Repository navigation
STM32 bench-pod firmware v3.10.7
BenchPod firmware 3.10.7: flash ESP32 targets, Wi-Fi updates go straight into the Wi-Fi chip, and a security hardening pass. Gateware v55 is unchanged. Ships Wi-Fi bridge firmware 0.1.3.
New
-
Flash ESP32-family targets over their UART bootloader. The pod drives the chip's own ROM bootloader the way esptool does, with no esptool and no USB cable to the target. Wire the target's TXD0 and RXD0 to two pod pins, IO0 (IO9 on the C3/C6) to a third, and EN to the pod's reset pin. Use
benchpod flash --target esp32 --rx <pin> --tx <pin> --boot <pin> --nreset(CLI 0.8.11) orflash(target="esp32", ...)(SDK 2.13.19). A.binor an ESP-IDFflasher_args.jsonworks, and the chip MD5-verifies every region. Written at 460800 baud. The ESP8266 is not supported. -
adc_readreads to the microvolt. The reply has a newuvfield.mvrounds to a whole millivolt, which hid most of what the ADC low range resolves (about 167 µV per count, and the reading's 16-sample average goes below a count). SDK 2.13.20 uses it foradc_read().
Changed
-
Wi-Fi updates go straight into the Wi-Fi chip. The pod no longer keeps a copy of that image in its external flash. An update is verified in memory, then written into the chip in the background (about 2 minutes; Wi-Fi is down meanwhile, and the update status reads
applying). Boards without the Wi-Fi chip no longer accept or store the image. CLI 0.8.11 waits through the write. -
The ±12 V output stays within ±5 V while the ADC is in the low range (analog module v3r4). A level past ±5 V on that output is refused, so are waveforms, replays and the control loop on it, and switching to the low range is refused while the output holds more than ±5 V. The refusal says what to do. In the normal range the output can still go to ±12 V; the other outputs and calibration are unchanged.
Security
- Updates over Ethernet and Wi-Fi need a good signature, whatever the signature policy says. Every published release is signed, so updating to a release works as before. An unsigned or developer build over the LAN is refused with
LAN needs a signed image; use USB. Install it over USB or the cloud instead, or allow it on the device (lan-ota anyon the console, orlan_unsignedover USB). - Wi-Fi chip images must be signed, from every connection, like the Wi-Fi bridge's firmware.
- LAN TLS settings (certificate upload, regenerate, HTTPS-only on or off) can be changed only over USB, the console or the cloud, not from Ethernet or Wi-Fi.
- The USB link is told apart by the interface a packet arrives on, so a LAN host can no longer reach the USB-only rights through the pod's Ethernet address.
- Genuine Wi-Fi bridges only. The pod sends Wi-Fi credentials only to a bridge that proves it is genuine: bridge firmware 0.1.3 with an identity certificate signed with our release key. A bridge without one keeps working with the credentials it already has, but the pod will not send it new ones.
- Wi-Fi bridge 0.1.3: learns the pod's address only from signed requests, the PAIR button also forgets the Wi-Fi credentials, and it refuses updates to an older version.
- A rollback floor for signed firmware exists but no release sets it yet.
- The console's
dfuis refused when the signature policy isrequired, release firmware no longer falls back to the old cloud login, and several command parsers check their lengths more strictly.
Tested
- ESP32 flashing on a BenchPod analog v3r2 with an ESP32-WROOM-32 board: 32 KB written and MD5-verified at 115200 and 460800 baud through the SDK and the CLI, and the board's own firmware still boots afterward.
- Wi-Fi image updates on a BenchPod analog v3r2: installs over the USB network and Ethernet, about 2 minutes each, MD5-verified, with other commands refused during the write, and again with this release's signature rules in place.
- Security: on a BenchPod analog v3r2, unsigned images over Ethernet are refused and the same image over USB is accepted, the LAN TLS changes are refused from Ethernet and work over USB, and signed releases install over Ethernet. The bridge identity checks are covered by host tests.
- The whole release on a BenchPod analog v3r2 (benchpod-30a9f1): LAN hardware tests 72 passed, 0 failed (9 skipped: features this bench has no wiring for), SDK end-to-end 158 passed, MCP 9, OpenHTF 2.
adc_readmicrovolts over Ethernet and the cloud on a BenchPod analog v3r2. The ±12 V limit: host tests, no change in the normal range on a BenchPod analog v3r2, and on a BenchPod analog v3r4 with the published release (±6 V refused in the low range, 4.9 V accepted, low-range readings in microvolts within 0.3 mV of the normal range from 1 to 4.5 V).