* [RFC -next 2/2] net: sched: red: don't reset the backlog on every stat dump
From: Jakub Kicinski @ 2018-01-04 20:18 UTC (permalink / raw)
To: john.fastabend, jiri, xiyou.wangcong, Nogah Frankel, Yuval Mintz
Cc: netdev, oss-drivers, edumazet, Jakub Kicinski
In-Reply-To: <20180104201851.6455-1-jakub.kicinski@netronome.com>
Commit 0dfb33a0d7e2 ("sch_red: report backlog information") copied
child's backlog into RED's backlog. Back then RED did not maintain
its own backlog counts. This has changed after commit 2ccccf5fb43f
("net_sched: update hierarchical backlog too") and commit d7f4f332f082
("sch_red: update backlog as well"). Copying is no longer necessary.
Tested:
$ tc -s qdisc show dev veth0
qdisc red 1: root refcnt 2 limit 400000b min 30000b max 30000b ecn
Sent 20942 bytes 221 pkt (dropped 0, overlimits 0 requeues 0)
backlog 1260b 14p requeues 14
marked 0 early 0 pdrop 0 other 0
qdisc tbf 2: parent 1: rate 1Kbit burst 15000b lat 3585.0s
Sent 20942 bytes 221 pkt (dropped 0, overlimits 138 requeues 0)
backlog 1260b 14p requeues 14
Recently RED offload was added. We need to make sure drivers don't
depend on resetting the stats. This means backlog should be treated
like any other statistic:
total_stat = new_hw_stat - prev_hw_stat;
Unlike for other statistics new_hw_stat < prev_hw_stat can be true.
Adjust mlxsw.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
---
drivers/net/ethernet/mellanox/mlxsw/spectrum.h | 1 +
drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c | 9 +++++++--
net/sched/sch_red.c | 1 -
3 files changed, 8 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
index ff8d32bc852c..6755050e4ee0 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
@@ -237,6 +237,7 @@ struct mlxsw_sp_qdisc {
u64 tx_packets;
u64 drops;
u64 overlimits;
+ u64 backlog;
};
/* No need an internal lock; At worse - miss a single periodic iteration */
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c
index c33beac5def0..d5091740bd40 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c
@@ -212,6 +212,7 @@ mlxsw_sp_qdisc_get_red_stats(struct mlxsw_sp_port *mlxsw_sp_port, u32 handle,
u64 tx_bytes, tx_packets, overlimits, drops;
struct mlxsw_sp_port_xstats *xstats;
struct rtnl_link_stats64 *stats;
+ s64 backlog;
if (mlxsw_sp_qdisc->handle != handle ||
mlxsw_sp_qdisc->type != MLXSW_SP_QDISC_RED)
@@ -226,17 +227,21 @@ mlxsw_sp_qdisc_get_red_stats(struct mlxsw_sp_port *mlxsw_sp_port, u32 handle,
mlxsw_sp_qdisc->overlimits;
drops = xstats->wred_drop[tclass_num] + xstats->tail_drop[tclass_num] -
mlxsw_sp_qdisc->drops;
+ backlog = mlxsw_sp_cells_bytes(mlxsw_sp_port->mlxsw_sp,
+ xstats->backlog[tclass_num]) -
+ mlxsw_sp_qdisc->backlog;
_bstats_update(res->bstats, tx_bytes, tx_packets);
res->qstats->overlimits += overlimits;
res->qstats->drops += drops;
- res->qstats->backlog += mlxsw_sp_cells_bytes(mlxsw_sp_port->mlxsw_sp,
- xstats->backlog[tclass_num]);
+ res->qstats->backlog += backlog;
mlxsw_sp_qdisc->drops += drops;
mlxsw_sp_qdisc->overlimits += overlimits;
mlxsw_sp_qdisc->tx_bytes += tx_bytes;
mlxsw_sp_qdisc->tx_packets += tx_packets;
+ mlxsw_sp_qdisc->backlog += backlog;
+
return 0;
}
diff --git a/net/sched/sch_red.c b/net/sched/sch_red.c
index a392eaa4a0b4..caebb7e37551 100644
--- a/net/sched/sch_red.c
+++ b/net/sched/sch_red.c
@@ -322,7 +322,6 @@ static int red_dump(struct Qdisc *sch, struct sk_buff *skb)
};
int err;
- sch->qstats.backlog = q->qdisc->qstats.backlog;
err = red_dump_offload_stats(sch, &opt);
if (err)
goto nla_put_failure;
--
2.15.1
^ permalink raw reply related
* Re: [net 02/14] Revert "mlx5: move affinity hints assignments to generic code"
From: Saeed Mahameed @ 2018-01-04 20:19 UTC (permalink / raw)
To: Sagi Grimberg, Saeed Mahameed
Cc: David S. Miller, Linux Netdev List, Thomas Gleixner, Jes Sorensen
In-Reply-To: <95a5c181-79fa-da60-1c94-766e19e97dd3@grimberg.me>
On 12/26/2017 9:14 AM, Sagi Grimberg wrote:
>
>> Are you sure it won't get populated at all ? even if you manually set
>> IRQ affinity via sysfs ?
>
> Yes, the msi_desc affinity is not initialized without the affinity
> descriptor passed (which is what PCI_IRQ_AFFINITY is for).
>
>> Anyway we can implement this driver helper function to return the IRQ
>> affinity hint stored in the driver:
>> "cpumask_first(mdev->priv.irq_info[vector].mask);"
>
> minus the cpumask_first, but yea. Please send a new patch so we can
> test it out.
Actually using mdev->priv.irq_info[vector].mask is wrong since it only
gives the initial hint and not the current actual affinity mask.
I found a better way to address this and return the actual dynamic
affinity of an interrupt vector.
I will send a patch to net soon.
^ permalink raw reply
* [net-next 00/10] net: create dynamic software irq moderation library
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
From: Andy Gospodarek <gospo@broadcom.com>
This converts the dynamic interrupt moderation library from the mlx5_en driver
into a library so it can be used by any driver. The penultimatepatch in this
set adds support for interrupt moderation in the bnxt_en driver and the last
patch creates an entry in the MAINTAINERS file.
The main purpose of this code in the mlx5_en driver is to allow an
administrator to make sure that default coalesce settings are optimized
for low latency, but quickly adapt to handle high throughput traffic and
optimize how many packets are received during each napi poll.
For any new driver the following changes would be needed to use this
library:
- add elements in ring struct to track items needed by this library
- create function that can be called to actually set coalesce settings
for the driver
Credit to Rob Rice and Lee Reed for doing some of the initial proof of
concept and testing for this patch and Tal Gilboa and Or Gerlitz for their
comments, etc on this set.
Andy Gospodarek (10):
net/mlx5e: move interrupt moderation structs to new file
net/mlx5e: move interrupt moderation forward declarations
net/mlx5e: remove rq references in mlx5e_rx_am
net/mlx5e: move AM logic enums
net/mlx5e: move generic functions to new file
net/mlx5e: change Mellanox references in DIM code
net: move dynamic interrpt coalescing code to include/linux
net/dim: use struct net_dim_sample as arg to net_dim
bnxt_en: add support for software dynamic interrupt moderation
MAINTAINERS: add entry for Dynamic Interrupt Moderation
MAINTAINERS | 5 +
drivers/net/ethernet/broadcom/bnxt/Makefile | 2 +-
drivers/net/ethernet/broadcom/bnxt/bnxt.c | 52 +++
drivers/net/ethernet/broadcom/bnxt/bnxt.h | 34 +-
drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c | 32 ++
drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 12 +
drivers/net/ethernet/mellanox/mlx5/core/Makefile | 2 +-
drivers/net/ethernet/mellanox/mlx5/core/en.h | 46 +--
drivers/net/ethernet/mellanox/mlx5/core/en_dim.c | 49 +++
.../net/ethernet/mellanox/mlx5/core/en_ethtool.c | 12 +-
drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 32 +-
drivers/net/ethernet/mellanox/mlx5/core/en_rep.c | 4 +-
drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c | 341 -------------------
drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 10 +-
drivers/net/ethernet/mellanox/mlx5/core/net_dim.h | 108 ++++++
include/linux/mlx5/mlx5_ifc.h | 6 -
include/linux/net_dim.h | 372 +++++++++++++++++++++
17 files changed, 693 insertions(+), 426 deletions(-)
create mode 100644 drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c
create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
delete mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/net_dim.h
create mode 100644 include/linux/net_dim.h
--
2.7.4
^ permalink raw reply
* [net-next 01/10] net/mlx5e: move interrupt moderation structs to new file
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
Create new header file to prepare to move code that handles irq
moderation to a library that lives in a header file.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Acked-by: Tal Gilboa <talgi@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en.h | 33 +----------
drivers/net/ethernet/mellanox/mlx5/core/en_dim.h | 75 ++++++++++++++++++++++++
include/linux/mlx5/mlx5_ifc.h | 6 --
3 files changed, 76 insertions(+), 38 deletions(-)
create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 543060c..ddb5429 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -49,6 +49,7 @@
#include "wq.h"
#include "mlx5_core.h"
#include "en_stats.h"
+#include "en_dim.h"
#define MLX5_SET_CFG(p, f, v) MLX5_SET(create_flow_group_in, p, f, v)
@@ -226,12 +227,6 @@ enum mlx5e_priv_flag {
#define MLX5E_MAX_BW_ALLOC 100 /* Max percentage of BW allocation */
#endif
-struct mlx5e_cq_moder {
- u16 usec;
- u16 pkts;
- u8 cq_period_mode;
-};
-
struct mlx5e_params {
u8 log_sq_size;
u8 rq_wq_type;
@@ -472,32 +467,6 @@ struct mlx5e_mpw_info {
u16 skbs_frags[MLX5_MPWRQ_PAGES_PER_WQE];
};
-struct mlx5e_rx_am_stats {
- int ppms; /* packets per msec */
- int bpms; /* bytes per msec */
- int epms; /* events per msec */
-};
-
-struct mlx5e_rx_am_sample {
- ktime_t time;
- u32 pkt_ctr;
- u32 byte_ctr;
- u16 event_ctr;
-};
-
-struct mlx5e_rx_am { /* Adaptive Moderation */
- u8 state;
- struct mlx5e_rx_am_stats prev_stats;
- struct mlx5e_rx_am_sample start_sample;
- struct work_struct work;
- u8 profile_ix;
- u8 mode;
- u8 tune_state;
- u8 steps_right;
- u8 steps_left;
- u8 tired;
-};
-
/* a single cache unit is capable to serve one napi call (for non-striding rq)
* or a MPWQE (for striding rq).
*/
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
new file mode 100644
index 0000000..84b8524
--- /dev/null
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies, Ltd. All rights reserved.
+ * Copyright (c) 2017, Broadcom Limited
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+*/
+
+#ifndef MLX5_AM_H
+#define MLX5_AM_H
+
+struct mlx5e_cq_moder {
+ u16 usec;
+ u16 pkts;
+ u8 cq_period_mode;
+};
+
+struct mlx5e_rx_am_sample {
+ ktime_t time;
+ u32 pkt_ctr;
+ u32 byte_ctr;
+ u16 event_ctr;
+};
+
+struct mlx5e_rx_am_stats {
+ int ppms; /* packets per msec */
+ int bpms; /* bytes per msec */
+ int epms; /* events per msec */
+};
+
+struct mlx5e_rx_am { /* Adaptive Moderation */
+ u8 state;
+ struct mlx5e_rx_am_stats prev_stats;
+ struct mlx5e_rx_am_sample start_sample;
+ struct work_struct work;
+ u8 profile_ix;
+ u8 mode;
+ u8 tune_state;
+ u8 steps_right;
+ u8 steps_left;
+ u8 tired;
+};
+
+enum {
+ MLX5_CQ_PERIOD_MODE_START_FROM_EQE = 0x0,
+ MLX5_CQ_PERIOD_MODE_START_FROM_CQE = 0x1,
+ MLX5_CQ_PERIOD_NUM_MODES
+};
+
+#endif /* MLX5_AM_H */
diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
index d44ec5f..60d17e5 100644
--- a/include/linux/mlx5/mlx5_ifc.h
+++ b/include/linux/mlx5/mlx5_ifc.h
@@ -2918,12 +2918,6 @@ enum {
MLX5_CQC_ST_FIRED = 0xa,
};
-enum {
- MLX5_CQ_PERIOD_MODE_START_FROM_EQE = 0x0,
- MLX5_CQ_PERIOD_MODE_START_FROM_CQE = 0x1,
- MLX5_CQ_PERIOD_NUM_MODES
-};
-
struct mlx5_ifc_cqc_bits {
u8 status[0x4];
u8 reserved_at_4[0x4];
--
2.7.4
^ permalink raw reply related
* [net-next 02/10] net/mlx5e: move interrupt moderation forward declarations
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
Move these to newly created file to prepare to move these functions to a
library.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Acked-by: Tal Gilboa <talgi@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en.h | 4 ----
drivers/net/ethernet/mellanox/mlx5/core/en_dim.h | 5 +++++
2 files changed, 5 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index ddb5429..2ccedf6 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -829,10 +829,6 @@ void mlx5e_dealloc_rx_wqe(struct mlx5e_rq *rq, u16 ix);
void mlx5e_dealloc_rx_mpwqe(struct mlx5e_rq *rq, u16 ix);
void mlx5e_free_rx_mpwqe(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi);
-void mlx5e_rx_am(struct mlx5e_rq *rq);
-void mlx5e_rx_am_work(struct work_struct *work);
-struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode);
-
void mlx5e_update_stats(struct mlx5e_priv *priv, bool full);
int mlx5e_create_flow_steering(struct mlx5e_priv *priv);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
index 84b8524..f5f6535 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
@@ -72,4 +72,9 @@ enum {
MLX5_CQ_PERIOD_NUM_MODES
};
+struct mlx5e_rq;
+void mlx5e_rx_am(struct mlx5e_rq *rq);
+void mlx5e_rx_am_work(struct work_struct *work);
+struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode);
+
#endif /* MLX5_AM_H */
--
2.7.4
^ permalink raw reply related
* [net-next 03/10] net/mlx5e: remove rq references in mlx5e_rx_am
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
This makes mlx5e_am_sample more generic so that it can be called easily
from a driver that does not use the same data structure to store these
values in a single structure.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Acked-by: Tal Gilboa <talgi@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en_dim.h | 6 ++++--
drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c | 22 +++++++++++++---------
drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 5 ++++-
3 files changed, 21 insertions(+), 12 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
index f5f6535..b676a057 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
@@ -72,8 +72,10 @@ enum {
MLX5_CQ_PERIOD_NUM_MODES
};
-struct mlx5e_rq;
-void mlx5e_rx_am(struct mlx5e_rq *rq);
+void mlx5e_rx_am(struct mlx5e_rx_am *am,
+ u16 event_ctr,
+ u64 packets,
+ u64 bytes);
void mlx5e_rx_am_work(struct work_struct *work);
struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
index e401d9d..1630076 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
@@ -264,13 +264,15 @@ static bool mlx5e_am_decision(struct mlx5e_rx_am_stats *curr_stats,
return am->profile_ix != prev_ix;
}
-static void mlx5e_am_sample(struct mlx5e_rq *rq,
+static void mlx5e_am_sample(u16 event_ctr,
+ u64 packets,
+ u64 bytes,
struct mlx5e_rx_am_sample *s)
{
s->time = ktime_get();
- s->pkt_ctr = rq->stats.packets;
- s->byte_ctr = rq->stats.bytes;
- s->event_ctr = rq->cq.event_ctr;
+ s->pkt_ctr = packets;
+ s->byte_ctr = bytes;
+ s->event_ctr = event_ctr;
}
#define MLX5E_AM_NEVENTS 64
@@ -309,20 +311,22 @@ void mlx5e_rx_am_work(struct work_struct *work)
am->state = MLX5E_AM_START_MEASURE;
}
-void mlx5e_rx_am(struct mlx5e_rq *rq)
+void mlx5e_rx_am(struct mlx5e_rx_am *am,
+ u16 event_ctr,
+ u64 packets,
+ u64 bytes)
{
- struct mlx5e_rx_am *am = &rq->am;
struct mlx5e_rx_am_sample end_sample;
struct mlx5e_rx_am_stats curr_stats;
u16 nevents;
switch (am->state) {
case MLX5E_AM_MEASURE_IN_PROGRESS:
- nevents = BIT_GAP(BITS_PER_TYPE(u16), rq->cq.event_ctr,
+ nevents = BIT_GAP(BITS_PER_TYPE(u16), event_ctr,
am->start_sample.event_ctr);
if (nevents < MLX5E_AM_NEVENTS)
break;
- mlx5e_am_sample(rq, &end_sample);
+ mlx5e_am_sample(event_ctr, packets, bytes, &end_sample);
mlx5e_am_calc_stats(&am->start_sample, &end_sample,
&curr_stats);
if (mlx5e_am_decision(&curr_stats, am)) {
@@ -332,7 +336,7 @@ void mlx5e_rx_am(struct mlx5e_rq *rq)
}
/* fall through */
case MLX5E_AM_START_MEASURE:
- mlx5e_am_sample(rq, &am->start_sample);
+ mlx5e_am_sample(event_ctr, packets, bytes, &am->start_sample);
am->state = MLX5E_AM_MEASURE_IN_PROGRESS;
break;
case MLX5E_AM_APPLY_NEW_PROFILE:
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
index ab92298..1849169 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
@@ -79,7 +79,10 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget)
mlx5e_cq_arm(&c->sq[i].cq);
if (MLX5E_TEST_BIT(c->rq.state, MLX5E_RQ_STATE_AM))
- mlx5e_rx_am(&c->rq);
+ mlx5e_rx_am(&c->rq.am,
+ c->rq.cq.event_ctr,
+ c->rq.stats.packets,
+ c->rq.stats.bytes);
mlx5e_cq_arm(&c->rq.cq);
mlx5e_cq_arm(&c->icosq.cq);
--
2.7.4
^ permalink raw reply related
* [net-next 04/10] net/mlx5e: move AM logic enums
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
More movement to help make this code more generic.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Acked-by: Tal Gilboa <talgi@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en_dim.h | 26 ++++++++++++++++++++++
drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c | 25 ---------------------
2 files changed, 26 insertions(+), 25 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
index b676a057..c9f0d05 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
@@ -72,6 +72,32 @@ enum {
MLX5_CQ_PERIOD_NUM_MODES
};
+/* Adaptive moderation logic */
+enum {
+ MLX5E_AM_START_MEASURE,
+ MLX5E_AM_MEASURE_IN_PROGRESS,
+ MLX5E_AM_APPLY_NEW_PROFILE,
+};
+
+enum {
+ MLX5E_AM_PARKING_ON_TOP,
+ MLX5E_AM_PARKING_TIRED,
+ MLX5E_AM_GOING_RIGHT,
+ MLX5E_AM_GOING_LEFT,
+};
+
+enum {
+ MLX5E_AM_STATS_WORSE,
+ MLX5E_AM_STATS_SAME,
+ MLX5E_AM_STATS_BETTER,
+};
+
+enum {
+ MLX5E_AM_STEPPED,
+ MLX5E_AM_TOO_TIRED,
+ MLX5E_AM_ON_EDGE,
+};
+
void mlx5e_rx_am(struct mlx5e_rx_am *am,
u16 event_ctr,
u64 packets,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
index 1630076..337dd60 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
@@ -82,31 +82,6 @@ struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode)
return mlx5e_am_get_profile(rx_cq_period_mode, default_profile_ix);
}
-/* Adaptive moderation logic */
-enum {
- MLX5E_AM_START_MEASURE,
- MLX5E_AM_MEASURE_IN_PROGRESS,
- MLX5E_AM_APPLY_NEW_PROFILE,
-};
-
-enum {
- MLX5E_AM_PARKING_ON_TOP,
- MLX5E_AM_PARKING_TIRED,
- MLX5E_AM_GOING_RIGHT,
- MLX5E_AM_GOING_LEFT,
-};
-
-enum {
- MLX5E_AM_STATS_WORSE,
- MLX5E_AM_STATS_SAME,
- MLX5E_AM_STATS_BETTER,
-};
-
-enum {
- MLX5E_AM_STEPPED,
- MLX5E_AM_TOO_TIRED,
- MLX5E_AM_ON_EDGE,
-};
static bool mlx5e_am_on_top(struct mlx5e_rx_am *am)
{
--
2.7.4
^ permalink raw reply related
* [net-next 05/10] net/mlx5e: move generic functions to new file
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
These functions were identified as ones that could be made generic and
used by multiple drivers. Most of the contents of en_rx_am.c are moved
to net_dim.c.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Acked-by: Tal Gilboa <talgi@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/Makefile | 4 +-
drivers/net/ethernet/mellanox/mlx5/core/en_dim.c | 48 ++++
drivers/net/ethernet/mellanox/mlx5/core/en_dim.h | 1 +
drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c | 320 ---------------------
drivers/net/ethernet/mellanox/mlx5/core/net_dim.c | 307 ++++++++++++++++++++
drivers/net/ethernet/mellanox/mlx5/core/net_dim.h | 109 +++++++
6 files changed, 467 insertions(+), 322 deletions(-)
create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
delete mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/net_dim.h
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/Makefile b/drivers/net/ethernet/mellanox/mlx5/core/Makefile
index 19b21b4..b46b6de2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/Makefile
+++ b/drivers/net/ethernet/mellanox/mlx5/core/Makefile
@@ -14,8 +14,8 @@ mlx5_core-$(CONFIG_MLX5_FPGA) += fpga/cmd.o fpga/core.o fpga/conn.o fpga/sdk.o \
fpga/ipsec.o
mlx5_core-$(CONFIG_MLX5_CORE_EN) += en_main.o en_common.o en_fs.o en_ethtool.o \
- en_tx.o en_rx.o en_rx_am.o en_txrx.o en_stats.o vxlan.o \
- en_arfs.o en_fs_ethtool.o en_selftest.o
+ en_tx.o en_rx.o en_dim.o en_txrx.o en_stats.o vxlan.o \
+ en_arfs.o en_fs_ethtool.o en_selftest.o net_dim.o
mlx5_core-$(CONFIG_MLX5_MPFS) += lib/mpfs.o
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
new file mode 100644
index 0000000..b9b434b
--- /dev/null
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
@@ -0,0 +1,48 @@
+/*
+ * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "en.h"
+
+void mlx5e_rx_am_work(struct work_struct *work)
+{
+ struct mlx5e_rx_am *am = container_of(work, struct mlx5e_rx_am,
+ work);
+ struct mlx5e_rq *rq = container_of(am, struct mlx5e_rq, am);
+ struct mlx5e_cq_moder cur_profile = mlx5e_am_get_profile(am->mode,
+ am->profile_ix);
+
+ mlx5_core_modify_cq_moderation(rq->mdev, &rq->cq.mcq,
+ cur_profile.usec, cur_profile.pkts);
+
+ am->state = MLX5E_AM_START_MEASURE;
+}
+
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
index c9f0d05..5ce8e54 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
@@ -104,5 +104,6 @@ void mlx5e_rx_am(struct mlx5e_rx_am *am,
u64 bytes);
void mlx5e_rx_am_work(struct work_struct *work);
struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode);
+struct mlx5e_cq_moder mlx5e_am_get_profile(u8 cq_period_mode, int ix);
#endif /* MLX5_AM_H */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
deleted file mode 100644
index 337dd60..0000000
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
+++ /dev/null
@@ -1,320 +0,0 @@
-/*
- * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses. You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- * Redistribution and use in source and binary forms, with or
- * without modification, are permitted provided that the following
- * conditions are met:
- *
- * - Redistributions of source code must retain the above
- * copyright notice, this list of conditions and the following
- * disclaimer.
