Re: [PATCH] char/mem: Make /dev/port less obviously broken (v0)

[Adam Jackson <ajax@redhat.com>]

On 2/6/12 7:15 PM, Greg KH wrote:

> Who would use this new ioctl?  And if it's been working ok until now,
> why is it needed?

I'll go out on a limb and say nobody's been seriously using /dev/port. 
libpciaccess would like to, since it avoids needing iopl, and also 
avoids duplicating all the kernel's per-arch logic for port access in 
userspace.  But if it's not going to give me the cycle size I asked for 
I need to fix it before I can use it.  Otherwise you get what I'm 
getting, which is a vesa driver that doesn't work.

Adding the ioctl was just me being polite and assuming that user-kernel 
ABI was a thing we actually believe in.  If it's not, great, let's just 
fix /dev/port to not be idiotic.  If it is, I'd prefer not wasting the 
memory on another inode.

> If you want something "new" like this, why not just create /dev/ioport
> or something like that to always use the proper alignment and not need
> an ioctl at all?

If you really want that shed painted a different color, fine.

- ajax