Xen devel - Aug 2009 - cannot access memory beyond end of bootstrap direct-map area

Hi,

Investigating the above message during xen 3.3 boot, found we left
a wrong assumption [modules_length == (mod[n].end - mod[0].start)]
in move_memory call, leading to an access above BOOTSTRAP_DIRECTMAP_END
and this panic on x86_32.

In my understanding Keir''s patch 18630 was fixing this kind of problem
and I''m wondering why this move_memory call didn''t go away
with it ?

If this were ok, attached patch fixes the panic on my system.

Cheers,
Pascal

-- 
\o/   Pascal Bouchareine - Gandi 
 g    0170393757           15, place de la Nation - 75011 Paris      


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

On 27/08/2009 18:28, "Pascal Bouchareine" <pascal@gandi.net>
wrote:
> Investigating the above message during xen 3.3 boot, found we left
> a wrong assumption [modules_length == (mod[n].end - mod[0].start)]
> in move_memory call, leading to an access above BOOTSTRAP_DIRECTMAP_END
> and this panic on x86_32.
> 
> In my understanding Keir''s patch 18630 was fixing this kind of
problem
> and I''m wondering why this move_memory call didn''t go
away with it ?
Looks like a backporting mistake. The original bug didn''t happen on my
system so the bad backport wasn''t noticed.
> If this were ok, attached patch fixes the panic on my system.
You should be able to remove the bogus move_memory() call, and change
absolutely nothing else. Does that also work for you? That''s the change
I
would check in.

 -- Keir



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

On Aug 27, 2009, at 8:55 PM, Keir Fraser wrote:
>> If this were ok, attached patch fixes the panic on my system.
>
> You should be able to remove the bogus move_memory() call, and change
> absolutely nothing else. Does that also work for you? That''s the  
> change I
> would check in.
Yup, I thought that''s what the patch did, modulo we need to keep track
of initial_images_start
and need "e" in place for the loop ? Or I need a look at the original
fix.

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

On 27/08/2009 22:41, "Pascal Bouchareine" <pascal@gandi.net>
wrote:
>> You should be able to remove the bogus move_memory() call, and change
>> absolutely nothing else. Does that also work for you? That''s
the
>> change I
>> would check in.
> 
> Yup, I thought that''s what the patch did, modulo we need to keep
track
> of initial_images_start
> and need "e" in place for the loop ? Or I need a look at the
original
> fix.
Initial_images_start and e would already be correct for the loop. Just the
move_memory() call outside the loop is completely unnecessary. Try removing
just that call and see if that works.

 -- Keir



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

On Aug 28, 2009, at 8:38 AM, Keir Fraser wrote:
> Initial_images_start and e would already be correct for the loop.  
> Just the
> move_memory() call outside the loop is completely unnecessary. Try  
> removing
> just that call and see if that works.
>
Ho, I have been misreading it. Indeed, just removing the call works too.





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