Mac mini and FreeBSD - buildworld

[Peter Grehan]

Hi Garance,

> Okay.  I added those two updates, and I was able to do a complete
> buildworld/installworld cycle.

  Great news :-)

> The only noticeable problem when I
> rebooted was a stream of complaints about ttya, so I just changed
> the entry for that from 'on' to 'off' in /etc/ttys.

  Yes, that's a relic of the recent tty subsystem changes. I'll fix that.

  I'll go through the mods:

>   lib/libstand/cd9660.c - something about HFS/ISO9660 hybrid CD's
>                           an Apple OpenFirmware 3

  The HFS/9660 hybrid CDs have an Apple partition map, with the
second partition starting 16 sectors into the CD. This is
what OpenFirmware sees when given 'cd:' without a partition
number.

  The mod to libstand was an attempt to modify the iso9660 offset 
calculations when given 'cd:', but since I put in the cmd-line argument 
to the loader, which allows 'cd:0' to be specified, this change is no 
longer relevant.

>   sys/dev/gem/if_gem.c - some changes wrt IFQ maxlen and dequeue.

  That's some patches I've been trying for ALTQ. Should be committed soon.

>   sys/dev/ofw/ofw_console.c - "ofw_nopoll" variable introduced to
>                   neutralize a while loop in ofw_timeout().

  This is to better support running the kernel in the gdb psim ppc 
simulator. Without nopoll, a read from the OpenFirmware console blocks,
so this has to be set to 1 inside the debugger to allow forward 
progress. I might commit this just to avoid keeping my diff alive.

>   sys/dev/zs/*  - several changes probably tied to tty's

  This was Marius Stobl's patch to the zs driver. I've not really 
exercised it.

>   sys/modules/Makefile - changes wrt the building of modules for
>                   'mem', 'uart', 'syscons' and 'netgraph' (it seems
>                   to completely drop 'mem', but I suspect that's a
>                   mistake...)

  Yeh, that was the mem module to-and-fro a while back. I'll bring this 
up to date and commit it, since modules should really be built by 
default from now on.

>   sys/powerpc/conf/GENERIC - a couple of changes.  I actually used
>                   this GENERIC (from the src tree on the ISO) for
>                   my own system build.

  Yep, just ALTQ and PF I think. These can be dumped into GENERIC now.
I'm going to pull EHCI since it doesn't work too well with external 
drive enclosures.

>   usr.sbin/sysinstall/config.c - minor change to recognize partitions
>                    of type = "apple" for powerpc.

  Without this one the /etc/fstab file isn't written, so I'll try and 
get this committed.

> Also, there were quite a few files which had changes that had
> something to do with "spinlocks" vs "critical" areas.  I didn't
> even want to think about pretending to figure those out...

  This was John Baldwin's patches. I've backed these out in my own tree, 
but I think they'll make an appearance soon in HEAD.

> The build that I did used the sources from Dec 27th, plus the two
> updates you mentioned.  That seems to be working okay, but I haven't
> pushed it too much.  Sometime this week I'm going to try another
> build with an up-to-date snapshot of HEAD, the two updates you
> pointed at, and perhaps some of the others listed above.

  Better get my X changes in then :)

later,

Peter.