Possible. I was using the OFED build environment in sles10sp1, and without the patch next_port sometimes gets a negative value. This might be the reason it was difficult to reproduce this. Anyway, in order to cover all possibilities (such as C99), I think that next_port should be unsigned.

The problem makes sense to me now, and it explains why it wasn't easily reproducible on other platforms. I'm not sure if we should convert next_port to an unsigned value, or just ensure that it's not negative. It's defined as an int since idr_get_new_above() expects an int. Do we need an explicit cast when calling idr_get_new_above(), or how about just casting next_port to unsigned when initializing it?

- Sean
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

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

Reply via email to