Re: [patch] epoll use a single inode ...

[Linus Torvalds]

On Tue, 6 Mar 2007, Davide Libenzi wrote:
> 
> I'm OK with everything that avoid code duplication due to those fake 
> inodes. The change can be localized inside the existing API, so it doesn't 
> really affect me externally.

Can you try with the first patch version that doesn't do anything special 
at all, and just uses a single dentry.

Yeah, the dentry name will be identical, and so you would see something 
like "7 -> signalfd:signalfd" when you do "ls -l /proc/<pid>/fd/" on a 
task that has such a special file descriptor (with no way to tell 
different timerfd's and signalfd's apart), but I think it's better to 
start off simple than to overdesign things.

And trying to tell them apart sounds a bit like overdesign, if only 
because I really don't see why anybody would really *care*. So it's a 
timer for poll/select/epoll - why care about anything else?

If it really really turns out that people care, we know how we can do it. 
We'd hook into "proc_fd_link()" and we'd allow a per-file mntget/dget that 
we could use to let special filesystems create fake entries on demand. So 
it's not impossible, it's just likely simply not needed.

			Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/