Re: [Xenomai] __get_user/__put_user

[Jan Kiszka <jan.kiszka@web.de>]

On 2013-05-09 21:16, Gilles Chanteperdrix wrote:
>>>> But why do we need to hold the nklock across the __rt_bind_helper in
>>>> __rt_queue_bind at all? Other binding services apparently don't have
>>>> this need.
>>>
>>>
>>> I suppose because we do not want the heap vanishing under our feet while
>>> we access its contents with the xnheap_* accessors.
>>
>> Right, but the general pattern for this is
>>
>> 	obj = xnregistry_fetch(handle)
>> 	get_nklock()
>> 	if (valid(obj))
>> 		do_something_with(obj)
>> 	put_nklock()
> 
> 
> Seems rather to be get_nklock before xnregistry_fetch.
> 

Only during heap and queue binding. All other native skin syscall
entries fetch from the registry without holding nklock. I suppose the
philosophy is that, if user space generates a race around picking up an
object and destroying / recreating a different one under the same name,
it will get what it deserves.

Jan


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 263 bytes
Desc: OpenPGP digital signature
URL: <http://www.xenomai.org/pipermail/xenomai/attachments/20130509/3834e7d7/attachment.pgp>