Re: [Xen-devel] PROBLEM: kernel panic xsave_init

[John Doe <securef33d@gmail.com>]

On 20/10/2015 16:27, Boris Ostrovsky wrote:
> On 10/20/2015 09:43 AM, Jan Beulich wrote:
>>>>> On 20.10.15 at 15:22, <boris.ostrovsky@oracle.com> wrote:
>>> The reason I think its this commit is that RAX, RDX and RCX look very
>>> much like arguments to xsetbv (which xstate_enable_boot_cpu() executes)
>>> and RAX value is 0x1f, which has two new bits that this commit defined.
>> That would be the two MPX related bits, yet us (luckily) white listing
>> leaf 7 in pv_cpuid(), it is quite easy to validate that we don't expose
>> this feature to PV(H) guests.
> 
> Oh, so something like
> 
> cpuid=['0x7:ebx=xxxxxxxxxxxxxxxxx0xxxxxxxxxxxxxx']
> 
> (bit 14 as zero) for John to try then.
> 
> 
> -boris
> 
>> Sadly the story is different for HVM
>> guests (where the leaf handling uses black listing), but the register
>> dump here clearly points to a PV guest (or Dom0).
>>
>> Jan
>>
> 
Jan the dump is taken from serial connection to Dom0, it does crash
during boot.

I just tried with xen 4.6.0 and it booted properly without xsave=0.
Running gdb against /proc/kcore, with a x/10x 0xffffffff81d58fad i just
get null bytes, with both xen4.4.3 (xsave=0) and 4.6.0.
Tomorrow i will send you the gdb output and i will try to run it during
the boot process.

J.