[Xen-ia64-devel] multi-domain fails on -sparse tree xenlinux

["Magenheimer, Dan (HP Labs Fort Collins)" <dan.magenheimer@xxxxxx>]

Hi Kevin --

Per the previous message, multiple domains is broken (again)
in the xenlinux now built from the -sparse tree.  The symptom
is that "xm create" starts the domain but I think when it
gets to blkif, it not only freezes but appears to kill dom0
also!  If you can take a look at this, I would appreciate it
as, once it is working from -sparse, I expect that all
users and developers will switch to it.

It may turn out to be something very simple.  I've been
spending so many hours on this, that I might have missed
the obvious.  If not,  I expect the problem will be in
one of these three areas:

1) I made every effort to ensure all of the patches went in.
If anything was not marked with __ia64__, it may not have
gone in.  It's also possible that I missed a patch if it
was not in drivers/xen.  It might be worthwhile to first
do a visual scan of all the patches and ensure they all
appear as you expected in the tree.  (See
linux-2.6-xen-sparse/arch/ia64/xen-mkbuildtree-pre for
all the steps that patch up the directories and files.) 

2) It may also be recent changes to the core.  In particular
blkback/interface.c has had some significant changes and two
lines that used to be replaced by ifdef _ia64__ are completely
gone.  There are new calls to *_vm_area() from interface.c (and
similarly from drivers/xen/net and drivers/xen/tpm) that are
defined in a new drivers/xen/utils.c file.  These routines
use a routine called generic_page_range() that is defined
in a Xen-specific addition to linux/mm/memory.c and
linux/include/mm.h.  All of this looked relatively generic
so I just used it directly, which may or may not have been
the right answer.

3) I also noticed a flurry of xen/common/grant_table.c changes
near tip in xen-unstable.hg (_after_ my merge point).  It's
possible that the grant
table mechanism was unstable at my merge point, but that seems
unlikely as we override the common grant_table with an
ia64-specific one (a source of future problems that we should
probably resolve).  Still, this might be worth looking at.

Thanks,
Dan

P.S. "xend start" puts some network-bridge error messages into
xend.log, but starting xm create does not result in anything
going into xend.log or xend-debug.log

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel