[openib-general] QP sq->tail always = 0

[Jeff Carr]

On 05/26/05 14:17, Jeff Carr wrote:

> When I run my test, the output for the server and client look similar.
> 
> For the server, it first runs: ib_post_recv() for N times (N=5 here)

sorry s/ib_post_recv/ib_post_send/

> ib_mthca 0000:09:00.0: SQ 050406 (1 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: SQ 050406 (2 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: SQ 050406 (3 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: SQ 050406 (4 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: SQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 0 tail, 5 max)
> 
> Then, the client sends N=5 responses via ib_post_recv() that the server 
> sees via ib_poll_cq().

Then the server runs ib_poll_cq() to recieve the responses from the 
client that the client sends vi ib_post_send(). ib_post_recv runs on 
both the client and server N times before any transfers begain.

> 
> ib_mthca 0000:09:00.0: SQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: SQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 1 tail, 5 max)
> ib_mthca 0000:09:00.0: SQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 2 tail, 5 max)
> ib_mthca 0000:09:00.0: SQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 3 tail, 5 max)
> ib_mthca 0000:09:00.0: SQ 050406 (5 head, 0 tail, 5 max)
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 4 tail, 5 max)
> 
> ib_mthca 0000:09:00.0: RQ 050406 (5 head, 5 tail, 5 max
> ib_mthca 0000:09:00.0: SQ 050406 full (5 head, 0 tail, 5 max, 0 nreq)
> 
> At this point, the RQ is empty: (head == tail)
> And at this point the SQ is full: ((head - tail) == max)

Regards,
Jeff