Xen devel - Apr 2009 - "Old method" bootloader failing with large ramdisk

I''m trying to boot a PV guest using the "old method"
of passing kernel= and ramdisk= and it appears to
work fine with a "small" initrd but not with a "large"
one. (Small is 4MB, large is 154MB.)  I''m sure both
of the initrd''s are properly gzip''ed etc.  Unpacked,
the large one approaches 400M.

By doing some kernel startup debugging, it appears
that the large initrd never finds its way into memory.
Or at least not completely or properly.

So I''m guessing that the "old method bootloader"
is having problems with the large initrd.  Silently,
I think, as I can''t find anything that looks like errors
in /var/log/xen/domain-builder-ng.log.

Should this be expected to work?  Has anyone else
been successful with a large initrd doing this?

Thanks,
Dan


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

I guess I should add that this is with a linux-2.6.29
PV kernel so might be something there too.

And "not completely or properly" means that "/init"
can''t be found (sys_access("/init") fails for the
large initrd
> -----Original Message-----
> From: Dan Magenheimer 
> Sent: Friday, April 24, 2009 5:10 PM
> To: Xen-Devel (E-mail)
> Subject: "Old method" bootloader failing with large ramdisk
> 
> 
> I''m trying to boot a PV guest using the "old method"
> of passing kernel= and ramdisk= and it appears to
> work fine with a "small" initrd but not with a "large"
> one. (Small is 4MB, large is 154MB.)  I''m sure both
> of the initrd''s are properly gzip''ed etc.  Unpacked,
> the large one approaches 400M.
> 
> By doing some kernel startup debugging, it appears
> that the large initrd never finds its way into memory.
> Or at least not completely or properly.
> 
> So I''m guessing that the "old method bootloader"
> is having problems with the large initrd.  Silently,
> I think, as I can''t find anything that looks like errors
> in /var/log/xen/domain-builder-ng.log.
> 
> Should this be expected to work?  Has anyone else
> been successful with a large initrd doing this?
> 
> Thanks,
> Dan
> 
> 
>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Don''t know if anybody cares about this, but it
definitely appears to be a Xen bug.  By adding
unused junk to a working initrd, I can cause a
boot failure.  When the initrd.gz is 109MB
it boots fine, but boot fails at 126MB.  This
is with a 1GB guest, so I don''t think it is
memory-size related.  Unfortunately, I didn''t
measure the unpacked sizes of these.

I''m off to work on other things now as I have
a workaround (smaller initrd).

Dan
> -----Original Message-----
> From: Dan Magenheimer 
> Sent: Friday, April 24, 2009 5:22 PM
> To: Dan Magenheimer; Xen-Devel (E-mail)
> Subject: RE: "Old method" bootloader failing with large ramdisk
> 
> 
> I guess I should add that this is with a linux-2.6.29
> PV kernel so might be something there too.
> 
> And "not completely or properly" means that "/init"
> can''t be found (sys_access("/init") fails for the
> large initrd
> 
> > -----Original Message-----
> > From: Dan Magenheimer 
> > Sent: Friday, April 24, 2009 5:10 PM
> > To: Xen-Devel (E-mail)
> > Subject: "Old method" bootloader failing with large ramdisk
> > 
> > 
> > I''m trying to boot a PV guest using the "old
method"
> > of passing kernel= and ramdisk= and it appears to
> > work fine with a "small" initrd but not with a
"large"
> > one. (Small is 4MB, large is 154MB.)  I''m sure both
> > of the initrd''s are properly gzip''ed etc.  Unpacked,
> > the large one approaches 400M.
> > 
> > By doing some kernel startup debugging, it appears
> > that the large initrd never finds its way into memory.
> > Or at least not completely or properly.
> > 
> > So I''m guessing that the "old method bootloader"
> > is having problems with the large initrd.  Silently,
> > I think, as I can''t find anything that looks like errors
> > in /var/log/xen/domain-builder-ng.log.
> > 
> > Should this be expected to work?  Has anyone else
> > been successful with a large initrd doing this?
> > 
> > Thanks,
> > Dan
> > 
> > 
> >
>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel