Re: Problems with gzipped tarballs on -current

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

Michael L. Hitch wrote:

> On Thu, 24 Dec 2009, John Klos wrote:
> 
> > Hmmm... Now, with a kernel built from yesterday's source, trying to do a
> > make update in pkgsrc reboots the machine when it gets to the extract
> > part. This is the m68k Mac, and it's remote, so I can't see what's going
> > on, but I'll try on the Amiga which is local.
> 
>   I was switching between 5.0_STABLE and -current to test out the
> raidframe parity map (a 6GB raid1 partition takes about an hour to
> rebuild parity, and I frequently get dirty parity on boot).  One of the
> things I was doing was extracting an old source file I had laying around
> on the disk.  It seems to be stable on 5.0_STABLE, but I've had some
> strange memory faults on -current.  A couple of them were 'illegal
> instruction' at weird places.  The last one was a fault in one of the pmap
> routines and appeared to have the wrong address for a page table entry
> pointer it was trying to update.

Well, most my changes are for pmap_bootstrap.c used on 4KB/page ports
and there are actually very few functional changes to amiga pmap
(using pmap_bootstrap_finalize to initialize several parameters,
 replace a magic constant with proper macro, fix NBSEG for 030,
 change one signed value to unsigned etc), so I wonder if
they could cause such occasional panic.

I've tried several tests (extract whole sets, build pkgsrc perl5 etc.)
on HP382 (68040) running NetBSD/hp300 around 20091213 (at the point
most my work was done) and there is no panic on DIAGNOSTIC kernel.

> Since my kernel was using a locally
> modified pmap_motorola.c, I presumed I hadn't managed to keep my changes
> in sync correctly with the recent pmap changes.

BTW, I also fixed all 4KB/page ports to have the same layout
for L1/L2 segements (and some of them work with 8KB/page settings)
so I guess now it's time to apply your pmap changes to other ports..
---
Izumi Tsutsui