Re: [PATCH 0/6] Add various undefined MMIO r/w functions

[Philippe Mathieu-Daudé]

On 6/17/20 4:05 PM, Alex Bennée wrote:
> 
> Philippe Mathieu-Daudé <philmd@redhat.com> writes:
> 
>> On 6/17/20 3:06 PM, Alex Williamson wrote:
>>> On Wed, 17 Jun 2020 16:39:56 +1000
>>> David Gibson <david@gibson.dropbear.id.au> wrote:
>>>
>>>> On Wed, Jun 17, 2020 at 11:09:27AM +0530, P J P wrote:
>>>>> From: Prasad J Pandit <pjp@fedoraproject.org>
>>>>>
>>>>> Hello,
>>>>>
>>>>> This series adds various undefined MMIO read/write functions
>>>>> to avoid potential guest crash via a NULL pointer dereference.  
>>>>
>>>> Hrm.  If this is such a common problem, maybe we should just add a
>>>> NULL check in the common paths.
>>>
>>> +1, clearly the behavior is already expected.  Thanks,
>>
>> 20 months ago Peter suggested:
>>
>> "assert that every MemoryRegionOps has pointers to callbacks
>>  in it, when it is registered in memory_region_init_io() and
>>  memory_region_init_rom_device_nomigrate()."
>>
>> https://www.mail-archive.com/qemu-devel@nongnu.org/msg573310.html
>>
>> Li Qiang refers to this post from Paolo:
>>
>>>  static const MemoryRegionOps notdirty_mem_ops = {
>>> +    .read = notdirty_mem_read,
>>>      .write = notdirty_mem_write,
>>>      .valid.accepts = notdirty_mem_accepts,
>>>      .endianness = DEVICE_NATIVE_ENDIAN,
>>
>> "This cannot happen, since TLB_NOTDIRTY is only added
>>  to the addr_write member (see accel/tcg/cputlb.c)."
>>
>> https://www.mail-archive.com/qemu-devel@nongnu.org/msg561345.html
> 
> What about catching it in memory_region_dispatch_write:
> 
>     if (mr->ops->write) {
>         return access_with_adjusted_size(addr, &data, size,
>                                          mr->ops->impl.min_access_size,
>                                          mr->ops->impl.max_access_size,
>                                          memory_region_write_accessor, mr,
>                                          attrs);
>     } else if (mr->ops->write_with_attrs) {
>         return
>             access_with_adjusted_size(addr, &data, size,
>                                       mr->ops->impl.min_access_size,
>                                       mr->ops->impl.max_access_size,
>                                       memory_region_write_with_attrs_accessor,
>                                       mr, attrs);
>     } else {
>         qemu_log_mask(LOG_UNIMP|LOG_GUEST_ERROR, "%s: %s un-handled write\n",
>                       __func__, mr->name);

The problem is what return value to return...
MEMTX_OK/MEMTX_ERROR/MEMTX_DECODE_ERROR? This is very
device-specific and can't be decided here for all the
cases.

Better to abort() and fix each device?

>     }
> 
>