Old code restricted the number of inspected SBALs to
QDIO_MAX_BUFFERS_PER_Q - 1, as otherwise the first_to_check and
first_to_kick cursors could overlap. Subsequent code would then assume that
there was no progress on the queue, when in fact _all_ SBALs on the queue
were ready-to-process.
This limitation no longer applies, so allow the queue-scan code to inspect
all SBALs on the queue. Note that qeth requires an additional patch
("s390/qeth: stop/wake TX queues based on their fill level"), to avoid
potential queue stalls when all 128 SBALs are reported as ready-to-process.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
multicast_outbound(q)))
qdio_siga_sync_q(q);
- /*
- * Don't check 128 buffers, as otherwise qdio_inbound_q_moved
- * would return 0.
- */
- count = min(atomic_read(&q->nr_buf_used), QDIO_MAX_BUFFERS_MASK);
+ count = atomic_read(&q->nr_buf_used);
if (!count)
return 0;