Hi Tony, > Before I go on debugging, a few questions, as my knowledge of the i2c > framework is quite limited: > > This particular bug seems to be triggered in i2c_dev_init when calling > i2c_add_driver. Were you able to reproduce this one as well? I couldn't. > 1. nvidiafb gets loaded before i2c_dev_init(), and when setting up > i2c, calls i2c_bit_add_bus. Is this safe to do? Looks OK to me. nvidiafb, like other framebuffer drivers, declares i2c busses. i2c-dev declares an i2c driver. Both are rather independent, and supposedly don't interract before an i2c client is created (that would happen in i2cdev_open()). i2c-dev isn't really part of i2c-core, it's more like a priviledged i2c pseudo-chip driver. > 2. In i2c_add_driver, there is a call to list_add_tail. Is this > acceptable without doing a LIST_HEAD_INIT first? Or is the list_head > initialized somewhere? The list is declared with: static LIST_HEAD(drivers); at the top of i2c-core.c. Looks OK to me but I am no list expert. Thanks, -- Jean Delvare
