On Sun, 21 Dec 2003, Jean Delvare wrote: > In the process of merging our i2c CVS repository changes into the Linux > 2.4 kernel, I saw that, back in January 2003, you removed bus scanning > from two drivers (i2c-algo-bit and i2c-algo-pcf). What was your reason > for doing so? Is it because it was unused, or because i2cdetect can be > used instead, or both, or any other reason I didn't think of? 1. If wanted in kernel code, have a single scanner in i2c-core instead. I2C address space is not that flat and some addresses have special meanings for 10-bit addressing, high-speed, routing/multiplexing. A simple scan 0x00-0x7f only gives you half the picture. 2. A simple SMBus Quick command may have effect, the most simple i2c chip is a relay acting on the R/W bit of the address. Some chip at 0x69 (?) locked the bus in the early days (-99), and I think it was solved using write, not read address in the bus scan. 3. The ThinkPad issue, corruption on bus scan. While it is a bug on the EEPROM fw, you well know that somehow this type of issue needs to be handled. The fewer places to fix the better. The sensors-detect prog has 24RF08 fix, how about adding that in i2cdetect too? 4. The in-module bus scan only outputs to log, the format may not be uniform cross algorithms. There is i2cdetect. > I took a look at the i2c drivers present in both Linux 2.4 and 2.6, and > it turns out that there are other drivers still having a bus scanning > parameter (i2c-algo-ite.c and i2c-algo-sibyte.c in 2.4, i2c-algo-ite.c > and i2c-ibm_iic.c in 2.6). As I will be submitting a patch to remove bus > scanning from "our" drivers (i2c-algo-bit and i2c-algo-pcf), I'm > wondering if I should do the same for these extra drivers as well. Pretty much everything gets copy-pasted from i2c-algo-bit and i2c-algo-pcf without much thought, the algorithm interface could be better so this wouldn't happen. I vote for removal of bus scan in all algorithm sources. > BTW, does i2cdetect work on non-i386 architectures? Its pretty straight-forward. If it doesn't work, problem is more likely in i2c-dev. -- Ky?sti M?lkki <kyosti.malkki at welho.com>
