[ofa-general] Re: IsSMdisabled and user_mad.c

[Hal Rosenstock]

On Wed, 2007-03-21 at 16:25, Roland Dreier wrote:
>  > > I guess it's OK, although I would also like to know how you plan to
>  > > handle the interaction between IsSM and IsSMDisabled -- eg what if a
>  > > process opens issm0 and then another process tries to open
>  > > issmdisabled0?  Or conversely if issmdisabled0 is open, what happens
>  > > when someone opens issm0?
>  > 
>  > I would think those are error cases. Does that make sense ? If so, what
>  > error makes most sense ? EINVAL or something else ?
> 
> That's not really in keeping with the current interface.  Right now if
> one process opens issm0 and then a second process tries to open, the
> second process blocks until the first one closes the file.  Would it
> make more sense to make the issmdisabled interface work in a similar
> way, i.e. only one of issm and issmdisabled can be open at any time,
> and an attempt to open both would block the second attempt?

Sure; it could work that way too. I'll work on a patch for this over the
next couple days.

-- Hal