Hi Sascha, On 23-08-08, Sascha Hauer wrote: > On Mon, Aug 07, 2023 at 07:07:43PM +0200, Marco Felsch wrote: > > Some vendors like Polyhex store the MAC address ASCII encoded instead of > > using the plain 6-byte MAC address. This commit adds the support to > > decode the 12-byte ASCII encoded MAC addresses. > > The upstream i.MX8MP dtsi files have "mac-address" nvmem cells described > in the device trees, but they point to a 6-byte long cell in ocotp. > These cells are not overwritten in the Polyhex dts files. How can there > be a 12-byte ASCII stored?
Please have a look at: https://lore.kernel.org/all/[email protected]/ and search for ethmac{1,2}. Once the devicetree is upstream I will sync our internal -upstream variant. > > Signed-off-by: Marco Felsch <[email protected]> > > --- > > drivers/of/of_net.c | 19 +++++++++++++++++++ > > 1 file changed, 19 insertions(+) > > > > diff --git a/drivers/of/of_net.c b/drivers/of/of_net.c > > index 75a24073da51..4e74986cdda8 100644 > > --- a/drivers/of/of_net.c > > +++ b/drivers/of/of_net.c > > @@ -79,6 +79,8 @@ static int of_get_mac_addr(struct device_node *np, const > > char *name, u8 *addr) > > return -ENODEV; > > } > > > > +#define ETH_ALEN_ASCII 12 > > + > > int of_get_mac_addr_nvmem(struct device_node *np, u8 *addr) > > { > > struct nvmem_cell *cell; > > @@ -98,6 +100,23 @@ int of_get_mac_addr_nvmem(struct device_node *np, u8 > > *addr) > > if (IS_ERR(mac)) > > return PTR_ERR(mac); > > > > + if (len == ETH_ALEN_ASCII) { > > I don't like this heuristic very much. If I understand the nvmem stuff > correctly then parsing of properties in non standard formats should be > fixed in a struct nvmem_cell_info::read_post_process hook. IMHO there is no standard to store MAC addresses, there is an easy way (raw address stored in 6-bytes in some nvmem reachable from the host) and a vendor-know-it-better way. While coding I was thinking about a property to indicate that the mac-address is stored in ascii like: &eeprom { macaddr1: mac-address@0 { reg = <0 0xc>; barebox,ascii-mac-address; } } Then I thought, if someone stores the mac-address in a 12-byte field this have to be an ascii encoded mac-address and I dropped the barebox,ascii-mac-address property. > > + u8 *mac_new; > > + int ret; > > + > > + mac_new = kzalloc(sizeof("xx:xx:xx:xx:xx:xx"), GFP_KERNEL); > > If anything, then sizeof("xxxxxxxxxxxx"), but what you want here is > ETH_ALEN. You're right. Thanks. Regards, Marco
