John,

There is an example in the 'AN 527 Selecting and Using GPIO Signals on DM3730 - AM3703 SOMs'  in appendix B showing how to enable the VPLL2 in the PMIC using i2cset and i2cget commands for I2C1 from within Linux console.

I have copied this information below.

Linux I2C Commands

The following commands can be run from the Linux prompt to enable VPLL2 at 1.8V.

DM-37x# i2cset -f -y 1 0x4b 0x91 0x5

DM-37x# i2cget -f -y 1 0x4b 0x91 0x05

DM-37x# i2cset -f -y 1 0x4b 0x8e 0x2e

DM-37x# i2cget -f -y 1 0x4b 0x8e 0x2e

Disabling the PKG_I2C_TOOLS package might allow you to use the ioctl control capability. 

Thanks,

Brad

Thank you guys for the guidances.  I think I finally made great progress today.

Adam, I looked through the board-omap3logic.c file and now I understand how each of our device nodes are being created.  I didn't see anything unusual with how the I2C1 node was being configured, other than the fact it is at a much higher clock rate (2600) than the other I2C2 and I2C3 nodes.  But I think this is made transparent to me at the user level anyways.

Anyways, I looked at the App Notes that Brad pointed out and also the Linux User Guide.  So this gave me an opportunity to play with the i2cset and i2cget command, which was just extremely helpful in troubleshooting my custom I2C driver code.

Here's what I noticed was my mistake.  I noticed in the App Notes, they use 'i2cset -f ..." (the FORCE option).  If I took out the -f option, things wouldn't work.  This hinted me to use ioctl(m_fd, I2C_SLAVE_FORCE, addr) in my code and I was no longer getting errors from my ioctl call.

I guess my first question is "Is FORCE the correct thing for me to do in my code"?  Doesn't seem like I have any other choice, right?

----------------------------------------

Then the other reason why my reads were coming back as 0 on ADC CH9 (Battery Backup Voltage) was because I forgot to do conversion on the ADC chip.  That was more of a "DUH" moment.  I found out the procedure on how to do this in the Linux BSP User Guide (Section 4.22 Analog-to-digital Converters).  They provided a sample script 'adc-script'.

But here's one other question:  In the TSP65950 (PMIC) datasheet Section 9.5 Monitor ADC Programming Model, they are very insistent on using the MADC_ISR register to determine if the Conversion is complete.  However, I noticed in the 'adc-script', it simply hardcodes a sleep of 1 second to wait for the conversion.  Then it reads the MADC CH9 registers.  I believe for my application, a sleep would work fine since I have no strict timing constraint.

However, would I need to worry about not clearing the interrupt bits in the MADC_ISR1 register?  Would this cause issues if I didn't?

I apologize for asking such a detailed low-level question.

------------------------------------

So far, I am able to move forward with all the learnings, so thank you so much.

John,

This manpage shows a warning when using the force option with i2cset command.

Force  access  to  the  device  even  if  it is already busy. By
              default, i2cset will refuse to access a device which is  already
              under  the  control  of  a  kernel  driver.  Using  this flag is
              dangerous,  it  can  seriously  confuse  the  kernel  driver  in
              question.  It  can  also  cause  i2cset to silently write to the
              wrong register. So use at your own risk and  only  if  you  know
              what you're doing.

Here is a document with information using the I2C device driver within Linux.  

Here is a write-up of someone else implementing I2C_SLAVE_FORCE.

Special caution needs to be taken when using the force option. 

I don't have any information that guarantees the conversion has completed after 1 second as seen in the adc-script example.

You might consider verifying the completion of the conversion by using the MADC_ISR register.

Regards,

Brad

Brad,

Yeah, I saw the notes about the -f option, which is why I was concerned about using the FORCE option, but it seems to be the only way I could get the PMIC I2C communication to work and it was used all throughout the LogicPD examples.  I don't know what my other alternative would be.  Would using a polling mechanism with no FORCE even work?

Did you mean to attach some links in your previous reply?

Anyways, thanks for all the help so far.

Oh wait, nm.  The links are there.  The lack of text coloring tricked my eyes.