Re: About the Darwin merge

On Wed, 23 Jun 2004, Gordon Bergling wrote:

> Xin LI wrote:
>
> >Hi, -CURRENT,
> >
> >From the 5.3-RELEASE todo page I saw:
> >
> >"Merge of Darwin msdosfs, other fixes

> >Is there someone actively working on this? BTW. A Darwin msdosfs

No me.  I have many local changes to merge first (mainly to speed up
writes by a factor of between 5 and 20 using VMIO and non-pessimal
cluster allocation).

> >merge patchset, which was originally posted as kern/55861 with
> >some minor modifications, was committed last December. Shall we
> >modify the status of this item?

That patch set took a lot of work to clean up and fix after it was
committed.  It remains incomplete.  E.g., fsck_msdosfs always (except
with -n of course) claims to fix up the dirty flag if it is set, but
doesn't actually do so unless the FAT needs to be written to fix some
other problem.

> A few weeks ago I had started to take an deeper look into the darwin sources
> especially into msdosfs. I had talked to scottl_at_ and if I remember
> correctly he had
> cc:ed rwadson_at_ about an initial diff I had created. This diff can be
> found at
> http://www.0xfce3.net/files/freebsd/01-msdosfs-darwin-merge-050604.diff
> (~ 240 KB).
>
> This diff is only a diff against HEADs msdosfs and the first release of
> apples msdosfs (-30).

More than 50% of it is to undo FreeBSD changes.  E.g., it reverses old
changes to remove __P(()) and atari support.  A diff relative to the
2000/08/28 version of -current is only 140K.

Bruce