Re: dom0 kernel immediate reboot?

[Pavel Cahyna <pavel%netbsd.org@localhost>]

On Tue, Sep 11, 2007 at 07:56:43AM +0200, Manuel Bouyer wrote:
> On Tue, Sep 11, 2007 at 09:28:53AM +1000, Sarton O'Brien wrote:
> > On Tue, 11 Sep 2007 06:20:53 am Manuel Bouyer wrote:
> > > On Mon, Sep 10, 2007 at 02:58:40PM -0500, david l goodrich wrote:
> > > > On Mon, 10 Sep 2007 20:47:53 +0200, Manuel Bouyer
> > > > <bouyer%antioche.eu.org@localhost>
> > > >
> > > > wrote:
> > > > > On Mon, Sep 10, 2007 at 09:54:25PM +1000, 
> > > > > bsd-xen%roguewrt.org@localhost wrote:
> > > > >> > this way.
> > > > >>
> > > > >> Heh, in the end mine is too :)
> > > > >>
> > > > >> I'm sure it used to boot from /usr though.
> > > > >
> > > > > Yes, for bigger filesystems it can work. It basically depends on the
> > > > > offset
> > > > > of the data blocks in the partition. If the block numbers are low 
> > > > > grubs
> > > > > works,
> > > > > if they are high grubs loads garbage. On a fresh filesystem the
> > > > > new files are likely to end up at the beggining of the partition
> > > > > (especially if the directory tree isn't too depth). After some use you
> > > > > copy a new xen.gz, it gets a high inode number and high blocks 
> > > > > numbers,
> > > > > and
> > > > > can't be loaded.
> > > >
> > > > I understand this, and understand that this is The Way It Is Now.
> > > >
> > > > But isn't this a ... bug?  Has anyone filed anything with grub?
> > >
> > > it's a bug, and no I didn't report it to grub ...
> > 
> > I might add, that this only seems to be a bug for compressed images as this 
> > is 
> > my existing grub config:
> > 
> > kernel (hd0,e)/pkg/xen3-kernel/xen dom0_mem=131072 vga=text-80x25
> > 
> > And so long as it's decompressed, no probs.
> 
> No, I think you've just been lucky. I've had problems with non-compressed
> images as well on large /

I also had problems only with compressed images (ffsv1). The read pattern is
different in the compressed case than in the uncompressed case.

Pavel