On Tue, 21 Apr 2015, Tom Yan wrote: > By blacklisting the hid-generic, I think I can confirm hid-generic is > the module that matters. Though I'm curious how it use the code from > usbhid/hid-core.c because I couldn't find a clue in the short > hid-generic.c. In fact, hid-generic.c contains no executable code at all! What matters is that it is a driver and it can bind to a device. Take a look at the hid_device_probe() routine in hid-core.c, and bear in mind that hid_hw_start() is an inline routine which calls hdev->ll_driver->start(). > Anyway I don't know much about coding so it's not really my concern. > But I think the fact hid-generic sets a default (no matter if it's > "enabled" or "disabled") brings a problem. The reason is, the wakeup > attribute is not "initialized" or "created" by hid-generic, so udev > can apply a rule related to the attribute before hid-generic get > loaded (which seems to be what it is doing, see > https://bugs.freedesktop.org/show_bug.cgi?id=90041). If the module > gets loaded or reloaded afterwards, the value would be set to the > hard-coded default anyway. This is a similar problem affecting all devices. The wakeup attribute in sysfs (along with all the other power-related sysfs attributes) gets created before the device is probed. Consequently, any changes made by drivers in their probe routines don't show up in the initial sysfs files. > Perhaps it is possible to change the behaviour by rewriting some bits > of udev, but I don't see the point of doing so. IMHO these hard-coded > default should be avoid from all drivers until it is really necessary, > and when it is needed, it's best to have the default set right at the > point when the attribute is initialized if possible. Setting the default when the attribute gets initialized is not possible in this case. The wakeup attribute belongs to the USB device, but the fact that the device can act as a keyboard is hidden away in one of the interfaces (and the interfaces are registered _after_ the device). Things would be a little better if the wakeup attribute were a property of the interface, but it isn't. Anyway, you're suggesting that drivers should never override sysfs attribute values. But there doesn't seem to be any other way to implement the kernel's policy that wakeup should be enabled by default for all keyboard devices. After all, only the driver knows whether or not the device it manages is a keyboard. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
