about .dump_station() callback

about .dump_station() callback

[Arend van Spriel]

Hi Johannes,

I am looking in get_station() and consequently dump_station() callback 
for our brcmfmac driver. The latter is called with index. Can I expect 
this to be starting at 0 and iterate through or can it be requested in 
any order.

Regards,
Arend

Re: about .dump_station() callback

[Johannes Berg]

On Tue, 2015-06-09 at 11:19 +0200, Arend van Spriel wrote:

> I am looking in get_station() and consequently dump_station() callback 
> for our brcmfmac driver. The latter is called with index. Can I expect 
> this to be starting at 0 and iterate through or can it be requested in 
> any order.

You can expect that, yes.

If it would help, we could instead give the driver one or two pointers
to unsigned long values that start out as 0 and can be controlled
entirely by the driver?

johannes

Re: about .dump_station() callback

[Arend van Spriel]

On 06/09/15 11:41, Johannes Berg wrote:
> On Tue, 2015-06-09 at 11:19 +0200, Arend van Spriel wrote:
>
>> I am looking in get_station() and consequently dump_station() callback
>> for our brcmfmac driver. The latter is called with index. Can I expect
>> this to be starting at 0 and iterate through or can it be requested in
>> any order.
>
> You can expect that, yes.

Thanks. Guess it is nl80211.c assuring that. I should have checked that 
first. See that I have to return -ENOENT when iteration is done.

> If it would help, we could instead give the driver one or two pointers
> to unsigned long values that start out as 0 and can be controlled
> entirely by the driver?

Not sure if I understand this. What would those pointers represent?

Regards,
Arend

Re: about .dump_station() callback

[Johannes Berg]

On Tue, 2015-06-09 at 12:52 +0200, Arend van Spriel wrote:

> Thanks. Guess it is nl80211.c assuring that. I should have checked that 
> first. See that I have to return -ENOENT when iteration is done.

Right.

> > If it would help, we could instead give the driver one or two pointers
> > to unsigned long values that start out as 0 and can be controlled
> > entirely by the driver?
> 
> Not sure if I understand this. What would those pointers represent?

Well, you could put in there whatever you like :)

Though realistically I wouldn't expect you to put anything other than
the sta_idx that you currently get.

johannes

Re: about .dump_station() callback

[Arend van Spriel]

On 06/09/15 13:01, Johannes Berg wrote:
> On Tue, 2015-06-09 at 12:52 +0200, Arend van Spriel wrote:
>
>> Thanks. Guess it is nl80211.c assuring that. I should have checked that
>> first. See that I have to return -ENOENT when iteration is done.
>
> Right.
>
>>> If it would help, we could instead give the driver one or two pointers
>>> to unsigned long values that start out as 0 and can be controlled
>>> entirely by the driver?
>>
>> Not sure if I understand this. What would those pointers represent?
>
> Well, you could put in there whatever you like :)
>
> Though realistically I wouldn't expect you to put anything other than
> the sta_idx that you currently get.

Well, I obtain (in ap mode) a list of associated clients and keep that 
list until iteration is done (or .dump_station() is called with sta_idx 
== 0). So I am fine with current api.

Regards,
Arend