[PATCH 3/5] staging: vme: add functions for bridge module refcounting

Martyn Welch martyn.welch at ge.com
Fri Aug 19 08:42:14 UTC 2011


On 19/08/11 09:32, Martyn Welch wrote:
> On 13/08/11 09:54, Emilio G. Cota wrote:
>> On Fri, Aug 12, 2011 at 12:30:49 +0200, Manohar Vanga wrote:
>>> This patch adds functions that allow for reference counting
>>> bridge modules. The patch introduces the functions
>>> 'vme_bridge_get()' and 'vme_bridge_put()'.
>> (snip)
>>> +int vme_bridge_get(unsigned int bus_id)
>> (snip)
>>> +void vme_bridge_put(struct vme_bridge *bridge)
>>
>> Note the input parameter imbalance; in fact this is serious
>> (see my email on patch 5) because _get() needs to acquire
>> vme_buses_lock, whereas _put() doesn't. Since a caller with
>> bridge has bridge->num, but the opposite doesn't hold (num
>> doesn't give you the bridge without acquiring vme_buses_lock),
>> it seems reasonable to me to take the bus_id as the input for
>> both functions, because the requirements on the caller are lower.
>>
> 
> Patch 4 makes changes the struct vme_bridge to struct vme_dev. Looking at the
> callers we are then effectively doing:
> 

Patch 5 even.

Martyn

> vme_bridge_get(vme_dev.id)
> 
> Then in vme_bridge_get(), looping through all the buses to find the one with
> the correct id...
> 
> We could just pass in the struct vme_dev to both functions.
> 
> Martyn
> 
>> But the locking needs to be handled with care, see my reply
>> to patch 5.
>>
>> 		Emilio
>>
>> _______________________________________________
>> devel mailing list
>> devel at linuxdriverproject.org
>> http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
> 
> 


-- 
Martyn Welch (Principal Software Engineer) | Registered in England and
GE Intelligent Platforms                   | Wales (3828642) at 100
T +44(0)127322748                          | Barbirolli Square, Manchester,
E martyn.welch at ge.com                      | M2 3AB  VAT:GB 927559189



More information about the devel mailing list