Re: [PATCH] of_net: add mtd-mac-address support to of_get_mac_address()

[Frank Rowand <frowand.list@gmail.com>]

Hi Rob,

On 4/16/19 5:29 PM, Florian Fainelli wrote:
> 
> 
> On 16/04/2019 13:05, Petr Štetiar wrote:
>> From: John Crispin <john@phrozen.org>
>>
>> Many embedded devices have information such as MAC addresses stored
>> inside MTD devices. This patch allows us to add a property inside a node
>> describing a network interface. The new property points at a MTD
>> partition with an offset where the MAC address can be found.
>>
>> This patch has originated in OpenWrt some time ago, so in order to
>> consider usefulness of this patch, here are some real-world numbers
>> which hopefully speak for themselves:
>>
>>   * mtd-mac-address                used 497 times in 357 device tree files
>>   * mtd-mac-address-increment      used  74 times in  58 device tree files
>>   * mtd-mac-address-increment-byte used   1 time  in   1 device tree file
>>
>> Signed-off-by: John Crispin <john@phrozen.org>
>> Signed-off-by: Felix Fietkau <nbd@nbd.name>
>> [cleanup of the patch for upstream submission]
>> Signed-off-by: Petr Štetiar <ynezz@true.cz>
>> ---
> 
> [snip]
> 
>> +static const void *of_get_mac_address_mtd(struct device_node *np)
>> +{
>> +#ifdef CONFIG_MTD
>> +    void *addr;
>> +    size_t retlen;
>> +    int size, ret;
>> +    u8 mac[ETH_ALEN];
>> +    phandle phandle;
>> +    const char *part;
>> +    const __be32 *list;
>> +    struct mtd_info *mtd;
>> +    struct property *prop;
>> +    u32 mac_inc = 0;
>> +    u32 inc_idx = ETH_ALEN-1;
>> +    struct device_node *mtd_np = NULL;
> 
> Reverse christmas tree would look a bit nicer here.

Do we a variable declaration format preference for drivers/of/*?

-Frank


> 
>> +
>> +    list = of_get_property(np, "mtd-mac-address", &size);
>> +    if (!list || (size != (2 * sizeof(*list))))
>> +        return NULL;
>> +
>> +    phandle = be32_to_cpup(list++);
>> +    if (phandle)
>> +        mtd_np = of_find_node_by_phandle(phandle);
>> +
>> +    if (!mtd_np)
>> +        return NULL;
>> +
>> +    part = of_get_property(mtd_np, "label", NULL);
>> +    if (!part)
>> +        part = mtd_np->name;
>> +
>> +    mtd = get_mtd_device_nm(part);
>> +    if (IS_ERR(mtd))
>> +        return NULL;
>> +
>> +    ret = mtd_read(mtd, be32_to_cpup(list), ETH_ALEN, &retlen, mac);
>> +    put_mtd_device(mtd);
>> +
>> +    of_property_read_u32(np, "mtd-mac-address-increment-byte", &inc_idx);
> 
> of_property_read_u8() would probably be good here since this can't be bigger than 5 anyway.
> 
>> +    if (inc_idx > ETH_ALEN-1)
>> +        return NULL; >
>> +    if (!of_property_read_u32(np, "mtd-mac-address-increment", &mac_inc))
>> +        mac[inc_idx] += mac_inc;
> 
> If I use a number greater than and included 128; this will cause a roll over, should this be range checked? Similarly, using of_property_read_u8() might be a better fit?
> 
> Other than those, LGTM