[okl4-developer] IDL & vtables

[Ryan Heffernan]

Actually I'm using the current OKL4 release, but I'm porting a server
written for the 2006 release to the current one, which is why I am
having this problem, sorry if i didn't make this clear. I've been able
to solve the problem though, a local IDL expert at Nicta straightened it
out. It looks like the Magpie-generated server_loop now does all the
work that previously had to be implemented by the server itself, which
is what all the vtable stuff was about. 

As far as the packet_t goes, it looks like the stream_pkt won't quite
work for what I want, but I think i might be able to incorporate what I
need into the driverv2 code. Thanks!

Ryan  

On Tue, 2008-03-04 at 19:18 +1100, Nelson Tam wrote:
> Hi Ryan,
> 
> Is there any reason why you're using the 2006 version?  The current  
> public release (1.5.2) doesn't use vtables anymore.  Is it possible  
> for you to switch to the latest release?
> 
> As for your second question, it seems like you're using 1.5.2 there (I  
> don't think device_core existed back in 2006).  You should use  
> driverv2, which is the new driver framework, because it has updated  
> drivers for supported platforms.  If you need to use packet_t, you're  
> better off copying the packet_t definition to libs/driverv2/include/ 
> driver.h.
> 
> Also, have a look at stream_pkt_t in libs/driverv2/include/stream.h  
> and see if that's suitable for your purposes.
> 
> On 04/03/2008, at 11:12 AM, Ryan Heffernan wrote:
> 
> > I'm currently dealing with some code written for an old (2006) version
> > of Nicta L4 and Iguana, and I'm at a standstill with the IDL compiler
> > and some vtables. Firstly, i wanted to ask if vtables are still in use
> > in OKL4/Iguana, and if so how they get generated? Looking at the timer
> > and serial main.c files it looks like they use vtables, but then I  
> > don't
> > think that code gets compiled because virtual drivers are used.
> >
> > Here's an example of the vtable use in my code:
> >
> > 	void *device_vtable[DEVICE_DEFAULT_VTABLE_SIZE] =
> > DEVICE_DEFAULT_VTABLE;
> >
> > and later on...
> >
> >
> > 	int a = (magpie_get_interface_bigid(&msgbuf));
> >                switch (a) {
> >                case 42:
> >                        idl4_process_request(&partner, &msgtag,  
> > &msgbuf,
> > &cnt, device_vtable[idl4_get_function_id(&msgtag) & DEVICE_FID_MASK]);
> >
> > The .idl4 file for this server does indeed contain a "device"  
> > interface,
> > but the build system gives the following complaints:
> >
> >  `DEVICE_DEFAULT_VTABLE_SIZE' undeclared here (not in a function)
> >  `DEVICE_DEFAULT_VTABLE' undeclared here (not in a function)
> >
> > If vtables are still being used, they must be defined and/or generated
> > differently than they were in 2006, so it would be a big help to me if
> > someone could explain how this process works or just simply how I can
> > work around it. Thanks!
> >
> > Also, my code needs to make use of the packet_t structure defined in
> > driver/include/driver.h, but the iguana SConstruct builds
> > driverv2/include/driver.h instead. So I hacked the following code in  
> > the
> > Sconstruct to make it build the driver files I want:
> >
> > if ig_env.machine.device_core is not None:
> >    ig_env.Package("libs/driver")#NEW
> >    ig_env.Package("drivers/drv_smc91x")#NEW
> >
> > Do you think this could eventually have disastrous consequences?
> 
> --
> (nt)
> 
> Nelson Tam
> nelson at ok-labs.com
> 
> 
>