Re: [Libusb-devel] Patch: libusb/os/linux_usbfs.c: Distingush between usbfs and other kernel mode drivers

[Xiaofan C. <xia...@gm...>]

On Wed, Oct 6, 2010 at 5:08 PM, Peter Stuge <pe...@st...> wrote:
> Xiaofan Chen wrote:
>> > It is really analogous to libusb0.sys in Windows.
>>
>> Totally different. If I need to use libusb0.sys, then I need to
>> attach it to the device (as a device driver or a filter).
>
> Exactly the same! usbfs must be attached to the device (well, to the
> interface) in order for libusb to work.
>
> The only difference is that Linux includes a generic USB driver
> (usbfs) already and Windows does not, and that when calling
> libusb_claim_interface() on Linux the driver is bound to the
> interface by the kernel. Windows binds the driver much more
> statically.
>

Okay this is a good point and I can agree with you it is
a kernel driver after all, albeit a generic one.

However, let's go back to the context of libusb. Say,
for libusb-0.1, what do you expect to return for the
usb_get_driver_np()?
http://libusb.sourceforge.net/doc/ref.nonportable.html

The typical code to call usb_detach_kernel_driver_np()
is to call usb_get_driver_np() first to see if there are
another driver is attached to the interface. If the existing
kernel driver (not usbfs) is already detached, you do
not need to detach the kernel driver.

Similar for libusb-1.0, I expect libusb_kernel_driver_active
to return one is only usbfs is bound to the interface.
http://libusb.sourceforge.net/api-1.0/group__dev.html#ga76e047ca95da2b32d0e7455da1804a49
And it is clear not the intention of libusb developers
to ask the user to attach the "usbfs" driver to the interface
for the libusb_attach_kernel_driver().

So my conclusion is still "usbfs" kernel driver is
equivalent to no kernel driver in the context of libusb.

Daniel: if I am wrong, please point out. Thanks.


-- 
Xiaofan