"Jeff V. Merkey" wrote:
> >
> > Have you considered using the md device or something similar as an
> > intermediate layer? That way you get a single linear descriptor
> > across all the partitions.
> >
> > It's not that "Linux always seems to assume that..."; rather, far too
> > often it's that Linux doesn't do things the way you seem to expect
> > them too. You should have noticed that by now.
>
> No, it does what I would expect a unix OS to do, including both the good
> and bad things about Unit. It's very similiar (though not identical) to
> the other Unixes, but with more of a PC flair.
>
Pretty much.
>
> The md device layer simply expands the "one partition - one filesystem"
> concept by saying. " ... here's a bunch of partitions that look like
> "one partition - one filesystem". This won't work with NetWare's
> segmentation model since volume segments are something the FS needs to
> know about. It could work this way in Linux with the mirroring (NetWare
> mirroring could be made to work with the current md architecture, but
> not without significant rework of the buffer cache to support logical
> rather than physical block mapping). Striping, however, is a different
> matter and I need to give this some thought. We could use the md
> devices to do it by concatenating NetWare partitions together, and
> telling all volumes they are only on one segment (and use md underneath
> to remap the partitions) but such a scheme would not be as efficient for
> stiping as what's currently there. Since the FS knows about segments,
> we simply round robin cluster allocation requests between segments, and
> get very good fie distribution across multi-segmented volumes. Using md
> to do this would not be as good since the FS would simply allocated
> until it spanned into the next partition -- NTFS model -- not as good as
> what's already there.
>
Well, as I said, "something similar" is the main thing. The basic issue
allowing you to set up a descriptor, call it a "device" and mount it.
If NWFS want to use additional information about the internal structure
of the device, it is welcome to. However, at least token separation of
the block layer from the filesystem layer is good practice, and makes
the bookkeeping cleaner.
-hpa
-- <hpa@transmeta.com> at work, <hpa@zytor.com> in private! "Unix gives you enough rope to shoot yourself in the foot."- 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 : Fri Apr 07 2000 - 21:00:19 EST