Re: [Xen-devel] [PATCH RFC] Persistent grant maps for xen blk drivers

[Roger Pau Monné]

On 23/10/12 20:50, Konrad Rzeszutek Wilk wrote:
> On Tue, Oct 23, 2012 at 08:09:27PM +0200, Roger Pau Monné wrote:
>> On 23/10/12 19:20, Konrad Rzeszutek Wilk wrote:
>>>>>> diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
>>>>>> index c6decb9..2b982b2 100644
>>>>>> --- a/drivers/block/xen-blkback/blkback.c
>>>>>> +++ b/drivers/block/xen-blkback/blkback.c
>>>>>> @@ -78,6 +78,7 @@ struct pending_req {
>>>>>>       unsigned short          operation;
>>>>>>       int                     status;
>>>>>>       struct list_head        free_list;
>>>>>> +     unsigned int            unmap_seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
>>
>> Should I change this to a bool? Since we are only setting it to 0 or 1.
> 
> I would just keep it as 'int'. Eventually we can replace this with a
> bit-map, but that can be done later.

I've already changed it to a bitmap.

>>>>> Perhaps there should be a #define for that array..
>>>>
>>>> Do you mean something like:
>>>>
>>>> #define unmap(req, i) req->unmap_seg[i]
>>>
>>> I was thinking that you just check for req->unamp_seg[i] to
>>> have an non-zero value. But since that array is just used as an check
>>> to see whether the functionality is enabled (or not), you might want
>>> to declerare the right values so:
>>> #define UNMAP_SG_ON 1
>>> #define UNMAP_SG_OFF 0
>>>
>>> or so.
>>
>> Agreed, will add the defines.
>>
>>>>>> +             if (persistent_gnts[i]) {
>>>>>> +                     if (!persistent_gnts[i]->handle) {
>>>>>> +                             /*
>>>>>> +                              * If this is a new persistent grant
>>>>>> +                              * save the handler
>>>>>> +                              */
>>>>>> +                             persistent_gnts[i]->handle = map[j].handle;
>>>>>> +                             persistent_gnts[i]->dev_bus_addr =
>>>>>> +                                     map[j++].dev_bus_addr;
>>>>>> +                     }
>>>>>> +                     pending_handle(pending_req, i) =
>>>>>> +                             persistent_gnts[i]->handle;
>>>>>> +                     pending_req->unmap_seg[i] = 0;
>>>>>
>>>>> Could we have a #define for that?
>>>>
>>>> Sure.
>>
>> I've used the previous macro, so it looks like:
>>
>> unmap(req, i) = UNMAP_SG_OFF;
>>
>> I'm not sure if this is what you meant, or if you where interested in
>> defining a set of macros like:
>>
>> #define check_unmap(req, i) req->unmap_seg[i]
>> #define unset_unmap(req, i) req->unmap_seg[i] = UNMAP_SG_OFF
>> #define set_unmap(req, i) req->unmap_seg[i] = UNMAP_SG_ON
>>
>> I would go for the first option (the unmap macro that can be used here
>> and in xen_blkbk_unmap).
> 
> I was just thinking something as simple as
> 
> 	if (reg->unmap_seg[i] == UNMAP_SG_OFF)
> 		continue;
> 
> And the #defines are just for the hard-coded values of 0 or 1.
> 
>>
>>>>> HA! By default, eh?
>>>>
>>>> Yes, you caught me, there's a paragraph in the commit message that
>>>> explains that we are using persistent grants in the frontend
>>>> unconditionally, since the protocol is compatible (you can have a
>>>> persistent blkfront and a non-persistent blkback). It simplifies the
>>>> logic in blkfront. Are you OK with it?
>>>
>>> It is OK, but you should be checking whether the backend supports it.
>>> I don't see it checking the info->feature_persistent_grant to print
>>> that.
>>
>> I don't understand why blkfront needs to check if the backend supports
>> persisten grants, blkfront is going to use persistent grants anyway.
> 
> What if it does not (say this guest runs on an older xen-blkback?)?
> Then you would be still printing 'persistent grants' in the blkfront.

Ok, I get your point. Now blkfront will only report the use of
persistent grants if the backend supports it.

If there are no further comments I will send v2 after doing some tests,
thanks for the reviews.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/