- *
- * - Redistributions in binary form must reproduce the above
- * copyright notice, this list of conditions and the following
- * disclaimer in the documentation and/or other materials
- * provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#include "en.h"
-
-/* Adaptive moderation profiles */
-#define MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
-#define MLX5E_RX_AM_DEF_PROFILE_CQE 1
-#define MLX5E_RX_AM_DEF_PROFILE_EQE 1
-#define MLX5E_PARAMS_AM_NUM_PROFILES 5
-
-/* All profiles sizes must be MLX5E_PARAMS_AM_NUM_PROFILES */
-#define MLX5_AM_EQE_PROFILES { \
- {1, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {8, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {64, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {128, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {256, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
-}
-
-#define MLX5_AM_CQE_PROFILES { \
- {2, 256}, \
- {8, 128}, \
- {16, 64}, \
- {32, 64}, \
- {64, 64} \
-}
-
-static const struct mlx5e_cq_moder
-profile[MLX5_CQ_PERIOD_NUM_MODES][MLX5E_PARAMS_AM_NUM_PROFILES] = {
- MLX5_AM_EQE_PROFILES,
- MLX5_AM_CQE_PROFILES,
-};
-
-static inline struct mlx5e_cq_moder mlx5e_am_get_profile(u8 cq_period_mode, int ix)
-{
- struct mlx5e_cq_moder cq_moder;
-
- cq_moder = profile[cq_period_mode][ix];
- cq_moder.cq_period_mode = cq_period_mode;
- return cq_moder;
-}
-
-struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode)
-{
- int default_profile_ix;
-
- if (rx_cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE)
- default_profile_ix = MLX5E_RX_AM_DEF_PROFILE_CQE;
- else /* MLX5_CQ_PERIOD_MODE_START_FROM_EQE */
- default_profile_ix = MLX5E_RX_AM_DEF_PROFILE_EQE;
-
- return mlx5e_am_get_profile(rx_cq_period_mode, default_profile_ix);
-}
-
-
-static bool mlx5e_am_on_top(struct mlx5e_rx_am *am)
-{
- switch (am->tune_state) {
- case MLX5E_AM_PARKING_ON_TOP:
- case MLX5E_AM_PARKING_TIRED:
- return true;
- case MLX5E_AM_GOING_RIGHT:
- return (am->steps_left > 1) && (am->steps_right == 1);
- default: /* MLX5E_AM_GOING_LEFT */
- return (am->steps_right > 1) && (am->steps_left == 1);
- }
-}
-
-static void mlx5e_am_turn(struct mlx5e_rx_am *am)
-{
- switch (am->tune_state) {
- case MLX5E_AM_PARKING_ON_TOP:
- case MLX5E_AM_PARKING_TIRED:
- break;
- case MLX5E_AM_GOING_RIGHT:
- am->tune_state = MLX5E_AM_GOING_LEFT;
- am->steps_left = 0;
- break;
- case MLX5E_AM_GOING_LEFT:
- am->tune_state = MLX5E_AM_GOING_RIGHT;
- am->steps_right = 0;
- break;
- }
-}
-
-static int mlx5e_am_step(struct mlx5e_rx_am *am)
-{
- if (am->tired == (MLX5E_PARAMS_AM_NUM_PROFILES * 2))
- return MLX5E_AM_TOO_TIRED;
-
- switch (am->tune_state) {
- case MLX5E_AM_PARKING_ON_TOP:
- case MLX5E_AM_PARKING_TIRED:
- break;
- case MLX5E_AM_GOING_RIGHT:
- if (am->profile_ix == (MLX5E_PARAMS_AM_NUM_PROFILES - 1))
- return MLX5E_AM_ON_EDGE;
- am->profile_ix++;
- am->steps_right++;
- break;
- case MLX5E_AM_GOING_LEFT:
- if (am->profile_ix == 0)
- return MLX5E_AM_ON_EDGE;
- am->profile_ix--;
- am->steps_left++;
- break;
- }
-
- am->tired++;
- return MLX5E_AM_STEPPED;
-}
-
-static void mlx5e_am_park_on_top(struct mlx5e_rx_am *am)
-{
- am->steps_right = 0;
- am->steps_left = 0;
- am->tired = 0;
- am->tune_state = MLX5E_AM_PARKING_ON_TOP;
-}
-
-static void mlx5e_am_park_tired(struct mlx5e_rx_am *am)
-{
- am->steps_right = 0;
- am->steps_left = 0;
- am->tune_state = MLX5E_AM_PARKING_TIRED;
-}
-
-static void mlx5e_am_exit_parking(struct mlx5e_rx_am *am)
-{
- am->tune_state = am->profile_ix ? MLX5E_AM_GOING_LEFT :
- MLX5E_AM_GOING_RIGHT;
- mlx5e_am_step(am);
-}
-
-#define IS_SIGNIFICANT_DIFF(val, ref) \
- (((100 * abs((val) - (ref))) / (ref)) > 10) /* more than 10% difference */
-
-static int mlx5e_am_stats_compare(struct mlx5e_rx_am_stats *curr,
- struct mlx5e_rx_am_stats *prev)
-{
- if (!prev->bpms)
- return curr->bpms ? MLX5E_AM_STATS_BETTER :
- MLX5E_AM_STATS_SAME;
-
- if (IS_SIGNIFICANT_DIFF(curr->bpms, prev->bpms))
- return (curr->bpms > prev->bpms) ? MLX5E_AM_STATS_BETTER :
- MLX5E_AM_STATS_WORSE;
-
- if (IS_SIGNIFICANT_DIFF(curr->ppms, prev->ppms))
- return (curr->ppms > prev->ppms) ? MLX5E_AM_STATS_BETTER :
- MLX5E_AM_STATS_WORSE;
-
- if (IS_SIGNIFICANT_DIFF(curr->epms, prev->epms))
- return (curr->epms < prev->epms) ? MLX5E_AM_STATS_BETTER :
- MLX5E_AM_STATS_WORSE;
-
- return MLX5E_AM_STATS_SAME;
-}
-
-static bool mlx5e_am_decision(struct mlx5e_rx_am_stats *curr_stats,
- struct mlx5e_rx_am *am)
-{
- int prev_state = am->tune_state;
- int prev_ix = am->profile_ix;
- int stats_res;
- int step_res;
-
- switch (am->tune_state) {
- case MLX5E_AM_PARKING_ON_TOP:
- stats_res = mlx5e_am_stats_compare(curr_stats, &am->prev_stats);
- if (stats_res != MLX5E_AM_STATS_SAME)
- mlx5e_am_exit_parking(am);
- break;
-
- case MLX5E_AM_PARKING_TIRED:
- am->tired--;
- if (!am->tired)
- mlx5e_am_exit_parking(am);
- break;
-
- case MLX5E_AM_GOING_RIGHT:
- case MLX5E_AM_GOING_LEFT:
- stats_res = mlx5e_am_stats_compare(curr_stats, &am->prev_stats);
- if (stats_res != MLX5E_AM_STATS_BETTER)
- mlx5e_am_turn(am);
-
- if (mlx5e_am_on_top(am)) {
- mlx5e_am_park_on_top(am);
- break;
- }
-
- step_res = mlx5e_am_step(am);
- switch (step_res) {
- case MLX5E_AM_ON_EDGE:
- mlx5e_am_park_on_top(am);
- break;
- case MLX5E_AM_TOO_TIRED:
- mlx5e_am_park_tired(am);
- break;
- }
-
- break;
- }
-
- if ((prev_state != MLX5E_AM_PARKING_ON_TOP) ||
- (am->tune_state != MLX5E_AM_PARKING_ON_TOP))
- am->prev_stats = *curr_stats;
-
- return am->profile_ix != prev_ix;
-}
-
-static void mlx5e_am_sample(u16 event_ctr,
- u64 packets,
- u64 bytes,
- struct mlx5e_rx_am_sample *s)
-{
- s->time = ktime_get();
- s->pkt_ctr = packets;
- s->byte_ctr = bytes;
- s->event_ctr = event_ctr;
-}
-
-#define MLX5E_AM_NEVENTS 64
-#define BITS_PER_TYPE(type) (sizeof(type) * BITS_PER_BYTE)
-#define BIT_GAP(bits, end, start) ((((end) - (start)) + BIT_ULL(bits)) & (BIT_ULL(bits) - 1))
-
-static void mlx5e_am_calc_stats(struct mlx5e_rx_am_sample *start,
- struct mlx5e_rx_am_sample *end,
- struct mlx5e_rx_am_stats *curr_stats)
-{
- /* u32 holds up to 71 minutes, should be enough */
- u32 delta_us = ktime_us_delta(end->time, start->time);
- u32 npkts = BIT_GAP(BITS_PER_TYPE(u32), end->pkt_ctr, start->pkt_ctr);
- u32 nbytes = BIT_GAP(BITS_PER_TYPE(u32), end->byte_ctr,
- start->byte_ctr);
-
- if (!delta_us)
- return;
-
- curr_stats->ppms = DIV_ROUND_UP(npkts * USEC_PER_MSEC, delta_us);
- curr_stats->bpms = DIV_ROUND_UP(nbytes * USEC_PER_MSEC, delta_us);
- curr_stats->epms = DIV_ROUND_UP(MLX5E_AM_NEVENTS * USEC_PER_MSEC,
- delta_us);
-}
-
-void mlx5e_rx_am_work(struct work_struct *work)
-{
- struct mlx5e_rx_am *am = container_of(work, struct mlx5e_rx_am,
- work);
- struct mlx5e_rq *rq = container_of(am, struct mlx5e_rq, am);
- struct mlx5e_cq_moder cur_profile = profile[am->mode][am->profile_ix];
-
- mlx5_core_modify_cq_moderation(rq->mdev, &rq->cq.mcq,
- cur_profile.usec, cur_profile.pkts);
-
- am->state = MLX5E_AM_START_MEASURE;
-}
-
-void mlx5e_rx_am(struct mlx5e_rx_am *am,
- u16 event_ctr,
- u64 packets,
- u64 bytes)
-{
- struct mlx5e_rx_am_sample end_sample;
- struct mlx5e_rx_am_stats curr_stats;
- u16 nevents;
-
- switch (am->state) {
- case MLX5E_AM_MEASURE_IN_PROGRESS:
- nevents = BIT_GAP(BITS_PER_TYPE(u16), event_ctr,
- am->start_sample.event_ctr);
- if (nevents < MLX5E_AM_NEVENTS)
- break;
- mlx5e_am_sample(event_ctr, packets, bytes, &end_sample);
- mlx5e_am_calc_stats(&am->start_sample, &end_sample,
- &curr_stats);
- if (mlx5e_am_decision(&curr_stats, am)) {
- am->state = MLX5E_AM_APPLY_NEW_PROFILE;
- schedule_work(&am->work);
- break;
- }
- /* fall through */
- case MLX5E_AM_START_MEASURE:
- mlx5e_am_sample(event_ctr, packets, bytes, &am->start_sample);
- am->state = MLX5E_AM_MEASURE_IN_PROGRESS;
- break;
- case MLX5E_AM_APPLY_NEW_PROFILE:
- break;
- }
-}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
new file mode 100644
index 0000000..ca05f4e
--- /dev/null
+++ b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
@@ -0,0 +1,307 @@
+/*
+ * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017, Broadcom Limiited. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "en.h"
+
+#define MLX5E_PARAMS_AM_NUM_PROFILES 5
+/* Adaptive moderation profiles */
+#define MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
+#define MLX5E_RX_AM_DEF_PROFILE_CQE 1
+#define MLX5E_RX_AM_DEF_PROFILE_EQE 1
+
+/* All profiles sizes must be MLX5E_PARAMS_AM_NUM_PROFILES */
+#define MLX5_AM_EQE_PROFILES { \
+ {1, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {8, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {64, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {128, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {256, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+}
+
+#define MLX5_AM_CQE_PROFILES { \
+ {2, 256}, \
+ {8, 128}, \
+ {16, 64}, \
+ {32, 64}, \
+ {64, 64} \
+}
+
+static const struct mlx5e_cq_moder
+profile[MLX5_CQ_PERIOD_NUM_MODES][MLX5E_PARAMS_AM_NUM_PROFILES] = {
+ MLX5_AM_EQE_PROFILES,
+ MLX5_AM_CQE_PROFILES,
+};
+
+struct mlx5e_cq_moder mlx5e_am_get_profile(u8 cq_period_mode, int ix)
+{
+ struct mlx5e_cq_moder cq_moder;
+
+ cq_moder = profile[cq_period_mode][ix];
+ cq_moder.cq_period_mode = cq_period_mode;
+ return cq_moder;
+}
+
+struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode)
+{
+ int default_profile_ix;
+
+ if (rx_cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE)
+ default_profile_ix = MLX5E_RX_AM_DEF_PROFILE_CQE;
+ else /* MLX5_CQ_PERIOD_MODE_START_FROM_EQE */
+ default_profile_ix = MLX5E_RX_AM_DEF_PROFILE_EQE;
+
+ return mlx5e_am_get_profile(rx_cq_period_mode, default_profile_ix);
+}
+
+static bool mlx5e_am_on_top(struct mlx5e_rx_am *am)
+{
+ switch (am->tune_state) {
+ case MLX5E_AM_PARKING_ON_TOP:
+ case MLX5E_AM_PARKING_TIRED:
+ return true;
+ case MLX5E_AM_GOING_RIGHT:
+ return (am->steps_left > 1) && (am->steps_right == 1);
+ default: /* MLX5E_AM_GOING_LEFT */
+ return (am->steps_right > 1) && (am->steps_left == 1);
+ }
+}
+
+static void mlx5e_am_turn(struct mlx5e_rx_am *am)
+{
+ switch (am->tune_state) {
+ case MLX5E_AM_PARKING_ON_TOP:
+ case MLX5E_AM_PARKING_TIRED:
+ break;
+ case MLX5E_AM_GOING_RIGHT:
+ am->tune_state = MLX5E_AM_GOING_LEFT;
+ am->steps_left = 0;
+ break;
+ case MLX5E_AM_GOING_LEFT:
+ am->tune_state = MLX5E_AM_GOING_RIGHT;
+ am->steps_right = 0;
+ break;
+ }
+}
+
+static int mlx5e_am_step(struct mlx5e_rx_am *am)
+{
+ if (am->tired == (MLX5E_PARAMS_AM_NUM_PROFILES * 2))
+ return MLX5E_AM_TOO_TIRED;
+
+ switch (am->tune_state) {
+ case MLX5E_AM_PARKING_ON_TOP:
+ case MLX5E_AM_PARKING_TIRED:
+ break;
+ case MLX5E_AM_GOING_RIGHT:
+ if (am->profile_ix == (MLX5E_PARAMS_AM_NUM_PROFILES - 1))
+ return MLX5E_AM_ON_EDGE;
+ am->profile_ix++;
+ am->steps_right++;
+ break;
+ case MLX5E_AM_GOING_LEFT:
+ if (am->profile_ix == 0)
+ return MLX5E_AM_ON_EDGE;
+ am->profile_ix--;
+ am->steps_left++;
+ break;
+ }
+
+ am->tired++;
+ return MLX5E_AM_STEPPED;
+}
+
+static void mlx5e_am_park_on_top(struct mlx5e_rx_am *am)
+{
+ am->steps_right = 0;
+ am->steps_left = 0;
+ am->tired = 0;
+ am->tune_state = MLX5E_AM_PARKING_ON_TOP;
+}
+
+static void mlx5e_am_park_tired(struct mlx5e_rx_am *am)
+{
+ am->steps_right = 0;
+ am->steps_left = 0;
+ am->tune_state = MLX5E_AM_PARKING_TIRED;
+}
+
+static void mlx5e_am_exit_parking(struct mlx5e_rx_am *am)
+{
+ am->tune_state = am->profile_ix ? MLX5E_AM_GOING_LEFT :
+ MLX5E_AM_GOING_RIGHT;
+ mlx5e_am_step(am);
+}
+
+#define IS_SIGNIFICANT_DIFF(val, ref) \
+ (((100 * abs((val) - (ref))) / (ref)) > 10) /* more than 10% difference */
+
+static int mlx5e_am_stats_compare(struct mlx5e_rx_am_stats *curr,
+ struct mlx5e_rx_am_stats *prev)
+{
+ if (!prev->bpms)
+ return curr->bpms ? MLX5E_AM_STATS_BETTER :
+ MLX5E_AM_STATS_SAME;
+
+ if (IS_SIGNIFICANT_DIFF(curr->bpms, prev->bpms))
+ return (curr->bpms > prev->bpms) ? MLX5E_AM_STATS_BETTER :
+ MLX5E_AM_STATS_WORSE;
+
+ if (IS_SIGNIFICANT_DIFF(curr->ppms, prev->ppms))
+ return (curr->ppms > prev->ppms) ? MLX5E_AM_STATS_BETTER :
+ MLX5E_AM_STATS_WORSE;
+
+ if (IS_SIGNIFICANT_DIFF(curr->epms, prev->epms))
+ return (curr->epms < prev->epms) ? MLX5E_AM_STATS_BETTER :
+ MLX5E_AM_STATS_WORSE;
+
+ return MLX5E_AM_STATS_SAME;
+}
+
+static bool mlx5e_am_decision(struct mlx5e_rx_am_stats *curr_stats,
+ struct mlx5e_rx_am *am)
+{
+ int prev_state = am->tune_state;
+ int prev_ix = am->profile_ix;
+ int stats_res;
+ int step_res;
+
+ switch (am->tune_state) {
+ case MLX5E_AM_PARKING_ON_TOP:
+ stats_res = mlx5e_am_stats_compare(curr_stats, &am->prev_stats);
+ if (stats_res != MLX5E_AM_STATS_SAME)
+ mlx5e_am_exit_parking(am);
+ break;
+
+ case MLX5E_AM_PARKING_TIRED:
+ am->tired--;
+ if (!am->tired)
+ mlx5e_am_exit_parking(am);
+ break;
+
+ case MLX5E_AM_GOING_RIGHT:
+ case MLX5E_AM_GOING_LEFT:
+ stats_res = mlx5e_am_stats_compare(curr_stats, &am->prev_stats);
+ if (stats_res != MLX5E_AM_STATS_BETTER)
+ mlx5e_am_turn(am);
+
+ if (mlx5e_am_on_top(am)) {
+ mlx5e_am_park_on_top(am);
+ break;
+ }
+
+ step_res = mlx5e_am_step(am);
+ switch (step_res) {
+ case MLX5E_AM_ON_EDGE:
+ mlx5e_am_park_on_top(am);
+ break;
+ case MLX5E_AM_TOO_TIRED:
+ mlx5e_am_park_tired(am);
+ break;
+ }
+
+ break;
+ }
+
+ if ((prev_state != MLX5E_AM_PARKING_ON_TOP) ||
+ (am->tune_state != MLX5E_AM_PARKING_ON_TOP))
+ am->prev_stats = *curr_stats;
+
+ return am->profile_ix != prev_ix;
+}
+
+static void mlx5e_am_sample(u16 event_ctr,
+ u64 packets,
+ u64 bytes,
+ struct mlx5e_rx_am_sample *s)
+{
+ s->time = ktime_get();
+ s->pkt_ctr = packets;
+ s->byte_ctr = bytes;
+ s->event_ctr = event_ctr;
+}
+
+#define MLX5E_AM_NEVENTS 64
+#define BITS_PER_TYPE(type) (sizeof(type) * BITS_PER_BYTE)
+#define BIT_GAP(bits, end, start) ((((end) - (start)) + BIT_ULL(bits)) & (BIT_ULL(bits) - 1))
+
+static void mlx5e_am_calc_stats(struct mlx5e_rx_am_sample *start,
+ struct mlx5e_rx_am_sample *end,
+ struct mlx5e_rx_am_stats *curr_stats)
+{
+ /* u32 holds up to 71 minutes, should be enough */
+ u32 delta_us = ktime_us_delta(end->time, start->time);
+ u32 npkts = BIT_GAP(BITS_PER_TYPE(u32), end->pkt_ctr, start->pkt_ctr);
+ u32 nbytes = BIT_GAP(BITS_PER_TYPE(u32), end->byte_ctr,
+ start->byte_ctr);
+
+ if (!delta_us)
+ return;
+
+ curr_stats->ppms = DIV_ROUND_UP(npkts * USEC_PER_MSEC, delta_us);
+ curr_stats->bpms = DIV_ROUND_UP(nbytes * USEC_PER_MSEC, delta_us);
+ curr_stats->epms = DIV_ROUND_UP(MLX5E_AM_NEVENTS * USEC_PER_MSEC,
+ delta_us);
+}
+
+void mlx5e_rx_am(struct mlx5e_rx_am *am,
+ u16 event_ctr,
+ u64 packets,
+ u64 bytes)
+{
+ struct mlx5e_rx_am_sample end_sample;
+ struct mlx5e_rx_am_stats curr_stats;
+ u16 nevents;
+
+ switch (am->state) {
+ case MLX5E_AM_MEASURE_IN_PROGRESS:
+ nevents = BIT_GAP(BITS_PER_TYPE(u16), event_ctr,
+ am->start_sample.event_ctr);
+ if (nevents < MLX5E_AM_NEVENTS)
+ break;
+ mlx5e_am_sample(event_ctr, packets, bytes, &end_sample);
+ mlx5e_am_calc_stats(&am->start_sample, &end_sample,
+ &curr_stats);
+ if (mlx5e_am_decision(&curr_stats, am)) {
+ am->state = MLX5E_AM_APPLY_NEW_PROFILE;
+ schedule_work(&am->work);
+ break;
+ }
+ /* fall through */
+ case MLX5E_AM_START_MEASURE:
+ mlx5e_am_sample(event_ctr, packets, bytes, &am->start_sample);
+ am->state = MLX5E_AM_MEASURE_IN_PROGRESS;
+ break;
+ case MLX5E_AM_APPLY_NEW_PROFILE:
+ break;
+ }
+}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/net_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.h
new file mode 100644
index 0000000..5ce8e54
--- /dev/null
+++ b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.h
@@ -0,0 +1,109 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies, Ltd. All rights reserved.
