Re: [patch 0/5] [PATCH,RFC] vfs: per-superblock unused dentries list (2nd version)

[Jan Blunck <jblunck@suse.de>]

On Sat, Jun 17, Neil Brown wrote:

> But I cannot see that the whole LRU list needs to be scanned during
> unmount.
> The only thing that does that is shrink_dcache_sb, which is used:
>   in do_remount_sb
>   in __invalidate_device
>   in a few filesystems (autofs, coda, smbfs)
>  and not when unmounting the filesystem (despite the comment).
> 
> (This is in 2.6.17-ec6-mm2).
> 
> I can see that shrink_dcache_sb could take a long time and should be
> fixed, which should be as simple as replacing it with
> shrink_dcache_parent; shrink_dcache_anon.

I don't remember exactly, maybe it was remounting instead of
unmounting. Although I believe that we should call shrink_dcache_sb() instead
of shrink_dcache_anon()+parent() when unmounting. I don't see any reason why we
should shrink the dcache with depth-first traversal instead of just killing
every unused dentry that we find (not even talking about that DCACHE_REFERENCE
handling is nonesense in that case).

> But I'm still puzzled as to why a long dcache LRU slows down
> unmounting. 
> 
> Can you give more details?

I think David already answered that.

Jan