On Mon, Jul 27, 2009 at 02:45:55PM +0800, Zhang Rui wrote: > Hi, Darrick, > > great job. :) Thanks. > we should at least send a netlink event for an ACPI power meter > notification, shouldn't we? Yes, an event would be useful at the very least for when the system starts capping, though unfortunately there doesn't seem to be a Notify event for when capping ends. > we should create the hwmon sysfs I/F for the hwmon device, > i.e. add the hwmon attributes > under /sys/devices/LINUXSYS:00/.../ACPI000D:00/hwmon0/ > rather than ?/sys/devices/LINUXSYS:00/.../ACPI000D:00/ Hee, this is a tricky matter. The other hwmon drivers create sysfs attributes under the sysfs object, not in a separate "hwmonX" object, and as I recall libsensors reads symlinks under the device object to figure out which bus the sensor device lives on. If it can't figure that out, it ignores the hwmon device. > if (acpi_disable) > return -ENODEV; Noted. > plus, _PMD is not supported in this driver, right? > I agree with Yakui that we can create some ACPI device sysfs attributes > besides the hwmon ones. e.g. exporting devices measured by the current > ACPI power meter device to user space. I agree. --D