+ * Copyright (c) 2017, Broadcom Limited
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+*/
+
+#ifndef MLX5_AM_H
+#define MLX5_AM_H
+
+struct mlx5e_cq_moder {
+ u16 usec;
+ u16 pkts;
+ u8 cq_period_mode;
+};
+
+struct mlx5e_rx_am_sample {
+ ktime_t time;
+ u32 pkt_ctr;
+ u32 byte_ctr;
+ u16 event_ctr;
+};
+
+struct mlx5e_rx_am_stats {
+ int ppms; /* packets per msec */
+ int bpms; /* bytes per msec */
+ int epms; /* events per msec */
+};
+
+struct mlx5e_rx_am { /* Adaptive Moderation */
+ u8 state;
+ struct mlx5e_rx_am_stats prev_stats;
+ struct mlx5e_rx_am_sample start_sample;
+ struct work_struct work;
+ u8 profile_ix;
+ u8 mode;
+ u8 tune_state;
+ u8 steps_right;
+ u8 steps_left;
+ u8 tired;
+};
+
+enum {
+ MLX5_CQ_PERIOD_MODE_START_FROM_EQE = 0x0,
+ MLX5_CQ_PERIOD_MODE_START_FROM_CQE = 0x1,
+ MLX5_CQ_PERIOD_NUM_MODES
+};
+
+/* Adaptive moderation logic */
+enum {
+ MLX5E_AM_START_MEASURE,
+ MLX5E_AM_MEASURE_IN_PROGRESS,
+ MLX5E_AM_APPLY_NEW_PROFILE,
+};
+
+enum {
+ MLX5E_AM_PARKING_ON_TOP,
+ MLX5E_AM_PARKING_TIRED,
+ MLX5E_AM_GOING_RIGHT,
+ MLX5E_AM_GOING_LEFT,
+};
+
+enum {
+ MLX5E_AM_STATS_WORSE,
+ MLX5E_AM_STATS_SAME,
+ MLX5E_AM_STATS_BETTER,
+};
+
+enum {
+ MLX5E_AM_STEPPED,
+ MLX5E_AM_TOO_TIRED,
+ MLX5E_AM_ON_EDGE,
+};
+
+void mlx5e_rx_am(struct mlx5e_rx_am *am,
+ u16 event_ctr,
+ u64 packets,
+ u64 bytes);
+void mlx5e_rx_am_work(struct work_struct *work);
+struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode);
+struct mlx5e_cq_moder mlx5e_am_get_profile(u8 cq_period_mode, int ix);
+
+#endif /* MLX5_AM_H */
--
2.7.4
^ permalink raw reply related
* [net-next 06/10] net/mlx5e: change Mellanox references in DIM code
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
Change all mlx5_am* and MLX_AM* references to net_dim and NET_DIM,
respectively, in code that handles dynamic interrupt moderation. Also
change all references from 'am' to 'dim' when used as local variables.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Acked-by: Tal Gilboa <talgi@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en.h | 10 +-
drivers/net/ethernet/mellanox/mlx5/core/en_dim.c | 14 +-
drivers/net/ethernet/mellanox/mlx5/core/en_dim.h | 20 +-
.../net/ethernet/mellanox/mlx5/core/en_ethtool.c | 12 +-
drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 32 +--
drivers/net/ethernet/mellanox/mlx5/core/en_rep.c | 4 +-
drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 2 +-
drivers/net/ethernet/mellanox/mlx5/core/net_dim.c | 284 ++++++++++-----------
drivers/net/ethernet/mellanox/mlx5/core/net_dim.h | 63 +++--
9 files changed, 219 insertions(+), 222 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 2ccedf6..da2d5e7 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -50,6 +50,7 @@
#include "mlx5_core.h"
#include "en_stats.h"
#include "en_dim.h"
+#include "net_dim.h"
#define MLX5_SET_CFG(p, f, v) MLX5_SET(create_flow_group_in, p, f, v)
@@ -237,8 +238,8 @@ struct mlx5e_params {
u16 num_channels;
u8 num_tc;
bool rx_cqe_compress_def;
- struct mlx5e_cq_moder rx_cq_moderation;
- struct mlx5e_cq_moder tx_cq_moderation;
+ struct net_dim_cq_moder rx_cq_moderation;
+ struct net_dim_cq_moder tx_cq_moderation;
bool lro_en;
u32 lro_wqe_sz;
u16 tx_max_inline;
@@ -248,7 +249,7 @@ struct mlx5e_params {
u32 indirection_rqt[MLX5E_INDIR_RQT_SIZE];
bool vlan_strip_disable;
bool scatter_fcs_en;
- bool rx_am_enabled;
+ bool rx_dim_enabled;
u32 lro_timeout;
u32 pflags;
struct bpf_prog *xdp_prog;
@@ -527,7 +528,7 @@ struct mlx5e_rq {
unsigned long state;
int ix;
- struct mlx5e_rx_am am; /* Adaptive Moderation */
+ struct net_dim dim; /* Dynamic Interrupt Moderation */
/* XDP */
struct bpf_prog *xdp_prog;
@@ -1075,4 +1076,5 @@ void mlx5e_build_nic_params(struct mlx5_core_dev *mdev,
struct mlx5e_params *params,
u16 max_channels);
u8 mlx5e_params_calculate_tx_min_inline(struct mlx5_core_dev *mdev);
+void mlx5e_rx_dim_work(struct work_struct *work);
#endif /* __MLX5_EN_H__ */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
index b9b434b..f620325 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
@@ -32,17 +32,17 @@
#include "en.h"
-void mlx5e_rx_am_work(struct work_struct *work)
+void mlx5e_rx_dim_work(struct work_struct *work)
{
- struct mlx5e_rx_am *am = container_of(work, struct mlx5e_rx_am,
- work);
- struct mlx5e_rq *rq = container_of(am, struct mlx5e_rq, am);
- struct mlx5e_cq_moder cur_profile = mlx5e_am_get_profile(am->mode,
- am->profile_ix);
+ struct net_dim *dim = container_of(work, struct net_dim,
+ work);
+ struct mlx5e_rq *rq = container_of(dim, struct mlx5e_rq, dim);
+ struct net_dim_cq_moder cur_profile = net_dim_get_profile(dim->mode,
+ dim->profile_ix);
mlx5_core_modify_cq_moderation(rq->mdev, &rq->cq.mcq,
cur_profile.usec, cur_profile.pkts);
- am->state = MLX5E_AM_START_MEASURE;
+ dim->state = NET_DIM_START_MEASURE;
}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
index 5ce8e54..21219de 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
@@ -40,23 +40,23 @@ struct mlx5e_cq_moder {
u8 cq_period_mode;
};
-struct mlx5e_rx_am_sample {
+struct mlx5e_rx_dim_sample {
ktime_t time;
u32 pkt_ctr;
u32 byte_ctr;
u16 event_ctr;
};
-struct mlx5e_rx_am_stats {
+struct mlx5e_rx_dim_stats {
int ppms; /* packets per msec */
int bpms; /* bytes per msec */
int epms; /* events per msec */
};
-struct mlx5e_rx_am { /* Adaptive Moderation */
+struct mlx5e_rx_dim { /* Adaptive Moderation */
u8 state;
- struct mlx5e_rx_am_stats prev_stats;
- struct mlx5e_rx_am_sample start_sample;
+ struct mlx5e_rx_dim_stats prev_stats;
+ struct mlx5e_rx_dim_sample start_sample;
struct work_struct work;
u8 profile_ix;
u8 mode;
@@ -98,12 +98,8 @@ enum {
MLX5E_AM_ON_EDGE,
};
-void mlx5e_rx_am(struct mlx5e_rx_am *am,
- u16 event_ctr,
- u64 packets,
- u64 bytes);
-void mlx5e_rx_am_work(struct work_struct *work);
-struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode);
-struct mlx5e_cq_moder mlx5e_am_get_profile(u8 cq_period_mode, int ix);
+void mlx5e_rx_dim_work(struct work_struct *work);
+struct mlx5e_cq_moder mlx5e_dim_get_def_profile(u8 rx_cq_period_mode);
+struct mlx5e_cq_moder mlx5e_dim_get_profile(u8 cq_period_mode, int ix);
#endif /* MLX5_AM_H */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
index 8f05efa..772d276 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
@@ -480,7 +480,7 @@ int mlx5e_ethtool_get_coalesce(struct mlx5e_priv *priv,
coal->rx_max_coalesced_frames = priv->channels.params.rx_cq_moderation.pkts;
coal->tx_coalesce_usecs = priv->channels.params.tx_cq_moderation.usec;
coal->tx_max_coalesced_frames = priv->channels.params.tx_cq_moderation.pkts;
- coal->use_adaptive_rx_coalesce = priv->channels.params.rx_am_enabled;
+ coal->use_adaptive_rx_coalesce = priv->channels.params.rx_dim_enabled;
return 0;
}
@@ -534,7 +534,7 @@ int mlx5e_ethtool_set_coalesce(struct mlx5e_priv *priv,
new_channels.params.tx_cq_moderation.pkts = coal->tx_max_coalesced_frames;
new_channels.params.rx_cq_moderation.usec = coal->rx_coalesce_usecs;
new_channels.params.rx_cq_moderation.pkts = coal->rx_max_coalesced_frames;
- new_channels.params.rx_am_enabled = !!coal->use_adaptive_rx_coalesce;
+ new_channels.params.rx_dim_enabled = !!coal->use_adaptive_rx_coalesce;
if (!test_bit(MLX5E_STATE_OPENED, &priv->state)) {
priv->channels.params = new_channels.params;
@@ -542,7 +542,7 @@ int mlx5e_ethtool_set_coalesce(struct mlx5e_priv *priv,
}
/* we are opened */
- reset = !!coal->use_adaptive_rx_coalesce != priv->channels.params.rx_am_enabled;
+ reset = !!coal->use_adaptive_rx_coalesce != priv->channels.params.rx_dim_enabled;
if (!reset) {
mlx5e_set_priv_channels_coalesce(priv, coal);
priv->channels.params = new_channels.params;
@@ -1465,14 +1465,14 @@ static int set_pflag_cqe_based_moder(struct net_device *netdev, bool enable,
int err = 0;
cq_period_mode = enable ?
- MLX5_CQ_PERIOD_MODE_START_FROM_CQE :
- MLX5_CQ_PERIOD_MODE_START_FROM_EQE;
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE :
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE;
current_cq_period_mode = is_rx_cq ?
priv->channels.params.rx_cq_moderation.cq_period_mode :
priv->channels.params.tx_cq_moderation.cq_period_mode;
mode_changed = cq_period_mode != current_cq_period_mode;
- if (cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE &&
+ if (cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE &&
!MLX5_CAP_GEN(mdev, cq_period_start_from_cqe))
return -EOPNOTSUPP;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 3aa1c90..edd4077 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -674,8 +674,8 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
wqe->data.lkey = rq->mkey_be;
}
- INIT_WORK(&rq->am.work, mlx5e_rx_am_work);
- rq->am.mode = params->rx_cq_moderation.cq_period_mode;
+ INIT_WORK(&rq->dim.work, mlx5e_rx_dim_work);
+ rq->dim.mode = params->rx_cq_moderation.cq_period_mode;
rq->page_cache.head = 0;
rq->page_cache.tail = 0;
@@ -919,7 +919,7 @@ static int mlx5e_open_rq(struct mlx5e_channel *c,
if (err)
goto err_destroy_rq;
- if (params->rx_am_enabled)
+ if (params->rx_dim_enabled)
c->rq.state |= BIT(MLX5E_RQ_STATE_AM);
return 0;
@@ -952,7 +952,7 @@ static void mlx5e_deactivate_rq(struct mlx5e_rq *rq)
static void mlx5e_close_rq(struct mlx5e_rq *rq)
{
- cancel_work_sync(&rq->am.work);
+ cancel_work_sync(&rq->dim.work);
mlx5e_destroy_rq(rq);
mlx5e_free_rx_descs(rq);
mlx5e_free_rq(rq);
@@ -1565,7 +1565,7 @@ static void mlx5e_destroy_cq(struct mlx5e_cq *cq)
}
static int mlx5e_open_cq(struct mlx5e_channel *c,
- struct mlx5e_cq_moder moder,
+ struct net_dim_cq_moder moder,
struct mlx5e_cq_param *param,
struct mlx5e_cq *cq)
{
@@ -1747,7 +1747,7 @@ static int mlx5e_open_channel(struct mlx5e_priv *priv, int ix,
struct mlx5e_channel_param *cparam,
struct mlx5e_channel **cp)
{
- struct mlx5e_cq_moder icocq_moder = {0, 0};
+ struct net_dim_cq_moder icocq_moder = {0, 0};
struct net_device *netdev = priv->netdev;
int cpu = mlx5e_get_cpu(priv, ix);
struct mlx5e_channel *c;
@@ -1999,7 +1999,7 @@ static void mlx5e_build_ico_cq_param(struct mlx5e_priv *priv,
mlx5e_build_common_cq_param(priv, param);
- param->cq_period_mode = MLX5_CQ_PERIOD_MODE_START_FROM_EQE;
+ param->cq_period_mode = NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE;
}
static void mlx5e_build_icosq_param(struct mlx5e_priv *priv,
@@ -4016,13 +4016,13 @@ void mlx5e_set_tx_cq_mode_params(struct mlx5e_params *params, u8 cq_period_mode)
params->tx_cq_moderation.usec =
MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_USEC;
- if (cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE)
+ if (cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE)
params->tx_cq_moderation.usec =
MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_USEC_FROM_CQE;
MLX5E_SET_PFLAG(params, MLX5E_PFLAG_TX_CQE_BASED_MODER,
params->tx_cq_moderation.cq_period_mode ==
- MLX5_CQ_PERIOD_MODE_START_FROM_CQE);
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE);
}
void mlx5e_set_rx_cq_mode_params(struct mlx5e_params *params, u8 cq_period_mode)
@@ -4034,17 +4034,17 @@ void mlx5e_set_rx_cq_mode_params(struct mlx5e_params *params, u8 cq_period_mode)
params->rx_cq_moderation.usec =
MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_USEC;
- if (cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE)
+ if (cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE)
params->rx_cq_moderation.usec =
MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_USEC_FROM_CQE;
- if (params->rx_am_enabled)
+ if (params->rx_dim_enabled)
params->rx_cq_moderation =
- mlx5e_am_get_def_profile(cq_period_mode);
+ net_dim_get_def_profile(cq_period_mode);
MLX5E_SET_PFLAG(params, MLX5E_PFLAG_RX_CQE_BASED_MODER,
params->rx_cq_moderation.cq_period_mode ==
- MLX5_CQ_PERIOD_MODE_START_FROM_CQE);
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE);
}
u32 mlx5e_choose_lro_timeout(struct mlx5_core_dev *mdev, u32 wanted_timeout)
@@ -4100,9 +4100,9 @@ void mlx5e_build_nic_params(struct mlx5_core_dev *mdev,
/* CQ moderation params */
cq_period_mode = MLX5_CAP_GEN(mdev, cq_period_start_from_cqe) ?
- MLX5_CQ_PERIOD_MODE_START_FROM_CQE :
- MLX5_CQ_PERIOD_MODE_START_FROM_EQE;
- params->rx_am_enabled = MLX5_CAP_GEN(mdev, cq_moderation);
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE :
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE;
+ params->rx_dim_enabled = MLX5_CAP_GEN(mdev, cq_moderation);
mlx5e_set_rx_cq_mode_params(params, cq_period_mode);
mlx5e_set_tx_cq_mode_params(params, cq_period_mode);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
index c6a77f8..ccb038f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
@@ -877,8 +877,8 @@ static void mlx5e_build_rep_params(struct mlx5_core_dev *mdev,
struct mlx5e_params *params)
{
u8 cq_period_mode = MLX5_CAP_GEN(mdev, cq_period_start_from_cqe) ?
