Skip to main content

v2n-secure-element-sign

Talk to the OPTIGA Trust M on V2N's BRD_I2C from a Linux/Yocto user-space app on the Cortex-A55 cluster: probe it, read its Coprocessor UID, and run a raw APDU session. The driver runs Infineon's host library through a PAL on the portable <alp/peripheral.h> I²C and timing calls. Nothing here writes to the chip — despite the directory name, this example does not sign anything.

Source: examples/v2n/v2n-secure-element-sign/.

Moved from the M33 to the A55 in alp-sdk v0.17.0

BRD_I2C is A55/Linux-exclusive — the Cortex-M33 must never master it. Earlier releases ran this example on m33_sm and issued a hand-rolled CalcSign APDU; the v0.17.0 example runs on the A55 and is read-only.

What it shows​

  1. Open BRD_I2C (bus_id = 8, Linux /dev/i2c-8) at 400 kHz and initialise optiga_trust_m_t. optiga_trust_m_init_with_reset reads the I2C_STATE register only.
  2. Reset hook for an idled part. A Trust M idle for more than about 10 s can stop ACKing until it is hardware-reset. SE_RST is driven by the GD32 supervisor, which the kernel's alplab,gd32-bridge-gpio driver owns, so the app finds the se-rst line on the gd32-bridge-gpio gpiochip and passes it to the driver as the reset hook. If the part is silent the driver pulses SE_RST once and probes again. An image whose kernel does not export the se-rst line gets no hook (plain probe).
  3. optiga_trust_m_read_product_info opens the Trust M application and reads the 27-byte Coprocessor UID (data object 0xE0C2).
  4. optiga_trust_m_send_apdu runs a raw session: the app sends OpenApplication itself, then GetDataObject(0xE0C2), and checks the bytes match step 3.

Failing after the reset retry means the chip is not on the bus, is not strapped to address 0x30, or is held in reset.

board.yaml​

som:
sku: E1M-V2N101

preset: e1m-x-evk

cores:
a55_cluster:
app: ./src
peripherals:
- i2c

chips:
- optiga_trust_m

diagnostics:
log_level: info

Expected output (OPTIGA-populated SoM)​

[se] I2C_STATE probe -> ALP_OK
[se] UID: cim CD platform 16 model 33 fw 80101071 build 2564
[se] raw APDU UID matches (27 bytes)
[se] RESULT PASS: Trust M probe, Coprocessor UID and raw APDU session work

Expected output (missing or held-in-reset chip)​

[se] RESULT FAIL: optiga_trust_m_init (Trust M not ACKing) -> -2

Wiring this into a real app​

The raw APDU session reaches every Trust M command (key generation, ECDSA, secure NVM) today, with the caller owning the command bytes and the application session. Typed calls for those commands, and a PSA driver registered with <alp/security.h>'s MbedTLS wrapper so the chip is used transparently, are still to come.

See also​

Questions about this page? Email contact@alplab.ai