Re: [RFC PATCH V2 1/8] xen/mem_event: Cleanup of mem_event structures

[Razvan Cojocaru <rcojocaru@bitdefender.com>]

On 01/22/2015 02:50 PM, Tamas K Lengyel wrote:
> On Thu, Jan 22, 2015 at 1:43 PM, Tim Deegan <tim@xen.org> wrote:
>> Hi,
>>
>> At 16:17 +0100 on 18 Jan (1421594274), Tamas K Lengyel wrote:
>>> From: Razvan Cojocaru <rcojocaru@bitdefender.com>
>>>
>>> The public mem_event structures used to communicate with helper applications via
>>> shared rings have been used in different settings. However, the variable names
>>> within this structure have not reflected this fact, resulting in the reuse of
>>> variables to mean different things under different scenarios.
>>>
>>> This patch remedies the issue by clearly defining the structure members based on
>>> the actual context within which the structure is used.
>>>
>>> Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com>
>>> Signed-off-by: Tamas K Lengyel <tamas.lengyel@zentific.com>
>>
>> This looks like a nice improvement.  If everyone who commented about
>> making ABI changes is happy with it, I am too.
>>
>> It would be nice if you could add some comments to the new struct
>> definitions saying what the various fields mean (e.g. when the event
>> triggers and what the fields will contain).
> 
> Ack, will do!
> 
>>
>> One nit in the patch itself:
>>
>>> @@ -592,13 +592,12 @@ int main(int argc, char *argv[])
>>>                  }
>>>
>>>
>>> -                rsp.gfn = req.gfn;
>>> -                rsp.p2mt = req.p2mt;
>>> +                rsp.mem_access_event.gfn = req.mem_access_event.gfn;
>>
>> You're dropping a p2mt update here.  Is that an oversight?
>> With the obvious equivalent update inserted,
> 
> No, it is not. That field is only used by mem_sharing and mem_paging,
> not by mem_access.

Ah, I stand corrected. I've checked agains my original patch, but the
series also separates the mentioned areas.


Sorry for the noise,
Razvan