From mboxrd@z Thu Jan  1 00:00:00 1970
From: swarren@wwwdotorg.org (Stephen Warren)
Date: Thu, 28 May 2015 16:09:12 -0600
Subject: [PATCH 2/3 v4] ARM: bcm2835: Add the Raspberry Pi firmware driver
In-Reply-To: <55678AFC.6020404@tronnes.org>
References: <20150528114500.GP11677@x1>
 <1432837987-22861-1-git-send-email-eric@anholt.net>
 <5567886F.5080108@wwwdotorg.org> <55678AFC.6020404@tronnes.org>
Message-ID: <55679208.3020509@wwwdotorg.org>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On 05/28/2015 03:39 PM, Noralf Tr?nnes wrote:
> 
> Den 28.05.2015 23:28, skrev Stephen Warren:
>> On 05/28/2015 12:33 PM, Eric Anholt wrote:
>>> This gives us a function for making mailbox property channel requests
>>> of the firmware, which is most notable in that it will let us get and
>>> set clock rates.
> ...
>>> +int rpi_firmware_property_list(struct device_node *of_node,
>>> +                   void *data, size_t tag_size)
>>> +{
>>> +    struct platform_device *pdev = of_find_device_by_node(of_node);
>> I would expect the of_node -> pdev mapping to happen at client device
>> probe time. Simplest for this driver would be if the
>> client-probe-time-mapping function returned the "struct rpi_firmware"
>> and the client passed that to this function.
> 
> What if the firmware driver/device is removed or reloaded?
> In that case the client has an invalid pointer. It wouldn't know about
> the change.

A provider can't be removed if it has clients. So, you'd have to remove
all clients first, then remove the provider. This may require adding
some code to increment/decrement the module use count when clients look
up the provider and are themselves removed.