[PATCH 1/3] spi: added spi_resource management

[Martin Sperl]

> On 02.12.2015, at 14:32, Mark Brown <broonie at kernel.org> wrote:
> 
> On Wed, Dec 02, 2015 at 02:04:24PM +0100, Martin Sperl wrote:
> 
>> The bigger question (based on your comments to Patch 2/3) is:
> 
> I haven't even looked at your reply ot that yet.
> 
>> Do you want to follow the devres approach (i.e: hiding
>> "struct spi_res" after allocation and returning "void *"
>> to the data-payload only in spi_res_alloc)?
> 
>> Or do you prefer to have "struct spi_res" as an explicit member of
>> a structure (i.e. in Patch 2/3 "struct spi_res_replaced_transfers")?
> 
> I wasn't aware that was an issue?

Here the section from your reply to the other patch I am referring to:
> On 01.12.2015, at 22:29, Mark Brown <broonie at kernel.org> wrote:
> 

>> 
>> +/* the spi_resource structure used */
>> +struct spi_res_replaced_transfers {
>> +	spi_res_release_t release;
>> +	struct list_head replaced_transfers;
>> +	int inserted;
>> +	struct spi_transfer xfers[];
>> +};
> 
> This quite clearly isn't a struct spi_resource, nor does it contain
> one…

As said: it could also get written as:
+/* the spi_resource structure used */
+struct spi_res_replaced_transfers {
+	struct spi_res resource;
+	spi_res_release_t release;
+	struct list_head replaced_transfers;
+	int inserted;
+	struct spi_transfer xfers[];
+};

But we loose the ability to just allocate memory the way we use
devm_kmalloc.

Martin