Re: [RFC] kern: simple futex for gnumach (version 10)

[Richard Braun]

On Tue, Jan 07, 2014 at 09:29:45AM +0100, Marin Ramesa wrote:
> On 01/06/2014 03:00:35 PM, Richard Braun wrote:
> >We also need to make sure nothing bad happens to an (object, offset)
> >pair when, say, copy-on-write is performed by the VM system...
> 
> I'll add a check_consistency() routine to futex_wake() which repeats the
> vm_map_lookup(). I'll use a sum of object address and offset for the
> event.

I'm not sure you get the point. The problem is that copy-on-write will
create shadow objects, possibly making your new lookup wrong. But
normally, such mappings are private, where a task-local address will
do the job whatever the mapping type, as long as it's valid.

> So, if the event is not the same, I'll just return from futex_wake().



> >All this looks quite tricky to me, and certainly not a "small hack
> >entry" :-/.
> 
> I won't give up. I've spent too much time on this already. I guess
> I'll just
> have to finish it.

Hum no, this is clearly not "too much time" yet. But I understand. Just
know that we may not be able to find the time to validate and debug your
code. I'm looking at it as a reviewer only, I will miss tricky details
too, and I won't test it since I have other things on my to do list.
Same goes for practically everyone else involved in the project. Expect
to be on your own for much of the development.

-- 
Richard Braun