On Wed, 18 Jul 2012, Oliver Neukum wrote: > On Wednesday 18 July 2012 11:42:50 Alan Stern wrote: > > On Wed, 18 Jul 2012, Oliver Neukum wrote: > > > > > On Wednesday 18 July 2012 10:54:20 Alan Stern wrote: > > > > On Wed, 18 Jul 2012, Lan Tianyu wrote: > > > > > > > > > On 2012/7/18 22:19, Alan Stern wrote: > > > > > > An alternative to her suggestion would be to disconnect the device when > > > > > > the port goes off. When the port is powered back on, the device will > > > > > > be rediscovered and probed again. > > > > > hi alan: > > > > > Reprobe may cause the devnum change. Userspace will treat it a > > > > > different device from previous one. e.g app held the device node fd, > > > > > After port went off and on, device reprobed and the fd became invalid. > > > > > > > > That's right. Your argument suggests that power-off should be > > > > controlled by userspace, not by the kernel, and it should happen only > > > > when no app is holding the device node open. > > > > > > That is a race. And it is basically unsolvable in user space if the device > > > is available to everybody. > > > > Everything userspace does is a race. > > Therefore some things should be done in kernel space ;-) Oliver, you seem to be arguing both sides of this discussion. You point out the the power-off operation is too dangerous in general for the kernel to do it, and now you say that it's too racy for userspace to do it. Are we to infer that you don't want it to be done at all? What about things like the ubiquitous USB webcams one finds in laptops these days? I practically never use the one in my laptop, so turning off its bus power seems like a good idea. Even if that means I have to turn the power on explicitly before using the webcam. 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
