Re: [PATCH 3/3] sanitize building of fs/compat_ioctl.c

[Andi Kleen <ak@suse.de>]

"David S. Miller" <davem@davemloft.net> writes:

> From: Matthew Wilcox <matthew@wil.cx>
> Date: Tue, 13 Dec 2005 10:34:34 -0700
> 
> > The 64-bit code doesn't compile because Andi keeps blocking the
> > is_compat_task() stuff.
> 
> The one place where I ever thought that was necessary, the
> USB async userspace I/O operation stuff, was solved much more
> cleanly with ->compat_ioctl() file_operations handlers.
> 
> What do you really still need it for at this point?

input needs it :/ Take a look at drivers/input/evdev.c:evdev_write_compat
Someone should roast in hell for that code.

> I also would like to avoid it if possible.

I have given in for now. Assuming the test is done on a flag that is only set
by the system call entry path. But I still think it will result in
a lot of ugly code. For for read/write it's hard to avoid because
there are so many variants and we have too many message passing
protocols now.

That said I have been too lazy so far to actually implement it:/

-Andi