[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] Re: Continuing problems booting



Gerd Hoffmann wrote:
> Jeremy Fitzhardinge wrote:
>> Gerd Hoffmann wrote:
>>> Jeremy Fitzhardinge wrote:
>>>  
>>>> Yes, I think the legacy interrupts are not being set up completely, but
>>>> I'm not quite sure how they should be set up.  Will look into it.
>>>>     
>>> FYI: Recently my dated, apic-less i386 laptop started to successfully
>>>      boot the pv_ops/dom0 kernel, all the way up to userspace.
>>>   
>> Do you get a vga console?  Can you start domains?
> 
> gfx console works (i.e. kernel /xen-3.3.gz vga=gfx-1024x768x16).
> vga text console didn't last time I tried.

Update: latest kernel (rc7 based) crashes.  rc6 from somewhen last week
is the working one.  rc7 messages:

unhandled page fault (ec=0003)
page table walk from c1c55000
 l1[0x055] = 9c55061 1c55

-> rw access to r/o page?

EIP c140f4c3

c140f2d2 <alloc_bootmem_core>:
[ ... ]
c140f4c3:       f3 ab                   rep stos %eax,%es:(%edi)

-> memset(page,0,PAGE_SIZE) ?

Dom0 domain builder says page tables are at c1c55000 -> c1c6a000

/me guesses the initial page tables are released to the page allocator,
but still they are mapped r/o  =>  boom as soon as one happens to get
allocated.  Which probably happens very soon on memory-constrained
machines like mine, while other might stay up longer and show strange
bugs later on ;)

BTW: The trick to see the messages on the laptop screen is:

 kernel /xen-3.3.gz vga=text-80x50,keep
 module /vmlinuz-2.6.29-rc7-tip-kraxel ro root=/dev/zen/rawhide \
        console=hvc0

>> x86-32 booting to usermode dom0, but only with serial console and domain
>> creation fails (SIGBUS in the domain builder, so I'm hoping its related
>> to the hvm qemu crash).
> 
> Didn't try yet, the machine is heavily underpowered for serious
> virtualization work, it has 192 MB RAM only.  And hvm doesn't work
> anyway because the box is way to old for that (Pentium III).

Doesn't work.  I get messages about failed multicalls with
remap_page_range and privcmd_ioctl in the stack trace.  Most likely
mapping the guest pages in the domain builder doesn't work.  No surprise
this leads to SIGBUS.

HTH,
  Gerd


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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.