Hi,

I would like to take advantages of the hardware TWI (I2C) functionality, which is available on the AT91SAM9G20. Therefore, I would like to write a Linux kernel module for it.

However, when reading the documentation in the Linux kernel source tree regarding TWI (Linux kernel 2.6.30.9), I see some "worrying" comments.

To start with, in the file /linux-2.6.30.9/drivers/i2c/busses/Kconfig, I read the following:



Next, when looking into the file /linux-2.6.30.9/arch/arm/mach-at91/at91sam9260_devices.c, I see the following comment:



Before I start to put some effort in that new driver, I would like to know if the above mentioned shortcomings are still valid.

Does anyone has any idea? If the shortcomings are still in place, are there ways to work around it (still using the HW TWI functionality available and not switching to a kind of bit-banging sSW TWI/I2C)?

Best rgds,

--Geert

I am also very interested in doing the same. I wouldn't mind not having the capability for repeated start conditions if that's a sacrifice necessary to use the hardware TWI. The AT91 kernel claims its fully supported (since .20 if I recall correctly), so I would imagine that it should work in some form, but I have searched pretty hard for a solid example of someone using it and turned up empty handed.

edit: I dug up this thread from the arm-linux-kernel mailing list. The response seems to indicate that the sam9g20 driver should work, but there doesn't appear to be one in the mainline kernel. It seems odd that there exists a file for the sam9g20-ek board, but none for the sam9g20 chip itself...seems kind of lazy, just to keep using the sam9620 drivers (and also defeats the purpose of using the updated chip?).