Xen devel - May 2006 - RE: RE: [Xen-changelog] [PAE] Allow pgdirs above 4GB for paravirt guests.

It''s same, but it can be exposing a bug with loading a wrong module.
Without the patch, it should complain like:

Loading scsi_mod.ko module
scsi_mod: version magic ''2.6.15-1.2054_FC5smp SMP 686 REGPARM 4KSTACKS
gcc-4.1'' should be ''2.6.16.13-xen0 SMP 686 gcc-3.4''
insmod: error inserting ''/lib/scsi_mod.ko'': -1 Invalid module
format

With that patch, it hangs like that:

Loading scsi_mod.ko module
(XEN)

I don''t see anything after (XEN). 

A quick debugging tells me that one processor is waiting at
smp_call_function and the other one has a bogus CR3, i.e. all four PDPT
entries are 0. I''ll take a look later today if I have time.

Jun
---
Intel Open Source Technology Center 

-----Original Message-----
From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk] 
Sent: Saturday, May 27, 2006 2:05 AM
To: Nakajima, Jun
Cc: xen-devel@lists.xensource.com
Subject: Re: [Xen-devel] RE: [Xen-changelog] [PAE] Allow pgdirs above
4GB for paravirt guests.


On 26 May 2006, at 23:31, Nakajima, Jun wrote:
> Dom0 hangs with this patch when loading modules. If I back it out, it
> boots fine. I configured dom0 as SMP (and the machine has >4GB).
Is the hang just of domain0, not Xen? What happens if you force 
emulation of writable pagetables (no batching)?

  -- Keir

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

On 27 May 2006, at 17:23, Nakajima, Jun wrote:
> With that patch, it hangs like that:
>
> Loading scsi_mod.ko module
> (XEN)
>
> I don''t see anything after (XEN).
>
> A quick debugging tells me that one processor is waiting at
> smp_call_function and the other one has a bogus CR3, i.e. all four PDPT
> entries are 0. I''ll take a look later today if I have time.
I haven''t yet reproduced this, but that may be because I''ve
just found
that SMP boot for domain0 without CONFIG_HOTPLUG_CPU is broken (I end 
up booting domain0 with a single processor, which isn''t the behaviour 
we used to have). I''ll fix that regression first and then when I have 
SMP domain0 hopefully I''ll be able to repro the bug.

  -- Keir


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

On 27 May 2006, at 17:23, Nakajima, Jun wrote:
> A quick debugging tells me that one processor is waiting at
> smp_call_function and the other one has a bogus CR3, i.e. all four PDPT
> entries are 0. I''ll take a look later today if I have time.
Okay, I can repro all sorts of weird crap with this new PAE patch: it''s
really very broken indeed. I''ll look into it myself. Thanks!

  -- Keir


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