On Mon, 3 Dec 2001, Ben Harris wrote:

> On Mon, 3 Dec 2001, Gavan Fantom wrote:
>
> > * The EtherH driver chokes on startup thanks to a bug in the podloader
> > stuff (see PR#14811)
>
> Thanks.  I've got a proper fix in my tree which I'll check in once it's
> tested.

What constitutes a "proper fix"? Is that changing which register
int_restore etc uses to save the state, or maybe even changing them to
macros (since IIRC they're only one/two instructions) with a register as a
parameter?

> > * NFS mounting anything complains with "Program version wrong" (ie a
> > mismatch in the version of the NFS protocol between client and server),
> > which seems odd given that the root filesystem is already nfs mounted, and
> > I can do an nfs v2 loopback mount on the nfs server.
>
> This can also mean a mismatch between mount_nfs and the kernel -- i.e.
> mount_nfs has asked for an NFSv3 mount, and the kernel doesn't support
> that.  Try specifying -o -2.

I did try specifying -2 in fstab, but that didn't appear to change
anything.

> > * Once at the login prompt, typing in a username causes a fair bit of
> > network traffic (as expected), followed by a hang. Logging in as root with
> > no password, so I don't know exactly where in the login procedure it's
> > hanging. The last network activity is again a readdir call.
>
> Hmm.  Does readdir work in general?  Have you got some swap by this point?
> Can you try booting single-user instead?

Yes, swap is working. Booting single-user hangs on trying to start the
shell.

> > * There does not appear to be any serial console code. A serial console
> > would complete my remote development environment (I already have
> > netbooting, and a remote power switch, serial console is the only bit
> > missing now), so would be very useful. There is already a serial driver,
> > so to allow the console to use it, where do I start?
>
> Not sure.  I've never even tested the serial driver code.  I know Reinoud
> added serial console support to acorn32 recently, so looking at his
> changes might be helpful.

Well, I could try putting a getty on tty00 to see if that works. I'll take
a look through acorn32 as well.

> > * The keyboard and mouse drivers are not sufficiently separable. Config
> > allows you to take out, for example, the mouse driver, but the compiler
> > throws up some errors. Is this a case of missing ifdefs, or should config
> > be disallowing the separation of these drivers?
>
> Almost all the code is shared between the keyboard and mouse drivers, so
> I'm not sure you'd save an awful lot by losing the mouse portion.  Still,
> I'll look into it.

Indeed. There's a comment in the actual driver that makes this pretty
clear. Maybe there should be a similar comment in the default config file.

-- 
Gillette - the best a man can forget