6.0R-alpha ISOs - not bootable?

[Wilko Bulte]

On Fri, Feb 10, 2006 at 10:28:59AM +0000, Doug Rabson wrote..
> On Monday 06 February 2006 21:07, Russell Howe wrote:
> > On Fri, Jan 20, 2006 at 12:50:16AM +0000, Russell Howe wrote:
> > > On Thu, Jan 19, 2006 at 04:48:03PM -0500, John Baldwin wrote:
> > > > Simple question I know, but have the MD5 sums been verified to
> > > > make sure the file wasn't downloaded as ASCII or something weird?
> > >
> > > I checked the MD5sum of the ISO image that I used to create the
> > > CDs, and it checked out. I haven't (yet) dd'd the filesystem off
> > > the CD and md5'd that to check that the CD was created correctly.
> > > I'll take it into the office tomorrow and check.
> >
> > OK, so it's not quite Jan 21st, but I've got my DVD/CD-RW into a
> > Linux box, and checked the discs. One of them came up with I/O errors
> > so I chucked it.
> >
> > I burnt a fresh disc of the 6.0R bootonly .iso image, and
> > 'md5sum /dev/cdrom' checks out OK against CHECKSUM.MD5 on
> > ftp.freebsd.org. The alpha currently has NetBSD on it, so I did an
> > md5 from there of the same disc and that came out OK too, but it
> > still failed to boot in the same way as before.
> >
> > > If all looks well, then I'll try and find out where things broke
> > > between 4.10R and 6.0R
> >
> > I burnt the 5.4 bootonly .iso image, checking the MD5sums on both the
> > Linux box and the NetBSD install on the alpha, and that gets past the
> > bootloader just fine, although it has the problem I always had with
> > FBSD 5.x of the box rebooting just as the SCSI controller (isp0) is
> > initialised.
> >
> > Is there anything in between 5.4 and 6.0R which I could test?
> 
> I suppose its possible that the part of the build process that makes the 
> image 'SRM bootable' was broken. This is done by setcdboot from 
> ports/sysutils/setcdboot and consists of finding the place on the CD 
> that contains the bootloader and marking the first sector of the CD 
> image so that SRM can find it. It writes to the last four uint64 fields 
> of the first 512 byte sector as follows:
> 
> 	uint64_t *lp = ... point to sector buffer ...;
> 	lp[60] = ... bootloader length in 512 byte blocks ...;
> 	lp[61] = ... bootloader start in 512 byte blocks from start of CD ...;
> 	lp[62] = 0;
> 	lp[63] = ... simple checksum of entire 512 byte sector ...;
> 
> If the last word is not correct, SRM will refuse to boot the disc.

On *every* Alpha system it will fail.  And it works on my Alphas (DS10,
164sx)

I run setcdboot manually on the Release ISOs I build, then burn these
ISOs and test boot them.

So I don't think this theory will hold.

-- 
Wilko Bulte				wilko at FreeBSD.org