Date:   Fri, 10 May 2019 13:31:55 +0200
From:   Maxime Ripard <maxime.ripard@...tlin.com>
To:     Petr Štetiar <ynezz@...e.cz>
Cc:     Andy Duan <fugang.duan@....com>,
        "davem@...emloft.net" <davem@...emloft.net>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "john@...ozen.org" <john@...ozen.org>,
        "bgolaszewski@...libre.com" <bgolaszewski@...libre.com>,
        Srinivas Kandagatla <srinivas.kandagatla@...aro.org>,
        Andrew Lunn <andrew@...n.ch>,
        Florian Fainelli <f.fainelli@...il.com>,
        Heiner Kallweit <hkallweit1@...il.com>,
        Rob Herring <robh+dt@...nel.org>,
        Frank Rowand <frowand.list@...il.com>,
        Mark Rutland <mark.rutland@....com>,
        Alban Bedel <albeu@...e.fr>, devicetree@...r.kernel.org
Subject: Re: [PATCH net 0/3] add property "nvmem_macaddr_swap" to swap
 macaddr bytes order

On Fri, May 10, 2019 at 01:28:22PM +0200, Petr Štetiar wrote:
> Andy Duan <fugang.duan@....com> [2019-05-10 08:23:58]:
>
> Hi Andy,
>
> you've probably forget to Cc some maintainers and mailing lists, so I'm
> adding them now to the Cc loop. This patch series should be posted against
> net-next tree as per netdev FAQ[0], but you've to wait little bit as
> net-next is currently closed for the new submissions and you would need to
> resend it anyway.
>
> 0. https://www.kernel.org/doc/html/latest/networking/netdev-FAQ.html
>
> > ethernet controller driver call .of_get_mac_address() to get
> > the mac address from devictree tree, if these properties are
> > not present, then try to read from nvmem. i.MX6x/7D/8MQ/8MM
> > platforms ethernet MAC address read from nvmem ocotp eFuses,
> > but it requires to swap the six bytes order.
>
> Thanks for bringing up this topic, as I would like to extend the
> functionality as well, but I'm still unsure how to tackle this and where,
> so I'll (ab)use this opportunity to bring other use cases I would like to
> cover in the future, so we could better understand the needs.
>
> This reverse byte order format/layout is one of a few other storage formats
> currently used by vendors, some other (creative) vendors are currently
> providing MAC addresses in NVMEMs as ASCII text in following two formats
> (hexdump -C /dev/mtdX):
>
>  a) 0090FEC9CBE5 - MAC address stored as ASCII without colon between octets
>
>   00000090  57 2e 4c 41 4e 2e 4d 41  43 2e 41 64 64 72 65 73  |W.LAN.MAC.Addres|
>   000000a0  73 3d 30 30 39 30 46 45  43 39 43 42 45 35 00 48  |s=0090FEC9CBE5.H|
>   000000b0  57 2e 4c 41 4e 2e 32 47  2e 30 2e 4d 41 43 2e 41  |W.LAN.2G.0.MAC.A|
>
>   (From https://github.com/openwrt/openwrt/pull/1448#issuecomment-442476695)
>
>  b) D4:EE:07:33:6C:20 - MAC address stored as ASCII with colon between octets
>
>   00000180  66 61 63 5f 6d 61 63 20  3d 20 44 34 3a 45 45 3a  |fac_mac = D4:EE:|
>   00000190  30 37 3a 33 33 3a 36 43  3a 32 30 0a 42 44 49 4e  |07:33:6C:20.BDIN|
>
>   (From https://github.com/openwrt/openwrt/pull/1906#issuecomment-483881911)
>
> > The patch set is to add property "nvmem_macaddr_swap" to swap
> > macaddr bytes order.
>
> so it would allow following DT construct (simplified):
>
>  &eth0 {
> 	nvmem-cells = <&eth0_addr>;
> 	nvmem-cell-names = "mac-address";
> 	nvmem_macaddr_swap;
>  };
>
> I'm not sure about the `nvmem_macaddr_swap` property name, as currently there
> are no other properties with underscores, so it should be probably named as
> `nvmem-macaddr-swap`. DT specs permits use of the underscores, but the
> estabilished convention is probably prefered.
>
> In order to cover all above mentioned use cases, it would make more sense
> to add a description of the MAC address layout to the DT and use this
> information to properly postprocess the NVMEM content into usable MAC
> address?
>
> Something like this?
>
>  - nvmem-cells: phandle, reference to an nvmem node for the MAC address
>  - nvmem-cell-names: string, should be "mac-address" if nvmem is to be used
>  - nvmem-mac-address-layout: string, specifies MAC address storage layout.
>    Supported values are: "binary", "binary-swapped", "ascii", "ascii-delimited".
>    "binary" is the default.
>
> Or perhaps something like this?
>
>  - nvmem-cells: phandle, reference to an nvmem node for the MAC address
>  - nvmem-cell-names: string, should be any of the supported values.
>    Supported values are: "mac-address", "mac-address-swapped",
>    "mac-address-ascii", "mac-address-ascii-delimited".
>
> I'm more inclined towards the first proposed solution, as I would like to
> propose MAC address octet incrementation feature in the future, so it would
> become:
>
>  - nvmem-cells: phandle, reference to an nvmem node for the MAC address
>  - nvmem-cell-names: string, should be "mac-address" if nvmem is to be used
>  - nvmem-mac-address-layout: string, specifies MAC address storage layout.
>    Supported values are: "binary", "binary-swapped", "ascii", "ascii-delimited".
>    "binary" is the default.
>  - nvmem-mac-address-increment: number, value by which should be
>    incremented MAC address octet, could be negative value as well.
>  - nvmem-mac-address-increment-octet: number, valid values 0-5, default is 5.
>    Specifies MAC address octet used for `nvmem-mac-address-increment`.
>
> What do you think?

It looks to me that it should be abstracted away by the nvmem
interface and done at the provider level, not the customer.

Maxime

--
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)

Powered by blists - more mailing lists