bcm2835-thermal and mailbox property retry

[notro]

The firmware driver is applied for v4.2, so we can start to use it.
Looking at the thermal driver I wonder if this retry loop is necessary?
I haven't seen any other use of the mailbox where this is done.

static int bcm2835_get_temp_or_max(struct thermal_zone_device *thermal_dev, unsigned long *temp, unsigned tag_id)
{
    int result = -1, retry = 3;
    print_debug("IN");

    *temp = 0;
    while (result != 0 && retry-- > 0) {
        /* wipe all previous message data */
        memset(&bcm2835_data.msg, 0, sizeof bcm2835_data.msg);

        /* prepare message */
        bcm2835_data.msg.msg_size = sizeof bcm2835_data.msg;
        bcm2835_data.msg.tag.buffer_size = 8;
        bcm2835_data.msg.tag.tag_id = tag_id;

        /* send the message */
        result = bcm_mailbox_property(&bcm2835_data.msg, sizeof bcm2835_data.msg);
        print_debug("Got %stemperature as %u (%d,%x)\n", tag_id==VC_TAG_GET_MAX_TEMP ? "max ":"", (uint)bcm2835_data.msg.tag.val, result, bcm2835_data.msg.request_code);
        if (!(bcm2835_data.msg.request_code & 0x80000000))
            result = -1;
    }

    /* check if it was all ok and return the rate in milli degrees C */
    if (result == 0)
        *temp = (uint)bcm2835_data.msg.tag.val;
    else
        print_err("Failed to get temperature! (%x:%d)\n", tag_id, result);
    print_debug("OUT");
    return result;
}

[popcornmix]

I did add that (over 2 years ago). A github issue reported occasional spurious readings (along with an mailbox error code). At the time retying was a quick fix. Not the right fix.

I've a suspicion this may have been fixed on the firmware side since then. I'm suspecting a latch_try (non-blocking mutex) with error if someone else was calling the function.

See: raspberrypi/firmware#132

So, feel free to remove the retry. I'd be interested if the issue reported still occurs:
raspberrypi/firmware#132 (comment)

[notro]

Thanks