net_dim() is currently passed a struct dim_sample argument by value.
struct dim_sample is 24 bytes. Since this is greater 16 bytes, x86-64
passes it on the stack. All callers have already initialized dim_sample
on the stack, so passing it by value requires pushing a duplicated copy
to the stack. Either witing to the stack and immediately reading it, or
perhaps dereferencing addresses relative to the stack pointer in a chain
of push instructions, seems to perform quite poorly.
In a heavy TCP workload, mlx5e_handle_rx_dim() consumes 3% of CPU time,
94% of which is attributed to the first push instruction to copy
dim_sample on the stack for the call to net_dim():
// Call ktime_get()
0.26 |4ead2: call 4ead7 <mlx5e_handle_rx_dim+0x47>
// Pass the address of struct dim in %rdi
|4ead7: lea 0x3d0(%rbx),%rdi
// Set dim_sample.pkt_ctr
|4eade: mov %r13d,0x8(%rsp)
// Set dim_sample.byte_ctr
|4eae3: mov %r12d,0xc(%rsp)
// Set dim_sample.event_ctr
0.15 |4eae8: mov %bp,0x10(%rsp)
// Duplicate dim_sample on the stack
94.16 |4eaed: push 0x10(%rsp)
2.79 |4eaf1: push 0x10(%rsp)
0.07 |4eaf5: push %rax
// Call net_dim()
0.21 |4eaf6: call 4eafb <mlx5e_handle_rx_dim+0x6b>
To allow the caller to reuse the struct dim_sample already on the stack,
pass the struct dim_sample by reference to net_dim().
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Link: https://patch.msgid.link/20241031002326.3426181-2-csander@purestorage.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
my_entity->bytes,
&dim_sample);
/* Call net DIM */
- net_dim(&my_entity->dim, dim_sample);
+ net_dim(&my_entity->dim, &dim_sample);
...
}
rx_ring->rx_stats.bytes,
&dim_sample);
- net_dim(&ena_napi->dim, dim_sample);
+ net_dim(&ena_napi->dim, &dim_sample);
rx_ring->per_napi_packets = 0;
}
if (priv->dim.use_dim) {
dim_update_sample(priv->dim.event_ctr, priv->dim.packets,
priv->dim.bytes, &dim_sample);
- net_dim(&priv->dim.dim, dim_sample);
+ net_dim(&priv->dim.dim, &dim_sample);
}
return work_done;
cpr->rx_packets,
cpr->rx_bytes,
&dim_sample);
- net_dim(&cpr->dim, dim_sample);
+ net_dim(&cpr->dim, &dim_sample);
}
return work_done;
}
cpr_rx->rx_packets,
cpr_rx->rx_bytes,
&dim_sample);
- net_dim(&cpr->dim, dim_sample);
+ net_dim(&cpr->dim, &dim_sample);
}
return work_done;
}
if (ring->dim.use_dim) {
dim_update_sample(ring->dim.event_ctr, ring->dim.packets,
ring->dim.bytes, &dim_sample);
- net_dim(&ring->dim.dim, dim_sample);
+ net_dim(&ring->dim.dim, &dim_sample);
}
return work_done;
v->rx_ring.stats.packets,
v->rx_ring.stats.bytes,
&dim_sample);
- net_dim(&v->rx_dim, dim_sample);
+ net_dim(&v->rx_dim, &dim_sample);
}
static int enetc_bd_ready_count(struct enetc_bdr *tx_ring, int ci)
dim_update_sample(tqp_vector->event_cnt, rx_group->total_packets,
rx_group->total_bytes, &sample);
- net_dim(&rx_group->dim, sample);
+ net_dim(&rx_group->dim, &sample);
}
static void hns3_update_tx_int_coalesce(struct hns3_enet_tqp_vector *tqp_vector)
dim_update_sample(tqp_vector->event_cnt, tx_group->total_packets,
tx_group->total_bytes, &sample);
- net_dim(&tx_group->dim, sample);
+ net_dim(&tx_group->dim, &sample);
}
static int hns3_nic_common_poll(struct napi_struct *napi, int budget)
struct dim_sample dim_sample;
__ice_update_sample(q_vector, tx, &dim_sample, true);
- net_dim(&tx->dim, dim_sample);
+ net_dim(&tx->dim, &dim_sample);
}
if (ITR_IS_DYNAMIC(rx)) {
struct dim_sample dim_sample;
__ice_update_sample(q_vector, rx, &dim_sample, false);
- net_dim(&rx->dim, dim_sample);
+ net_dim(&rx->dim, &dim_sample);
}
}
idpf_update_dim_sample(q_vector, &dim_sample, &q_vector->tx_dim,
packets, bytes);
- net_dim(&q_vector->tx_dim, dim_sample);
+ net_dim(&q_vector->tx_dim, &dim_sample);
check_rx_itr:
if (!IDPF_ITR_IS_DYNAMIC(q_vector->rx_intr_mode))
idpf_update_dim_sample(q_vector, &dim_sample, &q_vector->rx_dim,
packets, bytes);
- net_dim(&q_vector->rx_dim, dim_sample);
