Roland Dreier wrote:
> - wc->pkey_index = ntohl(cqe->immed_rss_invalid) >> 16;
> + wc->pkey_index = (uint16_t)(ntohl(cqe->immed_rss_invalid) &
0x7f);
This is pretty silly. We don't allow userspace to create QP1 anyway,
so is there any point setting the pkey_index field here at all?
You are absolutely right - QP1 is being used only in kernel level but i
wanted to be consistent
with the kernel level, so i fixed this line too.
I checked this issue and every (user) low level driver library, until today,
filled the value of the pkey_index and i didn't want that libmlx4 will
be different ...
Do you think that this code should be removed from all of the user level
libraries?
Dotan
_______________________________________________
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
