Re: [Qemu-devel] [RFC] qxl: don't panic on phys2virt

[Gerd Hoffmann]

On 03/30/12 21:27, Alon Levy wrote:
> On Fri, Mar 30, 2012 at 10:09:42AM +0200, Gerd Hoffmann wrote:
>> On 03/29/12 22:56, Alon Levy wrote:
>>> Issues a qxl_guest_bug from qxl_phys2virt. Everywhere else will fail.
>>
>> Nice start (there are more guest-triggerable asserts to be fixed this way).
>>
>>> qxl_phys2virt requires an additional argument because all it's possible
>>> return values are legit (well, I could use the fact it returns a pointer
>>> so it should be word aligned but I don't want to go there, this is
>>> totally internal).
>>
>> Really?  What is bad with returning and checking for NULL?  Guest
>> memslots are backed by guest ram and should never ever resolve to a NULL
>> pointer in qemu's process address space.
> 
> phys2virt actually returns an offset, not a real pointer iirc. I'll
> check it again.

It is a real pointer.
Do you have an updated patch?

cheers,
  Gerd