On Fri, Mar 25, 2005 at 02:18:13PM +0100, Jean Delvare wrote: > > Hi Ben, > > > This patch (as requested by J Delvare) adds the > > following bits which have been missing from the > > s3c2410-i2c.c bus driver, as well as a couple of > > other minor changes: > > > > 1) add a .functionality field to our algo, and > > report the facilities. > > Thanks, this will let me go on with core cleanups of the functionalities > handling. > > > 2) Add I2C_M_REV_DIR_ADDR handling for our start > > code, to better support our reported value > > from the .functionality probe (the use of > > I2C_FUNC_PROTOCOL_MANGLING) > > Hm, do you actually need mangling? Protocol mangling is something we > added because it was needed in some weird cases, not because we liked > it. It's not considered a wanted feature that i2c_adapters should > implement whenever possible. It should really only be implemented when > there is a concrete need. In my understanding, the need for mangling > means that some chip is not properly conforming with the I2C standard. I added it as the headers seem to suggest that I2C_FUNC_PROTOCOL_MANGLING is indicative of support for I2C_M_NOSTART. Addition of I2C_M_REV_DIR_ADDR was just to make this fuller. > > 3) change the call to dev_err() to dev_dbg() in > > the code that checks for an ack off the > > start condition. This avoids the `flood` of > > output when using i2cdetect > > Makes sense. > > > 4) Add a .algo_control field which points to a > > function returning zero. > > Hmm, why? The i2c-core deals properly with NULL algo_control. Poiting to > a function which returns 0 will be exactly the same than having no > function defined at all, but eats a few bytes of memory and wastes time > as well. Ok, just noticed at least one other bus driver seems to implement it (i2c-iop3xx). I can remove this easily. > > 5) Add an `.owner = THIS_MODULE` to the adapter > > I had not noticed it was missing, you're absolutely right. My home mailsystem is now fixed, so all messages to the list can be read again. Mail lost between 22nd and 24th. -- Ben (ben at fluff.org, http://www.fluff.org/) 'a smiley only costs 4 bytes'
