't need to.

	-hpa