Subject: Re: RFC: new mode bits in stat structure
To: Wolfgang Rupprecht <>
From: Jason Thorpe <>
List: tech-kern
Date: 06/26/1998 10:55:45
On 26 Jun 1998 09:35:07 -0700 
 Wolfgang Rupprecht <> wrote:

 > Just a few quick observations.  
 > 1) The flag bits method is not truly heirarchical.  It only allows a
 >    2-level storage system.

Yes, I understand this limitation...

 > 2) The mod-time, size etc. information of the disk and background
 >    storage can't both be presented as the same time.

This is something that has to be addressed by a different API...

 > When I think "heirarchical" my mental pricture is of something like
 > the union-filesystem with an arbitrary number of layers.  eg:
 > 	layer-1:	ramfs
 > 	layer-2:	disk
 > 	layer-3:	tape
 > The stat structure would perhaps return a new integer field, the
 > filesytem layer-number that this stat structure element referred to.

The problem here is that it doesn't provide any information about
multi-resident files... e.g. something that concurrently exists in
e.g. layer-2 and layer-3.

Hm... So, perhaps what we want is a new flags bit which indicates "resident
at this layer", and a new stat call which takes a "layer to stat" as an
additional argument.  What frightens me about this, however, is that it
is (1) a new API, requiring modification of potentially large numbers of
programs, and (2) potentially large amounts of additional overhead just
to do an ls.

Also, the additional layers might not be handled by the kernel at all,
whereas the ability for e.g. FFS to read the file from disk is indicated
by "non-resident bit clear".

Jason R. Thorpe                             
NASA Ames Research Center                            Home: +1 408 866 1912
NAS: M/S 258-5                                       Work: +1 650 604 0935
Moffett Field, CA 94035                             Pager: +1 650 428 6939