Set bluetooth mac address on ESP32?

[harbaum]

I am trying to act like a fischertechnik toy controller. The matching other device will only connect to devices of a specific vendor, in this case LNT (vendor prefix 10:45:F8).

Can I change the BT MAC address through micropython? I tried:

bluetooth.BLE().config(mac=(0, b'\x01\x02\x03\x04\x05\x06'))

But that results in:

ValueError: unknown config param

[puppet13th]

i think it's not possible to set specific bluetooth mac address. check doc about 'addr_mode'

[harbaum]

It seems the esp-idf only recently got the API to change individual mac addresses:

https://docs.espressif.com/projects/esp-idf/en/latest/esp32/api-reference/system/misc_system_api.html#mac-address
esp_err_t esp_iface_mac_addr_set(const uint8_t *mac, esp_mac_type_t type)

It's not even in the 5.0 release. It seems there were ways to change single or all mac on previous versions as well but this seems to be the first time there's a clean API to do this.

[puppet13th]

If custom mac address is supported in IDF, there is need to be implemented in mpy function. @dpgeorge @jimmo any comment on this?

[harbaum]

I really don't know how important this use case is. On the other hand many MPY/esp32 are probably used to emulate existing devices. And this may come with the need to change mac addresses.

[harbaum]

I just tested to add support for esp_base_mac_addr_set() to MPY and it works just fine. Downside of this approach is that it needs to be done before any network related code is run and it affects all mac addresses (Wifi station, wifi ap, bluetooth and ethernet).

I'll append a patch here.

[shr00mie]

Looks like there's more than one way to skin this particular cat:

• old api
  • esp_base_mac_addr_set()
    • sets the base mac address (for the session, from the looks of it), from which the other interface MACs are derived
• new api
  • esp_iface_mac_addr_set()
    • replacement for old API. used in conjunction with ESP_MAC_BASE argument to set the base MAC for the device (again, seems like for the duration of the session).
  • esp_netif_set_mac()
    • this seems to be the underlying function we would want exposed when calling IFACE.config(mac=''), especially because:
      "It is also possible to call the function esp_netif_set_mac() to set the specific MAC used by a network interface after network initialization."
    • better yet, introduce logic that checks if the interface in question, or any interface for that matter, has been initialized, and call esp_iface_mac_addr_set() with ESP_MAC_BASE if no interfaces have yet to be initialized, else fall back to esp_netif_set_mac(), which, at least per espressif's documentation, can be called even if the interface is already up.
    • i take that back.
      • seems like esp_iface_mac_addr_set() should be used when you want to, for example, maintain a certain manufacturer prefix, but don't care what the incremented interface addresses might be
        • this should probably be set somewhere in boot.py and be implemented under machine, as you have already done.
      • after initializing a given interface, if the config method is called with the mac argument, expose esp_netif_set_mac() should be exposed underneath, allowing those MACs to be set regardless of whether the associated interface has been brought up or not.

it's also entirely possible i have no idea what i'm saying.

LiNkY

p.s. it just seems like esp_iface_mac_addr_set() has the potential to be considerably more problematic due to, at least based on my reading, it setting the base device MAC, instead of the desired interface MAC, due to:

i suppose it's all well and good if you only ever plan on setting a custom mac for the esp32 as a wifi client, as that just happens to mirror the burnt in and/or custom base MAC, but for the remaining interfaces, it would seem that one would have to take the desired interfaces' last octet offset into consideration with additional code to offset the offset, which sounds about as appealing as jumping head first into a dumpster fire.

functionality has been around for a while. doesn't show up if you just change the esp-idf version via the dropdown. gotta go dig around for it. being able to set the hardware address on an interface seems like something fundamental, and has probably been around considerably longer than v4.4.

custom mac esp-idf v4.4

[harbaum]

So here's the little patch that allows to read and write the base mac address on esp32.

modmachine_mac_addr.patch.txt

Initially I hoped that I could use this to first read the base mac, then modify it and write it back. Turned out that you cannot read the base MAC before you set it. So you cannot use that to read the base mac set by the system itself. But unique_id() returns exactly that. So adding this to boot.py will move the mac address into the LNT range:

# modify mac for LNT vendor 10:45:F8 
mac = machine.unique_id()
machine.base_mac_addr(b'\x10\x45\xf8' + mac[3:])

Afterwards the device shows up using the 10:45:f8 vendor prefix for wifi as well as bluetooth.

[Temgle]

I find a way to set the first byte of mac address without restriction:
comment out the code line 152-155 in (esp-idf\components\esp_hw_support\mac_addr.c)

[luxlux]

So possibly this is a way to go. ... but keep in mind, this limiting rule could possibly make sense, because i guess also other things have to be fulfilled from ESP-IDF, for a multicast MAC address, which are possibly not fulfilled in the IDF code, so that they disallowed the use of such multicast MAC addresses as Base Mac Address.

[ctschach]

This is awesome....can someone create a full, updated patch for the current MicroPython version. Maybe even a pull request so that this goes into the official release?!

[luxlux]

i would say no. this patch works. but it is not really consequent to put it into "machine". for a patched solution this ok. But for a pull to the offical release, it would make more sense, to have realize this it in "bluetooth" or in "aioble".

[shr00mie]

Ideally, functionality should be exposed for interfaces to mirror espressif source. This hacky patching of machine address by offset to get the desired ble address is just silly. For some IOT functionality, being able to take advantage of address space in a direct manner is critical (beacons for example).

[shr00mie]

alright. so. i think i did a thing. i mean it's ghetto and i'd really love for this to be implemented the way it should be, at the BLE level, but this should work in the meantime.

per the documentation, given the base device MAC, the last octet of the BLE MAC is offset by 2 bits.

which presents itself as an issue if you try to set your desired BLE MAC using the base_mac_addr method:

from machine import base_mac_addr
import bluetooth as bt

# let's say this is the desired BLE mac
mac = bytearray([0x1C, 0x88, 0x3F, 0xE5, 0x7E, 0x8D])

# set device base mac addr
base_mac_addr(mac)

ble = bt.BLE()
ble.active(True)
ble.config('mac')
(0, b'\x1c\x88?\xe5~\x8f')

which can be programmed around like so:

from machine import base_mac_addr
import bluetooth as bt

mac = bytearray([0x1C, 0x88, 0x3F, 0xE5, 0x7E, 0x8D])

def base_ble_offset(mac):
	mac[5] = mac[5] - 2
	return mac

base_mac_addr(base_ble_offset(mac))

ble = bt.BLE()
ble.active(True)
ble.config('mac')
(0, b'\x1c\x88?\xe5~\x8d')

did i do it right?

[harbaum]

This offset is by purpose. The ESP32 needs several MACs for WiFi, Bluetooth etc and it derives them from the base address by just counting up.

[shr00mie]

...yeah. i get that part. the point i was trying to make was that it would be nice to be able to set the BT MAC directly via Python, similarly to how the capability is exposed with WiFi, instead of having to set the base address, with additional code in order to account for the offset.

[ctschach]

Is there a chance this patch will make it into the official MicroPython distribution?!