- MLX5_CQ_PERIOD_MODE_START_FROM_CQE :
- MLX5_CQ_PERIOD_MODE_START_FROM_EQE;
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE :
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE;
params->log_sq_size = MLX5E_PARAMS_MINIMUM_LOG_SQ_SIZE;
params->rq_wq_type = MLX5_WQ_TYPE_LINKED_LIST;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
index 1849169..dae77a9 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
@@ -79,7 +79,7 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget)
mlx5e_cq_arm(&c->sq[i].cq);
if (MLX5E_TEST_BIT(c->rq.state, MLX5E_RQ_STATE_AM))
- mlx5e_rx_am(&c->rq.am,
+ net_dim(&c->rq.dim,
c->rq.cq.event_ctr,
c->rq.stats.packets,
c->rq.stats.bytes);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
index ca05f4e..00b9ae3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
@@ -33,22 +33,22 @@
#include "en.h"
-#define MLX5E_PARAMS_AM_NUM_PROFILES 5
+#define NET_DIM_PARAMS_NUM_PROFILES 5
/* Adaptive moderation profiles */
-#define MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
-#define MLX5E_RX_AM_DEF_PROFILE_CQE 1
-#define MLX5E_RX_AM_DEF_PROFILE_EQE 1
-
-/* All profiles sizes must be MLX5E_PARAMS_AM_NUM_PROFILES */
-#define MLX5_AM_EQE_PROFILES { \
- {1, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {8, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {64, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {128, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {256, MLX5E_AM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
+#define NET_DIM_DEF_PROFILE_CQE 1
+#define NET_DIM_DEF_PROFILE_EQE 1
+
+/* All profiles sizes must be NET_PARAMS_DIM_NUM_PROFILES */
+#define NET_DIM_EQE_PROFILES { \
+ {1, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {8, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {64, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
}
-#define MLX5_AM_CQE_PROFILES { \
+#define NET_DIM_CQE_PROFILES { \
{2, 256}, \
{8, 128}, \
{16, 64}, \
@@ -56,193 +56,193 @@
{64, 64} \
}
-static const struct mlx5e_cq_moder
-profile[MLX5_CQ_PERIOD_NUM_MODES][MLX5E_PARAMS_AM_NUM_PROFILES] = {
- MLX5_AM_EQE_PROFILES,
- MLX5_AM_CQE_PROFILES,
+static const struct net_dim_cq_moder
+profile[NET_DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
+ NET_DIM_EQE_PROFILES,
+ NET_DIM_CQE_PROFILES,
};
-struct mlx5e_cq_moder mlx5e_am_get_profile(u8 cq_period_mode, int ix)
+struct net_dim_cq_moder net_dim_get_profile(u8 cq_period_mode, int ix)
{
- struct mlx5e_cq_moder cq_moder;
+ struct net_dim_cq_moder cq_moder;
cq_moder = profile[cq_period_mode][ix];
cq_moder.cq_period_mode = cq_period_mode;
return cq_moder;
}
-struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode)
+struct net_dim_cq_moder net_dim_get_def_profile(u8 rx_cq_period_mode)
{
int default_profile_ix;
- if (rx_cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE)
- default_profile_ix = MLX5E_RX_AM_DEF_PROFILE_CQE;
- else /* MLX5_CQ_PERIOD_MODE_START_FROM_EQE */
- default_profile_ix = MLX5E_RX_AM_DEF_PROFILE_EQE;
+ if (rx_cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE)
+ default_profile_ix = NET_DIM_DEF_PROFILE_CQE;
+ else /* NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE */
+ default_profile_ix = NET_DIM_DEF_PROFILE_EQE;
- return mlx5e_am_get_profile(rx_cq_period_mode, default_profile_ix);
+ return net_dim_get_profile(rx_cq_period_mode, default_profile_ix);
}
-static bool mlx5e_am_on_top(struct mlx5e_rx_am *am)
+static bool net_dim_on_top(struct net_dim *dim)
{
- switch (am->tune_state) {
- case MLX5E_AM_PARKING_ON_TOP:
- case MLX5E_AM_PARKING_TIRED:
+ switch (dim->tune_state) {
+ case NET_DIM_PARKING_ON_TOP:
+ case NET_DIM_PARKING_TIRED:
return true;
- case MLX5E_AM_GOING_RIGHT:
- return (am->steps_left > 1) && (am->steps_right == 1);
- default: /* MLX5E_AM_GOING_LEFT */
- return (am->steps_right > 1) && (am->steps_left == 1);
+ case NET_DIM_GOING_RIGHT:
+ return (dim->steps_left > 1) && (dim->steps_right == 1);
+ default: /* NET_DIM_GOING_LEFT */
+ return (dim->steps_right > 1) && (dim->steps_left == 1);
}
}
-static void mlx5e_am_turn(struct mlx5e_rx_am *am)
+static void net_dim_turn(struct net_dim *dim)
{
- switch (am->tune_state) {
- case MLX5E_AM_PARKING_ON_TOP:
- case MLX5E_AM_PARKING_TIRED:
+ switch (dim->tune_state) {
+ case NET_DIM_PARKING_ON_TOP:
+ case NET_DIM_PARKING_TIRED:
break;
- case MLX5E_AM_GOING_RIGHT:
- am->tune_state = MLX5E_AM_GOING_LEFT;
- am->steps_left = 0;
+ case NET_DIM_GOING_RIGHT:
+ dim->tune_state = NET_DIM_GOING_LEFT;
+ dim->steps_left = 0;
break;
- case MLX5E_AM_GOING_LEFT:
- am->tune_state = MLX5E_AM_GOING_RIGHT;
- am->steps_right = 0;
+ case NET_DIM_GOING_LEFT:
+ dim->tune_state = NET_DIM_GOING_RIGHT;
+ dim->steps_right = 0;
break;
}
}
-static int mlx5e_am_step(struct mlx5e_rx_am *am)
+static int net_dim_step(struct net_dim *dim)
{
- if (am->tired == (MLX5E_PARAMS_AM_NUM_PROFILES * 2))
- return MLX5E_AM_TOO_TIRED;
+ if (dim->tired == (NET_DIM_PARAMS_NUM_PROFILES * 2))
+ return NET_DIM_TOO_TIRED;
- switch (am->tune_state) {
- case MLX5E_AM_PARKING_ON_TOP:
- case MLX5E_AM_PARKING_TIRED:
+ switch (dim->tune_state) {
+ case NET_DIM_PARKING_ON_TOP:
+ case NET_DIM_PARKING_TIRED:
break;
- case MLX5E_AM_GOING_RIGHT:
- if (am->profile_ix == (MLX5E_PARAMS_AM_NUM_PROFILES - 1))
- return MLX5E_AM_ON_EDGE;
- am->profile_ix++;
- am->steps_right++;
+ case NET_DIM_GOING_RIGHT:
+ if (dim->profile_ix == (NET_DIM_PARAMS_NUM_PROFILES - 1))
+ return NET_DIM_ON_EDGE;
+ dim->profile_ix++;
+ dim->steps_right++;
break;
- case MLX5E_AM_GOING_LEFT:
- if (am->profile_ix == 0)
- return MLX5E_AM_ON_EDGE;
- am->profile_ix--;
- am->steps_left++;
+ case NET_DIM_GOING_LEFT:
+ if (dim->profile_ix == 0)
+ return NET_DIM_ON_EDGE;
+ dim->profile_ix--;
+ dim->steps_left++;
break;
}
- am->tired++;
- return MLX5E_AM_STEPPED;
+ dim->tired++;
+ return NET_DIM_STEPPED;
}
-static void mlx5e_am_park_on_top(struct mlx5e_rx_am *am)
+static void net_dim_park_on_top(struct net_dim *dim)
{
- am->steps_right = 0;
- am->steps_left = 0;
- am->tired = 0;
- am->tune_state = MLX5E_AM_PARKING_ON_TOP;
+ dim->steps_right = 0;
+ dim->steps_left = 0;
+ dim->tired = 0;
+ dim->tune_state = NET_DIM_PARKING_ON_TOP;
}
-static void mlx5e_am_park_tired(struct mlx5e_rx_am *am)
+static void net_dim_park_tired(struct net_dim *dim)
{
- am->steps_right = 0;
- am->steps_left = 0;
- am->tune_state = MLX5E_AM_PARKING_TIRED;
+ dim->steps_right = 0;
+ dim->steps_left = 0;
+ dim->tune_state = NET_DIM_PARKING_TIRED;
}
-static void mlx5e_am_exit_parking(struct mlx5e_rx_am *am)
+static void net_dim_exit_parking(struct net_dim *dim)
{
- am->tune_state = am->profile_ix ? MLX5E_AM_GOING_LEFT :
- MLX5E_AM_GOING_RIGHT;
- mlx5e_am_step(am);
+ dim->tune_state = dim->profile_ix ? NET_DIM_GOING_LEFT :
+ NET_DIM_GOING_RIGHT;
+ net_dim_step(dim);
}
#define IS_SIGNIFICANT_DIFF(val, ref) \
(((100 * abs((val) - (ref))) / (ref)) > 10) /* more than 10% difference */
-static int mlx5e_am_stats_compare(struct mlx5e_rx_am_stats *curr,
- struct mlx5e_rx_am_stats *prev)
+static int net_dim_stats_compare(struct net_dim_stats *curr,
+ struct net_dim_stats *prev)
{
if (!prev->bpms)
- return curr->bpms ? MLX5E_AM_STATS_BETTER :
- MLX5E_AM_STATS_SAME;
+ return curr->bpms ? NET_DIM_STATS_BETTER :
+ NET_DIM_STATS_SAME;
if (IS_SIGNIFICANT_DIFF(curr->bpms, prev->bpms))
- return (curr->bpms > prev->bpms) ? MLX5E_AM_STATS_BETTER :
- MLX5E_AM_STATS_WORSE;
+ return (curr->bpms > prev->bpms) ? NET_DIM_STATS_BETTER :
+ NET_DIM_STATS_WORSE;
if (IS_SIGNIFICANT_DIFF(curr->ppms, prev->ppms))
- return (curr->ppms > prev->ppms) ? MLX5E_AM_STATS_BETTER :
- MLX5E_AM_STATS_WORSE;
+ return (curr->ppms > prev->ppms) ? NET_DIM_STATS_BETTER :
+ NET_DIM_STATS_WORSE;
if (IS_SIGNIFICANT_DIFF(curr->epms, prev->epms))
- return (curr->epms < prev->epms) ? MLX5E_AM_STATS_BETTER :
- MLX5E_AM_STATS_WORSE;
+ return (curr->epms < prev->epms) ? NET_DIM_STATS_BETTER :
+ NET_DIM_STATS_WORSE;
- return MLX5E_AM_STATS_SAME;
+ return NET_DIM_STATS_SAME;
}
-static bool mlx5e_am_decision(struct mlx5e_rx_am_stats *curr_stats,
- struct mlx5e_rx_am *am)
+static bool net_dim_decision(struct net_dim_stats *curr_stats,
+ struct net_dim *dim)
{
- int prev_state = am->tune_state;
- int prev_ix = am->profile_ix;
+ int prev_state = dim->tune_state;
+ int prev_ix = dim->profile_ix;
int stats_res;
int step_res;
- switch (am->tune_state) {
- case MLX5E_AM_PARKING_ON_TOP:
- stats_res = mlx5e_am_stats_compare(curr_stats, &am->prev_stats);
- if (stats_res != MLX5E_AM_STATS_SAME)
- mlx5e_am_exit_parking(am);
+ switch (dim->tune_state) {
+ case NET_DIM_PARKING_ON_TOP:
+ stats_res = net_dim_stats_compare(curr_stats, &dim->prev_stats);
+ if (stats_res != NET_DIM_STATS_SAME)
+ net_dim_exit_parking(dim);
break;
- case MLX5E_AM_PARKING_TIRED:
- am->tired--;
- if (!am->tired)
- mlx5e_am_exit_parking(am);
+ case NET_DIM_PARKING_TIRED:
+ dim->tired--;
+ if (!dim->tired)
+ net_dim_exit_parking(dim);
break;
- case MLX5E_AM_GOING_RIGHT:
- case MLX5E_AM_GOING_LEFT:
- stats_res = mlx5e_am_stats_compare(curr_stats, &am->prev_stats);
- if (stats_res != MLX5E_AM_STATS_BETTER)
- mlx5e_am_turn(am);
+ case NET_DIM_GOING_RIGHT:
+ case NET_DIM_GOING_LEFT:
+ stats_res = net_dim_stats_compare(curr_stats, &dim->prev_stats);
+ if (stats_res != NET_DIM_STATS_BETTER)
+ net_dim_turn(dim);
- if (mlx5e_am_on_top(am)) {
- mlx5e_am_park_on_top(am);
+ if (net_dim_on_top(dim)) {
+ net_dim_park_on_top(dim);
break;
}
- step_res = mlx5e_am_step(am);
+ step_res = net_dim_step(dim);
switch (step_res) {
- case MLX5E_AM_ON_EDGE:
- mlx5e_am_park_on_top(am);
+ case NET_DIM_ON_EDGE:
+ net_dim_park_on_top(dim);
break;
- case MLX5E_AM_TOO_TIRED:
- mlx5e_am_park_tired(am);
+ case NET_DIM_TOO_TIRED:
+ net_dim_park_tired(dim);
break;
}
break;
}
- if ((prev_state != MLX5E_AM_PARKING_ON_TOP) ||
- (am->tune_state != MLX5E_AM_PARKING_ON_TOP))
- am->prev_stats = *curr_stats;
+ if ((prev_state != NET_DIM_PARKING_ON_TOP) ||
+ (dim->tune_state != NET_DIM_PARKING_ON_TOP))
+ dim->prev_stats = *curr_stats;
- return am->profile_ix != prev_ix;
+ return dim->profile_ix != prev_ix;
}
-static void mlx5e_am_sample(u16 event_ctr,
- u64 packets,
- u64 bytes,
- struct mlx5e_rx_am_sample *s)
+static void net_dim_sample(u16 event_ctr,
+ u64 packets,
+ u64 bytes,
+ struct net_dim_sample *s)
{
s->time = ktime_get();
s->pkt_ctr = packets;
@@ -250,13 +250,13 @@ static void mlx5e_am_sample(u16 event_ctr,
s->event_ctr = event_ctr;
}
-#define MLX5E_AM_NEVENTS 64
+#define NET_DIM_NEVENTS 64
#define BITS_PER_TYPE(type) (sizeof(type) * BITS_PER_BYTE)
#define BIT_GAP(bits, end, start) ((((end) - (start)) + BIT_ULL(bits)) & (BIT_ULL(bits) - 1))
-static void mlx5e_am_calc_stats(struct mlx5e_rx_am_sample *start,
- struct mlx5e_rx_am_sample *end,
- struct mlx5e_rx_am_stats *curr_stats)
+static void net_dim_calc_stats(struct net_dim_sample *start,
+ struct net_dim_sample *end,
+ struct net_dim_stats *curr_stats)
{
/* u32 holds up to 71 minutes, should be enough */
u32 delta_us = ktime_us_delta(end->time, start->time);
@@ -269,39 +269,39 @@ static void mlx5e_am_calc_stats(struct mlx5e_rx_am_sample *start,
curr_stats->ppms = DIV_ROUND_UP(npkts * USEC_PER_MSEC, delta_us);
curr_stats->bpms = DIV_ROUND_UP(nbytes * USEC_PER_MSEC, delta_us);
- curr_stats->epms = DIV_ROUND_UP(MLX5E_AM_NEVENTS * USEC_PER_MSEC,
+ curr_stats->epms = DIV_ROUND_UP(NET_DIM_NEVENTS * USEC_PER_MSEC,
delta_us);
}
-void mlx5e_rx_am(struct mlx5e_rx_am *am,
- u16 event_ctr,
- u64 packets,
- u64 bytes)
+void net_dim(struct net_dim *dim,
+ u16 event_ctr,
+ u64 packets,
+ u64 bytes)
{
- struct mlx5e_rx_am_sample end_sample;
- struct mlx5e_rx_am_stats curr_stats;
+ struct net_dim_sample end_sample;
+ struct net_dim_stats curr_stats;
u16 nevents;
- switch (am->state) {
- case MLX5E_AM_MEASURE_IN_PROGRESS:
+ switch (dim->state) {
+ case NET_DIM_MEASURE_IN_PROGRESS:
nevents = BIT_GAP(BITS_PER_TYPE(u16), event_ctr,
- am->start_sample.event_ctr);
- if (nevents < MLX5E_AM_NEVENTS)
+ dim->start_sample.event_ctr);
+ if (nevents < NET_DIM_NEVENTS)
break;
- mlx5e_am_sample(event_ctr, packets, bytes, &end_sample);
- mlx5e_am_calc_stats(&am->start_sample, &end_sample,
+ net_dim_sample(event_ctr, packets, bytes, &end_sample);
+ net_dim_calc_stats(&dim->start_sample, &end_sample,
&curr_stats);
- if (mlx5e_am_decision(&curr_stats, am)) {
- am->state = MLX5E_AM_APPLY_NEW_PROFILE;
- schedule_work(&am->work);
+ if (net_dim_decision(&curr_stats, dim)) {
+ dim->state = NET_DIM_APPLY_NEW_PROFILE;
+ schedule_work(&dim->work);
break;
}
/* fall through */
- case MLX5E_AM_START_MEASURE:
- mlx5e_am_sample(event_ctr, packets, bytes, &am->start_sample);
- am->state = MLX5E_AM_MEASURE_IN_PROGRESS;
+ case NET_DIM_START_MEASURE:
+ net_dim_sample(event_ctr, packets, bytes, &dim->start_sample);
+ dim->state = NET_DIM_MEASURE_IN_PROGRESS;
break;
- case MLX5E_AM_APPLY_NEW_PROFILE:
+ case NET_DIM_APPLY_NEW_PROFILE:
break;
}
}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/net_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.h
index 5ce8e54..a775c12 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/net_dim.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.h
@@ -31,32 +31,32 @@
* SOFTWARE.
