dql: dql_queued() should write first to reduce bus transactions

While doing high throughput test on a BQL enabled NIC,
I found a very high cost in ndo_start_xmit() when accessing BQL data.

It turned out the problem was caused by compiler trying to be
smart, but involving a bad MESI transaction :

0.05 │ mov 0xc0(%rax),%edi // LOAD dql->num_queued
0.48 │ mov %edx,0xc8(%rax) // STORE dql->last_obj_cnt = count
58.23 │ add %edx,%edi
0.58 │ cmp %edi,0xc4(%rax)
0.76 │ mov %edi,0xc0(%rax) // STORE dql->num_queued += count
0.72 │ js bd8

I got an incredible 10 % gain [1] by making sure cpu do not attempt
to get the cache line in Shared mode, but directly requests for
ownership.

New code :
mov %edx,0xc8(%rax) // STORE dql->last_obj_cnt = count
add %edx,0xc0(%rax) // RMW dql->num_queued += count
mov 0xc4(%rax),%ecx // LOAD dql->adj_limit
mov 0xc0(%rax),%edx // LOAD dql->num_queued
cmp %edx,%ecx

The TX completion was running from another cpu, with high interrupts
rate.

Note that I am using barrier() as a soft hint, as mb() here could be
too heavy cost.

[1] This was a netperf TCP_STREAM with TSO disabled, but GSO enabled.

Signed-off-by: Eric Dumazet <[e-mail]>
Acked-by: Jesper Dangaard Brouer <[e-mail]>
Signed-off-by: David S. Miller <[e-mail]>

include/linux/dynamic_queue_limits.h

···
 {
 	BUG_ON(count > DQL_MAX_OBJECT);
 
-	dql->num_queued += count;
 	dql->last_obj_cnt = count;
+
+	/* We want to force a write first, so that cpu do not attempt
+	 * to get cache line containing last_obj_cnt, num_queued, adj_limit
+	 * in Shared state, but directly does a Request For Ownership
+	 * It is only a hint, we use barrier() only.
+	 */
+	barrier();
+
+	dql->num_queued += count;
 }
 
 /* Returns how many objects can be queued, < 0 indicates over limit. */
 static inline int dql_avail(const struct dql *dql)
 {
-	return dql->adj_limit - dql->num_queued;
+	return ACCESS_ONCE(dql->adj_limit) - ACCESS_ONCE(dql->num_queued);
 }
 
 /* Record number of completed objects and recalculate the limit. */