[openib-general] Some Missing Features from mthca/user MAD access

[Hal Rosenstock]

On Thu, 2005-01-06 at 11:01, Roland Dreier wrote:
> > The simplest approach I think is, keep some file open, when its closed
> > (which happends automatically when the process dies) clean the is_sm bit.
> 
> This does seem to be the simplest approach.  However, there are two issues
> I'm still trying to figure out:
> 
>  - Where should the file be?  Do we really want to create a bunch of issm
>    files in /dev that only support open() and close()?  I can't think
> of anything
>    better, though.

There may be other is_xxx files too at some point.

>  - How does this interact with the kernel's own use of the IsSM bit?  Should
>    ib_port_modify() fail if an attempt to set the IsSM bit is made when the bit
>    is already sent?

The kernel has no need to set IsSM on its own; only on behalf of the SM.
I think the kernel just has to be trusted on this. I don't see a way for
the kernel to know whether the bit is being properly set or cleared for
that matter. I would presume one has to be root from user space to set
or clear this bit too so there is a level of trust there too.

-- Hal