cvs commit: src/sbin/growfs debug.c debug.h

[Nate Lawson]

On Thu, 14 Aug 2003, Lukas Ertl wrote:
> On Thu, 14 Aug 2003, Nate Lawson wrote:
> > The number one problem is the code sucks.  No one is willing to maintain
> > it because of this.
>
> Oh. :-)
>
> > The number two problem is that not enough testing has been done with
> > growfs on UFS2.  Feel free to do this and report any problems.
>
> I've successfully grown UFS2 filesystems residing on vinum volumes in the
> past, but I didn't do I regularly.

Have you done it with ACLs?  Snapshots?  I've done light testing using
md(4) but it's not enough.  This needs to be tested.

> > The actual problem I am addressing is larger than growfs.  Basically, any
> > time a sblock change is made, every UFS utility in the system has to be
> > updated and growfs has not kept up.  So I have extracted sblock updating
> > into a routine that is shared between the kernel and userland so that it's
> > only in one place in the code.  But more work is needed before it can be
> > committed.
>
> Well, I'd say that libufs is already a right step in this direction.

It also needs work but is outside the scope of what I'm doing.

> I
> would have already patched growfs to use libufs, but growfs somehow seems
> a little bit too much like voodoo. :-)

Back to my point at the top.

-Nate