Re: [PATCH v3 part1 1/4] sg_io: pass request_queue to blk_verify_command

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Fri, 2013-05-24 at 10:32 +0200, Paolo Bonzini wrote:
> Il 24/05/2013 10:03, James Bottomley ha scritto:
> >>>>> > >>> Does anyone in the real world actually care about this bug?
> >>>> > >>
> >>>> > >> Yes, or I would move on and not waste so much time on this.
> >>> > > 
> >>> > > Fine, so produce a simple fix for this bug which we can discuss that's
> >>> > > not tied to this feature.
> >> > 
> >> > Honestly, I have no idea how this is even possible.
> > Really?  It looks to me like a simple block on the commands for disk
> > devices in the opcode switch would do it (with a corresponding change to
> > sg.c:sg_allow_access).
> 
> Which switch?  What I can do is something like this in blk_verify_command:

not in blk_verify_command: outside of it, in the three places it's used.

>         if (q->sgio_type == TYPE_ROM)
> 		return 0;
> 	if (rq->cmd[0] == 0xA4)
> 		return -EPERM;
> 	if (!is_write &&
> 	    (req->cmd[0] == ... || rq->cmd[0] == ...))
> 		return -EPERM;
> 
> But then the particular patch you're replying to is still necessary,
> and you're slowing down blk_verify_command.

It's a set if if switches in non performance critical code.

>   It may be fine for stable
> and -rc, but IMHO it calls for a better implementation in 3.11.

What goes into stable should be what goes into the real kernel and it
helps separate the bug fix from feature argument.

James

> (Besides, I did it like this because it is what Tejun suggested).