*/
-#ifndef MLX5_AM_H
-#define MLX5_AM_H
+#ifndef NET_DIM_H
+#define NET_DIM_H
-struct mlx5e_cq_moder {
+struct net_dim_cq_moder {
u16 usec;
u16 pkts;
u8 cq_period_mode;
};
-struct mlx5e_rx_am_sample {
+struct net_dim_sample {
ktime_t time;
u32 pkt_ctr;
u32 byte_ctr;
u16 event_ctr;
};
-struct mlx5e_rx_am_stats {
+struct net_dim_stats {
int ppms; /* packets per msec */
int bpms; /* bytes per msec */
int epms; /* events per msec */
};
-struct mlx5e_rx_am { /* Adaptive Moderation */
+struct net_dim { /* Adaptive Moderation */
u8 state;
- struct mlx5e_rx_am_stats prev_stats;
- struct mlx5e_rx_am_sample start_sample;
+ struct net_dim_stats prev_stats;
+ struct net_dim_sample start_sample;
struct work_struct work;
u8 profile_ix;
u8 mode;
@@ -67,43 +67,42 @@ struct mlx5e_rx_am { /* Adaptive Moderation */
};
enum {
- MLX5_CQ_PERIOD_MODE_START_FROM_EQE = 0x0,
- MLX5_CQ_PERIOD_MODE_START_FROM_CQE = 0x1,
- MLX5_CQ_PERIOD_NUM_MODES
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE = 0x0,
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE = 0x1,
+ NET_DIM_CQ_PERIOD_NUM_MODES
};
/* Adaptive moderation logic */
enum {
- MLX5E_AM_START_MEASURE,
- MLX5E_AM_MEASURE_IN_PROGRESS,
- MLX5E_AM_APPLY_NEW_PROFILE,
+ NET_DIM_START_MEASURE,
+ NET_DIM_MEASURE_IN_PROGRESS,
+ NET_DIM_APPLY_NEW_PROFILE,
};
enum {
- MLX5E_AM_PARKING_ON_TOP,
- MLX5E_AM_PARKING_TIRED,
- MLX5E_AM_GOING_RIGHT,
- MLX5E_AM_GOING_LEFT,
+ NET_DIM_PARKING_ON_TOP,
+ NET_DIM_PARKING_TIRED,
+ NET_DIM_GOING_RIGHT,
+ NET_DIM_GOING_LEFT,
};
enum {
- MLX5E_AM_STATS_WORSE,
- MLX5E_AM_STATS_SAME,
- MLX5E_AM_STATS_BETTER,
+ NET_DIM_STATS_WORSE,
+ NET_DIM_STATS_SAME,
+ NET_DIM_STATS_BETTER,
};
enum {
- MLX5E_AM_STEPPED,
- MLX5E_AM_TOO_TIRED,
- MLX5E_AM_ON_EDGE,
+ NET_DIM_STEPPED,
+ NET_DIM_TOO_TIRED,
+ NET_DIM_ON_EDGE,
};
-void mlx5e_rx_am(struct mlx5e_rx_am *am,
- u16 event_ctr,
- u64 packets,
- u64 bytes);
-void mlx5e_rx_am_work(struct work_struct *work);
-struct mlx5e_cq_moder mlx5e_am_get_def_profile(u8 rx_cq_period_mode);
-struct mlx5e_cq_moder mlx5e_am_get_profile(u8 cq_period_mode, int ix);
+void net_dim(struct net_dim *dim,
+ u16 event_ctr,
+ u64 packets,
+ u64 bytes);
+struct net_dim_cq_moder net_dim_get_def_profile(u8 rx_cq_period_mode);
+struct net_dim_cq_moder net_dim_get_profile(u8 cq_period_mode, int ix);
-#endif /* MLX5_AM_H */
+#endif /* NET_DIM_H */
--
2.7.4
^ permalink raw reply related
* [net-next 07/10] net: move dynamic interrupt coalescing code to include/linux
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
This move allows drivers to add private structure elements to track the
number of packets, bytes, and interrupts events per ring. A driver
also defines a workqueue handler to act on this collected data once per
poll and modify the coalescing parameters per ring.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Acked-by: Tal Gilboa <talgi@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/Makefile | 2 +-
drivers/net/ethernet/mellanox/mlx5/core/en.h | 3 +-
drivers/net/ethernet/mellanox/mlx5/core/en_dim.c | 1 +
drivers/net/ethernet/mellanox/mlx5/core/en_dim.h | 105 ------
drivers/net/ethernet/mellanox/mlx5/core/net_dim.c | 307 ------------------
include/linux/net_dim.h | 376 ++++++++++++++++++++++
6 files changed, 379 insertions(+), 415 deletions(-)
delete mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
delete mode 100644 drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
create mode 100644 include/linux/net_dim.h
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/Makefile b/drivers/net/ethernet/mellanox/mlx5/core/Makefile
index b46b6de2..c805769 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/Makefile
+++ b/drivers/net/ethernet/mellanox/mlx5/core/Makefile
@@ -15,7 +15,7 @@ mlx5_core-$(CONFIG_MLX5_FPGA) += fpga/cmd.o fpga/core.o fpga/conn.o fpga/sdk.o \
mlx5_core-$(CONFIG_MLX5_CORE_EN) += en_main.o en_common.o en_fs.o en_ethtool.o \
en_tx.o en_rx.o en_dim.o en_txrx.o en_stats.o vxlan.o \
- en_arfs.o en_fs_ethtool.o en_selftest.o net_dim.o
+ en_arfs.o en_fs_ethtool.o en_selftest.o
mlx5_core-$(CONFIG_MLX5_MPFS) += lib/mpfs.o
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index da2d5e7..41e6783 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -46,11 +46,10 @@
#include <linux/mlx5/transobj.h>
#include <linux/rhashtable.h>
#include <net/switchdev.h>
+#include <linux/net_dim.h>
#include "wq.h"
#include "mlx5_core.h"
#include "en_stats.h"
-#include "en_dim.h"
-#include "net_dim.h"
#define MLX5_SET_CFG(p, f, v) MLX5_SET(create_flow_group_in, p, f, v)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
index f620325..2b89951 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
@@ -30,6 +30,7 @@
* SOFTWARE.
*/
+#include <linux/net_dim.h>
#include "en.h"
void mlx5e_rx_dim_work(struct work_struct *work)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
deleted file mode 100644
index 21219de..0000000
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.h
+++ /dev/null
@@ -1,105 +0,0 @@
-/*
- * Copyright (c) 2013-2015, Mellanox Technologies, Ltd. All rights reserved.
- * Copyright (c) 2017, Broadcom Limited
- *
- * This software is available to you under a choice of one of two
- * licenses. You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- * Redistribution and use in source and binary forms, with or
- * without modification, are permitted provided that the following
- * conditions are met:
- *
- * - Redistributions of source code must retain the above
- * copyright notice, this list of conditions and the following
- * disclaimer.
- *
- * - Redistributions in binary form must reproduce the above
- * copyright notice, this list of conditions and the following
- * disclaimer in the documentation and/or other materials
- * provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
-*/
-
-#ifndef MLX5_AM_H
-#define MLX5_AM_H
-
-struct mlx5e_cq_moder {
- u16 usec;
- u16 pkts;
- u8 cq_period_mode;
-};
-
-struct mlx5e_rx_dim_sample {
- ktime_t time;
- u32 pkt_ctr;
- u32 byte_ctr;
- u16 event_ctr;
-};
-
-struct mlx5e_rx_dim_stats {
- int ppms; /* packets per msec */
- int bpms; /* bytes per msec */
- int epms; /* events per msec */
-};
-
-struct mlx5e_rx_dim { /* Adaptive Moderation */
- u8 state;
- struct mlx5e_rx_dim_stats prev_stats;
- struct mlx5e_rx_dim_sample start_sample;
- struct work_struct work;
- u8 profile_ix;
- u8 mode;
- u8 tune_state;
- u8 steps_right;
- u8 steps_left;
- u8 tired;
-};
-
-enum {
- MLX5_CQ_PERIOD_MODE_START_FROM_EQE = 0x0,
- MLX5_CQ_PERIOD_MODE_START_FROM_CQE = 0x1,
- MLX5_CQ_PERIOD_NUM_MODES
-};
-
-/* Adaptive moderation logic */
-enum {
- MLX5E_AM_START_MEASURE,
- MLX5E_AM_MEASURE_IN_PROGRESS,
- MLX5E_AM_APPLY_NEW_PROFILE,
-};
-
-enum {
- MLX5E_AM_PARKING_ON_TOP,
- MLX5E_AM_PARKING_TIRED,
- MLX5E_AM_GOING_RIGHT,
- MLX5E_AM_GOING_LEFT,
-};
-
-enum {
- MLX5E_AM_STATS_WORSE,
- MLX5E_AM_STATS_SAME,
- MLX5E_AM_STATS_BETTER,
-};
-
-enum {
- MLX5E_AM_STEPPED,
- MLX5E_AM_TOO_TIRED,
- MLX5E_AM_ON_EDGE,
-};
-
-void mlx5e_rx_dim_work(struct work_struct *work);
-struct mlx5e_cq_moder mlx5e_dim_get_def_profile(u8 rx_cq_period_mode);
-struct mlx5e_cq_moder mlx5e_dim_get_profile(u8 cq_period_mode, int ix);
-
-#endif /* MLX5_AM_H */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c b/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
deleted file mode 100644
index 00b9ae3..0000000
--- a/drivers/net/ethernet/mellanox/mlx5/core/net_dim.c
+++ /dev/null
@@ -1,307 +0,0 @@
-/*
- * Copyright (c) 2016, Mellanox Technologies. All rights reserved.
- * Copyright (c) 2017, Broadcom Limiited. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses. You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- * Redistribution and use in source and binary forms, with or
- * without modification, are permitted provided that the following
- * conditions are met:
- *
- * - Redistributions of source code must retain the above
- * copyright notice, this list of conditions and the following
- * disclaimer.
- *
- * - Redistributions in binary form must reproduce the above
- * copyright notice, this list of conditions and the following
- * disclaimer in the documentation and/or other materials
- * provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-#include "en.h"
-
-#define NET_DIM_PARAMS_NUM_PROFILES 5
-/* Adaptive moderation profiles */
-#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
-#define NET_DIM_DEF_PROFILE_CQE 1
-#define NET_DIM_DEF_PROFILE_EQE 1
-
-/* All profiles sizes must be NET_PARAMS_DIM_NUM_PROFILES */
-#define NET_DIM_EQE_PROFILES { \
- {1, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {8, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {64, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
- {256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
-}
-
-#define NET_DIM_CQE_PROFILES { \
- {2, 256}, \
- {8, 128}, \
- {16, 64}, \
- {32, 64}, \
- {64, 64} \
-}
-
-static const struct net_dim_cq_moder
-profile[NET_DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
- NET_DIM_EQE_PROFILES,
- NET_DIM_CQE_PROFILES,
-};
-
-struct net_dim_cq_moder net_dim_get_profile(u8 cq_period_mode, int ix)
-{
- struct net_dim_cq_moder cq_moder;
-
- cq_moder = profile[cq_period_mode][ix];
- cq_moder.cq_period_mode = cq_period_mode;
- return cq_moder;
-}
-
-struct net_dim_cq_moder net_dim_get_def_profile(u8 rx_cq_period_mode)
-{
- int default_profile_ix;
-
- if (rx_cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE)
- default_profile_ix = NET_DIM_DEF_PROFILE_CQE;
- else /* NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE */
- default_profile_ix = NET_DIM_DEF_PROFILE_EQE;
-
- return net_dim_get_profile(rx_cq_period_mode, default_profile_ix);
-}
-
-static bool net_dim_on_top(struct net_dim *dim)
-{
- switch (dim->tune_state) {
- case NET_DIM_PARKING_ON_TOP:
- case NET_DIM_PARKING_TIRED:
- return true;
- case NET_DIM_GOING_RIGHT:
- return (dim->steps_left > 1) && (dim->steps_right == 1);
- default: /* NET_DIM_GOING_LEFT */
- return (dim->steps_right > 1) && (dim->steps_left == 1);
- }
-}
-
-static void net_dim_turn(struct net_dim *dim)
-{
- switch (dim->tune_state) {
- case NET_DIM_PARKING_ON_TOP:
- case NET_DIM_PARKING_TIRED:
- break;
- case NET_DIM_GOING_RIGHT:
- dim->tune_state = NET_DIM_GOING_LEFT;
- dim->steps_left = 0;
- break;
- case NET_DIM_GOING_LEFT:
- dim->tune_state = NET_DIM_GOING_RIGHT;
- dim->steps_right = 0;
- break;
- }
-}
-
-static int net_dim_step(struct net_dim *dim)
-{
- if (dim->tired == (NET_DIM_PARAMS_NUM_PROFILES * 2))
- return NET_DIM_TOO_TIRED;
-
- switch (dim->tune_state) {
- case NET_DIM_PARKING_ON_TOP:
- case NET_DIM_PARKING_TIRED:
- break;
- case NET_DIM_GOING_RIGHT:
- if (dim->profile_ix == (NET_DIM_PARAMS_NUM_PROFILES - 1))
- return NET_DIM_ON_EDGE;
- dim->profile_ix++;
- dim->steps_right++;
- break;
- case NET_DIM_GOING_LEFT:
- if (dim->profile_ix == 0)
- return NET_DIM_ON_EDGE;
- dim->profile_ix--;
- dim->steps_left++;
- break;
- }
-
- dim->tired++;
- return NET_DIM_STEPPED;
-}
-
-static void net_dim_park_on_top(struct net_dim *dim)
-{
- dim->steps_right = 0;
- dim->steps_left = 0;
- dim->tired = 0;
- dim->tune_state = NET_DIM_PARKING_ON_TOP;
-}
-
-static void net_dim_park_tired(struct net_dim *dim)
-{
- dim->steps_right = 0;
- dim->steps_left = 0;
- dim->tune_state = NET_DIM_PARKING_TIRED;
-}
-
-static void net_dim_exit_parking(struct net_dim *dim)
-{
- dim->tune_state = dim->profile_ix ? NET_DIM_GOING_LEFT :
- NET_DIM_GOING_RIGHT;
- net_dim_step(dim);
-}
-
-#define IS_SIGNIFICANT_DIFF(val, ref) \
- (((100 * abs((val) - (ref))) / (ref)) > 10) /* more than 10% difference */
-
-static int net_dim_stats_compare(struct net_dim_stats *curr,
- struct net_dim_stats *prev)
-{
- if (!prev->bpms)
- return curr->bpms ? NET_DIM_STATS_BETTER :
- NET_DIM_STATS_SAME;
-
- if (IS_SIGNIFICANT_DIFF(curr->bpms, prev->bpms))
- return (curr->bpms > prev->bpms) ? NET_DIM_STATS_BETTER :
- NET_DIM_STATS_WORSE;
-
- if (IS_SIGNIFICANT_DIFF(curr->ppms, prev->ppms))
- return (curr->ppms > prev->ppms) ? NET_DIM_STATS_BETTER :
- NET_DIM_STATS_WORSE;
-
- if (IS_SIGNIFICANT_DIFF(curr->epms, prev->epms))
- return (curr->epms < prev->epms) ? NET_DIM_STATS_BETTER :
- NET_DIM_STATS_WORSE;
-
- return NET_DIM_STATS_SAME;
-}
-
-static bool net_dim_decision(struct net_dim_stats *curr_stats,
- struct net_dim *dim)
-{
- int prev_state = dim->tune_state;
- int prev_ix = dim->profile_ix;
- int stats_res;
- int step_res;
-
- switch (dim->tune_state) {
- case NET_DIM_PARKING_ON_TOP:
- stats_res = net_dim_stats_compare(curr_stats, &dim->prev_stats);
- if (stats_res != NET_DIM_STATS_SAME)
- net_dim_exit_parking(dim);
- break;
-
- case NET_DIM_PARKING_TIRED:
- dim->tired--;
- if (!dim->tired)
- net_dim_exit_parking(dim);
- break;
-
- case NET_DIM_GOING_RIGHT:
- case NET_DIM_GOING_LEFT:
- stats_res = net_dim_stats_compare(curr_stats, &dim->prev_stats);
- if (stats_res != NET_DIM_STATS_BETTER)
- net_dim_turn(dim);
-
- if (net_dim_on_top(dim)) {
- net_dim_park_on_top(dim);
- break;
- }
-
- step_res = net_dim_step(dim);
- switch (step_res) {
- case NET_DIM_ON_EDGE:
- net_dim_park_on_top(dim);
- break;
- case NET_DIM_TOO_TIRED:
- net_dim_park_tired(dim);
- break;
- }
-
- break;
- }
-
- if ((prev_state != NET_DIM_PARKING_ON_TOP) ||
- (dim->tune_state != NET_DIM_PARKING_ON_TOP))
- dim->prev_stats = *curr_stats;
-
- return dim->profile_ix != prev_ix;
-}
-
-static void net_dim_sample(u16 event_ctr,
- u64 packets,
- u64 bytes,
- struct net_dim_sample *s)
-{
- s->time = ktime_get();
- s->pkt_ctr = packets;
- s->byte_ctr = bytes;
- s->event_ctr = event_ctr;
-}
-
-#define NET_DIM_NEVENTS 64
-#define BITS_PER_TYPE(type) (sizeof(type) * BITS_PER_BYTE)
-#define BIT_GAP(bits, end, start) ((((end) - (start)) + BIT_ULL(bits)) & (BIT_ULL(bits) - 1))
-
-static void net_dim_calc_stats(struct net_dim_sample *start,
- struct net_dim_sample *end,
- struct net_dim_stats *curr_stats)
-{
- /* u32 holds up to 71 minutes, should be enough */
- u32 delta_us = ktime_us_delta(end->time, start->time);
- u32 npkts = BIT_GAP(BITS_PER_TYPE(u32), end->pkt_ctr, start->pkt_ctr);
- u32 nbytes = BIT_GAP(BITS_PER_TYPE(u32), end->byte_ctr,
- start->byte_ctr);
-
- if (!delta_us)
- return;
-
- curr_stats->ppms = DIV_ROUND_UP(npkts * USEC_PER_MSEC, delta_us);
- curr_stats->bpms = DIV_ROUND_UP(nbytes * USEC_PER_MSEC, delta_us);
- curr_stats->epms = DIV_ROUND_UP(NET_DIM_NEVENTS * USEC_PER_MSEC,
- delta_us);
-}
-
-void net_dim(struct net_dim *dim,
- u16 event_ctr,
- u64 packets,
- u64 bytes)
-{
- struct net_dim_sample end_sample;
- struct net_dim_stats curr_stats;
- u16 nevents;
-
- switch (dim->state) {
- case NET_DIM_MEASURE_IN_PROGRESS:
- nevents = BIT_GAP(BITS_PER_TYPE(u16), event_ctr,
- dim->start_sample.event_ctr);
- if (nevents < NET_DIM_NEVENTS)
- break;
- net_dim_sample(event_ctr, packets, bytes, &end_sample);
- net_dim_calc_stats(&dim->start_sample, &end_sample,
- &curr_stats);
- if (net_dim_decision(&curr_stats, dim)) {
- dim->state = NET_DIM_APPLY_NEW_PROFILE;
- schedule_work(&dim->work);
- break;
- }
- /* fall through */
- case NET_DIM_START_MEASURE:
- net_dim_sample(event_ctr, packets, bytes, &dim->start_sample);
- dim->state = NET_DIM_MEASURE_IN_PROGRESS;
- break;
- case NET_DIM_APPLY_NEW_PROFILE:
- break;
- }
-}
diff --git a/include/linux/net_dim.h b/include/linux/net_dim.h
new file mode 100644
index 0000000..bb99073
--- /dev/null
+++ b/include/linux/net_dim.h
@@ -0,0 +1,376 @@
+/*
+ * Copyright (c) 2013-2015, Mellanox Technologies, Ltd. All rights reserved.
