Hi.
I believe that the same bug exactly exists in ibv_cmd_post_recv and
ibv_post_srq_recv.
thanks
Dotan
Ralph Campbell wrote:
There are error cases in ib_uverbs_post_send() where the return
value is negative (i.e., an error) and yet a non-zero resp.bad_wr
is not written back to ibv_cmd_post_send(). In this case,
ibv_cmd_post_send() should still set the bad_wr pointer.
It seems to me that this should go into OFED 1.3.
I will leave the mechanics of how that happens to Roland and Vlad.
Signed-off-by: Ralph Campbell <[EMAIL PROTECTED]>
diff --git a/src/cmd.c b/src/cmd.c
index 6d4331f..305e263 100644
--- a/src/cmd.c
+++ b/src/cmd.c
@@ -884,7 +884,8 @@ int ibv_cmd_post_send(struct ibv_qp *ibqp, struct
ibv_send_wr *wr,
while (--wr_count)
i = i->next;
*bad_wr = i;
- }
+ } else if (ret)
+ *bad_wr = wr;
return ret;
}
_______________________________________________
ewg mailing list
[EMAIL PROTECTED]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
_______________________________________________
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
