Re: [PATCH v6 1/3] memory: drop guest writes to read-only ram device regions

[Philippe Mathieu-Daudé]

On 5/25/20 12:20 PM, Paolo Bonzini wrote:
> On 25/05/20 03:18, Yan Zhao wrote:
>> On Thu, May 21, 2020 at 04:38:47PM +0200, Paolo Bonzini wrote:
>>> On 30/04/20 11:40, Peter Maydell wrote:
>>>>> This does not "drop" a write to a r/o region -- it causes it to generate
>>>>> whatever the guest architecture's equivalent of a bus error is (eg data
>>>>> abort on Arm).
>>>
>>>
>>>> More generally, this change seems a bit odd: currently we do not
>>>> check the mr->readonly flag here, but in general guests don't get
>>>> to write to ROM areas. Where is that check currently done
>>>
>>> Writes to ROM are directed to mr->ops unassigned_mem_ops.  Because _all_
>>> ram-device reads and writes go through the ops, for ram-device we have
>>> to stick the check for mr->readonly in the ops.
>>>
>>> On one hand, I was quite surprised to see that unassigned_mem_write does
>>> not return MEMTX_ERROR now that I looked at it.
>>>
>>> On the other hand, we should use MEMTX_ERROR in patch 2 as well, if we
>>> decide it's the way to go.
>>>
>>> (Sorry Yan for the late response).
>>>
>> hi Paolo,
>> thanks for your reply and never mind :)
>>
>> But there's one thing I just can't figure out the reason and eagerly need
>> your guide.
>>
>> why do we have to convert all .write operations to .write_with_attrs and
>> return MEMTX_ERROR? because of the handling of writes to read-only region?
> 
> Not all of them, only those that need to return MEMTX_ERROR.  I would
> like some guidance from Peter as to whether (or when) reads from ROMs
> should return MEMTX_ERROR.  This way, we can use that information to
> device  what the read-only ram-device regions should do.

Is it only device-specific or might it be partly arch/machine-specific
(depending on the bus it is mapped)?

Phil.