One last reply to a remark which I mist yesterday:
>> /* After this the ABIT_UGURU_DATA port should contain
>> ABIT_UGURU_STATUS_INPUT */
>> while(inb_p(client->addr+ABIT_UGURU_DATA) != ABIT_UGURU_STATUS_INPUT){
>> timeout--;
>> if (timeout == 0) {
>> #if ABIT_UGURU_DEBUG_LEVEL >= 1
>> printk(KERN_DEBUG "abituguru: error: state != more input after ready command\n");
>> #endif
> dev_dbg
>> return -EIO;
>> }
> Maybe you should also lock access (mutex
> or semaphore, see Linux device drivers third edition as pdf) to the chip in _read _write functions so no other
> concurrect access to achip from different part of driver could occur.
>
This is already taken care of. The detect_sensor_type function where
this comments was made is called before i2c_client_register, thus none
of the sysfs show/store will get called, so no race.
All the sysfs show/store functions either call abituguru_update_data
which does locking, or do the locking themselves directly. This locking
of the data thus protects against any races involving simultanious use
of read/write functions.
Regards,
Hans
