Thread: Performing File operations using File Inode number

I am 99% sure that there are no user-mode functions in Linux or Unix that take an inode as an input. There are ways (such as stat()) to find out which inode belongs to a particular file/directory, but that's it.

Of course, inside the kernel, inodes are used for many purposes, because the kernel uses inodes instead of filenames to reference a particular file. Even FAT filesystems which do not natively have an inode, have a way to provide a "fake" inode in Linux, because the upper layers of the filesystem requires all filesystems to have something like an inode that is unique for each file. [I seem to remember that FAT uses the first cluster of the file as an inode, and since no two files can use the same cluser, there's no problem with that].

--
Mats

Originally Posted by rak1986

I want to use file inodes to open, read, write, seek files etc. This can be used for file copying. The reason why I asked whether these FILE I/O operations can be performed using Inode number instead of file names is that it might give better performance since inode numbers are unique.

And my answer is that you can not do this (from user-mode, at the very least). In the whole scheme of things, finding the file is generally a small portion of the time needed to copy the file. If the file is bigger than a few kilobytes, it is almost certainly going to take 100x times longer to reach the files physical location on the hard-disk, than it takes to find the file within the directory. The inode would have to be validated (you wouldn't be allowed to just pass in any random number), so reading the directory for the relevant file would still be needed (even if that process also is about 100x slower than parsing through the directory to find the name).

So you are shooting mosquito's with a cannon, really.

--
Mats