+ net_dim(&q_vector->rx_dim, &dim_sample);
}
/**
rx_frames + tx_frames,
rx_bytes + tx_bytes,
&dim_sample);
- net_dim(&cq_poll->dim, dim_sample);
+ net_dim(&cq_poll->dim, &dim_sample);
}
int otx2_napi_handler(struct napi_struct *napi, int budget)
eth->rx_bytes += bytes;
dim_update_sample(eth->rx_events, eth->rx_packets, eth->rx_bytes,
&dim_sample);
- net_dim(ð->rx_dim, dim_sample);
+ net_dim(ð->rx_dim, &dim_sample);
if (xdp_flush)
xdp_do_flush();
dim_update_sample(eth->tx_events, eth->tx_packets, eth->tx_bytes,
&dim_sample);
- net_dim(ð->tx_dim, dim_sample);
+ net_dim(ð->tx_dim, &dim_sample);
if (mtk_queue_stopped(eth) &&
(atomic_read(&ring->free_count) > ring->thresh))
return;
dim_update_sample(sq->cq.event_ctr, stats->packets, stats->bytes, &dim_sample);
- net_dim(sq->dim, dim_sample);
+ net_dim(sq->dim, &dim_sample);
}
static void mlx5e_handle_rx_dim(struct mlx5e_rq *rq)
return;
dim_update_sample(rq->cq.event_ctr, stats->packets, stats->bytes, &dim_sample);
- net_dim(rq->dim, dim_sample);
+ net_dim(rq->dim, &dim_sample);
}
void mlx5e_trigger_irq(struct mlx5e_icosq *sq)
} while (u64_stats_fetch_retry(&r_vec->rx_sync, start));
dim_update_sample(r_vec->event_ctr, pkts, bytes, &dim_sample);
- net_dim(&r_vec->rx_dim, dim_sample);
+ net_dim(&r_vec->rx_dim, &dim_sample);
}
if (r_vec->nfp_net->tx_coalesce_adapt_on && r_vec->tx_ring) {
} while (u64_stats_fetch_retry(&r_vec->tx_sync, start));
dim_update_sample(r_vec->event_ctr, pkts, bytes, &dim_sample);
- net_dim(&r_vec->tx_dim, dim_sample);
+ net_dim(&r_vec->tx_dim, &dim_sample);
}
return pkts_polled;
} while (u64_stats_fetch_retry(&r_vec->rx_sync, start));
dim_update_sample(r_vec->event_ctr, pkts, bytes, &dim_sample);
- net_dim(&r_vec->rx_dim, dim_sample);
+ net_dim(&r_vec->rx_dim, &dim_sample);
}
if (r_vec->nfp_net->tx_coalesce_adapt_on && r_vec->tx_ring) {
} while (u64_stats_fetch_retry(&r_vec->tx_sync, start));
dim_update_sample(r_vec->event_ctr, pkts, bytes, &dim_sample);
- net_dim(&r_vec->tx_dim, dim_sample);
+ net_dim(&r_vec->tx_dim, &dim_sample);
}
return pkts_polled;
dim_update_sample(qcq->cq.bound_intr->rearm_count,
pkts, bytes, &dim_sample);
- net_dim(&qcq->dim, dim_sample);
+ net_dim(&qcq->dim, &dim_sample);
}
int ionic_tx_napi(struct napi_struct *napi, int budget)
u64_stats_read(&rq->stats.bytes),
&cur_sample);
- net_dim(&rq->dim, cur_sample);
+ net_dim(&rq->dim, &cur_sample);
rq->packets_in_napi = 0;
}
d->frames += frames;
dim_update_sample(d->event_ctr, d->frames, d->bytes, &dim_sample);
- net_dim(&d->rx_dim, dim_sample);
+ net_dim(&d->rx_dim, &dim_sample);
spin_unlock(&d->dim_lock);
}
* This is the main logic of the algorithm, where data is processed in order
* to decide on next required action.
*/
-void net_dim(struct dim *dim, struct dim_sample end_sample);
+void net_dim(struct dim *dim, const struct dim_sample *end_sample);
/* RDMA DIM */
return dim->profile_ix != prev_ix;
}
-void net_dim(struct dim *dim, struct dim_sample end_sample)
+void net_dim(struct dim *dim, const struct dim_sample *end_sample)
{
struct dim_stats curr_stats;
u16 nevents;
switch (dim->state) {
case DIM_MEASURE_IN_PROGRESS:
nevents = BIT_GAP(BITS_PER_TYPE(u16),
- end_sample.event_ctr,
+ end_sample->event_ctr,
dim->start_sample.event_ctr);
if (nevents < DIM_NEVENTS)
break;
- if (!dim_calc_stats(&dim->start_sample, &end_sample, &curr_stats))
+ if (!dim_calc_stats(&dim->start_sample, end_sample, &curr_stats))
break;
if (net_dim_decision(&curr_stats, dim)) {
dim->state = DIM_APPLY_NEW_PROFILE;
}
fallthrough;
case DIM_START_MEASURE:
- dim_update_sample(end_sample.event_ctr, end_sample.pkt_ctr,
- end_sample.byte_ctr, &dim->start_sample);
+ dim_update_sample(end_sample->event_ctr, end_sample->pkt_ctr,
+ end_sample->byte_ctr, &dim->start_sample);
dim->state = DIM_MEASURE_IN_PROGRESS;
break;
case DIM_APPLY_NEW_PROFILE:
projects / linux-block.git / commitdiff