Re: [Xen EFI] Impossible to limit the dom0 memory

[Juergen Gross <jgross@suse.com>]

On 25/01/18 11:40, Jan Beulich wrote:
>>>> On 25.01.18 at 11:25, <msd+xen-devel@msd.im> wrote:
>>>  Xen doesn't crash at all.
> 
> Did you read the above? it is being confirmed by ...
> 
>> With this file, it works, Xen boots :
>>
>> ```
>> [global]
>> default=xen
>>
>> [xen]
>> options=loglvl=all com1=115200,8n1 console=com1,vga
>> kernel=vmlinuz root=/dev/md2 ro rootdelay=10 noquiet nosplash 
>> earlyprintk=xen
>> ramdisk=initrd.img
>> ```
>>
>> With this file, I have just added "dom0_mem=1G,max:1G", Xen crashes :
>>
>> ```
>> [global]
>> default=xen
>>
>> [xen]
>> options=loglvl=all com1=115200,8n1 console=com1,vga dom0_mem=1G,max:1G
>> kernel=vmlinuz root=/dev/md2 ro rootdelay=10 noquiet nosplash 
>> earlyprintk=xen
>> ramdisk=initrd.img
>> ```
>>
>> I attached the boot logs "dom0_crash_with_dom0_memory.txt". The last 
>> line is "(XEN) Hardware Dom0 crashed: rebooting machine in 5 seconds."
> 
> ... this line you observe. Taking further lines from the log you've
> provided subsequently:
> 
> (XEN) d0v0: unhandled page fault (ec=0003)                                                                                                                                                    
> 
> This tells us that Dom0 didn't install a page fault handler yet.
> 
> (XEN) Pagetable walk from ffffffff82565200:                                                                                                                                                   
> (XEN)  L4[0x1ff] = 000000084600e067 000000000000200e                                                                                                                                          
> (XEN)  L3[0x1fe] = 000000084600f067 000000000000200f                                                                                                                                          
> (XEN)  L2[0x012] = 0000000846647067 0000000000002647                                                                                                                                          
> (XEN)  L1[0x165] = 0010000846565065 0000000000002565                                                                                                                                          
> 
> This tells us (together with the page fault error code) that the
> Dom0 kernel tried to provide memory as kernel stack which
> can't be written. This may be a Dom0 kernel stack overflow,

Really? Why do you think this is related to the stack?

The error code just tells us it was a write access. %rsp is
ffffffff82003cb0, while %cr2 is ffffffff82565200.

To me it seems as if the access was via %rbx.

It would really help to know which function caused the access. As the
stack contains some parts of the last messages issued I suspect this
is related to EFI code.

> but there may also be other reasons. At this point I can't
> exclude there being some root cause in Xen, but the issue
> needs to be investigated from the Dom0 kernel side.

So having a more readable backtrace would really help.

Guillaume, can you try to get symbol+offset for the values on the stack
looking like kernel code addresses (e.g. everything starting with
"ffffffff82")?


Juergen

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel