Hi Nicolas,

On Tue, Aug 03, 2010 at 01:21:14, Nicolas Luna wrote:
> Hi Sekhar,
>
> Yes there is an other device on the bus, but I removed it and I
> decreased the clock to 20khz. I cannot go below because the i2c seems
> not to work. It does solved the problem.
>
> I modified the driver to add a while(1) when it got an abnormal
> termination. I notified that every time the driver makes this error, I
> can see a clock stretch (with a scope) on the last i2c frame.
>
> - The PMIC (tps65070) is rated to work at 400khz, but even at 20khz it
> seems to need more time to response when it is addressed to much. I
> modified the touchscreen polling period to 100 msec and I still got
> errors. Maybe if the touchscreen driver was working by interruption that
> would leave a chance to the PMIC.

It should be possible to use an interrupt on a correctly wired hardware.
Can you try getting the interrupt mode working? Hopefully Todd or others
from list can help out in case you run into issues.

>
> - Why the i2c driver do not handles clock stretching?

It should and seems like there is a 1 second time out the driver has
waiting for transfer to complete. How long does the clock stretch in
your case? Please see DAVINCI_I2C_TIMEOUT in drivers/i2c/busses/i2c-davinci.c

> - On a test there was 2 successive reads to register 02h with a delay of
> 85 msec between. The first went right but while the next read, the salve
> put the clock low to stretch it. It sounds like a problem to me because
> the chip was not receiving a lot of messages and it stretched the clock.
> I cannot find any place in the datasheet that is talking about that
> issue.

Not sure regarding this. This will likely need some TPS hardware person to
analyze. Can you please post your observations to the PMU forum at:
http://e2e.ti.com/support/power_management/pmu/f/200.aspx

Thanks,
Sekhar

_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source

Reply via email to