I2C timeout is ignored on IDF 4.4 (IDFGH-7110)

[readmodifywrite]

IDF 4.4 ESP32

Using the i2c_master_cmd_begin() API:

i2c_master_cmd_begin( I2C_MASTER_PORT, handle, 50 / portTICK_RATE_MS );

This should set the timeout to 50 ms, and it does so on the IDF 3.3.x. On 4.4, it is ignoring the argument and using a 1000 ms timeout, which is resulting in a task watchdog reset.

[readmodifywrite]

I tried this as a workaround in my init code:

i2c_set_timeout( I2C_MASTER_PORT, 4000000 );

This results in an error on the trace log: E (9208) i2c: i2c_set_timeout(817): i2c timing value error

Followed by the watchdog reset.

[readmodifywrite]

i2c_set_timeout( I2C_MASTER_PORT, 50 )

This seems to actually work. However, the timeout setting is documented as APB bus cycles. This appears to actually result in something like 20 ms. I do not see the correlation of the number 50 to either 20 ms or to APB bus cycles.

[readmodifywrite]

Actually, that appears to time out more or less instantly, which kills all I2C comms.

What happened here? The I2C driver was fine the way it was.

[AxelLin]

@readmodifywrite Do you still hit issue with recent v4.4.4 release?

[readmodifywrite]

Sorry, I have no idea. I only very rarely update the IDF because lots of small things (just like this issue) break every time I do so. The amount of API churn and regressions (in really basic things, like I2C) is simply too much to keep up with.

[AxelLin]

I don't get it. You cannot get bug fix if you don't update your esp-idf. Should be no api change between v4.4 and v4.4.4.

[readmodifywrite]

I've already worked around the issue and I don't want to deal with whatever new stuff comes up. I won't get just a bug fix for the I2C, it's always 1000 other things.

Sorry, but I'm a busy engineer and I don't have time to debug this 13 month old issue plus all the other things I have to do every day. Every time I update the IDF it's a multi-day slog through multiple issues that were completely fine before. I can only update if I have an extremely compelling reason to. This is the practical reality of the profession.

[yuhongshuai]

I hit the same problem. find bug in idf i2c.c

    TickType_t wait_time = xTaskGetTickCount();
    if (wait_time - ticks_start > ticks_to_wait) { // out of time
        wait_time = I2C_CMD_ALIVE_INTERVAL_TICK;
    } else {
        wait_time = ticks_to_wait - (wait_time - ticks_start);
        if (wait_time < I2C_CMD_ALIVE_INTERVAL_TICK) {
            wait_time = I2C_CMD_ALIVE_INTERVAL_TICK;
        }
    }