version 1.1.7:
-change minimum accepted temperature value for pwm autopoints from 35 to 25,
 as some bioses allow setting 25.
-change sysfs alarm interface from one alarm file per sensortype to one alarm
 file per sensor. Now uguru implements the new alarm interface as documented in
 Documentation/hwmon/sysfs-interface
-All the sysfs entries are now generated automaticly from a template (one
 template for each sensor type) instead of using a huge static array as the
 huge array was becoming really huge and unmaintainable with the new alarm
 sysfs interface.

version 1.1.6:
-put the wait for alarm time in bank1 sensor type detect code for voltage
 sensor test back to 20ms, as according to the reporter only the temp test
 duration needs to be raised.
-except 0x40 (bit 6) being high as a valid fan sensor setting, I don't
 know what this bit does, but it is being reported high for fan1 on some
 motherboards. Currently this bit is only accepted for fan1, but that might
 change.
-use 77 (30%) as minimal allowed pwm%, but for pwm1 only.

version 1.1.5:
-converted from deprecated i2c-isa to a platform driver.
-renamed ABIT_UGURU_PRINTK to ABIT_UGURU_DEBUG, only use verbose for
 controlling debugging output always show info and warnings.
-many code-style cleanups to make upstream happy.
-raised the max number of supported pwm outputs to 5 to support the Abit
 AA8XE Fatal1ty.
-(attempt to) detect the number of pwm outputs and fan sensors, so that
 we do not give the user the ability to write pwm-settings for pwm's which
 aren't there, otherwise the user could cause us to overwrite random uGuru
 memory.
-raised wait for alarm time in bank1 sensor type detect code from 20 to 50
 ms, problems have been reported with 20 ms and aAA8XE Fatal1ty motherboard.
 The reporter didn't had problems with 100 ms but that seems excessive.

version 1.1.4:
-move the ChangeLog info to a seperate ChangeLog file.
-add a ABIT_UGURU_PRINTK macro which uses dev_info and friends and does all
 the verbose level handling
-add verbose param instead of define
-change from semaphores to mutexes
-minor cleanups as a result of reviews by upstream.

version 1.1.3:
-rename show/store_pwm_reg to show/store_pwm_setting for consistency.
-add comment about how exactly attr->nr and attr->index are used in all
 the different sysfs callback functions.

version 1.1.2:
-changeover from using the old? __SENSOR_DEVICE_ATTR2 macro to the new
 SENSOR_ATTR_2 macro. Note: the 5th and 6th param. have changed position!

version 1.1.1:
-Show voltages ona 0-3493 mV scale instead of as a registervalue multiplied
 by 1000.
-Change all the sysfs stuff from struct sensor_device_attribute to
 sensor_device_atribute_2 make settings array 2 dimensional as it ought to
 be and use index and nr to address the settings array. Remove all the
 glue functions for mask stuff and use attr->nr as mask.
-Split alarms_in in alarms_in_low and alarms_in_high, same
 for alarms_mask_in.
-Undo 1.1.0 change of not reading the settings in abituguru_update for
 bank1 settings, as the settings are needed to differentiate between low
 and high alarms.
-Rename pwm sysfs files to match sysfs-interface file from the kernel
 Documentation, rename mask files to match sysfs-interface-proposal.

