On 05/03/2010 05:23 PM, Gerry Reno wrote:
> On 05/03/2010 04:36 PM, Konrad Rzeszutek Wilk wrote:
>> On Mon, May 03, 2010 at 01:18:44PM -0400, Gerry Reno wrote:
>>> On 05/03/2010 12:18 PM, Konrad Rzeszutek Wilk wrote:
>>>> On Fri, Apr 23, 2010 at 01:50:40PM -0400, Gerry Reno wrote:
>>>>
>>>>> I previously had the pv_ops dom0 2.6.31.6 kernel built and running 
>>>>> on my
>>>>> hardware (amd phenom ii x4 quad-core) just fine under karmic.  But 
>>>>> that
>>>>> kernel had issues with rxchecksumming.  So I upgraded the server to
>>>>> lucid and am now trying to compile a new pv_ops dom0 kernel for it.
>>>>>
>>>>> Now, yesterday I carefully built the latest 2.6.32.10(xen/stable) 
>>>>> pv_ops
>>>>> dom0 kernel twice.  Once from the config in the email Lucid pv_ops 
>>>>> dom0
>>>>> howto and once from the lucid server config modified by adding all 
>>>>> the
>>>>>
>>>> Gerry,
>>>>
>>>> Can you attach your .config file ? I am not seeing it anywhere in the
>>>> thread. Also is this x86_64 or i386 build? If it is i386 do you have
>>>> CONFIG_NUMA enabled? If so, what happens if you disable that?
>>>>
>>>> Can you also tell me which Phenom it is? 940?
>>>>
>>>>
>>> Konrad, I attached my .config file and /proc/cpuinfo.  The build was
>>> x86_64.  The cpu is AMD Phenom II x4 940.
>> When I use your .config I get these:
>>
>> [   10.781056] ------------[ cut here ]------------
>> [   10.785736] WARNING: at
>> /home/konrad/git/neb.64/linux-up/mm/vmalloc.c:1AMD-BOOT #7
>> [   10.790016] Call Trace:
>> [   10.790016]  [<ffffffff810660cb>] warn_slowpath_common+0x7b/0xc0
>> [   10.790016]  [<ffffffff81066124>] warn_slowpath_null+0x14/0x20
>> [   10.790016]  [<ffffffff8111bb19>] vmap_page_range_noflush+0x2d9/0x370
>> [   10.790016]  [<ffffffff8111cfaf>] ? __vmalloc_area_node+0x5f/0x190
>> [   10.790016]  [<ffffffff8100fadd>] ?  
>> xen_force_evtchn_callback+0xd/0x10
>> [   10.790016]  [<ffffffff8111bbe2>] map_vm_area+0x32/0x50
>> [   10.790016]  [<ffffffff8111d050>] __vmalloc_area_node+0x100/0x190
>> [   10.790016]  [<ffffffff8111d0f7>] __vmalloc_area+0x17/0x20
>> [   10.790016]  [<ffffffff8103592e>] module_alloc+0x5e/0x60
>> [   10.790016]  [<ffffffff8109b8c6>] 
>> module_alloc_update_bounds+0x16/0x50
>> [   10.790016]  [<ffffffff8109e479>] load_module+0x989/0x1ce0
>> [   10.790016]  [<ffffffff8113f8a5>] ? vfs_read+0xb5/0x1a0
>> [   10.790016]  [<ffffffff8109f849>] sys_init_module+0x79/0x250
>> [   10.790016]  [<ffffffff810140f2>] system_call_fastpath+0x16/0x1b
>> [   10.790016] ---[ end trace a7919e7f17c0a727 ]---
>> FATAL: Error inserting xen_kbdfront
>> (/lib/modules/2.6.32.12AMD-BOOT/kernel/drivers/input/xen-kbdfront.ko):
>> Cannot allocate memory
>>
>> Is this about the same thing you see? (the calltrace being of interest
>> here).
>>
> The screen output goes by so fast I cannot tell if this call trace is 
> in it.  I'm still waiting for a serial card to show up so I can 
> capture the output.  When I boot the 2.6.32.10 kernel under Xen pv_ops 
> dom0 it will either start showing "soft lockup" bugs or it just hangs 
> after XENBUS: Unable to read cpu state.  If I boot the 2.6.32.10 
> kernel without Xen then it just hangs - no output at all.
>
> -Gerry
>
Finally, my serial port header cables showed up so I installed them and 
now have some console output.

Case 1: Xen 4.0.0-rc8 / pv_ops dom0 kernel 2.6.32.10

After playing around with the order of the consoles on the xen and 
kernel lines I am able to get a login prompt but only after quite a long 
time (4-5 mins).  It seems like something is stealing processor cycles.  
And the behavior is not consistent, on different boots I get different 
errors at times.


Case 1: pv_ops domU kernel 2.6.32.10  (no xen dom0)

The kernel shows a lot of errors and just hangs.



Attached: console output for both cases plus menu.lst

-Gerry