[ofw] RE: PATCH: Fix error handling of mlx4_bus driver.

Tzachi Dar tzachid at mellanox.co.il
Wed Feb 11 02:47:26 PST 2009


Applied on 1941.

As for Fab comment:
I have seen cases in which the cause was something like:

If ((prot !=6) &&
     ((tcp.src.port == 57) || (tcp.dst.port == 57)) &&
      (! Config.state != on)) {
	ASSERT(FALSE);
	return;
}

At this cases writing the assert correctly (and maintaining it) becomes
very hard.
So, I got my self a habit to use ASSERT(FALSE) which is very easy to
read / maintain.

In any case, the places that Fab has indicated were simple, so I have
changed them based
On his request.


Thanks
Tzachi 

> -----Original Message-----
> From: Fab Tillier [mailto:ftillier at windows.microsoft.com] 
> Sent: Monday, February 09, 2009 8:07 PM
> To: Tzachi Dar; ofw at lists.openfabrics.org
> Subject: RE: PATCH: Fix error handling of mlx4_bus driver.
> 
> >Index: Q:/projinf2/trunk/hw/mlx4/kernel/bus/core/device.c
> >===================================================================
> >--- Q:/projinf2/trunk/hw/mlx4/kernel/bus/core/device.c 
> (revision 3919)
> >+++ Q:/projinf2/trunk/hw/mlx4/kernel/bus/core/device.c 
> (revision 3920)
> >@@ -332,6 +332,11 @@
> >  struct ib_client_data *context, *tmp;
> >  unsigned long flags;
> >
> >+ if(device->reg_state != IB_DEV_REGISTERED) {  ASSERT(FALSE);
> 
> This would be better as ASSERT( device->reg_state == 
> IB_DEV_REGISTERED ) - that makes the assertion message a 
> little clearer, and will have the same effect.
> 
> >+  return;
> >+ }
> >+
> 



More information about the ofw mailing list