Re: [PATCH 2/3] block: fail SCSI passthrough ioctls on partition devices

[Paolo Bonzini <pbonzini@redhat.com>]

On 12/22/2011 08:44 PM, Linus Torvalds wrote:
>>> >>  This kind of crazy needs to go away.
>> >
>> >  What crazy?  It's not a permission problem.  Sending a SCSI command to a
>> >  partition makes no sense.  A permission problem implies that somehow you
>> >  should be able to fix it by granting additional permissions, which is not
>> >  the case here.
> Ahh, I misread the intention here, and didn't notice that it was doing
> it on the stupid SCSI ioctl commands, not the lowlevel SCSI "cmd". The
> fact that the changelog talked about sending read/write commands down
> to the disk confused me.
>
> But please do use ENOIOCTLCMD directly then, instead of using ENOTTY
> and turning it into ENOIOCTLCMD.

I disagree.  ENOTTY is perfect in all cases except the compat_ioctl 
(which I'm not denying is ugly, but beautifying it would make everything 
else ugly).

In fact ENOTTY means "fail", ENOIOCTLCMD means "handle this elsewhere". 
  Only with compat_ioctl it makes sense to "handle this elsewhere" (we 
know that we will get it again in the non-compat fallback path, and 
return -ENOTTY).

Secondarily, ENOIOCTLCMD is ultimately turned into EINVAL when the 
system call returns (not ENOTTY).

Paolo