Le Tue, 24 Feb 2009 09:11:53 +0800, Peter Teoh <htmldeveloper@xxxxxxxxx> a écrit : > But then it also puzzled me, because ioctl() in user space required a > open file descriptor as the first parameter, which is derived from > open() (as specified from "man ioctl"). so then how is the open() > from userspace passed down to kernel? When userspace uses the open() system call, the kernel function sys_open() gets called. This kernel function does all the job, and calls the ->open() operation of the device driver if it exists. Otherwise, it just doesn't call it, and it works perfectly fine. The sys_open() function does all what's needed, the ->open() operation is only here *if* your device driver needs to do something when the device is opened. This operation is optional. sys_open() http://lxr.free-electrons.com/source/fs/open.c#L1032 calls do_sys_open() http://lxr.free-electrons.com/source/fs/open.c#L1010 which calls do_filp_open() http://lxr.free-electrons.com/source/fs/namei.c#L1637 which calls nameidata_to_filp() http://lxr.free-electrons.com/source/fs/open.c#L918 which calls __dentry_open() http://lxr.free-electrons.com/source/fs/open.c#L793 which does 823 if (!open && f->f_op) 824 open = f->f_op->open; 825 if (open) { 826 error = open(inode, f); 827 if (error) 828 goto cleanup_all; 829 } (ouf) Sincerly, Thomas -- Thomas Petazzoni, Free Electrons Kernel, drivers and embedded Linux development, consulting, training and support. http://free-electrons.com -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
