Hi, Short selfintro for the watchdog people: I'm a FOSS developer / enthusiast for close to 15 years now. I've written Linux kernel drivers for a ton of webcams and various hwmon IC's. Two of the hwmon drivers I maintain are for Fujitsu manufacturered IC's which include a watchdog part in their hwmon solution. The fschmd driver for their family of i2c hwmon IC's already supports this using the standard watchdog API. While waiting for some input from Fujitsu needed to get the watchdog in their new SMSC SCH5636 superio (with a programmable micro controller) based hwmon solution working, I read the announcement of the watchdog timer driver core. I'm really happy to see the watchdog timer driver core, as I just copy and pasted a whole bunch of code from the fschmd driver to the sch5636 driver, having a standard watchdog core definitely is a good idea. So when working on finishing up my watchdog support for the sch5636 driver I decided to convert it to the new watchdog timer driver core. I hit 2 issues during this (and 2 minor others): 1) The driver core only works for 1 watchdog, but both my fschmd and my sch5636 test systems also have an iTCO watchdog and that tends to load first. Thus when writing the fschmd watchdog code I made it support loading either on the default watchdog minor, or on one of the additional reserved watchdog minors (212-215). The second patch in this series adds support for this to the watchdog timer driver core (which was quite straightforward to add). 2) Most Linux drivers use dynamically allocated memory for the driver data, which gets allocated on calling the probe function and released on driver unbind. This means that merely locking the module containing the driver as long as /dev/watchdog is open is not sufficient, since the driver can still be unbound. The third patch in this series fixes this. The other 2 patches to the watchdog core fix 2 minor issues and are self explanatory. The 5th patch in the series adds support for the integrated watchdog to the sch5636 driver. It seems this will be the first driver actually using the new core :) I think it is best for this entire series to go in through the watchdog tree, since the 5th patch depends on the others (even though the 5th patch touches hwmon code). Once this is upstream I'll also write a patch to convert the fschmd driver to use the watchdog core, and maybe also the iTCO driver. Regards, Hans _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
