Re: I2C bus probe-able?

[Eduardo Horvath <eeh%NetBSD.org@localhost>]

On Thu, 7 May 2009, Erik Fair wrote:

> And I wonder: is the iic (I2C) bus probe-able, 

Well, technically it is probe-able.  You can determine if a device exists 
at a specific address.  And since the original idea was to have Philips 
hand out specific addresses to specific devices, you could even identify 
devices based on their address.  

However, since then many vendors, including Philips, have added the 
ability to set I2C addresses of devices based on hardware strappings, and 
there are also I2C address translator devices which allow you to have 
multiple instances of the same device on different segments of the same 
I2C bus, the identification process becomes complicated.  Identifying a 
specific device would require issuing a specific sequence of I2C bus 
phases that the device would respond to in a specific way allowing the 
driver to determine if the device at that address is one it supports.  But 
since there is no standard for device access, unlike SMB where Intel 
specified only a subset of possible I2C transactions, the operations 
needed to identify one device can result in damage to some other type of 
device that might be sitting at that address.

So although it is technically probe-able, practically it's not probable 
since the probing process can cause damage to the system.

Eduardo