# Retained Core Dump and factory recovery This destructive dual-CDC USB example demonstrates Iris' default crash-loop recovery policy on a 16 MB layout. A normal OTA application deliberately crashes after startup, but the injection path does not increment a counter or select a partition. On the following boot, `esp_iris_boot_probe()` attributes the panic reset to the previously started application. At the configured Iris threshold (one for this example), Iris commits the failure record, selects the factory image, and performs a planned software restart. The original Flash Core Dump remains available in factory Recovery. The bootloader rollback feature is intentionally not used: a threshold greater than one must be allowed to restart the same application repeatedly. ## Standalone setup Create this example from the Component Registry, then run the commands below from its directory: ```bash idf.py create-project-from-example "espressif/esp_iris=0.2.0:crash_recovery" cd crash_recovery ``` With a source checkout, change to `ESP-Iris/components/esp_iris/examples/crash_recovery` from the repository root instead. These application USB examples require an SDK supporting ESP32-S31; the release matrix uses the pinned IDF 6.2 revision. After the first build downloads the component, prepare its bundled Gateway: ```bash python esp_iris.py setup python esp_iris.py doctor python esp_iris.py web ``` The launcher uses an isolated cached environment and does not change the IDF Python environment. Close the Gateway before flashing or monitoring its ports. ## Build and install factory recovery Verify the physical flash size and target before the first full flash: ```bash idf.py \ -B build-crash-recovery \ -D SDKCONFIG="$PWD/build-crash-recovery/sdkconfig" \ -D SDKCONFIG_DEFAULTS=sdkconfig.recovery.defaults build idf.py \ -B build-crash-recovery -p /dev/serial/by-id/<programming-port> flash ``` Build the crashing application: ```bash idf.py \ -B build-crash-application \ -D SDKCONFIG="$PWD/build-crash-application/sdkconfig" \ -D SDKCONFIG_DEFAULTS=sdkconfig.application.defaults build ``` With recovery connected, transfer `build-crash-application/esp_iris_crash_recovery.bin` using application execution mode. After the application crashes, the same device ID reconnects in factory recovery. ```bash IRIS=esp_iris.py python "$IRIS" ctl ota DEVICE_ID \ build-crash-application/esp_iris_crash_recovery.bin \ --elf build-crash-application/esp_iris_crash_recovery.elf \ --execution-mode application --wait python "$IRIS" ctl crash DEVICE_ID python "$IRIS" ctl coredump DEVICE_ID retained-core-dump.bin python "$IRIS" ctl rpc-raw DEVICE_ID 5120 1 ``` RPC `0x1400/1` returns four little-endian words: Iris count, Iris limit, last application address, and flags (`bit0=injection enabled`, `bit1=recovery pending`). RPC `0x1400/2` calls `esp_iris_crash_loop_reset()`, disables injection, and resumes the application. RPC `0x1400/3` clears the same Iris record, re-arms injection, and retries it. Both restart RPCs delay reboot long enough for their responses to drain. To validate a threshold of three, set `CONFIG_ESP_IRIS_CRASH_LOOP_LIMIT=3` in an ignored local defaults file for both profiles. Recovery never exits automatically. Return to the [example index](../README.md).
To create a project from this example, run:
idf.py create-project-from-example "espressif/esp_iris=0.2.0:crash_recovery"