On 25 November 2012 17:22, Paul Carpenter <paul@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote: > Viresh Kumar wrote: >> Add i2c bus recovery infrastructure to i2c adapters as specified in the >> i2c >> protocol Rev. 03 section 3.1.16 titled "Bus clear". >> >> http://www.nxp.com/documents/user_manual/UM10204.pdf > > You should also take note of section 3.1.14 and its notes on Software > Reset - > "Precautions must be taken to ensure that a device is not > pulling down the SDA or SCL line after applying the supply > voltage, since these low levels would block the bus" Hmm.. This patch has taken a very long time for being accepted, and the reason for that was, it was generalizing much more than what is required. And I am not sure, if checking SCL low would be considered like that only. :) @Wolfram: I would need your point of view on that, before trying it out. >> +static int i2c_get_gpios_for_recovery(struct i2c_adapter *adap) >> +{ >> + struct i2c_bus_recovery_info *bri = adap->bus_recovery_info; >> + struct device *dev = &adap->dev; >> + int ret = 0; > > Where is the check that SCL is NOT low (bus fault or device doing clock > stretching) Pending for Wolfram's point of view :) >> +static int i2c_generic_recovery(struct i2c_adapter *adap) >> +{ >> + /* >> + * By this time SCL is high, as we need to give 9 falling-rising >> edges >> + */ > > In my view that needs to be done by an actual check of the real SCL not > assumption. Checking that would be tricky. For GPIO recovery, we have to make it GPIO first in input mode and read its value. Right? >> + delay = DIV_ROUND_UP(delay, bri->clock_rate_khz * 2); >> + >> + for (i = 0; i < RECOVERY_CLK_CNT * 2; i++, val = !val) { >> + bri->set_scl(adap, val); >> + ndelay(delay); >> + >> + /* break if sda got high, check only when scl line is high */ >> + if (!bri->skip_sda_polling && val) > > Dont use 'val' read the actual SCL line which ensures you you are not > wasting your time because of hardware fault. Possibly saving your GPIO > from being stuck permanently. Again, wolfram can decide if it is really required :) -- viresh -- To unsubscribe from this list: send the line "unsubscribe linux-i2c" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
