Add a timeout to busio_i2c_scan

[prplz]

@UnexpectedMaker recently discovered an annoyance where i2c scan can hang for over a minute on esp32 (#5906).

It goes something like this:

• Construct an i2c with floating pins
• The i2c pullup check in common_hal_busio_i2c_construct passes most of the time despite floating pins
• Start an i2c scan
• busio_i2c_scan tries to probe 110 addresses, each probe times out after one second (see this comment), making cpy hang for an unacceptably long time

I have added a two five second deadline to busio_i2c_scan. I hope this will not effect normal operation and will only cover this strange case.

Tested on esp32-s3 hardware. With pullups the scan run in about one second, without the pullups it raises an exception after about two or three five or six seconds.

[prplz]

Force pushed: changed timeout to 5s.

[dhalbert]

We can also reduce the actual I2C timeouts to something much shorter.

[prplz]

Discussed with dan in discord and there's a lot more work to be done with i2c.

[dhalbert]

We can keep this open for discussion now. I will just make it a draft.

[jepler]

Another possibility is to check for KeyboardInterrupt at each trip through the loop, so that a scan can be ctrl-c'd

[tannewt]

Could we make the probe fast instead? https://github.com/adafruit/circuitpython/blob/main/ports/espressif/common-hal/busio/I2C.c#L144

We shouldn't need to wait at all to determine it.

[prplz]

• Remove busio_i2c_scan changes
• Change espressif i2c probe timeout from 1s to 10ms (1 freertos tick, unsure what behaviour zero would have)
• Change espressif i2c probe timeout from 10s to 1s and use freertos macro to avoid confusion

[prplz]

This is not working, it looks like i2c_master_cmd_begin caps the delay at 1 second minimum: https://github.com/adafruit/esp-idf/blob/circuitpython/components/driver/i2c.c#L1172

[microdev1]

esp-idf was updated recently, a merge from main should fix the CI.

[dhalbert]

This is not working, it looks like i2c_master_cmd_begin caps the delay at 1 second minimum: https://github.com/adafruit/esp-idf/blob/circuitpython/components/driver/i2c.c#L1172

This was a reported as a bug:
espressif/arduino-esp32#5934
espressif/esp-idf#4999

went back in time, and when I2C_CMD_ALIVE_INTERVAL_TICK was first introduced, it was a ceiling, not a floor:
https://github.com/espressif/esp-idf/blame/9075b507b54a99ce732fc2819ba17274d9a69ab5/components/driver/i2c.c

        TickType_t wait_time = (ticks_to_wait < (I2C_CMD_ALIVE_INTERVAL_TICK) ? ticks_to_wait : (I2C_CMD_ALIVE_INTERVAL_TICK));

(Commented as such here: espressif/esp-idf#4999 (comment))

[dhalbert]

I think the next step here is to submit one or more PR's (depending on versions) to es to fix espressif/esp-idf#4999. We can also patch this in https://github.com/adafruit/esp-idf for now, since we are already using that fork.

[dhalbert]

@prplz Shall we close this now that you fixed the pullup problem? I think it's still true that ESP-IDF really needs a fix for #5908 (comment). We could open a separate issue about that.

[prplz]

Sure