On Thu, 8 Jan 2004, Jean Delvare wrote: > > The GPIO pins could be bidirectional, but one way at a time. Meaning, > > the read value _GPIO_Y_0 is not valid when _GPIO_EN_0 is set. > > I did not see anything similar in Gatos' code. Change getscl/sda to test the _GPIO_EN_x bit. If set (pull low), return zero, else if unset (hi-Z) read and return _GPIO_Y_x. I wondered why there was I2C_ALGO_ATI in i2c-id.h : Katos. It claims to have working i2c kernel code, and does not use i2c-algo-bit. The code is from -99, so don't expect it to compile out-of-the box either. http://www.core.binghamton.edu/~insomnia/gatos/katos/src/katos-0.5.tar.gz > > Did you try without bit_test? You might need some <1 us delay between > > read and write to see the change, the bus has some capacitance even > > without monitor cable. > > Yes, I tried without bit_test=1. I2cdetect hangs using 100% CPU, which > makes me think my code is somehow broken, although I don't see where. > I've copied what I read from the gatos code, using i2c-savage4 as a > base, everything seems very simple, but it doesn't work. Would you have > any concrete suggestion of something I could try? Else I think I'll just > give up. I guess i2cdetect sleeps in sclhi() twice for each address to scan. Run with strace, does it return from syscalls or is it simply slow? Also have i2c-algo-bit debugging enabled, plz. At some point I got fed up tracing the failure paths and return codes in i2c-algo-bit. You could have better debugging experience using the i2c-algo-biths, just uncomment one COMPATIBILITY #define in the source. -- Ky?sti M?lkki <kyosti.malkki at welho.com>
