[PATCH 01/44] mtd: add new API for handling MTD registration

[Dmitry Eremin-Solenikov]

On 6/8/11, Jamie Iles <jamie at jamieiles.com> wrote:
> On Tue, Jun 07, 2011 at 07:48:58PM +0400, Dmitry Eremin-Solenikov wrote:
>> Lots (nearly all) mtd drivers contain nearly the similar code that
>> calls parse_mtd_partitions, provides some platform-default values, if
>> parsing fails, and registers  mtd device.
>>
>> This is an aim to provide single implementation of this scenario:
>> mtd_device_parse_register() which will handle all this parsing and
>> defaults.
>>
>> Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov at gmail.com>
>> ---
>>  drivers/mtd/mtdcore.c   |   47
>> +++++++++++++++++++++++++++++++++++++++++++++++
>>  include/linux/mtd/mtd.h |    5 +++++
>>  2 files changed, 52 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
>> index c510aff..d538e0a 100644
>> --- a/drivers/mtd/mtdcore.c
>> +++ b/drivers/mtd/mtdcore.c
>> @@ -452,6 +452,53 @@ int mtd_device_register(struct mtd_info *master,
>>  EXPORT_SYMBOL_GPL(mtd_device_register);
>>
>>  /**
>> + * mtd_device_parse_register - register an MTD device.
>> + *
>> + * @mtd: the MTD device to register
>> + * @part_probe_types: the list of MTD partition probes to try
>> + * @origin: start address of MTD device. =0 unless you are sure you need
>> this.
>> + * @defparts: default partition information to register. Only valid if
>> + *		defnr_parts > 0
>> + * @defnr_parts: the number of partitions in defparts.  If zero then the
>> full
>> + *		MTD device is registered if no partition info is found
>> + *
>> + * Extract partition info and register MTD device (partitions or a whole
>> device)
>> + * It calls parse_mtd_partitions(), checks the result. If there were no
>> + * partitions found, it uses default info specified (via
>> defparts/defnr_parts)
>> + * and then registers either partitions, or (if no partitions were
>> + * found/specified) the whow MTD.
>
> s/whow/whole ?

Yup.

>> + */
>> +int mtd_device_parse_register(struct mtd_info *mtd,
>> +			      const char **part_probe_types,
>> +			      unsigned long origin,
>> +			      const struct mtd_partition *defparts,
>> +			      int defnr_parts)
>> +{
>> +	int err;
>> +	struct mtd_partition *parts;
>> +
>> +	err = parse_mtd_partitions(mtd, part_probe_types, &parts, origin);
>> +	if (err <= 0 && defnr_parts) {
>> +		parts = kmemdup(defparts, sizeof(*parts) * defnr_parts,
>> +				GFP_KERNEL);
>> +		if (!parts)
>> +			err = -ENOMEM;
>> +	}
>
> I see that some of your patches for drivers (e.g. plat_nand) remove the
> kfree() of the partitions but this doesn't appear to be handled anywhere
> else afaict.  I don't think this can be done in mtd_device_unregister()
> as it doesn't have the mtd_partition array so probably needs to remain
> in the driver.

It's handled by kfree right aftert add_mtd_partitions. No mtd partitions
information is exported out of mtd_device_parse_register(). In fact I'll have
to check if I can remove all <mtd/partitions.h> includes from drivers.

Anyway I'll post a V2 later after fixing issues identified by you and Artem.

-- 
With best wishes
Dmitry