+ * Copyright (c) 2017, Broadcom Limited
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+*/
+
+#ifndef NET_DIM_H
+#define NET_DIM_H
+
+#include <linux/module.h>
+
+struct net_dim_cq_moder {
+ u16 usec;
+ u16 pkts;
+ u8 cq_period_mode;
+};
+
+struct net_dim_sample {
+ ktime_t time;
+ u32 pkt_ctr;
+ u32 byte_ctr;
+ u16 event_ctr;
+};
+
+struct net_dim_stats {
+ int ppms; /* packets per msec */
+ int bpms; /* bytes per msec */
+ int epms; /* events per msec */
+};
+
+struct net_dim { /* Adaptive Moderation */
+ u8 state;
+ struct net_dim_stats prev_stats;
+ struct net_dim_sample start_sample;
+ struct work_struct work;
+ u8 profile_ix;
+ u8 mode;
+ u8 tune_state;
+ u8 steps_right;
+ u8 steps_left;
+ u8 tired;
+};
+
+enum {
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE = 0x0,
+ NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE = 0x1,
+ NET_DIM_CQ_PERIOD_NUM_MODES
+};
+
+/* Adaptive moderation logic */
+enum {
+ NET_DIM_START_MEASURE,
+ NET_DIM_MEASURE_IN_PROGRESS,
+ NET_DIM_APPLY_NEW_PROFILE,
+};
+
+enum {
+ NET_DIM_PARKING_ON_TOP,
+ NET_DIM_PARKING_TIRED,
+ NET_DIM_GOING_RIGHT,
+ NET_DIM_GOING_LEFT,
+};
+
+enum {
+ NET_DIM_STATS_WORSE,
+ NET_DIM_STATS_SAME,
+ NET_DIM_STATS_BETTER,
+};
+
+enum {
+ NET_DIM_STEPPED,
+ NET_DIM_TOO_TIRED,
+ NET_DIM_ON_EDGE,
+};
+
+#define NET_DIM_PARAMS_NUM_PROFILES 5
+/* Adaptive moderation profiles */
+#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
+#define NET_DIM_DEF_PROFILE_CQE 1
+#define NET_DIM_DEF_PROFILE_EQE 1
+
+/* All profiles sizes must be NET_PARAMS_DIM_NUM_PROFILES */
+#define NET_DIM_EQE_PROFILES { \
+ {1, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {8, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {64, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+ {256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+}
+
+#define NET_DIM_CQE_PROFILES { \
+ {2, 256}, \
+ {8, 128}, \
+ {16, 64}, \
+ {32, 64}, \
+ {64, 64} \
+}
+
+static const struct net_dim_cq_moder
+profile[NET_DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
+ NET_DIM_EQE_PROFILES,
+ NET_DIM_CQE_PROFILES,
+};
+
+static inline struct net_dim_cq_moder net_dim_get_profile(u8 cq_period_mode, int ix)
+{
+ struct net_dim_cq_moder cq_moder;
+
+ cq_moder = profile[cq_period_mode][ix];
+ cq_moder.cq_period_mode = cq_period_mode;
+ return cq_moder;
+}
+
+static inline struct net_dim_cq_moder net_dim_get_def_profile(u8 rx_cq_period_mode)
+{
+ int default_profile_ix;
+
+ if (rx_cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE)
+ default_profile_ix = NET_DIM_DEF_PROFILE_CQE;
+ else /* NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE */
+ default_profile_ix = NET_DIM_DEF_PROFILE_EQE;
+
+ return net_dim_get_profile(rx_cq_period_mode, default_profile_ix);
+}
+
+static inline bool net_dim_on_top(struct net_dim *dim)
+{
+ switch (dim->tune_state) {
+ case NET_DIM_PARKING_ON_TOP:
+ case NET_DIM_PARKING_TIRED:
+ return true;
+ case NET_DIM_GOING_RIGHT:
+ return (dim->steps_left > 1) && (dim->steps_right == 1);
+ default: /* NET_DIM_GOING_LEFT */
+ return (dim->steps_right > 1) && (dim->steps_left == 1);
+ }
+}
+
+static inline void net_dim_turn(struct net_dim *dim)
+{
+ switch (dim->tune_state) {
+ case NET_DIM_PARKING_ON_TOP:
+ case NET_DIM_PARKING_TIRED:
+ break;
+ case NET_DIM_GOING_RIGHT:
+ dim->tune_state = NET_DIM_GOING_LEFT;
+ dim->steps_left = 0;
+ break;
+ case NET_DIM_GOING_LEFT:
+ dim->tune_state = NET_DIM_GOING_RIGHT;
+ dim->steps_right = 0;
+ break;
+ }
+}
+
+static inline int net_dim_step(struct net_dim *dim)
+{
+ if (dim->tired == (NET_DIM_PARAMS_NUM_PROFILES * 2))
+ return NET_DIM_TOO_TIRED;
+
+ switch (dim->tune_state) {
+ case NET_DIM_PARKING_ON_TOP:
+ case NET_DIM_PARKING_TIRED:
+ break;
+ case NET_DIM_GOING_RIGHT:
+ if (dim->profile_ix == (NET_DIM_PARAMS_NUM_PROFILES - 1))
+ return NET_DIM_ON_EDGE;
+ dim->profile_ix++;
+ dim->steps_right++;
+ break;
+ case NET_DIM_GOING_LEFT:
+ if (dim->profile_ix == 0)
+ return NET_DIM_ON_EDGE;
+ dim->profile_ix--;
+ dim->steps_left++;
+ break;
+ }
+
+ dim->tired++;
+ return NET_DIM_STEPPED;
+}
+
+static inline void net_dim_park_on_top(struct net_dim *dim)
+{
+ dim->steps_right = 0;
+ dim->steps_left = 0;
+ dim->tired = 0;
+ dim->tune_state = NET_DIM_PARKING_ON_TOP;
+}
+
+static inline void net_dim_park_tired(struct net_dim *dim)
+{
+ dim->steps_right = 0;
+ dim->steps_left = 0;
+ dim->tune_state = NET_DIM_PARKING_TIRED;
+}
+
+static inline void net_dim_exit_parking(struct net_dim *dim)
+{
+ dim->tune_state = dim->profile_ix ? NET_DIM_GOING_LEFT :
+ NET_DIM_GOING_RIGHT;
+ net_dim_step(dim);
+}
+
+#define IS_SIGNIFICANT_DIFF(val, ref) \
+ (((100 * abs((val) - (ref))) / (ref)) > 10) /* more than 10% difference */
+
+static inline int net_dim_stats_compare(struct net_dim_stats *curr,
+ struct net_dim_stats *prev)
+{
+ if (!prev->bpms)
+ return curr->bpms ? NET_DIM_STATS_BETTER :
+ NET_DIM_STATS_SAME;
+
+ if (IS_SIGNIFICANT_DIFF(curr->bpms, prev->bpms))
+ return (curr->bpms > prev->bpms) ? NET_DIM_STATS_BETTER :
+ NET_DIM_STATS_WORSE;
+
+ if (IS_SIGNIFICANT_DIFF(curr->ppms, prev->ppms))
+ return (curr->ppms > prev->ppms) ? NET_DIM_STATS_BETTER :
+ NET_DIM_STATS_WORSE;
+
+ if (IS_SIGNIFICANT_DIFF(curr->epms, prev->epms))
+ return (curr->epms < prev->epms) ? NET_DIM_STATS_BETTER :
+ NET_DIM_STATS_WORSE;
+
+ return NET_DIM_STATS_SAME;
+}
+
+static inline bool net_dim_decision(struct net_dim_stats *curr_stats,
+ struct net_dim *dim)
+{
+ int prev_state = dim->tune_state;
+ int prev_ix = dim->profile_ix;
+ int stats_res;
+ int step_res;
+
+ switch (dim->tune_state) {
+ case NET_DIM_PARKING_ON_TOP:
+ stats_res = net_dim_stats_compare(curr_stats, &dim->prev_stats);
+ if (stats_res != NET_DIM_STATS_SAME)
+ net_dim_exit_parking(dim);
+ break;
+
+ case NET_DIM_PARKING_TIRED:
+ dim->tired--;
+ if (!dim->tired)
+ net_dim_exit_parking(dim);
+ break;
+
+ case NET_DIM_GOING_RIGHT:
+ case NET_DIM_GOING_LEFT:
+ stats_res = net_dim_stats_compare(curr_stats, &dim->prev_stats);
+ if (stats_res != NET_DIM_STATS_BETTER)
+ net_dim_turn(dim);
+
+ if (net_dim_on_top(dim)) {
+ net_dim_park_on_top(dim);
+ break;
+ }
+
+ step_res = net_dim_step(dim);
+ switch (step_res) {
+ case NET_DIM_ON_EDGE:
+ net_dim_park_on_top(dim);
+ break;
+ case NET_DIM_TOO_TIRED:
+ net_dim_park_tired(dim);
+ break;
+ }
+
+ break;
+ }
+
+ if ((prev_state != NET_DIM_PARKING_ON_TOP) ||
+ (dim->tune_state != NET_DIM_PARKING_ON_TOP))
+ dim->prev_stats = *curr_stats;
+
+ return dim->profile_ix != prev_ix;
+}
+
+static inline void net_dim_sample(u16 event_ctr,
+ u64 packets,
+ u64 bytes,
+ struct net_dim_sample *s)
+{
+ s->time = ktime_get();
+ s->pkt_ctr = packets;
+ s->byte_ctr = bytes;
+ s->event_ctr = event_ctr;
+}
+
+#define NET_DIM_NEVENTS 64
+#define BITS_PER_TYPE(type) (sizeof(type) * BITS_PER_BYTE)
+#define BIT_GAP(bits, end, start) ((((end) - (start)) + BIT_ULL(bits)) & (BIT_ULL(bits) - 1))
+
+static inline void net_dim_calc_stats(struct net_dim_sample *start,
+ struct net_dim_sample *end,
+ struct net_dim_stats *curr_stats)
+{
+ /* u32 holds up to 71 minutes, should be enough */
+ u32 delta_us = ktime_us_delta(end->time, start->time);
+ u32 npkts = BIT_GAP(BITS_PER_TYPE(u32), end->pkt_ctr, start->pkt_ctr);
+ u32 nbytes = BIT_GAP(BITS_PER_TYPE(u32), end->byte_ctr,
+ start->byte_ctr);
+
+ if (!delta_us)
+ return;
+
+ curr_stats->ppms = DIV_ROUND_UP(npkts * USEC_PER_MSEC, delta_us);
+ curr_stats->bpms = DIV_ROUND_UP(nbytes * USEC_PER_MSEC, delta_us);
+ curr_stats->epms = DIV_ROUND_UP(NET_DIM_NEVENTS * USEC_PER_MSEC,
+ delta_us);
+}
+
+static inline void net_dim(struct net_dim *dim,
+ u16 event_ctr,
+ u64 packets,
+ u64 bytes)
+{
+ struct net_dim_sample end_sample;
+ struct net_dim_stats curr_stats;
+ u16 nevents;
+
+ switch (dim->state) {
+ case NET_DIM_MEASURE_IN_PROGRESS:
+ nevents = BIT_GAP(BITS_PER_TYPE(u16), event_ctr,
+ dim->start_sample.event_ctr);
+ if (nevents < NET_DIM_NEVENTS)
+ break;
+ net_dim_sample(event_ctr, packets, bytes, &end_sample);
+ net_dim_calc_stats(&dim->start_sample, &end_sample,
+ &curr_stats);
+ if (net_dim_decision(&curr_stats, dim)) {
+ dim->state = NET_DIM_APPLY_NEW_PROFILE;
+ schedule_work(&dim->work);
+ break;
+ }
+ /* fall through */
+ case NET_DIM_START_MEASURE:
+ net_dim_sample(event_ctr, packets, bytes, &dim->start_sample);
+ dim->state = NET_DIM_MEASURE_IN_PROGRESS;
+ break;
+ case NET_DIM_APPLY_NEW_PROFILE:
+ break;
+ }
+}
+
+#endif /* NET_DIM_H */
--
2.7.4
^ permalink raw reply related
* [net-next 08/10] net/dim: use struct net_dim_sample as arg to net_dim
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
Simplify the arguments net_dim() by formatting them into a struct
net_dim_sample before calling the function.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Suggested-by: Tal Gilboa <talgi@mellanox.com>
Acked-by: Tal Gilboa <talgi@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 13 ++++++++-----
include/linux/net_dim.h | 10 +++-------
2 files changed, 11 insertions(+), 12 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
index dae77a9..f292bb3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
@@ -78,11 +78,14 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget)
for (i = 0; i < c->num_tc; i++)
mlx5e_cq_arm(&c->sq[i].cq);
- if (MLX5E_TEST_BIT(c->rq.state, MLX5E_RQ_STATE_AM))
- net_dim(&c->rq.dim,
- c->rq.cq.event_ctr,
- c->rq.stats.packets,
- c->rq.stats.bytes);
+ if (MLX5E_TEST_BIT(c->rq.state, MLX5E_RQ_STATE_AM)) {
+ struct net_dim_sample dim_sample;
+ net_dim_sample(c->rq.cq.event_ctr,
+ c->rq.stats.packets,
+ c->rq.stats.bytes,
+ &dim_sample);
+ net_dim(&c->rq.dim, dim_sample);
+ }
mlx5e_cq_arm(&c->rq.cq);
mlx5e_cq_arm(&c->icosq.cq);
diff --git a/include/linux/net_dim.h b/include/linux/net_dim.h
index bb99073..2cceefa 100644
--- a/include/linux/net_dim.h
+++ b/include/linux/net_dim.h
@@ -341,21 +341,18 @@ static inline void net_dim_calc_stats(struct net_dim_sample *start,
}
static inline void net_dim(struct net_dim *dim,
- u16 event_ctr,
- u64 packets,
- u64 bytes)
+ struct net_dim_sample end_sample)
{
- struct net_dim_sample end_sample;
struct net_dim_stats curr_stats;
u16 nevents;
switch (dim->state) {
case NET_DIM_MEASURE_IN_PROGRESS:
- nevents = BIT_GAP(BITS_PER_TYPE(u16), event_ctr,
+ nevents = BIT_GAP(BITS_PER_TYPE(u16),
+ end_sample.event_ctr,
dim->start_sample.event_ctr);
if (nevents < NET_DIM_NEVENTS)
break;
- net_dim_sample(event_ctr, packets, bytes, &end_sample);
net_dim_calc_stats(&dim->start_sample, &end_sample,
&curr_stats);
if (net_dim_decision(&curr_stats, dim)) {
@@ -365,7 +362,6 @@ static inline void net_dim(struct net_dim *dim,
}
/* fall through */
case NET_DIM_START_MEASURE:
- net_dim_sample(event_ctr, packets, bytes, &dim->start_sample);
dim->state = NET_DIM_MEASURE_IN_PROGRESS;
break;
case NET_DIM_APPLY_NEW_PROFILE:
--
2.7.4
^ permalink raw reply related
* [net-next 09/10] bnxt_en: add support for software dynamic interrupt moderation
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
This implements the changes needed for the bnxt_en driver to add support
for dynamic interrupt moderation per ring.
This does add additional counters in the receive path, but testing shows
that any additional instructions are offset by throughput gain when the
default configuration is for low latency.
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
---
drivers/net/ethernet/broadcom/bnxt/Makefile | 2 +-
drivers/net/ethernet/broadcom/bnxt/bnxt.c | 52 +++++++++++++++++++++++
drivers/net/ethernet/broadcom/bnxt/bnxt.h | 34 ++++++++++-----
drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c | 32 ++++++++++++++
drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 12 ++++++
5 files changed, 120 insertions(+), 12 deletions(-)
create mode 100644 drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c
diff --git a/drivers/net/ethernet/broadcom/bnxt/Makefile b/drivers/net/ethernet/broadcom/bnxt/Makefile
index 59c8ec9..7c560d5 100644
--- a/drivers/net/ethernet/broadcom/bnxt/Makefile
+++ b/drivers/net/ethernet/broadcom/bnxt/Makefile
@@ -1,4 +1,4 @@
obj-$(CONFIG_BNXT) += bnxt_en.o
-bnxt_en-y := bnxt.o bnxt_sriov.o bnxt_ethtool.o bnxt_dcb.o bnxt_ulp.o bnxt_xdp.o bnxt_vfr.o bnxt_devlink.o
+bnxt_en-y := bnxt.o bnxt_sriov.o bnxt_ethtool.o bnxt_dcb.o bnxt_ulp.o bnxt_xdp.o bnxt_vfr.o bnxt_devlink.o bnxt_dim.o
bnxt_en-$(CONFIG_BNXT_FLOWER_OFFLOAD) += bnxt_tc.o
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 9efbdc6..3c5d2fa 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -1645,6 +1645,8 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_napi *bnapi, u32 *raw_cons,
rxr->rx_next_cons = NEXT_RX(cons);
next_rx_no_prod:
+ cpr->rx_packets += 1;
+ cpr->rx_bytes += len;
*raw_cons = tmp_raw_cons;
return rc;
@@ -1802,6 +1804,7 @@ static irqreturn_t bnxt_msix(int irq, void *dev_instance)
struct bnxt_cp_ring_info *cpr = &bnapi->cp_ring;
u32 cons = RING_CMP(cpr->cp_raw_cons);
+ cpr->event_ctr++;
prefetch(&cpr->cp_desc_ring[CP_RING(cons)][CP_IDX(cons)]);
napi_schedule(&bnapi->napi);
return IRQ_HANDLED;
@@ -2025,6 +2028,14 @@ static int bnxt_poll(struct napi_struct *napi, int budget)
break;
}
}
+ if (bp->flags & BNXT_FLAG_DIM) {
+ struct net_dim_sample dim_sample;
+ net_dim_sample(cpr->event_ctr,
+ cpr->rx_packets,
+ cpr->rx_bytes,
+ &dim_sample);
+ net_dim(&cpr->am, dim_sample);
+ }
mmiowb();
return work_done;
}
@@ -2610,6 +2621,8 @@ static void bnxt_init_cp_rings(struct bnxt *bp)
struct bnxt_ring_struct *ring = &cpr->cp_ring_struct;
ring->fw_ring_id = INVALID_HW_RING_ID;
+ cpr->rx_ring_coal.coal_ticks = bp->rx_coal.coal_ticks;
+ cpr->rx_ring_coal.coal_bufs = bp->rx_coal.coal_bufs;
}
}
@@ -4583,6 +4596,41 @@ static void bnxt_hwrm_set_coal_params(struct bnxt_coal *hw_coal,
req->flags = cpu_to_le16(flags);
}
+int bnxt_hwrm_set_ring_coal(struct bnxt *bp, struct bnxt_napi *bnapi)
+{
+ struct hwrm_ring_cmpl_ring_cfg_aggint_params_input req_rx = {0};
+ struct bnxt_cp_ring_info *cpr = &bnapi->cp_ring;
+ struct bnxt_coal coal;
+ unsigned int grp_idx;
+ int rc = 0;
+
+ /* Tick values in micro seconds.
+ * 1 coal_buf x bufs_per_record = 1 completion record.
