On Sun, 2006-03-05 at 11:59, Sasha Khapyorsky wrote:
> Hi again,
> 
> On 01:49 Fri 03 Mar     , Sasha Khapyorsky wrote:
> > 
> > complib's cl_atomic_() functions are not atomic in current implementation,
> > they use static per function locks so race is possible between different
> > functions (for example between _inc() and _dec()). Such races are pretty
> > reproducible in practice - for example it is used to inc/decrement mad
> > counters in opensm.
> > 
> > This patch is fast workaround. I think we will need to review usage and
> > implementation of atomic stuff in opensm for better solution.
> > 
> 
> The same patch but updated against recent SVN.
> 
> Sasha.
> 
> 
> This is workaround for broken cl_atomic_() stuff used in complib/OpenSM:
> it replaces per function static locks by one globally shared lock.
> In long term we will need better solution IMO.
> 
> Signed-off-by: Sasha Khapyorsky <[EMAIL PROTECTED]>

Thanks. Applied (to both trunk and 1.0 branch).

-- Hal


_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to