version 1.1.0:
-One test-user with an AN7 is still reporting the following error:
 abituguru: error: timeout exceeded waiting for more input state (bank: 32)
 Even with ridiciously high timeout (number of polls really) values this
 still happens occasionally. Here is a table of number of polls and
 occurence frequency of the error:
 polls:	error frequency:
 250	6 / min (1 / 10 sec)
 500	6 / min (1 / 10 sec)
 5000	3 / min
 50000	1 / min
 As if this version of the uGuru does something internally every 10 seconds
 which makes it unreachable for a while from the outside, making the
 timeout longer makes the chance smaller that our try to reach it window
 fully overlaps with its busy window. I've also had a report of this from
 another test-user but with a much lower frequency.
 All the changes below are mainly to cope with this behaviour of the uGuru.
-Add a retries parameter to abituguru_send_address and abituguru_read. With
 this parameter the number of retries to try the send adres step causing
 the problem can be specified, send_address will sleep before retrying.
-Try reading of a value in abituguru_update_device once and if it times out
 just keep the old value, instead of generating an error.
-Try all other reads and all writes multiple times.
-Count the number of timedout updates in abituguru_update_device, if this
 passes a certain treshold only then make it an error.
-Don't update all the settings everytime in abituguru_update_device,
 instead just read them once in our probe function, the used part of the
 settings won't change unless we write them, in which case we know what
 we've just written, so we still won't need to read them. This way we know
 we will always have a valid in memory copy of the settings, even though
 abituguru_update_device no longer logs every read timeout as an error.
 Note: This does mean that we do not have uptodate flags info in our in
 memory copy of the settings, this is not a problem since the flags part
 of the settings are only used by abituguru_detect_sensor_type, which
 does a read of the settings itself. BEWARE though, if we want to use the
 flags parts in the future in other places!
 As a nice side effect this makes abituguru_update_device much faster
 and generate a lot less ISA-IO cycles.
-I've also received one report of:
 abituguru: error: state != more input after ready command.
 Change the code to wait a while for DATA reg to hold 0x08 instead of
 reading and testing it only once as an attempt to fix this.

version 1.0.3:
-Fix the detect issue (needs force param) 1 test-user was reporting.
-Raise number of tries before giving up when waiting for CMD to hold
 0xAC after ready a little, this fixes the few remaining reports of:
 abituguru: error: CMD reg doesnot hold 0xAC after ready command

version 1.0.2:
-2 users are reporting the following error:
 abituguru: error: timeout exceeded waiting for more input state (bank: 32)
 This always happens with bank 32 (0x20 / alarm).
 Raise the time out in abituguru_wait as an attempt to fix this.
-2 users are reporting the following error:
 abituguru: error: CMD reg doesnot hold 0xAC after ready command
 Change the code to wait a while for CMD reg to hold 0xAC instead of
 reading and testing it only once as an attempt to fix this.
-1 user needs to use the force parameter at debugging output in case
 the normal detect code fails to help improve the detect code. 

version 1.0.1:
-Set min/max values during volt sensor type detect a little more sane
 (245 250 instead of 254 255), this seems to fix some detection issues.
-Fixed a possible race where we could be called during bank1 sensor type
 detection.
-Added debuging output all over the place since some testers are having
 trouble.
-Changed ABIT_UGURU_DEBUG define to ABIT_UGURU_DEBUG_LEVEL, which
 controls the amount of debugging output instead of just turning debugging
 on/off.

version 1.0.0:
-Added support for controling the fan PWM (speed control).
-Add force param to skip "detection" .

version 0.9.5:
-Reading/writing the settings (enable alarm, beep, shutdown) is now also
 supported.

version 0.9.4:
-reshuffle / cleanup (mainly bank1 handling)
-fix some race conditions

version 0.9.3:
-userspace expects temp and in values to be in milli units, so they are now
 multiplied by 1000 in show and divided by 1000 on set.
-Now that the min/max values can be written I've been able to trigger some
 false alarms and have managed to find out how the alarms work, so now
 reading the alarms is supported also.

version 0.9.2:
-Olle has now reverse engineered the writing of the flags and min/max
 values, so now writing the min/max values is also supported. 
-Now that we've seen testing of more mainboards, it turns out that
 the way the volt inputs are hooked up isn't identical for all mainboards,
 also causing the ranges for the volt inputs to be not constant. Because
 of this conversion for the voltage sensors has been removed, the raw
 register values are now returned. Sample entries for known mainboards
 will be added to sensors.conf .
-During this testing, it has also been shown that certain addresses in
 bank1 (0x0D) sometimes are attached to a temp sensor and sometimes to a
 volt sensor. Luckily with some tricks this can be detected, so now we
 detect if a sensor is connected and what type for all addresses in bank1.

version 0.9.1:
-Olle has done some more reverse engineering, so now reading the min/max
 values and the settings is also supported.
-modified the driver to work with the new hwmon seperation in 2.6.14,
 As a result this version requires 2.6.14 or higher.
-simplified the ready code.
-added a abituguru_wait function which waits for the uguru the reach
 a certain state, use this everywhere instead of having seperate loops
 with there own timeouts scatered over the place.
-use array sysfs initalisation instead of #define hell.
-fix rounding of values.

version 0.9:
-initial release, only reading the sensors is supported.