+ */
+ memcpy(&coal, &bp->rx_coal, sizeof(struct bnxt_coal));
+
+ coal.coal_ticks = cpr->rx_ring_coal.coal_ticks;
+ coal.coal_bufs = cpr->rx_ring_coal.coal_bufs;
+
+ if (!bnapi->rx_ring)
+ return -ENODEV;
+
+ bnxt_hwrm_cmd_hdr_init(bp, &req_rx,
+ HWRM_RING_CMPL_RING_CFG_AGGINT_PARAMS, -1, -1);
+
+ bnxt_hwrm_set_coal_params(&coal, &req_rx);
+
+ mutex_lock(&bp->hwrm_cmd_lock);
+ grp_idx = bnapi->index;
+
+ req_rx.ring_id = cpu_to_le16(bp->grp_info[grp_idx].cp_fw_ring_id);
+
+ rc = _hwrm_send_message(bp, &req_rx, sizeof(req_rx),
+ HWRM_CMD_TIMEOUT);
+ mutex_unlock(&bp->hwrm_cmd_lock);
+ return rc;
+}
+
int bnxt_hwrm_set_coal(struct bnxt *bp)
{
int i, rc = 0;
@@ -5705,7 +5753,11 @@ static void bnxt_enable_napi(struct bnxt *bp)
int i;
for (i = 0; i < bp->cp_nr_rings; i++) {
+ struct bnxt_cp_ring_info *cpr = &bp->bnapi[i]->cp_ring;
bp->bnapi[i]->in_reset = false;
+
+ INIT_WORK(&cpr->am.work, bnxt_dim_work);
+ cpr->am.mode = NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE;
napi_enable(&bp->bnapi[i]->napi);
}
}
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 5359a1f..b3d5a363 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -23,6 +23,7 @@
#include <net/devlink.h>
#include <net/dst_metadata.h>
#include <net/switchdev.h>
+#include <linux/net_dim.h>
struct tx_bd {
__le32 tx_bd_len_flags_type;
@@ -607,6 +608,17 @@ struct bnxt_tx_ring_info {
struct bnxt_ring_struct tx_ring_struct;
};
+struct bnxt_coal {
+ u16 coal_ticks;
+ u16 coal_ticks_irq;
+ u16 coal_bufs;
+ u16 coal_bufs_irq;
+ /* RING_IDLE enabled when coal ticks < idle_thresh */
+ u16 idle_thresh;
+ u8 bufs_per_record;
+ u8 budget;
+};
+
struct bnxt_tpa_info {
void *data;
u8 *data_ptr;
@@ -670,6 +682,13 @@ struct bnxt_cp_ring_info {
u32 cp_raw_cons;
void __iomem *cp_doorbell;
+ struct bnxt_coal rx_ring_coal;
+ u64 rx_packets;
+ u64 rx_bytes;
+ u64 event_ctr;
+
+ struct net_dim am;
+
struct tx_cmp *cp_desc_ring[MAX_CP_PAGES];
dma_addr_t cp_desc_mapping[MAX_CP_PAGES];
@@ -944,17 +963,6 @@ struct bnxt_test_info {
#define BNXT_CAG_REG_LEGACY_INT_STATUS 0x4014
#define BNXT_CAG_REG_BASE 0x300000
-struct bnxt_coal {
- u16 coal_ticks;
- u16 coal_ticks_irq;
- u16 coal_bufs;
- u16 coal_bufs_irq;
- /* RING_IDLE enabled when coal ticks < idle_thresh */
- u16 idle_thresh;
- u8 bufs_per_record;
- u8 budget;
-};
-
struct bnxt_tc_flow_stats {
u64 packets;
u64 bytes;
@@ -1126,6 +1134,7 @@ struct bnxt {
#define BNXT_FLAG_DOUBLE_DB 0x400000
#define BNXT_FLAG_FW_DCBX_AGENT 0x800000
#define BNXT_FLAG_CHIP_NITRO_A0 0x1000000
+ #define BNXT_FLAG_DIM 0x2000000
#define BNXT_FLAG_ALL_CONFIG_FEATS (BNXT_FLAG_TPA | \
BNXT_FLAG_RFS | \
@@ -1423,4 +1432,7 @@ int bnxt_setup_mq_tc(struct net_device *dev, u8 tc);
int bnxt_get_max_rings(struct bnxt *, int *, int *, bool);
void bnxt_restore_pf_fw_resources(struct bnxt *bp);
int bnxt_port_attr_get(struct bnxt *bp, struct switchdev_attr *attr);
+void bnxt_dim_work(struct work_struct *work);
+int bnxt_hwrm_set_ring_coal(struct bnxt *bp, struct bnxt_napi *bnapi);
+
#endif
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c
new file mode 100644
index 0000000..0eefe4b
--- /dev/null
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c
@@ -0,0 +1,32 @@
+/*
+ * Copyright (c) 2017 Broadcom Limited
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/net_dim.h>
+#include "bnxt_hsi.h"
+#include "bnxt.h"
+
+void bnxt_dim_work(struct work_struct *work)
+{
+ struct net_dim *am = container_of(work, struct net_dim,
+ work);
+ struct bnxt_cp_ring_info *cpr = container_of(am,
+ struct bnxt_cp_ring_info,
+ am);
+ struct bnxt_napi *bnapi = container_of(cpr,
+ struct bnxt_napi,
+ cp_ring);
+ struct net_dim_cq_moder cur_profile = net_dim_get_profile(am->mode,
+ am->profile_ix);
+
+ cpr->rx_ring_coal.coal_ticks = cur_profile.usec;
+ cpr->rx_ring_coal.coal_bufs = cur_profile.pkts;
+
+ bnxt_hwrm_set_ring_coal(bnapi->bp, bnapi);
+ am->state = NET_DIM_START_MEASURE;
+}
+
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index fe7599f..632e03b 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -49,6 +49,8 @@ static int bnxt_get_coalesce(struct net_device *dev,
memset(coal, 0, sizeof(*coal));
+ coal->use_adaptive_rx_coalesce = bp->flags & BNXT_FLAG_DIM;
+
hw_coal = &bp->rx_coal;
mult = hw_coal->bufs_per_record;
coal->rx_coalesce_usecs = hw_coal->coal_ticks;
@@ -77,6 +79,15 @@ static int bnxt_set_coalesce(struct net_device *dev,
int rc = 0;
u16 mult;
+ if (coal->use_adaptive_rx_coalesce)
+ bp->flags |= BNXT_FLAG_DIM;
+ else {
+ if (bp->flags & BNXT_FLAG_DIM) {
+ bp->flags &= ~(BNXT_FLAG_DIM);
+ goto reset_coalesce;
+ }
+ }
+
hw_coal = &bp->rx_coal;
mult = hw_coal->bufs_per_record;
hw_coal->coal_ticks = coal->rx_coalesce_usecs;
@@ -104,6 +115,7 @@ static int bnxt_set_coalesce(struct net_device *dev,
update_stats = true;
}
+reset_coalesce:
if (netif_running(dev)) {
if (update_stats) {
rc = bnxt_close_nic(bp, true, false);
--
2.7.4
^ permalink raw reply related
* [net-next 10/10] MAINTAINERS: add entry for Dynamic Interrupt Moderation
From: Andy Gospodarek @ 2018-01-04 20:21 UTC (permalink / raw)
To: netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
From: Andy Gospodarek <gospo@broadcom.com>
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Signed-off-by: Tal Gilboa <talgi@mellanox.com>
---
MAINTAINERS | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index 753799d..769857b 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -4944,6 +4944,11 @@ S: Maintained
F: lib/dynamic_debug.c
F: include/linux/dynamic_debug.h
+DYNAMIC INTERRUPT MODERATION
+M: Tal Gilboa <talgi@mellanox.com>
+S: Mainained
+F: include/linux/net_dim.h
+
DZ DECSTATION DZ11 SERIAL DRIVER
M: "Maciej W. Rozycki" <macro@linux-mips.org>
S: Maintained
--
2.7.4
^ permalink raw reply related
* Re: [net-next 00/10] net: create dynamic software irq moderation library
From: Or Gerlitz @ 2018-01-04 20:37 UTC (permalink / raw)
To: Andy Gospodarek
Cc: Linux Netdev List, mchan, Tal Gilboa, Or Gerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
> net/mlx5e: move interrupt moderation structs to new file
> net/mlx5e: move interrupt moderation forward declarations
> net/mlx5e: remove rq references in mlx5e_rx_am
> net/mlx5e: move AM logic enums
> net/mlx5e: move generic functions to new file
> net/mlx5e: change Mellanox references in DIM code
Hi, Andy && happy new 2018 -- this is indeed a nit, but I have
provided it to you twice (...),
please get the commit titles to align with what we do which is capital
letter after the net/mlx5e: prefix
from: net/mlx5e: move interrupt moderation structs to new file
to: net/mlx5e: Move interrupt moderation structs to new file
If you get other comments, just apply this for the next version, if everyone
is happy, that would be a very small effort to just fix and get that in..
^ permalink raw reply
* Re: [PATCH] ps3_gelic_net: Delete an error message for a failed memory allocation in gelic_descr_prepare_rx()
From: Geoff Levand @ 2018-01-04 20:44 UTC (permalink / raw)
To: SF Markus Elfring, netdev, linuxppc-dev, Benjamin Herrenschmidt,
Michael Ellerman, Paul Mackerras
Cc: kernel-janitors, LKML
In-Reply-To: <9613bfbf-11cc-1e66-484a-84fcda022861@users.sourceforge.net>
On 01/03/2018 06:40 AM, SF Markus Elfring wrote:
> Omit an extra message for a memory allocation failure in this function.
I applied this to my ps3-queue branch.
As I mentioned to you several times before, please keep
the commit subject line to less than 50 characters.
Also, in this case the prefix would be 'net/ps3_gelic_net:"
-Geoff
^ permalink raw reply
* Re: [PATCH net-next 0/5] Export SERDES stats via ethtool -S
From: Andrew Lunn @ 2018-01-04 21:07 UTC (permalink / raw)
To: Russell King - ARM Linux
Cc: David Miller, Vivien Didelot, Florian Fainelli, netdev
In-Reply-To: <20180103175832.GJ28752@n2100.armlinux.org.uk>
> Hi Andrew,
>
> Isn't the serdes interface just a normal PHY - the register set in
> Viviens debugfs patch certainly looks like an 88e1545 in one of the
> switches I've here (I think on the dev rev B). Is there a reason why
> this PHY isn't visible as a normal PHY, just like we export the other
> internal PHYs?
Humm, interesting idea. Yes, it looks like a normal PHY. It is however
not contiguous to the internal PHYs on the 6352. You probably can
however see it, if you look on the MDIO bus, at address 0xf.
Depends on what mode the ports are in, the SERDES can be connected to
either port 4 or 5. Port 5 does not have an internal PHY, but port 4
does. We could determine what port is using the SERDES, and redirect
PHY read/writes to the SERDES interface. But it gets a bit complex for
port 4, since it can use either copper or SERDES depending on which
comes up first. So suppose you need to be able to see both the copper
PHY and the SERDES, so you can initialize them both, in order for one
to actually come up.
Andrew
^ permalink raw reply
* Re: [PATCH bpf-next 1/2] bpf: implement syscall command BPF_MAP_GET_NEXT_KEY for stacktrace map
From: Jakub Kicinski @ 2018-01-04 21:08 UTC (permalink / raw)
To: Yonghong Song; +Cc: ast, daniel, netdev, kernel-team
In-Reply-To: <20180104072746.1569033-2-yhs@fb.com>
On Wed, 3 Jan 2018 23:27:45 -0800, Yonghong Song wrote:
> Currently, bpf syscall command BPF_MAP_GET_NEXT_KEY is not
> supported for stacktrace map. However, there are use cases where
> user space wants to enumerate all stacktrace map entries where
> BPF_MAP_GET_NEXT_KEY command will be really helpful.
> In addition, if user space wants to delete all map entries
> in order to save memory and does not want to close the
> map file descriptor, BPF_MAP_GET_NEXT_KEY may help improve
> performance if map entries are sparsely populated.
>
> The implementation follows the API specification of existing
> BPF_MAP_GET_NEXT_KEY implementation. If user provides
> an NULL key pointer, the first key is returned. Otherwise,
> the first valid key after the input parameter "key"
> is returned, or -ENOENT if no valid key can be found.
>
> Signed-off-by: Yonghong Song <yhs@fb.com>
> ---
> kernel/bpf/stackmap.c | 23 +++++++++++++++++++++--
> 1 file changed, 21 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
> index a15bc63..207b21c 100644
> --- a/kernel/bpf/stackmap.c
> +++ b/kernel/bpf/stackmap.c
> @@ -226,9 +226,28 @@ int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value)
> return 0;
> }
>
> -static int stack_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
> +static int stack_map_get_next_key(struct bpf_map *map, void *key,
> + void *next_key)
> {
> - return -EINVAL;
> + struct bpf_stack_map *smap = container_of(map,
> + struct bpf_stack_map, map);
> + u32 id;
> +
> + WARN_ON_ONCE(!rcu_read_lock_held());
> +
> + if (!key)
> + id = 0;
> + else
> + id = *(u32 *)key + 1;
> +
> + while (id < smap->n_buckets && !smap->buckets[id])
> + id++;
> +
> + if (id >= smap->n_buckets)
> + return -ENOENT;
AFAIU for hash maps the semantics of get next are as follows:
get_next(map, key) {
if (!key)
return get_first(map);
elem = lookup(map, key);
if (!elem) // <-- note this branch
return get_first(map);
if (elem->next)
return elem->next->key;
return -ENOENT;
}
For arrays elements always exist, hence the elem->next check is
omitted. Here you are, however, testing !smap->buckets[id] so I assume
elements may not exist.
Is there any precedent for get_next on non-existent key returning
element other than first? The stacktrace map is a bit special, and
returning id + 1 would defeat what you're trying to do here.. Is there
value in keeping the behaviour consistent across map types?
Anyway, you said in the commit message that "The implementation follows
the API specification of existing BPF_MAP_GET_NEXT_KEY implementation."
and I find that arguable :)
> + *(u32 *)next_key = id;
> + return 0;
> }
>
> static int stack_map_update_elem(struct bpf_map *map, void *key, void *value,
^ permalink raw reply
* [PATCH] sh_eth: remove sh_eth_plat_data::edmac_endian
From: Sergei Shtylyov @ 2018-01-04 21:26 UTC (permalink / raw)
To: Yoshinori Sato, Rich Felker, linux-sh, netdev
Cc: linux-renesas-soc, Sergei Shtylyov
[-- Attachment #1: sh_eth-remove-sh_eth_plat_data-edmac_endian.patch --]
[-- Type: text/plain, Size: 4872 bytes --]
Since the commit 888cc8c20cf ("sh_eth: remove EDMAC_BIG_ENDIAN") (geez,
I didn't realize that was 2 years ago!) the initializers in the SuperH
platform code for the 'sh_eth_plat_data::edmac_endian' stopped to matter,
so we can remove that field for good (not sure if it was ever useful --
SH7786 Ether has been reported to have the same EDMAC descriptor/register
endiannes as configured for the SuperH CPU)...
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
---
The patch is against DaveM's 'net-next.git' repo.
Not sure who should apply the patch -- will prolly be faster if DaveM does...
arch/sh/boards/board-espt.c | 1 -
arch/sh/boards/board-sh7757lcr.c | 4 ----
arch/sh/boards/mach-ecovec24/setup.c | 1 -
arch/sh/boards/mach-se/7724/setup.c | 1 -
arch/sh/boards/mach-sh7763rdp/setup.c | 1 -
arch/sh/kernel/cpu/sh2/setup-sh7619.c | 1 -
include/linux/sh_eth.h | 3 ---
7 files changed, 12 deletions(-)
Index: net-next/arch/sh/boards/board-espt.c
===================================================================
--- net-next.orig/arch/sh/boards/board-espt.c
+++ net-next/arch/sh/boards/board-espt.c
@@ -79,7 +79,6 @@ static struct resource sh_eth_resources[
static struct sh_eth_plat_data sh7763_eth_pdata = {
.phy = 0,
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.phy_interface = PHY_INTERFACE_MODE_MII,
};
Index: net-next/arch/sh/boards/board-sh7757lcr.c
===================================================================
--- net-next.orig/arch/sh/boards/board-sh7757lcr.c
+++ net-next/arch/sh/boards/board-sh7757lcr.c
@@ -76,7 +76,6 @@ static struct resource sh_eth0_resources
static struct sh_eth_plat_data sh7757_eth0_pdata = {
.phy = 1,
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.set_mdio_gate = sh7757_eth_set_mdio_gate,
};
@@ -104,7 +103,6 @@ static struct resource sh_eth1_resources
static struct sh_eth_plat_data sh7757_eth1_pdata = {
.phy = 1,
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.set_mdio_gate = sh7757_eth_set_mdio_gate,
};
@@ -148,7 +146,6 @@ static struct resource sh_eth_giga0_reso
static struct sh_eth_plat_data sh7757_eth_giga0_pdata = {
.phy = 18,
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.set_mdio_gate = sh7757_eth_giga_set_mdio_gate,
.phy_interface = PHY_INTERFACE_MODE_RGMII_ID,
};
@@ -182,7 +179,6 @@ static struct resource sh_eth_giga1_reso
static struct sh_eth_plat_data sh7757_eth_giga1_pdata = {
.phy = 19,
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.set_mdio_gate = sh7757_eth_giga_set_mdio_gate,
.phy_interface = PHY_INTERFACE_MODE_RGMII_ID,
};
Index: net-next/arch/sh/boards/mach-ecovec24/setup.c
===================================================================
--- net-next.orig/arch/sh/boards/mach-ecovec24/setup.c
+++ net-next/arch/sh/boards/mach-ecovec24/setup.c
@@ -159,7 +159,6 @@ static struct resource sh_eth_resources[
static struct sh_eth_plat_data sh_eth_plat = {
.phy = 0x1f, /* SMSC LAN8700 */
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.phy_interface = PHY_INTERFACE_MODE_MII,
.ether_link_active_low = 1
};
Index: net-next/arch/sh/boards/mach-se/7724/setup.c
===================================================================
--- net-next.orig/arch/sh/boards/mach-se/7724/setup.c
+++ net-next/arch/sh/boards/mach-se/7724/setup.c
@@ -374,7 +374,6 @@ static struct resource sh_eth_resources[
static struct sh_eth_plat_data sh_eth_plat = {
.phy = 0x1f, /* SMSC LAN8187 */
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.phy_interface = PHY_INTERFACE_MODE_MII,
};
Index: net-next/arch/sh/boards/mach-sh7763rdp/setup.c
===================================================================
--- net-next.orig/arch/sh/boards/mach-sh7763rdp/setup.c
+++ net-next/arch/sh/boards/mach-sh7763rdp/setup.c
@@ -87,7 +87,6 @@ static struct resource sh_eth_resources[
static struct sh_eth_plat_data sh7763_eth_pdata = {
.phy = 1,
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.phy_interface = PHY_INTERFACE_MODE_MII,
};
Index: net-next/arch/sh/kernel/cpu/sh2/setup-sh7619.c
===================================================================
--- net-next.orig/arch/sh/kernel/cpu/sh2/setup-sh7619.c
+++ net-next/arch/sh/kernel/cpu/sh2/setup-sh7619.c
@@ -122,7 +122,6 @@ static struct platform_device scif2_devi
static struct sh_eth_plat_data eth_platform_data = {
.phy = 1,
- .edmac_endian = EDMAC_LITTLE_ENDIAN,
.phy_interface = PHY_INTERFACE_MODE_MII,
};
Index: net-next/include/linux/sh_eth.h
===================================================================
--- net-next.orig/include/linux/sh_eth.h
+++ net-next/include/linux/sh_eth.h
@@ -5,12 +5,9 @@
#include <linux/phy.h>
#include <linux/if_ether.h>
-enum {EDMAC_LITTLE_ENDIAN};
-
struct sh_eth_plat_data {
int phy;
int phy_irq;
- int edmac_endian;
phy_interface_t phy_interface;
void (*set_mdio_gate)(void *addr);
^ permalink raw reply
* Re: [PATCH bpf-next 1/2] bpf: implement syscall command BPF_MAP_GET_NEXT_KEY for stacktrace map
From: Yonghong Song @ 2018-01-04 21:32 UTC (permalink / raw)
To: Jakub Kicinski; +Cc: ast, daniel, netdev, kernel-team
In-Reply-To: <20180104130822.377721d8@cakuba.netronome.com>
On 1/4/18 1:08 PM, Jakub Kicinski wrote:
> On Wed, 3 Jan 2018 23:27:45 -0800, Yonghong Song wrote:
>> Currently, bpf syscall command BPF_MAP_GET_NEXT_KEY is not
>> supported for stacktrace map. However, there are use cases where
>> user space wants to enumerate all stacktrace map entries where
>> BPF_MAP_GET_NEXT_KEY command will be really helpful.
>> In addition, if user space wants to delete all map entries
>> in order to save memory and does not want to close the
>> map file descriptor, BPF_MAP_GET_NEXT_KEY may help improve
>> performance if map entries are sparsely populated.
>>
>> The implementation follows the API specification of existing
>> BPF_MAP_GET_NEXT_KEY implementation. If user provides
>> an NULL key pointer, the first key is returned. Otherwise,
>> the first valid key after the input parameter "key"
>> is returned, or -ENOENT if no valid key can be found.
>>
>> Signed-off-by: Yonghong Song <yhs@fb.com>
>> ---
>> kernel/bpf/stackmap.c | 23 +++++++++++++++++++++--
>> 1 file changed, 21 insertions(+), 2 deletions(-)
>>
>> diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
>> index a15bc63..207b21c 100644
>> --- a/kernel/bpf/stackmap.c
>> +++ b/kernel/bpf/stackmap.c
>> @@ -226,9 +226,28 @@ int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value)
>> return 0;
>> }
>>
>> -static int stack_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
>> +static int stack_map_get_next_key(struct bpf_map *map, void *key,
>> + void *next_key)
>> {
>> - return -EINVAL;
>> + struct bpf_stack_map *smap = container_of(map,
>> + struct bpf_stack_map, map);
>> + u32 id;
>> +
>> + WARN_ON_ONCE(!rcu_read_lock_held());
>> +
>> + if (!key)
>> + id = 0;
>> + else
>> + id = *(u32 *)key + 1;
>> +
>> + while (id < smap->n_buckets && !smap->buckets[id])
>> + id++;
>> +
>> + if (id >= smap->n_buckets)
>> + return -ENOENT;
>
> AFAIU for hash maps the semantics of get next are as follows:
>
> get_next(map, key) {
> if (!key)
> return get_first(map);
>
> elem = lookup(map, key);
> if (!elem) // <-- note this branch
> return get_first(map);
> if (elem->next)
> return elem->next->key;
> return -ENOENT;
> }
>
> For arrays elements always exist, hence the elem->next check is
> omitted. Here you are, however, testing !smap->buckets[id] so I assume
> elements may not exist.
Right, buckets[id] could be NULL.
>
> Is there any precedent for get_next on non-existent key returning
> element other than first? The stacktrace map is a bit special, and
Sorry, I miss this. You are right. hashtable get_next_key will return
the first for non-existing key. And all other implemented get_next_key
is a variant of arrays where all keys already exist.
> returning id + 1 would defeat what you're trying to do here.. Is there
> value in keeping the behaviour consistent across map types?
Let us keep the behavior consistent with hashtable then.
>
> Anyway, you said in the commit message that "The implementation follows
> the API specification of existing BPF_MAP_GET_NEXT_KEY implementation."
> and I find that arguable :)
You are right. Will send v2 soon with re-wording of commit message as well.
>> + *(u32 *)next_key = id;
>> + return 0;
>> }
>>
>> static int stack_map_update_elem(struct bpf_map *map, void *key, void *value,
>
^ permalink raw reply
* Re: [net-next 00/10] net: create dynamic software irq moderation library
From: Saeed Mahameed @ 2018-01-04 21:37 UTC (permalink / raw)
To: Andy Gospodarek, netdev; +Cc: mchan, talgi, ogerlitz, Andy Gospodarek
In-Reply-To: <1515097290-17470-1-git-send-email-andy@greyhouse.net>
On 1/4/2018 12:21 PM, Andy Gospodarek wrote:
> From: Andy Gospodarek <gospo@broadcom.com>
>
> This converts the dynamic interrupt moderation library from the mlx5_en driver
> into a library so it can be used by any driver. The penultimatepatch in this
> set adds support for interrupt moderation in the bnxt_en driver and the last
> patch creates an entry in the MAINTAINERS file.
>
> The main purpose of this code in the mlx5_en driver is to allow an
> administrator to make sure that default coalesce settings are optimized
> for low latency, but quickly adapt to handle high throughput traffic and
> optimize how many packets are received during each napi poll.
