Re: amd64 boot failure (-current and 6.0_BETA)

[David Laight <david%l8s.co.uk@localhost>]

On Thu, Mar 29, 2012 at 03:48:07PM +0900, Joerg Sonnenberger wrote:
> On Thu, Mar 29, 2012 at 07:48:49AM +0200, dieter roelants wrote:
> > On Thu, 29 Mar 2012 01:02:14 +0900
> > fukumoto%imasy.or.jp@localhost wrote:
> > 
> > > amd64 -current and 6.0_BETA, boot loader fails to find /boot on my
> > > desktop machine (AMD Phenom II X4 905e and a Gigabyte GA-MA785GM-US2H
> > > motherboard).  Message is:
> > > 
> > > NetBSD/x86 ffsv2 Primary Bootstrap
> > > Boot failed (errno  5): Can't open /boot
> > 
> > Sounds like PR 45197. I've narrowed down the date of breakage to
> > 20110616.
> 
> That's still a lot of code, especially since I don't see any breakage.
> It shouldn't be hard to at least identify the exact commit that triggers
> it, there aren't too many.

That looks like the change to libsa that changed the calling convention.
I think that was done so it would compile with one of the other compilers.

Quite why that system doesn't like it is a different matter.
Especially since it doesn't die horribly.

Since it doesn't explode I'd guess the disk reads are either failing,
returning the wrong sector, or (maybe) writing to the wrong place.

I'd add some tracing around the disk sector reads.
No printf() in bootxx, but there is code to output strings and
hex-formatted numbers.

I really must apply some 'round tuits' to to bootxx_ffsv1 and bootxx_ffsv2.
7k really should be enough to support both fs types and 64k fragments.

        David

-- 
David Laight: david%l8s.co.uk@localhost