>
> For any new driver the following changes would be needed to use this
> library:
>
> - add elements in ring struct to track items needed by this library
> - create function that can be called to actually set coalesce settings
> for the driver
>
> Credit to Rob Rice and Lee Reed for doing some of the initial proof of
> concept and testing for this patch and Tal Gilboa and Or Gerlitz for their
> comments, etc on this set.
> > Andy Gospodarek (10):
> net/mlx5e: move interrupt moderation structs to new file
> net/mlx5e: move interrupt moderation forward declarations
> net/mlx5e: remove rq references in mlx5e_rx_am
> net/mlx5e: move AM logic enums
> net/mlx5e: move generic functions to new file
> net/mlx5e: change Mellanox references in DIM code
> net: move dynamic interrpt coalescing code to include/linux
> net/dim: use struct net_dim_sample as arg to net_dim
> bnxt_en: add support for software dynamic interrupt moderation
> MAINTAINERS: add entry for Dynamic Interrupt Moderation
Very clean and nice work!
Thank you Andy for following my suggestion of changing the API to be
static inline helper functions instead of function pointers provided by
device drivers, the current API is less demanding for the device drivers
and has no performance impact.
Awesome work.
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
>
> MAINTAINERS | 5 +
> drivers/net/ethernet/broadcom/bnxt/Makefile | 2 +-
> drivers/net/ethernet/broadcom/bnxt/bnxt.c | 52 +++
> drivers/net/ethernet/broadcom/bnxt/bnxt.h | 34 +-
> drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c | 32 ++
> drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 12 +
> drivers/net/ethernet/mellanox/mlx5/core/Makefile | 2 +-
> drivers/net/ethernet/mellanox/mlx5/core/en.h | 46 +--
> drivers/net/ethernet/mellanox/mlx5/core/en_dim.c | 49 +++
> .../net/ethernet/mellanox/mlx5/core/en_ethtool.c | 12 +-
> drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 32 +-
> drivers/net/ethernet/mellanox/mlx5/core/en_rep.c | 4 +-
> drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c | 341 -------------------
Goodbye mlx5/core/en_rx_am.c :)
> drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 10 +-
> drivers/net/ethernet/mellanox/mlx5/core/net_dim.h | 108 ++++++
> include/linux/mlx5/mlx5_ifc.h | 6 -
> include/linux/net_dim.h | 372 +++++++++++++++++++++
> 17 files changed, 693 insertions(+), 426 deletions(-)
> create mode 100644 drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c
> create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en_dim.c
> delete mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en_rx_am.c
> create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/net_dim.h
> create mode 100644 include/linux/net_dim.h
>
^ permalink raw reply
* Re: bonding: Completion of error handling around bond_update_slave_arr()
From: SF Markus Elfring @ 2018-01-04 21:41 UTC (permalink / raw)
To: Mahesh Bandewar (महेश बंडेवार),
linux-netdev
Cc: Andy Gospodarek, Jay Vosburgh, Veaceslav Falico, LKML,
kernel-janitors
In-Reply-To: <CAF2d9jjyfWT8Nm1V3fCKSrEE4Xs9gShs7nvj+dAYqXg42kUhvg@mail.gmail.com>
>>> If you see 8 out of 9 call sites in this file ignore the return value.
>>
>> How do you think about to fix error detection and corresponding
>> exception handling then?
>>
> If I understand your question correctly - not having memory is not a
> correctable error
I am unsure if it would be feasible to retry memory allocations for
this software module under other circumstances.
> and hence there are consequences.
Could one consequence be to let the error code “-ENOMEM” move through
the function call hierarchy?
Regards,
Markus
^ permalink raw reply
* [PATCH bpf-next v2 1/2] bpf: implement syscall command BPF_MAP_GET_NEXT_KEY for stacktrace map
From: Yonghong Song @ 2018-01-04 21:55 UTC (permalink / raw)
To: ast, daniel, netdev; +Cc: kernel-team
In-Reply-To: <20180104215504.2013475-1-yhs@fb.com>
Currently, bpf syscall command BPF_MAP_GET_NEXT_KEY is not
supported for stacktrace map. However, there are use cases where
user space wants to enumerate all stacktrace map entries where
BPF_MAP_GET_NEXT_KEY command will be really helpful.
In addition, if user space wants to delete all map entries
in order to save memory and does not want to close the
map file descriptor, BPF_MAP_GET_NEXT_KEY may help improve
performance if map entries are sparsely populated.
The implementation has similar behavior for
BPF_MAP_GET_NEXT_KEY implementation in hashtab. If user provides
a NULL key pointer or an invalid key, the first key is returned.
Otherwise, the first valid key after the input parameter "key"
is returned, or -ENOENT if no valid key can be found.
Signed-off-by: Yonghong Song <yhs@fb.com>
---
kernel/bpf/stackmap.c | 28 ++++++++++++++++++++++++++--
1 file changed, 26 insertions(+), 2 deletions(-)
diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
index a15bc63..6c63c22 100644
--- a/kernel/bpf/stackmap.c
+++ b/kernel/bpf/stackmap.c
@@ -226,9 +226,33 @@ int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value)
return 0;
}
-static int stack_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
+static int stack_map_get_next_key(struct bpf_map *map, void *key,
+ void *next_key)
{
- return -EINVAL;
+ struct bpf_stack_map *smap = container_of(map,
+ struct bpf_stack_map, map);
+ u32 id;
+
+ WARN_ON_ONCE(!rcu_read_lock_held());
+
+ if (!key) {
+ id = 0;
+ } else {
+ id = *(u32 *)key;
+ if (id >= smap->n_buckets || !smap->buckets[id])
+ id = 0;
+ else
+ id++;
+ }
+
+ while (id < smap->n_buckets && !smap->buckets[id])
+ id++;
+
+ if (id >= smap->n_buckets)
+ return -ENOENT;
+
+ *(u32 *)next_key = id;
+ return 0;
}
static int stack_map_update_elem(struct bpf_map *map, void *key, void *value,
--
2.9.5
^ permalink raw reply related
* [PATCH bpf-next v2 2/2] tools/bpf: add a bpf selftest for stacktrace
From: Yonghong Song @ 2018-01-04 21:55 UTC (permalink / raw)
To: ast, daniel, netdev; +Cc: kernel-team
In-Reply-To: <20180104215504.2013475-1-yhs@fb.com>
Added a bpf selftest in test_progs at tools directory for stacktrace.
The test will populate a hashtable map and a stacktrace map
at the same time with the same key, stackid.
The user space will compare both maps, using BPF_MAP_LOOKUP_ELEM
command and BPF_MAP_GET_NEXT_KEY command, to ensure that both have
the same set of keys.
Signed-off-by: Yonghong Song <yhs@fb.com>
---
tools/testing/selftests/bpf/Makefile | 2 +-
tools/testing/selftests/bpf/test_progs.c | 127 ++++++++++++++++++++++
tools/testing/selftests/bpf/test_stacktrace_map.c | 62 +++++++++++
3 files changed, 190 insertions(+), 1 deletion(-)
create mode 100644 tools/testing/selftests/bpf/test_stacktrace_map.c
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
index 1304753..a8aa7e2 100644
--- a/tools/testing/selftests/bpf/Makefile
+++ b/tools/testing/selftests/bpf/Makefile
@@ -19,7 +19,7 @@ TEST_GEN_PROGS = test_verifier test_tag test_maps test_lru_map test_lpm_map test
TEST_GEN_FILES = test_pkt_access.o test_xdp.o test_l4lb.o test_tcp_estats.o test_obj_id.o \
test_pkt_md_access.o test_xdp_redirect.o test_xdp_meta.o sockmap_parse_prog.o \
sockmap_verdict_prog.o dev_cgroup.o sample_ret0.o test_tracepoint.o \
- test_l4lb_noinline.o test_xdp_noinline.o
+ test_l4lb_noinline.o test_xdp_noinline.o test_stacktrace_map.o
TEST_PROGS := test_kmod.sh test_xdp_redirect.sh test_xdp_meta.sh \
test_offload.py
diff --git a/tools/testing/selftests/bpf/test_progs.c b/tools/testing/selftests/bpf/test_progs.c
index 09087ab..b549308 100644
--- a/tools/testing/selftests/bpf/test_progs.c
+++ b/tools/testing/selftests/bpf/test_progs.c
@@ -837,6 +837,132 @@ static void test_tp_attach_query(void)
free(query);
}
+static int compare_map_keys(int map1_fd, int map2_fd)
+{
+ __u32 key, next_key;
+ char val_buf[PERF_MAX_STACK_DEPTH * sizeof(__u64)];
+ int err;
+
+ err = bpf_map_get_next_key(map1_fd, NULL, &key);
+ if (err)
+ return err;
+ err = bpf_map_lookup_elem(map2_fd, &key, val_buf);
+ if (err)
+ return err;
+
+ while (bpf_map_get_next_key(map1_fd, &key, &next_key) == 0) {
+ err = bpf_map_lookup_elem(map2_fd, &next_key, val_buf);
+ if (err)
+ return err;
+
+ key = next_key;
+ }
+ if (errno != ENOENT)
+ return -1;
+
+ return 0;
+}
+
+static void test_stacktrace_map()
+{
+ int control_map_fd, stackid_hmap_fd, stackmap_fd;
+ const char *file = "./test_stacktrace_map.o";
+ int bytes, efd, err, pmu_fd, prog_fd;
+ struct perf_event_attr attr = {};
+ __u32 key, val, duration = 0;
+ struct bpf_object *obj;
+ char buf[256];
+
+ err = bpf_prog_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj, &prog_fd);
+ if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
+ goto out;
+
+ /* Get the ID for the sched/sched_switch tracepoint */
+ snprintf(buf, sizeof(buf),
+ "/sys/kernel/debug/tracing/events/sched/sched_switch/id");
+ efd = open(buf, O_RDONLY, 0);
+ if (CHECK(efd < 0, "open", "err %d errno %d\n", efd, errno))
+ goto close_prog;
+
+ bytes = read(efd, buf, sizeof(buf));
+ close(efd);
+ if (CHECK(bytes <= 0 || bytes >= sizeof(buf),
+ "read", "bytes %d errno %d\n", bytes, errno))
+ goto close_prog;
+
+ /* Open the perf event and attach bpf progrram */
+ attr.config = strtol(buf, NULL, 0);
+ attr.type = PERF_TYPE_TRACEPOINT;
+ attr.sample_type = PERF_SAMPLE_RAW | PERF_SAMPLE_CALLCHAIN;
+ attr.sample_period = 1;
+ attr.wakeup_events = 1;
+ pmu_fd = syscall(__NR_perf_event_open, &attr, -1 /* pid */,
+ 0 /* cpu 0 */, -1 /* group id */,
+ 0 /* flags */);
+ if (CHECK(pmu_fd < 0, "perf_event_open", "err %d errno %d\n",
+ pmu_fd, errno))
+ goto close_prog;
+
+ err = ioctl(pmu_fd, PERF_EVENT_IOC_ENABLE, 0);
+ if (CHECK(err, "perf_event_ioc_enable", "err %d errno %d\n",
+ err, errno))
+ goto close_pmu;
+
+ err = ioctl(pmu_fd, PERF_EVENT_IOC_SET_BPF, prog_fd);
+ if (CHECK(err, "perf_event_ioc_set_bpf", "err %d errno %d\n",
+ err, errno))
+ goto disable_pmu;
+
+ /* find map fds */
+ control_map_fd = bpf_find_map(__func__, obj, "control_map");
+ if (CHECK(control_map_fd < 0, "bpf_find_map control_map",
+ "err %d errno %d\n", err, errno))
+ goto disable_pmu;
+
+ stackid_hmap_fd = bpf_find_map(__func__, obj, "stackid_hmap");
+ if (CHECK(stackid_hmap_fd < 0, "bpf_find_map stackid_hmap",
+ "err %d errno %d\n", err, errno))
+ goto disable_pmu;
+
+ stackmap_fd = bpf_find_map(__func__, obj, "stackmap");
+ if (CHECK(stackmap_fd < 0, "bpf_find_map stackmap", "err %d errno %d\n",
+ err, errno))
+ goto disable_pmu;
+
+ /* give some time for bpf program run */
+ sleep(1);
+
+ /* disable stack trace collection */
+ key = 0;
+ val = 1;
+ bpf_map_update_elem(control_map_fd, &key, &val, 0);
+
+ /* for every element in stackid_hmap, we can find a corresponding one
+ * in stackmap, and vise versa.
+ */
+ err = compare_map_keys(stackid_hmap_fd, stackmap_fd);
+ if (CHECK(err, "compare_map_keys stackid_hmap vs. stackmap",
+ "err %d errno %d\n", err, errno))
+ goto disable_pmu;
+
+ err = compare_map_keys(stackmap_fd, stackid_hmap_fd);
+ if (CHECK(err, "compare_map_keys stackmap vs. stackid_hmap",
+ "err %d errno %d\n", err, errno))
+ ; /* fall through */
+
+disable_pmu:
+ ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
+
+close_pmu:
+ close(pmu_fd);
+
+close_prog:
+ bpf_object__close(obj);
+
+out:
+ return;
+}
+
int main(void)
{
struct rlimit rinf = { RLIM_INFINITY, RLIM_INFINITY };
@@ -852,6 +978,7 @@ int main(void)
test_pkt_md_access();
test_obj_name();
test_tp_attach_query();
+ test_stacktrace_map();
printf("Summary: %d PASSED, %d FAILED\n", pass_cnt, error_cnt);
return error_cnt ? EXIT_FAILURE : EXIT_SUCCESS;
diff --git a/tools/testing/selftests/bpf/test_stacktrace_map.c b/tools/testing/selftests/bpf/test_stacktrace_map.c
new file mode 100644
index 0000000..76d85c5d
--- /dev/null
+++ b/tools/testing/selftests/bpf/test_stacktrace_map.c
@@ -0,0 +1,62 @@
+// SPDX-License-Identifier: GPL-2.0
+// Copyright (c) 2018 Facebook
+
+#include <linux/bpf.h>
+#include "bpf_helpers.h"
+
+#ifndef PERF_MAX_STACK_DEPTH
+#define PERF_MAX_STACK_DEPTH 127
+#endif
+
+struct bpf_map_def SEC("maps") control_map = {
+ .type = BPF_MAP_TYPE_ARRAY,
+ .key_size = sizeof(__u32),
+ .value_size = sizeof(__u32),
+ .max_entries = 1,
+};
+
+struct bpf_map_def SEC("maps") stackid_hmap = {
+ .type = BPF_MAP_TYPE_HASH,
+ .key_size = sizeof(__u32),
+ .value_size = sizeof(__u32),
+ .max_entries = 10000,
+};
+
+struct bpf_map_def SEC("maps") stackmap = {
+ .type = BPF_MAP_TYPE_STACK_TRACE,
+ .key_size = sizeof(__u32),
+ .value_size = sizeof(__u64) * PERF_MAX_STACK_DEPTH,
+ .max_entries = 10000,
+};
+
+/* taken from /sys/kernel/debug/tracing/events/sched/sched_switch/format */
+struct sched_switch_args {
+ unsigned long long pad;
+ char prev_comm[16];
+ int prev_pid;
+ int prev_prio;
+ long long prev_state;
+ char next_comm[16];
+ int next_pid;
+ int next_prio;
+};
+
+SEC("tracepoint/sched/sched_switch")
+int oncpu(struct sched_switch_args *ctx)
+{
+ __u32 key = 0, val = 0, *value_p;
+
+ value_p = bpf_map_lookup_elem(&control_map, &key);
+ if (value_p && *value_p)
+ return 0; /* skip if non-zero *value_p */
+
+ /* The size of stackmap and stackid_hmap should be the same */
+ key = bpf_get_stackid(ctx, &stackmap, 0);
+ if ((int)key >= 0)
+ bpf_map_update_elem(&stackid_hmap, &key, &val, 0);
+
+ return 0;
+}
+
+char _license[] SEC("license") = "GPL";
+__u32 _version SEC("version") = 1; /* ignored by tracepoints, required by libbpf.a */
--
2.9.5
^ permalink raw reply related
* [PATCH bpf-next v2 0/2] bpf: implement syscall command BPF_MAP_GET_NEXT_KEY for stacktrace map
From: Yonghong Song @ 2018-01-04 21:55 UTC (permalink / raw)
To: ast, daniel, netdev; +Cc: kernel-team
The patch set implements bpf syscall command BPF_MAP_GET_NEXT_KEY
for stacktrace map. Patch #1 is the core implementation
and Patch #2 implements a bpf test at tools/testing/selftests/bpf
directory. Please see individual patch comments for details.
Changelog:
v1 -> v2:
- For invalid key (key pointer is non-NULL), sets next_key to be the first valid key.
Yonghong Song (2):
bpf: implement syscall command BPF_MAP_GET_NEXT_KEY for stacktrace map
tools/bpf: add a bpf selftest for stacktrace
kernel/bpf/stackmap.c | 28 ++++-
tools/testing/selftests/bpf/Makefile | 2 +-
tools/testing/selftests/bpf/test_progs.c | 127 ++++++++++++++++++++++
tools/testing/selftests/bpf/test_stacktrace_map.c | 62 +++++++++++
4 files changed, 216 insertions(+), 3 deletions(-)
create mode 100644 tools/testing/selftests/bpf/test_stacktrace_map.c
--
2.9.5
^ permalink raw reply
* [PATCH net-next] net: ipv6: Allow connect to linklocal address from socket bound to vrf
From: David Ahern @ 2018-01-04 22:03 UTC (permalink / raw)
To: netdev; +Cc: David Ahern
Allow a process bound to a VRF to connect to a linklocal address.
Currently, this fails because of a mismatch between the scope of the
linklocal address and the sk_bound_dev_if inherited by the VRF binding:
$ ssh -6 fe80::70b8:cff:fedd:ead8%eth1
ssh: connect to host fe80::70b8:cff:fedd:ead8%eth1 port 22: Invalid argument
Relax the scope check to allow the socket to be bound to the same L3
device as the scope id.
This makes ipv6 linklocal consistent with other relaxed checks enabled
by commits 1ff23beebdd3 ("net: l3mdev: Allow send on enslaved interface")
and 7bb387c5ab12a ("net: Allow IP_MULTICAST_IF to set index to L3 slave").
Signed-off-by: David Ahern <dsahern@gmail.com>
---
include/net/sock.h | 20 ++++++++++++++++++++
net/ipv6/datagram.c | 3 +--
net/ipv6/tcp_ipv6.c | 3 +--
3 files changed, 22 insertions(+), 4 deletions(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index 66fd3951e6f3..73b7830b0bb8 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -72,6 +72,7 @@
#include <net/tcp_states.h>
#include <linux/net_tstamp.h>
#include <net/smc.h>
+#include <net/l3mdev.h>
/*
* This structure really needs to be cleaned up.
@@ -2399,4 +2400,23 @@ static inline void sk_pacing_shift_update(struct sock *sk, int val)
sk->sk_pacing_shift = val;
}
+/* if a socket is bound to a device, check that the given device
+ * index is either the same or that the socket is bound to an L3
+ * master device and the given device index is also enslaved to
+ * that L3 master
+ */
+static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
+{
+ int mdif;
+
+ if (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif)
+ return true;
+
+ mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif);
+ if (mdif && mdif == sk->sk_bound_dev_if)
+ return true;
+
+ return false;
+}
+
#endif /* _SOCK_H */
diff --git a/net/ipv6/datagram.c b/net/ipv6/datagram.c
index a1f918713006..fbf08ce3f5ab 100644
--- a/net/ipv6/datagram.c
+++ b/net/ipv6/datagram.c
@@ -221,8 +221,7 @@ int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
if (__ipv6_addr_needs_scope_id(addr_type)) {
if (addr_len >= sizeof(struct sockaddr_in6) &&
usin->sin6_scope_id) {
- if (sk->sk_bound_dev_if &&
- sk->sk_bound_dev_if != usin->sin6_scope_id) {
+ if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) {
err = -EINVAL;
goto out;
}
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index aa12a26a96c6..c0f7e69f2e6c 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -176,8 +176,7 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
/* If interface is set while binding, indices
* must coincide.
*/
- if (sk->sk_bound_dev_if &&
- sk->sk_bound_dev_if != usin->sin6_scope_id)
+ if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id))
return -EINVAL;
sk->sk_bound_dev_if = usin->sin6_scope_id;
--
2.11.0
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox